This article is for educational purposes only. It is not investment advice and not a recommendation to buy or sell anything. Options and futures carry a high risk of loss, and leveraged products can lose more than the amount invested.

Nebius Group (NBIS) reports its third quarter results on Thursday, November 12, before the market opens, with a conference call at 8:00 a.m. New York time. On Wednesday I sold the earnings move with a short iron fly. This post explains why, step by step, before anyone knows the result. I will publish the review of the trade after the exit, profit or loss. The thesis in one sentence: the options pay for Nebius's average earnings move, but in a violent sector a few jumps make that average and the typical report moves about half as much, so I sell the move for one night and cap the jumps with wings. The context comes first, because it sets the thesis; then each piece of evidence, and what it implies for the trade.

What is the context?

AI neoclouds, the companies that rent out GPU computing power to AI developers, have just lived through a violent two weeks. On October 28, Microsoft's quarterly capital spending came in below expectations, and CoreWeave, Nebius and IREN fell after hours. From the October 28 close to the November 5 close, CoreWeave (CRWV) lost 31.7%, Nebius 34.0%, IREN 37.5% and Applied Digital (APLD), which builds data centers for these companies, 30.5%. Nebius then bounced 17.5% over the next two sessions and fell 5.3% on Tuesday, November 10, on analyst comments about its Microsoft partnership: over the ten sessions to Tuesday, it lost 21.4%, and Nvidia 22.5%.

Credit is the other half of the story. CoreWeave borrows heavily to buy chips. Some of its bonds yielded more than 16% before its results, and the cost of insuring its debt for five years reached 1,289 basis points a year on November 5, up from 452 in June, as I wrote in AI Capex Cut. I covered the other weak point of that debt, chips that lose value faster than the loans that finance them, in The AI Depreciation Bomb.

Then CoreWeave reported, on Tuesday evening. Third quarter revenue rose 168% to $3.66 billion, the revenue backlog reached $132.5 billion, and the company raised its 2026 revenue guidance to $13.0 to $13.3 billion, with capital spending of $37 to $39 billion. It also lost $709 million in the quarter, after $846 million of interest expense. Its finance chief said: "At current yields, we do not intend to issue additional parent-level unsecured debt through 2027", and new debt backed by customer contracts is to come in the next few weeks. Its most recent bond deal I could find dates back to September 18: $3.7 billion of convertible notes at 2.875%, due 2033.

On Wednesday the sector rebounded violently. At Wednesday's close, CoreWeave was up 24.7%, Applied Digital 7.4% and IREN 5.6%. Nebius closed up 15.4%; when I entered the trade at 12:56 p.m., it stood at 158.97, up 18.6% on Tuesday's close of 134.00. The only cause I found in the press is a sympathy rally after CoreWeave's results. I found no news specific to Nebius.

CoreWeave's credit relaxed but remains priced for a lot of risk. Its five-year credit default swap stood at 1,167 basis points on Wednesday, down from 1,289 on November 5. With the standard 40% recovery assumption, that spread implies a default intensity of about 19.5% a year (the spread divided by one minus the recovery rate), or a 63% probability of default by December 2031: a risk-neutral figure, which includes a risk premium and overstates the real-world odds, but shows how expensive the credit is. The curve still slopes upward, from about 950 basis points at six months, so the market prices costly credit, not an imminent default. A credit headline can move CoreWeave, and so Nebius, overnight: one more reason to own the wings.

Why does CoreWeave matter for a Nebius trade? Because Nebius moves with it. Over the past year, a regression of Nebius's daily returns on CoreWeave's gives a beta of 0.92 and an R² of 0.66 on 251 days: two thirds of the variance of Nebius's daily moves goes with CoreWeave's. Against the Nasdaq 100, over the same year, the R² is only 0.32. A credit scare or a sector shock can therefore produce a large Nebius move on any day, with or without a report.

So the context says two things: the sector moves violently in both directions, and much of what moves Nebius comes from outside the company, from CoreWeave, credit or Microsoft. Hence the thesis stated at the top: the options pay for Nebius's average earnings move, but in a violent sector a few jumps make that average and the typical report moves about half as much, so I sell the move for one night and cap the jumps with wings. The five steps below test each part of that sentence.

Step 1: What move were the options pricing?

The cleanest gauge is the straddle: buying the call and the put at the same strike, a bet that pays if the stock moves far in either direction. Its price is the market's estimate of the move. At 12:56 p.m., the 160 straddle expiring Friday, November 13, cost 21 per share, or 13.2% of the stock price.

That 13.2% covers the rest of the week, not just the report. Implied volatility, the annualized volatility built into option prices, shows how much of it belongs to Thursday. It stood at 221% for Friday's expiry, 136% for November 20 and 113% for November 27. The gap between the weeks is the earnings day. The straddle's price implies a standard deviation of 16.5% to Friday's close. Take out the ordinary sessions left, at Nebius's usual standard deviation of 6.6% a day outside reports over the past year, and the options priced an average earnings-day move of about 11.5%, up or down. Pricing those ordinary sessions at the volatility the later expiries imply once the earnings day is stripped out of them gives 12 to 13% instead: either way, at least the historical average.

The volatility surface shows how that event premium was built. Friday's at-the-money implied volatility rose by about 100 volatility points between November 4 and Wednesday morning, almost evenly across the strikes near the money: the same report, priced over fewer remaining days, weighs more and more in an annualized volatility.

So the trade targets this event premium: it uses Friday's expiry, and the exit is planned for Thursday, once the report is out and the event premium has drained, or not.

Step 2: How big is a typical Nebius earnings move, and how thick are the tails?

Next, compare that price with what the stock has actually done. I measured each of the last eight reports from the close before the release to the close of the first session after it: -3.1% (October 31, 2024), +3.2% (February 20, 2025), +4.2% (May 20, 2025), +18.6% (August 7, 2025), -7.0% (November 11, 2025), +1.3% (February 12, 2026), +15.7% (May 13, 2026) and +34.1% (August 12, 2026).

The average absolute move is 10.9%. The median is 5.6%. The options paid for about the average move, 11.5% against 10.9%, while the typical report moves half as much. Three of the eight reports moved more than 11.5%, all three upward, and five stayed inside. On its own, that proves little: even a fairly priced move is exceeded about four times in ten. The telling number is the median: if moves followed a normal curve, a price of 11.5% would imply a median move of about 9.7%; the actual median is 5.6%.

A gap that wide between the average and the median is the signature of thick tails. If earnings moves followed a normal distribution, the average absolute move would sit only 18% above the median. Here it is almost twice the median. Kurtosis measures the same thing on daily returns: excess kurtosis is zero for a normal distribution and rises when rare, large moves dominate.

Over the past year, the excess kurtosis of Nebius's daily returns, measured on log returns, is 1.3, and it falls to 0.5 once the four report days are removed. Those four days are 1.6% of the sessions but 9.5% of the year's variance, six times their weight. Over the longer history since October 2024, excess kurtosis reaches 5.8, and removing the eight report days does not lower it (6.0): the two largest daily moves of the period, -37.4% on January 27, 2025 and +49.4% on September 9, 2025, came on days without a report. Over the past year, the two isolated bars on the right of the histogram below are August 12, 2026 (+34.1%, a report) and July 30, 2026 (+27.1%, no report).

The screen's normality test passes over the year: it weighs the many ordinary days, and has little power in the tails, where each bar holds a single day. So the median supports selling the report move, but the tail is real, it has hit both sides, and it can come from news as well as from a report: the sale has to be capped on both sides.

The cap bounds the worst case. On the August 2026 move, the replay in Step 4 loses 12.22 per share with the wings and 32.24 without them, and the capped sale still made +2.35 per share on average over the eight reports, against +1.77 without wings.

Step 3: Is volatility expensive, and which side?

Thirty-day implied volatility stood at 109.5% on Wednesday. Over the past year it ranged from 71.5% in early October to 167.7% in July. Two gauges place today's level in that history. The rank, the position between the year's low and high, is 39%. The percentile, the share of the year's sessions with a lower implied volatility, is 79%; I measured it on the one-year chart below, sampled at 252 points, so it is accurate to about one point. The two disagree because of July's spike, which stretches the range and makes the rank look low. On most days of the year, implied volatility was lower than today: its median was about 95%.

Against what the stock actually does, the picture flips. Realized volatility, measured on daily closes up to Wednesday's and annualized, was 161% over the last 10 sessions, 119% over 30, 96% over 60 (close to the orange line on the chart) and 125% over 90. Options usually cost more than the moves that follow, a gap known as the volatility risk premium. On Nebius, one-month implied volatility at 109.5% sat below three of these four measures: it was not rich against what the stock had just done.

So implied volatility is high for Nebius's own year, but not high for what the stock is doing now. This is not a bet against volatility in general: with one-month implied volatility no higher than the stock's recent moves, a one-month volatility sale would have no clear edge here. The edge, if there is one, sits only in the event premium and in the shape of past moves: most reports are small, a few are huge, and the wings cap the huge ones.

Skew measures how much more puts cost than calls at the same distance from the stock. A steep put skew says the market fears a fall; a call skew says it fears a squeeze. Here it was nearly flat. Near the money, at 12:56 p.m., puts stood 0 to 3 volatility points above calls. Further out, on the surface above, Friday's implied volatility is only about 13 points higher at 80% of the stock price than at 120%, about 5% of its level; the pricer in Step 5, on the ask side for both, shows 12 points between the 130 put and the 190 call at 2:06 p.m., and by 3:14 p.m. the screen in Step 4 had the two wings within about 2 points of each other. Neither side was rich, so there was no reason to tilt the trade: a symmetric structure centered at the money, the 160 strike, was the natural shape. I explained how to read this gauge in Risk Reversal Skew Explained.

Step 4: Where do the options sit, and which structure?

The open options show where the hedging flows sit, and that matters for where the strikes go. Gamma is how fast an option's delta, its exposure to the stock, changes when the stock moves. The dealers on the other side of these options hedge by trading the stock, so the strikes with the most gamma are where that hedging is heaviest. I mapped it from open interest at Tuesday's close for the expiry of the trade, Friday, November 13, and for the monthly expiry a week later.

On Friday's expiry, about four fifths of the gamma sits between 130 and 190, and the two largest levels, on my measure and on the screen below, are 140 and 160. The 160 and 180 calls were the two most traded options of that expiry in my 2:48 p.m. count, with more than 10,000 contracts each by 3:14 p.m., and calls outnumbered puts more than two to one. On the monthly expiry, the 175 call strike holds the largest level, close to my upper break-even of 177.66, with the 150 put strike next.

Is a large level a magnet or an accelerator? A magnet when dealers own the options: they sell rallies and buy dips around the strike, and the price sticks to it. An accelerator when they are short them: their hedging buys strength and sells weakness, and the move runs. My chart uses the usual convention, dealers long the calls and short the puts, but with buyers chasing calls on Wednesday, the opposite is at least as likely, so I count on neither a pin nor a push. The map still shapes the trade in three ways. The short strike sits at 160, at the money, where the gamma and the trading are concentrated and where the options will be easiest to buy back. The wings sit at 130 and 190, at the edges of the zone that holds most of this week's gamma: the move history sets them (Step 2), and the map adds that beyond them little gamma is left, at today's price, to slow or feed a move. And the gamma of the 160 options grows fast into the expiry: with implied volatility at 95% after the report, my pricing puts it at nearly three times Wednesday's level with one session left, and about five times two hours before Friday's close. Being short the 160 strike on Friday, when that gamma peaks, is the gamma and pin risk I avoid by closing on Thursday.

The same screen confirms that the trade sells the event, not volatility in general: the options expiring this week hold 14% of the gamma of all Nebius options, but only 0.8% of their vega, the sensitivity to a change in implied volatility.

With the strikes set, I replayed five structures on the eight past reports. Each historical move was applied to 158.97, every position was valued at Thursday's close with implied volatility at 95% after the report, and entry prices were Wednesday's mid prices. Results per share:

• Short iron fly 160, wings 130 and 190: +2.35 on average, 5 winners out of 8, worst -12.22.

• Short straddle 160: +1.77 on average, 5 out of 8, worst -32.24 (August 2026), with no limit beyond.

• Calendar 160, sell November 13 and buy November 20: +1.04 on average, 5 out of 8, worst -6.40.

• Short iron condor 140/180, wings further out than the fly's: +0.69 on average, 5 out of 8, worst -10.56.

• Long straddle 160: -1.77 on average, 3 out of 8, worst -13.19.

In plain words: a short straddle sells the 160 call and the 160 put and keeps the premium if the stock stays near 160, with no limit on the loss. The iron fly is the same sale plus a 130 put and a 190 call bought as insurance, which caps the loss. An iron condor sells further away, here the 140 put and the 180 call, so it wins over a wider range but collects less. A calendar sells this week's option and buys the same strike for the following week, and earns as the earnings premium drains out of the front week.

The iron fly had the best average and a maximum loss known at entry. The short straddle earned less, because the iron fly's insurance costs little and pays heavily in a move like August 2026's +34.1%, and the straddle's loss has no floor: Step 2 rules it out. The calendar lost less in its worst case, but earned less than half as much on average, and its result depends on where next week's volatility settles after the report, a second unknown. The condor's short strikes at 140 and 180 still sat inside the range of the big moves.

Why wings at 130 and 190? They sit about 18% below and 20% above the entry price, well beyond the priced move of about 11.5%: far enough to keep most of the premium, close enough to cap a jump like August 2026's, a CoreWeave shock or another day like Wednesday. Together they cost about 3.3 per share, the gap between the straddle's 21 and the iron fly's 17.66 credit.

Eight reports are a small sample. The replay ranks structures; it does not prove an edge.

Step 5: Where do I get out, and what can I lose?

I sold the 160 iron fly expiring Friday, November 13: I sold the 160 call and the 160 put and bought the 130 put and the 190 call, for a credit of 17.66 per share, my fill (the replay above used mid prices). That is about 59% of the 30-point distance between the short strike and each wing. The maximum loss is 30 minus 17.66, or 12.34 per share, reached at expiry below 130 or above 190. The break-even points at expiry are 142.34 and 177.66, about 10.5% below and 11.8% above the entry price: almost exactly the priced move.

To check the price independently, here is the same structure in an option pricer at 2:06 p.m. New York time, with NBIS at 159.60, using the screen's volatilities: bid for the two options sold (about 241% on the 160 strikes), ask for the two bought. The credit read 18.52 per share, a little more than my 17.66 at 12:56 p.m., when I measured 221% at the money. An hour later the market paid slightly more for the same structure: my fill was one the market could give.

The exit was fixed before the entry. Nebius reports before the New York open on Thursday, with its call at 8:00 a.m. I keep the position through the release, because the premium I sold is the event premium, and it only collapses once the numbers are known. I buy the structure back on Thursday at about 3:25 p.m. New York, half an hour before the close, starting at the mid price and moving toward the market price until 3:57 p.m. at the latest. I will not hold it to Friday's expiry.

Why not at the open? The first hour after a report is the worst time to buy options back: spreads are at their widest, implied volatility is still being repriced and the stock is still looking for its level. The rest of Thursday also works for the seller, since a structure one day from expiry loses much of its remaining time value during the session if the stock stays in range. And Thursday's close is the exit I tested on the eight past reports. Why not on Friday? On expiry day the gamma of the 160 strike peaks, and a stock that ends near a short strike leaves you unsure which options will be exercised (pin risk). A second session is also a second chance for a big move: 16.1% on average over two days, against 10.9% on the first, on a stock that moved 10% or more on 33 days without a report over the past year. The risk I accept by staying until the afternoon is a move that keeps going during the session, since the structure loses when the stock moves; the 130 and 190 wings cap that loss.

If NBIS ends Thursday close to 158.97 and implied volatility falls to 95%, my pricing puts the gain at about 10 per share, not the full 17.66. The difference is Friday's time value, which I give up to avoid the expiry. The 95% is my assumption for implied volatility once the report is out, on a one-trading-day count; if it stayed at 130% instead, the gain would be about 7.3.

The pricer's scenario view draws the same structure's profit or loss for Wednesday, Thursday and Friday's expiry, along with the probability of each Friday price implied by option prices. Its Thursday curve keeps implied volatility at its current level, as if the report changed nothing, and starts from its own 18.52 credit: at 160 it shows a gain of only about 3.7 per share. My chart above assumes implied volatility falls to 95% after the report and starts from my 17.66 credit, which gives about 10. Apart from the 0.86 difference in credit, the gap between the two curves is the volatility drop the trade is selling.

What would prove me wrong?

A move beyond about 11.5%, like the +34.1% of August 2026. On this history, three reports out of eight would have lost money, the August 2026 move close to the maximum. The last two reports, both strong gains, could set the pattern again, and a stock that was up 18.6% on the day when I sold may be pricing good news, or not enough of it. A shock on CoreWeave's side, in its credit or its shares, would reach Nebius too, report or not. If implied volatility stays high after the report instead of falling to 95%, the gain at Thursday's close will be smaller. And if a median of 5.6% understates what Nebius now moves on a report, the thesis is wrong at its root.

Earnings trades are a repeated game. One result proves little either way. What matters is whether the price of the move keeps exceeding what the reports actually cost once the tail is capped. For another options case study built the same way, see my Brent options case study ahead of the midterms.

I will publish the full review of this trade after the exit on Thursday, profit or loss: the entry, the exit, what the report did to the stock and to implied volatility, and what the method got right or wrong.

This article is for educational purposes only. It is not investment advice and not a recommendation to buy or sell anything. Options and futures carry a high risk of loss, and leveraged products can lose more than the amount invested. Figures are as of November 11, 2026: 12:56 p.m. New York time for the trade, the close for the day's moves in the sector and for realized volatility, and the time given in each caption, unless stated otherwise.

Further reading

• The Volatility Risk Premium Explained: Why Implied Beats Realized

• Risk Reversal Skew Explained: Reading Put vs Call Implied Vol

• AI Capex Cut: How a 20% Cut Could Tip the US Into Recession in 2027

How to cite: Djellal Djouad, CrossVol Research, "How to Trade Earnings With Options: My Nebius (NBIS) Earnings Trade, Step by Step", 11 November 2026. https://newsletter.crossvolresearch.com/p/how-to-trade-earnings-with-options-nebius-nbis

This article is for educational purposes only. It is not investment advice and not a recommendation to buy or sell anything. Options and futures carry a high risk of loss, and leveraged products can lose more than the amount invested.

Bank of America has just written a sentence that every equity investor should read twice. Semiconductor earnings cuts "could start in 1Q if policy changes translate into 2H27 capex moderation. Under a scenario where 2027E AI capex is cut by 20%, [we] estimate 35-40%+ downside to AI Semis EPS. This alone would reduce 2027E S&P 500 EPS by 7-8%."

My view is that a cut of that size would tip the US economy into recession in the second quarter of 2027. Not through the arithmetic of capex alone, which is smaller than people think, but through three amplifiers: stock market multiples, AI credit and the Fed. Here are the numbers.

What did BofA actually say?

BofA tracks what analysts expect Microsoft, Amazon, Alphabet, Meta and Oracle to spend on capex, meaning data centers, chips and servers. Its data, as of November 6, show estimates for 2027 already revised up by more than 90% this year, the fastest pace in the series. Estimates for 2026 now stand roughly 160% above where they started two years earlier.

The bank's logic: AI chipmakers depend on a handful of buyers. If those buyers cut 2027 budgets by 20%, AI semiconductor earnings fall 35 to 40% or more, and since those companies are now a large slice of the index, S&P 500 earnings for 2027 fall 7 to 8%.

Would a 20% AI capex cut cause a recession on its own?

No. In a scenario Goldman Sachs cited on November 8, AI capex grows about 44% in 2027. Cut that by 20% and spending is still about 15% above 2026 (1.44 x 0.8 = 1.15). Growth collapses, the level does not fall.

The direct hit to the economy is real but limited. On our estimate, it takes about 0.6 percentage point off US growth, and 0.3 to 0.4 point once you remove the chips and servers that are imported. On earnings, the 2027 consensus for the S&P 500 is about +18%. Take 7 to 8% off the level and growth drops to between +8.6% (1.18 x 0.92) and +9.7% (1.18 x 0.93). That is a slowdown, not an earnings recession.

Our angle is that BofA counts only the chipmakers. If capex is cut because the returns on AI disappoint, the cloud companies will also have to write down the value of their GPU fleets, which would hit their own profits. The shock to S&P 500 earnings would then be larger than 7 to 8%. We showed on November 5 why the chips last three years while the loans last ten.

Michael Burry shared a chart over the weekend that frames the cycle: net investment by S&P 500 companies (capex minus depreciation) has reached 2.07% of nominal GDP, close to its 2000 levels. Past peaks of that ratio came with the Nasdaq top in March 2000, the S&P 500 top in late 2007 and the shale top in July 2014.

How much can multiple compression add to the fall?

Goldman Sachs puts the Nasdaq 100 at about 18.4 times expected earnings for the next twelve months. That looks cheap only if those earnings are right. A price is earnings multiplied by a multiple, so the two losses compound.

A worked example, with our assumptions written out. Earnings estimates fall 8%, in line with BofA. The multiple falls 15%, from 18.4 to about 15.6, because long-term rates rise or growth slows. The price falls by 1 minus (0.92 x 0.85), which is about 22%. Neither move needs to be extreme. Together they turn an earnings trim into a bear market.

Could a hot CPI and PPI this week bring back a Fed hike?

Yes, and this is the near-term trigger. October CPI comes out today, Tuesday, November 10, at 8:30 a.m. New York time (14:30 Paris). October PPI follows on Friday, November 13, at the same time, according to the official calendar of the Bureau of Labor Statistics.

The market did not wait for the number. This morning in London, before the release, Fed Governor Christopher Waller said he would back at least a quarter point hike at the December 8 and 9 meeting, and opened the door to a half point move if inflation runs hot. Overnight index swaps now price a December hike in full, at 100.4% against 83.1% earlier this morning.

With 165,000 jobs created in the latest payroll report, the labor market does not ask for help. Hot inflation figures would lock in that hike and make half a point a real option. The chain is then direct: higher long-term yields, lower multiples for tech, and more pressure on AI borrowers whose credit spreads have already widened sharply. We described on October 29 how the megacaps did not outrun the Fed, and on October 30 how credit, AI stocks and Treasuries already price a US slowdown.

What does it mean for 401(k)s and household spending?

This is where a market problem becomes an economic one. US retirement assets reached $51.2 trillion at the end of June 2026, according to the Investment Company Institute, including $19.9 trillion in IRAs and $10.8 trillion in 401(k) plans. A large share of that money sits in index and target date funds that follow the S&P 500 or the broad US market, and the ten largest companies now weigh close to 40% of the S&P 500. Retirement savers are, without choosing it, heavily invested in the AI complex.

Federal Reserve data for the second quarter show households holding $74.0 trillion of stocks directly and indirectly, about 48% of their financial assets on our calculation, and household net worth at a record 8.28 times disposable income. Moody's Analytics has estimated that the top 10% of earners account for about half of US consumer spending. Those households own most of the stocks. When earnings and multiples fall together, their 401(k) statements fall, confidence falls, and spending follows. That wealth effect is the bridge from a capex cut to a recession.

How does this become a recession by Q2 2027?

Through three channels, in this order.

First, markets. Earnings cuts for chipmakers can start in the first quarter, BofA says, while the capex slowdown itself would show in the second half of 2027. The market will not wait for the second half: it will price the cuts in the first half, and the wealth effect will start then.

Second, AI credit. The cost of insuring Oracle's debt for five years rose from about 84 basis points a year ago to 304 on November 5. CoreWeave's rose from 452 in June to 1,289 on November 5, its high. Ten-year loans backed by chips that last three years do not survive a spending cut. We explained in July why AI debt made the US investment grade index prettier and more fragile.

Third, the Fed. If inflation forces it to hike into a slowdown, it removes the cushion that usually stops a market fall from becoming a recession.

Each channel alone is a slowdown. Together, with a stock market this concentrated and households this exposed, they meet in the second quarter of 2027. Last Thursday's 6.4% fall in the Nasdaq 100 was the setup, as we wrote in Stock Market Crash: What Happens Next.

What would prove me wrong?

Soft CPI and PPI figures this week, which would take a December hike off the table. Cloud companies raising 2027 budgets with evidence that AI revenue covers the depreciation. And credit spreads on Oracle and CoreWeave tightening back. None of these is visible today.

This article is for educational purposes only. It is not investment advice and not a recommendation to buy or sell anything. Options and futures carry a high risk of loss, and leveraged products can lose more than the amount invested. Figures are as of November 9, 2026, unless stated otherwise.

Further reading

• How to Trade Earnings With Options: My Nebius (NBIS) Earnings Trade, Step by Step

• The AI Depreciation Bomb: The Chips Last Three Years, the Loans Last Ten

• Stock Market Crash: What Happens Next, From a Dead Cat Bounce to the AI Bomb

How to cite: Djellal Djouad, CrossVol Research, "AI Capex Cut: How a 20% Cut Could Tip the US Into Recession in 2027", 10 November 2026. https://newsletter.crossvolresearch.com/p/ai-capex-cut-recession-2027-stocks-401k

A CrossVol analysis by Djellal Djouad, built on the CrossVol terminal, our own options, positioning and market history data, and public filings.

This article is for educational purposes only. It is not investment advice and not a recommendation to buy or sell anything. Options and futures carry a high risk of loss, and leveraged products can lose more than the amount invested. Prices are closing prices of Friday, November 6, unless stated otherwise. Option surface figures marked "Friday midday" were taken around noon New York.

On Thursday the Nasdaq 100 fell 6.36%, its worst session since March 2020. About $2.3 trillion of market value disappeared in one day, the largest loss in dollars on record. Semiconductors lost 12.3%. On Friday the index rose 2.6% in the first twenty minutes, fell back to its low of the day when Bloomberg reported, early in the New York afternoon, that Microsoft is winding down an arrangement to rent AI computing capacity from Oracle, and closed up 1.35% at 26,990, with half of its morning gain gone.

Thursday's crash and Friday's dead cat bounce are the setup. What happens next is the bomb. Over the next two weeks, the whole chain that rents out AI computing power reports, one link after the other. The client, Microsoft, has just walked away from capacity it rented at Oracle. The landlords, Oracle, CoreWeave and Nebius, borrowed to build data centers: CoreWeave reports on Tuesday, Nebius on Thursday. The supplier, Nvidia, worth about $4.5 trillion, reports on November 17, three days before the monthly options expiry. And the market walks into it after selling its hedges on Wednesday, buying them back in a panic on Thursday, and bouncing on Friday mostly because volatility fell.

My reading: Friday's bounce is a dead cat bounce, the next two weeks carry the real danger, and next week could turn black. I tested that view against the options market, the positions reported to the US futures regulator, leveraged ETFs, dealer hedging, trend followers, our terminal and forty years of Nasdaq 100 history. The data agree with the first two parts. On the black week they are more careful than I am, and I will show you why.

In short:

• The bomb is the calendar. CoreWeave on Tuesday, Nebius on Thursday and Nvidia on November 17 report within days of Microsoft letting rented Oracle capacity go. Options price moves of about 21%, 19% and 11% by the expiry that follows each report. For Nvidia, that is about $500 billion of market value, up or down, three days before the November 20 expiry.

• The bounce has the profile of a dead cat bounce. It was narrow and mechanical: it came from dealer hedges unwinding as volatility fell, not from a wave of buyers.

• The market sold its hedges on Wednesday and bought them back on Thursday. The one-month QQQ skew went from 5.2 points, the 28th percentile of the past year, to 12.4, the 99th. Cboe counted 11.2 million puts traded on Thursday, against 7.7 million on Wednesday.

• The crash tail stayed expensive. On the bounce, investors sold back ordinary protection but kept their crash insurance near the highest price of the past year.

• Hedge funds were caught long. Leveraged funds went from their largest short in Nasdaq 100 futures in five years, in August, to a net long in late October, still in place on November 3, two days before the crash.

• Leveraged ETFs are forced sellers. On Thursday they had to sell about $40 billion into the close, more than half of it in chip funds. A 6% fall next week would force about $32 billion more.

• Below 26,600, then 26,000, there is a trap door. Dealers reach their shortest gamma, leveraged ETFs and volatility-targeting funds sell by construction, trend followers are already short on two horizons out of three, and the Fed is not coming to the rescue.

• A black week, down 5% or more, is not the central case. Options price it at about 14%, history at about one in four after a bounce like Friday's, and only inside crises already underway. It is a tail that options price below its historical frequency, and the next two weeks give it a fuse.

The week in five lines

• Monday to Wednesday, rotation then relief. The market read Tuesday's Democratic sweep of the midterms as gridlock. On Wednesday the S&P 500 gained 1.58% and Microsoft 6.7%.

• Thursday, the crash. Nasdaq 100 down 6.36%, S&P 500 down 4.36% with 426 members down and 72 up, semiconductors down 12.3%.

• Friday, a bounce that faded. Up 2.6% in the first twenty minutes, down to the low of the day on the Microsoft and Oracle headline, up 1.35% at the close. On the week the Nasdaq 100 lost 4.9% and the S&P 500 2.7%.

• Volatility woke up. The Nasdaq volatility index posted its biggest one-day jump since at least April 2020, and the VIX curve inverted.

• AI credit names kept falling. Oracle lost 11.2% on the week and CoreWeave 15.4%.

What happens next? The AI bomb of the next two weeks

On Thursday, less than half an hour after the open of what became the worst Nasdaq 100 session since March 2020, I published The AI Depreciation Bomb. Its argument was that the accounts assume GPUs keep their value far longer than they do, and that data center loans "look comfortable until the first contract renewal at a lower price." One day later, the most creditworthy tenant of the AI build-out passed on a renewal at Oracle. Over the next two weeks, the rest of the chain faces the same question in public, in New York time:

• Tuesday, November 10. October CPI at 8:30 am. After the close, CoreWeave reports its third quarter, with a call at 5 pm. Microsoft was CoreWeave's largest customer, 62% of its 2024 revenue according to its IPO prospectus: this is the first public test of how Microsoft treats the capacity it rents. Options expiring on Friday, November 13 price a move of about 21% either way.

• Thursday, November 12. Nebius reports before the open. In September 2025 it signed a contract to supply Microsoft with GPU capacity through 2031, worth $17.4 billion and up to $19.4 billion. Its options price about 19% by Friday. After the close come Cisco and Applied Materials, which sell the networks and the chipmaking tools of the build-out: about 10% and 11%. The 30-year Treasury auction falls in between, at 1 pm.

• Tuesday, November 17. Nvidia reports after the close. Options expiring on Friday, November 20 price a move of about 11% either way. On a market value of about $4.5 trillion, that is about $500 billion, up or down, on one stock.

• Wednesday, November 18. VIX options and futures expire.

• Friday, November 20. The monthly options expiry, three days after Nvidia. Dealers will have to rebuild a large part of their hedges in the same week as the most important report of the quarter.

Why call it a bomb rather than a calendar? Three reasons.

The chain runs on debt. The landlords borrow to buy the chips and build the halls, and lenders, as I wrote on Thursday, "lend mainly against the rental contracts," so that "the real collateral is then the tenant's promise to keep paying." A tenant that does not renew is that collateral leaving the building. The credit market had started to price it before Friday: the cost of insuring Oracle's debt for five years went from about 84 basis points a year ago to 304 on November 5, and CoreWeave's nearly tripled since June, from 452 to 1,289 on November 5, its highest of the past year. CoreWeave's results will be read for the renewal terms of its contracts, Nebius's for the pace of its Microsoft contract, and Nvidia's for whether the landlords keep ordering.

The market is not hedged for it. On Wednesday investors sold their hedges into the relief rally. On Thursday they bought them back at any price. On Friday they sold the ordinary protection again and kept only the crash wing, and the bounce itself came from dealer hedges unwinding as volatility fell, as I show below. That is the usual rhythm after a panic: the newly hedged sell into the bounce, then sell their puts once they feel safe, just before the next leg. A bounce built on falling volatility leaves the market less protected by the time the reports arrive, and the first disappointment pushes volatility back up, which turns Friday's mechanical support into selling.

The Fed is not coming to the rescue. In 1987, 1998 and 2020 the Fed stepped in after the panic. This time Friday's strong October jobs report left futures markets pricing a December 9 hike at more than 75%. A hot CPI on Tuesday morning, hours before CoreWeave reports, would make it worse.

Each of these moves is priced on its own. What options do not price is the chain: a disappointment at the first link raises volatility for the whole index just before the next link reports.

Was Thursday a stock market crash?

For the Nasdaq 100, yes. A fall of 6.36% in one session is a one-day crash by any usual measure. Since October 1985 the index had fallen 5% or more in a day 69 times, and only 30 of those falls were larger than Thursday's. All 30 came during known crises: 1987, the Asian and Russian crises of 1997 and 1998, the dot-com bubble and bust, 2008 and March 2020.

The week started as a rotation. On Monday the Nasdaq 100 lost 1.4% while the equal-weight S&P 500, where every stock counts the same, rose. On October 29, in The Megacaps Did Not Outrun the Fed, I wrote that "so far the market has rotated rather than broken." That still held on Wednesday.

On Thursday the same election result was read the other way. Newly elected Democrats could push local moratoriums on new data centers, and a large share of the US data center pipeline suddenly looked at risk. Then positioning took over: the stocks hedge funds owned most fell, the ones they were short rose, and forced selling fed on itself. The Nasdaq 100 closed at 26,629.81, after a low of 26,606.43. The Philadelphia Semiconductor Index lost 12.3%, its worst day since March 2020. Micron fell 17.5%, AMD 17.3%, Lam Research 16.4%, Broadcom 10.4% and Nvidia 9.3%. The S&P 500 lost 4.36%, its largest fall since April 2025.

On Friday, after a strong October jobs report, the Nasdaq 100 opened higher and reached its high of the day, 27,316, up 2.6%, at 9:50 am. It never got back there. It printed its low of the day, 26,888, at 12:50 pm, minutes after the Microsoft headline, and closed at 26,990, up 1.35%, keeping about half of its morning gain.

Is Friday's bounce a dead cat bounce?

A dead cat bounce is a short rebound inside a decline that does not change its direction. Four things make me think Friday is one.

It faded early and it was narrow. The stocks that had fallen most led the bounce: AMD closed up 11.0%, Lam Research 8.2%, Micron 5.1%, Nvidia 2.8%. Apple and Microsoft closed lower and Broadcom barely moved. A rebound led by the most beaten names, with the largest stocks absent, rarely marks a bottom.

It was mechanical. On Thursday investors bought puts in record numbers, and the dealers who sold them hedged by selling Nasdaq futures. On Friday morning implied volatility fell hard: the VXN, the Nasdaq 100 volatility index, went from 38.26 at Thursday's close to 32.25 at 9:50 am, six points lower. When volatility falls, an out-of-the-money put loses delta, its sensitivity to the index, so the dealers who sold it need fewer futures against it and buy some back. That effect is called vanna. The index peaked at 27,316 in the same five minutes as the volatility low. At 12:50 pm, on the Microsoft headline, the VXN rose to its high of the day, 35.43, and the index printed its low of the day, 26,888, again in the same five minutes. Volatility down, index up. Volatility up, index down. On our terminal, dealer vanna on the Nasdaq 100, QQQ and Nasdaq 100 futures still read as supportive at the close, concentrated around 26,500 on the index, the QQQ 650 strike and 27,500 on the futures. That support lasts as long as volatility keeps falling. It turns into selling when volatility rises, and the next two weeks are a calendar of reasons for it to rise.

Dealers did not change sides. Dealers are the banks and market makers on the other side of option trades, and they hedge by buying and selling the index. When they are short gamma, their hedging forces them to sell into declines and buy into rallies, which amplifies moves. On our terminal, dealer gamma on the Nasdaq 100 stays negative below 28,340, and on Nasdaq 100 futures below about 28,890. Our own calculation on Thursday night's positions puts the switch near 28,590. Either way, the index closed about 5% below it. On October 16, as I wrote two days later in The VIX Woke Up Loaded the Other Way, it was 4.1% below its switch level. The gap has widened.

Option buyers used the bounce to buy protection. On standard Nasdaq 100 index options, 2.4 puts traded for every call by midday on Friday, led by December puts at 26,000, 24,000 and 27,000. Our own scanners agree: almost all of the setups they flagged on Friday were put spread purchases.

History gives the same answer, as I show below: after a crash like Thursday's, a new closing low came within a week in 68% of all cases, and in 40% after a bounce of at least 1% like Friday's. On Friday's midday prices, the options market gave a probability of about 55% that the index trades back down to Thursday's low before next Friday's close. Our positioning indicators push me above the options: I put the odds of a retest of 26,600 by next Friday at 60% to 65%.

What is volatility saying, from one week to one year?

On October 4, in Volatility Is Starting to Talk Even While the VIX Sleeps, I wrote that the VIX "will be the last to admit it." This week it admitted it. The VIX measures the 30-day volatility of the S&P 500 implied by option prices. The VXN does the same for the Nasdaq 100.

On Thursday the VXN jumped 13.18 points to 38.26, up 52.6%, its largest one-day rise in points and in percent since at least April 2020. The VIX closed at 28.18, up 69.7%. The VIX rose above its three-month equivalent, VIX3M, to a ratio of 1.103, the highest of the past year, and the first VIX future settled 2.34 points above the second. When near-term protection costs more than protection three months out, the curve is inverted. In our review of October 19 to 23, I noted that the volatility curve never inverted during that week's AI selloff. This time it did.

The term structure tells the story by horizon. At-the-money implied volatility on QQQ, the Nasdaq 100 tracker:

• One week: 24.1% on Wednesday, 44.2% at Thursday's close, 34.3% at Friday midday.

• One month: 22.3%, then 33.6%, the highest of the past year, then 29.1%, still in the 97th percentile of the year.

• One year: 24.2%, then 26.5%, then 25.6%.

• The gap between one week and three months: +1.0 point on Wednesday, +15.2 on Thursday, +8.1 on Friday. At the peak of the April 2025 tariff shock it was +28.5.

So Thursday reached about half of the April 2025 stress at the front, and Friday gave half of it back. The one-year volatility barely moved. The market treats Thursday as an event, not as a change of regime. On Friday the VIX closed at 25.52, the VXN at 34.0 and the VIX to VIX3M ratio at 1.04: the curve is still inverted going into the weekend. At the close, the Nasdaq 100 surface told the same story: one-month at-the-money volatility at 28.55%, three months at 25.55% and one year at 23.54%, with the one-month 25 delta put at 35.10%.

The skew: sold on Wednesday, bought back on Thursday, and the tail never relaxed. Skew measures how much more investors pay for puts, protection against a fall, than for calls. I explained how to read it in Risk Reversal Skew Explained. On one-month QQQ options:

• The gap between 25 delta puts and calls, the standard skew measure, fell to 5.2 volatility points on Wednesday, the 28th percentile of the past year, below the 7.1 of the previous Friday: investors were selling protection into the relief rally. It jumped to 12.4 on Thursday, the 99th percentile, and eased to 9.9 at Friday midday, the 92nd.

• The premium of far out-of-the-money puts, the crash insurance, rose from 12.1 to 12.8 points, the 99th percentile.

Investors sold back ordinary protection on the bounce, and kept their crash insurance near the highest price of the year. That is not what a market convinced the bottom is in looks like.

Implied volatility is below realized volatility at the front. Normally implied volatility sits above the volatility the index then delivers. That gap is the volatility risk premium, the price of insurance, which I explained in The Volatility Risk Premium Explained. Today it is negative at the front. One-week implied volatility was 34% at Friday midday, while the Nasdaq 100 has delivered about 49% over the last five sessions. Implied stays below realized up to one month. Only from three months does the premium turn positive again, by about 3.5 points. The QQQ straddle for next Friday prices a move of about 3.8% either way.

Read from the prices of Nasdaq 100 puts at Friday midday, the options market gives a 28% chance that the index closes next Friday below Thursday's close, 14% for a fall of 5% or more, and 3.7% for a fall of 10% or more. These probabilities include the premium investors pay for puts, which usually makes them overstate the real odds of a fall. That is why I call the black week a tail, not a forecast.

Who holds the puts? Put and call positions on QQQ and the Nasdaq 100

Volume tells you what traded during the day. Open interest, the number of option contracts still open at the close, tells you who stays exposed overnight.

Thursday set a record for puts, but on QQQ calls traded almost as much. Cboe's daily statistics show 11.2 million puts traded on Cboe on Thursday, against 7.7 million on Wednesday, and 1.06 puts for every call, against 0.80. Bloomberg reported it as the most puts ever traded in one day. On QQQ, though, about 7.0 million puts traded against 6.8 million calls, 1.03 puts for every call, less than the 1.14 of Wednesday. The three most traded contracts were QQQ puts expiring the same day, at the 645, 650 and 640 strikes, with 478,946, 471,022 and 332,425 contracts.

The open positions lean toward puts. At Thursday's close, 7.53 million QQQ puts were open against 6.89 million calls, 1.09 puts per call. On Nasdaq 100 index options the ratio was 1.17, with 107,023 puts against 91,319 calls.

The puts sit at a few strikes. QQQ 650, about 26,725 on the index, holds 428,691 puts in expiries after Friday, and QQQ 600, about 24,670, holds 320,854. The largest single position is a long-dated one: 202,356 QQQ March 2027 puts at the 670 strike, protection on about $13 billion of Nasdaq 100 exposure. On the call side, the largest strikes are far above the market: QQQ 700 and 800, about 28,780 and 32,890 on the index, with about 245,000 calls each. On the November 20 expiry alone, 49,831 QQQ 650 puts were open after Thursday, 8,371 more than the day before, and another 16,771 traded on Friday.

Who sold and who bought. On Thursday, 62,022 QQQ November 20 puts at the 650 strike traded, $108.6 million of premium, mostly at the bid, which usually means sales, while 22,847 December puts at the 690 strike traded, $91.4 million, mostly at the ask and largely to close existing positions. Between Tuesday's close and Thursday's close, open interest in QQQ puts after March 2027 rose by 76,853: protection is moving to 2027.

What do swaps tell us about positioning?

Two kinds of swaps matter here: one can be read every day, the other can hardly be seen at all, which is a risk in itself.

Variance swaps: the price of the tails. A variance swap pays the difference between the volatility an index actually delivers and a level fixed at the start. The VXN is computed from the whole strip of out-of-the-money Nasdaq 100 options, which is how a 30-day variance swap is priced, so it is a good guide to that level. The gap between the VXN and one-month at-the-money QQQ volatility shows how much investors pay for the tails: 2.8 points on Wednesday, 25.08 against 22.3, then 4.7 on Thursday, 38.26 against 33.6, and about 4.3 at Friday midday, 33.4 against 29.1. The bounce took 4.5 to 5 points off both and left the gap almost intact: the tails kept their price. The VXN also closed more than 8 points above the VIX on Friday, against a median of 5.7 over the past year. The stress is concentrated in the Nasdaq.

A variance swap struck near the VXN's Friday close of 34 needs the Nasdaq 100 to move less than about 2.1% on a typical day over the next month for its seller to make money. Over the last five sessions the index delivered about 49% annualized, more than 3% a day. Sellers of variance are betting that Thursday was a one-off, just before the two weeks most likely to prove them wrong.

One old danger is absent. In February 2018, products linked to the VIX had to buy VIX futures into a spike, and one of them was wiped out. On our terminal, what these products would have to buy if the VIX rose another 10 points is a few percent of a normal day's volume, the lowest reading of the past year. They are not the weak link this time.

Equity swaps: the positions nobody can see. Hedge funds hold much of their stock exposure through total return swaps with their prime brokers: the bank owns the shares and the fund receives the performance, with leverage. New swap trades are reported without names, and the stock of positions is not reported at all. That is how Archegos built large, concentrated positions in 2021 without appearing in any public filing. What can be observed is indirect: on Thursday the stocks hedge funds own most fell hardest while the most shorted rose, the signature of a forced unwind of long and short books at the same time. What is not observed is the size of the swap books and where their margin calls start. That is why I treat the hedge fund unwind as a risk to watch in prices, not as a number.

What does the CFTC report say about hedge funds and asset managers?

Every Friday, the CFTC, the US futures regulator, publishes the positions held the previous Tuesday in financial futures, split by type of trader. For stock index futures, three groups matter. Asset managers are pension funds, mutual funds and insurers. Leveraged funds include hedge funds and trend followers. Dealers are the banks on the other side.

The latest report covers Tuesday, November 3, with the Nasdaq 100 at about 28,180. It was published on Friday at 3:30 pm New York. It does not show Thursday's selling. It shows who was exposed when it started.

Leveraged funds turned long at the wrong time. On August 11, leveraged funds were net short 89,125 E-mini Nasdaq 100 contracts, their largest short of the past five years. They turned net long on October 20, for the first time since November 2025, and held 10,073 contracts net long on November 3, after adding 7,772 in the last week. In twelve weeks they bought 99,198 contracts, about $56 billion of Nasdaq 100 exposure at November 3 prices. A small net long sounds harmless, but for this group it ranks in the 98th percentile of the past five years: leveraged funds are usually short Nasdaq futures, as a hedge against their stock portfolios or as a trend bet. They lifted the hedge, or bought the trend, near the top.

Dealers took the other side. Over the same twelve weeks, dealers went from a small net long to a net short of 112,529 contracts, their largest short of the past five years.

Asset managers did not run. Asset managers held a net long of 73,137 contracts, about $41 billion, in the 67th percentile of five years. In the week to November 3 they added 7,756 contracts, about $4.4 billion.

On the S&P 500, the picture is calmer. Asset managers held a net long of 928,584 E-mini S&P 500 contracts, about $351 billion, in the 69th percentile of five years. Leveraged funds were net short 372,318 contracts, about $141 billion, partly the futures leg of the basis trade, in which funds buy stocks and sell futures against them, so not a pure bet on a fall.

What it means for next week. The money that chased the Nasdaq into late October is the money that has to cut first. On our terminal's positioning monitor, those longs were built well above current prices and are now under water. The next report, for Tuesday, November 10, will be the first to show how much of it was sold in the crash.

What are leveraged ETFs forced to sell?

Leveraged ETFs promise two or three times the daily move of an index, or of a single stock, or the inverse of it. To keep that promise they rebalance every day at the close, and always in the direction of the market.

Take a fund with $100 of assets and three times leverage: it holds $300 of exposure. After a 6.36% fall, as on Thursday, it has $80.9 of assets and $280.9 of exposure, so it must sell $38.2 into the close to get back to three times. A fund that bets three times against the index must add to its short after a fall, which is also a sale. Long or short, leveraged ETFs sell after down days and buy after up days, in the last minutes of trading, all in the same direction.

How large is it this week? From the assets the issuers publish, I estimate that the leveraged and inverse ETFs on the Nasdaq 100, on chips and on single AI stocks had to sell about $40 billion of exposure at Thursday's close. About $16 billion came from the Nasdaq 100 funds, $13 billion of it from TQQQ alone. About $22 billion came from the chip funds, because semiconductors fell twice as much as the index. Funds on single stocks such as Nvidia, Micron and AMD added about $2 billion. On Friday's bounce, the same funds had to buy back about $10 billion at the close. Retail money keeps buying these dips with leverage: according to 247 Wall St., investors put $7.27 billion into SOXL and $3.49 billion into TQQQ in October, and Korean retail investors alone net bought $2.97 billion of SOXL.

Next week, at Friday's asset levels, a 3% fall of the Nasdaq 100 in one session would force about $16 billion of sales at the close, and a 6% fall, a repeat of Thursday, about $32 billion. A 3% rise would force about $16 billion of purchases. The group's assets have shrunk from about $85 billion at Wednesday's close to about $68 billion at Thursday's, so the same fall now forces about a fifth less selling than on Thursday, and chip funds still carry more than half of it. On a day when Nvidia or CoreWeave disappoints, the chip funds are the first to sell.

Where is the trap door below 26,600?

Five groups of mechanical sellers line up at the same levels. Above 27,000 they are a headwind. Below 26,600 they become a trap door.

Dealers. Our estimate of dealer gamma on Nasdaq 100 and QQQ options, on Thursday night's positions, is about minus $8.7 billion per 1% move around Friday's close. It gets worse as the index falls, reaching minus $10.1 billion between 26,000 and 26,300, and there is no positive gamma floor below. In plain terms, every 1% fall below 26,630 would force dealers to sell roughly $9 billion to $10 billion of Nasdaq exposure. Our terminal marks an acceleration zone near 25,850 on the index. Its DEFCON gauge, a five-level stress scale where 1 is the worst, swung between 3 and 2 on Friday, at 2 shortly after the Microsoft headline, and ended the week at 3.

The largest put strikes sit right there: QQQ 650, about 26,725 on the index, then 26,500, 26,000 and QQQ 600, about 24,670. In a market where dealers are short gamma, big put strikes act as magnets more than floors. They become support only when put buyers start taking profits.

Leveraged ETFs. They sell at the close of every down day: about $16 billion for a 3% fall of the Nasdaq 100, $32 billion for a 6% fall, more than half of it in chip funds.

Trend followers. CTAs, or commodity trading advisors, are systematic funds that buy futures when prices trend up and sell when they trend down, over short, medium and long horizons. On our trend model, the Nasdaq 100 futures signal is near its most negative level of the past year: the short and medium-term signals are already short, and the long-term signal is still slightly long and close to turning. When it turns, all three horizons sell together. Across US equity futures, trend followers have gone from a large long a year ago to a small short.

Volatility-targeting funds. These funds hold more stocks when markets are calm and fewer when they are volatile, to keep the risk of the portfolio constant. The realized volatility of the S&P 500 over 20 sessions has gone from 12.6% on Wednesday to 20.1%, and over 10 sessions from 15.0% to 26.2%. For a fund that targets 10% volatility, that means going from about 79% invested to about 50%. These funds spread the cut over several sessions, so part of that selling lands next week even without a new fall. Thursday stays inside the 20-session window until early December.

Hedge funds. The CFTC data above show leveraged funds net long Nasdaq futures two days before the crash, in the 98th percentile of five years. Our hedge fund monitor shows the same crowd from another angle: gross exposure near the top of its range, net exposure middling. Funds positioned like that protect themselves with index puts rather than by selling their longs, which fits Thursday's record put volume, and they sell the longs only when the puts stop being enough. Our deleveraging gauge, which tracks the pressure on hedge funds to cut, is in its high zone and rose again this week. A list of hedge fund losses circulated this week, which I could not verify: Arini was reported down 26%, VARA 33%, Tiger Global 8.5% and Lone Pine 8.8%. On October 25 I wrote that "the real risk sits in the longs." That is where it sits now.

One piece is still missing: capitulation. One-month implied correlation on the S&P 500 jumped from 9.5 on Wednesday to 22.8 on Thursday, but it remains far from its high of the year, 41.7, and the realized correlation between S&P 500 stocks is still low, around 0.17. Bottoms usually come when everything falls together and diversification stops working. If the selling spreads from AI to the whole market, correlation rises, and the systematic funds have to sell even more.

What happens after a stock market crash? Forty years of Nasdaq history

Before Thursday, the Nasdaq 100 had fallen 5% or more in one session 69 times since October 1985. Thursday was the first such fall since April 2025.

• Across all 69 cases, the index closed below the crash day's close within five sessions 68% of the time and within ten sessions 77% of the time. The median maximum loss over the next 20 sessions was 10.7%.

• When the next day bounced at least 1%, as on Friday, 25 cases, a new closing low came within the following five sessions 40% of the time, and within two weeks 52% of the time.

• The black week. In 6 of those 25 cases the week after the bounce fell 5% or more: October 1987, March 2001, September and October 2008, and twice in March 2020. Every one of them came inside a crisis that was already underway.

• Three months later, the median return after such a bounce was plus 13.8%, and the index was higher in 18 of the 25 cases.

History supports the dead cat bounce more than the black week. It also says that a crash like Thursday's has usually been bought within three months, unless the market is already in a crisis. The next two weeks will tell us which regime this is.

Why did Microsoft walk away from Oracle?

Bloomberg reported on Friday that Microsoft is winding down an arrangement to rent AI computing capacity from Oracle. According to the report, Microsoft chose not to renew the contract and has been switching that capacity off in recent months. People familiar with the matter said the rented capacity no longer met Microsoft's requirements, and several accounts of the story point to the older generation of Nvidia chips it ran on. The value of the arrangement was not disclosed, and Microsoft declined to comment. Oracle answered that demand for its AI infrastructure remains strong, including for earlier-generation GPUs.

The market reacted at once, then hesitated. Oracle fell to $109.21 at 12:38 pm, 5.2% below Thursday's close of $115.16. By the close it was back near $116, up about 0.8% on the day and still down 11.2% on the week. Microsoft closed down 1.0%. Options now price a move of about 10.6% for Oracle and 5.0% for Microsoft by November 20, the expiry that follows Nvidia's report.

The risk sits with whoever owns the chip when the contract ends. In The AI Depreciation Bomb I wrote that hardware "loses its value in steps, every time a new generation arrives." I also quoted Microsoft's chief financial officer, who said in January that a lot of the GPUs Microsoft buys are already contracted for most of their useful life. That protects Microsoft and moves the risk to its tenants. Friday showed the mirror image. When Microsoft is the tenant and has the choice, the owner of the hardware keeps the obsolescence. On September 25, after Oracle's force majeure notice on its New Mexico campus, I called it in The Crash Fuse Is Lit "the first real test of what an AI lease is worth." Friday was the second test, from the other side of the lease.

Old chips do find users, but not the most demanding ones. Oracle's answer, that demand stays strong even for earlier-generation GPUs, is Steve Eisman's point, which I cited on Thursday: older chips keep finding users. But as we argued on October 4 in AI inference has a floor, and it is priced in watts, the marginal cost of a token is dominated by electricity. An older chip needs more power for each token, which is why it loses value even while it still works.

For the Nasdaq, the question now is not Oracle alone. It is which other leases of older chips come up for renewal next, and at what price, and the next two weeks of reports are where the answer starts. On October 25 I recalled my verdict of October 13: the next phase would come "through credit. The indices will see it last." This week, the indices saw it.

Will the stock market crash again next week?

The sequence I watch is simple: a break below 26,600, Thursday's low, then below 26,000, where dealers are shortest gamma. Besides the reports above, three moments can start it, in New York time:

• The weekend. The Middle East, where talks between Iran and the United States resumed through Qatar while tanker fires were reported near the Strait of Hormuz. Then Asia on Monday morning: Japan's Nikkei and Taiwan's benchmark entered bear markets on Friday, joining Korea.

• Tuesday, November 10. October CPI at 8:30 am, the 10-year Treasury auction at 1 pm, CoreWeave after the close. Brent spent most of October above $100 a barrel, before this week's drop. A hot print would support a Fed hike on December 9, and equity weakness would then bring no rate relief. This is the most credible trigger of a black week.

• Friday, November 13. Weekly options expire, with a straddle of about 3.8% on the Nasdaq 100.

Wednesday is Veterans Day: the bond market is closed, stocks are open. And every day, watch Oracle and CoreWeave credit default swaps, the price of insurance against their default.

Levels to watch on the Nasdaq 100

• 28,340 to 28,590: dealer gamma turns positive. Above it, the bounce becomes credible.

• 27,250 to 27,270: the call wall of short-dated options.

• 26,990 to 27,050: Friday's close and pivot.

• 26,725: QQQ 650, the largest put strike.

• 26,606 to 26,630: Thursday's low and close.

• 26,000 to 26,300: dealers at their shortest gamma.

• 25,850: the acceleration zone on our terminal.

• 24,670: QQQ 600, the next large put strike.

The rest of the market in five lines

• Rates: the US 10-year yield fell from 4.93% to 4.86% on the week. On Thursday it fell about 6 basis points while stocks crashed. On October 18 I wrote that "AI specific sessions behave differently: on those days, yields fall with stocks." Thursday was one of them, and Friday's strong jobs report barely moved yields.

• Oil: Brent lost 13.3% on the week, to about $98, and WTI 12.8%, to about $91.

• Gold: up 4.5%, to about $4,075 an ounce on the December future.

• Dollar: the dollar index rose 0.4% to 102.3, the euro slipped 0.3% to 1.116 and the yen traded near 158.5 per dollar.

• Asia: Japan and Taiwan in bear markets, Korea already there. The selling started in US AI stocks and spread to the Asian supply chain.

What would prove me wrong

• A close above 27,270, then above 28,340 to 28,590. Dealers would stop amplifying moves and start damping them.

• Reassuring reports. If CoreWeave, Nebius and Nvidia reassure on renewals and orders, volatility falls further and the dealer support that lifted Friday keeps working for the bulls.

• A soft CPI on Tuesday. It would take the December hike off the table and remove the most credible trigger.

• The volatility sellers. On Thursday, put sellers dominated the flow in 62,022 QQQ November 20 puts at the 650 strike and 27,156 December puts at the 630 strike. If they are right, front-end volatility collapses and the bounce extends.

• History. Three months after a bounce like Friday's, the median return was plus 13.8%, and the index was higher in 18 of the 25 cases since 1985.

Bottom line

Friday's bounce has the features of a dead cat bounce: narrow, faded, mechanical, unconfirmed by dealers, and used by option buyers to add protection. I put the odds of a retest of 26,600 by next Friday at 60% to 65%.

The bigger risk is what happens next. Between Tuesday and November 20, the AI rental chain reports link by link: CoreWeave, Nebius, Cisco and Applied Materials, then Nvidia three days before the monthly expiry, just after Microsoft let rented Oracle capacity go. The market walks into it with dealers short gamma, leveraged ETFs that sell every down day, trend followers already short on two horizons, hedge funds caught long, and no Fed in sight. A black week remains the tail rather than the central case: options give it about one chance in seven, history about one in four after a bounce like Friday's, and only in crises already underway. But this time the tail has a date.

For an options trader, the structure of prices is clear. One to two-week volatility is the cheap part of the surface, below what the index is actually delivering. The crash wing is the expensive part. A put spread, buying a put near 25 delta and selling one near 10 delta, buys the cheap part and sells the expensive one. That is a description of option prices, not a recommendation.

The Microsoft and Oracle story tells us something larger than one session. The real risk of the AI cycle, how long the chips keep their value, has started to show in contracts. In the next two weeks it will show in earnings, then in credit, before it shows in the indices.

FAQ

What happens after a stock market crash? Since 1985, when a one-day fall of 5% or more in the Nasdaq 100 was followed by a bounce of at least 1%, a new closing low came within a week in 40% of cases. Three months later the index was higher in 18 of 25 cases, with a median gain of 13.8%, and the black weeks all came inside crises already underway.

Why did the Nasdaq crash on November 5, 2026? The Nasdaq 100 fell 6.36%, its worst day since March 2020. The market read the Democratic sweep of the midterms as a threat to the data center build-out, through local moratoriums, and crowded hedge fund positions in AI stocks were forced out. Semiconductors fell 12.3%.

Is Friday's rebound a dead cat bounce? On this week's evidence, probably: it was narrow, it faded, it came from dealer hedges unwinding as volatility fell, and option buyers used it to buy protection. The test is simple: a close back below Thursday's 26,630.

Will the stock market crash again next week? Nobody knows. Options price about a 14% chance that the Nasdaq 100 falls 5% or more by next Friday, and history about one in four after a bounce like Friday's. The most credible triggers are the October CPI and CoreWeave's results on Tuesday, then Nvidia on November 17.

Why did Oracle stock fall on November 6, 2026? Bloomberg reported that Microsoft is winding down an arrangement to rent AI computing capacity from Oracle. Oracle fell as much as 5.2%, to $109.21, then recovered to close up about 0.8%, and says demand for its AI infrastructure remains strong.

This article is for educational purposes only. It is not investment advice and not a recommendation to buy or sell anything. Past moves do not predict future ones. Options and futures carry a high risk of loss, and leveraged products can lose more than the amount invested.

How to cite: Djellal Djouad, CrossVol Research, "Stock Market Crash: What Happens Next, From a Dead Cat Bounce to the AI Bomb", 6 November 2026. https://newsletter.crossvolresearch.com/p/stock-market-crash-what-happens-next-nasdaq

Market commentary for educational purposes. This is not investment advice and not a recommendation to buy or sell anything. Figures are as of the dates given, markets move fast, and stocks, bonds and derivatives can lose far more than you expect.

For months I have told clients that the AI build-out carries a cost the market refuses to price. The chips bought in 2023, 2024 and 2025 keep wearing out, in economic terms, much faster than the accounts say. Every year adds a new layer of hardware to depreciate on top of the old one, and the bill grows faster than the revenue it is supposed to pay for.

This week the mood on Bloomberg is the opposite. Bloomberg Intelligence writes that much of the bad news already appears priced into tech credit spreads, that record order books show investors are paid for supply rather than for credit risk, and that rising Ebitda, earnings before interest, tax, depreciation and amortization, will keep leverage in check. I disagree. One of my clients, David Wolf, an engineer, built a cost model of an AI data center to test my thesis. I have checked it cell by cell and corrected it where it was wrong. The conclusion survives, and it comes down to one number: how long a GPU really lasts. At the end I explain why French banks sit on the wrong side of this.

What the accounts assume

Every hyperscaler depreciates its servers in a straight line, over lives it chooses itself. Microsoft, Alphabet and Oracle use six years, and so does CoreWeave, the largest of the GPU renters. Meta uses five and a half years since January 2025. Amazon is the only one that went the other way: from January 2025 it cut the life of part of its servers and network gear from six to five years, citing the faster pace of technology, particularly in artificial intelligence. None of them shows a separate life for GPUs. The accelerators sit in the same bucket as switches and storage.

This year the trend has been to stretch lives, not shorten them. Nebius, a GPU renter that used four years, moved to five from the first quarter of 2026. Microsoft now depreciates its data center buildings over 25 years instead of 15, from the fiscal year that began in July. Even on these long schedules the charge is exploding. Microsoft's depreciation expense rose 56% to $34.3 billion in the fiscal year to June. Oracle's nearly doubled to $7.6 billion. Meta's depreciation and amortization was up 44% in the third quarter.

Michael Burry argued a year ago that stretched lives would understate the industry's depreciation by $176 billion between 2026 and 2028. Steve Eisman answered last month that Burry is focusing on the wrong risk, because older chips keep finding users. Both have a point, and the market data shows why. An H100 rented for up to about $8 an hour at the 2023 peak. The Silicon Data H100 rental index, which Bloomberg carries, started at $3.40 in September 2024 and fell to $1.86 by February 2025, minus 45% in five months. It was still at $1.96 at the end of November 2025. Then it rebounded to $2.96 on November 4, up 51%, probably because power, not chips, became the bottleneck. Hardware loses its value in steps, every time a new generation arrives, and scarcity can lift the whole staircase for a while. It never makes a 2023 chip a 2026 chip.

The scale makes the question anything but academic. Bloomberg Intelligence's consensus puts 2026 capital spending by the largest US hyperscalers, now including SpaceX, at close to $960 billion: about $220 billion for Amazon, $195 billion to $205 billion for Alphabet, close to $200 billion for Microsoft. Bloomberg Intelligence sees Amazon and Alphabet near or above $300 billion each in 2027. Microsoft has put the share of its spending on short-lived assets, mostly chips and servers, at roughly half, and at about two-thirds in a recent quarter. Take only half of $960 billion as hardware and change nothing but the useful life: the yearly depreciation of this single vintage goes from $80 billion at six years to $160 billion at three. That is $80 billion a year of cost that a six-year schedule pushes into later years, for one year of spending.

David's $1 billion data center

David's model is simple, which is its strength. A $1 billion site: 60% for GPUs and servers, 7% for the network, 25% for power and cooling, 8% for the building. At $40,000 per GPU, server included, that buys 15,000 GPUs. The site draws 25 megawatts, pays Texas power at 6.5 cents per kilowatt-hour, employs 30 people, borrows at 8% and rents its GPUs at $3.90 an hour, 70% of the time. Depreciation is straight-line: five years for the GPUs and the network, fifteen for power and cooling, thirty for the building.

On those inputs the site costs $2.95 per GPU-hour, financing included, and earns a 24% margin. That is the number that goes on the cover of a pitch book. David did not stop there. An adjusted version adds the costs people forget (insurance, local property tax, software, sales and overheads) and lets the rental price fall 15% a year: the margin turns to minus 11.5%. A dynamic sheet starts from a $2.50 contract price, lets supply grow faster than demand and lets ageing GPUs lose 15% a year against new ones: the site loses money every year and never recovers its cumulative losses.

The arithmetic is right. Some inputs are not, and they do not all err in the same direction.

1. The price is too high for this chip. $40,000 per GPU and 25 megawatts for 15,000 of them describe a Hopper site, H100 or H200. $3.90 an hour is more than even an H200 rents for today, $3.48 on October 24. The H100 rents at $2.96, and David himself uses $2.50 for contracts in his dynamic sheet. At $2.50, his first sheet loses $42 million a year, a margin of minus 18%, before any forgotten cost.

2. The financing is charged too hard, and correcting it helps the operator. His adjusted version compares a price averaged over five years with the interest of the first year on the full $1 billion. Debt is repaid as the assets depreciate, so the interest falls every year. The clean method recovers each asset's cost over its life at the 8% borrowing rate, like a mortgage payment, and discounts the falling rental prices at the same rate, which is what I mean below by today's money. Done that way, his adjusted case roughly breaks even, a margin of about plus 2%, against a cost of $2.91 per GPU-hour on a five-year life. So his adjusted loss came from the financing method. The real losses come from the next point.

3. The GPUs do not last five years. This is the heart of the matter. A five-year schedule spreads the $600 million of GPUs at $120 million a year. If the chips are economically obsolete after three, the true cost is $200 million a year. The difference, $80 million a year on this one site, is $0.87 on every GPU-hour sold, pushed into the future by the accounting. On a three-year life the break-even price jumps to $3.83.

Put the corrected cost next to today's market and the whole case hangs on that one assumption. At the H100's $2.96, the site earns 9% of its revenue if the chips last six years, 2% if they last five, and loses 29% if they last three. Even at David's $3.90, held flat for three years, the margin is under 2%.

4. Failures are too low. David assumes 2% of the GPU fleet is replaced each year once the warranty has run out. When Meta trained Llama 3 on 16,384 H100s, it logged 419 unexpected interruptions in 54 days, including 148 faulty GPUs and 72 memory failures. Annualised, that is close to 10% of the fleet, though not every incident means a replacement. Something like 4% to 6% a year is more realistic. With 5% instead of 2%, the break-even on a five-year life rises from $2.91 to $3.11.

5. Prices do not glide down 15% a year. They move in steps: the H100 index lost 45% between September 2024 and February 2025, then gained 51% over the past year. Starting from David's own $2.50 with his 15% decline, the five-year price is worth $1.90 an hour in today's money, a third below the $2.91 that even a five-year life requires.

6. Power is billed at full load. David charges 25 megawatts around the clock while the GPUs work 70% of the time. Idle chips draw less, so this input is slightly harsh on the operator, by at most five cents per GPU-hour. It does not change the answer.

The loan outlives the chips

One more detail in David's dynamic sheet matters to whoever lends the money. He finances the site with a 10-year loan. The GPUs are depreciated over five years and, in my view, lose most of their economic value in three. The debt outlives the collateral.

Take the $600 million of GPUs. After three years, a 10-year loan still has $420 million outstanding. If the chips lose 25% to 35% of their value a year, they are worth between $165 million and $253 million. Even a five-year loan is worth more than the chips for the first two years at a 25% annual loss, four years at 35%. Matching the loan to a five-year life would raise the GPU debt service, with level payments, from $89 million to $150 million a year, another $0.66 per GPU-hour in cash. Lenders know this, which is why they lend mainly against the rental contracts. The real collateral is then the tenant's promise to keep paying, and several of the largest tenants of this cycle, OpenAI first, burn cash.

Why Bloomberg's calm does not reassure me

The Bloomberg Intelligence numbers are good. It is the reading I disagree with.

The credit market has already started to charge. On Bloomberg, the cost of insuring Oracle's debt for five years, its credit default swap, has gone from 82 basis points a year ago to 292 on November 4, after 329 on October 20. CoreWeave's has more than doubled since June, from 452 to 1,162 basis points on November 4, level with its October 22 high. These are builders and tenants of the AI boom, and the market already charges more to carry their risk.

Ebitda leaves out the one cost this trade is about. The argument that leverage stays contained because Ebitda grows about 29% a year through 2031 measures cash generation before depreciation. If the hardware has to be replaced every three years, that replacement is an operating cost in all but name, and Ebitda overstates what is left to service debt.

The supply is the symptom. Investment grade bond issuance by tech companies is about $400 billion this year, against about $250 billion in all of 2025. Tech bonds pay about 112 basis points more than Treasuries, against 92 for the whole investment grade index (a basis point is a hundredth of a percentage point), and their return over Treasuries is minus 170 basis points this year, against minus 15 for the index. Oracle and SpaceX trade closer to BB than to their ratings. Bloomberg reads this as compensation for supply. Supply is what capital intensity looks like when cash flow no longer covers it.

Free cash flow is turning negative. On Bloomberg Intelligence's own estimates, SpaceX, Alphabet, Meta, Oracle, Amazon, CoreWeave and Microsoft all have negative free cash flow in 2027 once dividends and buybacks are paid. Alphabet has already announced an $80 billion stock sale, including $10 billion placed with Berkshire Hathaway. Amazon's commitments, meaning debt, leases and purchase obligations, passed $650 billion at the end of June, $210 billion more than six months earlier. Meta's future AI and infrastructure commitments are about $825 billion, including $356 billion of leases that start as late as 2036.

A residual value guarantee is a bet on depreciation. More than $100 billion of special purpose vehicles and secured chip financings have been arranged recently. Four of the largest, Broadcom with Anthropic twice, Meta with Blue Owl and Meta with BlackRock, carry up to $113 billion of residual value guarantees. A residual value guarantee is a promise by the tenant or sponsor, Meta in the Meta deals, to pay the investors any shortfall if the asset is worth less than an agreed value when the lease ends. If data centers and their hardware lose value faster than planned, that promise becomes a debt.

Why French banks are on the wrong side

French bank shares bounced this morning, Société Générale by 2.3%, BNP Paribas and Crédit Agricole by 1.9% each in early trading, after Banque de France governor Emmanuel Moulin said in Paris that French banks and insurers are well equipped to absorb higher yields. Late on Tuesday, Fitch had moved the outlook on BNP Paribas and Crédit Agricole to negative, four days after doing the same for France itself. The reason is mechanical: Fitch rates both banks one notch above France, and wrote in May that they would be downgraded if France is. The ten-year OAT-Bund spread, the extra yield France pays over Germany, went from 85 basis points at the end of August to 165 on October 30, after a closing high of 174 on October 22, and stood near 151 this afternoon, with the French ten-year yield at 4.78% against 3.26% for Germany. I covered the spike on October 25 and, on October 31, how the stress had reached dollar funding.

Credit markets saw it first. BNP Paribas's five-year credit default swap almost doubled between February and October, from 37 to 73 basis points, and was still at 67 this afternoon.

Now look at where the same banks have been lending.

• Société Générale says it took part in a $2.6 billion delayed draw facility, a loan drawn as the chips are delivered, for CoreWeave, whose credit default swap now trades at 1,162 basis points, to buy Nvidia GB200 chips for OpenAI's workloads in Texas, which it calls one of the first GPU financing transactions in the global banking sector. It also led a $600 million financing for a Virginia data center leased by CoreWeave, and describes GPU renters as an asset class it expects to grow exponentially.

• BNP Paribas is one of the four administrative agents, the banks that run the loan for all the lenders, on the roughly $18 billion loan for the Oracle-linked campus in New Mexico, one of the large AI financings in Bloomberg Intelligence's own table.

• BNP Paribas, Crédit Agricole CIB, Natixis and La Banque Postale are among the seven lenders of Mistral AI's $830 million debt for a Paris-area data center holding 13,800 Nvidia GB300 chips.

• BNP Paribas, Crédit Agricole CIB and Société Générale were among the four lead banks of a £2.45 billion financing for Virtus Data Centres in September, and Société Générale and Crédit Agricole CIB were among the coordinators of a $1.24 billion loan for an AirTrunk campus in Tokyo in March.

To be fair, French banks are not the biggest lenders to US data centers. Japanese, American and Canadian banks lead those tables. The risk for a French bank is not the size of its AI book. It is that the AI exposure sits on a balance sheet already tied to its sovereign: according to Fitch, as reported this week, French government bonds held directly are worth 15% to 30% of the banks' core capital, and 30% to 60% with their insurance arms. Fitch calls that exposure moderate and estimates that a one point widening of the French spread cuts their core capital ratios by less than 0.1 point. Here is how faster depreciation reaches that balance sheet.

The ratios assume a five or six-year life. A data center loan is sized on the cash flows of the site, and its loan-to-value on what the site owns. With a five or six-year hardware life, both look comfortable until the first contract renewal at a lower price. Nothing shows in the accounts before that.

The collateral is the tenant. On the assumptions above, GPU-backed debt is under water against the hardware itself within a year or two. What protects the lender is the rental contract, and behind several of these contracts stands a tenant that burns cash, OpenAI first.

The big lenders are filling up. The $38 billion of construction loans for two Oracle-linked Vantage campuses took months to sell and, according to a Bloomberg report in October, banks' internal limits on concentrated AI exposure are becoming relevant. My guess is that the next loans go to the banks that still have room.

Dollars cost a French bank more. The US loans are in dollars. A French bank funds them in a market where its sovereign now pays more than Italy and where the euro-dollar basis, the extra cost of swapping euros into dollars, has turned against European borrowers.

The risk may not go far. Banks shift part of their credit risk through significant risk transfers, selling the risk of a loan portfolio to outside investors while keeping the loans, and Brussels is reviewing the rules when insurers take that risk. A French lender was reported in July to be preparing one on data center loans. Part of that risk may end up with insurers, and many French insurers belong to banking groups that hold the same government bonds.

Europe has no growth cushion. European shares fall more than world shares when oil rises, and their thin technology weight offers no growth story to offset it. French bank shares trade inside that index, with their sovereign on their books.

What would prove me wrong

First, the 2026 rebound. If power stays the bottleneck for years, old chips keep renting at decent prices and their economic life stretches toward the book life. The Democrats' sweep of Tuesday's midterm elections could slow permits for new sites and keep power scarce for longer, which would help.

Second, contracts. Microsoft's chief financial officer said in January that a lot of the GPUs it buys are already contracted for most of their useful life. That protects Microsoft. It does not remove the risk, it moves it to the tenant, and Steve Eisman, who disagrees with Burry on depreciation, says the real danger is the dependence on OpenAI and Anthropic.

Third, the site may be worth more than its chips. A campus with its power already connected keeps value even when its GPUs do not.

None of this changes the arithmetic of a GPU bought at today's prices and depreciated over six years. It changes who bears the loss, and when.

What I watch

• The H100, H200 and Blackwell rental indices, and how fast the next generation pushes them down.

• Any change of useful life in the next annual reports, due from late January.

• Residual value guarantees, lease commitments and stock sales in hyperscaler disclosures.

• French bank credit default swaps, the price of insurance against their default, the OAT-Bund spread and the next rating reviews of France.

My thanks to David Wolf, who built the model, tested my thesis with an engineer's eye and let me take his work apart in public.

Market commentary for educational purposes. This is not investment advice and not a recommendation to buy or sell anything. Past moves do not predict future ones, and leveraged products can lose more than the amount invested.

How to cite: Djellal Djouad, CrossVol Research, "The AI Depreciation Bomb: The Chips Last Three Years, the Loans Last Ten", 5 November 2026. https://newsletter.crossvolresearch.com/p/ai-bubble-gpu-depreciation-data-centers-french-banks

Market commentary for educational purposes. This is not investment advice and not a recommendation to buy or sell anything. Levels are as of the dates and times given, markets move fast, and currencies, bonds and derivatives can lose far more than you expect.

A note to readers: from next week, CrossVol Research becomes a paid publication, at $15 a month. More at the end of this piece.

Since the end of September I have told clients the same thing, several times and again this week: sell every EUR/JPY rally. The euro closed Friday at 175.78 yen. It was 178.35 on September 30 and 185.59 on August 31. It closed at 175.72 on October 23 and again on October 27, its lowest close in a year. Every bounce since September 30 has stalled below that day's close.

My view has not changed. The euro will keep falling, and mostly against the yen. What changed this week is where the stress shows up. The French fiscal scare started in government bonds, the OAT market I covered on October 25, in a piece whose figures FXStreet picked up. It has now reached one of the few markets that show who needs dollars and how badly: the cross-currency basis.

Options on EUR/USD do not confirm an imminent fall. Options on EUR/JPY still lean toward it. That gap is the trade, and the market data below, checked on Bloomberg after Friday's close, explain why.

The short version

• France's risk premium has nearly doubled. The 10-year OAT-Bund spread went from 85 bp on August 31 to 165 bp on October 30, after a closing high of 174 bp on October 22. France now pays 46 bp more than Italy to borrow for ten years.

• The stress has reached dollar funding. Measured against overnight rates, the 3-month EUR/USD cross-currency basis went from zero at the end of August to minus 2.5 bp at the end of September and minus 4.8 bp on October 23. It closed the week at minus 3.6 bp. On Bloomberg's dealer quotes, sterling's basis barely moved.

• The euro is weakest against the yen. Since August 31 the euro has lost 5.3% against the yen and 3.9% against the dollar. One-month EUR/JPY options charge 1.80 vol points more for protection against a falling euro than for a rising one. On EUR/USD the gap is 0.25.

• Real money is still long the euro. Asset managers held a net long of 173,249 euro futures on October 27, about €21.7 billion, down from 263,253 on September 1. Hedge funds are short and adding. In the yen, asset managers swung from 24,521 contracts short to 91,191 long.

• The chart agrees. EUR/JPY trades below its 50, 100 and 200-day moving averages, and the 50-day has crossed below the other two.

• What would change my mind: an OAT-Bund spread back below 150 bp, a 3-month basis back above zero against overnight rates, or a EUR/JPY close above its 50-day average, now 179.30.

Why is the euro falling? France's risk premium nearly doubled in two months

A currency also carries the credit of the governments that borrow in it. When investors start to doubt the second largest of them, the euro pays. The cleanest gauge of French risk is the OAT-Bund spread: the gap between the yield on 10-year French government bonds, called OATs, and the yield on 10-year German Bunds, the euro area's safe-haven bond.

On Bloomberg's generic 10-year yields, France closed October 30 at 4.972% and Germany at 3.322%, a spread of 165 bp. It was 85 bp on August 31 and 127 bp on September 30. It closed at 174 bp on October 22, and in the same week it reached 178 bp during the session, its highest since the 2011 euro debt crisis. Italy's 10-year yield closed at 4.513%. France pays 46 bp more than Italy.

The move came from both sides. In October the French 10-year yield rose 11.5 bp while the German 10-year fell 26.4 bp. Part of the Bund rally came with global bonds, which also rallied in October. The divergence is the point: French yields rose while German yields fell, inside the same currency. That is what a risk premium looks like.

The rating agencies have noticed. On Friday evening Fitch revised France's outlook to negative and kept the rating at A+, the grade S&P also gives it. Moody's kept France at Aa3 with a negative outlook on October 23. The background is the one I described on October 25: a deficit of 5.4% of GDP this year, no majority for the 2027 budget, and a presidential election in about six months.

The euro has traded this risk closely. Since August 31, the correlation between the daily closes of the OAT-Bund spread and EUR/USD is minus 0.94. A correlation of levels flatters two series that trend at the same time, so here is the stricter test: on daily changes it is minus 0.38. That is still a meaningful link for a currency pair that moves on dozens of drivers. Against EUR/JPY, shown below, the two figures are minus 0.79 and minus 0.24.

The stress has reached dollar funding

The second step matters more for the months ahead. Conks, whose work on dollar funding is among the most careful I read, flagged it this week: the French scare now shows up in the cross-currency basis. I checked it on Bloomberg.

What the basis is. A European bank or company that needs dollars can borrow them directly or swap euros into dollars, a route used on a huge scale. It hands over euros today, receives dollars, and reverses the exchange later at a price fixed today. In a frictionless market this swap would cost exactly the difference between euro and dollar interest rates. That rule is covered interest parity, which the Bank for International Settlements called "the closest thing to a physical law in international finance." The cross-currency basis measures how far the market is from it. When the basis is negative, whoever swaps euros into dollars accepts less than the euro interest rate to get them. Dollars carry a premium, and the more negative the basis, the higher the premium.

What Bloomberg shows. I computed the 3-month basis against overnight rates from Bloomberg's EUR/USD forward points and the 3-month SOFR and €STR swap rates. This is the measure closest to the pure dollar premium. It was zero at the end of August, at minus 0.03 bp. It turned negative in the last week of September, as the OAT spread jumped, and closed September at minus 2.5 bp. It reached minus 4.8 bp on October 23, the day after the OAT spread's closing high, and ended the week at minus 3.6 bp. These are end-of-day quotes from several screens, so a basis point of day-to-day noise is normal.

Two cautions, because precision matters here.

First, these are small numbers. Minus 4 bp is a first crack in the plumbing.

Second, the 3-month tenor now spans the end of the year, when banks shrink their balance sheets and dollar funding usually gets dearer around the year end. Part of the move is that calendar effect. The 1-month basis, which does not reach December 31 yet, was still positive on Friday, at plus 1.8 bp. But the turn did not wait for the calendar. The basis first dipped on September 23, the day the OAT spread jumped 6 bp, and it was already at minus 2.7 bp on September 28, before the 3-month window crossed December 31.

Is it the euro or the dollar? The test is to compare currencies. Bloomberg also carries dealer quotes for the standard 3-month basis swap. That contract is set against 3-month Euribor on the euro side rather than overnight rates, so its level sits higher, but its moves tell the same story. The euro quote fell from 14.5 bp on September 30 to 8.1 bp on October 30 on TradData's quotes, and from 12.7 bp to 8.6 bp on Parameta's. The sterling quote stayed put, at 14.5 bp then and 14.8 bp now, and the yen quote slipped from 2.5 bp to 0.2 bp (Parameta quotes for both). The euro moved most.

The same data hold one more clue. Euribor, the rate at which European banks lend to each other for three months, rose against the 3-month €STR swap rate, which tracks the expected overnight rate. The gap went from 9.6 bp on September 30 to 13.2 bp on October 30, after 16.3 bp on October 22. It was wider still in early September, so this is a hint rather than proof, but it points the same way: European banks are charging each other a little more for term money.

So this is not a global dollar squeeze. It is a relative bid for dollars coming from Europe. Conks expects dollar abundance to pull the short end of the basis back up, and that may well happen. Until it does, the basis tells us the French scare has moved from bond desks to funding desks. Funding desks are where forced sales of euros start.

Why the euro falls most against the yen

If Europe is the problem, why not simply sell EUR/USD? Because the yen has three things going for it that the dollar does not, and a fourth, carry, is turning its way.

The trend. Since August 31, EUR/JPY is down 5.3% and EUR/USD 3.9%. The dollar has lost ground against the yen too: USD/JPY went from 159.74 to 157.56. Since September 30 the euro has fallen by about the same amount against both, 1.4% against the yen and 1.5% against the dollar, but the yen leg sits on the longer and cleaner trend.

The carry. The cushion for euro holders is thin. One-month €STR swaps yield 2.44%, against the Bank of Japan's policy rate of 1.25%, so a long EUR/JPY position earns about 1.2 points a year, and the 1.4% fall since September 30 has already wiped out more than a year of it. Selling EUR/JPY does cost that carry, while selling EUR/USD earns about 1.5 points. But the cost is shrinking. The Bank of Japan held on October 30, one board member voted for a hike, the Bank repeated that it will keep raising rates, and swaps price a December hike at 81%.

The flows. Japanese investors were net sellers of foreign bonds for six weeks in a row through October 24, about ¥3.5 trillion in total, according to Japan's Ministry of Finance. The pace slowed each week, but the money is still leaving foreign bonds.

The options. A risk reversal compares the implied volatility of an option that pays if a currency pair falls with one that pays if it rises, both equally far out of the money (25 delta, in market terms). It is quoted in volatility points, call minus put. Negative means protection against a fall costs more. On Friday the one-month 25-delta risk reversal was minus 1.80 on EUR/JPY and minus 0.25 on EUR/USD. One-month at-the-money implied volatility was 7.31% on EUR/JPY and 5.84% on EUR/USD. The premium for euro downside over euro upside is seven times larger against the yen than against the dollar.

What options say, and why that is the opportunity

Options on EUR/USD are not priced for an imminent euro drop, and that deserves a careful reading.

On October 1, the day the OAT spread jumped 14 bp, from 127 bp to 141 bp, the one-month EUR/USD risk reversal fell to minus 1.47 and one-month implied volatility rose to 6.96%. Since then both have relaxed, to minus 0.25 and 5.84% on Friday. Yet the OAT spread is 24 bp wider than it was on October 1. The bond and funding markets kept the stress. The EUR/USD options market let it go. The 3-month risk reversal tells the same story, from minus 1.31 on October 1 to minus 0.52.

EUR/JPY options kept more of their skew. Since September 30 the one-month risk reversal went from minus 2.44 to minus 1.80, keeping about three quarters of its level, while the EUR/USD one went from minus 0.81 to minus 0.25. EUR/JPY implied volatility fell from 8.52% to 7.31%. The premium for euro downside against the yen is still there.

This is why I keep selling the rallies. The market is not paying up yet for the scenario where French stress keeps moving into funding. If the basis keeps falling, the repricing of euro downside in options is the next step, and it would most likely start where the skew already leans, in EUR/JPY.

Who still has to sell euros

The weekly positioning report of the Commodity Futures Trading Commission (CFTC) splits futures holders into categories. Two matter here: asset managers, the slow real money, and leveraged funds, mostly hedge funds.

In euro futures, asset managers held a net long of 173,249 contracts on October 27. That was 220,603 on September 29 and 263,253 on September 1. They have cut 90,004 contracts in eight weeks, yet they still hold about €21.7 billion of net long euro exposure. Leveraged funds were net short 71,574 contracts, against 39,265 on September 29. They have almost doubled their short in four weeks.

In yen futures the picture is the mirror image. Asset managers moved from a net short of 24,521 contracts on September 1 to a net long of 91,191 on October 27, about ¥1.14 trillion. Leveraged funds cut their short from 102,188 contracts to 14,630.

The euro long is the crowded position that is still unwinding. Each week it shrinks, someone has sold euros. The yen side has real money behind it. The risk on that side is the reverse: a yen long that keeps growing can unwind fast if the Bank of Japan disappoints in December.

The chart: EUR/JPY below every average

The Bloomberg chart at the top of this piece shows the full year. EUR/JPY peaked at 187.60 on April 15 and averaged 182.95 over the twelve months. It broke down at the start of September and has stayed below its 50-day moving average since. The 50-day average, now 179.30, has crossed below the 100-day (182.01) and the 200-day (183.25), the configuration chartists call a death cross.

The October rallies are the part I watch most closely. EUR/JPY bounced to a close of 178.00 on October 6 and of 177.73 on October 19 and 20. Both stalled below the September 30 close of 178.35. Lower highs, then a new low at 175.72: that is a trend.

Levels and what would change my mind

EUR/JPY. The October rallies stopped at 178.00 and 177.73, below the September 30 close of 178.35. That zone, then the 50-day average at 179.30, is where I have been selling. A close above 179.30 would put EUR/JPY back above its 50-day average and I would step aside. On the downside, 175.72, last week's close, is the one-year low.

OAT-Bund spread. A close above the October 22 closing high of 174 bp would mark the next leg. A close back below 150 bp would mean the political premium is leaking out, as I wrote on October 25.

3-month basis against overnight rates. A close below minus 5 bp outside a quarter end would show the funding stress spreading. Back above zero, it is fading.

EUR/USD options. A one-month risk reversal back below minus 1 would show the options market catching up with bonds and funding.

Positioning. The next CFTC report, for positions on November 3, comes out on Friday, November 6. Each weekly cut in the asset managers' euro long is euro supply.

FAQ

Why is the euro falling? Because investors are charging France a growing risk premium. The 10-year OAT-Bund spread nearly doubled, from 85 bp on August 31 to 165 bp on October 30, and France now pays more than Italy to borrow. The stress has spread to dollar funding, where the 3-month EUR/USD cross-currency basis, measured against overnight rates, turned negative.

Why is the euro falling against the yen? Holding euros against yen pays only about 1.2 points a year, the Bank of Japan is moving toward a hike, Japanese investors have been selling foreign bonds, and asset managers have swung to long yen. Options price far more euro downside against the yen than against the dollar.

What is the cross-currency basis? It is the extra cost, or saving, of getting dollars by swapping another currency rather than borrowing them directly. A negative EUR/USD basis means Europeans pay a premium for dollars. Measured against overnight rates, it went from zero at the end of August to minus 4.8 bp on October 23.

Do options confirm a euro sell-off? Not on EUR/USD, where the one-month risk reversal is back to minus 0.25 and implied volatility is 5.84%. Partly on EUR/JPY, where the risk reversal is still minus 1.80, seven times the EUR/USD reading.

What would stop the euro's slide? An OAT-Bund spread back below 150 bp, a 3-month basis back above zero against overnight rates, or a EUR/JPY close above its 50-day average at 179.30.

A note to readers. From next week, CrossVol Research becomes a paid publication, at $15 a month. I will give the details in a short note next week, including what stays free and what is for paid subscribers. Thank you to everyone who has read and shared these pieces, and to those who have already pledged to support them.

This article is for educational purposes only. It is not investment advice, a solicitation, or an offer to buy or sell any security or derivative. Levels are as of the dates and times stated and can change quickly. Currencies, bonds, options and futures carry a high risk of loss. The move in the cross-currency basis was flagged this week by Conks. The measurements, the reading and the views here are the author's own.

Market commentary for educational purposes. This is not investment advice and not a recommendation to buy or sell anything. Levels are as of the dates and times given, markets move fast, and bonds, credit, equities, currencies and derivatives can lose far more than you expect.

A note to readers: from next week, CrossVol Research becomes a paid publication, at $15 a month. More at the end of this piece.

For a month the market has been pricing a US slowdown one asset at a time. Credit went first, in the bonds of the companies that borrowed to build artificial intelligence. The AI stocks followed, on their results. Defensive sectors took the money, and Treasuries began to hedge equities again. One asset had held its lows: gold. It slipped through October and fell on all three sell off days, but it did not break. On Friday it broke. Spot gold closed at $3,874.62, down $146.90 on the day, below its one year low of $3,932.09 of November 4, 2025, and 28.5% below its record close of $5,417.21 on January 28.

That was the last piece, and it was sold for cash, which is how a slowdown starts. The market has started to price a US slowdown, led by AI, and what makes this one dangerous is what comes with it.

• Inflation is not under control. The PCE price index rose 3.6% in the year to September, and the energy shock behind it is still running.

• The Fed held on Wednesday with two votes for a hike, and futures still price a hike in December.

• The Bank of Japan held on Friday, but with one vote for a hike and a promise to keep raising rates, and swaps now price a December hike at 81%. That puts the yen carry trade on notice, and Japanese money has been leaving foreign bonds for six weeks.

• Credit is under stress where the leverage is, in AI borrowers and private credit, and after Friday's close S&P revised the outlook on Meta's AA- rating to negative, and Fitch did the same for France.

• Hedge funds are cutting their gross exposure. They came into October with gross leverage near the top of its five year range and a record share of their books in the Magnificent 7, and they are now shrinking those books, longs and shorts together.

• The hedge funds' Treasury basis trade stands in the way of a bond market that has become a haven again, pushed by trend followers covering their shorts and, I think, by pension money.

• On Tuesday the United States votes, and prediction markets give Congress to the Democrats. If President Trump loses it, I expect him to blame them for the slowdown.

Put together, these point to one risk for 2027: stagflation, an economy that stalls while inflation stays high. It is the worst mix for a central bank, because the tool that fights one feeds the other.

This piece takes them one by one, with the data, what our scanner confirms, what it does not confirm yet, and the levels I watch.

Where we were on October 13

On October 13 I mapped October 2026 onto the five acts of "The Script," Andy Constan's play about the end of a tightening cycle. The acts are his. Andy runs Damped Spring Advisors and writes here as . In his report of July 10, 2023, the hikes of Act 1 "don't achieve goal," in Act 2 "long end yields rise to new highs," in Act 3 higher yields compress valuations, in Act 4 the tightening hits demand and earnings contract, and in Act 5, "as equities sell off, companies fire workers." The reading of 2026 that follows is mine, not his.

My verdict then was: "early Act 4, through credit. The indices will see it last." Seventeen days later, here is what has played.

Act 2 reversed. The 10 year Treasury yield closed at 5.31% on October 5, its highest close since May 2002, and at 4.98% on October 29. The New York Fed's estimate of the 10 year term premium, the extra yield investors demand to lock their money up for ten years instead of rolling short bills, fell from 98.5 basis points on October 7 to 69.1 on October 28. A basis point, or bp, is one hundredth of a percentage point.

Act 4 moved from credit to stocks, inside AI. Oracle's five year credit default swap, the annual price of insurance against its default, set a record of 329 bp on October 20. CoreWeave's peaked at 1,163 bp on October 22. Then came the results. Since October 13 Alphabet has lost 20.9% and Meta 19.3%, Meta after its spending plan on October 22, Alphabet after its results on October 28.

The indices are only starting to see it. On October 29 the S&P 500 ETF (SPY) was only 1.1% below its October 13 close and 1.2% below its closing high of October 6. The equal weighted S&P 500 ETF (RSP), in which every stock counts the same, had gained 3.2% since October 13. On Friday the cap weighted index moved, after Amazon and Intel: the S&P 500 closed 1.82% lower and the Nasdaq 100 3.46% lower, and the VIX closed at 19.42 after touching 23.93. SPY ended the week 3.0% below its October 6 high, while RSP was still 2.3% above its October 13 close.

Act 5 has not started. Companies are not firing in large numbers yet: announced job cuts are down 39% from last year and jobless claims are near their lows. That is the part of the play that 2027 will decide, and I come back to it at the end.

Why gold is falling: the last piece

Gold is the last piece because of the order in which things break. In a slowdown the market does not sell gold because growth is weak. It sells gold because it needs cash, and gold, the great winner of 2025, is where the cash is. Gold slipped all through October, but it held its lows until Friday.

The levels at Friday's close. Spot gold ended below its 50, 100 and 200 day averages, at $4,254.37, $4,213.45 and $4,491.08. COMEX December gold futures closed at $3,884.90, below the contract's previous 2026 closing low of $4,007.90, set on October 23. The largest gold ETF, GLD, lost 3.5% to $355.78, 28.3% below its peak of January 29. It had gained 63.7% in 2025, against 16.4% for the S&P 500 ETF, and winners are what funds sell first when they need cash.

What did not sell it is just as telling. In the textbook, gold falls when real yields rise, because a bar of gold pays nothing and competes with the return of inflation protected Treasuries. A real yield is a Treasury yield after inflation, read on those inflation protected bonds, the TIPS. Since the peak in rates on October 5, the 10 year real yield fell from 2.95% to 2.67% on October 28, the inflation priced for the next five years rose from 2.36% to 2.47%, the dollar index eased from 102.10 to 101.85, and GLD still lost 2.9% to October 29, before Friday's 3.5% fall. Every textbook driver pointed up. Gold went down.

It went down with stocks. In 2026 GLD fell on 16 of the 21 sessions when the S&P 500 ETF lost 1% or more. In October, Treasuries rose on all four of those days and gold fell on all four, Friday included: SPY lost 1.8%, the long Treasury ETF (TLT) gained 0.4% and GLD lost 3.5%. On October 22, the day the 10 year yield fell 17 bp and the real yield 12 bp, GLD lost 2.6%.

Positioning explains the rest. On our trend model, the trend followers, the CTAs, are short gold on all three horizons, close to their most short of the year. Managed money, the CFTC's category for speculative funds, was still net long 104,206 contracts on October 20, worth about $43.2 billion, after eight weekly declines in a row. That long is what remains to be sold if the liquidation goes on.

On October 25 I wrote: "The signal I will watch most closely besides their results is gold: if it keeps falling with yields, liquidity rather than growth is driving the tape." It kept falling with yields. Liquidity is driving the tape. That does not contradict the slowdown. It is how the first phase of one looks.

Two precedents. In 2008, COMEX gold futures lost 29.7% between March 18 and November 13, from $1,003.20 to $704.90, as the recession deepened. They were back at $1,001.80 by February 20, 2009, after the Fed had cut its rate to zero and started to buy bonds. In March 2020, gold lost 11.8% in seven sessions, from $1,674.50 on March 9 to $1,477.30 on March 18, in the dash for cash, then set a record of $2,069.40 on August 6. In both cases gold turned around the time the Fed went all in: in 2008 it bottomed twelve days before the Fed announced its first bond purchases, and in 2020 three days after it cut to zero and restarted them.

This time the Fed is priced to hike in December. That is why gold was the last piece, and why it matters for 2027. The market is pricing a slowdown without a Fed put, the expectation that the central bank will cut rates to rescue markets.

What our scanner says

Our scanner runs every day on about 300 instruments: moves measured in standard deviations, trend signals, bank thematic baskets, credit, volatility and positioning. A standard deviation, or σ, measures the usual size of a move: a move of 2 σ or more, up or down, happens about one month in twenty if moves are normally distributed. I asked the scanner one question: does it confirm that the market is pricing a US slowdown? Its answer is yes for the first half of my thesis, and not yet for the second.

What it confirms:

Rates price slower growth. The 2 year Treasury yield, the one most tied to the Fed's path, fell from 4.92% on September 28 to 4.61% on October 29. The market also removed more than a third of the hikes it priced for the next three years. SOFR is the overnight rate on loans backed by Treasuries, and its futures price where it will be. The contract that covers mid 2029 implied a rate 92 bp above the overnight rate of the day on September 30, and 56 bp above it on October 29.

Defensives rallied by about two standard deviations. In the month to October 29, Goldman Sachs' defensives basket gained 8.5% and its bond proxies basket 8.1%, both about 2.1 σ, while its cyclicals versus defensives pair lost 8.8%, a 1.4 σ move. Utilities gained 8.6%, consumer staples 7.7% and health care 7.8%, while technology lost 3.0%.

AI trades as a growth risk. Goldman's AI beneficiaries versus AI at risk pair lost 17.1% in a month. The 50 day correlation between the relative performance of semiconductors (SMH against RSP) and long Treasuries (TLT) fell from plus 5% a month ago to minus 31%. AI underperformance now comes with bond rallies, which is the pattern of a growth scare.

Bonds are hedging again. The 50 day correlation between the S&P 500 and long Treasuries fell from 46.7% on October 13 to 16.9% on October 29. It is still positive, so the hedge is returning rather than back, but Treasuries rose on all four October sell offs, Friday included.

Credit sits at one year extremes where the leverage is. Oracle's and CoreWeave's CDS are in the top 5% of their year, and the US high yield spread touched its 99th percentile on October 22.

Positioning has turned defensive. Short interest in the Nasdaq 100 ETF (QQQ) was 82 million shares on October 15, above 96.6% of the readings of the last five years. October 22 was the largest degrossing day of the year in our hedge fund monitor, degrossing meaning funds cutting longs and shorts at the same time.

What it does not confirm yet:

A recession. Polymarket's contract on a US recession before the end of 2026 trades at 7.5%, against 43.5% a year ago. It has two months left to run, so it says more about the calendar than about 2027.

The index and its breadth. At Thursday's close the S&P 500 ETF was 1.2% below its high. 64.8% of the S&P 500 members were above their 50 day average on October 29, against 24.2% on September 29.

Copper. Copper futures were flat in the month to October 29, at $6.60 a pound, 4% below their September 9 high. A global slowdown would push them down. The copper to gold ratio is close to its one year high, but only because gold fell.

Inflation expectations. The 5 year breakeven, the inflation rate priced by the bond market for the next five years, rose 13 bp in a month to 2.47%. A disinflationary slowdown would push it down.

Broad credit. Bloomberg's high yield index was at 323 bp on October 29, below the 350 bp I named on October 13 as the next marker.

My reading: rates, sector rotation, AI and AI credit price a US slowdown. The index, the recession markets and broad credit do not, yet. That is the order I described on October 13. The indices see it last.

The AI trade is being de-crowded

Since Thursday the market has stopped rotating out of AI and started to empty it. On Thursday Alphabet fell 12.6% after its results, and Amazon and Intel reported after the close. On Friday the selling reached the whole chain. Intel lost 28.6%. The Philadelphia semiconductor index closed 7.2% lower, at 10,688.80, and the Nasdaq 100 3.5% lower, at 28,384.85. During the session the Nasdaq 100 had fallen 10% from its October high, the usual definition of a correction, and it closed 9.1% below it. Over the two sessions from Wednesday's close, Alphabet lost 13.9%, AMD 12.7%, the semiconductor ETF (SMH) 8.0%, Amazon 5.6%, Nvidia 5.0% and Microsoft 1.7%, while Meta gained 0.4%. The Nasdaq 100 ETF (QQQ) lost 4.8% while the equal weighted S&P 500 ETF (RSP) gained 0.4%. On Friday Goldman Sachs' AI beneficiaries versus AI at risk pair lost 6.1%, to 271.40, 3.7% above its October 23 low and 53.5% below its June 22 high. Goldman's Hedge Fund VIP basket, the stocks most widely held by hedge funds, lost 3.8%, to 15.5% below its June 30 high. Morgan Stanley's momentum pair lost 2.3%.

Crowded is the word. On the Goldman prime brokerage figures in our October 25 review, hedge funds held a record 22% of their exposure in the Magnificent 7. A trade that crowded is not sold on news alone. It is sold by the funds that have to cut their gross exposure.

Hedge funds are cutting gross exposure. Gross exposure is the size of a fund's book, its longs plus its shorts. Net exposure is the difference, its bet on the direction of the market. Degrossing cuts both at once, selling longs and buying back shorts, to shrink the risk without changing the bet. It started before the results. In the week to October 15, hedge funds sold technology stocks at the fastest rate in at least a decade, on Goldman Sachs' prime brokerage data. Goldman's figures then showed degrossing on 7 of 8 sessions, and October 22 was the fifth largest degrossing in its data, almost entirely long selling, and the strongest of the year in our hedge fund monitor.

Yet the books stayed big. Gross leverage, longs plus shorts against capital, was still 312.3% in Goldman's prime book, in the 96th percentile of five years, while the net leverage of US long short funds had fallen to 43.4%, the lowest in at least five years. The funds were hedged, not small, and a book that large has a long way to shrink. It is shrinking now. Our monitor flagged degrossing during Friday's session, though not at the close, after the strongest degrossing day of the year on October 22. On our estimate, gross leverage has come down since October 26 while net leverage has held steady: the funds are making their books smaller, not turning bearish.

Why now? Because volatility sets the size of a book. When volatility rises, the same gross carries more risk, and risk managers cut it to stay inside their limits, fastest at the multi-manager platforms, where the limits are tight. That is why degrossing feeds on itself: the selling raises volatility, which forces more selling, first in the stocks the funds own most. On Friday, as the VIP basket fell 3.8%, Goldman's pair of hedge fund favourites against the most shorted stocks lost only 0.5%, both legs falling together. And it does not stay in stocks. The largest multi-strategy funds run equity books and relative value trades in rates side by side, under one risk budget. When they cut gross, the Treasury basis trade is part of the gross.

For Act 4 the order matters. Credit went first, the AI leaders took the hit on their results, and now the funds that own them are reducing. After the close the story came back to credit, with S&P's negative outlook on Meta. The cap weighted index is the last step, and on Friday it started to move.

Rates: the haven is back

The 10 year yield closed Friday at 4.93% on Bloomberg, 38 bp below its October 5 high and below its 50 day average of 5.00%. The 30 year closed at 5.31%, against 5.67% on October 7. On the Treasury's own closing curve, from October 5 to October 29 the 2 year fell 23 bp, the 10 year 33 bp and the 30 year 34 bp.

Long yields fell more than short ones. That is a bull flattener: a rally in which the long end does most of the work, because the haven bid lands on duration while the front end, held up by a Fed that is still priced to hike, falls less. On the Treasury's curve the gap between the 10 year and the 2 year yields went from 47 bp on October 5 to 37 bp on October 29. On Bloomberg's generic yields it closed Friday at 33.7 bp, from 33.5 on Thursday.

Who bought? On the crash days themselves, the front end beat the long end. On October 22, when the semiconductor index lost 6.7%, the 2 year fell 16 bp and the 30 year only 13 bp, with the 10 year down 17 bp. In our intraday test during US hours, on October 16 and on October 22, the yield implied by the 2 year note future fell more than that of the ultra bond future. That fingerprint is a Fed repricing: when stocks fall, the market takes hikes out.

Between those days, the long end had a slower buyer, in my view: the pension money I described on October 15, which would show over weeks, not in a single session. At the end of September Goldman Sachs expected US pension funds to sell about $33 billion of equities around quarter end, at the 98th percentile of its estimates since 2000, with the proceeds going to bonds, and JPMorgan Asset Management expected the full effect in the early days of the fourth quarter. The 30 year peaked a week into the quarter, at 5.67% on October 7. I cannot show the pension flows yet, and our intraday test found the front end, not the long end, leading on both big down days since October 15. The 100 largest US corporate plans were 114.5% funded on September 30 according to Milliman, the best since June 2001, and at that level their glide path sells equities and buys long bonds. Milliman's October index, due in early November, will show where funded ratios stand, and the Treasury's October STRIPS data, due November 5, how much long duration was bought.

The trend followers added to the front end bid. Nomura's estimates, reported on October 24, showed that they had covered about $68 billion of bond futures and $38 billion of short rate futures in two weeks. Our trend model tells the same story: the largest single session of buying was October 22, mostly in the 2 year note, and very little in the long end. Their short term signals have turned long across the curve. Their medium and long term signals are still short.

That leaves a second stage loaded. Nomura's estimates, reported the same day, showed that a further 2% rally in bond futures would force about $256 billion of global bond buying by trend followers, and a 2% rally in short rate futures about $264 billion more, against only about $53 billion and $33 billion of selling on a 2% sell off. Our own model points the same way: the trend followers would buy far more on a rally than they would sell on a drop.

The basis trade sits in the way

The basis trade is one of the largest leveraged positions in the Treasury market. A hedge fund buys a Treasury bond, borrows the cash overnight in the repo market with the bond as collateral, and sells the matching futures contract. Asset managers like to hold long futures, which give them duration without tying up cash, so futures trade slightly rich to the bonds they track. The hedge fund collects that small gap until delivery. Because the gap is thin, the trade runs with leverage, often fifty times.

In the CFTC's weekly report on Treasury futures, leveraged funds were short about $474 million per basis point on October 20, and asset managers long about $652 million. That measure is the DV01, the dollar change in value for a one basis point move in yields. The leveraged short is 25% below its record of $632 million of November 18, 2025, but it has been growing again since October 6, when it stood at $458 million. On the 10 year note future alone, leveraged funds were short 2.1 million contracts, more than in 86.5% of the weeks of the last five years. The report for October 27, published on Friday, showed them adding about 120,000 contracts, to 2.23 million, shorter than in 90% of the weeks of the last five years. In Bloomberg's summary of the same report, speculators sold the front end ahead of the Fed, adding 153,000 contracts of shorts in 2 year notes, while buying 5 year notes, ultra 10 year notes and bonds.

The trend followers are a small part of that short, on our model. Most of the rest is the basis trade and other relative value positions.

On September 22 I wrote that the basis book had shrunk, to about $900 billion from $1.26 trillion at the start of the year, while its long end had grown: "The number got smaller. The risk got sharper." A blog post by two researchers of the Office of Financial Research, the Treasury's research office, estimated in August that hedge funds held about $2.0 trillion of cash Treasuries at the end of 2025, a record 7% of the marketable debt, and warned: "If repo rates spike or futures margin requirements increase abruptly, hedge funds may be forced to unwind positions rapidly." Here is why a haven bid is the danger now. The trade is hedged against the level of yields, not against the way they get there, and three things break it.

Futures that run ahead of cash. When the haven bid comes through futures, as it does when trend followers cover or asset managers add duration, futures rise faster than the bonds they hedge. The fund that is long the bond and short the future loses on the gap. It pays the full loss on its short futures in cash every day, as variation margin, while the gain on the bond comes back through the repo lender, which revalues the collateral every day but keeps a haircut. That is what the trend followers' buying of October 22 does, and it came mostly in the 2 and 5 year futures, where leveraged funds hold their largest shorts in face value: 1.29 million 2 year and 2.12 million 5 year contracts on October 20, about $470 billion.

Margins. Exchanges raise the margin required on futures when volatility jumps. The MOVE index, the implied volatility of Treasury options, went from 95.9 to 106.8 on October 22, fell back to 93.5 by October 29, and jumped to 100.2 on Friday.

Funding. The trade is financed overnight. If repo rates rise or lenders demand bigger haircuts, the carry disappears and the position has to shrink. SOFR printed 3.94% on October 28 and 29, near the top of its October range of 3.87% to 3.95%, and above the rate the Fed pays banks on their reserves, by more than usual.

The order of magnitude: with a short of $474 million per basis point, a fall of about 30 bp in yields across the curve, which is what October 5 to October 29 delivered, means about $14 billion of variation margin paid on the futures leg over those weeks, if the position had not changed. Most of it came back through the bonds, which are revalued every day in repo. The danger is the day the two legs stop matching: a squeeze in futures, a higher haircut or a spike in repo. In March 2020 the haven bid went to bills and futures while cash bonds were sold, the basis blew out, and the funds had to unwind into a market that could not absorb them. The Fed had to buy Treasuries on a scale never seen before.

Nothing in the bond market shows forced selling yet. The 30 year swap spread, the swap rate minus the Treasury yield, was minus 64.1 bp, inside its six month range of minus 79.9 to minus 59.2. Our hedge fund deleveraging gauge is high, a little below its peak of the year, after a sharp rise in October. What I watch: the next CFTC report, due on November 6, the Treasury's refunding announcement on November 4, which says how many more bonds it will sell, and SOFR against the rate on reserves into year end, when banks shrink their balance sheets.

Inflation is not cooperating, and the Fed knows it

The data first. The PCE price index rose 3.6% in the year to September, up from 3.4% in August. Core PCE, which excludes food and energy and is the Fed's preferred gauge, rose 3.1%. Consumer prices rose 3.7% in the year to September. The five year chart of core PCE shows the problem: the fall from the 2022 peak of 5.6% stalled in 2024, and the line has turned up since late 2025.

The source is energy. In a year, Brent crude has risen 72%, European gas (TTF) 207% and US diesel 71%. The January Brent contract, the front month since Friday's expiry, ended Friday's session at $112.33, against $110.80 on Thursday. The contract for delivery nine months ahead closed at $93.08 on Thursday, close to its one year high: the oil market prices expensive energy well into 2027. European gas closed at €103.19 per megawatt hour on Friday, up 4.7% on the day.

Oil is trading like a haven. On Friday, as AI stocks, gold and the hedge fund favourites were sold, two markets took the money: Treasuries and crude. The December Brent contract, which expired that day, traded as high as $118.00 and settled at $117.64, up $2.76, close to the year's highest close of $118.35 on March 31 and far above its one year average of $86.16. After the close, Bloomberg's Markets Live strategist Frank Monkam made the same observation: besides Treasuries, oil was arguably the only major market left to hide in, with the risk of a new escalation with Iran behind the bid. He also noted that the six month average of Brent's expiry spreads is near its highest in at least a decade: buyers pay up for barrels now rather than later, the sign of a market short of supply, not of demand. Bonds price the slowdown and oil prices the shortage. Together they are what stagflation looks like before it reaches the data. And they tie the Fed's hands. With inflation still elevated and energy adding to the upside risks, the same strategist wrote, the Fed has "considerably less room" to soften the blow.

What the bond market prices is much less. The 5 year breakeven, the inflation rate that equalizes a regular and an inflation protected Treasury, rose only 0.10 point in a year. The market priced the shock in real rates instead: the 10 year real yield rose 0.89 point. In other words, the bond market trusts the Fed to kill this inflation by keeping real rates high, and expects the slowdown to finish the job.

The breakeven is not asleep, though. At 2.47% at Friday's close it is 29 bp above its July 28 low of 2.18%, and below its May 4 high of 2.74%. And since October 5 the mix has turned: the 10 year real yield fell 28 bp while the 5 year breakeven rose 11 bp. Falling real yields with rising breakevens is how the bond market starts to doubt that the Fed will hold the line.

Wages are not the problem yet. The employment cost index rose 0.8% in the third quarter, in line with forecasts, after 0.9%, and 3.4% over a year. Prices paid by companies are. The ISM manufacturing prices index was 77.9 in September and is forecast at 77.1 for October. In services, the forecast is 74.5 after 74.0.

The Fed's choice. On Wednesday the Fed kept its target range at 3.75% to 4%, by 10 votes to 2. The two dissenters, Governor Lisa Cook and Cleveland Fed President Beth Hammack, wanted a quarter point hike. The statement says that activity "is expanding at a strong pace" and that "Inflation remains elevated." With the effective fed funds rate at 3.88% and PCE inflation at 3.6%, the real policy rate is about 0.3 point. That is barely restrictive, with energy near one year highs.

So why hold, with strong growth and elevated inflation? There are two readings. In the first, the Fed is behind the curve and knows it, which is what the two dissents say. In the second, the Fed sees the slowdown coming, in AI and in credit, and lets it do the work. Chair Warsh pointed that way, with a caveat. He said financial and credit conditions "have tightened for some businesses and households, much as we would expect," but that he was "unwilling and unable to call them tight." And: "You've seen credit spreads for certain assets widen out, gap out." This week the bond market chose the second reading. After the hold and a PCE print in line with forecasts, the 2 year yield fell 6 bp on Thursday, SOFR futures priced fewer hikes, and the 5 year breakeven eased.

But futures still price a hike in December. On Bloomberg's World Interest Rate Probability page at Friday's close, fed funds futures gave a 72.9% probability of a hike on December 9, and the implied rate peaks at 4.464% in September 2027, about 2.3 quarter point hikes above the effective rate of 3.88%. The Fed is not alone. Overnight index swaps, the contracts that price the average policy rate over a period, gave 82.3% to an ECB hike on December 17, 95.6% to a Bank of England hike on November 5, 80.3% to a Reserve Bank of Australia hike on November 3, and 80.9% to a Bank of Japan hike on December 18.

That is what an energy shock does. Almost every large central bank is priced to tighten while growth slows. Here is the bet I question. The bond market is betting the slowdown will do the Fed's job. If it does not, the Fed is late, with inflation at 3.6% and energy near one year highs, and 2027 starts with a central bank that has to choose between growth and prices.

Credit: stress where the leverage is

On October 13 I wrote that Act 4 would arrive through the credit that finances AI. It did, and it peaked a week later.

Bloomberg's US high yield index spread reached 338 bp on October 22, its widest of the year and 88 bp above its January 22 low of 250. The spread here is the OAS, the average extra yield over Treasuries after removing the value of the call options some bonds carry. It eased early this week, then widened again on Thursday, the day Alphabet fell 12.6%, to 323 bp. The investment grade spread is 87 bp, close to its one year high of 93 bp of March 16.

The stress sits where the leverage is.

AI borrowers. Oracle's five year CDS was 270.9 bp on Thursday, 3.4 times its level of a year ago, after its record of 329.1 bp on October 20. CoreWeave's was 1,016.7 bp, after 1,162.5 bp on October 22. Both are in the top 5% of their year. In a month the spread of Oracle's 4.95% bond of 2031 over Treasuries widened 61 bp to 232.9 bp, and that of CoreWeave's 9% bond of 2031 widened 237 bp to 971.3 bp.

Private credit. Blue Owl's two listed lending funds, OBDC and OTF, are business development companies, or BDCs, which lend to mid sized private firms. Both made one year closing lows on October 21, when OTF stood 34.5% below its January 5 high. They have bounced 6% to 9% since. Blue Owl itself is 37.6% below its 52 week high. In the unlisted funds, the exit door is narrow. In the second quarter, holders of Blue Owl's non traded technology lending fund (OTIC) asked to redeem 38.1% of their shares and received about 13% of what they asked (AltsWire). The third quarter tender, whose final results came on October 26, looked the same according to the Crowdfunded Wealth tracker, which compiles fund filings: 39% requested, 12.8% filled. These funds pay out pro rata at a 5% quarterly cap. It is not a suspension, but most of the money that wants out stays in.

Credit indices. On Thursday afternoon, the CDX investment grade index of North American credit default swaps traded at 60.1 bp, and Europe's high yield index, the iTraxx Crossover, at 312.3 bp.

What has not cracked: regional banks gained 4.3% in a month (KRE), and the broad high yield market is below the 350 bp I named on October 13 as the next marker. Credit stress is real on a one year view and concentrated in AI and private credit. It has not spread. The index, again, is the last to know.

The Bank of Japan sees it too. Its outlook report, published on Friday, lists "the profitability and financing of AI-related investment" and "developments in the foreign non-bank financial intermediary (NBFI) sector" among the financial risks it watches.

Why it matters for Act 4: lenders read cash flows, not stories. Alphabet reported about $9 billion of negative free cash flow and plans a "significant" increase in investment for 2027. When the largest AI spenders burn cash to build, the cost of the debt that finances the rest of the chain rises first.

Then the rating agencies. After the close on Friday, at 5:05 pm New York time, Bloomberg flashed that S&P had revised Meta's outlook to negative. The rating stays AA-, the fourth highest grade, and Moody's still rates Meta Aa3 with a stable outlook. An outlook change is not a downgrade. It is the agency's warning that one may come within six months to two years. But it shows how far Act 4 has travelled: what AI spending costs in cash is now a question for the credit rating of one of the strongest balance sheets in the world, not only for Oracle and CoreWeave.

And a sovereign. Seven minutes later, at 5:12 pm, Bloomberg flashed that Fitch had revised France's outlook to negative. Fitch keeps France at A+, the grade S&P also gives it, while Moody's has rated it Aa3 with a negative outlook since October 2025. I wrote on October 25 that France now pays more than Italy and Greece to borrow for ten years. On the same Friday evening, the agencies moved on the two places where debt pays for what income does not cover: AI spending and European budgets.

Japan: the Bank of Japan turns, and the carry trade listens

The Bank of Japan kept its policy rate at 1.25% on Friday, by 8 votes to 1. Board member Hajime Takata proposed a hike to 1.5%, arguing that with underlying inflation at the target the Bank must "step up the pace" toward the neutral rate. The majority judged underlying inflation "generally consistent with the price stability target of 2 percent" and wrote that "the Bank will continue to raise the policy interest rate." It also decided, unanimously, to sell its ETF holdings faster from January, about 860 billion yen a year instead of 330 billion. Governor Ueda named no month. The market did it for him: overnight index swaps price an 80.9% probability of a hike at the December 18 meeting, against about 57% before the decision according to investingLive, which may not use the same measure.

The yen firmed only a little. The dollar closed the week at 157.56 yen, against 157.71 at Thursday's close, after a low of 156.96 on Friday morning. The Nikkei 225 closed 2.12% lower, with Korea's Kospi down 3.48% and Taiwan down 2.49%: the AI sell off after Amazon's results came on top of the Bank of Japan.

The signal is in the bond market. The 10 year Japanese government bond (JGB) yield fell 6.4 bp to 2.782%, and the 30 year 6.7 bp to 3.850%. A hawkish turn usually lifts short yields more than long ones, and long yields can even fall if the market thinks the Bank will get ahead of inflation. Part of the move is that, part is the global haven bid, and part is money moving toward JGBs. On Bloomberg's generic yields, since its high of 3.125% on October 1, the 10 year JGB yield has fallen 34 bp, and the 30 year 40 bp since its high of 4.249% on October 6.

The Ministry of Finance's weekly flows show where it stands. Japanese investors were net sellers of foreign bonds for six weeks in a row through October 24, about 3.5 trillion yen in total, although the pace slowed each week, and Reuters reported in mid October that the selling came mostly from outside the life insurers. The buyers of JGBs have mostly been foreign: about 3.2 trillion yen in the four weeks to October 24. So money is leaving foreign bonds and money is moving into JGBs, but the large domestic holders, the life insurers, have not come home yet. A Bank of Japan that hikes in December is what could bring them.

Why money would come home. On September 25, in The Crash Fuse Is Lit, I wrote that "On a hedged basis, the Japanese institutional bid for US duration is effectively closed at current levels." A Japanese insurer that buys a 30 year Treasury and hedges the dollar pays the gap between US and Japanese short rates plus a cross currency basis, 2.93 points a year on September 25. In late September that left the hedged 30 year Treasury at 2.57% against 4.15% for the 30 year JGB. Both legs have moved since, but the arithmetic has not: at 5.31% for the Treasury and 3.85% for the JGB, a hedge cost close to 3 points still leaves the JGB about 1.5 points ahead, on my estimate. The insurers who do not hedge carry the currency risk instead, and a hawkish Bank of Japan is that risk. Japanese and Taiwanese insurers hedged only 41% of their foreign currency exposure at the end of June, the lowest since at least 2015.

The carry trade works the same way at a faster speed. Borrowing yen at about 1.25% to hold dollar assets at about 3.88% earns some 2.6 points a year. A 2.6% rise of the yen wipes out a year of it. The speculative part is small: leveraged funds were short only about $1.8 billion of yen futures on October 20, and on our model the trend followers have been cutting their yen short this week. The large carry is structural, in the unhedged foreign assets of Japanese investors and in everyone who funds in yen. Japan's holdings of Treasuries fell by $135 billion in the five months to July, the latest data I checked, to $1,104 billion.

The precedent is August 2024. Helped by Japanese intervention in mid July, the dollar had already fallen from almost 162 yen to about 153 when the Bank of Japan hiked on July 31. A weak US jobs report followed on August 2, and on August 5 the dollar touched 141.7 yen, the day the Nikkei lost 12.4%. That last leg, 11 yen in four sessions, was the mix of a Bank of Japan that tightens and a US economy that slows. That is the mix the market is starting to price again.

The weekly flows that cover the Bank of Japan meeting come out next week. They will tell whether Friday's decision moved Japanese money.

The midterms, and who gets the blame

On Tuesday Americans elect the whole House and about a third of the Senate. Bloomberg carries the Kalshi prediction markets. On Friday morning they gave the Democrats a 92% chance of winning the House and 69% for the Senate, up 6 points in five days, and 93% and 71% by Friday evening. A Democratic sweep of both chambers was the favourite at 68% on Friday morning, up 6 points in a month, and a Republican sweep stood at 7%. On Polymarket the Senate contract went from 28% on November 3, 2025 to 70.5% on Friday evening, and the House contract stood at 93.5%. Kalshi already gives the 2028 presidential election to the Democratic party at 61%.

My scenario. If the President loses Congress, which these markets now call likely, I expect him to blame the Democrats for the slowdown. The new Congress takes office on January 3, and from then on every weak number of 2027 will have a political owner. The blame itself is not a market price. Its consequences are, through three channels.

Fiscal. A divided Washington does not pass stimulus easily. If the slowdown arrives in 2027, it may find no fiscal cushion, and the burden falls on the Fed.

The Fed. The President has pushed for lower rates all year, and Chair Warsh is his own pick. The blame has already started there. At Wednesday's press conference a reporter read the Chair a remark the President made on October 7: "Kevin is great, but Kevin's one vote out of a Board with numerous people that were put on by Obama and Biden, and a couple by me. So you have a Board that would like to see the country do badly, in my opinion." Lisa Cook, the governor who voted for a hike that day with Cleveland Fed President Beth Hammack, faces a White House hearing on her future next week. A president who needs growth before 2028, and has someone to blame for its absence, will push harder. If the Fed eases with inflation at 3.6%, the inflation market reprices: that is the breakeven I watch.

Oil. On October 6, in Will Trump Strike Iran After the Midterms?, I wrote: "I do not think Trump resumes large scale strikes on Iran before November 3. I think he does after the vote, whether Republicans hold Congress or lose it." An escalation after the vote would be a second oil shock on top of Brent above $110.

Add the levers that need no Congress, tariffs first, and the three channels add up to the same mix: weaker growth and higher prices. Polymarket gives a recession before the end of 2026 only 6.5%, with two months to go. The vote will not change 2026. It will shape 2027.

2027: the stagflation case

Andy Constan's play ends on "Recession Island": equities sell off, companies fire workers, and the recession finally kills inflation. My worry for 2027 is another ending, the firing without the disinflation.

The growth side. The US economy grew at a 3.4% annual pace in the third quarter, after 2.2% in the second, on investment and household spending, and Chair Warsh put business investment at "about a double-digit pace." Hiring is another story. Payrolls rose only 29,000 in September, against 90,000 expected, July was revised to a loss of 10,000, and the unemployment rate rose to 4.2%. On Friday the Chicago PMI fell to 55.3 in October from 58.8, below the 56.0 expected. The engine that carries growth is investment, much of it in AI, and that is the one under pressure. In the month to October 29, Goldman's AI beneficiaries versus at risk pair lost 17.1%, Morgan Stanley's AI beneficiaries basket 11.3% and Goldman's AI data centers basket 9.8%. The market's question about AI has changed from how much these companies will earn to who pays for what they build. Alphabet answered on October 28 with negative free cash flow and more spending in 2027.

When companies spend more on machines and earn less, they cut people. In the Challenger report on announced job cuts, AI was the first reason given every month from March to July, and it remains the leading reason this year: 120,136 cuts through September, about a fifth of the total. Technology is the industry cutting the most, 165,925 jobs, up 54% from a year earlier. Netflix announced about 1,000 cuts on October 14, about 6% of its staff. Total announced cuts are still low, 43,281 in September, the lowest for a September since 2022, and initial jobless claims were 194,000 in the week to October 24. Hiring is what has stalled: hiring plans were the lowest for a September since 2011. Andy Challenger summed it up: "Companies are in a wait-and-see period right now." He pointed to high energy costs, an uncertain war in Iran and a rate hike that could make hiring more expensive. That is the point. Act 5 has not started. But the companies that drove 2025 and 2026 are now under pressure from their shareholders and their lenders at the same time, and they are among the largest employers in the country.

The inflation side. Oil is priced high into next year. The Fed is priced to raise rates to 4.464% by September 2027. Most large central banks are priced to hike. Prices paid by companies stand near 78 in manufacturing and 74 in services.

The mix. A slowdown that comes from AI and credit, inflation that comes from energy, and central banks that cannot cut because of the second while the first takes hold. That is stagflation. It is the case where stocks and bonds can lose together, because the bond hedge only works when inflation falls with growth. Treasuries hedge today because the market believes the slowdown will kill inflation. If it does not, the hedge fails, as it did from March to July, when Treasuries fell on 11 of the 13 equity sell offs.

What would tell me the stagflation case is taking over:

• Breakevens rising while real yields fall, further than since October 5 (breakeven up 11 bp, real yield down 28 bp): the market no longer trusts the Fed.

• The curve steepening from the long end while stocks fall.

• Treasuries falling again on the days stocks fall.

• Jobless claims rising while prices paid stay above 70.

• Gold turning up while the Fed is still on hold.

Gold would tell first, in the other direction. The day the market decides the Fed will choose growth over prices, gold stops being a source of cash and becomes a hedge again.

The levels I watch

Gold. Spot closed the week at $3,874.62, below the one year low of $3,932.09 of November 4, 2025: the liquidation is confirmed. A return above $3,932 would mean the break failed. On our trend model, gold needs a sharp rebound before the trend followers' short term signal turns.

Treasuries. The 10 year, 4.93% at Friday's close, against its 50 day average of 5.00%, and the 2s10s curve, 33.7 bp at the close. A further rally in the 10 year note future would turn the trend followers' medium term signal long and add to the haven bid.

Inflation pricing. The 5 year breakeven against about 2.50%, its late October high, and 2.74%, its May high.

The Fed. The December hike probability, 72.9% at Friday's close, and the implied peak, 4.464%.

Credit. Bloomberg high yield against 338 bp, then 350. Oracle's CDS against its 329 bp record, CoreWeave's against 1,163.

Hedge fund books. Our deleveraging gauge against its October peak, and another degrossing day in our monitor next week. Either would tell me the funds are no longer choosing to cut but are being made to.

The basis trade. The next CFTC report, due on November 6, SOFR against the rate on reserves at month end and into year end, the November 4 refunding, and the 30 year swap spread against its six month range of minus 79.9 to minus 59.2 bp.

Japan. USD/JPY at 155, then 150, the path of my first scenario of September 25. The 30 year JGB against 3.85% and its October high of 4.249%. The December 18 meeting, priced at 80.9%. The Ministry of Finance's weekly flows, next week.

Politics. Tuesday's result, the Senate at 71% on Kalshi on Friday evening, any escalation with Iran after the vote, and Brent against $118.35, its highest close of the year on March 31, which the expiring December contract nearly matched on Friday when it settled at $117.64.

Data. ISM manufacturing and prices paid on November 2, job openings (JOLTS) on November 3, ADP payrolls (67,000 expected, after 90,000) and ISM services on November 4, Challenger job cuts and jobless claims on November 5.

Earnings. Apple on November 2 after the close, Nvidia on November 18.

Key terms, in plain English

Stagflation. Growth that stalls while inflation stays high. Central banks cannot fight both, because cutting rates feeds inflation and raising them deepens the slowdown.

Basis point (bp). One hundredth of a percentage point. 100 bp make 1 percentage point.

Real yield and TIPS. The yield of a bond after inflation, read on Treasury Inflation Protected Securities, whose principal rises with consumer prices.

Breakeven inflation. The gap between a regular Treasury yield and a TIPS yield of the same maturity: the average inflation the bond market prices over that period.

Term premium. The extra yield investors ask to hold a long bond instead of rolling short term bills over the same period.

Bull flattener. A bond rally in which long yields fall more than short ones. A bear steepener is the opposite: long yields rising more than short ones in a sell off.

SOFR and IORB. SOFR is the overnight rate on loans backed by Treasuries, the repo market. IORB is the rate the Fed pays banks on their reserves. SOFR above IORB signals that cash is getting scarce in repo.

Basis trade. Buying Treasury bonds with borrowed money and selling the matching futures, to collect the small gap between the two.

DV01. The dollar change in the value of a position for a one basis point move in yields.

Variation margin. The cash a futures holder pays or receives every day as the price moves.

CTA. A commodity trading advisor, in practice a fund that follows price trends with futures.

Managed money. The CFTC's category for speculative funds in commodity futures.

OAS. The option adjusted spread: the average extra yield of a group of corporate bonds over Treasuries, after removing the value of the call options some of them carry.

CDS. A credit default swap, a contract that works like insurance against a company's default, quoted as an annual cost in basis points of the amount insured.

BDC. A business development company, a fund, listed or not, that lends to mid sized private companies.

Carry trade. Borrowing in a low yielding currency, such as the yen, to hold assets in a higher yielding one. It earns the rate gap, and loses when the funding currency rises.

JGB. A Japanese government bond.

Overnight index swap (OIS). A contract on the average overnight policy rate over a period, used to price the odds of a central bank move.

Prediction market. A market, such as Kalshi or Polymarket, where contracts pay out if an event happens, so their price reads as a probability.

Standard deviation (σ). The usual size of a move. A move of 2 σ or more, up or down, happens about once in twenty periods if moves are normally distributed.

Defensives and bond proxies. Stocks whose earnings hold up in a slowdown, such as utilities, consumer staples and health care. Bond proxies are those whose steady dividends make them trade like bonds.

Gross and net exposure. Gross is the size of a fund's book, its longs plus its shorts. Net is the longs minus the shorts, its bet on the direction of the market. Both are often given against the fund's capital, as gross and net leverage.

Degrossing. Funds cutting their long and short positions at the same time, to reduce risk.

Questions readers ask

Is the US in recession? No. GDP grew at a 3.4% annual pace in the third quarter, the Fed's statement says activity is expanding "at a strong pace," and jobless claims are low. Hiring is weak, though, with 29,000 payrolls in September. Markets price a slowdown ahead of the data, not a recession: Polymarket gives a recession before the end of 2026 only 6.5%, with two months left to run.

Why is gold falling if a slowdown is coming? Because the first phase of a slowdown is a hunt for cash, and gold was the great winner of 2025. Real yields fell and gold fell anyway. In 2008 and in March 2020 gold fell first and turned around the time the Fed went all in. This time the Fed is priced to hike.

Inflation is 3.6%. Why are bond yields falling? Because the bond market believes the slowdown will bring inflation down, and because bonds have become a haven again on the days stocks fall. That belief is the bet I question.

Why are hedge funds cutting their gross exposure? Because volatility sets the size of their books. They came into October with gross leverage near the top of its five year range and small net bets, so a jump in volatility makes the same book riskier, and risk managers shrink it. The selling hits the stocks they own most first, the AI leaders, and some of the same funds run the Treasury basis trade.

Why should anyone outside hedge funds care about the basis trade? Because it is part of the plumbing that finances the Treasury market. When it unwinds, Treasury yields can jump at the very moment they should fall, as they did in March 2020.

What does the Bank of Japan have to do with US markets? Japan is among the largest foreign holders of Treasuries, and the yen funds carry trades worldwide. A Bank of Japan that tightens while the US slows narrows the rate gap from both sides. In August 2024 that mix took the dollar from about 153 yen to 141.7 in four sessions.

What changes if the Democrats win Congress? A divided government, little room for fiscal support in a slowdown, more pressure on the Fed to cut, and a political fight over who owns the 2027 economy. Together they push toward weaker growth, and the pressure on the Fed and the oil risk also push toward higher prices.

What would prove you wrong? Three things. Core inflation turning down while energy eases, which would make the slowdown disinflationary and the Fed free to cut. AI earnings and investment holding up while credit spreads return to their January lows. And gold back at new highs with breakevens falling, which would mean its fall was positioning, not the last piece.

A note to readers. From next week, CrossVol Research becomes a paid publication, at $15 a month. I will give the details in a short note next week, including what stays free and what is for paid subscribers. Thank you to everyone who has read and shared these pieces, and to those who have already pledged to support them.

This article is for educational purposes only. It is not investment advice, a solicitation, or an offer to buy or sell any security or derivative. Levels are as of the dates and times stated and can change quickly. Bonds, credit, equities, currencies, options and futures carry a high risk of loss. The five act framework belongs to Andy Constan (The Damped Spring Report, "The Script," July 10, 2023). The reading of 2026 is the author's own and does not represent his views. Prediction market prices are probabilities set by traders, not forecasts of the author.

$GLD $TLT $SPY $HYG $ORCL $CRWV $FXY

Data: Bloomberg, CFTC, FRED, ICE, EIA, US Treasury, Federal Reserve, BEA, BLS, Bank of Japan, Japan Ministry of Finance, Challenger Gray & Christmas, Milliman, Kalshi, Polymarket.

Educational content, not investment advice. Options and futures carry a high risk of loss.

On Sunday, the last section of our weekly review came down to one question: can the megacaps outrun the Fed? Four sessions later, the answer is no.

What we wrote on October 25

• On the Fed: "Either way, the Fed is not the catalyst that rescues equities next week."

• On stocks: "Only the megacaps can rescue the market." The risk we named: "A repeat from Microsoft or Alphabet would remove the last support of the cap-weighted index."

• On rates: "A Fed hold is the trigger for the next leg of the Treasury squeeze."

The Fed held, and it did not help

On Wednesday the Fed kept its target range at 3.75% to 4% by 10 votes to 2. The two dissenters wanted a 25 bp hike.

Thursday's inflation data kept the pressure on. The PCE price index rose 0.5% in September and 3.6% over a year, up from 3.4% in August. Core PCE stands at 3.1%.

At Thursday's close, fed funds futures still priced a 74.6% chance of a hike on December 9, against 75% when we wrote. The path further out has come down a little: the implied rate now peaks at 4.46% in September 2027, against 4.56% in October 2027 on October 23.

The megacaps did not carry the index

Alphabet repeated the template Meta set on October 22: a miss in search advertising, about $9 billion of negative free cash flow and a "significant" increase in investment planned for 2027. The stock opened with a gap and closed down 12.64% at $283.82.

Microsoft reported Azure growth of 48% with spending below expectations. The shares rose after hours on Wednesday, then closed flat on Thursday, down 0.12%.

Amazon closed down 4.25% at $243.79 before its results. After the release it fell to $214.01 at 4:02 pm New York, then traded at $226.73 at 4:56 pm, 7.0% below its close.

Intel closed at $92.07, fell to $71.20 at 4:05 pm after its results and traded at $77.27 at 4:56 pm, 16.1% lower.

During the session, the damage stayed inside AI tech. The Nasdaq 100 lost 1.38% and the S&P 500 only 0.30%, while the Russell 2000 ETF gained 0.97%, Apple 1.48% and Meta 1.98%. Europe closed higher, with the DAX up 1.31% and the CAC 40 up 1.05%. Investors moved out of AI tech, not out of stocks. The VIX closed at 17.20.

After Amazon and Intel, equity futures took the next leg down. Nasdaq 100 futures settled at 29,562 at 4 pm, down 1.39%, then fell to 29,111 at 4:06 pm and stood at 29,218 when trading paused at 5 pm, 2.54% below Wednesday's settlement. S&P 500 futures went from a 7,754.25 settlement to about 7,710, 0.94% below Wednesday.

The Treasury leg is running

On October 18 we wrote that "AI specific sessions behave differently: on those days, yields fall with stocks." Thursday was one of those days. The 10-year Treasury yield fell 8.1 bp on the day to 4.94%, after a session high of 5.04%. Treasury bond futures gained about 0.9%.

The squeeze is starting where we expected it. On our trend-following model, the short-term signal has flipped to long on the whole US curve, from the 2-year to the 30-year, and on SOFR futures. The medium and long-term signals are still short, so trend followers remain net short for now. If the medium-term signals flip next, the buying becomes mechanical.

What comes next

• Positioning. On our October 25 numbers, hedge funds held a record 22% of their exposure in the Magnificent 7. After Alphabet and Amazon, de-grossing can resume.

• Apple reports on Monday, November 2, after the close. It is the next megacap test.

• The Bank of Japan decides on Friday. A December hike signal would lift the yen and pressure carry trades.

So far the market has rotated rather than broken: on the session, the S&P 500 lost 0.3% and the Nasdaq 100 1.4%. The question for Friday is whether the cap-weighted index can hold now that its last support is giving way.

Educational content, not investment advice. Options and futures carry a high risk of loss.

Weekly changes run from the Friday October 16 close to the Friday October 23 close, unless stated otherwise.

This article is for educational purposes only. It is not investment advice. Trading equities, bonds, futures and options involves substantial risk of loss.

The week in five lines

1. Equities: the main indices barely moved, but the AI trade broke underneath. Semiconductors lost 6.7% in one session while small caps and the average S&P 500 stock rose.

2. Bonds: US Treasuries rallied hardest on the day stocks fell, so bonds hedged stocks again after months of moving with them. Funds that were short Treasuries had to buy them back as yields dropped.

3. Credit: it kept leading. US high yield spreads reached a 17-month high, and the cost of insuring Oracle ($ORCL) and CoreWeave ($CRWV) debt widened sharply, while the Nasdaq 100 lost less than 1%.

4. Oil: Brent rose 6.2% to $116.72 on the Gulf escalation, with an extreme premium for prompt barrels and a $13 gap over WTI.

5. France: the gap between French and German 10-year yields widened 29 bp, the second biggest weekly move since 1990. France now borrows at higher rates than Italy and Greece. I cover it in a separate piece on the French bond crisis.

My reading: the week reset positioning without breaking the market. Hedge funds are at their most defensive in five years, crowded long books have been cut, and the volatility curve never inverted. The two themes I have described here since September both advanced: AI spending financed with debt widened its credit cracks, and long-term money kept choosing bonds over stocks. The large indices still show little of either.

A few units used throughout: a basis point (bp) is one hundredth of a percentage point. A CDS (credit default swap) is an insurance contract against a borrower's default, quoted as an annual premium in basis points. High yield bonds are bonds rated below investment grade.

Weekly scorecard: stocks, bonds, oil, gold and volatility

A note on method: several recaps circulating this weekend measure the week from Monday's close. Monday was itself an unusual session (the US 10-year closed at its weekly high and utilities fell 2%), so Monday-based numbers overstate several moves. Utilities, for example, are up 2.9% from Monday's close but only 0.8% from the previous Friday.

1. Equities: the AI trade broke under a calm surface

The Thursday shock

On Thursday, October 22, a supply warning from Lam Research ($LRCX) and Meta's plan to raise capital spending again, funded through a new share class, changed the market's question about AI from "how much will they earn?" to "who pays for it?". The Philadelphia Semiconductor Index fell 6.72% in one session. Meta lost 6.98% on twice its average volume. Nvidia lost 4.4%, AMD ($AMD) 8.1%, Micron ($MU) 6.3%. CoreWeave lost 10.1% and Nebius 7.2%.

The S&P 500 lost 1.21% that day, yet more of its stocks rose than fell: 335 against 161. The equal-weight S&P 500, which gives every member the same weight instead of weighting by market value, rose 0.73%, and consumer staples gained 3.3%. A few very large stocks pulled the index down while the typical stock went up.

Rotation, measured

Sectors, Friday to Friday:

Equal weight versus market-cap weight. The equal-weight S&P 500 ETF ($RSP) gained 1.1% while the standard S&P 500 ETF ($SPY) lost 0.4%, so the ratio between them rose 1.6%. When the average stock beats the index, money is leaving the largest names rather than leaving equities.

Bank baskets. Banks publish baskets of stocks built to isolate one theme or factor. Five-session moves to Friday's close:

Momentum is the strategy of owning what has risen most over the past year. On Thursday alone, the Nomura Pain Trade basket gained 12.0% and Morgan Stanley's momentum pair lost 8.1%. Over one month, "long value, short momentum" made almost 24%. The stocks everyone owned for the AI story were sold and the stocks everyone was short were bought back: a momentum crash, more than a sector rotation.

Positioning: the selling fuel is largely spent

Goldman Sachs prime brokerage data reported this week (a prime broker is the bank that finances and holds hedge funds' positions):

• Net leverage of US long/short hedge funds at 43.4%, the 0th percentile over one, three and five years. Net leverage is long exposure minus short exposure, relative to capital: these funds are at their most defensive in five years.

• Short leverage near a five-year high, at the 99th percentile, concentrated in AI infrastructure and semiconductors.

• Gross leverage of the prime book at 312.3% (96th percentile over five years). Gross leverage adds longs and shorts: funds are still heavily invested, on both sides.

• Goldman's basket of hedge fund favourite longs against their favourite shorts fell about 4.2% on Thursday, its worst day since 2011.

• De-grossing, the cutting of both longs and shorts, on 7 of the last 8 sessions. The October 22 session was the fifth-largest de-grossing in Goldman's prime book data, almost entirely long selling.

• Hedge fund exposure to the "Magnificent 7" (Apple $AAPL, Microsoft $MSFT, Alphabet $GOOGL, Amazon $AMZN, Nvidia $NVDA, Meta $META, Tesla $TSLA) at a record 22%.

My read on the "record shorts": mostly a financing artifact

Short interest charts are everywhere this weekend. Most of them miss the flows behind them.

Much of this short exposure is not an outright bet against the market. Prime brokers are rationing their balance sheets. Goldman Sachs said earlier this month that demand for equities financing "continued to exceed available industry capacity", and Bloomberg reported that Barclays and other banks worked with hedge fund clients to limit their balance sheet use, "encouraging them to take more short positions".

The sequence on the street looks like this:

1. A hedge fund manager asks for more financing.

2. The bank answers: public AI bets look risky and we already lend a lot against private AI bets. We will extend more financing, but only if you add short exposure to improve your collateral profile.

3. The manager shorts more stocks against the main bet.

That is how short exposure can look astronomically high while many end investors remain heavily long tech. The extra shorts finance the main bets. It also explains why equity financing rates have not blown out the way they did before the SpaceX IPO in June, even though lending is slowing.

Two consequences:

• The squeeze potential is mechanical. Shorts held as collateral can be covered quickly if prices bounce, which is why a strong week of results from the large cloud companies could trigger a sharp one to two day rally.

• The real risk sits in the longs. If AI spending guidance disappoints, the 22% Magnificent 7 exposure is the fuel, and these shorts will not provide a floor.

Breadth and volatility plumbing

• Friday breadth: 134 S&P 500 stocks up, 362 down, 73% of volume on the sell side. 58% of members trade above their 50-day moving average.

• Volatility-targeting funds, which cut exposure when market volatility rises, run at about 94% exposure. A 1% daily move would trigger about $15 billion of selling, and a 2% move about $69 billion.

• Dealer gamma. Option dealers hedge their books by trading the index. Below a "flip" level, their hedging amplifies moves instead of damping them. As of Thursday, the S&P 500 flip sat near 7,740, above the market, with large option positions at 7,600 and 7,800. Chart watchers also mark 7,616 as the neckline of a head-and-shoulders pattern.

2. Rates: the hedge came back, and short sellers were caught

The Thursday rally

The US 10-year yield fell from 5.12% on Wednesday to 4.95% on Thursday, a 16.6 bp drop in one session and the biggest since August 2025, as the AI selloff sent money into Treasuries. On the week, the 10-year lost 7.8 bp and the 30-year 9.9 bp.

Two correlations confirm that bonds hedged stocks again:

• the 50-day correlation between the S&P 500 and long-dated Treasuries (the $TLT ETF) fell from about 50% at the end of September to 19% on Friday;

• the 50-day correlation between AI stocks relative to the rest of the market ($SMH against $RSP) and long Treasuries reached -36%, near its lowest of the year.

This is the scenario I laid out in The Pension Valve Only Opens One Way (October 15): long-term money that does not come in to buy equities on a selloff is buying bonds instead. When that happens, stocks and bonds stop moving together and the bond hedge comes back. This week's falling correlation and falling term premium are what that flow looks like in the data.

On AI selloff days, yields now fall with stocks. On October 18, I wrote that "AI specific sessions behave differently: on those days, yields fall with stocks", and flagged the late-October AI earnings as the date to watch (The VIX Woke Up Loaded the Other Way). Thursday was that session.

Why the rally was so violent: positioning

CFTC data as of Tuesday, October 20 showed speculators heavily short Treasury futures: about 857,000 contracts net short on 2-year futures and 964,000 on 10-year futures. Those shorts had begun to shrink (the aggregate fell by about 93,000 contracts, measured in 10-year equivalents). On Thursday, open interest (the number of contracts outstanding) fell in 2-year and 5-year notes, a short-covering signal, while new positions were opened in 10-year notes.

CTAs (commodity trading advisors, systematic trend-following funds that buy what rises and sell what falls) entered the month maximum short bonds. The long end is the exception to the covering: as of Tuesday, speculators had added 30-year shorts for a second straight week (-16.7% week on week). Covering at the front, adding at the back: the market still bets on heavy bond supply at long maturities while buying short maturities on growth fears.

The squeeze is not over: half a trillion dollars of CTA buying is still loaded

The most important positioning number of the week did not come from the CFTC. It came from Nomura's cross-asset strategist Charlie McElligott, in a note reported on Saturday, October 24:

• CTAs covered $68 billion of bond futures and $38 billion of STIRs (short-term interest rate futures, such as SOFR contracts) in two weeks.

• And they are still short. Their net weight stands at -47.3% in bonds, the 12th percentile of history, and -32.8% in STIRs. In McElligott's words, much of the CTA signal went "from -100% Short to now -74% Short on BIG covering". They are not long. They are "less short".

• The fast windows have already flipped. Trend models blend several look-back periods. The two-week and one-month windows have turned long. The three-month window has not: it needs what McElligott calls a "substantial further rally from here".

• The asymmetry is extreme. Under Nomura's trigger estimates, a further 2% rally in bond futures would force CTAs to buy about $255.5 billion of global bonds, and a 2% rally in short-rate futures another $264.1 billion of STIRs: roughly half a trillion dollars of systematic buying. A 2% selloff would trigger only $52.6 billion and $33 billion of selling.

Two other details from the note matter for bonds. Spot and volatility moved together in Treasuries on Thursday: bond prices rose while bond volatility rose too, the signature of a short squeeze rather than calm buying (the MOVE index jumped from 95.9 to 106.8 that day). And CTAs lost 2.0% on the week, their 9th percentile, which pushes their risk models to cut the losing shorts faster.

This is why I do not read Friday's pause in the Treasury rally as the end of the move. Trend followers covered just enough to stop the bleeding, and the bulk of their short is still on the book. The fuel for a second leg is larger than the first one. ZeroHedge summed up the cross-asset link in one line: "the very thing creating the equity problem is solving the rates problem." An AI growth scare sends money into bonds, and every point of bond rally pulls the CTAs closer to the next trigger.

Term premium and the Fed

• The term premium is the extra yield investors demand to lock in money for ten years rather than rolling short-term bills. The New York Fed's estimate (the ACM model) fell from 98.5 bp on October 8 to 63.9 bp on October 22, and from 75.6 bp to 63.9 bp this week alone.

• The market removed part of the Fed's expected hikes. The gap between the SOFR futures contract three years out and today's SOFR rate (the benchmark overnight rate) fell from 76 bp on Monday to 58.5 bp on Friday, and December contracts stopped fully pricing another hike.

• The MOVE index of Treasury volatility swung from 95.9 on Wednesday to 106.8 on Thursday and closed the week at 103.3.

The long end did not trade on supply this week; it traded on who is buying. As long as equity selloffs send long-term money into bonds, rallies like Thursday's will keep coming on the bad days for stocks.

3. Credit: still leading

• US high yield spread: 340 bp, a 17-month high and the 99th percentile of the past year. The primary market saw the year's biggest wave of pulled deals.

• Oracle five-year CDS: 314 bp on October 21, from 301 on October 16, after a peak of 329 on October 20. On September 25, I noted it at 237 bp (The Crash Fuse Is Lit). On October 13, one of my markers for the next phase was "five year CDS above the 260.9 of October 8" (US Macro: Where Are We as of October 2026?). It is now 53 bp above that level.

• CoreWeave five-year CDS: about 1,100 bp, up from about 850 in early October. The spread of its 9% 2031 bond over Treasuries widened from 985 bp to 1,101 bp in one week.

• Oracle 4.95% 2031 bond: 260 bp over Treasuries, from 235 bp.

• SpaceX's $40 billion AI chip bond deal was snarled as borrowing costs across the sector rose.

• Blue Owl ($OWL), a bellwether of private credit, trades 45% below its 52-week high.

On October 13, my verdict was "early Act 4, through credit. The indices will see it last." This week, AI credit widened again and the Nasdaq 100 lost 0.8%. Credit is moving first, exactly in the order that piece described. The next markers to watch are cash high yield through 350 bp and the first decline in forward earnings estimates.

4. Volatility: a V inside an upward-sloping curve

Daily path, Monday to Friday (on Friday, October 16, the VIX closed at 19.99 and the oil volatility index at 50.7):

Definitions: the VIX is the 30-day implied volatility of the S&P 500, the market's price for expected swings. The VVIX is the implied volatility of the VIX itself; it rises when investors pay up for VIX calls, the classic institutional hedge against a volatility spike. The SKEW index measures the price of far out-of-the-money puts: above 150, options price a meaningful chance of a large drop. The put/call ratio compares put volumes to call volumes.

What the table says:

• Complacency peaked on Wednesday. VIX 16.68, VVIX 94.55, put/call 0.52, SKEW 150: every gauge pointed to a breakout the day before the shock.

• The volatility curve never inverted. The VIX stayed below its three-month equivalent (VIX3M) all week, a configuration called contango, with a ratio near 0.95 on Thursday. The market treated Thursday as a repricing within a range, not a crash.

• VVIX rose 16% in two sessions, from Wednesday's low to 109.7, while the VIX stayed below 20. In September, I wrote: "A VVIX that keeps climbing from here would be the options market saying it no longer trusts the calm" (The VIX Is Too Calm for a 3.4 Sigma Rates Shock). This week's tape is consistent with that.

• Downside protection is expensive. On Thursday, a "16-delta" S&P put, an option roughly one standard deviation out of the money, cost 1.8 times the equivalent call. The VIX traded 7.5 points above one-month realized volatility, the swings the index actually delivered.

• Implied correlation stays low. The Cboe one-month implied correlation index (COR1M) sits at 10.9: options price individual stocks as moving largely independently. That keeps index volatility cheap relative to single-stock volatility.

5. Oil: the risk premium sits in the curve

• Brent ($BNO) +6.24% to $116.72, near its March 31 peak of $118.35. WTI ($USO) +1.83% to $103.27. The Brent-WTI gap reached $13.45: the geopolitical premium sits in the international benchmark, closest to the Gulf.

• An extreme backwardation. The gap between the first and ninth Brent contracts widened from $17.65 to $24.51 in one week. Backwardation means near-term barrels cost more than later ones: buyers pay a premium for immediate supply.

• Oil volatility (OVX) rose from 50.7 to 60.0, +18% on the week.

• Strait of Hormuz: transits ran at about 7 ships a day, according to IMF PortWatch data. Tankers were hit, Aramco evacuated its Dhahran headquarters, and the US asked diplomats to leave Saudi Arabia.

• Positioning, in futures data released at the start of the week: money managers raised Brent net longs to a one-month high while WTI net longs were the smallest since early August.

According to export figures cited in the market this week, Gulf exports have recovered to 17 to 18 million barrels a day, above pre-war levels. Futures price the risk of disruption more than a measured shortage. That makes the downside violent in case of a ceasefire, and the upside violent in case of a real closure.

My October 17 piece (Oil Above $100) described energy "settling on a higher plateau", with trend followers already long and few sellers left to absorb a fresh shock. This week's $7 jump in Brent, met by almost no selling, is that mechanism at work.

6. Gold, the dollar and crypto

Gold ($GLD) is under pressure: -4.07% on the week, including 2.8% on Thursday, the same day Treasury yields dropped 16.6 bp. A haven asset falling on a flight-to-quality day usually points to liquidity needs: investors sell what they can to meet margin calls elsewhere, and on that day the haven of choice was Treasuries, not gold. I treat it as a warning sign for next week.

The euro lost 0.3% against the dollar as France weighed. Bitcoin fell to about $81,600 early in the week as the tech selloff cooled risk appetite.

7. Europe: France decouples

The French 10-year yield rose 21 bp to 5.02% while the German one fell 8 bp. The OAT-Bund spread (French minus German 10-year yield) widened 29 bp to 167 bp, after reaching 178 bp on Thursday. France now pays more than Italy (122 bp over Germany) and Greece (103 bp), and its five-year CDS (106 bp) is wider than Mexico's. Moody's kept France at Aa3 with a negative outlook on Friday evening, one notch above S&P and Fitch.

The CAC 40 lost 2.2% on the week and is the only major European index down for the year (-4.5%). Swiss stocks are becoming a refuge: the SMI is up 4.6% in October while the Euro Stoxx 50 is down 3.2%.

The street calendar is loaded: the yellow vests call for pot-banging protests in front of the National Assembly on October 30, and the unions have called a general strike starting November 5.

I have tracked this spread since September: 86 bp on September 4, 104.6 bp on September 18 (Dispersion and the OAT-Bund Channel), 141 bp on October 4 (Volatility Is Starting to Talk Even While the VIX Sleeps), 167 bp now. The full analysis, including the political and constitutional risk behind the move, is in today's separate piece.

8. Flows

• Equity funds took in $55.6 billion in the week to October 21, money market funds $28.8 billion, bonds $12.1 billion and gold $2.6 billion, according to BofA.

• China equity funds saw $17 billion of inflows and technology funds $18.8 billion, the largest since July: investors bought the dip.

• Emerging market bond funds recorded 12 straight weeks of inflows.

• ICI data show US mutual fund outflows at a record, while equity ETFs keep gathering assets: the shift from mutual funds to ETFs continues.

9. What this week confirmed

• The OAT-Bund widening is "a structural repricing, not a headline" (Sep 19). This week: 104.6 bp then, 167 bp now, 178 bp on Thursday.

• Cross-asset volatility talks before the VIX, with France at the epicenter (Oct 4). This week: France +29 bp, oil volatility +18%, VVIX +16% from Wednesday.

• Oracle credit insurance is the fuse of the AI financing story (Sep 25). This week: Oracle 5Y CDS 237 bp then, 314 bp now.

• The next phase arrives through AI credit and the indices see it last (Oct 13). This week: Oracle above the 260.9 marker, CoreWeave about 1,100, high yield at a 17-month high, Nasdaq 100 -0.8%.

• Long-term money that skips equities buys bonds, and the bond hedge returns (Oct 15). This week: Stock-bond correlation 50% to 19%, term premium 98.5 to 63.9 bp, Treasuries up on the worst equity day.

• On AI selloff days, yields fall with stocks (Oct 18). This week: Thursday: semiconductors -6.7%, 10-year yield -16.6 bp.

• A climbing VVIX means the options market no longer trusts the calm (Sep 23). This week: VVIX 94.5 to 109.7 while the VIX stayed under 20.

• Energy settles on a higher plateau, with few sellers left (Oct 17). This week: Brent +6.2% to $116.72, backwardation at $24.51.

Six weeks of posts, one consistent read: credit and cross-asset volatility move first, long-term money rotates from stocks into bonds, and the headline indices are the last to show it.

10. Next week: the Fed has no good option, and only the megacaps can save the market

The Fed's lose-lose

The futures market does not expect a Fed hike on Wednesday, October 28. Fed funds futures priced only a 9.4% probability of a hike at the October meeting on Friday's close, against 75% for December 9, and a policy rate peaking at 4.56% in October 2027, about 2.7 hikes above today's 4.00% target.

My view is that both outcomes are negative for stocks.

If the Fed holds, it falls further behind inflation. The data are moving the wrong way:

With the effective fed funds rate at 3.88%, the real policy rate, the policy rate minus inflation, is barely positive: about 0.2 point against CPI and 0.5 point against PCE. Policy is hardly restrictive while inflation is rising and oil is above $115. A hold on Wednesday, followed by a hot PCE print on Thursday, would tell markets that the Fed is behind the curve. That raises the inflation risk premium in long yields, and higher discount rates compress equity valuations, starting with the long-duration growth stocks that dominate the index.

If the Fed hikes by surprise, it tightens financial conditions into a fragile tape. With less than a 10% probability priced, a hike would be a genuine shock. Financial conditions are still easy: the Bloomberg US Financial Conditions Index stands at +0.80, above its long-run average, but it already fell from +1.21 on October 13 as credit spreads widened. A surprise hike would push the dollar and real yields up, widen credit spreads that are already at a 17-month high, and hit a market where volatility-targeting funds would sell about $15 billion of stocks for every 1% daily move.

Either way, the Fed is not the catalyst that rescues equities next week.

A Fed hold is the trigger for the next leg of the Treasury squeeze

The bond market reads the same decision very differently. A hike would push short rates up and extend the CTA short a little longer. A hold does the opposite. December contracts already stopped fully pricing another hike this week. A Fed that stays put on Wednesday, while growth stocks sell off on AI capex fears, lets front-end futures rally further. That is exactly the move that brings the three-month CTA window within reach.

Once that window flips, the buying is mechanical, not discretionary: about $264 billion of STIRs and $256 billion of bond futures for a 2% rally, against limited selling the other way. With CTAs still at the 12th percentile of their bond positioning, a hold could turn last Thursday's 16.6 bp drop in the 10-year yield into the first leg of a much larger squeeze, and probably a dramatic one.

That is the cruel symmetry of next week. The same hold that leaves the Fed behind inflation and hurts equity valuations would feed one of the largest forced bond-buying setups of the year. Bonds would keep doing the hedging job. Stocks would still need the megacaps.

Only the megacaps can rescue the market

If anything turns the tape, it is the largest companies in the index, and they report in the same 48 hours as the Fed, GDP, PCE and the ECB:

The setup favors a sharp move in either direction:

• Upside fuel. Hedge fund short leverage sits at the 99th percentile, concentrated in AI infrastructure. If Microsoft, Alphabet and Amazon defend their AI spending with visible revenue, the shorts held as financing collateral can be covered within one or two sessions.

• Downside risk. Hedge funds hold a record 22% of their exposure in the Magnificent 7. Meta showed the template on October 22: more capex without more revenue meant -7% in one session and a semiconductor crash. A repeat from Microsoft or Alphabet would remove the last support of the cap-weighted index.

So the week comes down to one question: can the megacaps outrun the Fed? If their results are strong enough to offset a Fed that is either late or hawkish, the index holds. If not, there is no other buyer in sight.

The full calendar

• Sun Oct 25. Brazil presidential runoff. What would move markets: A contested result would hit the real and emerging market carry trades.

• Tue Oct 27. Durable goods (Sep), Conference Board confidence (Oct), new home sales. Consensus: +1.9% / 80.9 / 650k.

• Wed Oct 28. FOMC decision. Consensus: Hold (9.4% hike probability). What would move markets: A hold with a dovish tone: behind the curve. A surprise hike: tighter conditions.

• Thu Oct 29. US Q3 GDP (advance), September PCE inflation. Consensus: GDP +3.2%, PCE 3.6% y/y, core 3.1%. What would move markets: Hot PCE: December hike priced fully, 10-year back above 5.10%.

• Thu Oct 29. Microsoft and Alphabet results. What would move markets: AI revenue versus AI spending.

• Thu Oct 29. ECB decision. Consensus: Pause. What would move markets: Any language on its bond backstop for France.

• Fri Oct 30. Amazon results.

• Fri Oct 30. Employment cost index (Q3), Chicago PMI. Consensus: +0.8% / 56.1. What would move markets: Wage inflation, a key Fed input.

• Fri Oct 30. Bank of Japan. What would move markets: A December hike signal would lift the yen and pressure carry trades.

• Fri Oct 30. Yellow vests "casserolades" (pot-banging protests) in front of the French National Assembly. What would move markets: First test of the protest bans.

• Mon Nov 2. Apple results.

• Tue Nov 3. US midterm elections. What would move markets: The window for an Iran strike after the vote is the oil market's binary risk.

• Thu Nov 5. General strike in France, called by the unions. What would move markets: The biggest street test for French spreads since the crisis began.

Consensus: Bloomberg surveys.

Bottom line. Markets enter the week with the VIX near 19.5, the MOVE near 103, oil volatility at 60, a broken AI momentum trade and a historically wide OAT-Bund spread. The Fed cannot help: holding leaves it behind rising inflation, and a surprise hike would tighten conditions into a fragile market. For bonds, a hold is the spark: CTAs are still short, with about half a trillion dollars of systematic buying waiting on a 2% rally. The only possible rescue for stocks comes from Microsoft, Alphabet and Amazon. The signal I will watch most closely besides their results is gold: if it keeps falling with yields, liquidity rather than growth is driving the tape.

FAQ: the week of October 19, 2026 in four questions

Why did semiconductor stocks fall on October 22, 2026? A supply warning from Lam Research and Meta's plan to raise AI capital spending again, funded through a new share class, made investors question who pays for the AI build-out. The Philadelphia Semiconductor Index fell 6.72% that day.

Why did Treasuries rally while stocks fell? Money left AI stocks for bonds, and speculators who were heavily short Treasury futures had to buy them back. The 10-year yield fell 16.6 bp on Thursday to 4.95%, and stocks and bonds hedged each other again. Trend-following funds are still net short bonds after covering, which leaves room for a larger squeeze.

Are hedge funds bearish? Their net exposure is the lowest in five years, but much of the record short position finances their long bets rather than expressing a view against the market. Their exposure to the Magnificent 7 is a record 22%.

Will the Fed hike on October 28, 2026? Futures priced only a 9.4% probability on October 23, against 75% for December. With CPI at 3.7% and the real policy rate barely positive, a hold risks leaving the Fed behind inflation, while a surprise hike would tighten financial conditions.

What are the key events for the week of October 26, 2026? The Fed on October 28; US GDP, PCE inflation, the ECB and results from Microsoft and Alphabet on October 29; Amazon and the Bank of Japan on October 30; Apple on November 2.

Further reading on CrossVol Research

• France Bond Crisis: The OAT-Bund Spread Hits 178 bp and France Now Pays More Than Italy and Greece (October 25, 2026)

• The Pension Valve Only Opens One Way (October 15, 2026)

• The VIX Woke Up Loaded the Other Way (October 18, 2026)

• Oil Above $100: What Futures Curves, Crack Spreads and Gas Spreads Say (October 17, 2026)

• US Macro: Where Are We as of October 2026? (October 13, 2026)

• Volatility Is Starting to Talk Even While the VIX Sleeps (October 4, 2026)

• The Crash Fuse Is Lit (September 25, 2026)

• The VIX Is Too Calm for a 3.4 Sigma Rates Shock (September 23, 2026)

• Dispersion and the OAT-Bund Channel (September 19, 2026)

This article is for educational purposes only and does not constitute investment advice or a recommendation to buy or sell any security. Equities, bonds, futures, CDS and options carry substantial risk, including the loss of more than the initial investment for leveraged instruments. Past patterns do not guarantee future results.

Market data as of the October 23 close.

This article is for educational purposes only. It is not investment advice. Trading bonds, futures and options involves substantial risk of loss.

France now borrows like the periphery

On Friday, October 23, 2026, France paid more to borrow for ten years than Italy, Greece, Spain and Portugal. The price of insurance against a French default closed above the price of insurance against a Mexican one.

Fifteen years ago, the euro crisis was a story of Greece, Italy, Spain and Portugal paying for their debt while France sat with Germany in the core. This week the order flipped. France pays 45 bp more than Italy and 64 bp more than Greece for ten-year money.

Key terms.

• A basis point (bp) is one hundredth of a percentage point: a yield moving from 5.00% to 5.01% has risen 1 bp.

• The OAT-Bund spread is the gap between the French 10-year yield (OATs are French government bonds) and the German 10-year yield (Bunds). Germany is the euro area's safe-haven borrower, so the spread measures the extra yield investors demand to hold French risk.

• The euro area core (Germany, the Netherlands, and until now France) borrows cheaply. The periphery (Italy, Spain, Portugal, Greece) historically paid a risk premium.

• A CDS (credit default swap) is an insurance contract against a borrower's default. At 106 bp, insuring 100 million of French debt for five years costs about 1.06 million a year.

Three months ago, French five-year insurance traded as low as 32 bp. It has more than tripled.

The week the French bond market broke a record

• OAT-Bund 10Y: +29 bp in one week, from 138 bp to 167 bp. According to Bloomberg, it is the second biggest weekly widening on records going back to 1990. On Thursday, October 22, the spread reached 178 bp, its highest since the 2011 euro debt crisis.

• French 10Y yield: 5.01%, after closing at 5.08% on Thursday, the highest in 24 years. In three months, the French yield is up 105 bp. The German one is up 17 bp.

• French 2Y yield: +22 bp to 3.80%, while the German 2Y fell 5 bp. The two-year gap between France and Germany went from 47 bp to 74 bp in five sessions.

• Euro-OAT futures, exchange-traded contracts on French government bonds, lost 1.9% on the week. Open interest, the number of contracts outstanding, rose by about 101,000 in the four days to Thursday, +15%: a record build-up of new positions.

• CAC 40 (tracked by the $EWQ ETF): -2.2% on the week and -4.5% year to date, -8.9% for a dollar-based investor. It is the only major European index down for the year. The DAX is up 2.8%, Milan 8.0%, Madrid 9.7%.

The French spread now sits 2.4 standard deviations above its three-month average. This newsletter has tracked it since September: 86 bp on September 4, 104.6 bp on September 18, when I called it "a structural repricing, not a headline", 141 bp on October 4, and 167 bp today.

Why the French 2-year yield is the real alarm

When investors worry about a country's long-term debt path, the long end of the curve, made of bonds with long maturities, usually sells off first. That is a term premium story: lenders want extra yield for tying up money for decades.

France did the opposite this week. Its 2-year yield rose as much as its 10-year (+22 bp and +21 bp) and its 30-year rose half as much (+11 bp). The gap between the French 2-year and 30-year shrank by 11 bp: a bear flattening, short rates rising faster than long rates.

The United States offers the contrast. The New York Fed's estimate of the US 10-year term premium (the ACM model) fell from 98.5 bp on October 8 to 63.9 bp on October 22, and Treasuries rallied across the curve.

A two-year bond matures before any fiscal plan can repair the debt ratio. Investors who dump it are pricing the next eighteen months: a budget without a majority, a presidential election in six months, and a street in open revolt. That is a political risk premium sitting at the front of the curve. Italy's curve showed the same front-end stress on its way into the 2011 crisis.

Moody's: a reprieve that reads like a warning

Late on Friday, Moody's kept France at Aa3 with a negative outlook. Aa3 is the fourth step of Moody's scale; each step is called a notch, and Moody's now rates France one notch above S&P and Fitch, which both cut it in 2025. The market had largely expected a downgrade. The statement still reads as a warning:

> "Slower growth and a higher deficit have pushed France's debt burden somewhat higher than previously expected. Bond yields have risen more rapidly than we expected, while widespread social protests have erupted which could constrain fiscal consolidation efforts."

And:

> "Despite the challenging political and social environment, our baseline assumption is that a deficit-reducing budget for 2027 will eventually be adopted, possibly by decree."

Source: Moody's statement as reported by Bloomberg, October 23, 2026.

Moody's base case already allows for a budget adopted by decree. The deficit is running at 5.4% of GDP this year, against a 5% target for next year. The government has no majority, and the National Assembly rejected the revenue part of the 2027 social security budget on Friday. Debt servicing costs are heading toward about 2.9% of GDP by 2027, against about 1.3% in 2020.

The reprieve should give French bonds a mild lift when markets reopen. The trajectory is unchanged, and France still sits one notch above where the two other major agencies already rate it.

The street has become a market variable

The bond market moved on the street this week, not on a fiscal data point.

• October 17-18: more than half a million people marched, public buildings were attacked, about 3,000 stores were looted and 2,700 people were arrested, according to Bloomberg.

• For October 24, authorities planned 20% more police than the previous Saturday, special intervention units, and soldiers to guard sensitive buildings. President Emmanuel Macron said only union-organized protests would be authorized, and that "the general recommendation is that there be no demonstrations in the coming days".

• October 23: a shooting in Trappes and La Verrière, west of Paris, left one dead and four wounded. The national anti-terrorism prosecutor took the case, Interior Minister Laurent Nuñez called it a terrorist crime, and press reports described a racist motive.

• October 24: the Interior Ministry counted 410,000 marchers (800,000 according to organizers), 1,314 arrests and 53 police officers injured. In Orléans, barricades were still up on Sunday.

• October 30: the "gilets jaunes" (yellow vests) are calling for "casserolades", pot-banging protests, in front of the National Assembly on Friday, in "defense of public freedoms" against what they call an authoritarian drift.

• November 5: the unions, led by the CGT, have called a general strike across the country, starting that day, covering transport, schools and hospitals.

Investors who used to buy France at 50 bp over Bunds "now won't even think about touching it", RBC BlueBay's chief investment officer told Bloomberg. When the natural buyers step back, the price is set by the marginal seller, and in a trending market that seller is the trend follower (or CTA, a systematic fund that buys what rises and sells what falls). The record build-up in OAT futures fits that picture.

The France Risk Gauge: a proprietary model, now at 81 out of 100

Standard risk dashboards treat France as a bond story. To capture the rest, I built the France Risk Gauge, a proprietary CrossVol model. The formulas stay in-house; here is what it does.

What it measures. One question: how much stress is France under, from the bond market to the street to the institutions? The answer is a single reading from 0 to 100, split into four zones: green (calm), amber (tension), orange (confidence crisis) and red (peril).

What goes in. Five blocks, each built from data that updates during the day:

1. The bond market: the OAT-Bund spread, the most direct price of French risk.

2. France against Italy: how France trades relative to the euro area's historical weak link.

3. Sovereign ratings: Moody's, S&P and Fitch, ratings and outlooks, updated automatically when an agency acts.

4. Political fear: daily counts of reported unrest and looting, attacks and far-right violence, and civil-war rhetoric, built from news wires and checked against official tallies (Interior Ministry, prosecutors) and weekly conflict-event data.

5. Constitutional risk: signals of governance by exception (protest bans, states of emergency, talk of constitutional revision, election delays or emergency powers), organized around three constitutional branches explained below.

How it reads. Each political and constitutional signal is compared with its own recent history, so a high reading means "unusual for France over the past six months", not just "a lot of headlines". The market blocks carry the most weight, because money is the final judge; the political blocks explain why money moves.

What it is for. Separating the fiscal part of the French risk premium from the political part, in real time.

On Sunday the gauge read 81 out of 100, in the red zone. The bond block alone reads 69. The political block pushes it into the red: over the past seven days, reports of unrest and looting ran at about six times their usual high level, and reports of attacks and far-right violence at about seven times.

The gauge does not try to predict riots. It measures how much of the move in French debt is political rather than fiscal. This week, most of it was.

The model: fear of the extremes sells the status quo

The gauge rests on a reading of how political shocks travel into prices.

Each shock gets attributed to one camp. Riots and looting feed the "danger from the left" story. An attack like Trappes feeds the "danger from the right" story. The two alternate, so the fear of extremes never comes back down. When fear is moderate, voters look for a providential figure from the center or the republican right: centrists such as Édouard Philippe or Gabriel Attal, and figures of the republican right such as Bruno Retailleau, Xavier Bertrand or David Lisnard. When fear becomes extreme, even a providential candidate looks like a gamble, and the status quo becomes the safe haven.

For a derivatives trader, the structure is familiar:

• fear of the extremes plays the role of implied volatility, the price the market puts on future swings;

• the providential figure is the protection you get sold;

• the OAT-Bund spread is the volatility index that confirms the fear;

• when fear reaches the tail, buyers stop paying for protection and go to cash, which here means the status quo.

The bond market prices this political volatility continuously and in size. That is why the spread works as France's political volatility index.

Constitutional risk: what Articles 16, 89 and 7 really allow

The constitutional block exists because the tail risk in France is governance by exception, not just a bad budget. The texts define which tails are real.

• Article 16 gives the President exceptional powers when the institutions, the independence of the nation, the integrity of its territory or its international commitments are under serious and immediate threat, and the regular functioning of the constitutional public authorities is interrupted. It gives powers, not time: it does not extend a mandate. Under the Constitutional Council's 1992 decision (92-312 DC), the Constitution cannot be revised while Article 16 is in force.

• Article 89 governs constitutional revision. The same text must pass both chambers, then either a referendum or a three-fifths majority of Parliament sitting together in Congress. No revision is allowed while the territory is under attack.

• Article 7 allows, and in some cases obliges, the Constitutional Council to postpone the presidential election only if a candidate dies or becomes incapacitated around the filing deadline or before the vote. Civil unrest is not a ground. During a vacancy of the presidency, the Constitution cannot be revised either.

The plausible tail is therefore the slow one: budgets passed by decree, protest bans, emergency measures, and an election campaign run under them. The gauge splits this into three branches (a revision under internal peril, an Article 7 delay, Article 16) and reports the most open one. This week, the revision branch saturated, led by protest bans: 22 reports a day, against a 90th percentile of 1.6.

For bond pricing, the slow version is enough. Institutional risk is what turns a deficit problem into a front-end problem.

Can the ECB save French bonds?

The obvious backstop is the European Central Bank. Its Transmission Protection Instrument (TPI) lets it buy a country's bonds, with no size limit set in advance, when spreads widen in a way it judges disorderly and unjustified by fundamentals. Purchases focus on maturities of 1 to 10 years, the exact segment under stress. Eligibility conditions include compliance with EU fiscal rules and an assessment of debt sustainability.

Two obstacles:

1. France has been under an excessive deficit procedure (the EU process for countries above the 3% deficit limit) since July 2024. Its fiscal path makes the "unjustified by fundamentals" test hard to pass, and the conditionality is political.

2. Former ECB president Mario Draghi said recently that ECB intervention for France is still a distant prospect. The race to succeed Christine Lagarde (Klaas Knot or Pablo Hernández de Cos) is increasingly framed around who can stand up to the bond vigilantes, investors who sell a government's debt to punish its fiscal policy. As Bloomberg put it on Sunday: "Never has market speculation focused so ominously on such a large and pivotal member of the currency bloc."

Until the ECB signals otherwise, trend followers remain the marginal price setters.

French bonds: the levels a quant trader watches

These are the levels and dates I am watching. They are a framework, not a recommendation.

• OAT-Bund 10Y. 167 bp (high 178). What would change the reading: A close above 180 bp opens a test of 200 bp, where the ECB intervention debate becomes unavoidable. A close back below 150 bp would mean the political premium is leaking out.

• France-Germany 2Y. 74 bp. What would change the reading: The cleanest gauge of political risk. If it keeps widening while the 10Y spread stabilizes, the market is pricing the next 18 months rather than the next decade.

• France 5Y CDS. 106 bp. What would change the reading: The gap to Italy (78 bp). France trading wider than Italy on CDS signals a lasting re-rating rather than a squeeze.

• CAC 40 vs DAX. CAC -4.5% YTD, DAX +2.8%. What would change the reading: French banks and domestic stocks carry the sovereign risk. A further spread widening without CAC underperformance would be an unusual divergence.

• National Assembly. Budget debate resumes next week. What would change the reading: Any use of decree powers is within Moody's base case. The market will read the deficit number, not the procedure.

• ECB decision. October 29. What would change the reading: Explicit language on TPI eligibility is the biggest potential spread compressor.

• Yellow vests "casserolades". October 30. What would change the reading: First test of the protest bans in front of Parliament.

• General strike. From November 5. What would change the reading: The biggest street test since the crisis began. The gauge's political block will measure it in real time.

Spreads already price a lot of political risk. But the natural buyers have left, positioning follows the trend, and the backstop is not in sight. In that setting, relief rallies like Friday's (-7 bp, the biggest daily tightening since October 6) can be sharp, and still short-lived.

FAQ: the French bond crisis in four questions

Why does France pay more than Italy to borrow? Because investors now price French political risk above Italian fiscal risk. Italy has a stable governing majority; France has a 5.4% deficit, no majority, an election in six months and mass protests. The 10-year gap reached 45 bp in France's disfavor on October 23, 2026.

What is the OAT-Bund spread today? 167 bp at the October 23, 2026 close, after a high of 178 bp on October 22, the widest since the 2011 euro crisis. It was 85 bp at the start of September.

Did Moody's downgrade France? No. On October 23, 2026, Moody's kept France at Aa3 with a negative outlook, one notch above S&P and Fitch, while warning that protests could constrain fiscal consolidation and that the 2027 budget may be adopted by decree.

Can the ECB stop the selloff? It has the tool, the TPI, but France's excessive deficit procedure complicates eligibility, and former ECB president Mario Draghi called intervention a distant prospect.

Further reading on CrossVol Research

• The OAT-Bund Spread Explained: France, Germany, and Sovereign Risk (September 19, 2026): the basics of reading the spread.

• Dispersion and the OAT-Bund Channel (September 19, 2026): written at 104.6 bp, "a structural repricing, not a headline".

• Volatility Is Starting to Talk Even While the VIX Sleeps (October 4, 2026): France as the epicenter of cross-asset volatility, at 141 bp.

• Stock Market Week in Review, October 19 to 23, 2026 (October 25, 2026): the AI selloff, the Treasury rally and oil at $117.

This article is for educational purposes only and does not constitute investment advice or a recommendation to buy or sell any security. Bonds, futures, CDS and options carry substantial risk, including the loss of more than the initial investment for leveraged instruments. Past patterns do not guarantee future results.

Data as of the close of Friday, October 16, 2026. VIX futures reopen Sunday evening, New York time.

Risk warning. This article is for education only. This is not investment advice. Futures and options, including VIX futures, VIX options and volatility exchange-traded products, can lose more than the initial investment. Leveraged and inverse products can lose most of their value in a single day.

Key takeaways

• The VIX closed Friday at 19.99, up 21% on the day. Some commentators have raised the prospect of a sharp Monday selloff. Our models say the setup matters more than the level.

• The VIX futures curve has inverted at the front. The October contract settled at 19.91, November at 19.24 and December at 18.95. One month ago the same months traded at 17.97, 18.65 and 19.03.

• Volatility exchange-traded products hold about 64,700 more VIX futures contracts long than short. In February 2018 the fragile side was short volatility. Today it is the other way around.

• If VIX futures jump 10 points in a session, our models put the forced buying by these products at about 18,300 contracts, or 8.4% of the recent average daily volume of the two front months.

• VIX calls outstanding with less than three months to expiry stand at 8.96 million contracts, near the highest level we have recorded. About 42% of it expires on Wednesday, October 21.

• The hedging of S&P 500 option dealers now amplifies moves, with the index 1.4% below the level where it would damp them.

1. Where we are

We have been writing about this for a month. On September 23, The VIX Is Too Calm for a 3.4 Sigma Rates Shock showed stocks and bonds falling together while equity volatility barely moved. On September 25, The Crash Fuse Is Lit argued that this cycle's fuse runs through AI spending financed by debt. On October 4, Volatility Is Starting to Talk Even While the VIX Sleeps flagged our cross asset volatility composite at systemic. On Friday, in Here We Are, equity volatility caught up.

The VIX measures the 30 day volatility of the S&P 500 that option prices imply. It closed Friday at 19.99, up 21% on the day, after trading as high as 22.31. VIX futures let investors buy or sell that number for a future date, one contract per month.

A futures curve lines up the price of every delivery month. In calm markets, later months cost more than near months, because investors pay up for protection further out. That shape is called contango. When the near months trade above the later months, the curve is in backwardation: investors pay more for protection now than later. That is what stress looks like on this curve.

One month ago, on September 17, the curve sloped up in the usual way: 17.97 for October, 18.65 for November, 19.03 for December. On Friday the front sat above the next months: 19.91 for October, 19.24 for November, 18.95 for December.

Two of our gauges confirm the stress. Our 30 day VIX futures gauge traded 0.68 points below the VIX index itself, one of its lowest readings of the past year. And November traded 3.4% below October, one of the steepest inversions of the past year. A future below the index itself means the market expects the spike to fade over the next month, while near-dated protection stays expensive.

2. Who is positioned, and how

Behind the VIX sit exchange-traded products that hold VIX futures. They let investors buy or sell volatility like a stock. Some are long volatility, such as $VXX, $VIXY, $UVXY (1.5 times leverage) and $UVIX (2 times). Others are short, such as $SVXY (minus 0.5 times) and $SVIX (minus 1 times). Each product tracks an index of short dated VIX futures. Every evening, the leveraged and inverse ones must reset their exposure to stay at their target leverage. That rebalancing is mechanical: it ignores the news.

Open interest is the number of contracts outstanding. Our models put these products at about 84,200 VIX futures contracts long and 19,900 contracts short on Friday, a net long position of about 64,700 contracts. Each contract moves $1,000 per VIX point, so the net position gains or loses about $65 million for each point the futures move.

This is the opposite of February 2018. On February 5, 2018, the risk sat with products that were short volatility. When the VIX exploded, they had to buy back futures into the close, and that buying fed the spike. The largest of them was terminated within days.

Today the short side is small, about 7% of the open interest of the two front months, and the long side dominates. That changes the shape of the risk, and some risk remains.

3. Why a shock would be amplified

Leveraged long products and inverse products both have to buy VIX futures after a day when volatility rises, and sell after a day when it falls. A product with twice the leverage that gains 20% in a day ends the day with too little exposure, so it buys. An inverse product that loses 20% ends the day with too much short exposure, so it buys back.

Our models put the buying needed after a 10 point jump in VIX futures at about 18,300 contracts. That is 8.4% of the recent average daily volume of the October and November contracts, and 6.3% of their open interest. On February 5, 2018, about 116,000 VIX futures contracts traded in a single minute just after the close, roughly a quarter of the whole market. We are well below that level. But forced buying grows with the size of the move, and in a disorderly market the volume available to absorb it tends to shrink.

The second layer is the options market on the VIX itself. Open interest in VIX calls expiring within three months reached about 8.96 million contracts on Friday, near the highest level we have recorded, and up 9.7% in five sessions. Call volume over the past week ran about 1.5 times its usual pace. Investors are buying protection against a volatility spike.

Dealers are the banks and market makers that sell options to investors and hedge the risk. The dealers who sold those calls hedge them with VIX futures. Gamma measures how fast an option's hedge changes when the underlying moves. Dealers who are short calls are short gamma: as the VIX rises, they must buy more VIX futures to stay hedged. That buying adds to the products' buying, on the same side, at the same time.

The third layer is the S&P 500. When option dealers on the S&P 500 hold a large long gamma position, they sell rallies and buy dips, which pins the index and keeps realized volatility, the volatility the index actually delivers, low. That damping is gone for now. Dealer gamma on both the S&P 500 and the Nasdaq 100 is negative in our model. The S&P 500 closed at 7,659, 1.4% below 7,766, the level where dealer hedging switches from damping moves to amplifying them. The Nasdaq 100 is 4.1% below its own switch level. Below that line, dealers sell into weakness and buy into strength.

4. What our models say about the next few days

These are scenarios, not forecasts. They describe what the positioning implies if a given move happens.

Scenario 1: the stress continues. If VIX futures add another 10 points, our models expect the products to buy about 18,300 contracts at the close. Dealers short VIX calls would add to that buying, and S&P 500 dealers would sell into weakness instead of absorbing it. Three mechanical buyers on the same side can make a volatility move overshoot. The short products have a hard limit: SVIX loses everything if its index doubles in a day, which would take a jump of about 19 points on VIX futures.

Scenario 2: the stress fades. If the VIX falls back, the same products sell futures at the close, about 6,000 contracts for a 10% drop in their benchmark index (about 2 points on VIX futures today). The curve can return to contango. The long products have their own limit: UVIX loses its entire value if its index falls 50% in a day, close to 9.6 points on VIX futures today. And the S&P 500 needs to climb back above 7,766 for dealer hedging to damp moves again. Until that line is recovered, a calmer market stays fragile.

The dates that matter:

• Tuesday, October 20. The products finish moving their exposure from the October to the November VIX future, at about 2,700 contracts a day in this final stretch.

• Wednesday, October 21. The October VIX futures and options expire: on expiry, a contract stops trading and is settled. About 3.8 million of the 8.96 million calls in our three month count are October calls. How much of that protection is rolled into November, rather than left to expire, will show how much fear remains.

• Late October. Earnings from the largest cloud and AI data center operators. The October 8 session, driven by OpenAI headlines, was a day our models flag as AI specific, with the Nasdaq 100 falling three times as much as the S&P 500 while Treasury yields fell.

• October 28. The Federal Reserve's policy decision.

5. The AI channel

Our stock and bond regime model is neutral. The recent correlation between Nasdaq 100 returns and changes in the 10 year yield is minus 0.18. A positive reading would mean stocks and yields fall together, so Treasuries hedge an equity portfolio. A negative reading means they do not. At minus 0.18, bonds are not yet a reliable hedge. AI specific sessions behave differently: on those days, yields fall with stocks. We have counted four of them in the past three months.

On the credit side, two of our five AI credit signals are on. $BIZD, a fund of business development companies that lend to mid-sized firms, is at its lowest level of the year. $ORCL is 59% below its 52 week high, with implied volatility near 60%. $CRWV fell 25% over five sessions. The US 10 year real yield, the Treasury yield after inflation, was at 2.78% on Thursday, near its highest level of the past year. Producer prices for semiconductors, a gauge of the AI spending race, dipped 0.34% in September after a 26% rise over twelve months. One month down is not a rollover, and we are watching the next print.

Bottom line

The VIX woke up on Friday, and it woke up loaded the other way. In 2018 the fragility sat with sellers of volatility. Today it sits with the mechanics that buy volatility when it rises: leveraged long products, dealers short VIX calls, and S&P 500 dealers who stopped damping moves. Each of these forces is moderate, and all three point in the same direction if the next shock comes. This week's expiry, next week's earnings and the Fed decision are the three dates to watch.

Risk warning. This article is for education only. This is not investment advice. Futures and options, including VIX futures, VIX options and volatility exchange-traded products, can lose more than the initial investment. Leveraged and inverse products can lose most of their value in a single day.

Prices are settlements of Friday, October 16, 2026.

Risk warning. This article is for education only. This is not investment advice. Futures, options and spreads can lose more than the initial margin. Past prices say nothing certain about future prices.

Key takeaways

• The November WTI contract rose 8.2% in five sessions, from $93.74 to $101.41 a barrel, and crossed $100 on Friday. December WTI rose 6.6%. December Brent rose 3.0%, to $109.86.

• Long-dated oil has started to move. Brent for November 2028 rose 6.5% in a month, to $78.46, its highest level since April.

• The geopolitical risk premium is largest in the front of the curve. December 2026 Brent trades $23.17 above December 2027, three times the gap of three months ago. December 2026 WTI trades $18.48 above December 2027.

• Diesel and gasoil (ICE gasoil, the European diesel benchmark) margins peaked in mid September and have eased since, while crude kept rising. Diesel remains the tightest product: December diesel costs $198 a barrel against $110 for Brent.

• Gasoline is the weak link. The gasoline margin lost more than a fifth of its value this week.

• European and Asian gas cost about $29 per million British thermal units, the US benchmark $3.12. The US curve shows its smallest end-of-winter spread since April.

• Energy equities outperformed while the S&P 500 had its worst week since June.

1. The week in numbers

A futures contract fixes today the price of a quantity delivered in a given month. The front contract is the nearest month still trading. WTI is the US crude benchmark, delivered in Cushing, Oklahoma. Brent is the benchmark for crude shipped by sea. ULSD (ultra-low sulfur diesel) and RBOB (a gasoline blend) are the US refined product futures. Henry Hub, in Louisiana, is the US natural gas benchmark. TTF is the Dutch gas hub that sets prices for Europe. Gas prices are quoted per million British thermal units (MMBtu), a unit of energy, or per megawatt hour (MWh) in Europe.

The November WTI contract stops trading on Tuesday, October 20. A contract this close to expiry can move more than the rest of the curve, so this article also shows December. Exchange-traded funds that hold these futures follow the same markets: $USO for WTI, $BNO for Brent, $UGA for gasoline and $UNG for Henry Hub gas.

Supply risk in the Gulf dominated the news flow. A building at Saudi Aramco's headquarters in Dhahran was reported hit on October 14, according to people familiar with the matter, and the crude rally accelerated on Thursday and Friday. Further attacks were reported in Yemen on October 17, the Strait of Hormuz stayed a focus of supply risk, and fresh disruptions hit Russian refineries. OPEC trimmed its 2026 demand growth forecast on October 13, citing attacks on Gulf shipping. In the other direction, the International Energy Agency reported record UAE output of 5.3 million barrels a day in September, condensates (light liquids produced with gas) included. US crude and diesel inventories fell again in the latest government data.

2. Brent and WTI futures curves: a shock priced for one year

A futures curve lines up the price of every delivery month. When near months cost more than later months, the curve is in backwardation: buyers pay a premium to receive oil now rather than later. Backwardation is the usual signature of a market short of physical barrels. The opposite shape, contango, appears when supply is ample and storage pays.

Both crude curves are in steep backwardation along their whole length:

A strip is the average price of a run of delivery months, here the twelve months of a calendar year. Brent's 2027 strip averages $95.4. For 2028, the contracts quoted so far (January to November) average $81.6. Two years out, Brent futures trade between $78 and $82. That anchor is itself moving. From April to mid September, the November 2028 Brent contract traded between $69.55 and $77.16 while the front month swung between $72 and $103. In the past month it rose 6.5%, to $78.46, its highest level since April. December 2027 gained 10.6% over the same month, slightly more than December 2026 (+9.0%). When the back of the curve rises with the front, the market is repricing the long-run cost of oil as well as the next deliveries.

The slope has moved fast. Three months ago, the Brent front month traded 15% above the contract twelve months later. One month ago the gap was 34%. Today it is 27%.

3. Prompt spreads and calendar spreads: where the premium sits

Two spreads measure the backwardation:

• the prompt spread compares the front month with the next one and measures the premium for barrels over the coming weeks;

• the calendar spread compares two contracts further apart. December 2026 against December 2027 is the reference one-year spread for hedgers and macro funds.

The expiring November WTI contract carried the sharpest prompt pressure: November traded $2.39 above December on Friday, against $0.88 on Monday. Buyers paid up for barrels deliverable at Cushing within weeks.

Further out, the picture is more nuanced. The December/January spreads are lower than in mid September. The one-year spreads, on the other hand, stay close to (Brent) or below (WTI) their September peaks:

• the WTI December/December spread widened from $14.14 on Monday to $18.48 on Friday, after a peak of $22.89 on September 15;

• the Brent December/December spread reached $23.17, close to its September 15 peak of $24.76, and in the top 2% of daily readings since April.

The premium has spread across the next twelve months. The market pays a premium for oil delivered at any point over the next year. The curve prices a supply shock concentrated in the next twelve months.

Some traders had positioned for the backwardation to narrow. In early October, options on WTI calendar spreads traded with put structures: 2,000 lots (contracts) of December/January 1x2 put spreads, which buy one put and sell two at a lower strike, with strikes at $0.50 and zero, and 500 lots of puts at -$0.50 on a January to June/July strip. A put spread on a calendar spread gains if the spread falls. The December/January spread is roughly unchanged since then, at $1.58 against $1.61 on October 2. The June/July 2027 spread rose from $1.06 to $1.50, against the bet.

4. Crack spreads: refining margins after the September peak

A refinery buys crude and sells products. The crack spread is the product price minus the crude price, both in dollars per barrel (a gallon price is multiplied by 42). It is a direct reading of the gross refining margin. The 3:2:1 crack combines them the way a typical US refinery does: two barrels of gasoline and one of diesel against three barrels of crude.

All legs on the December 2026 contracts ($/bbl):

From July to mid September, products led the move. The December diesel crack rose from $52 to $97 while Brent went from $72 to $101. Since then the distillate margins (diesel and gasoil) have eased by about $9 to $12 while Brent gained another $9. Gasoline and the combined 3:2:1 margin held up until early October, then dropped this week.

The week itself split the products:

• Gasoline. On the November contracts, the gasoline crack fell from $44.5 to $34.5, a loss of more than a fifth. The summer driving season is over, gasoline demand is seasonally weaker, and gasoline prices have not kept pace with crude. On December, the 3:2:1 margin fell $5.0 from Monday, and $6.4 from the previous Friday.

• Diesel. Against Brent, the December diesel crack edged up, from $86.0 on Monday to $88.5 on Friday. Against WTI, the November diesel crack was flat near $104. US distillate stocks remain near historical lows, and diesel futures jumped on Tuesday before partly retracing.

Diesel stays the tightest product. December diesel costs $198 a barrel once converted from its gallon price (November: $206), nearly twice the price of Brent. The third quarter was exceptional for refiners: ExxonMobil said higher refining margins added about $4.5 billion to its third-quarter results, after $4.1 billion in the second quarter, and TotalEnergies flagged strong downstream results.

5. The forward crack curve: a slow normalization of diesel

Futures also price refining margins for each delivery month. The diesel crack against Brent is priced at $83 for March 2027, $72 for June 2027 and $63 for December 2027. Gasoil follows the same path, from $74 for December 2026 to $49 a year later. Both stay above the levels of last June, when the December diesel crack traded between $45 and $58.

The gasoline crack shows its usual seasonal hump: about $33 through the winter, $40 to $45 for the April to September 2027 contracts, when summer-grade gasoline is more expensive to make and demand is higher, then $28 for December 2027.

The market therefore prices a gradual decline in diesel margins over more than a year, with no return to the spring levels before 2028.

6. Brent against WTI: the seaborne premium

Compared on the same December delivery, Brent traded about $4 to $7 above WTI through the summer. The gap widened to $12 to $14 in early October, as attention turned to risks to seaborne crude in the Gulf, and stood at $13.79 on Monday. It closed the week at $10.84. WTI caught up part of the gap, helped by US crude and diesel inventory draws, with strength spilling over from the expiring November contract.

7. Implied volatility: the OVX above 50

Implied volatility is the size of future price moves that option prices assume, expressed as an annual percentage. The OVX, published by Cboe, measures the 30-day implied volatility of crude oil. It rose in every session except Wednesday, from 48.52 to 50.74, with the largest jump on Friday as WTI broke $100. One-month implied volatility on WTI options reached 45.1% on Thursday, its highest from Monday to Thursday.

An OVX near 50 means option prices assume a typical daily move of about 3% over the next month (50% divided by the square root of 252 trading days). For comparison, the main equity volatility index usually sits between 12 and 20.

The volatility of the nearest option series appeared to fall on Thursday, from 40.1% to 35.4%. That day was the last trading day of the November WTI monthly options, so the reading moved from one option series to the next. It does not show a fall in volatility.

8. Listed options: where open interest sits

Open interest is the number of option contracts outstanding at each strike price. It shows where hedging flows can concentrate as prices move. A call gives the right to buy at the strike, a put the right to sell.

For the WTI weekly options expiring between October 19 and 22, all written on the December contract:

• call open interest is largest at $105, $100, $107.5 and $110, then at $98, $97.5 and $102.5;

• put open interest is largest at $92.5, $95 and $90;

• puts amount to about a third of calls on the nearest expiries, a put/call ratio of 0.37: open interest leans toward calls.

Gamma measures how fast an option's sensitivity to the underlying price changes as that price moves. Option dealers hedge that sensitivity by trading futures. If dealers are long the calls and short the puts (the usual convention, and an assumption, because open interest does not show who holds what), their hedging dampens moves above a certain price and adds to them below it. The largest gamma sits at the $98 strike of the December contract. The gamma flip, the price below which dealer hedging would start to add to moves rather than absorb them, is near $91.35.

Henry Hub gas shows the opposite set-up. For the November options expiring October 27, put open interest is heaviest at $2.75, $2.50 and $3.00, and call open interest at $3.50 and $3.25. The November contract, at $3.116, trades below its estimated gamma flip of about $3.28. On the same convention, hedging there would tend to add to price moves.

On ICE, Brent options turned back toward calls earlier in October after weeks of mostly bearish flows, with 17,000 lots of March $124/$125 and 13,000 lots of February $115/$116 call spreads. A call spread buys one call and sells a higher one, a capped bet on a rise often used by producers and macro funds.

9. Over-the-counter options: large puts far below the market

Options also trade over the counter (OTC), directly between banks and their clients, outside the exchanges.

Over the past 60 days, the WTI strikes with the most notional (the face value of the barrels covered) among OTC trades still live are puts: $216 million across 44 trades at $70, and $163 million across 16 trades at $60. Large out-of-the-money puts, with strikes well below the current price, are consistent with producers hedging future output. Call notional is spread from $76 to $110, largest at $90. Brent, TTF and NBP (the UK gas hub) options trade mostly in Europe and are not covered here.

10. Positioning: managed money and CTAs

The CFTC, the US derivatives regulator, publishes each week the futures positions of broad trader groups. Managed money covers hedge funds and commodity trading advisers. Swap dealers are mostly banks hedging deals with clients. Producers, merchants, processors and users are the commercial firms that produce, trade or consume the commodity.

In WTI, managed money cut its net long by about 13,000 contracts in the week to October 13. Swap dealers held a net short of 478,478 contracts, and commercial producers and users a net long of 353,143.

A report on October 13 put trend-following funds (CTAs, which buy and sell on price trends) at maximum long in oil, with their nearest selling trigger far below the market. Both readings can be true at once: trend followers are one part of managed money, and other funds in that group were reducing exposure. The CFTC data stops on Tuesday, October 13, so it does not show who bought during the Thursday and Friday surge. With trend followers already at maximum long, a fresh shock finds few sellers in that group, and a reversal of trend signals would be a risk on the way down.

Gasoline stands apart: managed money is heavily long while the gasoline crack is falling. In Henry Hub, funds remain net short and covered about 18,000 contracts.

11. Natural gas: Europe and Asia pay for winter, the US much less

Converted into the same unit, TTF at 87.55 euros per MWh equals about $28.8 per MMBtu. JKM, the benchmark for liquefied natural gas (LNG) delivered to Japan and Korea, trades at $29.9 for December. Henry Hub trades at $3.12.

Europe and Asia. TTF more than doubled since late June, from $13.6 to $28.8 per MMBtu, and gained 3.8% this week on tighter supply and stronger heating demand expectations. In September and early October, Europe paid more than Asia on several weeks; Asia is back at a $1.10 premium. LNG contracts indexed to oil usually price at 11% to 14% of the Brent price per barrel, a ratio called the slope. Spot TTF and JKM now trade at a slope of 26% to 27%, about double: spot gas is priced well above its usual link to oil.

The European curve. TTF stays between 85 and 88 euros through March 2027, then falls with the end of winter, to 74 euros for April 2027. The 2027 strip averages 71.0 euros. The 2028 contracts quoted so far (January to October) average 45.9 euros, about 35% lower: the curve prices much cheaper gas once this winter and the next have passed. The JKM curve shows the same shape, down to $14 for May 2028.

The United States. Henry Hub fell 1.6% on the week despite a brief spike on Thursday, when TC Energy declared force majeure (a supplier's legal release from delivery obligations after an event outside its control) on its Columbia Gas Transmission system following a leak on the Mountaineer Xpress line in West Virginia. Storage stands at 3,567 billion cubic feet after a 67 billion cubic feet injection, a build below expectations. The Henry Hub curve does carry a winter rise, from $3.12 for November to $3.76 for January. The seasonal spreads are small, though. The March/April 2027 spread, a classic bet on a cold end of winter nicknamed the widowmaker, trades at $0.08, its lowest level since April. US gas is abundant while Europe and Asia pay for winter.

12. Freight

The Baltic Dry Index, a gauge of the cost of shipping dry bulk cargo such as grains, coal and ore, fell 12% on the week. Rate indices for VLCC and Suezmax vessels (very large and mid-size crude tankers) were unchanged. The split is consistent with weaker dry bulk demand and with tankers kept busy by rerouted voyages.

13. Equities and rates

Energy stocks led: BP ($BP) +7.5%, ConocoPhillips ($COP) +6.5%, Shell ($SHEL) +5.0%, Chevron ($CVX) +4.7%, ExxonMobil ($XOM) +4.0%, TotalEnergies ($TTE) +3.4%, the energy sector fund $XLE +3.4% and the exploration and production fund $XOP +3.1%. Oil-field services lagged, with SLB ($SLB) up 1.5%. Options activity in the sector was heavily tilted to calls: on October 16, Occidental Petroleum ($OXY) traded 96,282 calls against 8,330 puts.

The S&P 500 had its worst week since June, with chipmakers down for an eighth session in a row, their longest losing streak since 2008. Higher fuel costs are feeding into prices: September consumer and producer prices both surprised to the upside, reinforcing expectations of further Federal Reserve rate increases. Long-dated Treasury yields rose more than short-dated ones on Friday, a steepening of the yield curve.

14. My read

The commodities strategist Jeff Currie wrote this week that "the macro repricing of energy is now underway", pointing to the move in long-dated oil. My reading of the curves is close to his.

For most of this year, the back of the oil curve ignored the front. Between April and mid September, December 2026 Brent swung across a $31 range while the November 2028 contract stayed within $7.6. A market that sees a temporary shock behaves exactly like that: it pays up for prompt barrels and leaves the long-run price alone. That changed over the past month. The November 2028 Brent contract broke above its range, and the December 2027 contract rose faster than December 2026 in percentage terms. When the whole curve lifts, the market is raising its estimate of what oil costs over the next several years, beyond the few missing cargoes of the moment.

The reasons are visible in the data of this article. Refining capacity disrupted in the Gulf and in Russia keeps diesel near twice the price of crude, and the forward diesel cracks stay above last June's levels into early 2028. Gas in Europe and Asia trades at double its usual link to oil. Hedgers and investors who need energy two and three years out are paying more for it, and they are typically the ones who move the long end.

I therefore see energy settling on a higher plateau: the spike in the front may fade, and the floor under it has moved up. The front of the curve can give back part of its premium, through calmer headlines or the expiry of crowded contracts, without changing that view. The signal I watch is the back end. If the 2028 contracts hold above their old range of $70 to $77 while the front corrects, the repricing is real. If they fall back into that range, the market will have treated this year as a shock after all.

Glossary

• Backwardation / contango: a futures curve where near months are more / less expensive than later months.

• Prompt spread: front-month price minus second-month price.

• Calendar spread: price difference between two delivery months, for example December 2026 minus December 2027.

• Strip: average price of a series of delivery months, often a calendar year.

• Crack spread: price of a refined product minus the price of crude, in dollars per barrel.

• 3:2:1 crack: margin of two barrels of gasoline and one of diesel against three barrels of crude.

• Implied volatility: expected size of price moves embedded in option prices, as an annual percentage.

• OVX: Cboe index of 30-day implied volatility on crude oil.

• Open interest: number of contracts outstanding.

• Gamma / gamma flip: rate of change of an option's sensitivity to the underlying price; the flip is the price where dealer hedging would switch from absorbing to adding to moves.

• Managed money / CTA: hedge funds and commodity trading advisers / trend-following funds.

• MMBtu: million British thermal units, the unit of energy used to price natural gas.

• TTF / JKM / Henry Hub: gas benchmarks for Europe, Northeast Asia LNG and the United States.

Further reading

• Will Trump Strike Iran After the Midterms? A Brent Options Case Study

• Energy Vol: TotalEnergies Skew and the Normalization Trade

• The Volatility Risk Premium Explained: Why Implied Beats Realized

Risk warning. This article is for education only. This is not investment advice. Futures, options and spreads can lose more than the initial margin. Past prices say nothing certain about future prices.

Data: CME, ICE, CFTC, EIA and Bloomberg. Calculations: CrossVol.

Educational content, not investment advice. Options and futures carry a high risk of loss.

Here we are.

As of this capture, with the US session still open, the S&P 500 is down 1.15%, the Nasdaq 100 down 2.29% and Russell 2000 futures down 1.70%. The VIX is at 22.24, up 35%. The VVIX, the volatility of the VIX itself, is at 121, up 31%. The Nations TailDex, which tracks the cost of 30 day puts three standard deviations out of the money on the S&P 500, is up 71%.

What we wrote before today

On September 23 we wrote that the VIX was too calm for a 3.4 sigma rates shock. Treasuries had sold off hard, stocks and bonds fell together, and equity implied volatility barely moved. On October 4, we wrote that volatility was starting to talk even while the VIX slept, with our cross asset vol composite already at systemic. Today equity vol caught up.

On September 25, The Crash Fuse Is Lit argued that this time the fuse runs through AI capex financed by debt. Today the Nasdaq 100 is falling twice as fast as the S&P 500.

The test we set ourselves

On October 15, The Pension Valve Only Opens One Way set a test. On a day US stocks fall more than 1%, the long end of the Treasury curve should do better than the 2 year during US hours. Today is the first live candidate. We will report the verdict after the close.

Educational content, not investment advice. Options and futures carry a high risk of loss.

Market commentary for educational purposes. This is not investment advice and not a recommendation to buy or sell anything. Levels are as of 4:00 am New York time on Thursday, October 15, 2026, unless another time is given, and markets move fast. Options, futures and bonds carry a high risk of loss.

At 8:30 am New York time on Wednesday, October 14, the September consumer price index came out hot. Prices rose 0.6 percent on the month and 3.7 percent on the year, against 3.6 percent expected. Core prices, which leave out food and energy, rose 0.3 percent on the month against 0.2 percent expected, and 2.6 percent on the year, up from 2.4. Treasuries were sold on the print, as they should be after a number like that. Then the long end came back, and it came back further than it had fallen. By 2:25 pm the ultra bond future, which tracks Treasuries maturing in about 25 to 30 years, had won back 133 percent of its post CPI loss. The 2 year note future had won back 89 percent. On the Treasury's own daily curve, set from prices at about 3:30 pm, the 2 year yield ended 3 basis points higher at 4.82 percent and the 30 year 4 basis points lower at 5.52 percent. The S&P 500 closed down 1.11 percent at 7,718.94.

In plain terms. A bond's price and its yield move in opposite directions: when buyers push a bond's price up, the yield a new buyer locks in goes down. A basis point is one hundredth of a percentage point. So a hot inflation number pushed short term yields up, as you would expect when the Fed may have to hike again, and yet the longest bonds ended the day more expensive than they were just before the number. Someone bought the long end, and mostly the long end, while stocks kept falling.

My reading is this. The 100 largest US corporate pension plans were 114.5 percent funded on September 30, meaning their assets covered 114.5 percent of the pensions they have promised, their best level since June 2001. At that level, the trigger of liability driven investing, or LDI, works only one way: sell equities, buy the long end, and keep cutting equity exposure. LDI is the practice of running a pension's assets so that they move like its promises to retirees, above all when interest rates move. October 14 showed that trigger at work during US trading hours, in three waves. Overnight, the unwind of the AI trade in Asia added equity selling of its own, and it will not bring the pensions back to buy.

What I wrote three weeks ago

On September 25, in The Crash Fuse Is Lit, I wrote that "US corporate pension funding ratios are near 25 year highs" and that selling equities to buy long duration bonds "is the mechanical output of every major pension's glide path policy." I also set the limits of that bid. "Glide paths are slow," I wrote. "It is a structural tailwind, not a tactical floor." On insurers: "they buy investment grade corporates, not Treasuries." And my conclusion: "The 'pension bid saves the trade' argument is a tailwind, not a backstop."

The same piece put a date on the risk. "Quarter end is not a tail scenario," I wrote. "It is the scheduled moment when every mark to market loss in this chain becomes a reported number." That line was about the chain of debt that finances AI. It holds just as well for pensions, because quarter end is also when the funded ratio becomes a reported number, the one investment committees act on.

Three things have changed since. The September 30 measurement came in at 114.5 percent, with the accounting discount rate at its highest since 2009, and that very likely put many plans across new steps of their glide paths, the schedules that cut equities as funding improves. The CPI day showed the bid during US hours, strongest in the long end, with stocks sold at the same time. And the AI trade is now unwinding, which adds equity selling that the pensions will not absorb. One thing has not changed. A tailwind is still not a backstop, and I come back to that at the end.

October 14, in three waves

I timed the day on CME Treasury futures and the S&P 500 future, December 2026 contracts, in 5 minute bars, New York time. Bond futures rise when yields fall. The six Treasury contracts run from the 2 year note (ZT) through the 5 year (ZF), the 10 year (ZN), the ultra 10 year (TN) and the bond (ZB) to the ultra bond (UB).

Wave 1, bonds bid into the print. From the reopening at 6:00 pm on Tuesday to 8:25 am on Wednesday, Treasury futures gained across the curve, about 2 basis points in yield terms: 0.375 point on the ultra bond, 0.17 on the 10 year, 0.05 on the 2 year. Bloomberg tied the move to a rally in UK gilts, and CME data showed new long positions added in 2 year and 5 year futures on Tuesday.

Wave 2, sold on the print. At 8:30 am the number hit, and every Treasury contract made its low at 8:55 am. The front end led. On Bloomberg, the 2 year yield jumped as much as 16 basis points, to about 4.95 percent, and the 30 year peaked at 5.62 percent shortly after 9:00 am. That is a bear flattener: yields rise, short yields rise more than long ones, and the curve between them flattens. The gap between the 30 year and the 2 year yields, the 2s30s curve, fell from about 77 basis points before the print to 67.2 at 9:00 am. After the print, the market put the odds of another Fed hike on October 28 at about 54 percent, according to Bloomberg. The Fed had already raised its target range by a quarter point on September 16, to 3.75 to 4.00 percent.

Wave 3, the long end comes back. From the 8:55 am low to a high at 2:25 pm, every Treasury future recovered, and the longer the maturity, the more it recovered. Measured against what each had lost between 8:25 and 8:55 am, the 2 year won back 89 percent, the 5 year 98 percent, the 10 year 105 percent, the ultra 10 year 108 percent, the bond 120 percent and the ultra bond 133 percent. Above 100 percent, a contract is back above its price from before the print. Bloomberg's headline on the move read "Treasuries Long-End Bounces Back, Leading a Post-CPI Recovery," and a US rates trader it quoted attributed the buying to both real money and fast money. Real money means long term investors without leverage who buy to hold, such as pensions, insurers and asset managers. Fast money means leveraged, short horizon traders such as hedge funds and CTAs, the trend following funds.

Stocks did the opposite. The S&P 500 future stood at 7,878.5 at 8:25 am and kept falling while the long end climbed, to a low of 7,741 at 2:00 pm, 1.7 percent below its level before the print. Between the 8:55 am low and the 2:25 pm high, the ultra bond future gained 1.9 points, or 1.8 percent. From 8:55 am to its 2:00 pm low, the S&P 500 future lost another 1.1 percent.

The day ended in a twist, short yields up and long yields down. On the Treasury's curve, set at about 3:30 pm, the 2 year yield rose 3 basis points to 4.82 percent and the 10 year fell 2 to 5.19 percent. The 20 year fell 3 and the 30 year 4, to 5.52 percent. In futures, the 2 year closed below its price from just before the print, and the ultra bond closed 0.25 point above it, 0.6 percent above Tuesday's close. A hot inflation number, a front end that won back most of its loss but not all of it, a long end that more than recovered, and stocks that stayed sold.

Why "opportunistic buying" does not fit

The first explanation you hear for a day like this is opportunistic buying: yields spiked, investors with cash saw value, and traders who were short took their profits. Some of that surely happened, and the trader Bloomberg quoted said fast money was there too. But three facts do not fit a story of value hunters alone.

First, the choice of maturity. The 2 year cheapened most after the print, about 16 basis points at the peak against roughly 7 for the 30 year, measured from Tuesday's close. A buyer looking only for value had more of it at the front end. The bid went to the long end instead, and the recovery rose step by step with maturity, from 89 to 133 percent. The new long positions added on Tuesday, the ones the print caught on the wrong side, were in 2 year and 5 year futures, and those are the two contracts that recovered least.

Second, the front end did not follow all the way. The 2 year never won back more than 89 percent of its loss and closed with a higher yield than before the print, and after the print the market priced slightly better than even odds of a hike on October 28. The long end did not rally because the market priced a softer Fed. It rallied because someone wanted long bonds. There is a macro reading of the same shape, in which a Fed that keeps hiking will slow the economy, so long yields fall while short yields rise. I do not rule it out, and it is probably part of the day. But it does not explain a bid that came in over five and a half hours, strongest in the longest contracts, two weeks after a quarter end that put many of the largest US pension plans across new derisking steps.

Third, stocks were sold while the long end was bought. On its own, that could be a plain flight to safety. Together with the first two facts, it is the signature of a buyer whose liabilities are long and whose equity target has just been cut.

Two pieces of evidence point the other way, and they deserve a straight answer.

ETF flows. Bloomberg data dated October 14 show 1.40 billion dollars out of SPY, the most traded S&P 500 fund, and outflows from bond funds too: 180 million dollars from TLT (Treasuries over 20 years), 269 million from IEF (7 to 10 years), 205 million from LQD (investment grade corporates) and 154 million from HYG (high yield). The only inflow among them was 16 million dollars into SHY, which holds 1 to 3 year Treasuries. If pensions were buying duration, meaning long bonds whose value moves most with rates, why did money leave long bond funds? Because pensions do not buy duration through ETFs. They buy it through separate accounts, cash bonds, STRIPS (zero coupon pieces of Treasury bonds), futures and swaps. Flows reported for a given day also often reflect orders from the day before. ETF flows neither prove nor disprove a pension bid. The shape of the curve during US hours is better evidence.

Swap spreads. A swap spread is the swap rate minus the Treasury yield of the same maturity. At 30 years it has been below zero for years, which means swaps yield less than Treasuries. When an LDI manager hedges with swaps, it receives the fixed rate, which pushes swap rates down and the spread further below zero. On October 14 the 30 year swap spread, on swaps tied to SOFR, the overnight rate published by the New York Fed, did the opposite. It closed at minus 63.38 basis points, its least negative level since January, and Bloomberg's headline that day read "Dollar Swap Spreads Back to Multi-Month Highs as Bonds Selloff." At 3:55 am this morning it stood at minus 63.39, against a one year range of minus 83.64 on March 13 to minus 62.75 on January 15. So there is no sign of a new wave of hedging through swaps. The move fits buying of cash bonds in the afternoon, which pushes Treasury yields down and lifts the spread, but the rise began during the selloff, so it does not prove it.

LDI in one page

A corporate defined benefit plan promises employees a pension for life. The plan's liabilities are the present value of those promises, called the PBO, for projected benefit obligation, in the accounts. The funded ratio, or funded status, is the plan's assets divided by those liabilities. Above 100 percent, the plan holds more than it owes, and the difference is its surplus.

The discount rate sets the size of the liabilities. Under US accounting rules (ASC 715, and IAS 19 in most other countries), the promises are discounted at the yield of high quality corporate bonds, in practice AA rated, on the measurement date. When that yield rises, the present value of the promises falls and the funded ratio improves. Milliman's discount rate was 6.47 percent on September 30.

One confusion is worth clearing up, because it is everywhere. You will read that the pension discount rate is set by the IRS segment rates. Those are corporate bond yields published by the IRS. For minimum funding contributions they are 24 month averages kept inside a corridor, and they move slowly by design. The premiums plans pay to the PBGC, the federal pension insurer, use the monthly rates instead. They are not what goes into the funded ratio a company reports, and they are not what an investment committee follows on its glide path.

The hedge ratio is the share of the liabilities' sensitivity to rates that the assets match. A plan hedged at 100 percent sees its assets and its liabilities move by the same dollar amount when rates move. The tools are long Treasuries, STRIPS (Treasury bonds cut into zero coupon pieces, the longest duration on offer), long corporate bonds and interest rate swaps. Treasuries alone are not enough, because the discount rate is a corporate yield: on September 30 the 30 year Treasury yielded 5.64 percent, 83 basis points less than the 6.47 percent discount rate. So LDI mixes Treasuries and STRIPS for the rate part of the discount rate with long corporate credit for the spread part.

The glide path is the plan's road map. It sets steps of funded ratio, every few points between roughly 90 and 120 percent. At each step crossed on the way up, the target for hedging assets rises and the target for return seeking assets, mostly equities, falls. One large US plan already held 78 percent in fixed income and 18 percent in equities by mid 2025, a plan near the end of the road. The end of the road is a pension risk transfer, or PRT: the plan pays an insurer to take over its promises, and the insurer invests the premium, mostly in investment grade corporate credit.

September 30 pulled the trigger

The Milliman 100 Pension Funding Index tracks the 100 largest US corporate defined benefit plans. On September 30 it stood at 114.5 percent, up from 112.2 percent at the end of August and 105.0 percent a year earlier. The surplus was 160 billion dollars, on 1,263 billion dollars of assets against 1,103 billion of liabilities. That is the best level since June 2001, when the index stood at 117.2 percent, and it is above the October 2007 peak of 108.1 percent.

How September got there matters. Plan investments returned minus 2.21 percent in the month, and assets fell by about 35 billion dollars, benefit payments included. But the discount rate rose 47 basis points to 6.47 percent, its highest since May 2009, and liabilities fell by 54 billion dollars. Net, funding improved by 19 billion dollars. Other trackers tell the same story at the same date: Wilshire 113.6 percent, Marsh for S&P 1500 sponsors 113 percent, LGIMA 112.3 percent, Aon for S&P 500 sponsors 109.8 percent and MetLife 109.7 percent.

This is the accounting link I pointed to in September. Plans check their funded ratio against their glide path on a schedule, monthly or quarterly for many, and quarter end is the measurement most committees look at. Many large plans also have their LDI manager watch the triggers daily, so part of the move happened in September itself. The index gained 2.3 points in September and 9.5 points in a year, and individual plans sit on both sides of the average, so the September 30 number put many plans across new steps. The orders that follow a step crossed at quarter end are executed in the weeks after it, which means October: sell equities, buy long Treasuries, STRIPS and long credit. That is consistent with an October 14 on which the long end was bought while stocks stayed sold.

Since the measurement, long yields have fallen. The 30 year went from 5.64 percent on September 30 to 5.52 percent on October 14 on the Treasury's curve. Bloomberg's US pension liability index, I02786, stood at 3,923 when I wrote in September. It fell in the last days of September, as yields rose into quarter end, and has gained 1.16 percent since, to 3,906 on October 14, because the value of the liabilities rises when yields fall. Its yield, 6.38 percent, sits close to Milliman's discount rate. So funded ratios have probably slipped a little in October. In a ratchet, that does not undo the step.

A ratchet, not a rebalance

This is the heart of my view, and it is where most commentary goes wrong. Two different mechanisms are both called pension rebalancing, and they push in opposite directions.

The first is rebalancing to fixed weights. A plan that holds, say, 60 percent equities and 40 percent bonds buys equities after a fall to get back to 60 percent. It is contrarian by construction. This is the quarter end pension rebalancing that strategists estimate every quarter, and it supports stocks after a drop.

The second is the funded ratio glide path. Each step crossed on the way up raises the hedging target and cuts the equity target. A later fall in equities lowers the funded ratio, but it does not lift the equity target back. Most glide paths are built as ratchets: they derisk, they do not re-risk.

At 114.5 percent, the second mechanism dominates for US corporate plans. Most of them are above the steps that matter, and their goal is no longer to beat a benchmark but to lock in the surplus and prepare a transfer to an insurer. The equity target was cut at the September 30 measurement, or at an earlier one, and a fall in stocks does not restore it. That is why, at 114.5 percent funded, the LDI trigger only works one way: sell equities, buy the long end, and keep reducing equity exposure.

Some plans still sit below their steps, and some run on fixed weights, as many public pensions do. They will buy stocks after a fall, and the net flow depends on the mix. For the largest corporate plans, the ones Milliman tracks, the direction is set.

Numbers help. Duration measures how much a value moves when rates move: liabilities with a duration of 12 years gain or lose about 1.2 percent for each 10 basis points. Take a plan at 114.5 percent. With 20 percent of its assets in equities, a 1 percent fall in the S&P 500 costs it about 0.23 point of funded ratio, 0.21 with 18 percent in equities and 0.25 with 22 percent. A 10 basis point rise in the discount rate adds 1.39 points to a plan with no rate hedge and 0.42 point to a plan hedged at 80 percent. It still adds 0.18 point at 100 percent, because a plan in surplus gains in ratio when its assets and liabilities fall by the same dollar amount.

October 14 itself lowered funded ratios. For the same plan, 80 percent hedged, the S&P 500 down 1.11 percent and rates down about 3 basis points cost about 0.38 point. So the trigger was not the day. It was the September 30 measurement. A 1 percent fall in equities does not force a pension to sell. It trims its ratio by about a fifth of a point. What sells is the target that was already cut. And for a plan that hedges most of its rate risk, equities are now the main risk left in the portfolio, which is the strongest reason to keep cutting them.

How large is the stock? Private US defined benefit plans held about 3.47 trillion dollars of assets at the end of June 2026, against 3.12 trillion dollars of promises, according to the Federal Reserve, a funded ratio of about 111 percent across all of them. If 18 to 22 percent of those assets is still in equities, close to the large plan above, that is about 0.62 to 0.76 trillion dollars of equities. That is a stock still held, not a sell order. But in a ratchet it can only go one way.

The night of October 15, and the AI loop

Overnight, the selling moved to Asia and to the AI trade. Korea's Kospi fell about 9 percent and triggered a trading halt, with Samsung Electronics and SK Hynix down as much as 11 percent. Japan's Nikkei and Taiwan's Taiex lost more than 4 percent each, and the MSCI Asia Pacific index fell 3.2 percent, according to Bloomberg. On Wednesday, CoreWeave and the other neoclouds, the companies that rent out AI computing capacity, had already fallen, and the Nasdaq 100 had lost about 2 percent. At 3:00 am New York time, the S&P 500 future was down 0.4 percent from Wednesday's close and the Nasdaq 100 future 1.2 percent.

Treasuries rallied, almost in parallel. Around 3:50 am, on Bloomberg, the 2 year yield was at 4.73 percent, 7 basis points below its Wednesday close of 4.80 percent, and the 30 year at 5.43 percent, 9 basis points below its close of 5.52 percent. Treasury futures prices rose across the curve, by 0.15 percent for the 2 year note and 1.4 percent for the ultra bond. The 2s30s curve swung between 69.4 and 75.6 basis points, the high in the thin hour after the evening reopening, and was at 70.7 at 3:54 am, little changed from its close of 71.9. A bull steepener, in which yields fall and short yields fall more, or a bull flattener, in which long yields fall more, would each tell a story. The curve told neither. It moved down nearly in one piece.

Two things follow. First, this is not the pensions' hour. US pension plans and their LDI managers trade mostly during US hours. The night belongs mostly to Asian and European banks, insurers, trend followers and macro hedgers, and nothing in its shape can be read as pension buying. The signature to look for was in US hours on October 14, and it was there.

Second, the night matters for pensions in another way. Lower stocks and lower yields both lower the funded ratio. On a ratchet, a lower ratio does not bring pensions back to buy equities. So the AI unwind adds equity selling, and the pensions, the natural dip buyers of a world run on fixed weights, do not step in. They keep executing the cut decided on September 30. On a down day, pensions are not forced sellers. They are missing buyers, and sellers on schedule. The AI unwind and the pension rotation push equities the same way.

How much is left

Most of the rotation has been done. The Milliman ratio went from 105.0 to 114.5 percent in a year, demand for STRIPS has slowed, and Bloomberg reported in June that pension demand for corporate bonds may be fading as plans are overfunded. I have seen the cycle described as 80 to 85 percent complete. Nobody can measure that, and I will not use a number.

STRIPS are the cleanest gauge of pension demand for very long duration, because their main buyers are pension funds and insurers matching very long liabilities. Net stripping, the monthly change in the amount of Treasury bonds held as STRIPS, averaged 6.1 billion dollars a month in 2023 and 5.6 billion in 2024, then 4.1 billion in 2025 and 3.3 billion from January to September 2026, with 2.8 billion a month in the third quarter and 2.1 billion in September. The flow is slowing. It has not reversed, even if single months have dipped below zero, as in February and June 2025 and in April 2026. October's number will be published in early November.

What remains is the end of the glide paths for plans still short of their targets, buying on yield spikes by plans already there, and pension risk transfers. Transfers reached 48 billion dollars in 2025, according to LIMRA, the fourth year above 45 billion. The first quarter of 2026 was slow at 3.8 billion, down 47 percent on a year earlier as deals were pulled into the fourth quarter of 2025, and Bloomberg Intelligence expects 40 to 50 billion dollars for the year. At 114.5 percent, many more plans can afford a transfer, and each one hands liabilities to an insurer that buys corporate credit with the premium. Pensions and insurers already hold about 6.4 trillion dollars of US dollar investment grade credit, roughly 40 percent of the market, according to a Goldman Sachs estimate reported by Bloomberg Intelligence.

The rest is smaller than in 2023 and 2024. But it goes one way. And the plans that gained most this year have the most left to do. October Three's model plan with a traditional 60/40 mix gained 13 points of funded ratio this year to September 30, against 3 points for its model plan with 80 percent in bonds. The less a plan was hedged, the more it gained from higher rates and rising stocks, and the more derisking it now has to do.

Insurers and the rest of the world

Insurers do not run glide paths. They match their assets to their liabilities, mostly annuities and the pension promises they take over in transfers. As I wrote in September, their demand "lands in 5 to 15 year corporates, not in long Treasuries." Through transfers, part of the pension bid moves from long Treasuries to corporate credit.

Not every pension buys. In Chile, the private pension fund managers, the AFPs, are sitting on losses of about 7.7 billion dollars on swaps in which they received a fixed US rate, a bet that rates would fall, according to Bloomberg Línea, and a new rule is forcing them to cut those positions. Unwinding a bet on falling rates means selling duration. The word pension does not make every flow a bid.

And as I wrote in September, "The US is the only major market where the LDI derisking trigger is live and mechanically compelling right now." European insurers are already matched under Solvency II, the EU's insurance capital rules, the Dutch pension reform is a reallocation that can add volatility rather than a bid, and the Japanese bid is opportunistic and depends on the currency. I see nothing since that changes it.

What would prove me wrong, and what I watch

A tailwind is still not a backstop. The pension bid did not stop the 30 year from reaching 5.67 percent on October 7, its 52 week high, about 15 basis points above where it closed on October 14. At Wednesday's close, the 30 year was still 158 basis points above its three year low of 3.94 percent, set on September 16, 2024, on the Treasury's curve. Treasury supply is still a multiple of any pension flow, and after the CPI the market saw about 54 percent odds of a Fed hike on October 28. The pension bid decides which way the flows go when stocks fall. It does not cap yields.

Three things would prove me wrong. The first would be the long end failing to outperform during US hours on the next big down day for stocks, or selling off with the front end. That would make October 14 a one off. The second would be net stripping turning negative in the October data and staying negative in the months after, which would mean the cleanest long duration bid has reversed. One negative month proves nothing. April 2026 was one, and so were February and June 2025. The third would be large plans adding to equities after a fall, which would mean the ratchet is not a ratchet. A fall in funded ratios alone would not prove me wrong. A ratchet that stops climbing does not climb back down.

What I watch:

Milliman's October index and the other trackers, published in early November.

Net stripping for October, in the Treasury's monthly statement of early November.

The Fed on October 28.

The 30 year against its 5.67 percent high, and the 30 year swap spread against minus 62.75 basis points, its January high.

The long end during US hours on the days stocks fall.

If I am right, every equity rally is a chance for pensions to sell, every spike in long yields a chance for them to buy the long end, and every sell off in stocks will find them absent. The valve opens one way.

Key terms, in plain English

Basis point. One hundredth of a percentage point. 100 basis points make 1 percentage point.

Yield and price. The yield is the annual return a buyer earns by holding a bond to maturity at today's price. Price and yield move in opposite directions.

LDI, liability driven investing. Running a pension's assets so that they move like its promises to retirees, above all when rates move.

Funded ratio. Also called funded status. Plan assets divided by the present value of the pensions promised. Above 100 percent, the plan is in surplus.

Accounting discount rate. The yield of AA rated corporate bonds used under ASC 715 to value the promises in the accounts. When it rises, liabilities fall.

IRS segment rates. Corporate yields published by the IRS, as 24 month averages for minimum funding contributions and as monthly rates for PBGC premiums. Not the rate behind the reported funded ratio.

Glide path. The road map that cuts equities and adds hedging bonds each time the funded ratio crosses a preset step.

Hedge ratio. The share of the liabilities' sensitivity to rates matched by bonds or swaps on the asset side.

Duration. How much a value moves when rates move. A duration of 12 years means about 1.2 percent per 10 basis points.

STRIPS. Treasury bonds cut into separate zero coupon pieces, the longest duration a pension can buy in Treasuries.

PRT, pension risk transfer. A plan pays an insurer to take over its pension promises. The last step of derisking.

Swap spread. The swap rate minus the Treasury yield of the same maturity. Negative at 30 years for years.

SOFR. The Secured Overnight Financing Rate, the overnight rate published by the New York Fed that US dollar swaps now reference.

Bear and bull flatteners and steepeners. Bear means yields rise, bull means they fall. Flattener means the gap between long and short yields narrows, steepener that it widens. A twist is when short and long yields move in opposite directions, as they did on October 14.

Real money and fast money. Long term investors who buy to hold without leverage, such as pensions and insurers, against leveraged short horizon traders, such as hedge funds and CTAs.

CTA. A commodity trading advisor, in practice a fund that follows price trends with futures.

ZT, ZF, ZN, TN, ZB, UB and ES. CME futures on 2, 5 and 10 year Treasury notes, ultra 10 year notes, bonds and ultra bonds, and on the S&P 500.

This article is for educational purposes only. It is not investment advice, a solicitation, or an offer to buy or sell any security or derivative. Levels are as of the dates and times stated and can change quickly. Options, futures and bonds carry a high risk of loss. The funded ratio sensitivities are illustrations built on stated assumptions, not the position of any actual plan.

$SPY $TLT $LQD $CRWV

Data: Bloomberg, Milliman, US Treasury, CME, Federal Reserve, LIMRA.

Market commentary for educational purposes. This is not investment advice and not a recommendation to buy or sell anything. Levels are as of the dates and times given, markets move fast, and credit, equities and derivatives can lose far more than you expect.

The first scene of Act 4 played today, and it played in credit. At 13:22 New York time, CoreWeave's 9 5/8 percent senior notes due July 15, 2032 were quoted at 79.82, down 4.63 points on the day, for a yield of 14.96 percent, a Z spread of 1,021 basis points and an I spread of 1,027. Oracle's 6.85 percent bond of 2066 traded at a G spread of 300 basis points, the widest since it was issued in February. Oracle's five year CDS closed on Monday at 251.9 basis points, 3.7 times its level of a year ago. In Seoul, SK hynix closed down 10.09 percent and Samsung Electronics down 7.39 percent, with CXMT, a Chinese memory maker, claiming a denser DRAM cell.

In plain terms. Bond prices are quoted per 100 of face value, so CoreWeave's notes, sold at 100 in June, now trade near 80 cents on the dollar. The yield is the annual return a buyer earns at that price if the company pays every coupon and repays the principal at maturity. A spread is the extra yield a company pays over a benchmark: US Treasuries for the G spread, interest rate swaps for the I and Z spreads. It is measured in basis points, hundredths of a percentage point. A spread of 1,000 basis points means the market demands about 10 percentage points a year more from CoreWeave than from the benchmark. A credit default swap, or CDS, works like insurance against default. Oracle's five year CDS cost 251.9 basis points a year at Monday's close, about 2.5 percent of the amount insured each year, for five years of cover.

The indices have not seen any of this. On my Bloomberg dashboard of October 7, S&P 500 forward earnings were at a record and US investment grade OAS was 80 basis points, tighter than in mid 2007. Forward earnings are the profits analysts expect over the next twelve months. Investment grade OAS, the option adjusted spread, is the average extra yield over Treasuries on US dollar corporate bonds rated BBB- or better, and the lower it is, the safer lenders feel. Act 4 is not arriving through rates. It is arriving through the credit that finances artificial intelligence, and the indices will see it last.

The Script, and whose it is

The acts are Andy Constan's. Andy runs Damped Spring Advisors and writes here on Substack as . Before that he worked at Brevan Howard, Bridgewater and Salomon. On July 10, 2023 he published a Damped Spring Report titled "The Script," subtitled "The only way to kill inflation," written as a play in five acts. It went to members first, and he made it public on July 25, 2023. You can read the original report and his post making it public.

Act 1, "Higherer for Longerer Island": "Hikes continue and don't achieve goal."

Act 2, "Long end yields rise to new highs": "Requires a supply catalyst."

Act 3, "Multiple compression": "Higher yields take the legs out of equity rally."

Act 4, "Earnings contraction": "The tightening of Act 2 and Act 3 hit demand."

Act 5, "Recession Island": "as equities sell off, companies fire workers."

Read as a chain, the script runs like this. Rate hikes alone fail to cool the economy. Long term bond yields then rise to new highs, which takes a catalyst such as a wave of new bond supply. Higher yields lower what investors will pay today for future profits, so stock valuations shrink. The tightening then hits demand, and profits fall. Only when share prices fall do companies cut jobs, and recession follows. The name of Act 1 plays on "higher for longer," the idea that rates would stay high for a long time.

The premise is that companies "won't fire workers until their share prices fall. Their share prices won't fall until long end yields rise." The script dates from July 2023 and that cycle. By March 2024 he was preparing to leave Higherer for Longerer Island on a journey that he said "may still end at Recession Island." The act by act reading of October 2026 that follows is mine, not his.

The five acts, one by one

Here is each act, what drives it, and where the data put us in October 2026.

Act 1, the Fed hikes and it does not work. The Fed raises its policy rate to cool demand and bring inflation down. In the script the hikes are not enough on their own, because companies do not fire workers as long as their share prices hold. The Fed held 5.25 to 5.50 percent from July 2023 to September 2024, cut to 3.50 to 3.75 percent on December 10, 2025, and hiked to 3.75 to 4.00 percent on September 16, 2026, by 12 votes to 0, the first hike under Chair Kevin Warsh and the first in more than three years. The fed funds rate is the overnight rate the Fed targets, and it anchors short term borrowing costs. The easing cycle was short and policy is tightening again, so Act 1 is playing a second time.

Act 2, long end yields at new highs. The Fed sets the overnight rate. It does not set the ten year yield. Investors do, and they ask for more to lend for ten or thirty years when the supply of bonds swells and patient buyers step back. That is why Constan wrote that Act 2 "requires a supply catalyst." The ten year closed at 5.31 percent on October 5 on the Treasury's official curve (5.3046 percent on Bloomberg's generic yield), the highest close since May 2002 and above the June 2007 high of 5.26 percent. The thirty year closed at 5.67 percent on October 7. The ten year Treasury yield is the benchmark for mortgage rates and long term corporate borrowing, and a higher yield means a lower bond price. The Fed cut by 175 basis points between September 2024 and December 2025 and has hiked only 25 since, yet Bloomberg's generic ten year rose about 137 basis points, from 3.94 percent on February 27, 2026 to its October 5 high. The long end is not following the Fed. It is following supply, which I take apart further down.

Act 3, multiple compression. When a bond with no default risk pays more, investors pay less for each dollar of future profit, because that profit now competes with a safe return. The S&P 500 forward P/E fell from 22.9 in late 2025 to 19.17 on October 7, 3.7 turns of compression. The forward P/E is the index price divided by the profits expected over the next twelve months. Each turn is one year of those profits: investors paid 22.9 years of them in late 2025 and 19.17 on October 7. Turn the ratio upside down and you get the earnings yield. At 19.17 it is about 5.2 percent, close to what the ten year Treasury pays with no default risk. At 22.9 it was 4.4 percent. The comparison flatters bonds, since profits can grow and a coupon cannot, but it shows how little extra return the index now offers for its risk. The multiple is still above the 16.8 of January 2023, so by that yardstick the compression is not extreme yet.

Act 4, earnings contraction. Tighter money and falling share prices hit demand, revenue slows, and the profit forecasts analysts publish start to fall. This is the act that turns a valuation problem into an earnings problem. It is not in the index data yet. Bloomberg's forward EPS for the S&P 500, the earnings per share analysts expect for the index over the next twelve months, was 385.6 dollars on October 7, a record. In the index data, the first sign would be a run of downward revisions, before any reported decline.

Act 5, recession. In the script, companies fire workers only once their share prices fall. Then unemployment rises and spending falls further. Unemployment rose from 3.4 percent in April 2023 to 4.5 percent in November 2025 and was 4.2 percent in September. So far the labor market has absorbed the shock.

On index data we sit between Acts 2 and 3. The question is whether the rate shock lasts long enough to break earnings, the bridge to Act 4. The four rate shocks before this one show what that takes.

Which cycle? Four rate shocks since 1999

All four pushed long yields up, three compressed the multiple, and only two reached Act 4. For each one, the same questions: what pushed long yields up, what the multiple did, what earnings did, and for the two that broke, what credit said first.

1999 to 2001, the full play. A booming economy and Fed hikes from June 1999 pushed the ten year from about 4.7 percent in January 1999 to a peak of 6.79 percent on January 20, 2000, and the Fed kept going, up to 6.50 percent in May 2000. The forward P/E fell from about 24 to about 15 by 2002. By September 2000, investment grade OAS was already near 165 basis points and high yield OAS, the same spread on bonds rated below investment grade, near 615, while forward EPS was still at its peak. Forward EPS then fell about a fifth, from about 63 dollars in mid 2000 to about 51 in late 2001, roughly 21 months after the yield peak. High Treasury yields alone did not do it. They met a capex bubble that was deflating, then the demand shock of September 2001.

2004 to 2007, the credit break. The Fed hiked 17 times, from 1 percent in June 2004 to 5.25 percent in June 2006, 425 basis points. At first the long end even fell, which Alan Greenspan called a "conundrum" in February 2005, the opposite of today, when the long end rose to new highs after the Fed had cut by 175 basis points. The ten year went from about 3.3 percent in June 2003 to a high of 5.26 percent in June 2007. The forward P/E slipped from about 17 to about 14, but earnings grew fast enough to carry the index higher. Investment grade OAS bottomed near 91 basis points and high yield OAS near 234 in June 2007. Both started to widen that summer, while earnings held through 2007, and then forward EPS fell about 40 percent into 2009. By December 2008 investment grade OAS was above 600. But the leverage of 2007 sat in housing and bank balance sheets. The leverage of this cycle sits in data centers, leases and vendor commitments.

2013, the taper tantrum. In May 2013 Ben Bernanke signaled that the Fed would slow its bond purchases, and the ten year jumped from 1.66 percent on May 1 to 2.98 percent on September 5. The forward P/E did not compress. It rose from about 14 to about 15, and forward EPS kept rising, from about 115 dollars to about 130 through 2014. In September the Fed delayed the taper. The shock was sharp, then it faded. The ten year ended 2013 at 3.04 percent and 2014 at 2.17. Duration matters as much as size: a spike that reverses does not have time to choke demand.

2018, the pivot. 2018 is the comfortable answer. The Fed hiked four times that year. The ten year rose from about 2.5 percent at the start of 2018 to about 3.2 percent in the autumn, the forward P/E lost about four turns into December, from about 18.5 to about 14.5, forward EPS kept rising, and the Fed turned in January 2019, signaling it would stop hiking. The S&P 500 fell 19.8 percent between its close of September 20 and its close of December 24. Forward EPS went from about 157 dollars to about 175 through 2018 and 2019. The pivot was the circuit breaker. That analogue needs a Fed that turns. This one hiked four weeks ago and buys no duration, meaning it does not buy long term bonds in the market.

The pattern. In none of the four did rates alone break earnings. Earnings broke when three things met: yields high for long enough, an excess financed with debt that higher rates exposed, and a Fed that turned too late. The Fed held 6.50 percent until January 2001 and 5.25 percent until September 2007, after credit had cracked, and its cuts did not save earnings. 2013 was too short. 2018 had its pivot, in time. 2007 broke through housing and the banks. 2000 broke through a capex boom financed with debt. In 2000 and in 2007, credit moved before earnings. Today investment grade OAS, at 80, is tighter than in June 2007, and high yield, at 298, is wider than its June 2007 low but less than half its level of September 2000. The index is calm. The single names are not, as the credit section below shows.

2000 is my answer. 2026 already has two of the three conditions, a capex boom increasingly financed with debt and a Fed that hikes instead of turning. The remaining one, time, is the open question, and the next section explains why I think the long end stays high. The engine in 2000 was a capex cycle financed with debt. Carriers borrowed to lay fiber, and depreciation hit income statements two to three years later, just as revenue rolled over. Capex, capital expenditure, is money spent on long lived assets such as fiber, data centers or chips. The cash goes out at once, but the accounting cost, depreciation, is spread over the asset's useful life, so heavy spending today becomes a rising expense for years, whatever revenue does. For AT&T, Verizon, WorldCom, Qwest and Sprint, depreciation and amortization went from 14.0 percent of revenue in 1997 to 20.8 percent around 2001, and stayed above 19 percent for three years. WorldCom, which had booked operating costs as capital spending, filed for bankruptcy in July 2002. Credit broke first, in the most levered names. The index learned last.

The hyperscalers, the companies that build and run the largest data centers, are on the same path. My basket of six also includes Apple and Nvidia. For Amazon, Microsoft, Alphabet, Meta, Apple and Nvidia combined, depreciation was 7.3 percent of revenue in 2025 and consensus takes it to 14.4 percent in 2028. In dollars it goes from 162 billion in 2025 to 499 billion in 2029, a tripling. Energy in 2014 to 2016 is the false analogue. Its ratio went from 7.1 to 15.2 percent because revenue fell 47 percent, from 716 to 382 billion dollars, while depreciation rose 14 percent. The hyperscaler ratio is a numerator story, like telecom: it rises because depreciation grows, not because revenue collapses.

The long end has no buyer of last resort

Act 2 needed a supply catalyst, a reason for more long bonds to come to market than buyers want at current prices. This cycle has four, and none of them is cyclical.

The deficit. Bloomberg's data put the 2026 federal deficit at 1.97 trillion dollars and federal debt at 38.5 trillion at the end of 2025, against 22 trillion in 2018.

Corporate duration. 41 percent of the hyperscalers' 2025 and 2026 dollar issuance has a maturity beyond 15 years, against 16 percent for the ICE BofA US Corporate Index (TwentyFour Asset Management, September 3). The long bond needs the same pension and insurance money, the investors who buy thirty year paper to match obligations that run for decades.

China. Mainland China held 618.0 billion dollars of Treasuries in July 2026, down 77.6 billion, or 11.2 percent, in a year, against a peak of 1,316.7 billion in November 2013 (Treasury TIC data, the monthly report on foreign holdings). Its total reserves, gold included, were 3,789 billion at the end of September, of which foreign exchange 3,400 billion (SAFE, China's foreign exchange regulator).

The Fed. Quantitative tightening, the Fed shrinking its bond holdings by letting them mature without replacing them, ended on December 1, 2025, after more than 2.2 trillion dollars of runoff since June 2022. The New York Fed scheduled no reserve management purchases, the bill purchases it uses to keep bank reserves ample, from August 14 to October 14, and 15.6 billion of reinvestments for September 15 to October 14. Fannie Mae and Freddie Mac were net buyers of only about 50 billion of mortgage bonds through mid year against a 200 billion directive, by Goldman Sachs' estimate, reported by Bloomberg in July. No official buyer is taking duration.

The price shows. The 30 year real yield, the return after inflation on inflation protected Treasuries, reached 3.007 percent on July 31, the highest since 2008. The 10 year SOFR swap spread was minus 39.7 basis points, off a 2025 low of minus 54.3. The swap spread is the 10 year swap rate minus the 10 year Treasury yield. Below zero, Treasuries yield more than swaps, a sign that dealers struggle to carry all the government paper on offer. The MOVE index, the bond market's equivalent of the VIX, measured from Treasury options, rose from 72.0 to 110.5 in the third quarter.

The 30 year mortgage rate hit 7.49 percent, the highest since 2023. American homeowners can refinance whenever rates fall, so mortgage bonds shorten when yields drop and lengthen when yields rise. When yields rise, refinancing stops, the bonds behave like longer debt, and holders who hedge sell Treasuries or their swap equivalent to stay balanced, which pushes yields higher again.

The 20 year. The market now discusses cutting or scrapping the 20 year bond. Citi expects 3 billion dollars less per 20 and 30 year auction, replaced by bills, debt that matures within a year (Bloomberg, October 9 and 12), and Jason Williams, Citi's head of US rates strategy, said Bessent "may point to poor pension-fund Treasury demand, perhaps caused by increased long-end corporate bond issuance," as a reason to reduce or cancel the 20 year. That is the AI channel reaching the auction calendar. Each quarter the Treasury's refunding statement says how much it plans to sell at each maturity. In its August 5 statement, the Treasury said it expects to keep coupon sizes, the amounts sold at each note and bond auction, "for at least the next several quarters," but it now evaluates future "changes" to them, no longer "increases," which Bloomberg reads as opening the door to cuts in 20 and 30 year supply. The 20 year touched 5.75 percent on October 5, its highest since it returned in 2020 (its highest close on Bloomberg's generic yield, 5.709 percent, came on October 7), and was at 5.614 percent at 14:39 today.

Last week's 10 and 30 year auctions stopped at the highest yields since 2000. The September 15 reopening of the 20 year, 13 billion dollars, tailed 2.0 basis points, 5.420 percent against 5.400 when issued, bid to cover 2.57, with dealers taking 16.85 percent of competitive awards against 12.49 in August and indirect bidders down to 52.47 percent from 62.93. A tail means the auction cleared at a higher yield than the market yield just before it, a sign of weak demand. Bid to cover is the amount bid per dollar sold. Dealers are the primary dealers, firms obliged to bid at every auction, so they absorb what others leave. Indirect bidders bid through an intermediary, a dealer or the New York Fed, and include foreign central banks. More for dealers and less for indirects means thinner real demand. The next reopening is on October 21.

BNP Paribas strategists led by Guneet Dhingra wrote in a note reported by Bloomberg on October 6 that "eliminating the 20-year could be perceived as panic, and signal an exhaustion of the Treasury's toolkit, encouraging bond vigilantes," called the Treasury's moves "a 'band-aid on a gunshot wound'," and kept a short in the 30 year with a 5.8 percent target. Bond vigilantes are investors who sell government debt to push back against deficits, and BNP's short is a bet that the 30 year yield rises toward that target.

The yield curve plots yields against maturities, and it is concave at the long end. At 14:43 today the 10s20s30s butterfly, twice the 20 year minus the 10 and the 30 year, was 43.7 basis points, so the 20 year yields 21.85 basis points, half the butterfly, more than the average of its neighbors, down from 61.6 on March 20 and against a one year average of 52.8. The 20s30s spread, the 30 year minus the 20 year, was minus 4.79 basis points, a one year z score of minus 1.99, about two standard deviations below its average of the past year, a rare level. Cutting the 20 year would richen it further, that is, lower its yield as supply shrinks, but the borrowing would move to other tenors, the 10 and 30 year or bills, which BNP calls expensive while the Fed is hiking.

Harley Bassman, the Convexity Maven, warned today that the jump in yields has made the longest bonds the ones delivered into futures, and that a 20 year above 6 percent would be a "nightmare scenario," with selling pressure equal to a 110 billion dollar auction (Bloomberg). Treasury bond futures are settled by delivering whichever eligible bond is cheapest for the seller to deliver. This year's rise in yields cut the prices of the longest bonds the most, so the cheapest to deliver is now the longest eligible bond, and the futures take on its large sensitivity to rates. He puts the negative convexity of bond futures at its highest since 2007. Negative convexity means a position becomes more sensitive to rates as yields rise, so losses accelerate and hedgers must sell more. That is the mortgage loop again, duration that grows as yields rise.

The AI financing chain

Issuance. In 2025, Amazon, Alphabet, Meta, Microsoft and Oracle issued 121 billion dollars of US corporate bonds, against 28 billion a year on average from 2020 to 2024 (BofA Securities, via Reuters). By August 10, 2026, hyperscaler debt issuance had reached about 220 billion dollars for the year, against 12.5 billion in the same period of 2025 (BNP Paribas data, via Reuters). More than 120 billion of data center spending by Oracle, Meta, xAI and CoreWeave sits in off balance sheet vehicles (Financial Times, late 2025). These are separate companies, often co-owned with private capital funds, that borrow to build a campus and lease it to the tech company, so the debt sits outside the tech company's balance sheet.

Off balance sheet. A lease enters the balance sheet only when it starts, when the building is handed over. Until then it is disclosed in the notes to the accounts. Leases signed but not yet commenced, from the latest filings: Microsoft 329.1 billion dollars, Oracle 288, Meta 279.0, Amazon 137.2 and Alphabet 85.2, about 1.12 trillion in total (June 30, August 31 for Oracle). Microsoft's alone, in undiscounted payments, the simple sum of future rents, are 3.7 times the 88.5 billion of operating and finance lease liabilities on its balance sheet. Purchase commitments and other contractual obligations add 811.0 billion at Alphabet, and non-cancelable contractual commitments 349.3 billion at Meta, both reported separately from leases. Moody's, in a July 22 note, counted 1.2 trillion of lease commitments across six hyperscalers including CoreWeave, more than 820 billion not yet started, and treats them as debt-equivalent liabilities, against about 460 billion of direct debt (CNBC). Meta's Hyperion campus in Louisiana was financed by Beignet Investor LLC, a Blue Owl vehicle, with 27.3 billion of senior secured notes, bonds that are paid first and backed by pledged assets, at 6.581 percent due 2049, the largest single bond tranche on record (IFR).

Nvidia is the knot. Its 10-Q, the quarterly report filed with the SEC, for the quarter ended July 26, 2026 lists 366 billion dollars of future commitments: 279 billion of supply and capacity, 29 billion of cloud service agreements, 25 billion of data center leases not commenced, 25 billion of promised equity investments "in AI model makers, infrastructure financiers, and other private companies," and 8 billion of capex. In August, after the quarter closed, it added guarantees capped at 105 billion to support a land, power and shell buildout with SB Energy affiliates "on behalf of a customer, an affiliate of OpenAI," for about 4.25 gigawatts in Pike County, Ohio, a site that will "exclusively host NVIDIA AI infrastructure, subject to limited exceptions." Here the guarantee means Nvidia pays part of the rent and power bills, up to the cap, if the OpenAI affiliate cannot. Nvidia's maximum exposure under guarantees is 108.5 billion, against 33.4 billion of debt on the balance sheet. The loop is explicit. A letter of intent to invest up to 100 billion in OpenAI (September 22, 2025) was replaced by a 30 billion investment, and OpenAI was set to use much of that capital to buy Nvidia chips (Reuters, February 20, 2026). In effect, the supplier is financing demand for its own chips. If that demand disappoints, the same exposure surfaces on several balance sheets at once.

Depreciation against monetization. Consensus capex for the six names, the average of analysts' forecasts, goes from about 370 billion dollars in 2025 to 695 billion in 2026 and 1.109 trillion in 2028, and depreciation from 162 billion in 2025 to 499 billion in 2029. Clear Street estimated AI specific service revenue at about 25 billion dollars in 2025, against 382 billion of capex at the five largest spenders in Clear Street's count. An expert interview (In Practise, a former AWS executive) put one gigawatt at about 30 billion of capex for 6 to 7 billion of annual revenue and 3 to 3.5 billion of annual cash contribution at today's rates, a payback of about nine to ten years, the time it takes for the cash generated to repay the investment. He gets to five to six years only if revenue ramps up and so does utilization, the share of capacity actually sold. Depreciation on what is built is locked in. Revenue is not. That is how earnings break by accounting, before any recession.

China, energy and the consumer hit the same revenue line and calendar. Cheaper Chinese models cap the price of inference, running trained AI models to answer queries. Sightline Climate data cited by Bloomberg on April 1 had close to half of the US data center builds planned for 2026 projected to be delayed or canceled, mostly for lack of electrical equipment such as transformers, switchgear and batteries, a count whose denominator SemiAnalysis disputed on June 18. Oracle cited force majeure, a contract clause that excuses delays caused by events outside a party's control, on Project Jupiter in New Mexico (Bloomberg, September 24), after New Mexico's State Land Office twice refused a route for the gas pipeline meant to supply the site. That is a power delay written into a lease. And spending by the top income households, who own most of the stock market, rides on equity valuations that Act 3 is compressing.

Credit, where Act 4 starts

CoreWeave. The 9 5/8s of 2032 were sold at par in June. Today's quote at 13:22: price 79.82, yield 14.96 percent, G spread 990.3 basis points, Z spread 1,021.4, I spread 1,027.2, modified duration 3.93, DV01 326 dollars per million of face. Modified duration says the bond's value, accrued interest included, falls about 3.93 percent for each percentage point rise in yield, and DV01 is the dollar change per million of face value for a one basis point move. The 1,000 basis point line is crossed in Z and I, for the first time since issue, not yet in G, which hit 993.6 at 13:27 against a low of 519.1 on June 15.

Oracle. S&P cut it to BBB- on July 9, one notch above high yield. BBB- is the lowest investment grade rating. One more notch down would put Oracle in high yield at S&P, a step toward losing investment grade, and many funds are not allowed to hold high yield bonds. The 6.85s of 2066, a 2.75 billion tranche, one maturity of a larger deal, sold on February 2 at 99.81, were at 80.74 at 13:48, a yield of 8.56 percent, G spread 299.7, Z 370.8, I 383.1, modified duration 11.31, DV01 928 dollars per million of face. The G spread has gone from 192.2 on February 3 to 300.1 today, a record for the line. The five year CDS went through its previous record, near 220 to 230 basis points, in late September, rose to 260.9 on October 8 and closed at 251.9 on Monday, against 68.0 a year earlier. ZeroHedge wrote on X today that Oracle unsecureds, the bonds with no specific collateral behind them, trade "like deep junk." At 300 over, that is a phrase, not a measurement.

G spread is measured over the interpolated Treasury curve, I spread over the interpolated swap curve, Z spread over the zero coupon swap curve. The I spread sits above the G spread on both bonds, by 37 basis points for CoreWeave and 83 for Oracle, because swap rates trade below Treasury yields.

Beignet. The 6.581s of 2049 were at 91.44 at 13:51, down 0.78 point, G spread 219.6, Z 263.9, I 273.0, duration 9.29. That is close to the widest since issue, but 10 basis points inside the record of 229.1 set on July 24.

Bloomberg Law noted on July 21 that "nearly 80% of data-center securities sold since early last year are now quoted at a wider spread than at issuance."

The indices lag. Investment grade OAS was 80.01 basis points and high yield 298.49 on October 7, high yield up about 41 from its May low near 257. CDX IG was 58.94 at 13:57 and CDX HY 341.69 at 13:58 today, against a one year range of 288.4 to 406.4 for CDX HY. CDX IG and CDX HY are traded indices of credit default swaps on 125 investment grade and 100 high yield North American companies, the quickest gauge of credit risk.

Memory. At a conference in Shanghai on October 9 and 10, Cao Kanyu, the president of CXMT, announced progress on a 4F² DRAM cell, two thirds of the area of today's 6F² designs, and a new generation of DDR5 server modules by the end of 2026. F is the smallest feature a chip process can print, so at the same F a 4F² cell packs more memory into each wafer, which can lower the cost per bit. CXMT was already the world's fourth largest DRAM maker in the second quarter according to Counterpoint, listed on July 27 with a first day gain of 466 percent and 8.6 billion dollars raised, and Washington held off on adding it to the Entity List, the US export blacklist, to protect the trade truce (Export Compliance Daily, June 18). Its prospectus had no dedicated HBM project, the high bandwidth memory stacked next to AI processors, and a Morningstar analyst told Bloomberg to expect a meaningful time gap before commercial yields, here the share of usable chips per wafer. But memory pricing power is part of the margin the AI chain is built on. TrendForce wrote in June that conventional DRAM prices had risen sharply since the second half of 2025 in a tight market, strengthening suppliers' pricing power in HBM. A fourth supplier with a denser cell attacks exactly that.

Verdict, and the levels

My verdict: early Act 4, through credit. The indices will see it last.

The rates channel did its work in Acts 2 and 3. Act 4 is arriving where leverage is highest and the accounting is most fragile, in the bonds of the companies that borrowed to build AI and in the suppliers whose margins depend on it. That is the 2000 pattern, not 2018 and not 2007. This is not a crash call.

The levels I watch:

CoreWeave 9 5/8 2032, a G spread close above 1,000 basis points (993.6 today).

Oracle 6.85 2066 beyond 300 basis points of G spread, five year CDS above the 260.9 of October 8, and any move from BBB- to high yield at S&P.

Beignet 2049 above 229.1, its record.

CDX HY above 406.4, the one year high, and cash high yield OAS through 350, then 400.

The ten year above 5.31 percent, the thirty year above 5.67, the 20 year at 6 percent, Bassman's line, the October 21 reopening of the 20 year and the November 4 refunding.

Friday's TIC data for China and the New York Fed's purchase schedule from October 15.

For the indices, the first monthly decline in forward EPS and unemployment above the 4.5 percent of November 2025.

Andy's play ends with companies firing workers. In this cycle, the first scene of Act 4 is already playing in the bond market.

Key terms, in plain English

Basis point. One hundredth of a percentage point. 100 basis points make 1 percentage point.

Yield. The annual return a buyer earns by holding a bond to maturity at its current price, if every coupon is paid and the principal is repaid. Prices and yields move in opposite directions.

Spread. The extra yield a bond pays over a benchmark. The G spread is measured over US Treasuries, the I spread over interest rate swaps, the Z spread over the whole zero coupon swap curve. Wider spreads mean lenders see more risk.

OAS. The option adjusted spread, the spread of a bond or an index after removing the value of the call options some bonds carry. The standard gauge for investment grade and high yield indices.

Investment grade and high yield. Ratings of BBB- or better are investment grade. Anything lower is high yield, also called junk.

CDS. A credit default swap, a contract that works like insurance against a company's default, quoted as an annual cost in basis points of the amount insured.

CDX IG and CDX HY. Traded indices of credit default swaps on 125 investment grade and 100 high yield North American companies.

Duration and DV01. Modified duration measures how much a bond's value moves when its yield moves, in percent per percentage point. DV01 is the dollar change for a one basis point move, here per million of face value.

Convexity. How duration itself changes as yields move. Negative convexity means a position becomes more sensitive to rates as yields rise, so losses accelerate.

Butterfly. A measure that compares a middle maturity with the two around it. Here, twice the 20 year yield minus the 10 and the 30 year.

Swap spread. The swap rate minus the Treasury yield of the same maturity. Negative means Treasuries yield more than swaps.

Real yield. The yield after inflation, read on inflation protected Treasuries.

MOVE index. The implied volatility of Treasury options, the bond market's VIX.

Quantitative tightening. The Fed shrinking its balance sheet by letting bonds mature without buying new ones.

Refunding. The Treasury's quarterly announcement of its borrowing plans, including how much it will sell at each maturity.

Auction tail and bid to cover. A tail is the amount by which an auction clears above the market yield just before it. Bid to cover is total bids divided by the amount sold.

Forward EPS and forward P/E. Forward EPS is the profit per share analysts expect over the next twelve months. The forward P/E is the price divided by it.

Capex and depreciation. Capex is spending on long lived assets. Depreciation spreads that cost over the asset's useful life in the income statement.

Leases not yet commenced. Leases signed but not started. They are disclosed in the notes to the accounts, outside the balance sheet, until the asset is delivered.

Off balance sheet vehicle. A separate company that borrows to build an asset and leases it to the user, so the debt does not show on the user's balance sheet.

TIC data. The Treasury's monthly report on foreign holdings of US securities.

Hyperscalers. The companies that run the largest cloud and AI data centers, such as Amazon, Microsoft, Alphabet, Meta and Oracle.

This article is for educational purposes only. It is not investment advice, a solicitation, or an offer to buy or sell any security or derivative. Levels are as of the dates and times stated and can change quickly. Credit, equities and derivatives carry a high risk of loss. The five act framework belongs to Andy Constan (The Damped Spring Report, "The Script," July 10, 2023). The act by act reading of October 2026 is the author's own and does not represent his views.

$CRWV $ORCL $NVDA $META $MU $TLT

Data: Bloomberg.

Educational case study. This is not investment advice and not a recommendation to trade. Prices are October 5, 2026 settlements, the book is illustrative, and options and futures can lose far more than you expect.

Brent settled at 100.32 on Monday for the December contract. The thirty year Treasury yields 5.66 percent, after touching 5.70 percent on Monday, a level last seen in 2002. A gallon of regular gasoline costs 4.37 dollars, up almost 40 percent from a year ago. The midterms are four weeks away, on November 3.

The war with Iran has been running since February 28. The spring ceasefire collapsed on July 8, and since then Washington has kept to limited strikes while a naval blockade squeezes Iranian exports. The President has also been unusually explicit about timing. Asked whether the bombing could ramp up after the midterms, he told TIME, in an interview published on October 1, that it was possible. At a rally in Mobile, Alabama, the next day he said the war would probably end right after the midterms and that oil prices would come tumbling down.

My view is simple to state. I do not think Trump resumes large scale strikes on Iran before November 3. I think he does after the vote, whether Republicans hold Congress or lose it. This piece is about how to express that calendar with options, and about the two parts of the structure that do most of the work, a put ratio before the vote and a tail hedge underneath it.

Why the vote sets the oil calendar

Gasoline is the most visible price in American politics, and this White House is already spending its ammunition to hold it down. The 172 million barrel release from the Strategic Petroleum Reserve agreed in March ends with a final offer of up to 40 million barrels, made on September 29, and the reserve already sits at its lowest level since 1982. The pattern is not new. In 2022 the last 15 million barrels of that year's release were announced on October 19, three weeks before the midterms. A president four weeks from an election, with a 4 dollar gallon, does not choose to add a Gulf escalation on top of it.

After the vote the logic flips, and it flips whichever way the vote goes. If Republicans hold Congress, the electoral cost of escalation disappears for two years. If they lose, and Democrats lead the generic ballot by eight to nine points in the main averages, the clock that matters becomes the lame duck. The House already passed an Iran war powers resolution 220 to 204 on September 15, with seven Republicans in favor, and a new Congress takes office on January 3. That is a reason to act sooner, not later. The buildup points the same way. The carrier Theodore Roosevelt left San Diego on September 27, an amphibious group with about 2,000 Marines sailed the next day, and US officials put the extra forces at nine to ten thousand, due in the region by late November.

The physical market explains why I expect a drift lower before the vote rather than a collapse. Crude flows through Hormuz are back near prewar levels, but product flows are still a fraction of normal, China has halted October fuel exports, and global inventories drew more than 500 million barrels between February and August. OPEC+ held its November targets on October 4 and meets again on November 1. Before the vote the ceiling on Brent is political and the floor is physical. After the vote the distribution turns binary, an escalation that spikes oil or the deal the President keeps promising, which would send it lower.

What the options market already prices

The October 27 expiry is the last monthly Brent option before the vote. At the money vol is 47.3. The 25 delta put trades at 50.1 against 48.5 for the 25 delta call, and the 10 delta put at 54.1 against 52.5 for the 10 delta call.

So the risk reversal is only about 1.6 vols in delta terms. Both wings are bid over at the money, the puts by about 7 vols and the calls by about 5. Measured by strike the tilt is larger, a put 15 to 20 percent out of the money sits 3.5 to 4 vols above a call the same distance away. The market is pricing two tails at once. Since mid July, traders have bought large volumes of cheap downside put spreads as protection against an abrupt end to the war, and that demand is part of what keeps the puts bid.

The November 25 expiry, which covers the January contract and the weeks after the vote, is more balanced still. There the 110 call trades at 51.1 vol and the 85 put at 50.4. After the vote the market already pays as much for upside as for downside.

The ratio logic

Before the vote I want to own a drift lower, not a crash. That is exactly what a put ratio is built for. The structure is a 3 by 5 on the October 27 expiry. Buy three puts at 91, sell five puts at 77.

Each strike has a job. The 91 put settled at 1.47 dollars a barrel with about a 20 delta. It is the exposure. It starts paying as soon as Brent slips under 91, which is where SPR barrels, steadier Gulf flows and jawboning from the White House could take it before November 3. The five 77 puts settled at 0.16 with a 3 delta. They are the statement that a collapse before the vote is unlikely while the physical floor holds, and they shape the payoff so that it peaks exactly at 77.

The arithmetic deserves a line of its own. Three times 1.47 is 4.41. Five times 0.16 is 0.80. Each 3 by 5 unit costs 3.61 dollars a barrel. Selling more options than you buy does not make a structure a credit, and a 3 delta option pays very little in dollars even when its vol looks rich. The true 10 delta put is the 85 strike at 0.63, and a 1 by 2 in 91 and 85 would cost only about 0.21. I still prefer 77 for the short strike, and the reason is the thesis itself. A short strike at 85 sits inside the range I expect oil to visit before the vote. A short strike at 77 sits below it.

Sized at 100 lots on the short leg, about 10 million dollars of Brent, the ratio is long 60 puts at 91 and short 100 puts at 77, for 72 thousand dollars. At expiry it is worth nothing above 91, 360 thousand at 85, 660 thousand at 80, and it peaks at 840 thousand at 77. Below 77 it gives money back at 40 thousand dollars per dollar of Brent, because 40 more puts are sold than bought, and below about 57.80 the plain ratio starts to lose, with no limit.

The tail hedge logic

Every ratio has a weak spot, the part you are net short. Here it is 40 puts below 77. Before the vote the event that reaches that zone is not a slow recession, it is the war ending early, the same deal the President promises for after the election, only sooner. That is the gap a tail hedge exists to close.

Buying 40 puts at 65 for 0.02 costs 800 dollars, about one percent of the ratio's premium. It does three things. It limits the give back to the 12 dollars between 77 and 65. It turns the floor into a gain, because below 65 the payoff locks at 360 thousand however far Brent falls. And it removes the open ended short, which means the structure can be held through headlines and margin calls rather than cut at the worst moment.

Why 65 rather than 70 or 60? A tail hedge is chosen for its strike, not its delta. At 65 it is close to free and the worst give back is bounded at 12 dollars. Further down the protection is cheaper but the unprotected zone gets wider. Closer in it starts to cost real money and eats the payoff the ratio was built for. With the tail in place, the full put structure costs 73 thousand dollars, is never worth less than zero at expiry, and makes money net of premium anywhere below about 89.80.

The bond leg, and an exact DV01

The macro leg is short 10 million face of the on the run thirty year, the 5.125 percent of August 2056, which closed at 92.273, or 92 and 8¾ thirty seconds, to yield 5.664 percent. A post vote escalation means a second oil shock on top of a 4 dollar gallon, with the long end already near 5.7 percent. I want to be short duration into that.

Every hedge ratio downstream is built on this leg's DV01, so it has to be exact. DV01 is the change in the bond's full price, accrued interest included, for a one basis point move in yield. The textbook form is modified duration times full price times 0.0001. For an October 6 settlement, accrued interest for 52 of the 184 days in the coupon period is 0.724, so the full price is 92.998. Modified duration is 14.499. The product gives 0.1348 points per 100 of face for one basis point, which is 1,348 dollars per million of face and about 13,480 dollars per basis point for the 10 million short. Repricing every cash flow with the yield bumped half a basis point either side gives the same 1,348 dollars per million. A risk field quoted around 13.48 is therefore dollars per basis point per 10,000 of face. To get a million you multiply by 100, not by 10,000.

For large moves DV01 is not enough, so every scenario below reprices the bond in full. A 100 basis point rise earns about 1.21 million. A 100 basis point fall costs about 1.51 million, and the gap is convexity working against a short. Carry is a small cost, a 5.125 percent coupon paid against about 3.92 percent of repo earned, roughly 20 thousand dollars into November 25, which the scenarios below leave out.

Hedging the scenario that hurts both legs

The scenario that hurts the whole book is the one the President describes, a deal, oil tumbling, yields following it down. Correlation cannot hedge it. Brent and the thirty year yield have moved with a 0.63 correlation over three months of daily data but only 0.35 over a year, which explains about 12 percent of the variance. The hedge has to be direct, calls on Treasury bond futures.

Two choices make it work. The expiry is the December options, which expire on November 20, after the vote and inside the window where a deal would land. And the size comes from DV01, not notional. A bond future moves like its cheapest to deliver bond divided by its conversion factor. The rule of thumb says that with yields below the 6 percent notional coupon the lowest duration bond in the basket is cheapest, here a 2042 issue. The delivery screen says otherwise. The 2044 to 2046 bonds trade about 5 basis points cheaper than the 2043 issues, and the 2.5 percent of February 2045 ranks first, with a conversion factor of 0.6179.

With a duration near 13 years rather than 11, that bond makes the future move about 137 dollars per basis point per contract. So 98 contracts at delta one carry about 13,400 dollars per basis point, the same as the short. Sized on the shorter bond, the hedge would have been about 20 percent too large.

The 104 calls settled at 1 and 16/64, or 1.25 points, 1,250 dollars a contract, on 13.9 vol and a 37 delta, so 98 of them cost 122.5 thousand. On day one they offset about 37 percent of the short's DV01, and the book keeps most of its short duration. If yields collapse and the future runs through 104, the calls take over close to one for one. A short bond plus calls on matching DV01 is a synthetic put, a short with a floor that starts once the future trades through 104. The surface helps here too, because on bond options the downside carries the premium, with out of the money puts on the December expiry about 1 to 1.5 vols over equidistant calls. One refinement for anyone running the true thirty year at size. The classic contract tracks bonds of 15 to 25 years, while the Ultra Bond, whose cheapest to deliver is currently the 2.25 percent of February 2052 and which moves about 170 dollars per basis point per contract, matches a thirty year short much more closely.

After the vote, the calls

The post vote leg is 100 calls at 105 on the January contract, expiring November 25, three weeks after the vote. They settled at 4.37 dollars a barrel, 437 thousand dollars, the most expensive line in the book, which is fair, because this is the bet. Two refinements are worth weighing. A 105 and 113 call spread costs 1.73 instead of 4.37 but caps the payoff, which matters if an escalation closes the Strait again. And November 25 may be early if the window that counts is the lame duck before January 3, in which case an expiry in late December fits better.

Every leg at its own expiry

Every price in the book is the October 5 settlement, as Bloomberg's historical price screens show.

Together the book pays 632.5 thousand dollars of premium. Priced with every leg at its own expiry, and before the 20 thousand of carry, the grid has the shape of the thesis.

Yields up and oil up is where it earns, about 1.79 million with Brent up 20 percent and the long end up 100 basis points. With Brent flat it earns 579 thousand if yields rise 100 basis points and loses 862 thousand if they fall 100, the worst cell on this grid, because the bond calls put a floor under the short. The floor is not perfect. In a 200 basis point rally the loss grows to about 1.04 million, because the thirty year gains convexity faster than the bond the future tracks. A 20 percent Brent selloff with flat yields is roughly flat, the ratio pays for the calls.

In the order the legs expire

The calendar matters more than the grid. The Brent puts settle on October 27, the bond calls on November 20, the Brent calls on November 25, and the short is marked there.

If the thesis plays out, with Brent down 10 percent by October 27, then 20 percent above today's level by November 25 and the thirty year up 50 basis points, the book makes about 1.26 million. Quiet into the vote and the same spike after makes about 1.22 million. An escalation before the vote lands in the same place at expiry, so being early on timing is not fatal. A Gulf strike that sends Brent up 25 percent while bonds rally on fear still makes about 870 thousand, and an oil selloff with yields higher makes about 650 thousand.

If the President gets his deal instead, and Brent falls 25 percent after the vote with yields down 50 basis points, the book loses about 824 thousand. That is the price of being wrong, and it is bounded, because the Brent calls can only lose their premium and the bond calls absorb more than half of the bond loss. The put structure, tail included, has already expired by then. The risk off cases before the vote cost between 70 thousand and 494 thousand. If nothing moves at all, the book loses its premium, 632.5 thousand, plus about 20 thousand of carry.

Five checks before trusting any scenario grid

Check every DV01 against modified duration times full price, and read the cheapest to deliver from the delivery screen rather than from a rule of thumb. Check put call parity on the settlements. At the 100 strike the call minus the put should equal the future minus the strike, and here it does to the cent at 0.32. Read deltas from the chain rather than from the label a strike was given. Recompute the net premium from settlements to see whether a structure is a credit or a debit. And price each leg at its own date, or say clearly that you have not.

The political clock does not make a trade by itself. It tells you where the risk sits in time. Before November 3 the risk is a drift lower and an early deal, which the ratio and its tail hedge are built for. After November 3 it is the escalation I expect, or the deal I do not, and the book is shaped so that the first pays and the second is survivable.

Update, October 17, 2026: the follow-up on this week in energy, Oil Above $100: What Futures Curves, Crack Spreads and Gas Spreads Say.

This article is for educational purposes only. It is not investment advice, a solicitation, or an offer to buy or sell any security or derivative. The positions are hypothetical, priced from October 5, 2026 settlements, and ignore transaction costs, margin, liquidity and taxes. Options and futures carry a high risk of loss, and short option positions can lose more than the premium received.

$BNO $USO $TLT

Data: Bloomberg, CME.

The VIX closed the week at 15.31. On its own that number says nothing is wrong. Equity vol is low, range bound, and as calm as it has been all autumn. If the VIX were the only instrument on the desk you would conclude the market is comfortable with multi decade high yields, a fresh war front in the Middle East, and 119bn dollars of Treasury supply landing in five sessions. It is not comfortable. It is just looking in the wrong place.

The signal this week is not in equity vol. It is everywhere else. The MOVE index sits at 107, up almost 47 percent in a single month, and VXTLT confirms it with a 49 percent jump of its own, so this is not a MOVE anomaly, it is the whole long end repricing. VSTOXX and VDAX are both up around 20 to 23 percent, spiking to 20.7 and 20.1 on the first of October. One month euro and sterling implied vol surged from the 22nd of September, 28 percent on EURUSD and 18 percent on GBPUSD. Only one major vol surface on the board did not move, and it is the one everyone quotes. Volatility is starting to talk. Most people are listening to the single index that has nothing to say.

The ratio flags it, the composite confirms it

The quick way to see the divergence is the MOVE over VIX ratio. It went from 5.0 to 7.0 in a month, a 39 percent widening, and 7.0 is a cycle high. That tells you bonds are stressed relative to stocks. It is a real signal, but it is a bilateral one. It cannot tell you whether this is a rates only event or something broader, because it only looks at two surfaces.

So we track a cross asset vol composite instead. It is an equal weighted average of the z scores of eight vol series, VIX, MOVE, VSTOXX, VDAX, VXTLT, OVX, and one month EURUSD and GBPUSD implied vol, each normalized to its own trailing mean and standard deviation. A reading of zero means everything sits at its own average. Positive means broad based stress.

The path through September is the whole story. The composite started the month at minus 1.30 on the 4th, every vol suppressed at once, the false calm. It poked positive to plus 0.70 on the 15th when oil and European equity vol jumped on the energy shock. Then it fell back to minus 0.88 on the 22nd as equity and FX vol briefly compressed. That is the important moment. The MOVE over VIX ratio alone would have kept flashing stress through that dip, and it would have been wrong, because the stress was not yet broad. The composite got it right by staying negative.

From the 23rd it turned. By the 1st of October the composite hit plus 1.48, the high of the month, as MOVE, VXTLT, VSTOXX, and euro and sterling vol all spiked together. That simultaneity is the signal the ratio cannot give you. It is the kind of co movement that historically shows up before forced de risking and liquidity events, not just a rates repricing. The slight pullback to plus 0.78 on the 2nd is partial mean reversion in FX and equity vol, not an all clear. We enter the week with the composite still well above average and the ratio at a cycle high. That is a fragile base, not a calm one.

Why rates vol is screaming

The driver is not a mystery. Ten year yields are at 5.30 percent and the thirty year is at 5.65 percent, the highest since 2007. Into that you drop the supply gauntlet. The Treasury brings 58bn in threes on Tuesday, 39bn in tens on Wednesday, and 22bn in thirties on Thursday. 119bn in a single week, at the cheapest levels in nearly two decades, with sentiment on government bonds already fragile and swap spreads at risk of widening.

This is where last week ties into this one. The soft September payroll let the market price out the urgency of follow up hikes, and that was supposed to be the relief. It was not, because the back end does not trade on the next labor print. It trades on supply and term premium, and both are pushing the wrong way. We argued through the week that the durable buyer of this duration is the liability matcher, the pension running LDI that buys because yields are high rather than in spite of it, with the Milliman 100 funded ratio north of 114 percent. This week is the live test of that thesis. The 119bn of supply either gets absorbed by real money into the auctions, which caps the move, or it does not, which lets term premium win and keeps the bid for protection alive. Watch the ten year and thirty year auction tails on Wednesday and Thursday. They matter more than the minutes.

Europe is pricing it louder

VSTOXX and VDAX up 20 plus percent with spikes near 21 is not a US story imported. Europe has its own compounding risk and it is pricing it honestly, because the backdrop is genuinely stagflationary. Growth is running near 0.9 percent this year, down from 1.4, with inflation still around 3 and the region a whisker from a technical recession in the second quarter. Energy inflation at 9 percent year on year is the tax underneath all of it, with Brent above 100 and the oil to equity correlation deeply negative since the US Iran conflict began, so higher crude is a direct headwind for European equities rather than a rotation. Into that the ECB hiked again in September and is shrinking its balance sheet, adding sovereign supply to a market that has to absorb it without the old spread compression tailwind. October odds have been trimmed to around 20 percent, but a December hike is still live.

The credit tape makes the stress concrete, and the timing is the tell. The iTraxx Main widened 26 percent in a month and Crossover 16 percent, and both peaked on the first of October, the exact day the cross asset composite hit its plus 1.48 high. That synchronicity is the point. Credit did not wobble on its own idiosyncratic story, it moved in lockstep with the whole vol complex. Euro AT1 spreads widened 12 basis points last week with yields at 6.25 percent, the highest since March. The FX channel is live too, with EURUSD down 3.1 percent on the month to its weakest since May 2025 and one month euro vol up 28 percent, peaking at 6.96 on the first, while EURCHF at 0.9326 edges toward the level where the SNB starts to matter.

France is the epicenter

The single most acute idiosyncratic risk in Europe is French government debt. The OAT Bund spread has gone from 86 basis points on the fourth of September to 141 on the fourth of October, 65 basis points of widening in a month, back toward euro crisis peaks. It is not a rates move dressed up as a spread. The French ten year rose 67 basis points on the month against Germany's 12, so this is France specific. You can see it in the plumbing. French banks and large corporates are now issuing below their own sovereign, which is what acute stress looks like, the CDS redenomination spread is back at 2017 levels, and liquidity has thinned. The ratings are already AA3 and A plus with a negative Moody's outlook, the 2027 budget and election are the overhang, and QT has left France leaning more on foreign buyers, which makes the yield more sentiment driven, not less. The hierarchy has even inverted, with the Italian BTP Bund spread at 114 now sitting inside the French 141, something that would have read as a typo a year ago.

Underneath, the banks are the forward worry rather than the current one. The average euro area NPL ratio is still a benign 1.8 percent, but the guidance is unanimously deteriorating, French NPLs are rising on soft consumption and construction refinancing, and lending standards are tightening at a historically aggressive pace. Put the five fronts together, stagflation, France, credit repricing, a banking book that is turning on a lag, and a weakening euro that enlarges the energy bill and narrows the ECB's room to pause, and the one hedge that works across all of them is European energy, the only sector where this scenario is a tailwind.

The crack in credit

The number that deserves more attention than it is getting sits at the bottom of the quality stack. The CCC segment of US high yield is trading at distressed levels for the first time since the 2023 banking crisis. Distress at the junkiest end does not stay contained by itself. It is where the next credit hammer tends to fall first, and it is doing this while equity vol prices nothing. A VIX at 15 and CCC at distressed is not a stable pair.

Positioning is the accelerant

The dangerous part is not the catalysts. It is who is holding what into them. Discretionary investors are at the bottom of their long run sentiment range, already defensive. Systematic strategies are doing the opposite, adding equity exposure precisely because realized vol has been low. That bifurcation is the fragility. Low vol pulls the systematic community long, and a vol spike forces that same community to sell into a market the discretionary side has already stepped back from. The fuel and the match are in the same room.

Underneath, momentum has posted three straight weeks of gains and S and P dispersion is grinding lower. In rates, SOFR options price two live paths, one more hike then a long hold, or a drift back toward 5.125 with little easing after, and an October hike is now more likely than not, with the September CPI on the 14th the input that settles it. In FX the defensive posture is already on, long dollar, deeper short euro and sterling, the yen flipping hard from short to long on normalization and intervention risk, and EM carry longs starting to moderate as higher for longer raises the cost. In commodities, WTI open interest is around 276 thousand contracts, gold cannot hold gains even after softer core PCE because the strong dollar and high real yields pin it, and metals are the only major group at a new high this year, doing it straight into a ten year above 5 percent, with copper carrying the same pump then dump risk that already played out in gold, silver, platinum and iron ore.

Where the spike lands

If the composite pushes back toward plus 1.5, the damage is not evenly spread, and the sort is almost entirely about duration and beta. Rank the S and P sectors by exposure to a further composite spike and Technology sits at the top by a wide margin, with the highest beta at 1.50, the second longest equity duration at 22.5 years, and a dividend yield of 0.42 percent, a 488 basis point deficit to the ten year. That is the long duration growth trap in one line. Every leg higher in real yields compresses the present value of its cash flows more than anything else in the index, which is why semis led the drawdown on a vol adjusted basis the last time the ten year pressed 5 percent in 2023. Consumer Discretionary is next, high beta at 1.21 and directly in the path of the oil to consumer squeeze with Brent above 100. Communication Services follows, carrying the widest credit spreads in the index, up 25 basis points this year, so the bond market is already writing stress into its capital structure. Industrials round out the high exposure group on duration and energy pass through.

The sleeper is Real Estate. It has the longest duration in the index at 23.7 years and is already down 7 percent in the third quarter with spreads widening, so it does not need much to extend the move. The other side of the ledger is Energy, and it is the one place a composite spike is a tailwind rather than a threat. Its beta to the S and P is negative at minus 0.36, so a composite driven by rates and geopolitics tends to push oil and the energy sector higher while it drags everything else down. The defensives behave as you would expect, Staples near zero beta, Health Care the cleanest hedge on growth weakness, and Financials shielded by the lowest implied vol and the shortest duration in the index, with net interest margin offsetting part of the rate drag.

What matters this week

Three catalysts sit above the rest. The 119bn of Treasury supply is the most mechanical, and the auctions will tell you whether the duration bid is real. The Fed and ECB minutes on Wednesday and Thursday are the second, parsed for any hawkish residue now that hike urgency has faded. The fresh Middle East escalation is the third, with Yemen launching a full operation against Houthi areas on the 4th and OPEC plus holding November output unchanged, keeping supply tight into the risk. Around those, ISM Services on Monday, China back from Golden Week on Thursday with USD CNH as the pressure valve, and the University of Michigan five year inflation expectation on Friday.

Put it together and the composite does the summarizing for us. It enters the week at plus 0.78 with the MOVE over VIX ratio at a cycle high of 7.0, which is a market already stretched before a single catalyst lands. An auction tail, a hawkish line in the minutes, or another leg in the Middle East pushes the composite back toward plus 1.5 and beyond, and that is the zone where stress stops being a rates story and becomes a cross asset one. The VIX will be the last to admit it. When it finally does speak, it will not be adding new information. It will be catching up to what the rest of the vol complex has been saying all month.

This article is for educational purposes only. It is not investment advice, a solicitation, or an offer to buy or sell any security or derivative. Options, futures and leveraged products carry a high risk of loss.

$TLT $XLK $XLY $XLRE $XLE $SPY

Data: Bloomberg.

The sell-side has two ways of talking about AI infrastructure, and both of them hide the same fact.

The first is capex in absolute terms. Six hundred billion here, seven hundred billion there, a number large enough to end the conversation. The second is inference output in narrative terms. The price of intelligence is collapsing, tokens are getting cheaper every quarter, so the unit economics will take care of themselves. One framing is too big to argue with. The other is too smooth to question. Neither one asks what a token actually costs to produce.

A token is a manufactured good. Its marginal cost is dominated by one input, and that input is electricity. Once you price inference off the electricity floor instead of off a narrative, the floor does not move the way the consensus assumes. Models that ignore it are not off by a rounding error. They are off by an order of magnitude the market has not yet been forced to confront.

That is the argument of "Tokens, Watts, and Geography: A Four-Tier Framework for Pricing the AI Inference Trade, 2026-2030". The framework sorts the world not by GDP and not by chip access, but by the thing that actually sets the floor: the price and availability of power at the interconnect.

Tier 1, Sovereign Capable. China, the Gulf, the Nordics, France. Power is cheap, dispatchable, and politically backed. These are the places that can run inference at the floor and export the surplus.

Tier 2, Emerging Stretched. India, South Korea, Japan. Demand is real, the grid is not ready, and the gap is the story.

Tier 3, Power Constrained Wealthy. The United States on PJM, Germany, the United Kingdom. Capital is abundant and electrons are not. This is the tier that is pricing inference as if the floor were someone else's problem.

Tier 4, Arbitrage Suppliers. Russian Siberia, Canadian hydro, Brazilian Norte and Nordeste, Australian Pilbara. Stranded power looking for a load. The release valve, and the swing factor nobody is modeling.

The reason this paper matters is not that it has a view on AI. Everyone has a view on AI. It matters because it supplies the missing floor in a larger piece of work.

"The AI Infrastructure Repricing" argues that the capex cycle is being financed in ways that look a lot like past credit cycles. "Convergent Faults" maps how private credit has synchronized that risk across the system. "The China AI Disruption Thesis" takes the geography to its conclusion. Each of those rests on an assumption about what inference costs, and until now that assumption was borrowed from the same narrative this paper takes apart. "Tokens, Watts, and Geography" is where the floor gets built. The rest of the program stands on it.

If you only read one thing from the research program this quarter, read this one first, because the others assume it.

Read the paper: https://djellaldjouad.com/papers/tokens-watts-and-geography/

$NVDA $VST $CEG $TLN $NRG $MSFT $GOOGL $AMZN $TLT $SPY

Further reading

• The AI Infrastructure Repricing: https://djellaldjouad.com/papers/the-ai-infrastructure-repricing/

• Convergent Faults: https://crossvolresearch.com/

• Working papers and books: https://crossvolresearch.com/

Data as of Friday, September 25, 2026.

I smell smoke.

On Thursday, September 24, Oracle sent a force majeure notice to the developer of Project Jupiter, the AI data center campus it is leasing in New Mexico. Oracle is not walking away. It is protecting its right to defer rent if the campus misses its planned 2028 opening. By Friday, Oracle's five year credit default swap was at 237 basis points, 51 more than two weeks earlier. The roughly $18 billion of loans on the campus had already been quoted at 89 to 91 cents on the dollar on September 18, six days before the notice. ARM, the collateral behind SoftBank's $25 billion margin loan, had lost 6.7% in two sessions. Japan's Financial Services Agency said it was stepping up scrutiny of how the country's biggest banks and life insurers finance AI data centers, with a focus on US projects. And the US Treasury Secretary spent the same Friday telling the world that the yen should be stronger.

Each of those facts has a story of its own. Put them next to each other and you get a fuse. Japan is the largest foreign holder of US Treasuries, a lender to US data centers through private credit, and SoftBank's home market. Japanese and Taiwanese insurers hedge only 41% of their foreign currency exposure. Washington wants the yen higher. Oracle wants more time. The quarter ends on Wednesday.

I have been following this fuse since the spring. In May, the CrossVol Research book I co-authored called Oracle "the acute case, and the one to watch as a leading indicator". In June, my paper on private credit mapped a convergence window that "opens in Q3 2026". In August I wrote here that "this time the fuse runs through AI capex financed by debt". The receipts, with dates and links, are further down.

This note started as a simple trade on the Treasury curve. Every time I asked what could go wrong, the answer led to the same place. I am publishing the whole chain, with the numbers I pulled from Bloomberg this week and corrected where my first pass was wrong, because the people who most need to read it are marking their books this weekend.

Two warnings up front. To risk managers: the party is already over, your marks just have not heard it yet. To investors: the last chart in this note shows the S&P 500 earning 1.39 points less than a 10 year Treasury. The party is over there too.

It Started With a Belly Trade

A tweet this week made the case for the belly of the Treasury curve. Real yields near 2.7% on the 5 and 7 year. A position that makes 10% if yields fall 1% and loses nothing if they rise 1%. Payoff ratios of 1 to 7.5 on the 10 year and 1 to 3 on the 20 year. I answered it with the math, starting with the inputs.

The inputs were close in the belly and wrong at the front. Against Bloomberg's curve for September 24, the curve I first worked from understated the 2 year by 35 basis points and the 1 year by 13, which matters for any carry or roll analysis anchored at the front end. From 3 years to 30 years, it was within 7 basis points.

The real yield holds. With the 7 year at 5.08% and a 7 year breakeven of about 2.36% to 2.37%, interpolated between the 5 year (about 2.40%, which Bloomberg's generic series could not confirm on the day) and the 10 year (2.33%), the real 7 year yield is about 2.71% to 2.72%. The real 5 year is about 2.60%. On my first pass I had 2.66% and 2.59%. Either way, 2.7% real in the belly is right, and you can check it against TIPS.

The payoff mostly holds too. On my first pass curve, a 7 year bought at 5.03% has a modified duration of about 5.8. Carry plus about 3 basis points of rolldown gives 5.06% a year. If yields rise 100 basis points over the year, the P&L is 5.06 plus 0.2 of convexity minus 5.79, about -0.5%. If they fall 100 basis points, it is about +11.1%. So "make 10 / lose nothing" is structurally fair, as long as you read "lose nothing" as "lose 0.5%". With those inputs, the one year breakeven on the 7 year is about 90 basis points, which is the "±1%" of the pitch.

The 20 year ratio is right. The 10 year ratio is too generous. The 10 year has a duration of 7.7 and carries 5.12%. Up 100 basis points, it loses 2.2%. Down 100, it makes 13.2%. That is 1 to 5.9 on unrounded numbers, not 1 to 7.5. To get to 7.5 you would need a duration of 7.2 or a carry of about 5.5%, so the pitch was about 25% too optimistic. The 20 year has a duration of 12.5 and carries 5.55%, and its convexity is worth about a full point. Up 100, it loses 5.9%. Down 100, it makes 19.0%. That is 1 to 3.2, so 1 to 3 is fair.

Then comes the part the pitch leaves out. Plus or minus 1% is not an extreme scenario. It is one standard deviation. Annual volatility on the 7 year is about 80 to 100 basis points, so the pitch describes the payoff at one sigma and sells it as insurance.

In this regime, a 150 basis point rise is believable. The 10 year has already added 37 basis points since September 1, from 4.80% to 5.17%, after touching 5.20% on September 24. The Fed hiked on September 16 and leaned hawkish. A 37 basis point move in less than four weeks is not 150, so the tail case needs its own justification. The rest of this note is that justification. At +150 basis points on the 7 year, the same arithmetic gives about -3.2%. "Lose nothing" dies past about 90 basis points. It is a window, not a floor.

So I asked the obvious question. What could go wrong?

The Bid That Is Supposed to Save Bonds

The standard answer is that someone big will buy duration at these yields, and pensions and insurers are the usual candidates. The relative value case for them is real. The 10 year Treasury yields 48 basis points more than the S&P 500's forward earnings, the 20 year 87 more, and investment grade corporates 126 more.

Pensions first. US corporate pension funding ratios are near 25 year highs, with the rolling z score above +2 sigma. At or above full funding, the rational liability driven investment (LDI) response is to derisk: sell equities and buy long duration bonds that match the liabilities. That is the mechanical output of every major pension's glide path policy, and the incentive has rarely been stronger.

It is not that simple. The Dutch pension system is working through a €1.6 trillion structural reallocation, a reminder that large pension rotations can create volatility instead of a stabilizing bid, depending on sequencing and market depth. And some US state pensions have been adding to private credit, treating the volatility there as a buying opportunity. The rotation is not monolithic. It depends on funded status, liability duration and governance constraints. Keep the private credit detail in mind when we get to Oracle.

Insurers second. Life insurers and annuity writers are natural buyers at these levels, but they buy investment grade corporates, not Treasuries. The Bloomberg US IG Corporate Index yields 5.95% at an option adjusted spread of 77 basis points. That is the instrument that matters for insurer asset liability management, not the on the run 10 year. At 5.95% with investment grade quality, new annuity business is very profitable and insurers are writing it. The demand is real, but it lands in 5 to 15 year corporates, not in long Treasuries.

The institutional bid has four limits.

• Funded status is rate sensitive. When yields rise, liabilities shrink, and so does the duration a plan needs to buy to hedge them. The bid fades exactly when the market needs it.

• Glide paths are slow. Pension derisking happens over quarters, not days. It is a structural tailwind, not a tactical floor.

• Supply overwhelms the bid. Even if every overfunded US pension rotated fully into long bonds tomorrow, annual Treasury net issuance would be a multiple of that flow. The bid is necessary but not sufficient to cap yields.

• Insurers are duration matched, not duration extending. They buy to match liabilities, not to bet on rates. Once they are matched, the marginal bid stops.

The rotation into bonds is real. The equity risk premium is negative, funding ratios are near records and investment grade yields are at 5.95%. But the bid is slow, partial and supply constrained. It may put a floor under duration at the margin, as a Bloomberg analysis argued on September 22, but it does not cap yields in a regime where fiscal supply is structural and the Fed is still hawkish. The "pension bid saves the trade" argument is a tailwind, not a backstop.

The Liability Side in One Index

To see why US pensions are in derisking mode, look at their liabilities. I02786, the Bloomberg US Pension Liability Index, tracks the present value of US corporate pension liabilities. When yields rise, the index falls, liabilities shrink and funded ratios improve mechanically. The 2022 and 2023 rate shock was the largest liability compression in decades, and it drove the biggest improvement in funded ratios since the early 2000s.

Three levels matter. The peak was 5,147 in July 2020, with rates near zero and liabilities at their maximum. The trough was 3,459 in October 2023, after the hiking cycle had compressed liabilities by about 33%. In September 2026 the index stands at 3,923, about 13% above that trough and still about 24% below the 2020 peak. Funded status is a two sided equation, and this index only shows the liability side. The asset side is the one a crash would hit.

Bloomberg does not publish a continuous funded ratio series for non US pension systems comparable to I02786. The best available proxies are listed company universes, and they are point in time.

Europe and Asia Are Not on the Same Clock

Are pensions and insurers in Europe and Asia on the same derisking horizon? No, and the divergence matters. Each region sits at a different point in the cycle, with its own rate regime, regulation and liability structure.

Europe is ahead of the US, and structurally disrupted.

• UK. The 2022 LDI crisis forced an accelerated derisking that the US is only now approaching. UK corporate pensions are about 102% funded, and many have already locked in duration matches or are in runoff. The marginal LDI bid from UK pensions is largely spent.

• Netherlands. The €1.6 trillion Dutch system is in the middle of its move from defined benefit to defined contribution under the Wet Toekomst Pensioenen reform. At 91% funded on the AEX proxy, it is not in derisking mode. It is in structural reallocation mode, which can create volatility instead of a stabilizing bond bid. The sequencing of that transition is a source of duration supply risk, not demand.

• European insurers. Under Solvency II they are already duration matched by regulation. Their marginal bid is limited to new premium inflows.

Japan is the divergence that matters most.

• Japanese life insurers hold 388 trillion yen of combined assets, about $2.5 trillion at 157 yen to the dollar, and they are the largest institutional bond buyers outside the Fed and the BoJ. Their behavior is the key variable for global duration.

• The picture is split. Meiji Yasuda doubled its plan for super long JGB purchases in the fiscal year to March 2027 to more than 2 trillion yen, calling 30 year yields of 3.7% to 3.9% "an excellent buying opportunity". Japanese insurers as a group bought the most super long JGBs in three years in June 2026, a net 630.5 billion yen.

• But in their April 2026 plans, 5 of 10 large insurers expected the 10 year JGB yield to climb another 50 basis points to 3%, which kept them on the sidelines. They were right: the 10 year JGB is at 3.07%. Fukoku Mutual explicitly avoided 30 and 40 year bonds.

• GPIF, the world's largest pension fund, has grown its fixed income holdings from 71 trillion yen to 129 trillion yen since fiscal 2019. That was proportional growth, not a derisking rotation. Societe Generale estimates GPIF could buy up to 12.3 trillion yen more JGBs without changing its allocation mix. That is a potential bid, not a committed one.

• The critical constraint: Japanese and Taiwanese insurers hedged only 41% of their foreign currency exposure as of June 30, 2026, the lowest since at least 2015. Dollar weakness accelerates the repatriation of foreign bonds into JGBs, and it also crystallizes losses on unhedged Treasury positions. For the US long end, that cuts both ways.

China and India are moving the other way.

• Chinese insurers are adding equities, not derisking into bonds. Their equity holdings rose by 640 billion yuan in the first half of 2025 to 3.1 trillion yuan, the highest since 2022. At the same time, Beijing is recapitalizing its largest insurers with 200 billion yuan of special government bonds, about $29 billion. That is a solvency backstop, not an LDI signal.

• LIC, India's largest insurer, is preparing to buy zero coupon sovereign STRIPS for the first time. That is early stage duration extension, years behind the US and UK cycle.

The JGB curve has become a global duration signal in its own right. The spread between the 2 year and the 10 year was about 105 basis points when a Bloomberg Macro View flagged it on September 7, and it is 114 now, from 1.93% to 3.07%. Treasury and gilt curves are biased toward the same steepening, as fiscal and duration risks outweigh short rate expectations. The loop runs both ways. If Japanese insurers extend duration at home, they take pressure off Treasuries. If they stay short and repatriate dollar assets, they add to it.

The US is the only major market where the LDI derisking trigger is live and mechanically compelling right now. Europe's bid is largely behind us. Asia's bid is opportunistic, rate conditional and complicated by currency. None of it is a structural floor under US duration.

What It Costs a Japanese Insurer to Own a Treasury

The next question decides the Japanese bid. What does it cost a Japanese insurer to buy Treasuries with the currency hedged, and does the trade beat JGBs?

Bloomberg's FXHCUSJP index, the standard 3 month rolling FX hedge cost for a yen based investor in dollar assets, stood at 2.93% on September 25, 2026. That is the all in annualized cost of selling dollars forward against yen, rolled every 3 months.

It has come down a long way, from 5.58% in January 2024, a decline of 265 basis points. About 147 basis points of that came from the yen side: as the BoJ hiked, the JPY 3 month implied yield rose from -0.35% to +1.12%. The rest came from the dollar side, where short rates are lower than in January 2024.

Three year swap rates tell the same story further out. The 3 year USD SOFR swap is at 4.74% against 2.16% for the 3 year JPY OIS, a gap of 258 basis points. The cross currency basis (DBXSJP3M) is at -17.3, compressed from -54 at its October 2024 extreme. A negative basis is an extra cost for a yen investor who funds dollars through FX swaps, and at -17.3 it adds about 17 basis points a year.

The compression of the basis is a signal in itself. When Japanese demand for dollar assets was intense, from 2022 to 2024, the basis was deeply negative and yen investors paid a large premium to access dollars. At -17, the market is already pricing less Japanese buying.

Now put the hedged Treasury next to the JGB.

Beyond 2 years, hedged Treasuries yield less than JGBs at every tenor, and at 2 years it is a wash. The gap widens dramatically at the long end. A Japanese insurer buying 30 year Treasuries fully hedged earns 2.57%, 159 basis points less than it gets on a 30 year JGB at 4.15%. That is the structural reason Japanese life insurers have been rotating into super long JGBs at home instead of extending into US duration.

Three caveats.

• Rolling hedge against a matched hedge. The 2.93% is the rolling 3 month cost. A life insurer matching 20 to 30 year liabilities would use a long dated cross currency swap, not a rolling 3 month forward. The 1 year forward points of -471 work out to about 3.0% a year (4.71 yen on a 157.3 spot), a touch above the 3 month cost, so hedging longer does not close the gap. The 1 year basis index (DBXSJP1Y) did not return data, so the longer end is an approximation.

• Unhedged and partially hedged books. At a 41% hedge ratio, some insurers are deliberately running FX risk and betting on a weak yen. Unhedged, the 30 year Treasury at 5.50% beats the 30 year JGB at 4.15% by 135 basis points. That is a currency bet, not an asset liability trade.

• The direction of the basis. If the BoJ keeps hiking and the basis compresses toward zero, the hedge cost falls further and hedged Treasuries become more competitive. But the JGB curve is steepening at the same time, as BoJ hikes push short rates up and fiscal supply weighs on the long end. To reach parity, the hedge cost would have to fall another 84 basis points at 10 years and 159 at 30 years.

At a 2.93% hedge cost, hedged Treasuries do not beat JGBs at any tenor. That is the quantitative reason Japanese buying of Treasuries is opportunistic and unhedged, not a structural asset liability bid. Hedged Treasuries only become attractive if further BoJ hikes compress the differential while JGB yields fall, two forces that partly contradict each other. On a hedged basis, the Japanese institutional bid for US duration is effectively closed at current levels.

Japan Is Already Going Home

Which leads to the risk that is growing: Japan reducing its Treasury holdings, and doing it faster if the yen appreciates.

The chart shows the core dynamic. Japanese Treasury holdings and USDJPY tend to move in the same direction. When the yen weakens, the unhedged dollar book gains in yen terms and the carry feels free. When the yen strengthens, the same book loses in yen terms and the incentive to repatriate accelerates.

Japan's holdings (US Treasury TIC data, HOLDJN on Bloomberg) peaked at $1,325 billion in November 2021, troughed at $1,062 billion in December 2024, and are falling again: from $1,239 billion in February 2026 to $1,104 billion in July 2026, a drop of $135 billion in five months. That is already the second largest five month drawdown on record, behind only the 2022 rate shock.

USDJPY stood at 157.30 on Friday, after peaking at 163.86 in July 2026. The yen has gained 1.2% over the past month and 2.8% over three months, so the trend is already in motion. One month implied volatility is 8.89%, which prices a one standard deviation move of about 4 yen over the next month and about 14 yen over a year. A return to 140 to 145, where USDJPY traded in 2023, is a 0.9 to 1.2 sigma move on a one year view. Getting there within a month would take a 3.0 to 4.3 sigma move.

The repatriation runs through three channels.

1. Stealth repatriation, already underway. TIC data show Japanese investors unwinding through Treasury bill maturities rather than outright sales of long bonds. It is quiet but cumulative: every maturing bill that is not rolled is a permanent reduction in the Japanese bid for US duration. Bloomberg analysis links a $100 billion reduction in Japanese holdings to sustained upward pressure on Treasury yields.

2. The unhedged exposure. Japanese and Taiwanese insurers hedged only 41% of their foreign currency exposure as of June 30, 2026. At 157, the unhedged dollar book sits on large unrealized FX gains accumulated since 2022. A move to 140, an 11.0% drop in USDJPY or a 12.4% rise in the yen, would erase a large part of those gains and turn positions bought above 140 into losses. The rational response is to close the position: sell Treasuries, repatriate, buy JGBs. The trigger is not a policy decision. It is a mark to market threshold.

3. FX intervention. Japan's Ministry of Finance spent an estimated $34.5 billion on intervention in May 2026. Intervention draws on FX reserves, which are held largely in Treasuries, so each episode is both a direct cut in Japanese Treasury demand and a signal that Tokyo wants a stronger yen. The two channels reinforce each other.

GPIF, with $1.8 trillion of assets, is the wild card and the most consequential single actor. Speculation is growing that it could raise its JGB allocation target, which would come with a lower foreign bond allocation. Tokyo is pushing pension funds and households to invest more at home, although analysts broadly see that as a long term shift, not a near term catalyst. Goldman Sachs expects life insurers to "probably lower USD exposure after raising it in recent years", while calling significant near term repatriation unlikely. The key words are near term. The structural direction is not in doubt.

How big could it get? The table below is illustrative. It applies the yen move to the whole $1.1 trillion of Japanese holdings, which is an upper bound, since insurers hedge part of their books and official reserves do not mark to market the same way.

The selling and yield figures are my illustrative assumptions, about 10 basis points per $100 billion of selling. Bloomberg's analysis only says that a $100 billion drop in Japanese holdings goes with sustained upward pressure on yields, without a number. The effects would be structural, not transient: a permanent reduction in the marginal buyer base, not a one off shock that reverses.

Four reasons this is non linear.

• The carry unwind amplifies the move. Yen appreciation triggers an unwind across the whole yen funded book, not just Treasuries. In August 2024, USDJPY moved 10 yen in days and forced deleveraging across global risk assets at the same time.

• JGB yields near three decade highs make home genuinely attractive. Markets may be underpricing how fast the repatriation math can shift. The 30 year JGB at 4.15% against a hedged 30 year Treasury at 2.57% is a 159 basis point argument for staying home, and it widens with every BoJ hike.

• The contagion is already visible in emerging markets. Malaysian bonds face Japanese outflows as their yield premium over JGBs has shrunk from 278 basis points, the five year average, to about 115. Treasuries are next in line, not first.

• The short JGB trade is structurally over. Bloomberg Intelligence notes that a more credible BoJ and potentially large inflows would turn long end JGBs from laggards into outperformers. The consensus short that kept Japanese capital offshore is unwinding.

The repatriation risk is already in the data. Holdings are down $135 billion in five months. The mechanism is stealth, bill runoff rather than long bond sales, so it will not show up in prices until the flow is big enough to overwhelm the marginal buyer. The yen is the accelerant, with only 41% of the insurers' foreign exposure hedged. Add the 159 basis point hedged gap at 30 years and the direction of travel is clear. The question is pace, not direction. For the belly trade at the top of this note, this is the tail risk that makes the "lose nothing" window fragile. A yen episode does not just move yields. It removes the marginal buyer exactly when the supply calendar is heaviest.

Bessent, the House That Tells Its Best Customer to Cash Out

Which brings me to Scott Bessent, who has decided to help.

On Friday, September 25, the Treasury Secretary posted on X that he had discussed "the desirability of a strong yen that reflects Japan's strong economic fundamentals" with Finance Minister Katayama. Both sides reaffirmed that the undervaluation of the yen is "a matter of concern". Prime Minister Takaichi had told President Trump in a meeting earlier in the week that an undervalued currency is "problematic". The yen rose as much as 1.2% to 156.94 per dollar on Friday, its best day in nearly three weeks and the best performer in the G10.

It did not come out of nowhere.

• June 22. Katayama and Bessent agreed to take "bold steps" on currencies if needed, and the two countries were described as increasingly "aligned" on FX policy, with USDJPY near 161.

• July. USDJPY peaked at 163.86, the weakest yen in four decades.

• Early August. The US and Japan carried out their first joint FX intervention in 15 years. Japan alone is estimated to have spent about $74 billion. Bessent let photographers catch his notepad, which showed a plan to buy $10 billion of yen.

• September 8. At a Southern Methodist University event in Texas, Bessent told traders who they were up against: "I am the house now, so when we intervene with the Japanese yen, I have pretty good insight into what the Japanese, what the Bank of Japan is going to do, what Japanese policymakers are going to do."

• September 9. The Treasury tripled its long dated buyback to $6 billion, from $2 billion. It did not stop the bond selloff.

• September 25. Verbal intervention again, this time with explicit "strong yen" language from both sides.

That is casino language. Bessent is claiming the house edge through inside knowledge of coordinated US and Japanese policy, and the message to carry traders is that they are betting against a counterparty who knows the next move before they do. Casinos usually keep their edge quiet. This one announced it on stage, tripled its bet the next day, lost the hand anyway, and is now asking its biggest customer to cash out.

Because that is the problem. Bessent runs a Treasury issuance calendar of more than $2 trillion a year, and he is also a public advocate of a stronger yen. Those two jobs are at war.

Japanese and Taiwanese insurers hedge only 41% of their foreign currency exposure. On the full $1.1 trillion of Japanese holdings, a 10% drop in USDJPY is up to $110 billion of FX losses. Bessent is, in effect, pulling the trigger on that scenario himself.

The real variable, and the real risk, is the BoJ. Fitch says further yen appreciation will likely require BoJ rate hikes, and that yen weakness "does not appear to be primarily due to relative stances of US and Japanese monetary policy". That is the crux. Verbal intervention, even joint intervention, is not enough without the BoJ. The BoJ raised its policy rate from 1.00% to 1.25% on September 18, the highest since 1995. SOFR was 3.88% on September 24. The gap of about 263 basis points is a structural anchor of the carry trade.

The BoJ has moved at a deliberately slow and well signaled pace to avoid a disorderly carry unwind. A BIS paper estimates that a 25 basis point policy surprise, when carry trade activity is elevated, can lead to almost 10% yen appreciation. That is the August 2024 playbook, when a modest BoJ surprise triggered a violent unwind. A Bloomberg Macro View on August 5, 2026 described the path to USDJPY 142 as likely to be "rapid and disorderly".

Wall Street is split. Wells Fargo argues the BoJ may struggle to deliver more hikes than markets already price. JPMorgan notes that a stronger yen could itself reduce the BoJ's incentive to hike, a self limiting dynamic. And on September 25, Morgan Stanley abandoned its long held call for dollar weakness in the second half of 2026, citing wider rate differentials and robust US growth, which runs straight into the strong yen narrative.

The fire Bessent is playing with, in order of probability:

1. Verbal intervention works too well (near term). USDJPY breaks 155, then 150. Japanese insurers sitting on unhedged dollar books start repatriating systematically. Bill runoff turns into outright long bond sales, and the $135 billion drawdown of the past five months becomes $200 billion or more.

2. The BoJ hikes in response to US pressure (medium term). A 25 to 50 basis point surprise, politically easier with US backing, triggers the carry unwind. USDJPY moves 10 yen or more in days, as in August 2024. Global risk assets reprice at the same time, and the 7 and 10 year Treasuries lose their marginal buyer at the worst possible moment, in peak supply season.

3. The trade deal becomes a currency clause (structural). The $550 billion US and Japan investment deal already embeds implicit FX expectations. If currency commitments become explicit, as has been rumored in the bilateral negotiations, Japan's capital allocation becomes a geopolitical variable instead of a market one. That would be a permanent shift in the Treasury buyer base.

4. Intervention reserves are finite. Japan spent an estimated $74 billion or more in the August joint intervention. If reserves are deployed at that pace, they last weeks, not months, and every intervention that fails to hold the level destroys credibility and invites carry traders to reload.

The paradox fits in one sentence. Bessent needs a strong yen for the trade deficit and the political mandate, and a strong yen accelerates the repatriation of the $1.1 trillion Japanese bid his issuance calendar leans on. He is asking Japan to do the one thing that undermines his ability to fund the US government at acceptable rates. The 30 year Treasury at 5.50% is the market's answer, and it was set before Friday's statement.

A strong yen, cheap long term funding and a loyal Japanese buyer. Pick two.

Yes, he is playing with fire. This time the JGB alternative is genuinely attractive, at 4.15% for 30 years. The hedge cost leaves hedged Treasuries 159 basis points behind at 30 years. And the unhedged share of the insurers' books is at its highest since at least 2015. The conditions for a disorderly repatriation have never been better aligned.

Oracle's Act of God Is an Air Permit

Then Oracle lit the fuse.

On September 24, Oracle sent a force majeure notice to Blue Owl Capital's STACK Infrastructure unit, the developer of Project Jupiter, a huge AI data center campus in Doña Ana County, New Mexico. Oracle is the main tenant. It is not trying to exit the project. It is protecting its contractual position to defer rent payments if the campus misses its planned 2028 opening.

Force majeure used to mean wars, earthquakes and pandemics. Here is the list of calamities behind this one, accumulated since August:

• an air permit application paused after environmental lawsuits

• natural gas pipeline permits blocked

• a New Mexico Supreme Court stay on the air permit proceeding for the adjacent microgrid

• community opposition and local regulatory resistance

Oracle told Reuters that force majeure notices are "commonplace" in deals like Jupiter. So are smoke detectors. The market did not find it commonplace. Oracle fell 4% to 7% on September 24, Blue Owl fell 3.4% to 3.6%, and Bloom Energy, which was supposed to power the campus with fuel cells, fell as much as 8.7%. Oracle is down 31% this year.

The credit market was ahead of the press release. Oracle's five year CDS has widened by 51 basis points in 10 trading days, from 186 on September 11 to 237 on September 25, and it has risen every day since Monday.

Against its hyperscaler peers, Oracle is an outlier. Its CDS is 2.4 times Meta's and 4.6 times Microsoft's.

There is no contagion to the core hyperscalers yet. But Oracle's CDS now trades at levels you would associate with a BB credit, not an investment grade technology company. The CDS market is pricing something the equity market is only starting to acknowledge.

This is not a simple corporate credit story. Oracle's data center build is financed through a layered private credit architecture that spreads the risk across the global institutional system.

• Project Jupiter (New Mexico, STACK Infrastructure, Blue Owl): about $18 billion of loans, quoted at 89 to 91 cents on the dollar by the syndicate banks Santander and Jefferies as of September 18, according to the FT. That is a 9 to 11 point mark to market loss on a campus that has not opened, recorded six days before the notice.

• The Michigan campus in Saline Township: PIMCO was reportedly in talks with Bank of America in April 2026 to provide about $14 billion of debt financing, potentially structured as a bond syndicated to other investors. Where that deal stands after the force majeure is unknown.

• Ares and Vantage Data Centers: Ares committed $2.4 billion to Vantage, part of it to fund infrastructure supporting Oracle's partnership with OpenAI.

In March 2026, the BIS warned explicitly that hyperscalers have turned to off balance sheet arrangements to finance their infrastructure expansion, "often in partnership with private credit firms", increasing the exposure of insurers and private credit funds in ways that are difficult to track. Oracle's force majeure is the first live stress test of that architecture.

The Japanese exposure runs through three interlocking channels.

1. SoftBank, the most leveraged node. SoftBank is not a passive investor. It is the most leveraged single point of failure in the AI financing chain.

• Its margin loan against ARM shares was raised to $25 billion on September 18, up $5 billion.

• Apollo expanded a NAV loan backed by Vision Fund 2 assets to $9.2 billion, the largest NAV loan in the world.

• SoftBank is reportedly planning another jumbo bond deal of $10 to $20 billion next week.

• Its total investment in OpenAI is heading toward about $65 billion.

• SoftBank shares slid 13% on September 14, their biggest drop in nearly three months, after the heads of Anthropic and OpenAI sounded the alarm on AI safety.

SoftBank is a tenant side risk through its Stargate and OpenAI data center commitments, and a creditor side risk through its financing of the wider AI infrastructure ecosystem. Stress in Oracle linked projects propagates straight into SoftBank's collateral values.

2. Japanese banks and insurers, with the FSA already watching. On September 25, the day after Oracle's notice, Japan's Financial Services Agency said it is stepping up scrutiny of AI data center financing by the country's biggest banks and life insurers, examining their risk management frameworks with a specific focus on US data center projects. The timing is hard to miss.

It follows a deteriorating trend. Japanese banks were standout underperformers in March 2026 amid private credit jitters, and the head of Japan's banking lobby said at the time that "we haven't reached the bottom yet" on US private credit failures. Morgan Stanley capped redemptions from a private credit fund. JPMorgan restricted lending to such funds after marking down loan values. Japanese bank stocks fell about 14% from their year high in that episode.

3. The opacity of off balance sheet debt. This is the most important structural point in the BIS warning. Because hyperscaler data center debt sits off balance sheet, in special purpose vehicles, project finance structures and private credit funds, the real exposure of Japanese banks and insurers to Oracle linked risk is not visible in standard regulatory filings. The FSA's scrutiny is an admission that regulators themselves do not have the full picture.

Every thread in this note converges on the same date: the September 30 quarter end.

Three features make this quarter end non linear, not just risky.

1. The opacity is the risk. Because the exposure is off balance sheet and spread through private credit structures, nobody knows the aggregate Japanese institutional exposure to Oracle linked debt. When marks move, the response is simultaneous and uncoordinated. Everyone discovers their exposure at the same time.

2. The collateral chains are circular. SoftBank's margin loan is secured on ARM shares. ARM's valuation depends on AI infrastructure demand. That demand depends on hyperscaler capex commitments. Oracle's force majeure is the first public signal that those commitments are not ironclad. The loop closes on itself.

3. Everyone uses the same exit. Japanese banks cutting private credit, Japanese insurers repatriating unhedged dollars and SoftBank managing its margin loan all sell the same thing, US risk assets, in the same quarter end window, while the Treasury market absorbs peak supply with a smaller foreign buyer base.

CVC Marathon's chief executive said this week that investment grade data center bonds are "the most compelling investment opportunity in credit", while explicitly warning about the lower rated end. Project Jupiter loans at 89 to 91 cents are sub investment grade in everything but name. The line between investment grade and the rest is the fault line in AI infrastructure debt, and Oracle just drew it in public.

It is not explosive yet, but the detonator is live. Oracle's notice is the first domino that makes private credit marks undeniable at quarter end. Whether it cascades depends on whether the Project Jupiter loans hold 89 to 91 cents or gap lower when Q3 books close on Wednesday.

Oracle calls it commonplace. For everyone who lent against its rent, it is the first real test of what an AI lease is worth.

ARM, the Collateral Pin

If you want one price to watch, it is ARM.

ARM fell 7.9% on September 24, its largest one day drop in months, on volume 37% above the previous session, in apparent sympathy with Oracle and Blue Owl. It gapped down at the open, $319.23 against a $332.56 close the day before, sold off to an intraday low of $303.65 and closed at $306.34. Friday's rebound to $310.32 was shallow. ARM is still $22.24, or 6.7%, below its close before the notice, and after hours trading at $311.49 suggests no catalyst overnight.

The price move is not just an equity story. SoftBank's $25 billion ARM margin loan, raised by $5 billion on September 18, creates a direct mechanical link between ARM's share price and SoftBank's financial stability.

At current prices the loan is well within safe territory on a pure loan to value basis. SoftBank holds about 90% of ARM's shares, so the collateral pool is enormous relative to $25 billion. Even on SoftBank's stake alone, the loan is 8.55% of the collateral. The thresholds in the table are illustrative, since the terms of the loan are not public, and ARM would have to fall to about $159, or about $177 measured on SoftBank's stake, before the loan reached 15% of the collateral.

The risk is not an immediate margin call. The risk is velocity and correlation. ARM fell 7.9% in one day on news that is structurally negative for the whole AI infrastructure thesis, and Oracle is down 31% this year. If the AI capex story deteriorates further, ARM, whose valuation depends on AI infrastructure demand, is no safe haven.

The margin loan is one of five pressures on SoftBank at the same time.

• The $25 billion ARM margin loan, with ARM's market value down about $23 billion in two sessions, about $21 billion on SoftBank's stake.

• The $9.2 billion Apollo NAV loan backed by Vision Fund 2 assets, the largest NAV loan in the world, secured on private AI company valuations that are increasingly hard to mark.

• A $6.5 billion revolving credit facility, expanded on September 18, due to expire and be redrawn.

• About $65 billion of total OpenAI commitments, with bond investors and lenders already demanding higher rates on what Bloomberg described as "an increasingly risky gamble".

• A planned jumbo bond deal of $10 to $20 billion next week, marketed into a market where SoftBank's debt costs are "soaring" and where, according to Morgan Stanley, Oracle's force majeure has just put AI infrastructure debt under fresh scrutiny.

The timing of that bond deal is acute. Morgan Stanley said on September 25 that the Oracle notice is "putting loan and lease documents under fresh scrutiny and adding to the challenges facing companies looking to tap debt markets to finance the AI buildout." SoftBank is trying to raise $10 to $20 billion in exactly that market, next week, with its main collateral asset down 6.7% in two sessions.

ARM sits at the center of a circular collateral chain that connects every thread in this note.

1. ARM falls, SoftBank's collateral erodes, its borrowing capacity tightens, and it is forced to scale back AI investment commitments.

2. SoftBank scales back, the demand signal from OpenAI and Stargate data centers weakens, the hyperscaler capex story deteriorates, and Oracle style force majeure events multiply.

3. More force majeure events push private credit data center loans lower, Japanese bank and insurer balance sheets take hits, FSA scrutiny intensifies, and US private credit positions are deleveraged.

4. Japanese deleveraging sells dollar assets, the proceeds go home, USDJPY falls, Bessent's strong yen push accelerates the cycle, and the marginal Treasury buyer disappears at peak supply.

ARM is not just an equity position. It is the collateral pin holding the SoftBank financing architecture together, and a 7.9% drop in a single day on one tenant's force majeure notice shows how fast that chain can reprice when the AI infrastructure narrative shifts.

Three trading days are left before Q3 books close on Wednesday, September 30.

• Project Jupiter loans are marked at 89 to 91 cents, so Q3 write downs are unavoidable for holders who mark at fair value.

• ARM is down 6.7% from before the notice, so SoftBank's NAV marks deteriorate.

• Oracle's CDS is at 237 basis points and rising, and its bonds traded at 2.8 times average volume on September 24.

• The FSA announced its scrutiny of Japanese bank and insurer exposure to US data centers on September 25.

• SoftBank's bond deal tries to price next week, right around quarter end.

Quarter end is not a tail scenario. It is the scheduled moment when every mark to market loss in this chain becomes a reported number, and when institutional risk managers, already under FSA scrutiny, set their Q4 allocations. Oracle's notice arrived six days before quarter end. For the AI infrastructure financing complex, the timing could hardly have been worse.

To the Risk Managers: The Party Is Already Over

The music stopped on Thursday. Your marks have until Wednesday to hear it.

Here is how the decision is being made right now, between Friday evening, September 25, and Wednesday's close. The process is already underway, from Friday night through the weekend, not next week.

Step 1: the Q3 marks are being finalized this weekend.

• Project Jupiter loans at 89 to 91 cents. Holders who carry them at fair value, the syndicate banks holding loans for distribution and the private credit funds, book a 9 to 11 point loss on face value, and their auditors will insist. A bank or insurer that holds the loan at amortized cost books a credit loss allowance instead of a mark.

• Oracle bonds. CDS at 237 basis points and bonds trading at elevated volume. Investment grade paper trading at near high yield CDS levels forces a conversation in every risk committee about whether the internal rating still holds.

• ARM, down 6.7% from before the notice. Anyone with ARM as collateral, including the lenders on SoftBank's $25 billion margin loan, marks the collateral pool.

• Treasury duration. The 10 year is up 37 basis points since September 1. A 10 year position with a duration of 7.7 has lost close to 3% in price since then. Every pension and insurer with unhedged duration books that loss.

Step 2: the FSA changes the risk appetite function.

The FSA's announcement on September 25 is not background noise. When a regulator says it is stepping up scrutiny of a specific exposure, the institutional response tends to be immediate and mechanical.

• Expect risk limits to be tightened preemptively. Chief risk officers do not wait for an examination to conclude. They reduce the exposure before the examiners arrive.

• Expect new commitments to pause. Any Japanese bank or insurer in the pipeline for a new AI data center loan has every reason to wait until the FSA framework is clear. That covers the Michigan campus, any deal after Oracle, and deals like Vantage Data Centers, which was seeking $2 billion of loans from Pimco and PGIM as recently as September 10.

• Daiichi Life's posture is the template. In April 2026, Daiichi was already tightening its selection of private credit managers after "several high-profile defaults overseas", saying "now's the time" to be more selective. That was before Oracle. Every other Japanese insurer is now having the conversation Daiichi had five months ago.

In April 2026, Finance Minister Katayama said private credit was not a major issue in Japan, citing limited exposure. The same day, the FSA said it was watching private credit risks and saw limited exposure. Five months later, the same regulator is stepping up scrutiny of AI data center financing. That is an escalation in official language, and compliance teams cannot ignore it.

Step 3: the Q4 decision tree. By Wednesday's close, every major Japanese institutional investor faces the same four questions.

On current data, every answer points the same way: reduce US private credit exposure, add JGB duration, repatriate dollar proceeds. The only question is pace.

Step 4: the house meets the allocation committee. Bessent's "I am the house" was about yen intervention. Friday's statement with Katayama on the undervaluation of the yen is the policy signal risk managers will be reading while they finalize Q4 allocations this weekend. The message they receive is that the US Treasury Secretary is actively working to strengthen the yen. The allocation response to that message, less unhedged dollar exposure, repatriation and more JGBs, is exactly what undermines the Treasury market Bessent is simultaneously trying to support with buybacks. The house is playing both sides of the same table.

Stanley Druckenmiller, Bessent's early mentor, called the bond buybacks a mistake in August, his point being that governments defending prices against fundamentals always lose. The risk managers deciding Q4 this weekend are reading the same fundamentals: Oracle's CDS at 237 basis points, Project Jupiter at 89 to 91 cents, ARM down 6.7%, the FSA on the case, 30 year JGBs at 4.15% against hedged 30 year Treasuries at 2.57%. Bessent's edge is informational. Their math is arithmetic. Arithmetic wins.

The SoftBank bond deal, reportedly $10 to $20 billion and trying to price next week, is the canary. If it prices at acceptable spreads next week, the market is saying the house narrative holds. If it is pulled, delayed or priced with a significant concession, it confirms that Oracle's force majeure has repriced the whole AI infrastructure debt complex, and that Bessent's informational edge has limits that arithmetic does not.

The Credit Market Has Already Read the Memo

Equities have not flinched. Credit has.

Since the start of August, the Bloomberg US CCC option adjusted spread (BCAUOAS) has widened 154 basis points, from 8.14% to 9.68%, while the S&P 500 rose 1.9%, from 7,600 to 7,743. The divergence is sharpest in the final week. CCC spreads widened 37 basis points over September 23 to 25, from 9.31% to 9.68%, while the S&P 500 sat only 0.7% below its August 13 peak of 7,799. The equity market is pricing a soft landing. The CCC market is pricing something else entirely.

The move is concentrated at the bottom of the quality spectrum. The broad high yield index (LF98TRUU) widened only from 260 basis points on August 28 to 294. This is a quality tiering event, not a broad credit selloff. Yet. In past credit dislocations the weakest credits have often moved first, investment grade next and equities last, as in the 2015 energy episode. The equity market's complacency today is historically consistent, and historically wrong at this stage of the cycle.

Oracle is not in a CCC index, since it is still rated investment grade. But its CDS and the CCC index are being repriced against the same backdrop: AI infrastructure debt, Project Jupiter loans at 89 to 91 cents, and a quarter end that forces those marks into reported numbers on Wednesday. At 237 basis points, Oracle's CDS is 3.2 times the average of its four hyperscaler peers.

The rest of the backdrop points the same way. The S&P 500's forward earnings yield of 4.69% is below every Treasury yield from 2 years out: 4.86% on the 2 year, 5.17% on the 10 year, 5.56% on the 20 year and 5.50% on the 30 year. A negative equity risk premium has not historically been immediately fatal for equities, but it removes the "there is no alternative" bid that carried the multiple expansion of 2020 and 2021. And the 10 year is up 37 basis points since September 1. When rates rise this fast, the first casualties are the most leveraged, least liquid credits, which is exactly the CCC cohort. The AI infrastructure loans look like the 2026 version of energy high yield in 2015: a sector stress that shows the wider leverage cycle has turned.

Every point on the CCC chart is a mark. On Wednesday, September 30, those marks become Q3 reported numbers, and what was a paper loss becomes a constraint on Q4 risk appetite. The marks behind that 9.68% will flow into the Q3 numbers of those who hold this risk, Japanese banks, insurers and private credit funds included. The FSA announced its scrutiny of exactly this kind of exposure on September 25. The chart is not showing a risk. It is showing the leading indicator that Q4 allocation decisions will be made from.

The closest precedent is the fall of 2018. The S&P 500 set a record on September 20, 2018, leveraged credit repriced, and the index fell 19.8% into Christmas Eve as the earnings revision cycle caught up with the credit signal. The difference today is that the credit repricing is happening together with a negative equity risk premium, a Treasury supply calendar of more than $2 trillion, a Japanese repatriation risk, a strong yen push from Washington and a forced quarter end mark. In 2018, only one of those five was present. The CCC market is not predicting a crash. It is predicting that the S&P 500 at 7,743 has not yet read the same document that the credit market finished reading on September 24.

The cleanest signal is the split inside high yield.

What did not happen to BB spreads matters most. From June 17 to September 25, BB spreads moved only 23 basis points, from 1.50% to 1.73%, while CCC spreads rose 202 basis points, from 7.66% to 9.68%. The gap between the two widened from 6.16 to 7.95 percentage points, 179 basis points in 100 days. This is not a broad credit selloff. It is a surgical repricing of tail risk at the bottom of the quality spectrum, while the credits closest to investment grade are largely untouched.

The CCC bond index widened 17 basis points on Thursday alone, to 1,010 basis points, with a yield to worst of 15.04% and a return of -2.15% for the month to date. In loans, CCC discount margins trade 4.5 times wider than single B and more than 12 times wider than double B, and that gap has widened materially over the past year.

What does that imply for defaults? The standard conversion divides the spread by one minus the recovery rate. At a 40% recovery, a 9.68% spread implies an annual default rate of about 16.1%, and the CCC bond index at Thursday's 10.10% implies about 16.8%. Bloomberg's model puts the one year default probability of the CCC cohort at 7.87% to 7.88%, and at 3.70% for high yield as a whole. The market is pricing about twice the model's base case. The gap is the risk premium, the compensation for the timing, clustering and recovery of defaults, not just their expected number. With private credit marks opaque and quarter end forcing simultaneous write downs, that premium is rational.

Perspective matters. At 968 basis points, CCC spreads are still well short of the crisis peaks of 2016 and 2020, when they were much wider. This looks like a sector stress, AI infrastructure and private credit, not yet a systemic crisis. For reference, CCC default rates peaked at about 12% after the 2015 and 2016 energy bust and at about 16% to 18% in 2020. In 2008 and 2009, CCC spreads went well above 2,000 basis points and defaults ran above 30%. What stands out today is the speed, from 563 basis points on February 2 to 968, a 72% rise in less than eight months.

Bloomberg's MLIV team noted on September 14 that distress is "not elevated enough to trigger a slowdown in growth", which suggests yields have more room to run. That is the nuance. The CCC market is in distress, but the distress is not yet broad enough to feed back into the real economy through tighter lending standards and capex cuts. That feedback typically takes two to three quarters to show up once CCC spreads break 900 basis points.

There is a counterintuitive argument going around. In a Bloomberg Macro View on September 24, Sebastian Boyd noted that CCC bonds, precisely because their duration is short, offer relative protection against further rate rises compared with investment grade or higher rated high yield. At a 15% yield to worst, the carry absorbs a large rate move before total return turns negative. That is arithmetically correct and strategically dangerous here. The risk in CCC is not duration. It is default. Oracle's force majeure is not a default, but it is the first public sign that AI infrastructure cash flows can be deferred, and deferral is the first thing a lender prices.

A gap of 7.95 points between CCC and BB, nearly double its February low of 4.08, is the credit market saying that the bottom of the quality spectrum is being repriced for a specific, identifiable stress while the rest of the market has not caught up. In past episodes when the gap was this wide and moving this fast, realized CCC default rates have tended to rise toward the market's implied level within about two quarters. Bloomberg's 7.87% base case and the market's 16% will converge. The only question is whether they converge up, with defaults accelerating, or down, with the stress contained. The September 30 marks are the first hard data point.

To Investors: The Party Is Over

And now the warning for everyone else.

This chart takes the S&P 500's earnings yield and subtracts the 10 year Treasury yield, every day since January 1990. Green means stocks pay you more than bonds. Red means they pay you less. On September 25 it stood at -1.39 points, the deepest reading since the early 2000s. On forward earnings the gap is smaller but it has the same sign: the S&P 500's forward earnings yield is 4.69%, against 5.17% for the 10 year, 5.56% for the 20 year and 5.95% for investment grade corporates.

For roughly two decades, from the early 2000s until the last few years, you were paid a premium to own equities over Treasuries. That premium is gone. You are now paid less to own the S&P 500 than to lend to the US government for ten years, at the precise moment when the largest foreign buyer of that government's debt is being told to go home, when pension funds have every reason to sell equities and buy bonds, and when the AI capex story that carries the index has just hit its first force majeure.

The last time this spread sat this deep in the red was the late 1990s and the early 2000s. We know how that ended. I am not calling the day. I am telling you the party is over, and that the people who run the biggest balance sheets in the world spend this weekend deciding how fast they leave.

I Wrote It Down Before the Smoke

Most of this chain is not new to readers of my work. Here is what I published, with dates, so you can judge for yourself.

• The China AI Disruption Thesis, the CrossVol Research book I co-authored, May 23, 2026: "Oracle is the acute case, and the one to watch as a leading indicator."

• My MPRA working paper adapted from it, posted June 12, 2026: "Long Oracle CDS: BBB- watch, negative FCF, debt-to-equity approaching 500%. Prime candidate for first credit event in the AI infrastructure complex." In the predictions I timestamped on OSF on June 4, I listed "CDS spread widening >50bps on hyperscaler names" for Q4 2026 to Q1 2027. Oracle's CDS went from 186 to 237 basis points between September 11 and September 25. That is one name, it is early, and it is not a credit event. It is the direction I flagged.

• Convergent Faults, my paper on private credit's synchronized systemic risk, first posted on Zenodo on June 5, 2026 and later on SSRN and SocArXiv: "The five channels are coupled, and the joint stress horizon is short. Under realistic macroeconomic conditions, each channel tightens the others." The convergence window, I wrote, "opens in Q3 2026", and it centres on the first half of 2027. On the mechanism that quarter end now puts to the test: "When the bank reassesses the collateral, either through a routine quarterly review or in response to an external event such as a software-loan default, the advance rate falls. The platform either has to post additional collateral or to repay a portion of the outstanding draw. If neither is feasible quickly, the platform sells loans into a thin secondary market to raise cash, which itself generates a further mark-down event."

• The Coming Shadow Banking Crash, my book, June 8, 2026. From its published description: "Three trillion dollars of private credit now sits outside the reach of bank regulators. It is fed by pension funds reaching for yield, by insurance balance sheets reaching for spread, and by an artificial intelligence capital expenditure cycle whose cash flows may not arrive in time to service the debt." In the book, on the data centers: "A meaningful share of the announced builds will be delayed, scaled back, or relocated." On the selling: "If multiple regional banks issue margin calls on multiple private credit funds in the same week, the asset sales required will be concentrated in time and concentrated in asset type. The market impact will be larger than the aggregate exposure would suggest." The book names no company, by design, and it does not predict a date.

• Here on Substack. On August 20: "this time the fuse runs through AI capex financed by debt." On September 19, the BoJ hike "threatens to pull Japanese money home and steepen global curves from the long end."

What I did not put in print is the Japanese leg in the exact form it is taking: SoftBank's margin loan on ARM, and Japanese banks and insurers exposed to US data center debt. The books mapped the architecture. This week showed who is standing on it.

What I Am Watching

• SoftBank's bond deal: size, price and concession, or whether it gets pulled.

• Project Jupiter loan quotes once the September 30 marks are in.

• Oracle's five year CDS against 237 basis points.

• ARM against its September 24 close of $306.34.

• USDJPY against Friday's 157.30 and its intraday low of 156.94, and any word on intervention.

• The 30 year JGB against 4.15%, and the 10 year Treasury against 5.17%.

• Any follow up from the FSA on bank and insurer exposure.

• The Fed on October 28 and the BoJ on October 30.

The CrossVol terminal tracks index volatility, skew, term structure and dealer positioning live, at crossvol.com for $99 a month.

This note is for information only and is not investment advice.

Further reading

• The Treasury Basis Trade Is Losing on Three Fronts

• The AI Debt Trojan: Why the US IG Index Just Got Prettier and More Fragile

• What JPM Just Confirmed About 2027 AI Data Centers: PJM Grid Constraint and Hyperscaler CDS Divergence

On Wednesday, September 23, the Treasury market delivered the kind of move that usually shows up everywhere at once. The 10 year yield rose 15.9 basis points to 5.125%, which Bloomberg puts at a 3.39 sigma day against its own 30 day realized volatility. The 5 year rose 17.2 basis points. Brent traded above $102, Fed Governor Barr signaled that further tightening is likely needed, and equities reversed lower. Stocks and bonds sold off together. The VIX rose 0.75 to 14.96. That last number is the story. Index volatility registered the shock the way a sleeping market registers a noise, with a flinch, and the reason it can afford to stay that calm is the same reason the calm is fragile. Implied correlation across the S&P 500 sits near its lows, large cap single stocks have become about as volatile as small caps, dealers were last reported short vega on the downside, and the bond hedge that would normally cushion an equity drawdown has not worked since March. This note walks through the equity vol complex first, then the correlation channel that connects it to today's rates shock, and ends with a one page risk map.

The VIX Rose, Quietly

The move in spot VIX needs to be read against where it came from. The index peaked at 17.84 on September 10, its highest close since late July, printed 17.71 on September 16 and then collapsed. It lost 3.50 points in four sessions to close at 14.21 on Tuesday, September 22, about two thirds of it on Thursday, September 17, the day before the September 18 triple witching expiry. Today it rose to 14.96, up 0.75, or a little over 5%. For a three standard deviation day in the 10 year, that is a muted response, but it is about what the VIX usually does when the S&P 500 slips by less than 1%, which is all stocks have given up so far. Index vol is tracking spot, and it is the equity market that has barely priced the rates shock. The rest of the complex explains why that can change quickly.

The VVIX, the implied volatility of the VIX itself, rose from 83.17 to 87.79, in step with spot. The more telling number is where it started. The VVIX peaked at 102.66 on September 10 alongside the VIX, then fell to 83.17 on Tuesday, only 0.27 above its August low of 82.90, set on August 27, and its third lowest close of the past year. Protection on the VIX itself is cheap by the standards of the past year, which fits a market that is calm on the surface. Vol of vol often turns before spot vol at inflection points, so the level to watch is the floor near 83. A VVIX that keeps climbing from here would be the options market saying it no longer trusts the calm.

Tail Demand Is Still There

SKEW closed at 144.80 on September 22, and the September 23 print was not yet published at the time of writing. It peaked at 154.49 on September 11, its highest close since early July, and has since fallen back to about its median of the past year, near 144. That is still well above the 130 to 135 area that would describe a relaxed tail market, and SKEW was at or below 135 on five sessions in August. A SKEW in the mid 140s means the market is still paying up for deep out of the money puts relative to at the money options. Put that next to a VIX that spent the previous week falling and you get the familiar signature of complacency in the at the money strip with residual fear in the wings.

The Curve Is Steep, and the Front End Just Moved

The VIX term structure is in steep contango. On Bloomberg's intraday prints, the 9 day VIX sits at 13.16, the 30 day VIX at 14.96, the 3 month at 18.00 and the 6 month at 20.00, a spread of 6.84 points from 9 days to 6 months. Near dated implied volatility collapsed into the expiry, with the 9 day VIX down 4.01 points on September 17 alone, while the 6 month index, at 20, sits near its lowest level of the past year and right on its median since 2012. The curve is steep because the front end is depressed.

The detail that matters today is how the curve moved. At Tuesday's close the same four points stood at 12.13, 14.21, 17.61 and 19.79. The 9 day rose 1.03, the VIX 0.75, the 3 month 0.39 and the 6 month 0.21, and the spread from 9 days to 6 months narrowed from 7.66 to 6.84 points in a single session. Contango is still steep. The front end rises the most on almost every day the VIX rises, so the flattening on its own is ordinary. The confirmation of stress would be a 9 day VIX trading above the 30 day, which last happened on September 15.

The futures curve puts a price on the same shape. The October contract trades at 17.70 against spot at 14.96, a basis of 2.74 points. November is at 18.45, December 18.80, January 19.55, February 20.00, March 20.25, April 20.60 and May 20.80, the high of the curve, before June dips back to 20.60. The back months sit within about a point of each other, so the 0.20 dip from May to June says little about when uncertainty peaks. The basis has two readings. For anyone long volatility through futures, it is a steep carry cost, since the October contract converges to wherever the VIX sits at its October 21 expiry. For anyone short volatility through futures, it is income, collected in exchange for exposure to exactly the kind of spike that a rates shock can trigger. Either way, the basis is mostly the premium the market charges for owning volatility, plus event risk. The October contract settles on 30 day implied volatility from October 21, a window that includes the November 3 midterm elections.

Dealers Were Short the Wrong Kind of Vega

The positioning picture makes the calm more fragile. In late August, UBS derivatives strategists noted that record buying of S&P 500 index options had left dealers short net vega, particularly on the downside, and described the vanna profile, the rate at which dealers get shorter vega as spot falls, as historically high. In that configuration a selloff forces dealers to buy volatility to rebalance, and they buy it into a rising VIX. The hedging flow amplifies the move it is responding to.

Three things add to that fragility. First, the September 18 triple witching expiry, one of the largest on record at an estimated $7 trillion of notional according to Citadel Securities, removed a large block of the gamma that had been dampening realized moves. With that cushion gone, the index is more directionally sensitive. The UBS read predates that expiry, so some of the short vega it described may have rolled off with it. Second, the leveraged and inverse single stock ETF complex keeps growing, with Jane Street now providing about $1.2 billion of notional swaps to roughly 75 such products according to Bloomberg, which adds a structural source of forced delta and vega hedging. Third, banks have reportedly been offloading their exposure to those products through exotic crash puts, which moves the tail risk around the system without removing it.

What Barclays Saw, and What the Index Shows

On September 22, Barclays flagged that the traditional volatility discount of large cap stocks against small caps, historically about 10 points, has effectively disappeared. The observation is about single stocks. Large cap names are now about as volatile as small cap names, which removes the low volatility edge that made large caps the calm part of the market.

It is tempting to read that straight into the indices, and over the past month the index data says something different. The RVX, the Russell 2000 volatility index, stands at 20.00 against a VIX of 14.96, a gap of 5.04 points. On August 24 the same gap was 3.81, with the RVX at 19.66 and the VIX at 15.85. At the index level the small cap premium has widened over the past month, after shrinking over the summer from an average of about 6.3 points in May and June to 3.7 in August, which is the direction Barclays describes. Today's 5.04 is back near its median since 2015 of about 4.8. The gap moved between 3.75 and 4.60 over the past month and was 4.05 on September 21. It then jumped to 4.57 on September 22, a day the VIX fell, and to 5.04 today, with small cap volatility leading today's rise, the RVX up 1.22 against 0.75 for the VIX.

The two observations fit together once you add correlation. If large cap single stocks are as volatile as small caps while the S&P 500 index stays calm, the only thing reconciling the two is that those stocks are moving independently of each other. The diversification inside the index is doing all the work. That is the condition Bloomberg Intelligence flagged on September 8, when it warned that historically low stock correlation was suppressing index volatility and that the VIX may be understating true risk.

Correlation Is the Variable

The Cboe 3 month implied correlation index, COR3M, closed at 9.53 on September 22, matching its 30 day low from September 4, after peaking at 13.07 on September 16. Today it trades at 10.06, which is an implied correlation of about 0.10. To a first approximation, index implied volatility is the average single stock volatility scaled by the square root of average correlation. Hold single stock volatility fixed and correlation becomes the main lever on the VIX. COR3M is a 3 month measure, so its natural partner is the 3 month VIX, at 18.00 today. With COR3M near 10, the market is pricing constituent moves as largely offsetting, and index volatility reflects a diversified portfolio. The stress inside the names stays hidden.

The transmission from here is mechanical. Large cap single stocks are already about as volatile as small caps, according to Barclays. Low correlation hides it, because the index nets the moves out. A macro shock that hits every sector at once pushes correlation toward one, and a rates move like today's is the textbook case. The VIX then has to reprice toward the volatility of its constituents, and dealers who are still short vega on the downside have to buy that repricing. Today's rise in COR3M, off a 30 day low on the day of a 3.39 sigma rates move, is what the first step of that sequence looks like. One day does not make a regime, but this is the variable to watch.

A 3.39 Sigma Day in Treasuries

The Treasury selloff is the macro catalyst that tests the correlation regime. The 2 year yield rose 15 basis points to 4.908%. The 5 year rose 17.2 to 5.010%, the largest of the four benchmark tenors. The 10 year rose 15.9 to 5.125%, and the 30 year rose 11 to 5.413%. Over the past 30 days the 10 year has realized 4.70 basis points a day, or 75.8 annualized, and Bloomberg puts today's move at 3.39 sigma. Under a normal distribution, a daily move that large happens less than 0.1% of the time. That sigma figure is specific to the 10 year and its own realized volatility, since each tenor has its own.

The belly led. The 5 year moved the most, consistent with a market repricing the path of Fed policy after Governor Barr's comments and a strong PMI. The shape of the curve moved with it. The spread between the 5 year and the 30 year flattened by 6.2 basis points, and the 2s5s10s butterfly cheapened the belly by 3.5 basis points. This is a policy repricing concentrated where the Fed path lives, and it is exactly the kind of shock that reaches every equity sector through the discount rate at the same time.

Stocks and Bonds Are Moving Together

The 60 day rolling correlation between daily S&P 500 returns and daily changes in the 10 year yield stands at -0.593, about 2.7 times its one year average of -0.217. The sign matters, and it is easy to read backward. The series correlates stock returns with yield changes. A negative number means stocks fall when yields rise, and since bond prices fall when yields rise, stocks and bonds tend to move together, gaining and losing on the same days. That is the regime in which Treasuries stop hedging equities.

On the chart, the correlation crossed below zero in mid March and has stayed there for six months, with its most negative reading, -0.720, in early June. The positive stretches of the past year, from October to early December 2025 and from February to mid March 2026, were the periods when the bond hedge worked. Today's joint selloff, with stocks lower and yields up 15 to 17 basis points from the 2 year to the 10 year, is that regime expressing itself on a large day. The practical consequence is that the diversification benefit of a balanced stock and bond portfolio is weak right now, and cross asset vega exposure adds up.

Rates Volatility Has Room to Run

The MOVE index closed at 78.56 on September 22, with a 90 day z score of +0.88, above its recent average but well short of extreme, and down from a peak of 83.90 on September 14. That reading predates today's 3.39 sigma move. A day like today usually lifts implied rates volatility, and a MOVE heading back toward or through its September 14 peak while the VIX sits near 15 would be a dangerous combination. Rates volatility rising while equity volatility stays suppressed is the environment in which correlation resets tend to be most violent, because the equity market has yet to price the shock that the bond market is already trading.

The Small Cap Credit Channel

Barclays' factor research adds a fundamental layer. About 32% of Russell 2000 constituents are unprofitable, and leverage runs at 3.5 times net debt to EBITDA. With the 5 year Treasury at 5.01%, a sustained rate shock goes straight at the refinancing capacity of that part of the market. That is a credit channel that feeds back into equity volatility through earnings risk, and it is consistent with the RVX leading today's move.

The Risk Map

The picture is a fragile low volatility equilibrium. On the surface, the VIX is near 15 after falling 3.50 points in the four sessions to Tuesday. Underneath, protection on the VIX itself is cheap, with the VVIX near its lowest levels of the year. SKEW is back at its one year median but still above the relaxed zone. The VIX curve is steep because the front end is depressed. Dealers were short vega on the downside in UBS's late August read, and the expiry cushion is gone. Implied correlation is coming off its lows, rates volatility was already above its 90 day average before today's shock, and since March stocks and bonds have tended to move in the same direction, so Treasuries have stopped hedging equities. The core risk is a correlation convergence event. If the rates shock keeps hitting every sector at once, correlation and the VIX reprice together, and any short dealer vega left on the downside makes that move nonlinear.

Five signals are worth watching from here. The first is COR3M, and whether today's rise off the low continues. The second is the 9 day VIX against the 30 day, for a new inversion at the front of the curve. The third is the VVIX floor near 83. The fourth is the next MOVE prints, for whether rates volatility follows the 3.39 sigma day. The fifth is the correlation between stocks and the 10 year yield, because as long as it stays this negative, the usual hedge will not be there when the VIX finally catches up.

The CrossVol terminal tracks the VIX term structure, skew, implied correlation and the correlation between stocks and yields live, at crossvol.com for $99 a month.

Djellal Djouad

Related: The Crash Fuse Is Lit

Sources

Bloomberg News markets coverage of the Treasury selloff, oil and Fed commentary, September 23, 2026.

Bloomberg Intelligence on S&P 500 volatility and the risk of a correlation reset, September 8, 2026.

Bloomberg on large cap stocks losing their low volatility edge, citing Barclays, September 22, 2026.

Barclays Equity Factor Insights, September 2026, published September 14, 2026.

Bloomberg on dealer vega exposure and VIX upside, citing UBS derivatives strategists, August 25, 2026.

Bloomberg on Jane Street joining the swap dealers behind leveraged single stock ETFs, September 20, 2026.

Bloomberg on banks offloading leveraged ETF risk through exotic crash puts, August 2, 2026.

Citadel Securities estimate of the options notional expiring on September 18, 2026.

Cboe daily index history for VIX, VVIX, SKEW, VIX9D, VIX3M, VIX6M, RVX and COR3M.

Bloomberg data for the September 23 intraday prints, VIX futures, Treasury yields, the MOVE index and the correlation between S&P 500 returns and 10 year yield changes.

This note is for information only and is not investment advice.

Further reading

• The Market Doesn't Price One Distribution Anymore. It Prices Three.

• The Treasury Basis Trade Is Losing on Three Fronts

• The AI Debt Trojan: Why the US IG Index Just Got Prettier and More Fragile

Wall Street spent this week reassuring itself. The Treasury basis trade, the leveraged bet that hedge funds run between cash bonds and futures, has fallen to roughly 900 billion dollars from 1.26 trillion at the start of the year. Strategists from the large dealers lined up to call it benign. Fewer dislocations to exploit. A naturally self-correcting trade. No evidence of stress. One head of rates strategy went further and argued that a shrinking opportunity set means underlying Treasury demand is stronger than the market thinks.

All of that can be true and still miss the point. A smaller number is not the same as a smaller risk. Read where the book actually shrank, and where it did not, and the reassuring story turns into a map of exactly where the next accident sits.

What left, and what stayed

The decline is concentrated in the two and five year part of the curve. That is where the arbitrage compressed, because asset managers cut their net long futures positions as the policy outlook flipped from cuts to hikes after oil surged. Less futures demand, smaller gap to cash, smaller trade. Fine.

But the basis book tied to the 25 to 30 year sector did not shrink. It grew. The leverage did not leave the system. It migrated to the longest, most convex, least liquid part of the curve. That is the single most important sentence in the entire story, and it is the one the benign framing skips. You did not de-risk. You moved the risk to the place where it is hardest to hedge and hardest to exit.

Why convexity breaks the hedge

The basis trade is sold as a near riskless convergence play. Buy the cash bond, sell the future, earn the small spread, scale it with borrowed cash. The problem is that the hedge is only clean when the curve moves in parallel. On a 100 million dollar position, a 10 basis point parallel shock leaves only about 20 thousand dollars unhedged. That is the number that makes people comfortable running the trade at fifty times leverage.

Now break the parallel assumption. Double an adverse shock from 50 to 100 basis points and the residual loss does not double, it more than triples. Two forces turn against the hedger at once. Convexity means the price response is not linear, so a hedge ratio calibrated to small moves is wrong in large ones. And the future carries an embedded option over which bond is cheapest to deliver. In a bear flattener, that cheapest to deliver bond switches, the deliverable basket shifts, and the ratio that was correct at inception is now simply wrong. The hedge stops being a hedge at the exact moment you need it. That failure is largest in the long end, which is precisely where the book just grew.

The dealers are short convexity too

This is not only a hedge fund problem. Look at positioning. Speculators sit extreme short the front end. Dealers are the other side, extreme long. A dealer community that is long the front end into a rising yield tape is effectively short convexity. As yields rise, dealers lose on the mark and have to hedge by selling more futures, which pushes yields higher still. The hedge feeds the move rather than damping it. You have leveraged funds and their dealer counterparties both positioned so that stress makes them sell the same direction. That is the definition of a crowded, reflexive trade.

Liquidity is the accelerant

Now put that positioning into today's order book. Top of book in the market is near one million dollars, the lowest reading on the series. That is nothing relative to the size of the leveraged positions sitting on top of it.

When depth is that thin, the basis does not widen a few basis points in an orderly drift. It gaps. A spread that was five basis points becomes 50 to 100 in minutes because there is no depth to absorb the flow. And the leverage turns that gap into a solvency event. On a 100 million dollar position earning an 85 basis point spread, daily carry is around 2,300 dollars. A 50 basis point adverse move marks the bond leg near 4.1 million dollars, which is a margin call on the order of 1,700 times the daily carry. You do not carry your way through a call that size. At fifty times leverage the position that survives a 50 basis point day is the one whose margin buffer exceeds the move, and at fifty times that buffer does not exist. The trade is not calm because it is safe. It is calm because nothing has forced it yet.

The tail that comes

Here is the part the desk should say out loud. The basis trade and the dispersion trade are not two different risks. They are the same tail seen from two desks. Dispersion is long single name volatility and short index volatility, and it dies when correlation goes to one. Basis is a convergent spread run on heavy leverage, and it dies when a correlated selloff forces same direction liquidation into a market with no depth. Both are killed by the same two part failure. Correlation goes to one and liquidity goes to zero at the same instant.

The macro backdrop is set up for exactly that. The stock bond correlation is now positive, around plus 0.61. The bond leg is no longer a diversifier, it is a second correlated exposure. Supply is heavy, with large auctions rolling and the Treasury now a far larger share of total issuance than a decade ago. The regime just flipped to hikes into a supply and geopolitics driven vol backdrop, which is the environment in which repo tightens and the long end gaps. Repo is flush with cash today. It was flush before it wasn't, in March 2020, when funds unwound basis into a vacuum and it took Federal Reserve bond buying and repo intervention to stop the cascade. That was not a freak event. It was this book.

What to watch

Do not watch the headline size. A shrinking notional is being read as an all clear when it is really a change of address. Watch the long end basis, the 25 to 30 year sector where the leverage concentrated. Watch top of book depth and repo, because the trade is fine until funding tightens or vol spikes, and then it is not fine all at once. And watch the language. Comfortable with the resilience, and repo is flush with cash, are the exact words you hear right before the resilience is tested.

The basis trade did its job for years. It provided liquidity and it convergence traded away small dislocations. But the reduction in its size this year did not remove the tail. It concentrated it in the most convex, least liquid corner of the largest bond market in the world, into a regime where bonds no longer hedge stocks. The number got smaller. The risk got sharper.

Related: The Crash Fuse Is Lit

Further reading

• The Treasury Basis Trade Is Losing on Three Fronts

• Same Symptoms, Different Disease: why 2026 is not a rerun of 2022

• Validated and capped