BofA says a 20% cut in 2027 AI spending would take 7 to 8% off S&P 500 earnings. Alone, that is a slowdown. Add falling multiples, a hot CPI and 401(k) losses, and a Q2 2027 recession becomes real.
Un focus sur les marges exceptionnelles des boites Semis. Si réduction des Capex + arrivée de nouveaux concurrents plus compétitifs, elles seront contraintes de brader leurs prix pour faire du volume. C’est un principe de commerce de base et les marges diminueront pour retrouver des niveaux d’avant super cycles. Il y a une erreur de grille de lecture : l'industrie des semi-conducteurs a toujours obéie à une logique cyclique par le passé (suivant les cycles industriels et automobiles).
Le piège des multiples est bien présent et vouloir évaluer les géants des puces et de l'intelligence artificielle comme des valeurs de croissance pures à long terme occulte le fait qu'ils traversent actuellement un super-cycle d'investissement (les Capex massifs dans l'IA).
Le retournement de marge est inéluctable. Une fois que les besoins massifs d'infrastructure (les data centers) seront en grande partie satisfaits ou que la concurrence s'intensifiera, ces entreprises connaîtront mécaniquement un retour de balancier sur leurs volumes et leurs marges bénéficiaires exceptionnelles (qui atteignent parfois des sommets de $70% à $80%). C’est de la simple logique au final.
The Fed may be the weakest of your three amplifiers. In 1994 and 2018, both tech-led selloffs under a hiking Fed, the market bottomed in December and had recovered by mid-February once the Fed changed course, so policy ended up absorbing the shock rather than spreading it. A 20% capex cut would itself pull the hike odds down, since it removes the demand the hawks are leaning against. What would the Fed have to do in your scenario to become an amplifier, keep hiking into Q1?
Yes, keep hiking into Q1, and that is already priced. after today’s CPI, swaps price hikes for December and one more by January 27, better than even odds of a second move.
I think that the différence with 1994 and 2018 is inflation. Core CPI was under 3% and falling in 1994, and core PCE was below 2% in late 2018, so the Fed could turn fast, within two weeks of the December 2018 low.
Today core CPI has printed 0.3% three months in a row, about 3.7% annualized, with energy passing into services. A capex cut hits earnings and credit spreads within weeks but inflation only with a lag, so the Fed keeps tightening while the shock spreads.
For me that looks more like 2022 in my view, when the Nasdaq lost a third of its value.
I imagine that even a pause may not help the long end. Late 2024 the Fed cut by a full point and the 10yyield rose almost as much.
The amplifier I'd add is where the money hides. In October investors already rotated into the old economy: US energy rose 13.9%, utilities 10.1% and financials 5.8%, while chips fell 15.4%. That shelter works while growth holds. If a capex cut tips growth over, no corner is spared, the shelter included. More on the rotation: https://floodlinesresearch.substack.com/p/a-short-trip-back-to-the-steam-age
Un focus sur les marges exceptionnelles des boites Semis. Si réduction des Capex + arrivée de nouveaux concurrents plus compétitifs, elles seront contraintes de brader leurs prix pour faire du volume. C’est un principe de commerce de base et les marges diminueront pour retrouver des niveaux d’avant super cycles. Il y a une erreur de grille de lecture : l'industrie des semi-conducteurs a toujours obéie à une logique cyclique par le passé (suivant les cycles industriels et automobiles).
Le piège des multiples est bien présent et vouloir évaluer les géants des puces et de l'intelligence artificielle comme des valeurs de croissance pures à long terme occulte le fait qu'ils traversent actuellement un super-cycle d'investissement (les Capex massifs dans l'IA).
Le retournement de marge est inéluctable. Une fois que les besoins massifs d'infrastructure (les data centers) seront en grande partie satisfaits ou que la concurrence s'intensifiera, ces entreprises connaîtront mécaniquement un retour de balancier sur leurs volumes et leurs marges bénéficiaires exceptionnelles (qui atteignent parfois des sommets de $70% à $80%). C’est de la simple logique au final.
The Fed may be the weakest of your three amplifiers. In 1994 and 2018, both tech-led selloffs under a hiking Fed, the market bottomed in December and had recovered by mid-February once the Fed changed course, so policy ended up absorbing the shock rather than spreading it. A 20% capex cut would itself pull the hike odds down, since it removes the demand the hawks are leaning against. What would the Fed have to do in your scenario to become an amplifier, keep hiking into Q1?
Thanks David, fair challenge.
Yes, keep hiking into Q1, and that is already priced. after today’s CPI, swaps price hikes for December and one more by January 27, better than even odds of a second move.
I think that the différence with 1994 and 2018 is inflation. Core CPI was under 3% and falling in 1994, and core PCE was below 2% in late 2018, so the Fed could turn fast, within two weeks of the December 2018 low.
Today core CPI has printed 0.3% three months in a row, about 3.7% annualized, with energy passing into services. A capex cut hits earnings and credit spreads within weeks but inflation only with a lag, so the Fed keeps tightening while the shock spreads.
For me that looks more like 2022 in my view, when the Nasdaq lost a third of its value.
I imagine that even a pause may not help the long end. Late 2024 the Fed cut by a full point and the 10yyield rose almost as much.
The amplifier I'd add is where the money hides. In October investors already rotated into the old economy: US energy rose 13.9%, utilities 10.1% and financials 5.8%, while chips fell 15.4%. That shelter works while growth holds. If a capex cut tips growth over, no corner is spared, the shelter included. More on the rotation: https://floodlinesresearch.substack.com/p/a-short-trip-back-to-the-steam-age