The AI-Slowdown Debate Is Splitting the Basket: Warsh Fed Hikes Into AI Capex, Dispersion Widens

The AI-Slowdown Debate Is Splitting the Basket: Warsh Fed Hikes Into AI Capex, Dispersion Widens

The Fed delivered the hike markets had priced: on September 16, 2026 the FOMC voted 12–0 to raise the federal funds target range 25bp to 3.75%–4.00% — the first hike in three years. The 10-year Treasury has since pushed to a 24-year high: "The 10-year Treasury yield just hit a 24-year high above 5.36%, up from 4.75% in August" (CNBC, Oct 7; the FRED daily print was 5.22% on Oct 8), with core PCE at 3% in August and CNBC noting "a few more rate hikes may be needed." Into that backdrop, the AI-slowdown debate got its sharpest datapoint — and its sharpest counter-datapoint — in the same week.

The Oct 8 OpenAI revenue scare — the demand signal under the whole complex was grossed up

The Financial Times reported Oct 8 that OpenAI told investors its September annualized revenue was "approaching $50 billion at the end of September, rather than the $70 billion previously reported." Per Yahoo/Benzinga: "AI stocks fell Thursday as new figures put OpenAI's annualized revenue roughly $20 billion below previous reports, raising questions about the sector's growth." The Nasdaq fell as much as 1.4% intraday; Oracle dropped more than 5%, with NVDA, AMD and MSFT all lower — and AMD fell 6.36% over three sessions (see /markets/META/2026/10/12 news feed). Bloomberg's Lisa Abramowicz "pointed to the market reaction as evidence of how dependent the technology sector has become on OpenAI and rival Anthropic."

The mechanics matter as much as the miss. Reuters: "The discrepancy mainly arose from an attempt to produce a direct comparison with figures from rival Anthropic," and "OpenAI, unlike Anthropic, does not include, for instance, revenue from sales via cloud partners such as Amazon's AWS and Alphabet's Google Cloud." Axios sourced the accounting: "The $70 billion ARR figure for OpenAI was an attempt to 'gross up' its revenue number based on an interest from investors in having an apples-to-apples figure using Anthropic's methodology, sources said" — Anthropic records the full $100 of a partner-channel sale as revenue and the cloud provider's cut as an expense, while OpenAI records only its share. Reuters adds the dependence math: "Anthropic pays cloud partners about 16% of every dollar earned through them, which last year accounted for half of revenue, a Reuters analysis showed."

The repricing ripple: OpenAI is seeking fresh funding "of at least $30 billion at a reported $1.4 trillion valuation. That represents roughly 28 times its newly reported annualized revenue, compared with 20 times the previous figure." Polymarket odds of an OpenAI IPO by end-2027 closing at ≥$1.5T fell from ~80% in September to ~58%, with no-IPO-by-2027 at ~23%. The WSJ separately reported Broadcom "has been working to arrange more than $50 billion in financing for custom chips developed with OpenAI" — the vendor-financed demand chain now has a visible seam.

Yet the revisions into Q3 prints are still going UP

The same week, CNBC's earnings-season preview (Oct 7) showed consensus moving the other way: "Analysts expect nearly 30% year-over-year earnings growth for the S&P 500, up from 26.7% on June 30th," and "Estimated EPS growth for the tech sector has increased to 65% today from 57% on June 30, thanks partly to upward revisions for Nvidia and Micron Technology" (FactSet). Micron "delivered a knockout quarter." Meta's Muse launch "has set off an arms race to capture AI-enabled consumer e-commerce."

The most important divergence datapoint in that preview is a crossover: "While those companies are expected to notch an average 20% growth, the other 493 stocks in the S&P 500 are forecast to deliver 27% year-over-year gains, according to Russell Investments.¹" The Mag-7 are no longer the index's growth leaders — the rest of the market is expected to outgrow them this earnings season, even as the basket still accounts for the bulk of the tech sector's 65% EPS-growth estimate.

What it means

The scare demonstrated the correlation is still real in drawdowns — the whole complex (NVDA, ORCL, AVGO, MSFT, AMD) sold together on one private-company disclosure. But the fundamentals underneath are diverging three ways: (1) the demand metrics the AI trade was priced on are method-dependent gross-ups, now being corrected in public; (2) consensus revisions are still rising into the Q3 prints, so the debate gets its test over the last two weeks of October; (3) the basket's growth premium over the rest of the market has inverted (20% vs 27%), while within the basket the spread is extreme — MSFT +33.7% over 3 months at 0.5% from its high vs GOOGL -0.7% over 3 months and 12.5% below its high, and TSLA 21.9% below its high (/markets/MSFT/2026/10/12, /markets/GOOGL/2026/10/12, /markets/TSLA/2026/10/12). The paper-gain quality problem compounds this — see The Paper-Gain Ledger: Alphabet's $98B SpaceX Mark and Amazon's $53B Anthropic Mark — Now Facing IPO Price Discovery — and the capital competing for the same cash is in Capital Allocation and FCF Divergence: Free Cash Flow Squeeze vs. Shareholder Returns — and Alphabet Now Selling Equity.