TL;DR
The Fed hiked into the AI capex boom — the first hike in three years, to 3.75%–4.00% — and instead of hitting "tech" as a block, it split the Mag 7 into four different trades. Meta found the story the market wasn't pricing (Muse), Alphabet became the first Mag 7 company to sell equity to fund its buildout, and Tesla finally shipped a new business line — seven years late. The correlated-basket trade is now contradicted by every cross-section: growth, margins, funding model, and momentum.
The rate shock landed, and the basket shattered along funding lines
The Warsh Fed's September 16 hike was the test the thesis needed, and the results were unambiguous: self-funding compounders absorbed it, externally-funded builders lagged. Apple closed at $341.07 at its 52-week high while Alphabet sat 14.5% below its high at a 17.27 P/E — cheapest in the basket despite 24.2% TTM revenue growth ai-slowdown-debate-rotation.
"Inflation remains elevated. Today's policy action will support a timelier return to the Committee's 2 percent goal. The Committee will deliver price stability." — FOMC statement, Sept 16, 2026
With the 10-year at 5.18% on September 24, the discount-rate regime lands hardest on the names funding capex from outside their own cash generation capital-allocation-and-free-cash-flow-divergence. Consensus targets now embed the dispersion: roughly 46% implied upside to Nvidia, 32% to Amazon, 25% to Alphabet — but negative upside for Apple and a Hold on Tesla ai-slowdown-debate-rotation
.
What to watch: Whether the 10-year holds above 5% — the funding cost of the negative-FCF tier is now the single most important macro variable for this group.
Alphabet is selling equity, not buying it
The capital-allocation inversion crossed a line this quarter: Alphabet raised external equity while running negative FCF. Berkshire Hathaway made a $10 billion private-placement purchase of Alphabet shares directly from the company in June, at a discount to trading prices, lifting its stake to roughly $38 billion capital-allocation-and-free-cash-flow-divergence.
"This private-placement purchase, coupled with additional open-market stock buys, increased the conglomerate's position in the tech giant by $17 billion to around $36.6 billion during the second quarter." — Motley Fool, Sept 27, 2026
The context: Alphabet ran −$5.86B in quarterly FCF with $44.92B of capex against $39.07B of operating cash flow, while Amazon posted the group's largest capex at $54.21B against $223B of total debt capital-allocation-and-free-cash-flow-divergence. A company that spent 2022–24 as one of history's largest buyback machines is now issuing discounted equity to part-fund AI capex — the exact inversion of shareholder-return orthodoxy, at the precise moment the Fed hikes into it.
What to watch: The Q3 print in late October — whether Berkshire-backed equity capital funds a capex raise, and whether buyback pace slows.
Meta found the pivot the market wasn't pricing — and it's aimed inside the basket
Muse, Meta's consumer AI agent launched September 8, shot to the top of Apple's App Store and reached roughly 700,000 U.S. daily users in under a month — and the stock re-rated +31.7% in a month on a Q2 EPS miss meta-muse-consumer-ai-agent.
"In the coming years, I expect that Muse is going to grow into the personal superintelligence that billions of people around the world are going to use to accomplish their goals and improve their lives." — CNBC
The collision is intra-basket: Amazon blocked Muse from making purchases on its marketplace (consistent with its Perplexity lawsuit and its blocking of OpenAI and Google agents), while Alphabet had a down week as Muse emerged as a potential rival search channel meta-muse-consumer-ai-agent. The Meta side repriced in weeks; the loser side — Alphabet's Search channel, Amazon's purchase funnel — is still repricing, and it's not in GOOGL's 17.3x multiple yet.
What to watch: Whether sell-side analysts start cutting Alphabet Search revenue estimates citing agentic competition — the market's repricing of the loser side of Muse.
Tesla shipped its first pivot product — seven years late
The Semi factory in Sparks, Nevada opened for high-volume production on September 24, with capacity of up to 50,000 trucks a year and deliveries to PepsiCo and DHL — but as Electrek put it, the program is "7 years behind schedule" tesla-existential-autonomy-capex-pivot.
"Tesla Semi finally enters volume production, 7 years behind schedule." — Electrek, Sept 25, 2026
The problem is what the Semi is: a capital-intensive, low-margin industrial leg — the opposite of the software-margin story a P/E of 344.55 depends on — funded by a core auto business running a 1.4% operating margin and −$1.09B quarterly FCF tesla-existential-autonomy-capex-pivot. Every new plant dollar now competes directly with the FSD/robotaxi narrative while the stock sits 24% below its high, below its 200-day average.
What to watch: Whether Musk's "very near future" self-driving teases convert to shipped product before the auto margin forces a capital rethink.
What surprised us
- The hike helped some AI names. The rate shock that was supposed to hit "AI capex" as a block instead lifted Apple to its 52-week high and Microsoft +34.6% over three months, while the externally-funded builders lagged — funding model, not AI exposure, was the differentiator ai-slowdown-debate-rotation
.
- Apple ignored a record $5.7B patent verdict. A jury found its Taptic Engine infringed two Taction haptics patents on September 26 — the largest US patent award on record — and the stock, at its high, simply didn't care ai-slowdown-debate-rotation
- Berkshire's $10B placement is closer to a rescue-era deal than a portfolio buy. The Fool notes Berkshire "never made a direct equity investment in Apple" — Greg Abel underwrote Alphabet's capex program itself, not just the stock capital-allocation-and-free-cash-flow-divergence
- Nvidia has a China catalyst nobody's trading. China's MIIT signaled intent to approve RTX Pro 5500 purchases by ByteDance and Alibaba — ByteDance reportedly weighing ~1 million units — yet the stock is roughly flat over a month ai-slowdown-debate-rotation