No Fed Put: November 5's 5.5% Nasdaq Rout Prices What Warsh's October Hold Does to AI Capex

There was no November FOMC meeting — the 2026 calendar runs from October 27–28 straight to December 8–9. The decision that reset the AI rate debate came on October 28, when the Committee held the funds rate at 3.75%–4.00%, upgraded activity to 'expanding at a strong pace', and left inflation 'elevated'. Two dissents for a quarter-point hike turned that hold into the committee's own signal that December is live.

The projections make December consequential. September's medians put the 2026 year-end funds rate at 4.1% — above the current range — with 2026 PCE inflation at 3.7% and 2026 growth at 2.3%. Investors already expect a December increase, and Warsh's own path to 'a timely return of inflation to the 2 percent goal' runs through higher rates, not lower ones. The committee's own median now points to a December hike rather than a rescue.

Democrats flipped seven Senate seats on November 3 — in Iowa, Kansas, Texas, South Carolina, Ohio, North Carolina and Georgia — and are heading into the new Congress with a House majority above 250. For the AI trade this cut two ways at once: politics now raises the odds of data-centre moratoria while it drains the White House's push for lower rates. Reuters reported that candidates who campaigned against new data-centre development won in Texas and Georgia, and the Financial Times described the session as a 'blue wave' that upended AI bets.

A 25 basis point decision now moves the AI complex because the AI complex was financed with debt. Goldman Sachs Research estimates nearly $500 billion of AI-related debt issuance in 2026, with roughly a third of hyperscaler capex debt-financed this year and 35% expected in 2027. Debt-financed capex converts a policy debate into a credit-spread debate — and credit spreads are set by lenders, not by the FOMC.

Removing the Fed put sorts the AI complex by balance sheet, not by AI exposure. Over the trailing twelve months, Microsoft generated $60.0 billion of free cash flow while spending $135.9 billion on capex and repaying $7.9 billion of debt; Meta produced $28.7 billion and Alphabet $19.7 billion. Amazon's numbers are the outlier: it ran a $46.9 billion free cash flow deficit and issued $124.6 billion of debt to cover $215.2 billion of capex.

Oracle's numbers are the hinge, because its demand is not the problem. The company reported first-quarter fiscal 2027 revenue of $19.3 billion, up 30%, cloud infrastructure revenue of $7.4 billion, up 121%, and remaining performance obligations of $664 billion — while free cash flow was negative $5 billion. Its trailing interest expense of $5.10 billion already consumes 22% of operating income, and its $155.9 billion of debt sits against $37.1 billion of cash and short-term investments. A 25 basis point move — which Warsh's own projections make more likely — would add roughly $390 million a year to Oracle's interest bill if the full debt stack repriced.

Upstream is where the AI cycle's profits actually accrued, and it is where the first crack appeared. SK hynix reported third-quarter revenue of 104.6 trillion won, up 328% year over year, with operating profit of 83.3 trillion won; Micron's fiscal fourth quarter delivered revenue of $54.23 billion against $11.32 billion a year earlier. Earning roughly four-fifths of revenue as operating profit is the most expensive place in the chain to be wrong about demand.

With the Fed on hold and its own projections pointing to 4.1% at year-end, the marginal AI dollar has to be underwritten rather than regulated. JPMorgan Chase's third-quarter results showed $15.3 billion of net income on $50.2 billion of revenue, $5.1 trillion of total assets and a 13.9% standardized CET1 ratio, and the firm's own research puts 2026 hyperscaler capex at $697 billion. Whoever writes the loan absorbs the difference between a 4% policy rate and a 3% one.

The transmission is already visible in the least-financeable corner of the complex. CoreWeave's shares fell more than 9% on November 4 and its bonds tumbled, with the market explicitly attributing the move to the election result; the company carries $45.2 billion of net debt against $5.0 billion of stockholders' equity, and trailing interest expense of $1.87 billion against $3.94 billion of EBITDA. When the equity is thin, a wider spread becomes an equity problem rather than a financing inconvenience.

Over one to three years the outcome is binary, and the difference is who funds the last tranche. Goldman Sachs expects 35% of hyperscaler capex to be debt-financed in 2027, alongside roughly $300 billion of project finance and data-centre transactions — which means the next twelve months of AI capex are already committed and the question is who absorbs the refinancing. A cycle funded at 4% by levered balance sheets ends earlier than a cycle funded by cash flow, and trailing operating margins near 75% at SK hynix and a 87% gross margin at Micron are what a cycle nearing its end looks like.

The investable conclusion is narrow and testable. If AI capex is financed by cash flow, Warsh's hawkishness is irrelevant to the buildout and the trade is purely about demand. If it is financed by debt, the Fed's own projection of 4.1% at year-end is a direct tax on that buildout, and November 5 was the market noticing. Own the balance sheets that can fund the boom, and be careful with the ones that need the boom to keep going.