Almost Every Fed Official Discussed the Risk of an AI Bust, Minutes Show
This post may contain links from our sponsors and affiliates, and Flywheel Publishing may receive compensation for actions taken through them.
At 2:00 p.m. ET on November 18, 2026, the Federal Reserve released minutes of its October 27-28[1] meeting. Artificial intelligence took over the financial-stability discussion. Almost all participants commented on risks of a scenario in which returns on AI investment disappoint investors.
That scope stands out because it is rare for nearly every official to raise the same scenario in a stability review, and the Fed tracks scenarios like this one because a sudden repricing of assets can reach the lenders behind them. Policy did not change. The committee held the federal funds rate at 3.75% to 4.00%[1] by a 10-2 vote, and Lisa Cook and Beth Hammack dissented in favor of a quarter-point rise.
On the same day, NVIDIA (NASDAQ:NVDA | NVDA Price Prediction) fell 2.89% to $150.96, its lowest close since June 24, 2025. That close left it with a market value of $3.645 trillion. The Invesco QQQ Trust (NASDAQ:QQQ) rose 0.39% to $659.51. No source links either move to the minutes, but the AI bellwether fell while the broad Nasdaq-100 fund rose.
What the Minutes Say
The minutes record a discussion of a risk scenario. Most participants noted that AI-related debt had grown rapidly, much of it through complex, opaque financing structures. A few said AI exposures were becoming concentrated among a small number of firms.
Fed staff said off-balance-sheet leasing commitments and circular financing arrangements among AI-related firms could amplify losses if returns disappoint. In circular financing, money moves in a loop between suppliers, investors, and customers. An off-balance-sheet lease keeps a long-term payment out of reported debt, so a company can carry more leverage than its balance sheet shows. The minutes describe these risks in general terms.
How Losses Could Travel
The New York Fed’s markets desk said risk sentiment worsened after two leading AI firms reported revenue growth below expectations and a hyperscaler’s spending plans needed more financing than investors expected.
High-yield bonds sold to fund data center construction saw an especially notable widening in spreads. A spread is the extra yield investors demand over Treasuries.
Several participants were concerned that supervisors may miss banks’ indirect exposure through credit lines to private credit funds.
Read-Through for QQQ and NVIDIA
The concentration warning applies directly to QQQ. As of June 30, 2026[2], NVIDIA was its largest holding at about 7.6% of net assets. Apple (NASDAQ:AAPL), Micron Technology (NASDAQ:MU) and Microsoft (NASDAQ:MSFT) came next. Over the past month, the fund lost 8.19%, while NVIDIA lost 30.46%.
NVIDIA’s latest quarter brought in $115.84 billion[3] of revenue, up 103.2%, yet its filing carries the exposures the Fed described: $390 billion in supply commitments and guarantees of up to $118 billion that would bite if returns disappoint, plus three customers at 14%, 11% and 11% of revenue.
How AI Spending Could Shape Rates
The rate effect works in both directions. Many called the AI buildout a material source of inflation risk, but a number said closer scrutiny of AI returns and financing could slow AI spending and ease inflation. A slowdown that hurts NVIDIA could help the rate-sensitive stocks inside QQQ.
Interim CEO Colette Kress said on the November 17 earnings call that “demand continues to meaningfully outstrip supply.”[4] That assessment comes from the seller of the chips, so it likely carries less weight than the Fed’s warning for investors considering the risk.
Where AI Stocks Stand After the Fed Minutes
The bull case rests on cash. NVIDIA trades at about 10 times forward earnings, and the average analyst target is $265.23.
The bear case is that this cash depends on financing the Fed now calls opaque and concentrated. If returns disappoint, those commitments and guarantees turn into losses. Riding a mania is fine as long as you plan the exit, which is the whole subject of our free bubble survivor’s handbook.
The AI theme still holds up better than the minutes mean, because revenue is still running ahead of the debt the Fed is worried about. That view fails if data center bond spreads keep widening before the December 8-9 meeting, or if NVIDIA closes below its 52-week low of $148.80.
Contact [email protected] for any questions or corrections.








