The Federal Reserve’s November 2026 Financial Stability Report, updated on November 13, puts artificial intelligence at the top of the risks cited by market contacts who were asked about threats to the U.S. financial system. The finding arrives as the AI infrastructure buildout draws more borrowing, complex leases, guarantees, and project financing. It does not mean the central bank has forecast an AI crash. It does show that the financing and security consequences of the AI boom have become a serious subject for financial-stability monitoring.
What the Fed survey actually found
In the report’s fall 2026 survey, 87% of respondents cited AI as a salient risk, compared with 50% in the spring survey. The fall survey was based on responses from 23 contacts gathered by Federal Reserve Bank of New York staff during September and October. The contacts included professionals at broker-dealers, banks, investment funds, and advisory firms. The report explicitly says the summary should not be interpreted as representing the views of the Federal Reserve Board or the New York Fed.
Respondents pointed to three broad AI-related concerns: elevated equity valuations that could fall if earnings disappoint, cybersecurity risks involving sophisticated attacks or unintended effects from frontier models, and complex financing structures tied to AI investment. The survey is a measure of what a small group of market professionals is watching, not a poll of every bank, investor, or business in the country.
The data-center financing numbers are getting larger
A separate section of the report examines debt issued to finance advanced-computing infrastructure. The Fed estimates that this issuance exceeded $680 billion by October 2026, about five times the $130 billion issued during all of 2023. Advanced-computing borrowing represented roughly 31% of nonfinancial investment-grade corporate debt issuance in 2026, up from less than 6% in 2023. The report estimates more than $1.1 trillion in outstanding public debt related to advanced computing as of October.
Debt is only part of the picture. Large technology companies are also making future lease commitments and providing guarantees or other contractual support for projects financed by smaller third parties. The report says undiscounted values of leases that have not yet commenced exceeded $1 trillion, while other obligations such as take-or-pay contracts and residual-value guarantees exceeded $2 trillion. These commitments are not the same as on-balance-sheet debt and should not simply be added to debt balances as if they were identical liabilities. They do, however, show why analysts are looking beyond headline capital-expenditure totals.
Why AI investment can become a financial-system issue
Data centers require expensive buildings, power connections, cooling, networking equipment, and advanced chips. Developers may use long-term customer contracts, leases, guarantees, project finance, and equity support to bring capacity online. Those structures can spread investment across several companies. They can also make it harder to see where losses would land if demand, utilization, or future AI revenue falls short of expectations.
The Fed notes that these interconnections could allow large technology firms to absorb some losses during a downturn, limiting wider damage. But they may also leave those firms with exposures that are larger than their on-balance-sheet liabilities suggest. A sudden slowdown in AI-related earnings or a sharp reassessment of asset valuations could affect lenders, suppliers, investors, and planned infrastructure projects at the same time.
What the report does—and does not—say
The report is a monitoring document, not a warning that a crisis is inevitable or an announcement of new AI regulation. Its general assessment reflects information available through October 28, 2026, while the report itself was published in November. The survey records concerns raised by 23 market contacts; it does not establish that 87% of the financial industry believes AI is dangerous. The distinction matters when interpreting a striking percentage.
For companies building or buying AI capacity, the practical takeaway is to assess more than model performance. Buyers and investors should understand who owns the infrastructure, how utilization is contracted, which party carries power and lease costs, how guarantees are structured, and whether expected demand supports the financing. The Fed’s report makes clear that the AI story is now also an infrastructure-credit story—and that transparency around those commitments matters to the wider economy.
