AdvertiseFee Deal Desk

Macro

Fed Puts Large-Bank AI Exposure Near $1 Trillion in Stability Report

The Federal Reserve estimates that large US banks' financial exposure to the artificial intelligence industry stands at nearly $1 trillion, and it warns…

Be a creator
November 17, 2026, 04:07 AM UTC5 min read
AI SummaryAI
  • The Federal Reserve estimated large US banks' AI-related financial exposure at nearly $1 trillion in its November 13 report.
  • Seven technology firms including NVIDIA, Microsoft and Oracle carry over $3 trillion in leases, contracts and guarantees.
  • Federal Reserve survey respondents naming AI a major financial risk jumped to 87% from 50% six months earlier.
  • The New York Fed found three of four bank vulnerability indicators rising in its November 12 analysis.

Fed Estimates Bank AI Exposure Near $1 Trillion

The Federal Reserve estimates that large US banks' financial exposure to the artificial intelligence industry stands at nearly $1 trillion, and it warns that losses at technology firms could spread to banks and asset managers if returns on AI investment fall short of expectations. The estimate appears in the Fed's semiannual Financial Stability Report, published on November 13, and marks the first time the central bank has treated AI infrastructure borrowing and its financial risks as a stand-alone item.

The exposure figure covers more than direct lending. It includes financing extended to technology companies, data center developers, special-purpose vehicles tied to AI infrastructure, private credit managers and power suppliers. The Fed notes the number represents exposure, not booked losses, and that the risk is also assessed in the central bank's annual bank stress tests. Borrowing has scaled quickly: debt issued between January and October to fund data centers and advanced computing infrastructure exceeded $680 billion, roughly 5.2 times the $130 billion raised across all of 2023. Outstanding public bonds tied to advanced computing topped $1.1 trillion as of October, and AI-related borrowing climbed from under 6% of investment-grade corporate issuance in 2023 to about 31% this year.

Off-balance-sheet commitments grew even faster. Seven major technology firms, including NVIDIA (NVDA), Microsoft and Oracle, carry more than $3 trillion in future lease payments, long-term purchase contracts and financial guarantees. Leases that have not yet started exceed $1 trillion, and long-term contracts plus residual value guarantees top $2 trillion. Such obligations do not appear in full on corporate balance sheets today, but they would weigh on earnings if AI demand disappoints. Financing structures have also shifted: where large tech firms once built data centers with their own funds or corporate bonds, projects now draw more on private credit, project finance and structured vehicles. Concern inside the system is measurable. In the Fed's survey of financial stability contacts, 87% of respondents named AI a major financial risk, up from 50% six months earlier, while 52% flagged the possibility of broad declines in risk assets.

New York Fed Flags Rising Bank Vulnerability

Separate work on the banking system itself points in the same direction. A New York Federal Reserve Bank analysis published on November 12 found that three of the four indicators measuring US bank vulnerability are on a rising trend, covering capital shortfalls, liquidity strain, deposit outflows and the risk of forced asset sales. The New York Fed judged current vulnerability below the levels seen just before the 2008-2009 crisis, though above the low readings recorded between 2015 and 2020. Under a repeat of the sharp 2022 rate-rise shock, it estimated banks would need about $151.1 billion of additional capital to hold their ratios at target.

The Fed's report also describes how the credit chain forms: a cloud company signs a long-term lease with a data center developer, the developer borrows against that contract, and a chipmaker guarantees the residual value of customer equipment. When demand matches forecasts, the structure works and every party earns its return. When returns fall short, losses travel along the same links to banks, private credit funds and insurers. The Fed added that large technology firms are deepening financial ties with smaller, lower-rated counterparties, and that the risk these firms actually carry may exceed the debt shown on their statements.

Fed Chair Kevin Warsh added an official warning on November 16. Speaking at the opening of the productivity and jobs task force conference in San Francisco, he said the central bank has begun scenario analysis of AI's economic impact and that recognizing the implications of so-called superintelligence, the term the Trump administration uses for the technology, is a Fed responsibility. Warsh laid out both paths. If AI delivers less than expected, returns on large capital spending fall, equity values decline, and slower growth pushes unemployment higher. If productivity improves decisively, the economy could sustain above-average growth while staying near full employment. He cautioned that the scenarios are possibilities to be studied, not an official economic forecast. One timing caveat matters: the report draws on information gathered through October 28, before NVIDIA's leadership change, so the uncertainty that followed is not reflected in the assessment.

The $1 trillion headline understates where the risk now sits. Booked bank exposure is large, but the $3 trillion in lease, contract and guarantee obligations is growing faster and sits mostly off balance sheet, so reported corporate debt captures only part of the AI build-out's financing. The sharpest single datapoint in the report is the sentiment shift: a 37-point jump in six months among respondents naming AI a top financial risk. For markets pricing risk appetite across assets, from credit spreads to crypto derivatives funding rate signals, the figures to watch are the ones that convert into cash calls: the $151.1 billion capital estimate under a rate shock, and the lease obligations that come due as real AI demand is tested.

COINOTAG's editorial and research desk.

AI-Assisted

AI-generated, AI-reviewed, under COINOTAG editorial oversight.

Price-Impacting News

Live
All breaking news
Breaking News