Governor Barr on AI’s Potential Risks to Financial Stability

8 Min Read

Federal Reserve staff are flagging a financing question that could matter as much as the AI spending story itself. The latest Financial Stability Report, issued in November, examines how the AI infrastructure buildout is being funded, and its findings point to a growing reliance on debt, including obligations that sit off company balance sheets. For traders, the message is that a sharp repricing of AI-related assets could travel through leverage, collateral and counterparty channels rather than stay confined to a single sector.

Debt, Leverage and the Shift in How AI Gets Financed

The starting point is valuation. The report notes that the equity premium, a rough gauge of the extra return investors demand for holding stocks instead of risk-free bonds, sits at low levels relative to history. It is not as stretched as during the dot-com boom, but it leaves less cushion if expectations about AI returns fall short. The Federal Reserve is explicit that policymakers do not target market price levels. Its concern is what a repricing would do to the real economy, through lower investment and through wealth effects that weigh on consumer and business spending.

The bigger shift is in funding. On-balance-sheet debt at the seven largest AI firms has been rising sharply during 2026 and is expected to surpass $800 billion before year-end. The report draws a comparison with the telecom investment cycle of the late 1990s and early 2000s, although hyperscaler spending and debt issuance since 2020 have not yet exceeded the earlier broadband comparators. The recent acceleration in capital expenditure suggests that gap could close quickly.

Off-balance-sheet obligations are the more unusual feature. Long-term contracts, lease commitments and residual value guarantees have grown faster than reported debt, and the report estimates the potential maximum nominal value of these liabilities at more than $3 trillion. That figure is not discounted, which makes it hard to compare directly with on-balance-sheet debt. Strong-credit hyperscalers are also indirectly supporting projects involving weaker borrowers, and lower-credit firms that depend most on outside funding have seen their cost of capital climb significantly in recent months. Those firms are turning to asset-backed financing, which adds another layer of exposure.

Circular financing deepens the link. Hyperscalers and GPU producers have committed more than $100 billion in cumulative equity to large language model labs and cloud start-ups. Those smaller firms often use the money to buy compute from their investors. The sums are small relative to the investors’ balance sheets but critical to the recipients. Collateral backstops add another dependency: larger firms guarantee resale values for the data centers and chips pledged against asset-backed loans. The report treats those guarantees as off-balance-sheet liabilities because they show how much losses the stronger firms could absorb in a default and forced liquidation.

Cross-Asset Transmission: Equities, Collateral Values and Dealer Exposure

The source material contains no live prices, index levels or chart data, so there are no technical levels to map here. What it does provide is a transmission map. A reversal in AI-linked equities would reach bond and credit markets through the financing structures described above. The equity premium is defined relative to risk-free bonds, which means a sharp equity selloff could change how investors compare stocks with Treasuries, even though the report does not forecast any yield move.

Leverage is the amplifier. The report describes hedge fund leverage and broader risk appetite as high, and it warns that a sharp reversal could trigger rapid deleveraging and potentially self-reinforcing fire sales. Equity strategies have also increased their use of leverage, partly through prime brokerage services offered by large dealers. Capital at the largest banks is described as a bulwark against spillovers, but counterparty exposures could rise during volatility. Positions in AI-related names are therefore worth tracking as dealer exposures, not only as sector exposures.

Collateral quality is the second channel. Much of a data center’s cost sits in GPUs and specialized chips, which can lose economic value as newer generations arrive. The report notes that lease rates for older chips are currently holding up, which suggests some residual demand. It also flags that other parts of the build, such as substations, high-voltage connections and grid interconnection, may take years to find use if demand disappoints. Past infrastructure booms show that excess capacity can sit idle for long periods, so asset values tied to physical capacity could be slower to adjust but harder to recover.

Market structure adds a sentiment layer. The report highlights herding risk if many traders rely on similar AI models trained on similar data, and it cites theoretical research on algorithmic collusion and spoofing. None of that is presented as a current market fact, but it helps explain why a sudden shift in sentiment could produce synchronized reactions.

Trader Checklist: What Would Change the Read

The base case from this report is not a call for a crash. It is a warning that financing has become a transmission channel. Next Move Markets reads the material as a reason to watch balance sheets, collateral and dealer exposures alongside earnings and capex announcements. Several developments would challenge this interpretation. Stronger-than-expected returns on AI investment would ease the repricing concern. Stabilizing funding costs for lower-credit firms would suggest the asset-backed shift is not under strain. Evidence that older chips keep generating lease income would support collateral values. A widening gap between reported debt and off-balance-sheet obligations would point the other way.

  • Track the growth of on-balance-sheet debt at the largest AI firms against the expected $800 billion level, and watch whether capital expenditure keeps accelerating.
  • Monitor funding costs and issuance for lower-credit firms moving into asset-backed financing, since rising costs there would signal stress earlier than headline equity moves.
  • Watch the equity premium and dealer leverage data for signs that risk appetite is stretching further, particularly in prime brokerage exposures tied to AI names.
  • Track lease rates for older GPU generations and any changes in guarantees on data center collateral, as weakening resale values would challenge the assumption that the buildout’s assets hold their worth.

Editorial note: This article is market intelligence for educational purposes and is not investment advice.

Source: FRB: Speeches (2026-11-16 10:00:00). Prepared by Next Move Markets from the cited source.

Published by the Global Research Desk under the editorial responsibility of Youssef Elhouari, Founder & Editor-in-Chief. For information and education only, not investment advice. See our editorial policy.

Share This Article
The Next Move Markets Global Research Desk comprises market analysts and financial editors specializing in macroeconomic drivers, central bank policy (Fed, ECB, BOE, BOJ), forex technical analysis, energy markets, and global equity developments. The team delivers real-time market insights and educational analysis for active market participants.
Leave a Comment
Join us on Telegram