
THE WIRE / FINANCIAL
Fed Governor Michael Barr says AI agents could turn bank runs into events that unfold in minutes, and AI debt could amplify a market drop
In a Nov. 16 speech, Barr said he is less worried about direct losses from the AI boom for now, but warned that leverage, opaque off-balance-sheet deals and autonomous AI agents could make a sharp repricing spread faster.
What happened
On Monday, Nov. 16, Federal Reserve Governor Michael S. Barr spoke at the Seventh Annual Association for Computing Machinery International Conference on Artificial Intelligence in Finance in Milan, Italy. The Fed published the full text the same day. Barr said it was the fourth speech he had given in the past year devoted entirely to AI, and he opened by calling himself an optimist about the technology's long-term promise.
Most of the speech walked through the Federal Reserve's latest Financial Stability Report, released the previous Friday, which focuses on the AI investment boom. Barr said on-balance-sheet debt at the seven largest AI firms has grown significantly in 2026 and is expected to exceed $800 billion before the end of the year. He said the report estimates that the potential maximum value of these firms' off-balance-sheet obligations, such as long-term contracts, data center and compute leases, energy supply contracts and residual value guarantees, now exceeds $3 trillion, and that this complexity can be opaque to investors.
He also described more than $100 billion in cumulative equity investments by hyperscalers and graphics chip makers in AI labs and cloud start-ups, a form of "circular financing" in which the smaller firms use the money to rent or buy chips and computing power from the companies that invested in them.
The line that drew the headlines came near the end. Barr said major banks now report finding AI agents attempting to take actions in retail customer accounts without the bank being able to verify authenticity, consent and alignment with customer instructions. He added that AI agents could "instantaneously detect yield differentials or stability signals and autonomously move funds at high speed, potentially accelerating runs that unfold in minutes," overwhelming traditional liquidity management and regulatory timelines. Yahoo Finance, Investing.com and CNBC TV18 reported the remarks.
Background and history
The Federal Reserve publishes its Financial Stability Report twice a year. Barr said this edition centres on what it calls "advanced computing debt" and on three ways it could affect financial stability: the rapid growth of that debt, circular financing networks, and other complex financial obligations. CNBC TV18 reported that 87% of the contacts surveyed for the report named AI as a salient risk.
Barr compared the debt-funded buildout with the telecom investment cycle of the late 1990s and early 2000s. He said the report compares hyperscalers' cumulative capital spending and debt issuance since 2020 with the whole telecom industry's during the broadband cycle that began in 1995, and that while the hyperscalers have not yet passed that mark, their recent acceleration suggests they may do so soon. He also cited an estimate that AI-related capital investment could rival or exceed earlier infrastructure booms in railroads in the 1870s, highways in the 1950s and 1960s, and fiber-optic cable in the 1990s, noting that some of those booms fizzled with consequences for the wider financial system.
He pointed to other open questions: the chips that make up most of a data center's cost may lose value quickly as newer models arrive, though lease rates for older chips appear to be holding up for now; and power and grid investments could sit idle for years if demand disappoints, as "dark fiber" did after the 1990s. Beyond lending, he warned that AI trading systems built on similar models could herd together, and that research shows some self-learning trading algorithms can develop collusive or manipulative behaviour without explicit intent.
Our analysis
This is a warning from a senior regulator about how stress could spread, not a report of stress happening now. Barr said that given the still-limited amount of debt behind the boom, he is less concerned about direct losses. His concern is the plumbing: with hedge-fund leverage and risk appetite "apparently quite high," a sharp reversal could trigger rapid deleveraging and "self-reinforcing fire sales," and banks that lend to leveraged equity strategies could see counterparty exposures rise. He said the high capital levels at the largest banks are a bulwark against spillovers.
The AI-agent point is new territory. Bank runs have sped up before when people could move money from a phone; Barr's argument is that software acting on its own, reacting to the same signals at once, could compress a run into minutes, faster than banks or regulators are set up to respond. He framed it as a reason for guardrails such as strict authentication, explicit consent for specific actions and technical limits on what agents can do.
Our scanner put this near the top because it comes straight from the Fed's own published text and was picked up by several outlets. We would place it a little lower for an ordinary reader: it describes possibilities, Barr himself allowed that there may be "no AI investment fizzle but a continued boom," and nothing in the speech points to an immediate threat to deposits.
What to watch next
No action is needed in the United States.
Signposts: any proposal from banking regulators on authentication and consent rules for AI agents acting in customer accounts; closer supervisory attention to AI-related debt, leases and collateral guarantees; and how markets handle this week's Nvidia earnings and Anthropic's market debut, which Yahoo Finance flagged as tests of investor sentiment. A sharp, sustained fall in AI-linked stocks alongside reports of forced selling by leveraged funds would be the scenario Barr described.
For readers: keep bank deposits within FDIC insurance limits, and be careful about giving any AI tool or agent the ability to move money in your accounts.
Confidence and rank
Our confidence is high. Quotes and figures come from the Federal Reserve's published speech text and matching coverage by Yahoo Finance, Investing.com and CNBC TV18; the risks described are possibilities, not forecasts.
Our read of the risk to a general reader is medium — A senior regulator's warning about possible future stress; no bank run or market break is happening, and Barr said big banks' capital is a bulwark.
How we rank
Ranks run from 0 to 10 and follow published scoring rules, including how widely a story reaches and how many independent sources report it. A high rank is not the same as a high risk to you. How we score →
More on The Wire