Featured image: the Marriner S. Eccles Federal Reserve Board Building in Washington, in an August 2008 file photo. Photo: AgnosticPreachersKid, Wikimedia Commons, CC BY-SA 3.0. Cropped to 16:9.
Federal Reserve Governor Michael S. Barr said Monday, Nov. 16, 2026, that a sharp reversal in the AI investment boom could lead to rapid deleveraging and “self-reinforcing fire sales.” He spoke in Milan at the Association for Computing Machinery’s conference on AI in finance, in remarks the Fed published the same day.
Barr pointed to Friday’s Financial Stability Report and then went further, describing how leverage, circular financing and off-balance-sheet obligations could amplify a repricing. He also warned that AI systems used in trading and in retail banking could create new ways for stress to spread, including faster deposit outflows.
The survey shares are from the Federal Reserve’s November 2026 Financial Stability Report. The debt figures are from Barr’s Nov. 16 speech, which cites that report. Off-balance-sheet amounts are undiscounted, which Barr noted complicates comparisons with on-balance-sheet debt.
What Barr said about an AI investment reversal
Barr opened with a disclaimer that Fed policymakers “do not target levels of financial markets” and that he would not comment on short-term ups and downs. He said the latest report instead looks at longer-term Federal Reserve questions about financial stability, including whether investors will see returns “on a timeline and to an extent consistent with their expectations.”
In one scenario he has used before, even if AI later proves transformative, a mismatch between those expectations and actual returns could still produce a large repricing, he said. The potential investment losses, he added, could lower output growth in two ways: from a direct drop in investment, and from the effect of lost wealth on consumer and business spending.
He said the equity premium — a rough measure of the extra return investors require for holding stocks rather than risk-free bonds — is “at low levels relative to historical norms, though not nearly as low as the levels reached during the dot-com stock market boom.”
Barr’s warning about fire sales was tied to leverage outside the core AI firms themselves. “Given the still-limited amount of debt backing the current boom, I’m less concerned about direct losses,” he said. “However, with leverage at hedge funds and broader risk appetite apparently quite high, a sharp reversal could lead to rapid deleveraging and, potentially, self-reinforcing fire sales.” He said some of that equity-strategy leverage is provided through the prime brokerage services of large dealers, and that counterparty exposures could rise in a period of equity-market volatility even though high capital at the largest banks is a “bulwark against spillovers.”
By the numbers: the Fed survey Barr recapped
Barr did not repeat the survey percentages in the published text of the speech, but he told the audience that Friday’s Financial Stability Report remains a main source on AI and financial stability. The Boston Report covered that report when it was released, including the finding that AI had become the risk cited most often by market contacts.
In the report’s overview graphic, 87% of contacts in the September and October Survey of Salient Risks cited artificial intelligence, up from 50% in the spring 2026 survey. The next most-cited items in the fall results were a risk-asset correction, at 52%; higher long-term rates, at 48%; and cyberattacks and fiscal debt sustainability, each at 43%. In the spring survey, those four items were cited by 35%, 20%, 20% and 30% of contacts, respectively. A Boston Report calculation from the two AI figures is a 37-percentage-point increase, or 74% more than the spring share (37 divided by 50).

Those shares measure how often contacts named a risk, not the probability that any one shock will occur. The report said respondents pointed to three AI-related worries: elevated equity valuations and the possibility that weaker-than-expected earnings could lead to a market correction; cybersecurity risk from sophisticated attackers or from frontier models themselves; and risks from complex AI-related financing.
The $800 billion and $3 trillion figures
Barr said on-balance-sheet debt of the seven largest AI firms “has been growing significantly in 2026 and is expected to exceed $800 billion before year’s end.” He compared that growth to the telecom investment cycle of the late 1990s and early 2000s, and said that while hyperscalers have not yet surpassed that earlier cycle on the report’s comparison of cumulative capital spending and debt issuance, “the recent acceleration in their capital expenditures suggests they may do so soon.”
The same speech, citing the report, said other obligations — long-term contracts, lease commitments and residual-value guarantees — “have grown even more rapidly.” The report estimates that the potential maximum nominal value of those off-balance-sheet liabilities “now exceeds $3 trillion,” Barr said. In a footnote he cautioned that firms generally present such liabilities before discounting, “which complicates comparisons to on-balance-sheet numbers, which are discounted.”
That $3 trillion figure is not the same as the report’s separate estimate that outstanding public debt related to advanced computing was more than $1.1 trillion as of October 2026, or its finding that advanced computing-related debt issuance had risen to more than $680 billion by October, about five times the $130 billion issued in all of 2023. Those are different measures, from the report’s data-center box, and they should not be added together.
Barr also said hyperscalers and large graphics-processor producers have “over $100 billion in cumulative equity investments in large language model labs and cloud-computing start-ups.” He described those stakes as a form of circular financing, in which the larger firms invest in smaller ones that then rent or buy computing from the investor. The IMF’s managing director, in remarks the Boston Report reported separately, has also flagged circular AI financing; Barr’s figure refers specifically to those cumulative equity investments, which he said are a relatively small share of the larger firms’ balance sheets but are “critical for the smaller firms receiving the investments.”
Trading, cyber risk and faster bank runs
Beyond the investment boom, Barr turned to how generative AI could change wholesale trading. He said it is his understanding that generative AI “is not yet being used for autonomous trading at large scale,” but that similar models trained on similar data could produce herding, synchronized reactions to news, and, if they share a flaw, failures “with systemic consequences.” He cited research that some reinforcement-learning algorithms can develop collusive or manipulative behavior, including spoofing, without explicit intent.
On cybersecurity, he said incidents this year have shown models treating security boundaries “as obstacles to overcome rather than rules to be followed.” Defenders are also using AI, he said, but “attackers only need to be right once, while defenders need to be right all the time.”
In consumer finance, 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 “may accelerate deposit outflows during stress periods,” because they could “instantaneously detect yield differentials or stability signals and autonomously move funds at high speed, potentially accelerating runs that unfold in minutes.”
The speech is part of a stretch of business news about how regulators describe the AI buildout. Barr said he remains “an optimist about the long-term promise of AI,” but that AI can strengthen the financial system “only if it also supports financial stability.” Readers can follow continuing coverage in our AI news center.
Sources
- Federal Reserve: Speech by Governor Michael S. Barr, “Some Potential Financial Stability Implications of Artificial Intelligence,” Nov. 16, 2026, Milan
- Federal Reserve: Financial Stability Report, November 2026 (PDF), including the Survey of Salient Risks graphic and Box 2.2, “Leverage in the Data Center Buildout”
- Reuters: “AI rockets to top of worries about financial system in latest Fed survey,” Nov. 13, 2026
- CNBC: “AI risks are biggest threat to the financial system, Fed survey finds,” Nov. 13, 2026
Read next in Artificial Intelligence
Artificial IntelligenceAlibaba Releases Qwen4-Max, Pricing It at $1.50 per Million Input Tokens in Singapore
Artificial IntelligenceU.S. and Nine Partners Sign San Francisco Declaration to Safeguard AI Supply Chains
Artificial IntelligenceGoogle Launches Gemini for Healthcare, With a Massachusetts Insurer Among Early Users
Most read this week
- Finland Confirms Strike on Russian Warship in Gulf of Finland; Ukraine’s SBU Claims Drone Attack
- U.S. Envoys Witkoff and Kushner Hold Talks With Putin in Moscow, Then Head to Dublin to Brief Zelensky
- Israeli Strike on Gaza’s Nasser Hospital Kills Six, Including a Surgeon, Health Ministry Says
- Nvidia Board Hires Wachtell’s Joon Kim to Lead Huang Probe as Shares Fall 5.47% and an Investor Sues
- Drones Kill Workers at Zaporizhzhia Nuclear and Thermal Plants, IAEA Says, as Russian Strikes Hit 5 Substations
Latest
- Mamdani Disbands NYPD’s Strategic Response Group, Shifts Protest Response to a Civilian Unit
- A Sudan Army Drone Strike Killed 15, Including 4 Children, a Survivor and a Monitor Say
- Russia Says It Intercepted 1,336 Ukrainian Drones as Crimea Officials Halt Kerch Bridge Traffic
- Apple Opens a 200,000-Person U.S. Study of How People Use Health AI
- Katherine Clark, Poised to Be the First Woman House Majority Leader, Lays Out Democrats’ Plans
Get The Boston Report Briefing
Our free weekly email with the most important stories from Boston and the world. Unsubscribe anytime.
Free. We never sell your address. Privacy policy