The Bank of England has documented a hedge-fund-leveraged AI bubble, named the mechanism, and chosen to publish a report about it.
The Bank of England’s Financial Policy Committee published its twice-yearly Financial Stability Report on Tuesday. The FPC named the mechanism. Hedge fund leverage in equity markets has risen “significantly,” concentrated in AI-related companies, transmitted through prime brokers — the banks that intermediate the leveraged trades. The FPC named the cross-asset channel. Hedge funds also hold government bonds; an equity unwind forces bond selling, broadening the impact into the gilt market. The FPC named the catalyst. “Historically unprecedented levels of investment” in AI could be reassessed; concentrated, momentum-driven positions amplify the volatility; the leverage accelerates the decline.
The FPC has the macroprudential tools. The countercyclical capital buffer — extra bank capital that builds up in good times and releases in bad — can be raised. The leverage ratio framework — the rule that limits how much banks can borrow against their equity — can be tightened. Sectoral capital requirements — extra capital banks must hold against loans to specific sectors — can be applied to bank exposures to highly leveraged funds. The Prudential Regulation Authority, the Bank’s bank-supervision arm, can be directed to act. The FPC has used the CCyB in recent years, raising and holding the buffer as conditions required. The tools are not theoretical. The tools are not absent. The tools are in the drawer. The drawer is unlocked. The FPC chose not to open it.
The friction is real. Treasury coordination, bank industry resistance, the lag between action and effect — these are constraints on the FPC. They are not prohibitions. The “preparing changes” language is the tell. The FPC said it is “preparing changes to the rules that govern how much capital U.K. banks must set aside against potential losses.” “Preparing” is a process word. The reforms will address “unintended consequences in the leverage framework” — the leverage framework that is currently transmitting the hedge fund build-up into the bank balance sheets the FPC is supposed to oversee. “Strengthen the releasability and usability of buffers” is the language of a committee adding tools to a toolkit it has decided not to deploy.
The cross-asset contagion the FPC names is the one that surfaced in the 2022 gilt dysfunction, when leveraged liability-driven-investment funds were forced to sell gilts into a falling market. The same architecture is in place. Today, bond markets are again sending warning signals as yields jump and stocks slump, and the exact same architecture is in place. Hedge funds “which also invest in government bonds could face pressure to sell those assets, broadening the impact of the setback.” A leveraged unwind in AI-related equities forces selling in the UK gilt market. The FPC is identifying, in its own document, the mechanism by which a hedge fund deleveraging transmits into the yield curve it is responsible for stabilizing.
The public liability extends from the funding to the code. The FPC explicitly judged that recent advances in frontier AI pose material risks to financial stability — specifically naming Anthropic’s Mythos and models developed in China that can rapidly identify software vulnerabilities. When frontier models can identify and exploit bugs faster than firms can patch them, the resulting operational contagion becomes another vector for systemic failure, inevitably demanding public backstops when the dust settles.
The Bank for International Settlements warned last month of the same dynamic from a different cut. The BIS named the firm-level problem: fierce competition to dominate AI “risks driving investment spending to excessive levels, threatening the profitability of leading firms.” The FPC names the system-level problem: the leverage funding the investment, the cross-asset contagion when the investment disappoints, the prime broker channel transmitting both. Two central-bank bodies, one month apart, name the same risk from opposite ends. Neither has used the powers it holds.
The pattern is documented. The 1998 LTCM episode established the post-Bretton Woods precedent: leveraged nonbank positions, a perceived near-failure of a large hedge fund, ad-hoc central bank coordination to contain the contagion. The Federal Reserve’s role in organizing the private-sector rescue, in which fourteen banks contributed roughly $3.6 billion to recapitalize the fund, was the precedent for treating hedge fund near-failures as a stability concern. The institutional memory is in the public record.
The 2021 Archegos episode documented the prime broker channel: concentrated leveraged positions in a handful of banks, sudden margin calls, and contagion through the prime broker balance sheet — with Credit Suisse and Nomura absorbing multi-billion-dollar losses between them. The BOE is naming, in 2026, the same mechanism Archegos already demonstrated. The institutional response is “preparing changes.”
The substantive frame: a macroprudential authority that names the bubble, names the channel, names the catalyst, names the cross-asset contagion, names the operational risk, and publishes a twice-yearly report while the position builds. The substantive wrong is not inattention. The substantive wrong is naming the risk and declining to use the powers that exist to address it.
The architecture of the next crisis is already on the balance sheet. When the mechanism is tested, the public will once again finance the gap between a prime broker’s risk model and reality. The score is the score. The Bank of England identified the risk. The Bank of England chose to warn.