Andrew Bailey, governor of the Bank of England and chair of the Financial Stability Board, has put G20 finance ministers on notice: the AI boom is starting to look like a classic late-cycle credit event. In a letter sent ahead of upcoming G20 meetings, Bailey flagged inflated AI valuations, surging leverage, and the systemic risks posed by frontier AI models as the top threats to global financial stability.
The mechanics are familiar to anyone who lived through 2008. Investors are piling into leveraged ETFs and trend-following strategies, often with borrowed money. Hedge funds hold positions across both equities and government bonds, creating a transmission belt for shocks. Layer that on top of historically high valuations and heavy market concentration in a handful of AI names, and you have a powder keg. Bailey’s warning is pointed: “As we have seen multiple times in the past, rising leverage is a feature of a maturing financial cycle. While it can reinforce rising markets, it can also intensify declines when sentiment turns, as recent weeks have demonstrated.”
The cross-investment web between AI companies and hyperscalers compounds the problem. If one major AI player stumbles, the contagion could drag down tech giants and spill into the broader market. This isn’t hypothetical doom-mongering. NYU finance professor Aswath Damodaran recently argued that an AI crash could hit harder than the dot-com bust because today’s AI buildout demands massive spending on physical infrastructure, much of it debt-financed. In the 2000 crash, losses mostly hit shareholders. This time, lenders and the real economy could take the hit too.
Bailey’s second concern is less visible but arguably more dangerous. Frontier AI models are gaining autonomous capabilities that could “materially alter the speed, scale and economics of cyber risk.” Attacks would get cheaper, faster, and more frequent. And because banks worldwide depend on a handful of large tech providers, a single successful attack could knock out institutions across multiple countries simultaneously. Bailey notes that uneven cyber defenses between nations create vulnerabilities in their own right — a point that gets less attention than it deserves. Many countries still have no rules for developing, releasing, or deploying advanced AI models, a regulatory vacuum that extends far beyond finance. The FSB is now studying how financial firms can safely use frontier models for cyber defense, but the clock is ticking. The uncomfortable truth is that the financial system has never faced a risk quite like this — one where the same technology driving record valuations could also become the weapon that destabilizes them.
