IMF warns: AI company debt burden is more dangerous than stock overvaluation

IMF warns: AI companies' debt burden is more dangerous than stock overvaluation

Tobias Adrian, financial advisor and director of the International Monetary Fund's Department of Monetary and Capital Markets, said that the growth of debt financing in the artificial intelligence industry poses a more serious threat to financial stability than overvalued valuations of technology company stocks.

Why debt worries the IMF more than stocks

Conversations about an "AI bubble" over the past two years have centered primarily on the stock prices of companies like Nvidia, Microsoft, Meta, and OpenAI partners: if expectations for AI revenue do not materialize, the value of these securities could plummet sharply. Adrian shifts focus to another part of the balance sheet — debt. In his view, what matters for the financial system is not how much stocks are worth, but who borrowed money and on what terms to build AI infrastructure.

The difference is fundamental: a decline in stock prices dilutes shareholder capital but does not necessarily trigger a chain reaction. Overdue or restructured debt is another matter: it hits banks, bond funds, and insurers that hold these obligations on their balance sheets, and can spread through the system much faster.

The IMF regularly publishes reports on global financial stability in which it assesses risks accumulating in different market segments — from sovereign debt to corporate lending. The emergence of a separate topic of "debt for AI infrastructure" in the rhetoric of senior representatives of the fund suggests that the scale of such financing is already sufficient to figure in such analysis alongside more traditional risk categories.

How the AI boom is financed with debt

Building data centers, purchasing Nvidia accelerators, and renting computing power from cloud providers require enormous capital expenditures. Some large technology companies and specialized data center operators in recent years have increasingly resorted not only to their own profits but also to bonds, bank loans, and structured deals with private lenders to finance this construction.

It is precisely this model — capex financed by debt rather than free cash flow — that concerns representatives of the IMF and other financial regulators. If demand for AI services or revenue growth rates do not match the forecasts embedded in credit models, servicing this debt could become a problem not only for the technology companies themselves but also for their lenders.

How this differs from the classic dotcom bubble

The analogy with the dotcom bubble of the late 1990s, often recalled in relation to AI, works only partially. Back then, it was primarily stock valuations that were inflated — companies raised capital through public offerings, and the collapse hit equity investors first. Today's AI infrastructure build-out is substantially financed by borrowed money, and the holders of this risk are not only shareholders but also banks, pension funds, and insurance companies that bought corresponding bonds. This is structurally a different channel for problem transmission in case of a market reversal, and that is precisely why IMF representatives insist that debt on the AI boom be monitored separately from the stock market.

What this means

The IMF's position adds a new layer of regulatory attention to the AI sector: now important are not only stock multiples but also the quality of debt financing the data center boom. For investors and regulators, this is a signal to monitor the debt burden of hyperscalers and data center operators as closely as the quotations of their stocks.