In September and October, 87% of financial market representatives surveyed by the US Federal Reserve System (Fed) called artificial intelligence a significant risk to the stability of the financial system. By comparison, in the May survey, 50% of respondents expressed such an assessment. In the report, the central bank also drew attention to the rapid growth of debts related to the construction of data centers and other computing infrastructure.

Reuters reports this.

Why artificial intelligence worries the financial market

The survey results were published in the Fed's semiannual "Financial Stability Report" — it summarized the opinions of more than two dozen representatives of banks, brokerage firms, investment funds, and consulting companies.

Respondents consider both the AI technology itself and the extremely rapid pace of investment in this sector to be the main threats.

In particular, survey participants are concerned about the possibility that a slowdown in profit growth in the AI sector will trigger a large-scale market correction.

Among other threats to the financial system, survey participants named:

The Fed report pays special attention to the financing of data centers and other infrastructure for advanced computing.

By October 2026, the volume of debt obligations issued in this segment reached 680 billion dollars — already five times more than for all of 2023 (about 130 billion dollars).

Currently, such borrowings account for almost a third — about 31% — of the total volume of investment-grade corporate debt outside the financial sector. Three years ago, its share was less than 6%.

The Fed also paid special attention to the complex financing schemes for part of these investments.

Tools such as long-term contracts, lease agreements, and guarantees allow operators of large-scale cloud infrastructures ("hyperscalers") and other large corporations to maintain financing for projects implemented by smaller third-party companies.

In addition, hyperscalers and GPU developers often invest in such projects themselves by purchasing equity stakes.

According to the Fed, the total volume of debt obligations, guarantees, and lease agreements that have not yet come into force regarding this infrastructure may exceed 3 trillion dollars.

The regulator notes that such a model of "cyclical financing" could allow the largest companies to absorb losses in the event of an economic downturn, minimizing the blow to the financial system as a whole.

At the same time, this means that the real vulnerability of these corporations to a possible downturn may be significantly higher than follows from the volume of liabilities reflected directly in their balance sheets.

"The complex interrelationships between data center development companies, public credit market participants, private lenders, and special financial structures create vulnerabilities that require close monitoring," the report says.

Valuations across asset classes remain high — partly due to investor enthusiasm for AI technologies.

The document emphasizes that investments in this sector became a significant driver of US economic growth this year, however, an unexpected drop in corporate profits could provoke a market downturn, tighter lending conditions, and a reduction in investments.

The situation could also be complicated by the high level of borrowed money usage (leverage) in certain segments of the financial sector.

In particular, according to the Fed, the level of leverage among hedge funds — especially the largest of them — remains the highest over the entire observation period since 2013.