Corporate AI Hardware Debt Poses Greater Stability Risk Than Stock Prices, IMF's Adrian Warns

Tech companies are borrowing heavily to invest in artificial intelligence hardware. Adrian highlighted that there may be a discrepancy between how long AI hardware remains useful and the repayment period of the loans taken to acquire it.

Tobias Adrian, director of the Monetary and Capital Markets Department at the International Monetary Fund, says that gap - not high stock market prices - is the real worry for financial stability.

This warning follows a Sunday statement from the Bank for International Settlements that identified AI as one of four key vulnerabilities to global economic prosperity. Adrian's remarks at the ECB symposium in Sintra, Portugal, on Tuesday underscore growing unease over how tech firms are financing their AI push.

Adrian noted that major tech firms are starting to leverage up.

The Borrowing Binge

Adrian said that from a financial stability perspective, "what is quite worrisome … is that the major tech firms are starting to leverage up themselves."

Get your free investing masterclass bonus when you join Market Briefs, our free daily newsletter

He added that chips can age "fairly quickly" while firms are also issuing medium or long-term debt. If the assets stop earning money, the debt stays.

The core issue is the timing mismatch: AI chips, such as those used for training large language models, can become obsolete within a few years as new, more efficient hardware emerges. Meanwhile, corporate bonds and loans taken to finance these purchases may stretch over five to ten years. If the chips lose value or fail to generate projected revenue before the debt is repaid, firms could face refinancing difficulties or defaults.

The danger is not a classic stock bubble. Adrian said, "Last week's stock selloff has shifted the picture somewhat." "There's not as much valuation pressure anymore that you saw recently," he said. "Valuations are very much related to price-earnings ratios, and so it's both the prices that have come off a little bit, but it's earnings that continue to surprise on the upside as well."

He observed that while investors have had "extremely aggressive" "expectations for AI profitability, earnings have actually outperformed, suggesting that market dynamics are" "really quite different from sort of bubble-like behavior."

Earnings Are the Real Test

So far, the story is good. AI earnings have been strong. But Adrian said the core question is whether those profits will last. "As long as profitability continues, as long as corporates and individuals are paying, you know, for the frontier-model costs, that is probably fine," he explained. "But at some point, of course there could be a disappointment to earnings."

If earnings drop, the large borrowing by tech firms could become a big problem. "It's really the underlying profitability that is the key financial stability issue at this point," Adrian said.

Adrian said a true bubble would show "quite a bit of a dilution of market forces." He does not see that yet. He observed that the divergence between rising chipmaker valuations and declining software stocks could be viewed as a healthy market signal.

What to Watch

Adrian is watching the leverage. However, a shortfall in earnings might pose difficulties.

Subscribe to Market Briefs, our free daily newsletter, and claim your bonus investing masterclass