US banks' $1 trillion AI exposure raises risk questions

Approximately $1 trillion in exposure to companies involved in the AI infrastructure build-out is held by the biggest US bank holding companies, according to the Federal Reserve's November Financial Stability Report.

The banking system remains sound with high capital ratios, the Fed stated.

In a New York Fed survey, 87% of market contacts cited AI as a risk, compared with 50% in the spring. The question is whether banks can withstand a downturn if the AI boom falters.

How Banks Got Pulled Into the Data Center Race

The answer lies in how the buildout is financed, the Fed notes. Borrowing for advanced computing infrastructure exceeded $680 billion by October this year, roughly five times the $130 billion raised in all of 2023.

Large technology companies also support smaller, riskier developers via leases and guarantees, which do not immediately affect the giants' balance sheets.

Developers use leases from cloud providers to obtain construction financing. Chipmakers guarantee GPU resale values to support chip purchases.

If the big tech firms also take equity stakes, the Fed notes, this creates a "circular financing" dynamic.

These off-balance-sheet commitments have increased faster than the tech giants' debt. Leases not yet commenced surpass $1 trillion, while guarantees and contract obligations exceed $2 trillion.

The Fed highlighted a trade-off in these relationships.

"Via these connections, the largest technology firms may absorb losses in the event of a future downturn, mitigating their effects on the rest of the financial system. However, it also suggests that these firms' economic exposure to such a downturn may be substantially larger than suggested by their on-balance-sheet liabilities," the Fed noted.

The biggest US banks are also part of this web. Their exposure includes technology giants, data center developers, and utility companies.

Additionally, it encompasses private credit vehicles and computing-related special purpose entities (SPEs).

What an AI Slowdown Would Mean for Lenders

These connections become critical if AI spending fails to deliver returns. Lower-than-anticipated earnings could depress asset prices, tighten credit conditions, and lead to reduced capital expenditure, according to the Fed.

A weaker economy would then strain highly leveraged companies and increase loan losses for banks.

The riskiest portions of these arrangements reside in high-yield funds, life insurance companies, and private debt vehicles.

Investors saw a preview of the risks in early November when US stocks lost $2 trillion in a single hour.

Respondents to the Fed survey cautioned that elevated valuations and debt-laden financing could magnify losses in a more severe market correction.

Not all observers are concerned. Macro investor Raoul Pal has described fears over Big Tech debt as overdone. Additionally, the Fed noted that banks report robust profits and their AI-related exposures are subject to the annual supervisory stress test.

According to the Fed's scenario, the chain of events begins with an unexpected decline in corporate earnings. This sets the stage for tech earnings to become the next key indicator for banks.