AI debt may exceed $800 billion as Bowman flags equity risk

By Sophie Lindqvist

AI-related debt across corporate bonds, private credit and project finance is on track to exceed $800 billion this year, Federal Reserve Vice Chair for Supervision Michelle Bowman said Tuesday.

In a Council on Foreign Relations speech in Washington, released at 3:30 p.m. ET, Bowman warned of "a large correction in equity valuations, if returns on AI investment fall short of expectations."

She called that risk a main threat to growth. Lower equity prices could undermine the resilience of consumer spending, she said. The speech came hours after October consumer prices rose more than expected.

What makes AI financing vulnerable?

Financing costs have risen notably as investors perceive more risk, Bowman said. Her $800 billion estimate covers corporate bonds, private credit and project finance.

Lease commitments, contract commitments and residual value guarantees could add potential payments of more than three times that amount. Bowman's $800 billion figure is a projection, not a confirmed issuance total.

Bowman said: "AI revenues have been rising quickly but are not yet sufficient to pay for the expense of maintaining existing and developing new AI models and data centers."

The gap requires financing before returns can cover spending on models and data centers. The reported remarks offered no chip shipment forecast, process node, wafer size or capacity estimate. Bowman named no chip designers, hyperscalers or data center developers.

How circular is the AI financing?

Bowman pointed to hyperscalers' off-balance-sheet support for smaller data center developers. She also cited equity stakes taken by chip designers and hyperscalers in their own customers.

Together, she said, these links create the appearance of circular financing. Such arrangements complicate any assessment of independent customer demand. She offered no issuer-by-issuer exposure estimates.

Her downside path was specific: a sharp correction in AI firms' equity prices, higher financing costs and reduced access to credit. Those outcomes would slow the AI buildout and soften both growth and inflation, she said.

She said some reassessment is already under way, citing lower valuations, wider corporate bond spreads and higher credit default swap prices for some AI companies. These indicators show a market reassessment, not proof of a completed investment downturn.

What does Bowman signal for December rates?

Bowman estimated core PCE inflation at 3.3% in October. Tariffs, energy and AI-related price pressures each accounted for about one-quarter of above-target core inflation over the 12 months to September.

She expects core PCE to move closer to 2% next year as those shocks fade. She said further work remained to bring inflation closer to 2%. That forecast still depends on those three price pressures easing rather than worsening.

She also said financial conditions have tightened through lower stock prices, higher private interest rates and a stronger dollar. She said the benefit of additional policy tightening must be weighed against its costs for interest-rate-sensitive sectors.

Bowman did not call for a December hike. Governor Christopher Waller backed a December hike and said a half-point move should be considered. Dallas Fed President Lorie Logan said at least two more quarter-point hikes may be needed.

After the consumer-price data, futures fully priced at least a quarter-point December hike and put the odds of a half-point move at 11.8%, according to CME FedWatch data cited by Reuters. The October PCE report and November jobs report will arrive before the Dec. 8-9 meeting, Bowman said.

Her position remains conditional rather than dovish. Inflation pressures persist, labor conditions appear firm, and another hike carries costs. Stocks closed lower on Tuesday. The market summary cited no index levels.

The next test for the AI buildout is credit pricing: tighter financing could slow data center capacity additions, even if AI revenue continues to rise quickly.