Federal Reserve Vice Chair for Supervision Michelle W. Bowman used a speech on 10 November 2026 to set out, in unusual detail, how the artificial intelligence buildout is being paid for, and what happens to the wider economy if the money does not come back. The remarks, titled The U.S. Economy and Considerations for Monetary Policy, were delivered at the C. Peter McColough Series on International Economics at the Council on Foreign Relations in Washington, D.C. The Board of Governors published the text the same day and noted that Bowman presented identical remarks at a second event, held by Unleash Prosperity Now, later that day.
Bowman described an investment boom that has become a financing structure. Investment in data centers has surged, she said, and growth in real business fixed investment has strengthened to a pace she compared with the 1990s technology boom, with high-tech equipment, software and data center construction accounting for the large majority of growth in business investment since 2025. The problem she identified is not the spending but the gap between the spending and the revenue: “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.”
As capital spending has risen, she said, hyperscalers and other companies have increasingly sourced funds from credit and equity markets, and the cost of that financing has been rising notably as investors price in more risk. She put a figure on the borrowing: “AI-related debt has been rising rapidly, with notable gains in corporate bonds, private credit, and project finance, and is on track to exceed $800 billion this year.” Beyond the debt, she said, long-term lease and contract commitments and residual value guarantees have also surged, and their potential payments are estimated at more than three times the amount of the debt.
The speech then sets out how the pieces are connected. Bowman said hyperscalers and large technology firms, drawing on their strong credit ratings, have assumed off-balance-sheet liabilities so that smaller and riskier companies can use them as collateral in asset-based financing for data center construction and the purchase of accelerated computing equipment. She also said hyperscalers and chip designers have made equity investments in the smaller companies that buy their products, “creating the appearance of circular financing.”
The exposure is not confined to the technology companies. Bowman said the largest U.S. bank holding companies have some exposure to AI-related financing, which is scrutinised in annual supervisory stress tests, and that life insurers, pension funds, private credit funds and mutual funds hold exposure too. Most of those entities are well capitalised and have stable funding, she said, but she added that the complexity of the financing structures makes counterparty risk hard to assess properly. She located the concentration of risk in a few large firms that provide AI services, including AI labs and the hyperscalers.
That concentration produces the scenario the speech is built around. “If AI-related revenues fall short of sharply rising costs, especially as public sentiment and local opposition could lead to increased regulation and delays, we could see a sharp correction in equity prices for AI firms, higher financing costs, and reduced access to credit.” The consequence she drew is macroeconomic rather than corporate: such a shock would likely slow the buildout and soften both economic activity and inflation.
She also suggested the reassessment may already have begun. “We may have already seen some reassessment of expected returns play out in recent weeks as equity valuations dropped, corporate bond spreads widened, and credit default swaps increased for some companies involved in the AI buildout.”
The surrounding economic picture in the speech is less dramatic than the financing section. Bowman said core PCE inflation likely rose to 3.3 percent in October, that tariffs, energy and AI-related demand each contributed to above-target core inflation, and that price pressures from memory chips, semiconductors and computers have been sharp, in a few categories by double digits. She put AI use at “about 25 percent of businesses and roughly 10 percent of workers reporting using AI every day in their work”, with adoption concentrated in larger firms and in information and business services. She also said she thinks some workers will be displaced as firms adopt the technology even as others see productivity and wages rise, and that monetary policy cannot address that transition directly.
On policy she was careful to separate the AI channel from her main concern. She said “Monetary policy also has limited influence on AI-related price pressures”, though she added that in some segments of AI-intensive sectors, particularly among newer and medium-sized firms, higher interest rates appear to be holding back investment and purchasing decisions. Separately, she cited the October Senior Loan Officer Opinion Survey on Bank Lending Practices, saying it suggested that the concentration of demand weakness in real estate loan categories reflected existing headwinds in those sectors. She said she expects the October PCE report and the November employment report before the December meeting, and repeated that policy is not on a preset course. The Board's text carries her standard disclaimer that the views are her own and not necessarily those of her colleagues on the Federal Open Market Committee or the Board of Governors.
What the speech does not contain is as notable as what it does. It announces no supervisory action, names no firm, gives no estimate of potential losses and sets no threshold at which the correction scenario would trigger. It is a risk assessment by a regulator, not a finding.