Same investors, two borrowers: Barr and Vujčić on how AI financing meets public debt
Two central bankers spoke about the financing of the AI build-out on 16 November, from different angles that fit together.
Michael Barr, citing the Fed's new Financial Stability Report, put numbers on the structure: on-balance-sheet debt of the seven largest AI firms expected to exceed $800 billion by year-end, potential off-balance-sheet obligations of more than $3 trillion, and over $100 billion of equity stakes in AI labs that buy compute from their investors. He said he is less concerned about direct losses given the still-limited debt, and more about how these arrangements would transmit a repricing.
Boris Vujčić, previewing the ECB's review, described the same shift - record debt issuance by cloud and AI infrastructure firms, capital expenditure projected at $5.5 trillion by 2030, opaque off-balance-sheet structures - and then connected it to sovereigns. Euro area governments need to borrow more, the investor base has become more price-sensitive, and he asked whether AI-related issuance could push up financing costs or crowd out other investment.
Neither speech claims that a repricing is coming. Both describe where losses would travel if it did: through guarantees and leases in Barr's account, through funds, hedge funds and bond markets in Vujčić's.

Why it matters
For rate watchers the link is the supply of bonds. When companies building data centres and governments facing higher deficits borrow from the same investors in the same years, the price of that money is set by both - which is why a financial stability review in Frankfurt now discusses US technology capex.