The Financial Stability Board has warned that the private-credit industry's growing role in funding the AI build-out could trigger "sizeable" investor losses if model demand or compute economics shift sharply. In a new report on vulnerabilities in private credit published this week, the global watchdog — which coordinates with central banks across two dozen jurisdictions — flagged AI infrastructure financing as a concentration risk for funds that are increasingly outside traditional banking supervision.
The single most striking number is the speed of the shift. AI firms accounted for more than a third of all private-credit deals in 2025, the FSB found, up from 17% across the previous five years combined. Tech is now one of the three largest sectors borrowing from private credit, alongside healthcare and services. "This focus on specific sectors may leave private credit funds exposed to idiosyncratic risks … [and] increase exposure to region or industry-specific shocks," the report said.
The FSB's specific worry on AI loans is a "sharp correction in asset valuations, which have increased rapidly", and which "could lead to sizeable credit losses to private credit investors". Two scenarios are flagged. The first is supply-side: a "significant shortfall in the supply of electricity, a critical factor in the construction and operation of datacentres", causing project delays or cancellations. The second is demand-side: an oversupply of datacentre capacity that outpaces realised AI demand and depresses returns.
The bank-private-credit handshake
The other thread in the report — and the one most often missed in headline coverage — is how exposed regulated banks have become to the supposedly unregulated private-credit channel. Banks now lend directly to private-credit funds, finance their riskier portfolios, lend to firms that simultaneously borrow from those funds, and increasingly co-originate deals with asset managers. The FSB notes that this leaves banks exposed to an "opaque sector where lenders may have only partial information about borrowers".
The watchdog cites two test cases from last year: the collapses of US auto firms Tricolor and First Brands, both private-credit-backed and both since hit with fraud allegations. JP Morgan and Barclays took losses on Tricolor; UBS and Jefferies disclosed material exposures to the failures. Together, the FSB writes, those bankruptcies showed "how tightly integrated banks can be in the intricate web of exposures in corporate credit". Several private-credit funds also recently capped client withdrawals after a multibillion-pound run on redemptions.
Why it matters
Two things are worth holding in mind together. First, the AI build-out is real, the GPUs are on order, and the datacentres are under construction — there is genuine demand at the bottom of the stack. Second, the financing structure layered on top of that demand has migrated from the bank-supervised system, where it would have shown up in stress tests and capital ratios, into a parallel channel where the supervisors get a partial view. The FSB's report is, more than anything, a request for that view to be widened before the cycle turns.
The Tricolor and First Brands precedents matter less for their headline losses than for their shape. Both were borrowers that traditional banks were not lending to directly, but to which traditional banks turned out to be exposed several layers deep. If a single AI-infrastructure issuer goes the same way — a builder over-extended on hyperscaler off-take that fails to materialise, say, or a sponsor undone by an electricity-supply contract — the contagion path is structurally similar and the data on who-owes-what-to-whom is structurally worse.
Primary source: FSB — Report on Vulnerabilities in Private Credit (5 May 2026).
