Private credit’s huge AI footprint obscures growing risks: Altaf Kassam
XLF•AI-related deals made up 34% of global private credit transactions in 2025, while roughly $700 billion of data-centre spending over 2026-28 is estimated to be financed by private credit. The author warns that model-based valuations may understate risks as chip collateral values fall and investors seek redemptions.
1. AI lending expands
Private credit is financing a growing share of AI infrastructure, often through funds, infrastructure vehicles and joint ventures whose holdings are valued by managers using models rather than market prices. The OECD estimated AI-related deals accounted for 34% of global private credit transactions in 2025, twice the five-year average; the Bank of England estimated about $700 billion of data-centre spending over 2026-28 will be financed by private credit.
2. Collateral valuations
The author cautions that valuations may lag changes in asset values. Data-centre buildings typically account for around 30% of an asset’s cost, with chips and equipment making up the rest, according to JPMorgan Asset Management. Rental rates for Nvidia H100 chips have fallen by 25% to 50% across regions since early 2025, according to Silicon Data.
3. Redemption pressure
Blue Owl Capital, a major data-centre lender, capped redemptions from its private credit funds at 5% twice this year after heavy withdrawal requests. The article says semi-liquid funds that allow redemptions at stated asset values can give investors an incentive to withdraw before loan marks adjust, while private-market investments are increasingly reaching pension savers.




