Private credit’s huge AI footprint obscures growing risks: Altaf Kassam

Much of the financing for AI infrastructure is taking place through private credit funds, infrastructure vehicles and joint ventures rather than public debt markets. The OECD estimates AI-related deals accounted for 34% of global private credit transactions in 2025, twice the five-year average. The Bank of England estimates about $700 billion of data-centre spending over 2026-28 will be financed by private credit; the Financial Stability Board valued the market at $1.5 trillion to $2 trillion at the end of 2024.

Private lenders generally value holdings quarterly using cash-flow models and comparisons with similar deals, which can produce smoother marks than market prices. Data-centre buildings account for around 30% of asset cost, while chips and equipment make up the remaining 70%; lenders increasingly accept chips as collateral. Rental rates for Nvidia H100 chips have fallen 25% to 50% across regions since early 2025, according to Silicon Data.

The article notes that private credit default rates remain low and that infrequent valuations can suit investors who hold loans to maturity, but argues the approach is less suited to funds offering withdrawals. Blue Owl Capital twice capped redemptions from its private credit funds at 5% this year after heavy requests. The FSB has warned that investors redeeming at stated asset values before loans are marked down can leave the shortfall to those who remain; the article says growing retail access to private markets may expose pension savers to this risk.