Private credit’s huge AI footprint obscures growing risks

Altaf Kassam, CFA, founder of Reference Portfolio Partners and former head of Investment Strategy at State Street Investment Management, argues in a Reuters commentary that much of the AI infrastructure build-out is being financed through instruments that do not trade. While hyperscalers such as Meta Platforms, Alphabet and Amazon have tapped public debt markets, much of the sector's funding flows through private credit funds, infrastructure vehicles and joint ventures.

Because these holdings lack a market price, managers value them, usually quarterly, using expected cash-flow models and comparisons with similar deals. Kassam argues this can produce 'smoothed valuations' that keep volatility artificially low and make private credit appear to offer higher returns with lower risk, even though the underlying risk persists and is simply not accurately measured.

The scale of AI private lending is growing quickly. The OECD reckons AI-related deals made up 34% of global private credit transactions in 2025, and the Bank of England estimates roughly $700 billion of data-centre spending over 2026-28 will be privately financed. Kassam notes this compares with a private credit market the Financial Stability Board estimated at $1.5 trillion to $2 trillion at the end of 2024.

Kassam questions the robustness of valuation assumptions. A data centre's building typically accounts for only around 30% of cost, per JPMorgan Asset Management, with chips and equipment making up the remaining 70%, and lenders increasingly accept those chips as security. Rental rates for Nvidia's H100 have fallen 25% to 50% across regions since early 2025, according to Silicon Data, with smaller specialist providers' rates lower still. The column acknowledges counterarguments: major AI tenants are highly creditworthy and private credit default rates remain low, and long-horizon holders may reasonably ignore interim marks. However, Kassam notes this logic works less well when investors can redeem. Blue Owl Capital, one of the largest data-centre lenders, twice capped redemptions from its private credit funds at 5% this year. The FSB's 'first-mover incentive' describes how semi-liquid funds valuing loans at par allow early redeemers to exit at stated value while the shortfall falls on remaining investors.

Kassam also flags a distribution risk: UK pension providers have pledged under the Mansion House Accord to place 10% of default funds in private markets by 2030, US 401(k) plans are being opened to alternatives, and the FSB estimated retail investors held around 13% of global private credit in 2024. Because most pension savers rely on default or target-date strategies, Kassam warns the eventual marginal buyer of AI's least transparent debt could be savers who will not see losses as they occur.