COMMENTARY: Private credit's huge AI footprint obscures growing risks

Private credit is playing an outsized role in financing today’s AI infrastructure build-out, but it may also be masking how much risk investors are actually taking. Unlike publicly traded instruments, many AI-related holdings in private credit funds are valued by managers using models of expected cash flows and comparisons with similar deals, usually quarterly. Those “mark-to-model” valuations can smooth volatility and make returns look steadier than markets would. The problem is growing as AI becomes a major share of private lending: OECD estimates AI deals were 34% of global private credit transactions in 2025, doubling the five-year average. The Bank of England projects about $700 billion of data-centre spending through 2026-28 financed by private credit, within an estimated $1.5 trillion to $2 trillion market range. Key underwriting risks include collateral assumptions, since chips drive value and chip prices can fall sharply.