AI infrastructure losses could reach $140B

AI infrastructure credit losses could reach $140 billion in a severe debt re-rating, according to a Chicago Booth stress test of the financing stack. The analysis maps roughly $520 billion in senior unsecured hyperscaler bonds, $250 billion in project and data-center finance, and about $200 billion in estimated private credit; Goldman Sachs’ credit-market review describes the buildout as a multiyear borrowing cycle as capital spending approaches operating cash flow. Infrastructure buyers should underwrite counterparty, collateral-depreciation, and refinancing risk—not just the advertised compute price—alongside the archive’s circular-financing analysis.