S&P Global Flags Potential Undisclosed AI-Related Debt Among US Hyperscalers

Mark Habib, head of digital infrastructure at S&P Global, said at a roundtable in London that US hyperscalers may carry substantial undisclosed AI-related debt, largely stemming from implicit guarantee arrangements for data center construction or smaller AI companies.

Habib noted that interconnected financing structures between large tech firms and AI projects could create off-balance-sheet liabilities. He said S&P has no immediate plans to adjust ratings, as it sees no major disconnect between market pricing and ratings, but a rating change could follow if current spending levels are deemed unsustainable.

S&P estimates that six major US hyperscalers—Amazon, Microsoft, Alphabet, Oracle, Meta, and SpaceX—will spend more than $7 trillion combined on data centers and other AI-related capital expenditures from 2025 to 2030, while remaining in negative free cash flow at least until next year. In July, S&P downgraded Oracle's credit rating to BBB- due to high AI spending and uncertainty in its profitability path.