S&P warns big tech's AI liabilities may exceed disclosed amounts

The largest technology companies likely have AI-related liabilities beyond those currently visible, S&P Global said on October 19, though its lead sector analyst sees no immediate reason to adjust their credit ratings.

Mark Habib, S&P's Global Head for Digital Infrastructure, told a London roundtable that financing arrangements involving hyperscalers can effectively guarantee large data center projects or smaller AI businesses. He said S&P may lack visibility when companies do not disclose those obligations or when the agency does not rate the relevant transactions.

S&P has not estimated how much extra debt these interconnected arrangements could put on companies' balance sheets, Habib said. The agency expects Amazon, Microsoft, Alphabet, Oracle, Meta and SpaceX to spend more than $7 trillion between 2025 and 2030 on data centers and other AI-related capital expenditure.

Debt from AI infrastructure providers and neocloud companies, including Nvidia-backed CoreWeave, has come under increasing stress amid concerns over borrowing levels. Asked whether the bond selloff suggested investors viewed the debt as riskier than ratings implied, Habib said S&P did not see a major mismatch requiring rating changes.

Habib identified a failure to deliver expected profits as the main ratings risk. Spending becoming unsustainable could also alter S&P's assessment, he said. In July, the agency cut Oracle to BBB-, its lowest investment-grade rating, citing AI expenditure and uncertainty over profitability.