S&P Global warns: AI-related liabilities of US tech giants may exceed known scale

S&P Global has warned that the scale of AI-related liabilities at the world's largest technology companies may far exceed disclosed levels, but the agency currently has no plans to revise ratings.

Mark Habib, S&P Global's Head of Digital Infrastructure, stated at a London roundtable that hyperscalers may carry significant undisclosed AI-related debt, primarily originating from implicit guarantees for data center construction or smaller AI enterprises. He said:

"It is reasonable to believe there is more (debt) that we cannot see."

This statement comes as the market grows increasingly cautious about the prospects for financing AI infrastructure. Bond prices issued recently by AI infrastructure providers and neocloud companies such as CoreWeave, supported by Nvidia, continue to fall, reflecting investors' concerns over aggressive borrowing.

Mark Habib focused on the "interconnected financing structures" between hyperscalers and AI projects—large technology companies providing de facto guarantees for major data center construction or smaller AI businesses, resulting in potential off-balance-sheet liabilities.

"If these liabilities are not disclosed, they will not appear in hyperscalers’ financial disclosures," Habib said. "If the transactions these liabilities may generate do not fall within our ratings scope, we may also be unable to assess them."

He stated that there is currently no specific estimate for how much additional debt these circular financing structures might add to companies' balance sheets, but S&P is working towards a holistic evaluation.

S&P forecasts that Amazon, Microsoft, Alphabet, Oracle, Meta, and SpaceX—the six leading hyperscalers in America—will collectively spend over $7 trillion on data centers and other AI-related capital expenditures between 2025 and 2030, and will remain in negative free cash flow at least through the next year.

Mark Habib noted that the greatest risk facing hyperscaler ratings remains whether expected profitability can be realized on schedule.

In July this year, S&P lowered Oracle's credit rating to BBB-, the lowest investment grade, citing excessive AI spending and "uncertainty around the path to profitability."

In response to recent market signals stemming from AI-related bond sell-offs, Mark Habib stated that S&P currently does not see a significant disconnect between market pricing and ratings that would necessitate forced correction.

"At present, we do not think there is a major gap that would force us to adjust ratings," he said, explaining that S&P's rating methodology takes a longer-term perspective.

However, he clarified conditions that could trigger a ratings reassessment: if current spending levels are deemed unsustainable, S&P’s rating stance may change. "This could come from a variety of factors—AI companies ramping up investments even further, or competition within the industry becoming so intense that it impacts firms’ revenue growth."