S&P Global warns top U.S. hyperscalers may carry undisclosed AI-linked liabilities
Heavy spending on artificial intelligence infrastructure is increasing scrutiny of how the world's largest technology groups fund data centers and related expansion. S&P Global says some obligations tied to that buildout may sit outside what is currently visible in hyperscalers' reported disclosures, even as it sees no immediate need to change ratings.
Highlights
- S&P Global warns top U.S. hyperscalers like Amazon, Microsoft, and Alphabet may carry undisclosed AI-related liabilities via opaque financing arrangements and guaranteed obligations.
- S&P expects the top six hyperscalers to spend over $7 trillion on AI and data center investment between 2025 and 2030, likely resulting in negative free cash flow at least through next year.
- Market concerns about high leverage have increased pressure on AI infrastructure provider debt, exemplified by Oracle’s July downgrade to BBB- due to heavy AI investment and uncertain profitability.
AI financing exposure and ratings outlook
As reported by Reuters, S&P Global estimates the biggest technology companies are likely to have more AI-linked liabilities than are currently known, partly because some financing arrangements tied to data centers or smaller AI firms may not be fully visible. Mark Habib, S&P's global head for digital infrastructure, says the main risk to hyperscaler ratings remains whether expected profits from the AI buildout fail to materialise.Speaking at a roundtable in London, Habib says it is safe to assume there is additional debt the market is not seeing. He points to so-called interconnected financing structures, in which hyperscalers effectively guarantee large data center developments or support borrowing by smaller AI companies, creating potential obligations that may not appear clearly in disclosures or in transactions S&P does not rate.
Habib says S&P does not yet have an estimate for how much extra debt those circular financing structures could add to balance sheets, but the agency is trying to take a holistic view. He also says S&P does not currently see a major disconnect between market pricing and its ratings that would force immediate action, because its credit assessments are designed to reflect a longer-term outlook.
Spending scale and wider market pressure
S&P expects the top six U.S. hyperscalers, Amazon, Microsoft, Alphabet, Oracle, Meta and SpaceX, to spend more than $7 trillion on data centers and other AI-related capital expenditure between 2025 and 2030. It also expects those companies to post negative free cash flow until at least next year, underscoring the scale of investment now being absorbed by the sector.Debt issued by AI infrastructure providers and neocloud companies such as Nvidia-backed CoreWeave has come under growing pressure in recent weeks as investors worry about the aggressive pace of borrowing. That market tension has raised questions over whether bond investors view AI-linked issuers as riskier than current ratings suggest.
Habib says that view could change if spending levels are no longer sustainable. He says further increases in AI investment or intensifying competition that weakens revenue generation could alter S&P's assessment, and notes that Oracle was downgraded in July to BBB-, the lowest rung of investment grade, over concerns about its AI spending and what S&P called an uncertain path to profitability.
Our earlier article on U.S. AI data center expansion highlighted how soaring financing costs, construction delays and power constraints are starting to challenge the $10 trillion buildout. We noted that only a minority of planned projects were viewed as credible, while examples such as CoreWeave volatility and Cipher’s cost overruns underscored growing investor and funding risks around AI infrastructure.