Credit stress builds around AI data centers as construction continues

The cost of insuring Oracle debt against default rose 26 basis points overnight to a record 288 basis points, according to S&P Global Market Intelligence data reported by Reuters Oct. 15. The move adds to signs of strain in the financing behind the AI data center buildout, even as demand for computing capacity and space remains strong.

Oracle has drawn scrutiny as it borrows to build capacity ahead of expected AI revenue. Moody’s rates Oracle Baa2 with a negative outlook. Oracle had $129.5 billion in outstanding debt as of May 31, 2026. S&P cut Oracle in July to BBB-minus, one notch above speculative grade.

The concern is the gap between spending and returns. Oracle must commit large sums before it knows when new facilities will generate enough revenue to cover their costs. That leaves it reliant on more borrowing and on major customers, including OpenAI.

A review of 18 market signals over the past week points to broader pressure on AI infrastructure financing. More of the debt and lending measures have deteriorated as investors demand higher returns and lenders grow more selective.

CoreWeave’s 2032 bonds traded at about 77 cents on the dollar, reflecting concern about repayment. The cost of insuring debt issued by Amazon, Alphabet, Meta and Nvidia also reached records, according to Reuters, though their levels remain well below Oracle’s.

Pressure extends across the financing chain. Several banks that lent to large data center projects, including SocGen, SMBC and MUFG, are pulling back, The Information reported Sept. 30. Global AI-related debt issuance fell from $113 billion in June to $23 billion in September, the Financial Times reported, citing Morgan Stanley data, while some private credit funds have limited investor withdrawals. The independent monitor of PJM, the largest U.S. grid, told federal regulators on Sept. 25 that about 42 percent of the forecast growth in data center power demand has been delayed or canceled. Chip stocks fell for six straight sessions through Oct. 14, new AI stock offerings have struggled, and insiders at recently public AI companies have sold shares.

The demand measures tell a different story. Prices to rent AI chips are rising, data center space remains nearly full, and construction spending is sharply higher than a year ago. Large technology companies have maintained or raised their spending plans. Microsoft kept its 2026 plan unchanged in July after raising it in April. For now, the financing is under more strain than the demand driving construction.

The next test comes in late October, when the largest technology companies report earnings. A cut in capital spending or a decline in AI chip rental prices would show that the pressure has reached demand as well as credit markets.