Borrowing costs for artificial intelligence data center companies have surged, with CoreWeave's bonds trading at yields above 16 percent, signalling growing anxiety in credit markets about the future of the sector. The company's five-year credit default swaps spiked above 1,289 basis points on Thursday, widening beyond their October peak, according to Deutsche Bank, before easing to around 1,155 on Friday. The stress reflects concerns that new Democratic governors and a Democratic-controlled Congress will impose tougher restrictions on data center development in key states.
Investors are demanding higher returns to protect against default risk across the entire sector, even among the largest and healthiest companies. Credit default swap premiums on Nvidia debt have risen 76 percent over the last month, while Broadcom's have gained 49 percent and Amazon's have climbed 44 percent, according to the latest figures. These increases signal that markets are pricing in greater uncertainty about future financing costs across the industry. Jim Reid, Deutsche Bank's global head of macro and thematic research, noted that debt and equities are now moving more in sync, creating additional pressure on borrowing conditions.
Political developments are reshaping expectations for data center deployment. Moratoriums in Ohio, Texas, and Georgia could impact 45 percent of planned US capacity, Goldman Sachs said. Texas governor-elect Gina Hinojosa has indicated she wants a moratorium until her state legislature passes community protections, though she has said she is not opposed to data centers outright. The Trump administration is pushing back against these concerns, with National Economic Council Director Kevin Hassett telling CNBC on Friday that he expects data center growth to continue as forecast despite Democrats' recent election victories.
Despite the credit market stress, debt issuance has not collapsed. Goldman Sachs reported five AI bond deals totalling 67 billion dollars since October 29, bringing total hyperscaler debt issuance for the year to 296 billion dollars. The bank forecasts 440 billion dollars of hyperscaler debt issuance for next year, plus an additional 300 billion dollars in project-level issuance. However, a key vulnerability has emerged: hyperscalers collectively reported negative free cash flow in the third quarter as capital spending exceeded operating cash flow, Goldman analysts said.
Experts warn that the debt market trajectory may be reaching a turning point. Bank of America analysts estimate AI-related credit issuance will rise 31 percent from 2026 to 2027, hitting 685 billion dollars, before falling 23 percent to 531 billion dollars in 2028. The bank cautioned that downside risks exist from a potential slowdown in AI infrastructure projects driven by Democratic policies, which would reduce capital spending needs and debt market funding. Christopher Wood, Jefferies' global head of equity strategy, suggested that rising interest rates and potential credit limits represent the most plausible way the AI investment boom could eventually end.
For borrowers seeking to finance data center projects, the changing credit landscape means significantly higher interest rates and potentially tighter lending conditions, while savers holding corporate bonds face heightened credit risk across the technology and infrastructure sectors.