Bond investors have shifted their focus from inflation risk to the possibility of an AI-driven credit bust, with recent trading showing that concerns over leveraged AI infrastructure can temporarily outweigh price pressures. Despite stronger-than-expected inflation data, demand for U.S. Treasurys increased as investors moved out of AI-linked corporate debt, pushing government yields lower while credit costs for certain issuers rose.
CoreWeave, an AI cloud provider that has borrowed heavily to fund data center development, has become the focal point of this repricing. Its five-year credit default swaps climbed past 1,100 basis points, implying roughly a 64% chance of default over five years, up from about 30% a year earlier. In cash markets, the company’s 9.625% notes due 2032 dropped to about $0.79 on the dollar, with yields touching approximately 16% before easing toward 14%, underscoring how exposed its expansion strategy is to shifting financing conditions.
Stress has spread beyond the weakest names. Oracle’s default swaps approached 300 basis points, SoftBank’s widened by 24 basis points in a single session to 424, and protection on Amazon, Alphabet and Meta reached record levels, albeit at much lower levels than for more leveraged borrowers. The episode illustrates that lower Treasury yields do not automatically translate into relief for corporate borrowers when widening spreads more than offset the decline in benchmark rates.
The pressure is highly concentrated in AI-related high-yield debt. Dollar high-yield AI spreads widened 25 basis points in one day to a record 360 basis points over Treasurys, while comparable non-AI high-yield spreads moved only two basis points and remained below their March peak. What the market is reassessing is a debt-funded buildout: according to BofA Securities, the five largest hyperscalers sold $121 billion of U.S. corporate bonds in 2025, far above the $28 billion average annual issuance from 2020 to 2024. If investors now require materially higher yields, some projects that penciled out under cheaper capital may no longer meet return thresholds.
A protracted pullback in funding could cool demand for data center construction, semiconductors, electrical equipment and specialized labor. Earlier AI investment has strained supply across chips, electricity, construction and skilled workers; softer capital flows could ease some of these pressures, though existing contracts and multi-year build timelines might delay the impact. At the same time, technology, media and telecoms bond issuance had already tripled to $330 billion year to date by October 1, and rising long-dated corporate supply has been competing for investor capital.
Inflation expectations offer a reality check on the notion of an outright deflationary shock. Ten-year breakeven inflation stood at 2.32% on October 15, nearly unchanged from 2.33% on October 9, having moved only a few basis points over the period when nominal and real yields drifted lower. This pattern supports the idea that markets are repricing growth and credit risks around AI infrastructure, rather than signaling a broad collapse in future consumer prices. For the Federal Reserve, the policy challenge is to distinguish a sector-specific financing shock in AI-linked credit from a broader tightening in overall financial conditions.
