The data center bond market is undergoing a sweeping selloff, with Gartner predicting that 80% of new cloud computing companies will cease independent operations by 2030, while Allianz's chief investment officer warns that data center bonds are entering a "natural selection" moment.
According to a Thursday report, prices have fallen into stressed territory across a wide range of debt instruments — from specialized financing bonds for large data centers leased to Meta, Microsoft, and Nvidia, to debt issued by emerging cloud computing companies like CoreWeave — reflecting deep investor skepticism about the sustainability of returns on AI infrastructure investment.
The immediate triggers for this selloff include: OpenAI delaying its IPO to next year, Anthropic's quarterly revenue growth slowing, and OpenAI's disclosure last week that its actual annualized revenue was far below prior expectations. These signals have intensified market doubts about AI monetization capabilities and triggered correlated declines in both stocks and bonds. Nvidia's market value has evaporated by approximately $270 billion over the past week, CoreWeave's stock has fallen about 17% during the week, and competitor Nebius has dropped 18%.
Concerns among asset management institutions are spreading. Allianz Group's Chief Investment Officer Ludovic Subran stated: "Data center bonds are entering a 'natural selection' moment, and the weakest credit entities will struggle to survive. The surge in credit spreads triggered by power and construction delays indicates that data center debt was previously priced too leniently for execution risk."
Bond prices fall across the board, some hitting distressed levels
The scope of this selloff has exceeded prior expectations, extending beyond emerging cloud computing companies to multiple data center financing projects backed by large technology giants as tenants.
Bonds issued by Beignet Investor, a special purpose vehicle jointly funded by Blue Owl and Meta for the Hyperion project, have fallen from above par in early July to approximately 92 cents. Bonds for two projects developed by Hut 8 in Louisiana and Texas (the latter leased by Nvidia) have dropped to approximately 94 cents and 92 cents respectively. Bonds for Blackstone-owned QTS Data Centers' "Project Excalibur" in Georgia (with Microsoft as the tenant) have fallen below 88 cents.
Emerging cloud computing companies have been hit even harder. Bonds issued by CoreWeave have fallen below 80 cents — a level typically viewed by the market as the hallmark threshold for distressed debt — with yields briefly approaching 15% on Thursday. According to a previous report, approximately $18 billion in loans related to Oracle's "Project Jupiter" data center in New Mexico have also slid into stressed territory due to power supply permit delays amid local opposition. Oracle's construction bonds for its Michigan AI campus have also fallen from par to 87 cents.
A trillion-dollar gap: AI revenue far from sufficient to cover capital expenditure
The deeper logic behind the shift in market sentiment lies in the enormous gap between AI infrastructure investment and foreseeable returns.
Technology consulting firm Gartner released a report this week predicting that 80% of emerging cloud computing companies will "cease independent operations" by 2030 due to a multi-trillion-dollar gap between the pace of capital expenditure and AI return expectations.
Gartner estimates that AI data center operators need to generate cumulative revenue of $8.5 trillion by the end of 2030 to cover capital and financing costs and achieve reasonable returns. However, it predicts that consumer and enterprise AI service spending flowing to data center providers "will not exceed $3.6 trillion." Gartner warned: "If AI infrastructure investment maintains its current trajectory, more than $4 trillion in assets could face impairment risk by 2032."
Nvidia CEO Jensen Huang said in a Bloomberg interview on Thursday that he does not believe CoreWeave needs financial support from Nvidia. "I'm optimistic. Their cost of capital is actually declining," he said, adding that CoreWeave's strength is "improving every day."
Regulatory headwinds and circular financing: compounding risks that cannot be ignored
Beyond fundamental pressures, two types of structural risks are drawing increasing attention from investors.
The first is political resistance from local communities. Backlash against data centers has spread across multiple locations in the United States, triggered by concerns over water consumption, noise, and air pollution, and has already stalled some projects. New York State and Texas have imposed strict restrictions on new projects, and other states are considering similar measures. Tom Murphy, Head of U.S. Investment Grade Credit at Columbia Threadneedle, said the opposition to data centers "is something we are closely monitoring. I wouldn't say it has changed our investment decisions, but it does make data centers that are already built and operational more valuable."
The second is circular financing risk. The Bank for International Settlements recently published a report noting that "a significant proportion of financing and commercial relationships in the AI sector are self-referential in nature" — meaning customers and suppliers finance each other. This structural characteristic means that if any link in the chain encounters problems, it could trigger a chain reaction, making risk transmission harder to predict.
Market divergence: operational assets relatively resilient
Despite overall pressure, signs of divergence are emerging within the market.
Tom Murphy's remarks suggest that data center assets that have already been built and are operational may actually command a relative premium under current headwinds due to supply contraction. This aligns with the logic of Gartner's report — new projects face a triple squeeze of permitting delays, rising financing costs, and revenue uncertainty, while the scarcity of existing high-quality assets will become more pronounced.
For bond investors, the most critical question now is: as yields on bonds from companies like CoreWeave approach 15%, has the pricing of distressed assets sufficiently reflected execution risk and industry cycle risk? The answer to this question may gradually emerge in the coming months from revenue data on the AI demand side, including from OpenAI and Anthropic.
Comments