The rapid expansion of artificial intelligence (AI) has been fueled by trillions of dollars in borrowed capital, but recent market developments have raised concerns about the sustainability of this debt-driven growth. Investor confidence waned sharply last week following a selloff in bonds issued by CoreWeave, a New Jersey-based data center operator carrying $72 billion in liabilities to power 51 data centers dedicated to AI workloads. CoreWeave’s bond prices collapsed, with investors assigning more than a 60% probability that the company might fail to meet its debt obligations.
This turbulence has extended, to a lesser degree, to other major players in the AI ecosystem. Bonds linked to Oracle, a company owned by Larry Ellison that also supports AI projects, declined in value, although not as dramatically as CoreWeave’s. Meanwhile, Wall Street investors have increased demand for credit default swaps—financial instruments designed to insure corporate debt against default—on companies like SoftBank, the Japanese conglomerate heavily invested in AI ventures through its partnership with OpenAI.
These trends pose a risk of financial losses for current bondholders and raise the likelihood that these companies may face steeper borrowing costs in the future, potentially slowing the pace of AI development. Barclays, an international investment bank, reported that the spread, or risk premium, on AI-related debt has surged by more than 50% over the past six months, reflecting heightened market anxiety.
Analysts point to a fragile market environment amid a continuation of robust borrowing. “The market is on a knife’s edge everywhere,” said Alexandra Coleman, founder of the data center risk research firm Azraq. This apprehension comes as financial institutions and private equity firms forecast continually rising capital needs to support AI infrastructure, seen by many as central to future technological advancements.
A contributing factor to market unease is the Federal Reserve’s recent interest rate hikes, aimed at curbing inflation partly driven by the booming demand for data centers. According to David Munves, head of credit strategy at AIR Platforms, there was a “collective realization” last week that the risks connected with AI investments may be greater than initially anticipated. He noted a disconnect between the funds companies are currently able to raise and their potential to generate sufficient revenues going forward.
CoreWeave’s bond yield recently climbed above 15%, landing it squarely in the category of “junk bonds.” The company’s stock plunged 24% over the week, with a 12% drop on Friday alone. CoreWeave’s business model is complex and has drawn comparisons to WeWork, the office leasing company known for aggressive and ultimately unsustainable financial arrangements. CoreWeave purchases AI chips from manufacturers such as Nvidia and leases data center space on long-term contracts, often exceeding 15 years, from third-party owners. It then subleases this space to clients like Meta, Microsoft, and OpenAI on shorter-term agreements. This leaves CoreWeave responsible for high fixed costs even if customer demand fluctuates.
CoreWeave denied the WeWork analogy. Lia Davis, a spokesperson for the company, stated, “The complexity of our financing reflects discipline, which is a point of pride for us.”
As the AI sector continues to rely heavily on debt to fund its expansion, market participants are closely watching whether companies like CoreWeave, Oracle, and SoftBank are isolated cases or indications of broader financial challenges in the rapidly evolving AI landscape.