The rapid expansion of artificial intelligence (AI) technologies has been fueled by extensive borrowing, but recent developments have raised concerns among investors about the sustainability of this debt-driven growth. Last week, market confidence sharply declined in CoreWeave, a major data center company heavily involved in AI infrastructure. CoreWeave, which has accumulated around $72 billion in liabilities to operate 51 data centers, saw its bond prices plummet and the perceived risk of default rise to over 60 percent. This downturn has spurred wider unease about the financial health of other AI-related firms, including Oracle and SoftBank, although the decline in bond values for these companies has been less severe.

The rise in borrowing costs and debt concerns coincides with a broader tightening of monetary policy as the Federal Reserve has begun raising interest rates to combat inflation, which has been partly driven by the rapid growth of data centers. Investors are increasingly demanding higher premiums for lending to AI-related companies, with the spread on risky AI debt increasing more than 50 percent over the past six months, according to data from Barclays.

CoreWeave’s business model has drawn particular scrutiny. The company leases data center space long-term from property owners and then subleases it to clients such as Meta, Microsoft, and OpenAI for shorter durations. This structure creates financial risk because CoreWeave is committed to fixed costs on leases even if customer demand falls off or contracts are not renewed. Some investors have expressed concerns that major cloud providers might reduce their usage of CoreWeave’s facilities as their own data center capabilities expand, potentially leaving CoreWeave with underutilized assets. In response, CoreWeave has rejected comparisons to troubled companies like WeWork and emphasized confidence in its financial discipline and business strategy.

Oracle, another prominent player in AI infrastructure, has faced delays with some of its data center projects, including a significant initiative in Wisconsin called Project Lighthouse. Regulatory hurdles, particularly from state power authorities, have postponed progress, complicating Oracle’s ability to generate projected returns and repay debt. Oracle had previously secured $38 billion in loans for these developments, anticipating revenue from one facility would support debt repayment for the other.

SoftBank, a major Japanese conglomerate heavily invested in AI companies including OpenAI, has also seen rising costs related to its borrowing. The company recently completed a $30 billion investment in OpenAI, financed partly by an $11 billion debt issuance with interest rates close to 10 percent. Insurance costs for SoftBank’s debt have increased substantially, indicating elevated market concerns about its credit risk. SoftBank executives have maintained that their debt levels remain proportional to the company’s overall asset growth.

Despite these challenges, some investors view the current cautious lending environment as potentially beneficial. A slower pace of new financing could enhance the value of existing AI infrastructure by limiting oversupply and encouraging more disciplined investment. However, industry participants acknowledge that the future of the AI build-out will depend significantly on the ability of companies to manage high borrowing costs while generating sustainable returns on their substantial capital investments.