AI · October 19, 2026

AI Demand Surges, but Debt Concerns Spark Market Turmoil

Close-up of a green circuit board with components
Akshat Sharma / Unsplash

Global markets have seen a sharp decline in AI-related stocks over the past few days. Semiconductor companies, including Nvidia, have been hit hard, with the market capitalization of U.S. semiconductor firms expected to drop by up to 2 trillion dollars due to recent selling pressure. The once-high-flying stocks of AI investment darlings have suddenly become targets for sell-offs. The issue, however, is not a decline in AI demand. For instance, TSMC, the world's largest semiconductor foundry, has reported record-high earnings for six consecutive quarters, driven by strong demand for AI chips. Data center companies specializing in AI have also secured multi-year contracts. The problem lies in financing. AI companies have taken on massive debt to secure data centers and GPUs, and markets are now questioning whether these firms can repay their debts. A notable example is CoreWeave, an AI cloud company backed by Nvidia. CoreWeave has rapidly grown by leasing AI computing power to Microsoft, OpenAI, and others. Despite securing contracts worth over 100 billion dollars, the company carries around 50 billion dollars in debt and has yet to turn a profit. This has led to a sharp decline in CoreWeave's stock and bond prices, with some bond yields exceeding 16 percent. This indicates that markets believe lenders must charge higher interest rates to compensate for the risk. The price of credit default swaps, a type of insurance against default, has surged to reflect a market assessment that the probability of default within five years is around 63 percent.

While this does not indicate an actual default probability, it highlights the growing unease among investors. The anxiety is not confined to CoreWeave. Financial Times reports that bonds used to finance data centers housing Meta and Oracle are also being heavily sold. As investors flock to safer assets like U.S. Treasuries, their prices have risen. Notably, this unease has spread from the stock market to the bond market. Previously, investors primarily worried that AI companies' future growth expectations and stock valuations were excessive. Now, even lenders are questioning whether they will receive their principal and interest payments. This shift signals a change in the nature of concerns surrounding AI investments. Steve Sosnick, chief market analyst at Interactive Brokers, pointed out that the focus has shifted to the speed, efficiency, and profitability of expanding AI infrastructure. If data center construction slows, spending on semiconductors and equipment could also decrease. There is, however, significant counterargument. Jensen Huang, CEO of Nvidia, stated that CoreWeave's financial situation is improving daily and expressed optimism, noting that capital raising costs are decreasing. He clarified that Nvidia does not need to intervene. Deutsche Bank also viewed CoreWeave's long-term contracts as generating enough revenue to cover related investment and operating costs, even turning a profit. The bank assessed the recent stock decline as a buying opportunity. Michael Intrater, CEO of CoreWeave, emphasized in a memo to employees that the secured contracts would enable the company to manage its debt.

Ultimately, this situation differs slightly from questioning the growth potential of the AI industry itself. Even as the AI industry continues to expand, the question is whether all companies that have aggressively expanded through debt will survive. Financial Times described this as a 'Darwin moment' for AI infrastructure companies. The era where any investment was justified solely by increasing AI demand has passed, entering a 'survival of the fittest' phase where only financially sound companies with actual profitability will thrive. VeslAI, a company that has secured 2 trillion won in GPU financing, has emphasized that its financial risks differ from those of CoreWeave, citing customer diversification and flexible contract structures. Jensen Huang's remarks on potentially halting chip supplies to irresponsible AI developers are seen as emphasizing that companies should prioritize developing safe AI models over relying on regulations. Huang has consistently argued that the development of AI without ensuring safety should be halted. SpaceXSI has showcased the high performance of its 'Grok Voice 3.0' by highlighting its success in handling approximately 20,000 calls daily at the Starlink customer service center, resolving 70 percent without human intervention. This demonstrates a competitive edge over other AI-focused companies by leveraging real-world applications through existing businesses.