While AI infrastructure stocks face a sharp sell-off due to credit stress and rising rates, physical construction continues. Hut 8's Louisiana project highlights the disconnect between market sentiment and on-the-ground expansion.
AI infrastructure stocks are experiencing a significant sell-off, yet the physical construction of data centers continues unabated. This disconnect between Wall Street sentiment and on-the-ground activity was highlighted by Yahoo Finance Executive Editor Brian Sozzi and EMJ Capital Founder and President Eric Jackson. They pointed to Hut 8's (HUT) Louisiana data center project as a prime example of the ongoing build-out, even as market values for related equities decline.
The recent equity downturn is driven by deepening credit stress and concerns over the financing of massive AI capital expenditures. According to recent market data, CoreWeave, a major neocloud provider, has accumulated over $51 billion in debt and obligations to expand its AI infrastructure. Investors are increasingly wary of this leverage, with CoreWeave's five-year credit default swap (CDS) spread spiking above 1,140 basis points at one point on October 15, a level S&P Global Market Intelligence associates with distressed credit. This anxiety has rippled across the sector, contributing to a broader rout that wiped out over $1 trillion in market value from a basket of 20 leading AI and chip companies this week.
Despite the financial headwinds, the physical expansion of AI infrastructure remains robust. The sell-off has been exacerbated by rising U.S. Treasury yields, which have topped 5%, increasing the cost of capital for these energy-intensive projects. Broadcom CEO Hock Tan and Intel executives have warned that funding availability could become a greater bottleneck than chip or power shortages. Nevertheless, companies are proceeding with construction plans, suggesting that long-term demand for compute power remains strong despite short-term financing pressures.
Analysts note that while the market is punishing high-leverage AI debt, the underlying demand for infrastructure is not in question. Deutsche Bank reported that CoreWeave's total backlog exceeds $100 billion, with a significant portion tied to investment-grade counterparties such as Microsoft and Meta. JPMorgan research suggests that the shortage of powered data-center capacity could reach approximately 19 gigawatts next year, indicating that the physical supply of compute infrastructure is still lagging behind demand. The current market correction may reflect a reassessment of risk rather than a fundamental shift in the AI build-out trajectory.