Debt investors become cautious as AI businesses borrow billions

Debt Markets Turn Cautious Even as AI Stocks Stay Strong

Investors in stock markets continue to push AI-related shares to new highs, largely brushing aside warnings of a potential technology bubble. But debt investors are sending a more cautious signal, as borrowing costs rise sharply for newer and riskier AI-focused companies.

Young AI and data-centre firms raising money through bonds are being forced to pay much higher interest rates, reflecting scepticism about their long-term stability. In one case, data-centre builder Applied Digital had to pay interest about 70% higher than similarly rated companies, showing investor concern about risk.

Several warning signs are emerging in debt markets. Some AI-related bonds have fallen in price shortly after issuance, while the cost of credit default swaps — insurance against bond losses — has increased. Investors are worried that construction delays at large data centres could delay revenues, or that demand for AI computing power may eventually fall, leaving excess capacity and debt defaults.

Debt investors tend to be more cautious than equity investors because they do not benefit from upside gains. As AllianceBernstein portfolio manager Will Smith said, equity investors accept higher risk because they benefit if things go well, while bond investors focus mainly on avoiding losses.

AI-related companies have raised more than $100 billion in debt this year, according to Refinitiv, much of it by large, established firms such as Amazon, which raised $15 billion to fund Amazon Web Services. However, newer and smaller companies are paying significantly higher rates.

In October, Wulf Compute, part of Terawulf, raised $3.2 billion in bonds paying 7.75% interest, well above the sector average of about 5.5%. Similarly, Cipher Compute, a unit of Cipher Mining, raised $1.7 billion at a rate just above 7%.

Smaller firms often rely on contracts with large customers such as Meta, OpenAI and Microsoft to support their credit ratings. But this dependence worries investors. Applied Digital, which raised $2.35 billion at a steep 9.25%, depends heavily on CoreWeave as its main tenant. CoreWeave itself borrowed $1.75 billion at around 9%, without offering collateral, making the debt riskier.

CoreWeave’s shares surged after its March stock market listing but later fell sharply, while its bonds now trade at a discount and yield over 12%, far above the average for similarly rated companies. Applied Digital’s bond yields have also risen above 10%, signalling growing concern among debt investors.

Analysts say that while large technology firms are unlikely to be seriously harmed, smaller AI infrastructure companies could face stress if expectations fail to match reality. As S&P Global’s Naveen Sarma noted, the key risk lies in construction delays and uncertainty over future demand.

Recent market moves suggest sentiment is shifting. Earlier AI debt deals are being repriced lower, while newer ones are being issued at higher yields. Equity markets have also pulled back from their highs, especially among more speculative AI stocks.

“The debt markets, like the equity markets, are moving with sentiment,” said CKC Capital co-founder K.C. Baer. “Investors are trying to judge how much risk they are really taking in a very uncertain environment.”