Companies that build data centers and lease computing capacity for artificial intelligence are approaching a massive wave of repayment of loans raised at record-low rates. They will have to return tens of billions of dollars, and refinancing will cost significantly more. For shareholders, this is a direct risk of lower profits, and CoreWeave bonds are already yielding buyers more than 16% per annum.
The problem extends far beyond a single industry: it is precisely the AI boom in recent years that has kept the US stock market at its highs. If data center operators have to cut capital expenditures to service debt, shares of technology giants and mutual funds of private investors will come under pressure.
The era of cheap money for AI is ending
According to calculations based on BloombergNEF data, starting in 2027, data center builders will have to repay obligations worth $13.4 billion. When refinanced at current rates, interest expenses will rise by 2.85 percentage points, and additional costs will amount to at least $380 million.
This is only the first wave. Tax-exempt companies leasing AI chips — primarily CoreWeave and Nebius — will have to return $49 billion to creditors in 2030–2032. The peak of payments falls in 2030: almost $20 billion, of which about half are CoreWeave's obligations. Among the borrowers are also mining companies Iren, TeraWulf, and Hut 8, so the debt difficulties of the AI sector could directly affect the crypto market.
Creditors are already showing caution. CoreWeave bonds maturing in 2032 trade with a yield above 16%, while the yield on 10-year US Treasury securities has again exceeded 5%. Such a high premium attracts debt buyers, but at the same time signals deep market doubts about the company's ability to pay.
The key risk for the industry is the need to refinance hundreds of billions of dollars at the highest rates in decades. At the same time, the collateral is chips, which are becoming cheaper every year. The logical question: if creditors are already alarmed now, who will refinance these obligations in the future?
Why AI debts concern the entire market
Debt load is growing against the backdrop of record construction: spending on building data centers in the US in September reached $96.5 billion on an annualized basis — 90% more than a year earlier. The more facilities the industry builds on credit, the harder it will be to service debt at high rates.
Pressure is also mounting from another side: Ethereum co-founder Vitalik Buterin supported the movement against data center construction in order to buy time to prepare for AI risks — from cyberattacks to biosecurity threats. If such calls turn into restrictions, it will become harder for borrowers to recoup the loans they have already taken.
Private investors will also feel the blow: US equity funds have attracted $2.1 trillion since 2020, and Nvidia securities account for the largest share in the S&P 500 and Nasdaq indices. The more concentrated money is in the AI sector, the more painful a possible debt crisis in the industry will be for holders of index funds.
Cryptalist analysis: The debt overhang of the AI sector creates a systemic risk that the market still underestimates: the high yield on CoreWeave bonds reflects the real probability of refinancing at prohibitive rates, and the link through miners Iren, TeraWulf, and Hut 8 transmits this risk directly into the crypto industry. Institutional investors should view 2027–2032 as a window of heightened volatility at the intersection of technology and digital assets.