Artificial intelligence infrastructure builders are confronting a sharpening debt refinancing challenge as billions of dollars in fixed-rate obligations approach maturity. BloombergNEF estimates that about $13.4 billion in fixed-rate data center debt comes due in 2027. Refinancing at recent Treasury and bond yields could lift borrowing costs by roughly 2.85 percentage points and add at least $380 million in annual interest expenses.
The pressure is even larger for neocloud companies, which face approximately $49 billion in maturities from 2030 to 2032, with annual peaks near $20 billion. At the same time, at least $1 trillion in data center-related debt is scheduled to mature over the next five years, much of it concentrated among a relatively small group of firms.
Credit markets are turning more cautious after a sharp selloff in technology equities. Some banks that had recently been active in AI-related financing have become hesitant to commit to new data center deals, including arrangements backed or guaranteed by leading technology companies. Lenders are stress-testing cash flows under weaker economic scenarios and scrutinizing customer commitments and operational timelines more closely.
Goldman Sachs Research still expects large hyperscale cloud operators to issue around $420 billion in debt in 2027, up from an estimated $250 billion in 2026. However, elevated U.S. Treasury yields and tighter lending conditions may raise refinancing costs and slow construction activity, potentially delaying projects and constraining new AI capacity.
