AI infrastructure debt boom fuels spending, analysts warn on demand

Big technology companies and a growing cohort of smaller firms have taken on billions of dollars in debt to build artificial intelligence infrastructure over the past year, driving rapid stock gains for some companies even as analysts warn the spending may outpace demand and create financial, labor and environmental strains. The surge in AI-related investment accelerated through 2024 and into 2025, with industry estimates ranging from roughly $250 billion spent on AI in 2024 to projections above $350 billion in 2025 and forecasts of much larger cumulative spending by 2030.

Companies are financing massive data center and chip builds through traditional borrowing and creative deals that pledge future customer revenue as collateral. Oracle, which has reported about $82 billion in long-term debt, entered what has been described as a multiyear contract with a major AI developer that analysts say would require roughly $25 billion in annual borrowing to service associated commitments. Smaller infrastructure providers have taken similar bets: CoreWeave and Nebius have reportedly borrowed against multibillion-dollar customer contracts, locking their balance sheets to the ability of a...