Big Tech borrows heavily to build AI data centers worldwide

In short: Big tech companies and data center operators are borrowing huge sums to build AI data centers, and some experts worry the debt could strain the wider economy.

What's going on

Companies are racing to build AI-ready data centers, which are the warehouse-like buildings filled with computers that run AI systems. These projects are expensive, so many firms are paying for them with borrowed money, not just cash on hand.

Industry data cited by analysts shows data center debt issuance rose from about $92 billion in 2024 to roughly $182 to $183 billion in 2025. Hyperscalers, meaning the biggest cloud companies like Amazon, Microsoft, Google, Meta, and Oracle, issued around $121 billion in bonds in 2025 alone. Bonds are IOUs that companies sell to investors and then pay back over time, plus interest.

Oracle is a high-profile example, but it is not the only one. Reporting says Oracle issued about $25.75 billion in bonds in a single year to support its AI data center build-out, and analysts have estimated it may need to borrow about $25 billion per year for several years. Some borrowing is also harder to spot, because it can sit in long-term leases and other commitments that do not look like regular debt (like promising to pay rent for years on a building you do not own).

What to watch

The big question is whether AI services will earn enough money to cover these fixed payments. If demand is weaker than expected, or if interest rates stay high, refinancing could get more expensive, and problems could spread to lenders and investors like pension funds.

Source: NYTimes