Financial Times highlights hidden lease risks in AI data center boom

In short: As big tech firms build more AI data centers, creditors are focusing on large lease commitments that can be hard to spot in headline debt numbers.

What's going on

A Financial Times opinion column argues that investors who lend money, like bond buyers, are taking a closer look at how much major tech companies may owe for AI infrastructure. This includes not just traditional debt, like bonds, but also long-term leases on data centers (large buildings full of computers that run online services and AI).

The column says leases can act a lot like debt because they often cannot be canceled. It compares lease payments to other fixed payments, like interest on a loan. Credit analysts often adjust company finances to treat these leases as part of the company’s real obligations, even if they are not fully visible on the main balance sheet.

Meta is used as a key example. The column notes Meta issued $25 billion in bonds earlier this year, which is easy for investors to count. But Meta also disclosed about $347 billion in future lease obligations for projects that have not started yet, compared with $84 billion of on-balance sheet debt reported at the end of 2025.

What to watch

The column warns that lease details are often clearer only in annual reports, and many data center projects are still being built. That means future filings could reveal larger commitments than some investors expect, which could affect how expensive it is for these companies to borrow money.

Source: Financial Times