Artificial intelligence needs vast data centres full of expensive chips, and building them costs more than even the richest technology companies want to fund from their own cash alone. Increasingly, the money is borrowed. That borrowing, often called AI debt, has become one of the questions investors are asking about how long the AI boom can last.

Where the money comes from

Large technology companies issue bonds to fund data centre building. Specialist cloud companies that rent out computing power borrow heavily, sometimes using the chips themselves as security. Private credit funds lend to data centre projects. Some deals are structured through separate companies so that the debt does not sit directly on the parent’s balance sheet.

Why companies borrow

Demand for AI computing has grown faster than companies can build. Borrowing lets them move quickly without waiting for profits to pile up. When interest rates were low, debt was cheap. With rates higher, it costs more, but many companies still see AI as too important to slow down. Our explainer on AI capex covers the spending side.

Why markets worry

The first concern is whether the returns will justify the cost. If AI revenue grows more slowly than expected, companies may struggle to earn enough to service their debt. The second is that the assets backing some loans, such as chips, can lose value quickly as newer models arrive. The third is complexity: when debt is spread across funds, partners and separate vehicles, it is harder for investors to see the full picture.

Higher interest rates add to the pressure, because refinancing debt becomes more expensive.

Who is exposed

Bondholders, private credit investors and banks that lend to the sector are directly exposed. Shareholders are exposed too, because heavy debt can limit a company’s flexibility. Suppliers, including chipmakers, rely on customers continuing to spend, so trouble at heavily indebted buyers could reduce orders. Our guide to how banks are exposed to the AI boom covers the lending side.

Signs to watch

Watch bond yields on technology company debt compared with government bonds, comments on financing in company results, any delays or cancellations of data centre projects and warnings from regulators about financial stability.

The takeaway

AI debt is the borrowing that funds the AI build-out. It is not a problem in itself, but if returns disappoint or rates stay high, it could turn a slowdown into something more painful.

Understanding where the risks sit beneath a popular theme is a valuable skill. Our free Trader Assessment can help you see how well you spot them.

Sign up to Our Mailing List

Join our mailing list to gain access to the latest news & research.

    Samuel & Co. In The News