AI’s data-centre boom is piling up debt, and investors are getting nervous

The AI build-out is increasingly being financed with debt, not just the cash generated by the biggest technology companies. The Economist reports that nearly $850bn in AI-related borrowing was issued this year across companies tied to the boom, raising questions about who ultimately carries the risk.

Watch Desk analysis

What happened

The figures reported by The Economist span neocloud providers, major cloud companies, data-centre projects, chipmakers and utilities. Their combined balance-sheets carry more than $3tn of debt, the article says. For ten publicly listed neoclouds, borrowings total about $100bn, roughly three times annual revenue, and only two are profitable.

The article also describes financing that can be harder to follow than a straightforward company bond. Data centres, their chips and even power sources may be financed separately; some projects use special-purpose vehicles, while chip-backed loans use computing hardware as collateral. It cites $1.2tn in agreed leases that have yet to commence, alongside $1.5tn in supplier purchase commitments.

Why it matters

AI infrastructure is expensive, and borrowing spreads the bill across lenders and investors as well as the companies building it. Separate financing structures and guarantees between companies can make it harder to see where the exposure sits if projects fall short or borrowing costs rise.

That is not evidence that an AI crash is imminent. It is a reason to look beyond headline spending plans and ask whether the infrastructure can generate enough revenue to service the obligations attached to it.

Our read

The striking figure is not a prediction of collapse; it is the scale and complexity of the financing. When the boom’s foundations include leases, project vehicles and chip-backed loans, “how much debt?” is only the first question. “Who owes it, and who has promised to catch it?” is the more revealing one.

What to watch

• Whether borrowing costs keep rising for neocloud providers and other infrastructure builders.
• How much planned AI capacity translates into paying demand.
• Whether company guarantees and project financing make the eventual exposure clearer, or murkier.

Discussion spark: Should investors judge the AI infrastructure boom mainly by future demand, or by how much debt and contingent liability is already supporting it?

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.