The AI boom increasingly runs through a data-center debt machine

Vantage Data Centers is seeking roughly $2bn in additional loans after borrowing about $48bn since 2025, according to the Financial Times. The numbers make visible an underappreciated feature of the AI boom: cloud demand is being translated into a vast, leveraged real-estate and energy buildout.

The financing works because data centers can resemble infrastructure assets. Long contracts with large technology tenants support predictable cash flows, while buildings, equipment and land offer collateral. But AI changes the familiar risk model. Compute hardware ages quickly; power can arrive later than the building; cooling requirements shift; and a small group of hyperscalers accounts for much of the demand.

That loop accelerates capacity, but it can also conceal correlated assumptions. If tenants slow spending, renegotiate, consolidate workloads or favor a different technical design, many projects may reprice at once. A contract reduces demand risk; it does not abolish counterparty, refinancing or obsolescence risk.

The key AI indicator is no longer venture funding alone. Watch loan covenants, tenant concentration, contracted-versus-energized capacity, refinancing schedules and who ultimately holds the debt. The software narrative is being financed like infrastructure, while the underlying technology still changes like software.