Big Tech's off-balance-sheet AI debt reaches an estimated $1.65 trillion, analysis finds

Off-balance-sheet AI liabilities at five US hyperscalers — Alphabet, Amazon, Meta, Microsoft and Oracle — have swelled roughly eightfold in about four years to an estimated $1.65 trillion, a new financial analysis found. That figure exceeds the five companies’ combined reported on-balance-sheet debt of about $1.35 trillion.

The liabilities stem largely from long-term data-center lease commitments and GPU and server supply contracts that have been signed but not yet delivered or brought online. Under current accounting rules, those obligations do not appear on the balance sheet, leaving investors with an incomplete view of how much the AI build-out has been financed on credit.

Meta carries the starkest gap, with about $420 billion in off-balance-sheet debt, nearly triple its reported figure, according to the analysis.

The finding matters because it reframes the AI capital-spending boom as a leverage story rather than a pure cash-flow one. The same five companies are still guiding to a combined roughly $725 billion in AI capital spending in 2026, up about 77 percent from 2025. Those commitments keep the off-balance-sheet total climbing.

The estimate is an outside analysis of public filings, not a figure the companies report or endorse, and accounting treatment of leases and supply deals can vary by contract. None of the five has disputed the calculation.

Whether these liabilities stay in the footnotes or force clearer disclosure will depend on how regulators and auditors treat the swelling lease-and-supply commitments as the 2026 spending lands.