Nine major technology companies, including Alphabet, Amazon, Meta, Microsoft, and Oracle, have accumulated approximately $3 trillion in AI-related financial commitments. These obligations are not recorded as liabilities on balance sheets but are disclosed in financial footnotes.
The total figure comprises $1.2 trillion in unstarted data-center leases and $1.9 trillion in purchase agreements for chips and infrastructure. This volume is five times higher than the $600 billion in capital expenditures reported by these firms over the past 12 months.
Companies utilize GAAP-compliant reporting and special-purpose vehicles to manage these obligations. For instance, Meta used a structure financed by Blue Owl Capital for its Hyperion project. A Nikkei Asia analysis noted that five hyperscalers hold $1.65 trillion in such hidden obligations, exceeding their $1.35 trillion in reported debt.
Individual company disclosures show significant growth in these commitments. Alphabet reported $811 billion in obligations as of June 30, 2026, up from $332 billion three months prior. Oracle’s commitments reached $273 billion, a 30-fold increase over four years, while Meta’s share is estimated at $420 billion.
Analysts warn that these off-balance-sheet commitments create refinancing risks and pressure on free cash flow. Because standard leverage metrics often exclude these footnote disclosures, the true financial position of these companies may be understated, especially if AI revenue growth fails to match infrastructure spending.
