Big Tech’s artificial intelligence infrastructure race has quietly accumulated $1.65 trillion in future lease and purchase obligations that most investors are not fully accounting for.
A Nikkei investigation of SEC filings found that Microsoft (NASDAQ: MSFT), Meta Platforms (NASDAQ: META), Amazon (NASDAQ: AMZN), Alphabet (NASDAQ: GOOG), and Oracle (NASDAQ: ORCL) have each built enormous off-balance-sheet commitments tied to data centers, networking equipment, and power infrastructure.
These obligations are perfectly legal under U.S. accounting rules because they represent future contractual commitments rather than traditional borrowings, meaning they receive far less investor attention than reported debt figures.
The scale of the gap between reported and off-balance-sheet debt is striking across all five companies, with some showing obligations several times larger than what appears on their formal balance sheets.
Meta Platforms carries approximately $140 billion in reported debt, while its estimated off-balance-sheet obligations reach roughly $420 billion, a figure approximately three times its disclosed liabilities.
Oracle’s off-balance-sheet commitments have expanded roughly 30-fold over just four years, growing to an estimated $273 billion against approximately $100 billion in reported debt as the company races to build AI capacity.
Microsoft and Amazon each carry an estimated $350 billion in off-balance-sheet obligations respectively, while Alphabet’s commitments reach roughly $250 billion despite only $30 billion in reported debt.
Much of this financing flows through long-term leases, project financing, and private credit arrangements that spread obligations across developers, insurers, and institutional lenders rather than sitting directly on corporate balance sheets.
That structure allows technology companies to keep leverage ratios appearing manageable at the precise moment their capital spending is reaching historically unprecedented levels.
The accounting treatment shifts once facilities begin operating, at which point lease obligations move onto financial statements, depreciation begins, and any underutilized infrastructure can generate impairment charges if AI demand falls short of projections.
These companies do generate substantial cash flows that provide meaningful support for long-term commitments, with Microsoft alone producing nearly $100 billion in operating cash flow over the past year.
The concern is not solvency but transparency, as investors relying solely on headline debt figures are missing a significant portion of the total financial commitments already made to AI infrastructure.
If enterprise AI adoption continues expanding at its current pace, the investments could generate attractive returns that justify every dollar committed, which is precisely what management teams across all five companies are betting on.
If demand disappoints, however, data centers that never earn projected returns would generate write-downs affecting shareholders, while lenders, insurers, and private-credit investors financing the construction would also absorb losses.
As earnings season unfolds for Meta, Microsoft, Alphabet, and Amazon within days, investors who read only the headline balance sheet figures will see a comfortable picture that the footnotes of SEC filings tell very differently.
