Five US tech giants are sitting on $1.65T in hidden AI debt, Nikkei study finds

Meta's $420 billion in off-balance-sheet AI debt leads $1.65 trillion hidden total at five tech giants

Across five major US tech companies, off-balance-sheet obligations have reached an estimated $1.65 trillion, a total that has grown eightfold in roughly four years and is driven almost entirely by the AI infrastructure buildout, according to a Nikkei study published July 21. 1

These hidden obligations now exceed the five companies' actual on-balance-sheet debt, making it harder for investors to assess their real financial exposure. 1 The Nikkei study names Meta and Oracle among the companies carrying the largest off-balance-sheet loads. 1

When a company signs a multiyear lease on data center capacity or locks in GPU supply contracts, those commitments can sit outside the balance sheet. They appear as operating obligations or footnotes in regulatory filings rather than as debt. The cash has not been spent yet. But the obligation is binding, and it is growing at a pace that has no precedent in these companies' financial history.

Eightfold in four years. That growth curve maps directly onto the generative AI spending cycle that began in 2022. The companies at the center of the Nikkei analysis are building out GPU clusters, commissioning dedicated data centers, and committing to years of hardware and facility costs through the contracts driving this number.

Meta's situation is the most extreme. Its off-balance-sheet obligations are nearly triple its transparent debt, meaning the number investors see when they check Meta's reported borrowings captures less than a third of what the company has committed to pay. 1

For the investing public, this is a structural visibility problem. Balance sheets exist so that creditors and shareholders can quantify what a company owes. Debt-to-equity ratios, interest coverage metrics, and credit ratings all derive from that information. When $1.65 trillion in commitments across five companies lives outside that framework, every standard tool for measuring financial risk operates with an incomplete picture.

The Nikkei study frames this as a transparency concern. 1 That framing is accurate but understates what is at stake. The current AI infrastructure buildout is being financed partly through obligations that do not appear where investors are trained to look.

Consider the other side of these transactions. When NVIDIA books GPU revenue, the buyers of those GPUs absorb multiyear payment commitments. The Nikkei analysis indicates that five of those buyers have structured their obligations in ways that keep the full magnitude off their balance sheets. 1 The revenue shows up in earnings reports. The corresponding liability does not.

This is the cost dimension of AI spending that most coverage skips. The revenue side dominates headlines: new AI products, ad targeting improvements, enterprise contracts. The infrastructure commitments driving all of that accumulate in accounting structures that keep them separate from reported debt.

These contracts do not expire if AI revenue disappoints. Data center leases run for years. GPU supply agreements lock in pricing and volume. If returns arrive more slowly than expected, the obligations remain.

For anyone tracking AI investment through public filings, the gap between what these companies have spent and what they have committed to spend is the number that matters. The reported figures tell one story. The off-balance-sheet commitments tell another. Across five companies, that second story totals $1.65 trillion and counting.

The Nikkei study does not argue that AI spending is misplaced. It shows that the current accounting framework makes it impossible to judge. When off-balance-sheet obligations dwarf what appears on the balance sheet, the question is not whether AI will deliver returns. It is whether anyone holding the stock will see the bill coming.