Big Tech's AI Spending Runs on $1.65 Trillion in Off-Balance-Sheet Debt, While Small Businesses Spend Almost Nothing on AI

The Numbers Don't Match

Big Tech is spending close to $1 trillion this year on AI infrastructure, according to OilPrice.com, with even more projected for next year. A lot of that is funded by debt. Most of it isn't showing up where investors are looking.

A Nikkei investigation, reported by The Wire, found that five companies leading the AI race, Alphabet, Microsoft, Amazon, Meta, and Oracle, are sitting on roughly $1.65 trillion in obligations that do not appear on their balance sheets. That's more than the $1.35 trillion in debt these same companies officially report combined.

None of it is illegal. None of it is secretly hidden either. It's disclosed, according to The Wire, just buried in footnotes rather than the headline balance sheet numbers most analysts and reporters actually check.

Where the Debt Goes to Hide

The mechanism is simple once you understand it, per The Wire's reporting on the Nikkei findings. A company that wants to build a $10 billion data center has two choices: borrow the money directly and put it on the books, or set up a joint venture, take a minority stake, let a private credit fund finance construction, and sign a long-term lease to use the facility once it's built.

The second option functions almost identically to debt. The company is contractually obligated to make payments for years either way. But because it doesn't hold majority ownership, current accounting rules let the obligation sit off the primary balance sheet until the facility actually goes live.

The company-by-company numbers are eye-popping. Meta's off-balance-sheet exposure is roughly $420 billion, nearly triple its officially reported debt of about $140 billion, a figure Nikkei confirmed directly. Oracle's hidden obligations have grown more than thirtyfold in four years to roughly $273 billion, driven largely by its Stargate data center partnership with OpenAI, against reported debt of around $100 billion.

Amazon, Microsoft, and Alphabet's figures, roughly $350 billion, $350 billion, and $250 billion respectively, are estimates based on the pattern Nikkei established from the two companies it could verify directly. All five companies declined to comment on the findings, according to The Wire.

Meanwhile, Small Businesses Aren't Buying What Big Tech Is Selling

Data from Capital on Tap, one of the UK's biggest SME lenders, tracked spending across tens of billions of pounds in transactions and found AI adoption among small businesses, while growing, is nowhere near the scale needed to justify this spending.

Just 3.2% of UK small businesses spent money on AI company services in the second quarter of 2024. That's quadrupled to 12.8% by the second quarter of 2026, according to OilPrice.com's reporting on the Capital on Tap data.

Mean average spend per business jumped from £93 to £288 over that period. But the median, a better measure of the typical business, is just £75.60, because the mean gets pulled up by a small group of heavy spenders, the top 1% averaging £3,159 each.

Across all the transaction data Capital on Tap processes, AI tools account for just 0.1% of total card spending value. Extrapolated to the UK economy, that puts total AI services spending at roughly £4 billion a year, less than a tenth of Tesco's annual turnover.

Global AI spending estimates run somewhere between £100 billion and £150 billion a year, per OilPrice.com. That's real money, but far short of the near-$1 trillion Big Tech is set to spend on infrastructure this year alone, let alone the larger sum expected next year.

The Bull Case Deserves a Fair Hearing

The strongest argument in Big Tech's defense is that this is early-stage infrastructure investment, not unlike the railroads or the fiber-optic buildout of the dot-com era. Companies like Meta and Oracle are betting that enterprise and consumer AI usage will scale dramatically as models get cheaper and more capable, and that revenue will eventually catch up to capacity. Off-balance-sheet financing through joint ventures is also a long-established, legal structure used across industries, from airlines financing planes to utilities financing power plants, not a novel accounting trick invented for AI.

The mismatch between current usage and current spending is significant. Businesses' AI spending would need to grow by orders of magnitude, not incrementally, to make these infrastructure bets pay off, according to OilPrice.com's analysis of the Capital on Tap figures.

It is growing. It's growing fast. It just isn't growing at the pace $1.65 trillion in obligations requires.

No regulator has announced an investigation into how these five companies structure or disclose this debt. No accounting body has proposed a rule change forcing these obligations onto primary balance sheets. The question for investors, and for anyone watching where this money eventually has to come from, is what happens if enterprise AI adoption stalls before these multi-year lease and capacity contracts start coming due.