Buried in Meta Platforms’ audited financials is a number that does not appear on the balance sheet: roughly $420 billion in AI-related obligations, according to an analysis by the Wall Street Journal published Aug. 17. That is nearly three times the $83.7 billion of debt the company reports on its books, and auditor Ernst & Young flagged the structure as the largest of its kind it has encountered, people familiar with the matter said.
The gap between the two numbers is the story. Meta, like its hyperscale peers, has been financing its AI buildout through finance leases and supplier financing, arrangements that keep the obligations off the debt line that investors watch most closely. The reported balance sheet looks clean; the footnotes tell a different story, and the difference is now large enough that auditors are calling it out in the most conspicuous terms available to them.
The mechanics are familiar to anyone who has followed the data center boom. Companies lease servers, buildings, and power infrastructure from specialized owners rather than buying them outright, and suppliers extend credit for equipment purchases with repayment tied to usage. The structures let a company expand capacity at speed without inflating the debt ratios that credit ratings and loan covenants are built on. What is unusual here is the size: $420 billion is a figure that dwarfs the entire annual revenue of most companies in the world.
The obligations are real contracts, not accounting fiction. Meta is on the hook for lease payments, equipment costs, and supplier financing across years, and the total exposure has grown as the company has committed to data centers and chips at a pace that few expected even eighteen months ago. The question is not whether the money will be paid, but when the market will start pricing the payments into Meta’s valuation.
Investors have been slow to adjust. The headline balance sheet still shows $83.7 billion in debt, a manageable figure for a company of Meta’s cash flow, and the stock trades on the strength of its advertising business and its AI ambitions. But analysts said the true ratio of obligations to cash flow, the one that counts everything, is several times higher than the reported one, and that difference becomes material the moment growth slows or interest rates stay high.
The timing adds pressure. Meta is defending itself in California courts against lawsuits over its handling of minors on its platforms, a legal front that carries its own costs and reputational risk. Fighting that fight while carrying nearly half a trillion dollars in obligations off the books is a combination that executives would prefer not to discuss in the same sentence, but the two are now connected in the minds of analysts who track the company’s risk profile.
The pattern is industry-wide. Microsoft, Amazon, and Google have all used similar financing structures to fund data centers without loading their balance sheets, and the AI boom has pushed every major player deeper into the same playbook. What EY’s flag on Meta says is that the practice has reached a scale where the difference between reported and actual obligations is no longer a footnote-level detail but a structural feature of how the industry is financed.
Regulators and standard-setters are watching. Accounting rules already require disclosure of off-balance-sheet arrangements, but the disclosure lives in the notes, where few investors read it. The EY flag, which describes the structure as the largest the firm has seen, is a signal that auditors themselves are uneasy about how the obligations are presented, and that unease tends to flow into future rule changes.
For Meta, the $420 billion figure is a warning about the shape of its future. The company’s AI spending has bought it a position in the race for frontier models and a reason for investors to stay patient with its core business. It has also bought a mountain of obligations that will be paid for years, whatever happens to the models. The market will eventually price the difference between the two books, and the only question is when.
The accounting debate turns on how the obligations are classified. Lease standards require capitalization of most long-term leases, but structures can be designed so that commitments fall outside the thresholds, and supplier financing arrangements sit in the payables rather than the debt line, which keeps them out of the ratios that rating agencies and lenders watch. The result is a reported balance sheet that understates the company’s true commitments, and the EY flag is an auditor’s way of saying the gap has become material.
The comparison with peers sharpens the question. Microsoft, Amazon, and Google have all used similar financing, and their investors have generally accepted it as a feature of the industry’s buildout, but none has carried obligations of this relative size. If Meta’s commitments were added to its reported debt, its debt profile would move from conservative to aggressive by most measures, and the shift in perception would change the cost of its future borrowing.
What would change the picture? A slowdown in AI demand, a rise in financing costs, or a downturn that leaves the leased assets worth less than the obligations would turn the footnotes into a balance-sheet problem. Analysts said the arrangements are manageable as long as growth continues, which is another way of saying the $420 billion is a bet on the same AI expansion that Meta is spending to create.
