Michael Burry Says Big Tech Is Sitting on a $3 Trillion AI Time Bomb

A Familiar Warning From a Familiar Face

The man who predicted the 2008 housing collapse is sounding the alarm again — and this time, his target is the artificial intelligence spending frenzy gripping the world's biggest technology companies. Michael Burry, the hedge fund investor immortalized in The Big Short, published a detailed blog post this week arguing that the AI boom has quietly produced a mountain of financial risk that Wall Street is choosing to ignore.

Burry argues that Amazon, Meta Platforms, Alphabet, Microsoft, and Oracle have quietly amassed more than $3 trillion in largely off-balance-sheet commitments — risks he says the market has broadly overlooked. These aren't small accounting footnotes. They represent a staggering web of future financial obligations that don't always appear prominently on company balance sheets, making them easy for casual investors — and even professional analysts — to miss.

Breaking Down the $3 Trillion

Burry estimates the five hyperscalers have nearly $1.2 trillion in uncommenced lease commitments and more than $1.5 trillion in purchase commitments. Including special-purpose vehicles, guarantees, contingent backstops, and other obligations, he estimates the total exceeds $3 trillion. To put that in perspective, $3 trillion is larger than the entire GDP of France.

Burry wrote that "these liabilities, I say, are, in essence, in hypergrowth mode," arguing that they are growing much faster than the companies themselves. That's a crucial distinction. When a company's debts and obligations grow faster than its revenues or assets, the gap eventually has to close — and the closing is rarely painless. Burry also argues that circular financing among hyperscalers, AI labs, and chipmakers could artificially reinforce demand and revenue, creating a feedback loop that inflates the appearance of a healthy market.

The Hardware Problem Nobody Wants to Talk About

Burry's central concern is the mismatch between the long lifespan of data-center infrastructure and the rapid pace of AI chip development. Data centers can take three to five years to build and often involve lease terms stretching 13 to 20 years, while AI accelerators can see major changes in power density and cooling requirements within 12 to 18 months. In other words, companies are locking themselves into decades-long real estate commitments to house hardware that may be obsolete in a year.

Mainstream cloud providers typically set AI server depreciation lives at five to six years, but given that Nvidia's chip iteration cycle has compressed from 18–24 months to roughly 12 months, Burry believes the actual economic life of GPUs is closer to two to three years. He previously estimated that by stretching GPU useful lives to five to six years, hyperscale cloud providers would understate depreciation expenses by approximately $176 billion between 2026 and 2028. Meanwhile, Meta extended its server depreciation period from 4–5 years to 5.5 years — a move that alone added approximately $2.9 billion in book profits. Burry sees that kind of accounting flexibility as masking true costs.

Wall Street Is Not Doing the Math

Turning his attention to Wall Street, Burry critiques analysts for their overly optimistic assessments of these hyperscalers. He emphasizes that many have failed to properly account for maintenance capital expenditures and the economic depreciation of AI infrastructure, and specifically points to Oracle, whose stock price projections he believes may reflect an unrealistic level of optimism. The spending pace itself is breathtaking: institutional estimates project the five companies' combined 2026 capital expenditures will reach $769.2 billion — nearly double the 2025 figure — and could surpass $1 trillion by 2027. Microsoft alone is expected to spend $190 billion in capital expenditures in 2026, a year-over-year increase of roughly 130%.

Burry does not accuse the companies of fraudulent behavior but maintains that the current frenzy surrounding AI infrastructure poses substantial financial risks if demand begins to wane. "When the music's over, these off-balance sheet commitments become real liabilities very quickly," he warns, stressing that the resultant impairments, termination costs, and unused capacity could exert considerable financial stress on these firms. Given his track record, those are words worth taking seriously — even if Wall Street, for now, is still dancing.