The total includes $1.2 trillion in unstarted leases and $1.9 trillion in purchase commitments.
- Burry alleges hyperscalers use overly long depreciation schedules for rapidly aging AI chips and servers.
- He warns of three-way “compression” across AI demand, corporate earnings and financing.
- Burry also argues that circular financing among hyperscalers, AI labs and chipmakers could artificially reinforce demand and revenue.
Michael Burry revived his warnings about the AI boom after an analysis found nine technology giants had amassed around $3 trillion in off-balance-sheet commitments, arguing he had highlighted the risks months before they reached the mainstream.
Big Tech’s Hidden $3 Trillion Tab
“Well, you could have heard it first, months ago, 2025 even,” the “Big Short” investor said on X while sharing a WSJ report. Alphabet, Amazon, Meta, Microsoft, Oracle, Nvidia, Broadcom, SpaceX and Advanced Micro Devices together reportedly disclosed $3 trillion in commitments largely for AI infrastructure.
The obligations are disclosed in filing footnotes but are not yet recognized as liabilities on the face of corporate balance sheets. They include $1.2 trillion in leases that have not started and $1.9 trillion in purchase commitments covering chips, energy and data-center infrastructure. This compares with $600 billion in capex over the companies’ latest reported 12-month periods.
Alphabet alone reported $811 billion in purchase and contractual commitments as of June 30, while Meta disclosed $347 billion in leases that have yet to commence. Many of these agreements are apparently difficult to cancel. If AI demand disappoints, the companies could be left paying for expensive infrastructure that fails to generate sufficient returns.
Why Burry Thinks AI Spending Could Become An Earnings Problem
Burry returned to X in late October 2025 after a roughly two-year hiatus and, by early November, had disclosed or hinted at bearish put positions involving Nvidia and Palantir. He alleged that hyperscalers were understating depreciation by assigning five- to six-year useful lives to Nvidia chips and servers whose economic cycle, in his view, is closer to two or three years.
Burry called it “one of the more common frauds of the modern era,” estimating that the practice could suppress depreciation expenses by $176 billion between 2026 and 2028. By his calculations, it could inflate Oracle’s 2028 earnings by 26.9% and Meta’s by 20.8%.
Burry Warns Of 3-Way AI Compression
Through late 2025 and earlier this year, Burry’s focus widened from chip depreciation to the financial structure supporting the AI buildout: circular financing, weakening cash flow and increasingly complex borrowing. He asked hyperscalers: “When does the spending for AI data center buildout actually end?”
In May, Burry published a Substack post titled, “Tracepalooza & the Bezzle,” arguing that training and benchmarking activity could exaggerate near-term AI demand and create a bullwhip effect. He highlighted Nvidia’s $182 billion in forward purchase commitments, offshore financing, construction assets that do not begin depreciating until placed in service, and hundreds of billions of dollars in hyperscaler off-balance-sheet exposure.
Burry called the potential reckoning a three-way “compression”: demand weakens, earnings fall as expenses catch up and financing tightens as capital becomes more expensive.
Oracle And Nebius Are Burry’s Latest AI Shorts
Recently, Burry was again targeting circular financing, arguing that capital moving among hyperscalers, AI labs and chipmakers could artificially reinforce revenue and demand. He also disclosed that he had shorted Oracle at around $145 per share and taken a larger short position in Nebius at $212 per share.
Burry said AI companies had “gorged themselves” on off-balance-sheet liabilities, including backstops, uncommenced leases and purchase commitments, leaving them vulnerable if demand, earnings and financing compress simultaneously.
For updates and corrections, email newsroom[at]stocktwits[dot]com.<