Michael Burry isn't backing down. The investor who famously called the housing bubble is now going after Big Tech's accounting practices, and he's getting louder about it. In his latest salvo, Burry accused major AI companies and hyperscalers like Meta Platforms Inc. (META) and Oracle Corp. (ORCL) of inflating their earnings through what he calls accounting gimmicks.

The Fraud Accusation

On Wednesday, Burry took to X and escalated things considerably. He shared a chart showing relationships between AI companies and accused them of engineering revenue through circular funding loops. According to Burry, this chart will eventually be seen "as a picture of fraud, not a flywheel."

"Every company listed below has suspicious revenue recognition," he wrote, claiming that "true end demand is ridiculously small" and that "almost all customers are funded by their dealers." In other words, Burry thinks the AI boom is built on companies essentially funding their own customers to buy their products.

The Depreciation Debate Gets Technical

Burry also doubled down on his earlier criticism that big-tech companies are understating depreciation on their AI capital expenditures. His argument? Just because old chips are still running doesn't mean they're worth what companies say they're worth on the balance sheet.

"The idea of a useful life for depreciation being longer because chips from more than 3-4 years ago are fully booked confuses physical utilization with value creation," he explained. "Just because something is used does not mean it is profitable. GAAP refers to economic benefits."

To illustrate his point, Burry compared AI chips to aging airplanes. Airlines keep old planes around for overflow during Thanksgiving or Christmas, he noted, but those assets are only "marginally profitable" and not worth all that much.

He got specific with the numbers. Nvidia Corp.'s (NVDA) A100 GPUs, released in 2020, "take 2-3x more power per FLOP (compute unit) so cost 2-3x more in electricity alone than H100s." And Nvidia claims the H100 is 25x less energy efficient than Blackwell for inference, which underscores how quickly the economics deteriorate as chip generations advance. This, Burry argues, makes it hard to justify those longer depreciation timelines for older hardware.

The Pushback From Prominent Voices

Earlier this month, Burry revealed he was betting against Palantir Technologies Inc. (PLTR) and Nvidia with put options. Initial reports wildly overstated the position as $912 million against the AI trade, but Burry clarified his actual exposure is $9.2 million. He holds 50,000 put options against Palantir with a $50 strike price, expiring in 2027.

Since then, leading analysts and market experts have questioned whether Burry's depreciation concerns really justify these trades. Steve Eisman, who shared the spotlight with Burry in "The Big Short," was blunt in his assessment: "I just don't think his concerns matter that much." Eisman downplayed the impact of extending depreciation schedules, saying it won't meaningfully affect the long-term outcomes of AI investments.

David Friedberg, a former Alphabet Inc. (GOOG) executive, was even more direct. He said Burry has "got this wrong," pointing out that changes to depreciation schedules simply reflect the reality of how data centers operate today. Chips that are seven and eight years old are still running at full utilization, Friedberg noted, which suggests the extended depreciation timelines might actually be appropriate.

So who's right? Burry sees accounting manipulation and a house of cards. His critics see normal business evolution and legitimate use of aging hardware. The debate cuts to the heart of how we value AI infrastructure and whether the boom is built on solid ground or creative bookkeeping. Either way, with millions on the line in his put positions, Burry is clearly confident in his thesis.