‘Big Short’ Michael Burley warns about NVIDIA stock and AI ‘Tokenmaxxing’

AI For Business


Michael Burry warns that NVIDIA stock looks vulnerable and AI’s “tokenmaxxing” trend won’t continue in two recent Substack posts.

“The conditions for an aggressive decline are as strong as they have been in the history of stocks,” Barry wrote in a post on Monday.

The investor of “The Big Short” fame — known for making far-sighted decisions and dire predictions — added that Nvidia’s next decline could be “even more dramatic” than the previous three crashes (56% in 2018, 67% in 2021, and 43% in 2025).

Berry said his views are based not only on fundamentals but also on options and stock market volume. He noted that Nvidia’s stock volume on a 50-day moving average basis is “the lowest since 1999.” He added that there is a “lack of hedging activity” because it is cheaper to buy Nvidia put options than similar stocks.

He said that if Nvidia stock actually falls, he expects there will be “very few buyers on the way down” because there is insufficient structural demand and market makers have to pull out.

Burley warned in a Friday post that Nvidia’s customer concentration is “off the charts” and that the company could take a big hit to its bottom line if its biggest buyers walk away.

He noted that Nvidia’s recent earnings showed its top three customers accounted for 64% of its accounts receivable, up from 56% in the previous quarter.

Barry also noted that for the first time in 13 quarters, the company’s No. 1 customer accounted for more of Nvidia’s accounts receivable, but its share of sales declined.

“This is not quite a smoke bomb, but more like a finding that there was a finger on the trigger,” he wrote.

Burry suggested that Microsoft brought forward inventory that it “doesn’t really need” as it slows data center construction, either because it wanted to maintain priority for Nvidia’s next chip, or because Nvidia brought forward inventory to improve its bottom line.

He said it could pave the way for “evil bullwhip,” referring to how small changes in customer demand can reverberate through the supply chain and have a more pronounced impact on suppliers.

Mr. Barry has previously said he is betting on Nvidia and other high-flying AI stocks, and has repeatedly named the chipmaker as particularly vulnerable if the AI ​​boom falters.

Demand is moderate

Burley called “tokenmaxxing,” or companies encouraging employees to use AI models as much as possible, “a crazy, rushed, temporary phase.”

“Tokenmaxxing is not just a heavy use of AI, and it’s certainly not a sustainable use of AI,” he said. “It’s quota-driven, leaderboard-driven, executive-mandated overconsumption.”

Burley argued that this trend will not support the AI ​​industry in the long term, as “everyone is already using AI, and many are using it as desperately and as expensively as possible in this massive training phase.”

He warned that “the market is using the most expensive stage of AI deployment as if it were normal and indicative of future demand.”

“NVIDIA is releasing new chips every year, data centers are being built as quickly as possible, power plants are being built as quickly as possible, and external demand has already been compressed,” he added.

Nvidia did not immediately respond to a request for comment outside of normal business hours.