The fight between Nvidia and one of its loudest naysayers, investor Michael Burry, escalated.
Following the “Big Short” investor's social media posts in which he argued that the artificial intelligence investment boom was replicating the 1990s dot-com bubble with Nvidia at the center, the chipmaker quietly distributed a private memo to analysts that specifically named Burry to rebut many of his claims.
Nvidia's seven-page response to “Questions and Claims We Received” began by citing “Michael Burry on Twitter/X” as the first collection of source documents that the company sought to refute.
Burry, for his part, responded in a post on Substack: “Nvidia emailed a memo to Wall Street sell-side analysts rejecting my arguments.” [stock-based compensation] and depreciation… I stand by my analysis. I'm not saying that Nvidia is Enron. It's clear Cisco.”
Burry has repeatedly warned that today's hype around AI infrastructure reflects the expansion of telecommunications in the late 1990s far more than what investors remember from the dot-com disaster. He pointed to massive capital spending plans, extended depreciation schedules and rising valuations as evidence that markets are once again confusing a supply boom with sustained demand.
The Nvidia memo, first published by Barron's, responded to Burry's criticism of Nvidia's stock-based compensation dilution and share buybacks.
“NVIDIA has repurchased $91 billion of stock since 2018, not $112.5 billion; Mr. Burry appears to have improperly accounted for RSU taxes,” the memo said, referring to restricted stock units. “Employee equity awards should not be confused with buyback program performance. NVIDIA's employee compensation is consistent with competitor compensation. Employees benefiting from a rising stock price does not mean that the original equity awards were inflated at the time of grant.”
The memo also disputed Burry's claims about the depreciation period. To Burry's accusation that customers overestimate the useful life of Nvidia's graphics processors to justify rapid capital expenditure, Nvidia counters that its customers depreciate GPUs over four to six years based on real life longevity and usage patterns.
Nvidia added that older GPUs like the A100 that launched in 2020 continue to run at high utilization rates and retain significant economic value well beyond the two to three years claimed by critics.
The memo also rejects Burry's suggestion of “circular financing,” explaining that Nvidia's strategic investments represent only a small portion of revenue and that AI startups predominantly raise capital from outside investors.
The Cisco of today
Burry said he believes Nvidia now occupies the same position as Cisco – the key hardware supplier that drove a massive capital investment cycle – in 1999-2000.
Just as telecom companies have spent tens of billions of dollars laying fiber-optic cables and buying Cisco equipment based on projections that “internet traffic is doubling every 100 days,” today's hyperscalers are promising nearly $3 trillion in AI infrastructure spending over the next three years, Burry said in a Substack newsletter.
The gist of his Cisco analogy is that excessive supply meets far less demand than expected. In the early 2000s, less than 5% of U.S. fiber capacity was operational, Burry said. Today, he believes the industry's belief in limitless AI demand is based on similarly optimistic assumptions about data center performance and GPU longevity, he said.
“And once again there's a Cisco at the center of it all, with the picks and shovels for everyone and the expansive vision to go with it. His name is Nvidia,” Burry wrote.
—CNBC's Michael Bloom contributed reporting.