The battle between Nvidia and one of its most vocal naysayers, investor Michael Burley, is heating up.
Following a series of social media posts by “Big Short” investors claiming that the artificial intelligence investment boom is replicating the 1990s dot-com bubble centered on Nvidia, the chipmaker secretly circulated a private memo to analysts that apparently name-checked Berry in an effort to push back on many of Berry’s claims.
“NVIDIA emailed a memo to Wall Street sell-side analysts pushing back on my arguments regarding SBC and depreciation. I stand by my analysis,” Barry said in a post on Substack, referring to stock compensation. “I don’t claim that Nvidia is Enron. It’s clearly Cisco.”
Barry has repeatedly warned that today’s AI infrastructure frenzy reflects the telecom expansion of the late 1990s far more than dot-com sweepstakes investors remember. He cited large capital spending plans, longer depreciation schedules and soaring valuations as evidence that the market is once again misinterpreting a supply boom as durable demand.
The NVIDIA memo, first reported by Barron’s, disputed Berry’s claims about the depreciation period.
In response to Burry’s accusation that customers are exaggerating the useful life of Nvidia’s graphics processing units to justify runaway capital expenditures, Nvidia counters that customers depreciate their GPUs over four to six years based on their actual lifespan and usage patterns.
Nvidia added that older GPUs, such as the A100 released in 2020, continue to operate at high utilization rates and retain meaningful economic value well beyond the two to three years that critics claim.
The memo also rejects Berry’s suggestion of “rotating funding,” saying Nvidia’s strategic investments represent a small portion of its revenue and that the AI startup is primarily funded by outside investors.
Cisco today
Burry said he believes NVIDIA now occupies the same position as Cisco held from 1999 to 2000 as a major hardware supplier that drove a major capital investment cycle.
Just as telecommunications companies spent tens of billions of dollars installing fiber optic cables and purchasing Cisco equipment based on predictions that “Internet traffic will double every 100 days,” today’s hyperscalers are pledging about $3 trillion in spending on AI infrastructure over the next three years, Barry said in a Substack newsletter.
The crux of his Cisco analogy is that supply has been overbuilt in response to demand that was much lower than expected. In the early 2000s, less than 5% of the U.S.’s fiber-optic capacity was operational, Barry said. The industry’s current belief in limitless AI demand is based on similarly optimistic assumptions about data center power and GPU longevity, he said.
“And once again, there is Cisco at the center of it all, with the picks and shovels for everyone and the expansive vision that goes with it. Its name is NVIDIA,” Barry wrote.
—CNBC’s Michael Bloom contributed reporting.
