Michael Burry says Nvidia (NVDA) is sidestepping the most important questions facing AI investors, noting that the company responded to criticisms he never made while avoiding the core issue of how rapidly its chips lose economic value.
In a new thread on X, the “Big Short” investor responds to an Nvidia memo distributed to institutions and circulating on social media platforms that addresses Burry’s claims on chip depreciation.
In the memo, Nvidia says that Burry claimed that Nvidia is depreciating power, plant and equipment (PPE) more slowly than peers to understate depreciation expense and boost net income. The chipmaker highlights that it discloses useful life estimates consistent with those of peers. Nvidia also says that its customers depreciate GPUs over four to six years based on real-world longevity and utilization patterns.
“Older GPUs such as A100s (released in 2020) continue to run at high utilization and generate strong contribution margins, retaining meaningful economic value well beyond the 2 to 3 years claimed by some commentators.”
But Burry says he never called out Nvidia on the PPE depreciation front.
“I made no such claim because no such claim is plausible. Nvidia is a fabless company that only designs chips. It has very little capital expenditure and very little depreciation. No one cares about Nvidia’s own depreciation. One straw man burnt.”
While Burry did short NVDA, he said hyperscalers like Meta are the ones understating the depreciation of Nvidia chips, not the chipmaker itself.
The famed short-seller also says Nvidia set up a second misdirection by defending the longevity of older GPUs. He says the response ignores the economic reality that matters to investors today.
“The second rebuttal is about Nvidia’s customers using 4 to 6-year-old chips. Well, another straw man, as the chart I posted above and on X is clearly about 2026 to 2028.”
According to Burry, the real risk to investors today is economic obsolescence. He points to Nvidia’s own claims that Blackwell is far more efficient than prior chips, which he says renders older hardware costly and difficult to justify, regardless of whether it still functions inside a data center.
“The 2020 to 2024 Nvidia A100 chip takes 2 to 3 times more power per FLOP than a 2022 to 2025 H100 chip. In turn, Nvidia itself claims that the 2024 to 2025 Blackwell is 25 times more energy efficient than the H100.”
Burry adds that next-generation data centers are being built for specific cooling and electrical specifications, narrowing the conditions under which older GPUs can operate and further reducing their economic lifespan.
“From 2025 on, expensive data centers are being built with very specific cooling and electrical specs for a specific generation of Nvidia chips…
An older chip will be a residual value unit that is an energy hog and extremely costly, not to mention difficult to run outside a same-vintage data center. Still, one can say they are ‘fully utilized’ to the degree they can be today. Still that is the direction you are going, chances are, you are forced to do it, and it is not pleasant.”
Disclaimer: Opinions expressed at CapitalAI Daily are not investment advice. Investors should do their own due diligence before making any decisions involving securities, cryptocurrencies, or digital assets. Your transfers and trades are at your own risk, and any losses you may incur are your responsibility. CapitalAI Daily does not recommend the buying or selling of any assets, nor is CapitalAI Daily an investment advisor. See our Editorial Standards and Terms of Use.