Meet the man behind Michael Burry’s big AI short

Michael Burry is turning his attention to artificial intelligence, and new reporting from CNBC’s Yun Li shows how that view is being shaped by research from Phil Clifton, his long-time colleague and the successor to his investment operation.

In a segment aired on Nov. 25, Li outlined Clifton’s argument that while AI adoption is real and widespread, the underlying economics of the infrastructure boom are not yet justified.

Li noted that Burry, known for calling the U.S. housing collapse before 2008, recently disclosed bearish positions against Nvidia and Palantir, buying thousands of put options at the end of September.

His posts on X have been warning about “big tech euphoria” and the pace at which valuations have expanded.

Palantir CEO Alex Karp pushed back in an interview: “As far as I can tell, the two companies he’s shorting are the ones making all the money, which is super weird,” he said. “The idea that chip sentimentology is what you want to short is bat-shit crazy.”

Li said Burry hasn’t shifted, even after Nvidia reported another very strong quarter. Instead, he is leaning into Clifton’s thesis. Clifton argues that while tools like ChatGPT are widely used, the revenue side of the equation is still small relative to the spending required to support AI workloads.

Li summarized the disconnect: “OpenAI is expected to bring in a little over $20-billion this year… but big tech companies have quadrupled their investment to nearly $400-billion to build out AI.”

Clifton’s view, she said, is that: “Just because a technology is good for society or revolutionizes the world doesn’t mean it’s a good business proposition.”

Clifton draws a comparison to the early-2000s telecom boom, where companies laid far more fibre-optic capacity than demand ultimately required. He believes cloud providers may be repeating that pattern, pouring capital into GPUs and data-center infrastructure long before there is evidence that revenue will catch up. Nvidia’s share price has risen sharply as hyperscalers place record orders for advanced chips, but Clifton raises the question of whether those customers will earn adequate returns on that spending.

A key point of concern is depreciation -namely, how long cloud platforms assume their servers and GPUs will remain productive. Large tech firms generally depreciate servers over roughly six years, but Nvidia introduces new chips annually, each more powerful and energy-efficient than the last. That cadence, Clifton argues, means older chips become uneconomic long before they are fully depreciated.

Nvidia disputes that view and says its hardware remains productive for far longer, but Burry is focused on what Li called a “contradiction”: the company promotes the superiority of its newest chips while assuring customers that earlier generations remain viable for years.

Li added that Burry recently deregistered Scion Asset Management, meaning he no longer has to report positions publicly. However, his recent puts against Nvidia and Palantir, and Clifton’s underlying research, suggest an emerging negative thesis on what many investors still see as the market’s most powerful long-term theme.

-30-