AI chipmaker Nvidia beat Wall Street expectations this week, announcing soaring revenues while downplaying growing fears of an AI bubble.
The company’s chips have been at the center of the ongoing AI boom, with AI companies’ valuations soaring to astronomical levels while their revenues continue to lag far behind.
Despite plenty of shakiness in the market, investors still firmly believe that the trillions of dollars companies like OpenAI, Microsoft, and Oracle are planning to spend on AI infrastructure buildouts will one day deliver a return.
But at the same time, economists are warning that those investors are overlooking some glaring problems with their plans to pile graphics processing units into enormous data centers to train and run immensely power-hungry AI models. For one, all of that hardware is degrading fast while running around the clock.
“You’re investing in something that is a perishable good,” economist and author David McWilliams told Fortune, calling AI hardware “digital lettuce” that’s “going to go off now.”
Worse yet, GPUs could also become obsolete as the technology improves.
“Technological change suggests that if you buy a GPU today, the chip is going to be outdated next year,” McWilliams added.
To McWilliams, it’s a recipe for disaster. He argues that the AI trade is “undoubtedly going to crash” — and points out that the tech isn’t creating any jobs, either.
McWilliams isn’t the first to point out some eyebrow-raising accounting in the sector. Michael Burry, who famously shorted the US housing market before its collapse in 2008, similarly argued earlier this month that AI companies’ projected growth could be massively exaggerated, accusing them in a November 10 tweet of assuming it will take far longer for AI hardware to depreciate to boost the numbers.
“By 2028, Oracle will overstate earnings 26.9 percent, Meta by 20.8 percent, etc.,” he wrote. “But it gets worse,” he added cryptically.
... continue reading