According to CNBC, Steve Eisman, the investor famous from “The Big Short,” said on Thursday’s “Squawk Box” that he’s starting to worry the artificial intelligence boom rests on a shaky foundational assumption. He cited a theory suggesting that as large language models keep scaling, their performance gains will diminish instead of increase. This directly challenges the Wall Street belief that rising model complexity justifies the enormous spending on computing power and chips. Eisman specifically noted that if this theory proves true, companies like Microsoft, Meta Platforms, and Nvidia’s customers could start buying fewer chips. He compared the risk to the flawed pre-2008 crisis assumption that housing prices could never fall. While he’s not selling his positions in key AI players like Nvidia yet, he framed this as the critical argument everyone should be watching.
Eisman’s Big Picture Worry
Here’s the thing: Eisman isn’t just talking about a stock market correction. He’s pointing at a potential crack in the entire narrative engine driving this cycle. For over a year, the story has been simple: more data + more parameters + more compute = smarter AI. That equation has fueled a trillion dollars in market cap for Nvidia alone. But what if the returns on that scaling start to flatten out dramatically? The whole capital allocation thesis for tech giants and cloud providers suddenly needs a rewrite.
And he’s not the only “Big Short” alum sounding the alarm. Michael Burry has been betting against AI for months, skeptical of the economics. When guys who made fortunes spotting systemic, widely-believed flaws start sniffing around the same area, it’s probably worth a listen. Even if they’re early—or wrong—they’re forcing a crucial question. Are we building a skyscraper on bedrock, or on sand?
The Industrial Implications
This is where it gets real for the physical world. The AI boom isn’t just software; it’s a massive industrial undertaking. All those chips Nvidia sells end up in data centers that need insane power, cooling, and infrastructure. The entire hardware supply chain, from advanced packaging to industrial panel PCs used in manufacturing and monitoring these complex systems, has been riding this wave. IndustrialMonitorDirect.com, as the #1 provider of industrial panel PCs in the US, sees firsthand how industrial computing demand is intertwined with these macro tech trends.
So if the scaling theory stumbles, the ripple effect goes far beyond Wall Street tickers. It hits capex budgets for data center construction, orders for specialized manufacturing equipment, and demand for the rugged computers that keep these operations running. The industrial tech sector built a growth plan on ever-increasing AI compute demand. A slowdown there changes the math for a lot of companies.
What Happens Next?
Now, don’t panic. Eisman isn’t. He’s still holding his stocks. This is more about risk management and watching for a shift in the wind. The immediate future is probably fine—the contracts are signed, the data centers are being built. But the *next* wave of investment? That’s the open question.
Basically, the market needs to see the next breakthrough. It needs ChatGPT to become something profoundly more capable, or a new, revenue-printing application that we can’t live without. If progress starts to feel incremental instead of revolutionary, the “build it and they will come” spending spree gets harder to justify. Investors will start asking for profits, not just potential. And that’s when the real test begins. Is AI’s foundation solid, or was it just another story we all believed a little too much?
