The AI Bubble That Wasn’t: What Michael Burry Got Wrong About Artificial Intelligence

Not long ago, news broke that Michael Burry — the same investor immortalized in The Big Short — had taken a position against the market. Almost instantly, financial social media and commentary shows began sounding alarms: “The AI bubble is popping.” Shares in some of the market’s most innovation-driven companies tumbled, and panic set in as investors wondered if one of Wall Street’s most famous contrarians was seeing something the rest of us weren’t.

But as with many market narratives, the dramatic headline was more powerful than the underlying truth.

However, the size of Burry’s bet was significantly overstated. Early coverage claimed that he had placed “hundreds of millions” in short positions against tech and artificial intelligence. The reality, once filings were fully reviewed, was far more modest. Burry himself later clarified that he had spent $9.2 million—not $912 million—on one of his short positions specifically. The math didn’t justify the market reaction, but by the time those corrections surfaced, the damage had already been done. Fear had outrun facts.

Adding to the complexity, market analysts quickly noted that Burry’s trades were likely underwater from the start. The stocks he targeted remained higher than at any point during the third quarter when the positions were established, and would need to drop another 5-7% just for his trades to break even—not accounting for the erosion in option value from time decay.

The Depreciation Argument

So what was Burry’s argument? In essence, he suggested that tech companies have been artificially inflating their profitability by extending the expected useful life of their equipment—particularly their computing hardware. He put specific numbers to this claim: Burry estimated that Big Tech will understate depreciation by $176 billion between 2026 and 2028, inflating reported profits by double-digit percentages at several major technology companies.

And there is some evidence supporting his concern. Several large technology firms have extended their depreciation schedules from 4-5 years to 5.5 years in recent years. Burry called this practice “one of the most common frauds in the modern era”—stretching the depreciation schedule to boost near-term earnings while pushing costs into later years.

He implied that the accounting shift made companies appear more profitable in the short term than they truly are, and that when the real depreciation hits, earnings would fall sharply.

That argument might hold water if we were still operating within the traditional technology cycle—where hardware depreciates rapidly, often losing most of its productive value in five years or less. That’s historically true for office PCs, servers, and general IT systems.

But it does not hold true for GPU-powered artificial intelligence infrastructure.

Why AI Infrastructure Is Different

The mistake lies in assuming that a high-performance, AI-optimized GPU is just an expensive computer. It’s not. If AI infrastructure were a machine, it wouldn’t be a laptop — it would be a hydroelectric power plant. You don’t toss it aside when something newer comes along. You maintain it, upgrade certain elements, refuel it, cool it, optimize it, and in doing so, extend its productive life far beyond initial projections.

Leading GPU platforms, which power much of today’s large-scale AI training and inference, are proving far more durable than originally estimated. Advances in cooling technologies, modular component upgrades, and ongoing software refinements are actively extending their functional lifespan. Multiple AI labs are now reporting that their infrastructure remains economically viable longer than expected when originally purchased. In effect, the technology is improving with use—not degrading.

This is not a speculative phenomenon driven by hype. It reflects a structural shift in the underlying economics of computing and machine learning. AI is no longer a cyclical product; it is a long-term platform on which other innovations are being built. The utility of the hardware compounds over time rather than diminishes.

The Market’s Reaction

So why did investors react so dramatically to Burry’s positioning? Because markets respond first to narrative and only later to analysis. A headline that reads, “The Big Short investor is betting against AI” sparks emotional recoil. Investors think, “If he saw the housing collapse before anyone else, maybe he sees the AI collapse, too.” And in that moment, positions are liquidated, not because fundamentals have changed but because sentiment has.

That reflex was psychological, not financial. It wasn’t a commentary on the worth of AI—it was a commentary on our instinct to avoid being the last person holding a hot potato.

Even CEOs of targeted companies pushed back forcefully, with one calling Burry’s moves “super weird” and “batshit crazy.” The executive pointed out that “the two companies he’s shorting are the ones making all the money”—questioning why anyone would bet against the actual profit generators in the AI ecosystem.

The Valuation Question

It’s worth noting that some of Burry’s concerns about valuation aren’t entirely without merit. Some AI stocks have reached price-to-sales ratios exceeding 150—far beyond the 30-40 range that typically served as a ceiling during the dot-com bubble. These are stratospheric multiples by any historical standard, and they do suggest that certain AI stocks have gotten ahead of themselves in the near term.

However, extreme valuations and long-term fundamental value are not the same thing. A stock can be temporarily overpriced while the underlying technology remains transformational. The dot-com bubble produced several of today’s most valuable companies and a genuine internet revolution—even though most of the bubble companies failed.

Burry Steps Away

Perhaps most telling is what happened after Burry made his bet. Shortly after disclosing these positions, Burry deregistered Scion Asset Management and stepped away from managing outside money. In a letter to investors, he stated that his “estimation of value in securities is not now, and has not been for some time, in sync with the markets.”

Market analysts interpreted this not as a dire prediction of imminent collapse, but rather as an acknowledgment that he was stepping away “from a game he believes is fundamentally rigged”—a market environment where traditional valuation metrics no longer seem to apply in the same way they once did.

The Long-Term Investment Case

From Drexel & Co.’s perspective, artificial intelligence is not following the arc of a bubble. It is reshaping corporate operations across industries—from logistics and manufacturing to finance and healthcare. Long-term investors must recognize the difference between hype-driven volatility and platform-level innovation.

What does this mean in practice?

First, durability matters more than outdated depreciation assumptions. AI systems are not consumer electronics with expiration dates; they are enterprise infrastructure that becomes more powerful—and more integrated—over time.

Second, capital investment is still accelerating. Leading companies are expanding AI budgets year over year, not pulling back. They are embedding AI into strategy, decision-making, productivity, and customer engagement. You don’t increase investment in something you believe is peaking.

Third, innovation cycles are extending profitability beyond original estimates. Historically, technology adoption has occurred in waves spanning three to five years. AI is already showing signs of building in continuous cycles, where improvements stack rather than reset.

The Bottom Line

The bottom line is simple:

AI wasn’t overbought. It was under-understood.

Markets didn’t sell off because AI’s potential was suddenly invalidated. They sold off because a famous skeptic raised his eyebrows at the wrong assumptions—and then promptly closed his hedge fund when the market didn’t agree with his timing. This phase is not a collapse; it’s a recalibration as businesses integrate AI deeper into their core operations and the investment community catches up to what this technology really represents.

The real risk for investors today is not being exposed to AI too early.

It’s being positioned too late.