Michael Burry's 2008 Windfall Proves His AI Bubble Thesis Is Outdated

Michael Burry earned himself a legendary status in financial circles through one of the most prescient calls in market history. In 2008, he made roughly $100 million in personal profit while generating $700 million for his Scion Capital investors by correctly predicting the subprime mortgage crisis. However, the very track record that made him famous is now being weaponized to support arguments that seem increasingly disconnected from current market realities.

How Michael Burry Built His 2008 Legacy

The inspiration for Christian Bale’s character in “The Big Short,” Michael Burry rose to prominence by identifying the housing market collapse before it happened. His $700 million windfall for investors—combined with his personal $100 million gain—cemented him as a genuine market prophet. But that success in 2008 created a particular mindset: the ability to see what others couldn’t.

The problem is that pattern recognition from one crisis doesn’t automatically transfer to another market cycle, especially in an entirely different asset class like artificial intelligence.

Where Michael Burry’s Recent Track Record Breaks Down

Since his 2008 triumph, Michael Burry’s market calls have become noticeably inconsistent. As U.S. equities have advanced over the past several years, he has repeatedly warned of impending collapse with little accuracy. In fact, he closed Scion Capital in the past year, citing misalignment with market direction.

This track record matters because Burry’s latest prediction is arguably his most aggressive: that AI stocks represent a 1999-style speculative bubble destined to crash like the dot-com era.

The Three Pillars of Michael Burry’s AI Bear Case—And Why They Crumble

Michael Burry’s recession hypothesis hinges on three central claims about why AI infrastructure and AI-enabling companies are overvalued. Each one, however, confronts substantial evidence to the contrary.

Claim 1: Tech Companies Are Manipulating Depreciation Schedules

Burry argues that Meta, Microsoft, and Alphabet are artificially inflating earnings by depreciating server infrastructure over four to six years instead of shorter timeframes. The implication: their true profitability is much lower than reported.

The data contradicts this. While newer GPUs do degrade faster than traditional servers, most AI infrastructure carries a useful lifespan of 15 to 20 years. More importantly, older-generation GPUs don’t become worthless when new chips launch. Legacy hardware remains valuable for inference workloads—the actual deployment of AI models for end users rather than training. This secondary market for used semiconductors significantly extends overall asset value.

Claim 2: Massive CAPEX Spending Is Crushing Cash Flow and Returns

Michael Burry contends that hyperscalers are burning cash at unsustainable rates without adequate returns on their infrastructure investments.

Reality shows the opposite trajectory. Companies like Alphabet have increased operating cash flow from below $100 billion to $164 billion in recent periods. Simultaneously, operating margins are expanding across the tech sector. Companies running AI-scaled operations report returns exceeding $3 for every $1 deployed in infrastructure.

The newest AI wave—agentic AI systems that autonomously perform complex tasks—is reportedly delivering 25% or greater cost savings to enterprises. These aren’t speculative promises; they’re showing up in actual financial results.

Claim 3: NVIDIA’s Valuation Mirrors Cisco at the Peak

Michael Burry’s most eye-catching comparison is between NVIDIA today and Cisco in March 2000, when Cisco’s price-to-earnings ratio exceeded 200. He suggests NVIDIA is similarly overpriced and destined for a two-decade decline.

The comparison doesn’t withstand scrutiny. NVIDIA’s current P/E ratio stands at 47—less than one-quarter of Cisco’s peak valuation. The gap between 47x and 200x earnings is not a matter of degree; it’s a fundamental difference in valuation extremism. NVIDIA’s multiple may be elevated, but it bears no resemblance to the internet bubble’s most excessive multiples.

Market Signals Point in the Opposite Direction

Physical market dynamics reinforce why Michael Burry’s thesis struggles to gain traction among active traders. NVIDIA’s H100 GPU rental prices have surged approximately 17% since mid-December, signaling persistent scarcity and robust demand for AI compute capacity. This price action doesn’t reflect a speculative top—it reflects constrained supply meeting powerful end-user demand.

Institutional and sophisticated investors are voting with their capital. Large options traders placed bold bullish bets this week: one trader deployed roughly $1 million on Bloom Energy calls as H100 scarcity threatens to become an energy bottleneck; simultaneously, a major options player risked approximately $9 million on March NVIDIA calls at the $205 strike. These aren’t panic bets; they’re convictions backed by serious capital.

The Verdict: Michael Burry’s AI Prediction Lacks Supporting Evidence

Michael Burry’s legacy as a crisis forecaster remains unquestionable. The $100 million he made in 2008, combined with the $700 million he generated for his investors, will forever cement his place in market history. But that success created a framework—a conviction that major dislocations can be identified in advance—that doesn’t automatically apply to artificial intelligence infrastructure in 2026.

The evidence overwhelmingly suggests the opposite. Cash flows are expanding, not contracting. Returns are rising, not falling. Valuations are elevated but not bubble-level extreme. GPU demand is accelerating, not plateauing.

Michael Burry may eventually be proven correct about something. But current data trends don’t support his AI bubble narrative.