원문 보기
November 4th, 2025. A single SEC filing drops. Michael Burry, the man who made $700 betting against the housing market in 2008, has just put $1.1 billion on the line. His target? The two hottest AI stocks in the world, Nvidia and Palantir. If he's right, millions of retail investors are about to lose everything.
If he's wrong, he goes down in history as the guy who missed the AI revolution. This is the story of the biggest contrarian bet in modern finance. Meet John. He's a 34-year-old software engineer from Austin. In February 2024, he bought 50 shares of Nvidia at $880 per share, $44,000, nearly his entire savings.
By November, his investment had doubled. He felt like a genius. Then Burry's filing hit the news. Nvidia dropped 40% in 2 weeks. John's watching his profits evaporate, and he's asking himself the same question everyone else is asking, "Does Michael Burry know something we don't?" If you don't know Burry, here's what matters.
He's the investor from The Big Short, the guy who bet against the entire US housing market in 2005 when everyone said he was insane. We made $725 million, I think, on the funds in 2007. And in the first 6 months of 2008, there was about $730 million in withdrawals. He profited when Cathie Wood's Ark Innovation ETF collapsed in 2021.
His track record isn't perfect. He's often early, but when he bets big, markets pay attention. And this bet? This is massive. So, what does Burry actually see that everyone else is missing? It comes down to three things, and they're more dangerous than they sound. The first one is simple. The AI boom isn't driven by real customer demand.
It's driven by something else entirely, and this is where it gets weird. Money starts going in circles. Nvidia invests $100 million in OpenAI. OpenAI uses that money to buy Nvidia chips. Oracle spends tens of billions on Nvidia hardware to build data centers. Then OpenAI signs a $300 billion cloud deal with Oracle.
Oracle takes that revenue and spends it back on more Nvidia chips. Around and around it goes. Burry calls this circular financing, and he's not wrong. Look at the investments. Nvidia has funded startups like CoreWeave, Figure AI, and Psi. Those same startups then turn around and spend billions on Nvidia's GPUs.
Nvidia isn't just selling the tools. It's also lending money to buyers so they can purchase more tools. The result? Nvidia's revenue looks explosive, but it's not coming from real end users. It's coming from companies funded by Nvidia itself. This matters because Wall Street is valuing these companies as if this growth is organic and sustainable.
But if the money is just spinning in a loop, what happens when one link in the chain breaks? The whole thing collapses. Burry shared a chart on Twitter showing the web of investments between Nvidia, OpenAI, Microsoft, Oracle, and AMD. He wrote, "The future will regard this as a picture of fraud, not a flywheel.
True end demand is ridiculously small." But it doesn't stop there. Burry's second point cuts even deeper. Cloud computing, the backbone of AI, is slowing down. A few years ago, Amazon Web Services was growing 40% year over year. Google Cloud was up 45%. Microsoft Azure was still above 20%, but today, AWS growth has dropped into the teens.
Google Cloud has slowed significantly. Even Microsoft, the biggest beneficiary of OpenAI, is seeing weaker growth in its cloud segment. Think about that for a second. If AI workloads were genuinely driving a once-in-a-generation boom, cloud revenue should be exploding. Instead, the growth rate is falling.
Burry posted a chart comparing cloud growth from 2018 to 2022 versus 2023 to 2025. The difference is stark. The black line shows strong growth. The red line shows the slowdown. His message is clear. The AI hype isn't translating into real sustained demand. Some investors argue this is just a short-term supply problem.
Data centers take years to build. The recent spending surge hasn't fully come online yet. Maybe in 3 years, growth will accelerate again. But Burry isn't buying it. He thinks the slowdown is structural, not temporary. And if he's right, companies like Nvidia and Palantir are being valued for growth that will never materialize.
Now, here's what makes this worse. Burry's third argument is technical, but it's also where his case gets strongest. Big tech companies are quietly manipulating their earnings. They're doing it by extending how long their AI hardware is supposed to last. Here's how it works. When a company buys expensive server or chips, it doesn't expense the full cost immediately.
It spreads that cost over the expected lifespan of the equipment. That's called depreciation. If you depreciate a chip over 3 years, your annual expense is high. But if you suddenly decide to depreciate over 6 years, your annual expense gets cut in half. Your profit magically goes up, even though nothing in the business has actually changed.
Burry noticed that Meta, Google, Microsoft, and Oracle have all extended their depreciation schedules in recent years. They're claiming their hardware lasts longer now. But here's the problem. Nvidia is releasing new GPU generations every 12 to 18 months. Each one vastly more powerful and efficient than the last.
If new chips are two to three times better, older ones don't last longer. They become obsolete faster. So, why are companies pretending their equipment lasts longer? Burry goes as far as to suggest this could be seen as a form of accounting manipulation. He points to ByteDance, TikTok's parent company, as a perfect example.
ByteDance extended the useful life of its servers from 4 years to 5. That instantly added $15 billion to its profits. Then they extended it again from 5 to 6 years. Earnings jumped again. Meanwhile, in the same filing, ByteDance admitted that a lot of its old hardware was already obsolete and had to be written down.
The contradiction is obvious. Hardware is becoming obsolete faster while companies are claiming it lasts longer. Burry estimates there's a hidden hole of $176 billion in unrecognized depreciation across the tech sector. When that cost finally hits the books, it will crush earnings. And when earnings fall, stock prices follow.
Burry's thesis didn't go unnoticed. Please, given that there may be disruption. But in general, what's very, very special, I believe, about our our partnership is we're going to be able to move 2x, and I think in the future even faster. So, you get a much, much faster, and actually, because you He said, "The idea that you want to short chips is batshit crazy." He even suggested Burry was screwing with the economy to get out of his position.
It was emotional and not particularly convincing. Nvidia responded, too, but in a more corporate way. They sent a seven-page memo to Wall Street analysts mentioning Burry by name. They argued that older chips like the A100 still run well and provide strong margins, but they didn't provide data to prove it.
They also said their investments are small relative to global capital markets, which misses Burry's point entirely. The issue isn't the size of the investments. It's that the investments create artificial demand that inflates Nvidia's revenue. Burry's response? He doubled down. On his Substack, Cassandra Unchained, he wrote, "I stand by every point." He clarified that he's not calling Nvidia the next Enron.
He's calling it the next Cisco, a perfectly legitimate, profitable company that still became the poster child of overbuild during the dot-com bubble. And while all of this is happening, people like John are still holding their shares, watching the numbers, trying to decide what to do next. So, back to John, our retail investor.
He's sitting on Nvidia shares that are down from their peak. Should he sell? Should he hold? Should he follow Burry and bet against AI? Here's the reality. Burry's timing might be off. When he bet against the housing market in 2005, he was right, but he almost lost everything before the crash finally happened in 2008.
His investors nearly pulled their money. Timing the market is incredibly hard, even for someone as smart as Burry. And for retail investors like John, shorting stocks or buying put options is even riskier. You can be right about the direction and still lose money if the timing isn't perfect. But Burry's core message is worth listening to.
The AI boom is built on massive spending without proven demand. Companies are using accounting tricks to inflate their earnings, growth is slowing, and when the market realizes this, the correction could be brutal. John doesn't need to short Nvidia, but he should probably ask himself one question. You're not just investing, you're making a bet.
Am I betting on a real business, or am I betting on hype? And by the time most people realize what's happening, it's already too late to get out. Because in every bubble, there's a moment where everything still looks fine, the numbers keep going up, the headlines stay positive, everyone's making money, right before it isn't.
Comments 0