저자: TokenEconomyAi채널진행자 · 외국인
요약 보기
마이클 베리의 AI 버블 붕괴 주장을 반박하는 영상. 1999년 닷컴과 달리 현재 AI 기업은 실제 매출·이익이 큼(엔비디아 연매출 216억달러, 나스닥 PER 26). 빅테크 7250억달러 투자는 전력·칩이라는 물리적 병목에 따른 실물 인프라 투자이며, 버블보다는 집중 리스크와 자기강화적 투자 순환이 실질 위험이라고 봄. 베리의 과거 예측은 대부분 빗나갔다고 주장. 조정 가능성은 인정하나 닷컴식 붕괴는 부정.
다듬은 글
Michael Burry just bet millions that AI is the next dot-com crash. He's wrong. And I can prove it with three numbers he's ignoring. On July 22nd, the man who predicted the 2008 housing collapse doubled his short position against Nvidia. He posted on X that markets are "acting like the last months of 1999." A video comparing AI to the dot-com bubble just hit 190,000 views in nine hours.
The narrative is everywhere: AI is a bubble. Big Tech is spending 700 billion dollars while firing 168,000 workers. OpenAI just lost 38 billion dollars. The crash is coming. But here's what none of these predictions are telling you. The data that made Burry right in 2008 is the same data that makes him wrong in 2026.
And by the end of this video, you'll understand why the AI "bubble" comparison isn't just misleading — it's mathematically backwards. --- Let's start with something uncomfortable for the AI bears. Michael Burry has been predicting market crashes for 18 years. He's been right exactly once. In 2019, Burry warned that passive investing was a bubble "worse than subprime." The S&P 500 went up 29 percent...
that year. In 2020, he predicted a massive market collapse from COVID. The market recovered within five months and hit all-time highs. In 2022, he called for a prolonged recession and bear market. The market bottomed and began a two-year bull run. In late 2025, Burry actually closed his hedge fund. His stated reason was that he was "misaligned with markets." Those are his words, not mine.
Here's the pattern that matters: Since 2015, the S&P 500 has posted positive returns after every single Burry bearish prediction. Every one. Being right once doesn't make you a prophet. It makes you someone who was right once. The 2008 call was brilliant. The decade and a half since has been a masterclass in being early to the point of being wrong.
But let's assume Burry's track record doesn't matter. Let's look at the actual argument: Is AI really the dot-com bubble? --- The dot-com bubble comparison rests on one assumption: that AI companies today look like internet companies in 1999. They don't. And here's where the math gets brutal for the bears.
In March 2000, the Nasdaq traded at a price-to-earnings ratio of 60. Today, it trades at 26. That's not a bubble multiple. That's a discount to historical averages. But P/E ratios are just the start. In 1999, corporate profits as a share of GDP were declining. Companies were burning cash with no revenue.
Pets.com raised 82 million dollars and had revenue of 619,000 dollars before going bankrupt. In 2026, US corporate profits sit at 12.4 percent... of GDP. That's near all-time highs. The companies driving the AI narrative aren't burning cash with no revenue. They're generating the largest profits in corporate history.
Nvidia just reported 216 billion dollars in annual revenue. Not projected revenue. Actual revenue. Up 65 percent... year over year. Microsoft's AI business hit a 37-billion-dollar run rate. They have 20 million paid enterprise Copilot seats. Accenture alone signed a deal for 740,000 seats. Even the OpenAI "loss" that everyone's citing is misleading.
The 38.5 billion dollar figure includes a 41.5 billion dollar one-time accounting charge from their nonprofit-to-profit conversion. That's not cash leaving the building. It's an accounting adjustment. Their actual cash burn is 3 to 5 billion per year against 13 billion in 2025 revenue and an annualized run rate now approaching 24 billion.
Pets.com had 619,000 dollars in revenue. OpenAI has 24 billion. These are not the same thing. --- So if Burry is wrong, who's right? Meet Gavin Baker. He runs Atreides Management. He was early and correct on Nvidia, Astera Labs, Coherent, and Cerebras. His track record on AI investments over the past three years has been nearly flawless.
Baker's thesis is simple: AI is not a bubble. It's a supercycle constrained by physics. Here's what he means. Bubbles happen when capital floods into assets with no underlying scarcity. Dot-com was unlimited domain names and server space with no demand. AI infrastructure is the opposite. There are two physical bottlenecks constraining the entire industry: power and chips.
You cannot build a data center without electricity. You cannot train a model without GPUs. Both are genuinely scarce. Both take years to expand. The 725 billion dollars that Big Tech is spending isn't speculation. It's buying physical assets: data centers, power infrastructure, silicon. When Amazon, Microsoft, Google, and Meta spend 725 billion in a single year, that money doesn't evaporate like dot-com VC funding.
It builds real infrastructure that generates real revenue. Baker's conclusion: "Tech trades at roughly the same multiple it did five or six years ago. Multiples have compressed since the start of 2025. On valuation, there is no bubble." --- Now let me be honest about what the bulls are getting wrong.
The AI infrastructure buildout is real. The revenue is real. The profits are real. But that doesn't mean there's zero risk. The biggest legitimate concern isn't fraud or speculation. It's concentration. Four companies are responsible for 725 billion dollars in AI infrastructure spending. If any one of them decides the ROI isn't there, the entire supply chain feels it.
The second risk is the self-reinforcing investment loop. Companies are spending to keep pace with competitors. Microsoft spends, so Google spends, so Meta spends. Each company's capex justifies the others. If that loop breaks — if one major player pulls back — the cascade could be real. But here's the critical difference from dot-com: Even in a pullback scenario, the infrastructure exists.
The data centers don't disappear. The chips don't evaporate. The enterprise customers don't un-sign their contracts. A correction is possible. A dot-com style 78 percent... collapse where the underlying value was always zero? The data doesn't support it. --- So who benefits from understanding this? The winners are value investors buying the fear.
Every time Burry makes a prediction, institutional money gets cautious. Valuations compress. Patient capital accumulates. The infrastructure players win regardless of which AI model dominates. Nvidia, AMD, the power companies, the data center REITs — they're selling shovels in a gold rush that's actually producing gold.
Enterprise software companies with real AI integration win. Microsoft's 20 million Copilot seats aren't speculation. They're recurring revenue locked into multi-year contracts. The losers are retail investors who sell on fear headlines. Every Burry prediction has been followed by market gains. Panic selling at the bottom isn't a strategy.
The losers are also the AI wrapper startups with no moat. When correction pressure comes, the companies without real technology or real revenue get wiped first. That's not a bubble popping. That's market discipline working correctly. --- Michael Burry was right once, 18 years ago. Since then, he's predicted crashes that didn't come, closed his fund because he was "misaligned with markets," and now he's betting against the largest infrastructure buildout in human history.
The AI bubble narrative compares 2026 to 1999 while ignoring every metric that made 1999 dangerous: sky-high P/E ratios, declining corporate profits, companies with zero revenue. Today we have compressed multiples, record corporate profits, and companies generating hundreds of billions in actual revenue.
Could AI have a correction? Absolutely. Concentration risk is real. The capex loop could slow. But a dot-com style collapse where the underlying value was always speculative? The numbers don't support it. Burry is betting against Nvidia, against Microsoft, against the infrastructure backbone of the entire AI economy.
He's betting that 216 billion dollars in annual revenue is somehow equivalent to Pets.com. That's not a prediction. That's a misreading of the data that made him famous in the first place. The question isn't whether AI is a bubble. The question is whether you understand the difference between speculation and infrastructure. Do you think the AI correction is coming? Tell me in the comments. ---
Comments 0