Why I Hope Michael Burry Is Wrong: AI Bubble Warning and My Contrarian Bet

 

Why I Hope Michael Burry Is Wrong: AI Bubble Warning and My Contrarian Bet

Open YouTube these days and you'll see Michael Burry content everywhere. "The Big Short Legend Warns of AI Bubble!" with clickbait thumbnails. Click on them and they're all saying the same thing: hyperscalers are pouring insane amounts of money into AI infrastructure, manipulating depreciation accounting, and a major catastrophe is coming.

Burry is even uploading this in parts - Lesson One, Lesson Two, and so on.

Watching these videos, I felt something strange. Not anxiety, but curiosity. Because the story was completely opposite to my investment direction. I'm betting on utility sectors - indirect beneficiaries of AI - and Tesla. I also hold ETFs like ARKG and ICLN. I have almost no direct exposure to Nvidia or Meta, but my portfolio clearly stands on the premise that "AI will continue to grow."

If Burry is right, I'm wrong. That's why I wanted to properly understand his argument. What is he seeing? What am I missing? This essay is a record of that process.

Understanding Burry as an Investor and His Methodology

To understand Michael Burry, you first need to know what type of investor he is. I once wrote that there are two types of big investors.

The first type is like Warren Buffett. He constantly revises his opinions and changes his perspective with the times. Buffett used to hate tech stocks but bet big on Apple, and while he preaches "hold forever," he'll cut losses decisively when needed. His methodology is consistent, but his application is flexible.

The second type is like Michael Burry. He pushes one methodology to its extreme. For him, "numbers" matter more than "narratives." That's how it was during the 2008 subprime crisis. When everyone believed in the housing market's growth, he dug into mortgage bond accounting and found the rot. The structural contradictions of CDOs, the fiction of credit ratings. He trusted balance sheets, not market sentiment, and that's why he won.

Burry's strength is clear: he doesn't get swept up in crowd psychology. When everyone's cheering, he coldly looks at the numbers. But his weakness is equally clear: once he's convinced of a framework, he doesn't change it. That's why he failed to predict the post-2008 housing boom. He still believed housing was overvalued and focused on "real assets" like farmland and water. But the government never let the housing market collapse. Burry's logic was correct, but reality didn't follow his logic.

This AI warning seems to follow the same pattern. Once again, he's dug into accounting books and found numerical contradictions. And he's applied the frame: "This is the 2000 dot-com bubble." But is it really?

Burry's AI Warning: Four Core Arguments

Here's Burry's argument in summary:

First, depreciation manipulation. Hyperscalers (Meta, Google, Microsoft) have extended the depreciation period for AI chips and servers to 5-6 years. Tech equipment should normally be depreciated over about 3 years, but extending it artificially inflates pre-tax profits. His evidence: Google's 2023 policy change created an additional $3.9 billion in profit.

Second, the gap between technological acceleration and accounting. Nvidia GPUs get new models every year. H100 to Blackwell, Blackwell to Blackwell Ultra. Vera Rubin in 2026. But depreciation over 5-6 years? This doesn't make sense. The actual tech lifespan and accounting lifespan are completely different.

Third, even executives are worried. Microsoft CEO Satya Nadella reportedly said in internal meetings that "we need to shorten chip replacement cycles." Amazon actually reduced its depreciation period. Burry sees this as a signal that "insiders know the risk."

Fourth, dependence on private credit. A significant portion of AI infrastructure investment is funded by private fund debt. If interest rates rise, this structure could collapse.

His conclusion is simple: massive asset write-downs will occur soon, and the AI bubble will burst like the dot-com bubble.

The Different Picture I See: Supply Shortage and Competitive Structure

I followed Burry's logic and then stopped. Something didn't add up. His framework assumes "oversupply." Like the dot-com bubble when 95% of fiber optic cables sat unused. A scenario where everyone over-invests in the same thing and ultimately no one makes money.

But the current AI market is the exact opposite. Supply is short. Order an Nvidia GPU now and you'll wait 6-12 months. Meta and Google extended depreciation periods not because they want to "use chips slowly" but because "there are no chips to replace them with right now." If H100s were piled up in warehouses, they'd obviously switch to Blackwell immediately. But they can't.

And the competitive structure is different. During the dot-com bubble, everyone competed with the same infrastructure (fiber optic cables) for the same market. It was a winner-takes-all game. Amazon and Google won, and everyone else died.

But AI? It splits into GPU, TPU, NPU. Nvidia's GPU strength is versatility. Google's TPU strength is efficiency. Apple's NPU strength is low-power on-device processing. The market is divided three ways.

In fact, on November 25, 2025, when news broke that Meta was adopting Google's TPU, Nvidia's stock dropped 2.6-6% while Google's rose. By Burry's logic, Nvidia shouldn't be in a decline phase yet. But the market is already pricing in "substitute competition."

The Multi-AI Era: Winner-Takes-All Is Over

I currently use two AIs: Grok and Claude. Their characteristics are completely different. Grok is fast and intuitive. Good for search-like use. Claude is deep and thoughtful. I use it for writing long pieces or refining complex logic.

This is the core of what I sense in the AI market: one AI isn't enough. As of November 2025, 64% of individual users simultaneously use 3 or more AIs. 68% of enterprises use multi-vendor strategies. 72% of developers switch between 3 or more LLMs per week.

In the dot-com era, "Google was enough." Search on Google, shop on Amazon. Done. But AI? Brainstorm with ChatGPT, write long pieces with Claude, search with Gemini, solve math with Grok. Each excels at different things, so I believe they'll all survive.

According to McKinsey surveys, 60% of companies generate revenue through "specialized AI applications." While general-purpose AI (OpenAI, Google) dominates some areas, medical AI, manufacturing AI, and legal AI are protected by their own data and regulatory barriers. The manufacturing AI market alone is projected to grow from $7.6 billion in 2025 to $62.3 billion in 2032.

This is a completely different picture from dot-com. Dot-com was "one winner." AI is "multiple winners."

So Why Will Burry Be Wrong?

Burry's weakness is the same as his strength. He trusts accounting books. If numbers look wrong, there's a problem. And he applies past patterns (dot-com, subprime) to the present.

But this time there's a different variable beyond the numbers: structural difference.

Dot-com was over-investment in identical infrastructure. Fiber optic cables sat 95% idle. AI is competitive investment in different infrastructures. GPU, TPU, NPU each compete in their own markets.

Dot-com was winner-takes-all. Network effects were strong. But AI is multi-platform. Users simultaneously use multiple AIs.

Dot-com had unclear revenue models. Pets.com sold dog food but couldn't make money. But AI already generates revenue. Google is preparing to raise ad revenue with Gemini, and OpenAI makes $2 billion annually from subscriptions.

Burry looks at "numbers," but I look at "structure." That's why I bet on a different path than him.

My Bet: Indirect Beneficiaries and Utilities

I didn't buy Nvidia. I thought direct exposure was risky. Whether Burry's warning is right or wrong, Nvidia will be volatile. Instead, I focused on indirect AI beneficiaries.

The utility sector. AI data centers consume enormous power. Running ChatGPT once takes 10 times the electricity of a Google search. The more hyperscalers pour money into AI, the more power demand grows. Even if Burry is right and the AI bubble bursts, data centers will run and electricity will sell until then.

Tesla follows similar logic. Tesla is a car company but also an AI company. FSD (Full Self-Driving) is a chunk of AI technology. And Tesla doesn't use Nvidia chips - it makes its own (Dojo). A beneficiary of hardware diversification.

ARKG is a gene therapy and biotech ETF. AI accelerates drug development. ICLN is a clean energy ETF. I accumulated it expecting renewable energy demand to grow as data center carbon emissions become a bigger issue.

My portfolio bets on "AI growth" but doesn't directly invest in "AI chip companies." I take indirect routes. Even if Burry's warning is partially right, I believe my investments can withstand it.

And Burry is a veteran. Honestly, he's more likely to be right. But even so, I have to trust my investment view and the average purchase prices I've built up over time. If Burry is right, I need to prepare countermeasures accordingly.

The 2027 Test

Of course, Burry could be right. His timeline is roughly 2027. When AI capex slows, depreciation issues explode, and asset impairments occur. Analysts like Seeking Alpha also raise the possibility of 2027 capex slowdown.

If that happens? I'll take losses. Utilities, Tesla, ARKG will all drop. But I think I'll be okay anyway.

Why? Because I'm not trying to time things. Burry tries to figure out "when will it burst?" I look at "what direction long-term?" Even if AI stumbles in 2027, I believe it'll grow again by 2030. Electricity will keep selling, autonomous driving will keep advancing, drug development will keep accelerating.

Burry is an investor who pushes one methodology to the extreme. I don't do that. I keep multiple scenarios open, diversify, and take indirect routes. That's why we take different paths.

In a way, Burry invests in a point in time, while I invest in the flow of time.

Conclusion: Why I Hope He's Wrong

Watching Burry's warnings plastered across YouTube, I thought: I hope he's wrong. Partly for personal reasons - I don't want my investments to lose money. But more importantly, because I believe AI really is different.

Dot-com was a vague expectation that "the internet will change everything." True, but too early. There was no revenue model.

AI already generates revenue. Already increases productivity. And most importantly, the competitive structure is different. Not winner-takes-all but multi-platform. One AI isn't enough. Each excels at different things, and users move between multiple AIs.

Burry looks at numbers. I look at structure. That's why we bet on opposite directions.

By 2027, we'll know who was right. Maybe we're both wrong. Markets love mocking our predictions.

But right now, at this moment, I'm confident in the picture I see. And I've bet money on that confidence. I hope Burry is wrong. I hope I'm right. But most of all, I hope AI really does change the world. And I hope the world recognizes that I was right. Who knows, I might get famous 😂

That's why I invested.