So Michael Burry's been making headlines again with his latest prediction - that AI is basically Dot Com 2.0. Here's the thing though: his track record since 2008 has been pretty rough.

Don't get me wrong, the guy made $100 million personally and $700 million for investors by nailing the subprime crisis. That's legendary. But ever since? He's been consistently early and wrong on the bearish side. He even closed his hedge fund last year because he couldn't align with where markets were actually going. That's a pretty big tell.

Now Michael Burry is arguing that big tech firms like Meta and Microsoft are cooking their books with unrealistic depreciation schedules to pump earnings. He's saying AI infrastructure has shorter lifespans than traditional servers, so the ROI doesn't add up.

Here's why that doesn't really hold up: Yes, GPUs have shorter cycles, but most AI infrastructure actually lasts 15-20 years. Plus, older GPUs don't just become worthless. They're still useful for inference - running trained models for end users. That's a whole different revenue stream.

Then there's his cash flow concern - the idea that massive CAPEX spending will tank margins. But look at what's actually happening. Alphabet's operating cash flow jumped from under $100 billion to $164 billion. That's not a strain. Margins are expanding across the board. And companies scaling AI are reportedly seeing returns of over $3 for every $1 invested. The latest wave, agentic AI, is supposedly cutting costs by 25% or more.

Michael Burry also compares NVIDIA to Cisco in 2000, saying both are overvalued. But the numbers don't match. Cisco's P/E in March 2000 was over 200. NVIDIA's current P/E is 47. That's not even close to the same situation.

What's interesting right now is the actual market signal. H100 GPU rental prices spiked about 17% since mid-December. That's showing real scarcity and demand, especially from agentic AI usage ramping up. Infrastructure plays like Nebius Group, CoreWeave, and IREN are benefiting. Energy companies like Bloom Energy are seeing action too - they solve the power bottleneck that's becoming the real constraint.

The options market is putting real money behind this thesis too. Saw some serious whale activity on NVIDIA calls and Bloom Energy, with one trader dropping $9 million on March calls. Bloom jumped 8% Monday while most tech was struggling.

Look, Michael Burry's contrarian credentials are solid. But his AI thesis is getting hammered by actual data. Between the rental price surge, the cash flow expansion, and the efficiency gains from agentic AI, the bearish narrative just isn't holding up. Sometimes the market really is just better than the bear case.