The man who predicted the 2008 crisis now says the AI infrastructure boom is the real deal, even if the market stumbles along the way.
The Summary
- Nouriel Roubini told Bloomberg he sees AI as "a secular increase in capex and potential growth," not a bubble destined to pop
- "Dr. Doom" acknowledges corrections could occur but frames them as normal market volatility, not systemic collapse
- The signal: when the most famous bear on Wall Street turns bull on infrastructure spending, something fundamental has shifted
The Signal
Nouriel Roubini earned his nickname the hard way. He called the housing crash when everyone else was still partying. So when he tells Bloomberg he's not worried about an AI bubble, the market should pay attention. Not because he's always right, but because pessimism is his brand.
His thesis is straightforward: the money pouring into AI infrastructure represents a secular shift in capital expenditure, not speculative froth. Data centers, chips, power infrastructure. The stuff agents need to run.
"I think it is a secular increase in capex and potential growth."
This matters because the infrastructure layer is where Web4 gets built. Companies are spending real money on compute, not just bidding up tokens or chasing the next foundation model. The capex cycle Roubini identifies is the same one that built railroads, electrified cities, and wired the internet. These buildouts create decades of economic returns, even when individual companies flame out.
But Roubini isn't saying the ride will be smooth. He told Bloomberg that "some corrections could occur." Translation: valuations will swing. Some big names will stumble. Entire subcategories of AI tooling will get commoditized overnight. That's different from a bubble popping. A bubble implies the whole premise was fake. Roubini is saying the premise is real, but the path is messy.
The timing of this matters. We're deep enough into the AI infrastructure cycle that the skeptics have data. Utilization rates, power consumption, actual revenue from inference workloads. The numbers justify the spending in a way they didn't 18 months ago. Roubini is reading the same data everyone else has access to, and he's seeing substance.
The Implication
If Roubini is right, the current AI infrastructure wave isn't speculation. It's foundation-laying. That means two things for anyone building in Web4: first, the capital will keep flowing into the picks-and-shovels layer. Compute, storage, networking. Second, corrections won't kill the trend. They'll just shake out the companies that mistook narrative for business model.
Watch the capex numbers. If Roubini's thesis holds, spending should stay elevated even through market turbulence. That's the tell. Bubbles deflate. Infrastructure cycles grind forward.