Anthropic spent billions building Claude. A Chinese company built something nearly identical for pocket change. Tech stocks dropped because investors just realized they paid premium prices for what turns out to be a commodity.
This is the second time in months that a Chinese AI model showed up and made American valuations look stupid. DeepSeek did it in January. Now another one does it again. Analysts keep saying "nearly as good" like that's supposed to be comforting. Nearly as good for a fraction of the price is called getting f*cked.
The sell-off happened because the market finally understood what a moat actually is. It's not proprietary technology when someone in Shenzhen can replicate your billion-dollar model for the cost of a used Honda Civic. It's not brand loyalty when the products are functionally interchangeable. It's nothing.
Retail traders bought Anthropic proxies at all-time highs because they read that AI was the future. They were correct about the future part. They were wrong about who gets to profit from it. Turns out building expensive things that can be copied cheaply is not a business model. It's a charity program for engineers.
The "DeepSeek moment" framing is generous. It suggests this is a one-time shock that the market will digest and move past. What if it's not a moment? What if it's just how AI works now? Chinese labs keep shipping comparable models for less money. American companies keep pretending their costs are justified. The market keeps learning the same lesson.
Analysts will explain this away with something about quality differences or enterprise trust or regulatory moats. None of that matters when the price gap is this wide. Customers will beta-test the cheap option, find out it works fine, and never come back.
Every tech investor now owns stock in companies that spent billions to get Xeroxed by a team working out of a building with no heat.
Photo by Solen Feyissa on Unsplash
Leave a Comment