Deepseek crash 2.0?
This day started with dramatic news from China.
Beijing launched GLM-5.2, an open-source model that not only stands shoulder to shoulder with giants like OpenAI and Anthropic, but does so at a tenth of the cost.
The market got used to the idea that big money always wins thanks to more computing and expensive chips, but now it may be asking itself if all this investment is worthwhile against a model that proves the path can be brutally shortened.
The real drama with GLM-5.2 is not just its computing capabilities, but its background story:
This model was trained on Chinese chips manufactured by Huawei.
Not Nvidia
Not AMD
And not any other leading American company!
Investors on Wall Street are asking themselves if, while the West builds walls of sanctions on chip exports, China has simply built itself an independent alternative that is many times better than what the West offers?
This model is completely exposed, it is open-source, public, accessibility is high, and the message is clear:
China's technological independence is no longer in the experimental stage, it is in the mass production stage.
The market priced American tech giants based on the assumption that their technological advantage is unassailable...
The moment an open-source model emerges that competes with the performance of the big names and costs almost nothing, the profitability model of the entire industry is called into question...
If the technological advantage shrinks faster than expected, and if Western companies have to start lowering prices to fight a wave of cheap and accessible models coming from the East, the margins of these companies may start to erode.
The market, which detests uncertainty, may discover that the wall it built around Silicon Valley turned out to be a drywall?