
DeepSeek Just Ditched Nvidia Entirely. Its Next Model Runs on Huawei Chips.
China's most celebrated AI lab is building its next model without a single Nvidia GPU. The chip war just got personal.
The AI Post newsroom — delivering AI news at the speed of intelligence.
DeepSeek, the Chinese AI lab that sent shockwaves through Silicon Valley earlier this year by matching frontier models at a fraction of the cost, just made its boldest move yet. Its next-generation model, V4, will run entirely on Huawei's latest AI chips. Not a single Nvidia GPU in the stack.
The Information reported Friday that Chinese tech giants including Alibaba have ordered hundreds of thousands of Huawei's Ascend processors in preparation for V4's launch. Reuters confirmed the report. This is not a hedge or a secondary supply chain. This is a full breakup with the Western chip ecosystem.
Here is why this matters more than any benchmark score: the entire premise of US export controls on advanced chips was that China could not train frontier AI models without access to Nvidia's hardware. DeepSeek V3 already poked holes in that theory. V4 tears it apart entirely.
Nvidia should be paying very close attention. Chinese chipmakers already seized 41% of their domestic AI chip market, a number we reported this week. Now the country's flagship AI lab is proving you can build state-of-the-art models without touching American silicon. Every percentage point of domestic chip adoption in China is a revenue line Nvidia never gets back.
The geopolitical implications are staggering. If V4 performs anywhere close to what DeepSeek has delivered before, the US export control strategy will need a fundamental rethink. You cannot contain an industry that no longer needs your products.
Watch for V4 benchmarks. If they are competitive with GPT-5 or Claude Mythos on Huawei hardware alone, the chip war narrative flips overnight. And Nvidia's $2.8 trillion valuation starts looking a lot more fragile.