DeepSeek V4 designed for Huawei chips challenges US AI dominance
DeepSeek has released version 4 of its AI model, designed to run on Huawei's GPU chips, demonstrating China's technological autonomy. The launch threatens the market position of US firms OpenAI and Anthropic, which plan IPOs in late 2026. OpenAI and Anthropic have accused DeepSeek of using "distillation" to train its models on responses from ChatGPT and Claude.
DeepSeek has released version 4 of its artificial intelligence model, designed to run on GPU chips from Chinese technology group Huawei, in a move that underscores Beijing's push for technological autonomy and challenges the dominance of US AI firms.
The launch threatens the market position of OpenAI and Anthropic, both of which plan to go public in late 2026. OpenAI and Anthropic have accused DeepSeek of using "distillation" — siphoning responses from their models ChatGPT and Claude to train its own systems.
DeepSeek offers its model in open source but keeps training data sealed, a stance that has drawn criticism from US rivals. The company's decision to tie its AI to Huawei's processors provides a concrete alternative to the infrastructure controlled by American technology giants.
The development comes amid broader tensions over AI leadership. In recent weeks, the White House has moved to reintegrate Anthropic amid Pentagon opposition over its Mythos AI model, and US officials have considered regulating AI after concerns over that model.
Separately, two US citizens were killed by federal agents in Minneapolis in January, an incident that has fueled opposition to the use of surveillance algorithms by immigration enforcement authorities.