DeepSeek Excludes US Chipmakers From V4 Model Optimization

The Chinese artificial intelligence lab DeepSeek excluded U.S. chipmakers Nvidia and Advanced Micro Devices (AMD) from the performance optimization phase of its upcoming flagship V4 model. Departing from industry standards, DeepSeek granted early access to domestic suppliers, including Huawei Technologies and its chip designer HiSilicon, providing them a multi-week head start to optimize software for Chinese processors.

This strategic shift aligns with Beijing's push for technological self-sufficiency and follows a period of tightening U.S. export controls on advanced AI hardware. The move potentially accelerates the development of a parallel Chinese technology supply chain, a risk previously identified by Nvidia CEO Jensen Huang.

Simultaneously, a senior Trump administration official alleged that DeepSeek trained its latest model using Nvidia Blackwell chips in mainland China, which would violate U.S. export laws. The official claimed DeepSeek might attempt to conceal the use of American hardware by falsely attributing the training to Huawei systems. Nvidia and AMD declined to comment on these allegations. DeepSeek previously disrupted the AI market in January 2025 with its R1 reasoning model, which achieved high performance using older Nvidia chips and fewer compute resources.