Chinese artificial intelligence lab DeepSeek has not provided early access to its upcoming flagship model, V4, to major U.S. chipmakers Nvidia and AMD, according to two sources familiar with the matter.
The decision marks a departure from standard industry practice, where AI developers typically share pre-release versions of major models with leading chipmakers to ensure optimal performance on widely used hardware.
Instead, DeepSeek granted early access to domestic suppliers, including Huawei Technologies, the sources said. The move gives Chinese chipmakers several weeks to optimise the software for their processors ahead of the model’s expected release.
Nvidia and AMD declined to comment, while DeepSeek and Huawei did not respond to requests for comment.
The reasons behind the decision were not immediately clear. However, industry analysts suggest the move could align with broader Chinese government efforts to reduce reliance on U.S. semiconductor technology.
The development comes amid heightened scrutiny over U.S. export controls. A senior Trump administration official recently told Reuters that DeepSeek’s latest AI model was trained using Nvidia’s most advanced Blackwell chips in mainland China, a potential violation of U.S. export restrictions. U.S. policy currently bars shipments of certain advanced AI chips to China.
DeepSeek’s models have been downloaded more than 75 million times on the open-source platform Hugging Face since the company gained global attention in early 2025. Chinese open-source AI models have seen rapid adoption, surpassing downloads of models from other countries on the platform over the past year.
U.S. authorities previously allowed Nvidia’s H20 and AMD’s MI308 chips, designed for AI inference, to resume shipments to China, though licenses for more advanced processors remain restricted. It is unclear whether DeepSeek has obtained approval to purchase those chips.
Analysts say the immediate commercial impact on Nvidia and AMD may be limited, as DeepSeek’s models primarily serve as benchmarks rather than enterprise deployment standards. However, the decision underscores growing technological and geopolitical tensions as Chinese AI firms prepare to release new models in the coming months.
