News
DeepSeek restricts latest AI model access for Nvidia and other US chipmakers
Thursday, February 26, 2026 at 01:04 AM
Reports indicate that DeepSeek is restricting access to its latest AI models for major American chip manufacturers, including Nvidia. This move highlights potential friction in the software-hardware feedback loop essential for optimizing next-generation AI infrastructure and chip architectures.
Context
Chinese AI lab DeepSeek has broken standard industry protocol by withholding its upcoming flagship model, V4, from U.S. chipmakers including Nvidia and AMD. Instead, the lab granted early access to domestic suppliers like Huawei Technologies, providing them with a several-week head start to optimize software for Chinese processors. This shift follows the massive success of DeepSeek-R1, which reached 75 million downloads and previously triggered a $600 billion temporary sell-off in Nvidia market value.
The decision is viewed as a strategic move to disadvantage U.S. hardware within the Chinese market. Compounding the friction, U.S. officials allege the model was trained on Nvidia’s restricted Blackwell chips in mainland China, potentially violating export controls. While analysts suggest the immediate technical impact on Nvidia may be limited due to advanced coding tools, the exclusion marks a significant departure from collaborative norms and intensifies the geopolitical divide in the semiconductor supply chain.
Related Companies
Nvidia
NVDA