DeepSeek Keeps Latest AI Model "Undisclosed" to U.S. Chipmakers Like Nvidia

Chinese artificial intelligence (AI) startup DeepSeek has reportedly not shared its next flagship model with U.S. semiconductor companies for performance optimization.

Quoting sources, Reuters reported that this was a move that "deviates from the usual industry practice that typically takes place ahead of a major model update."

According to the sources, instead of U.S. firms, DeepSeek first granted access to domestic Chinese suppliers, including Huawei Technologies. DeepSeek shocked the global market last year with a low-cost AI model. It is expected to launch its next-generation large-scale update, "V4," in the near future.

AI developers generally share pre-release versions of their key models with companies such as Nvidia and AMD before the official launch. This is to ensure that their software runs efficiently on widely used hardware. DeepSeek had also worked closely with Nvidia engineers in the past.

Reuters reported that Nvidia, AMD, DeepSeek, and Huawei did not respond to requests for comment on the matter.

Ben Bajarin, Chief Executive Officer (CEO) of Creative Strategies, said, "In terms of general-purpose data accelerators, the impact on Nvidia and AMD is minimal," explaining that "most companies do not run DeepSeek, and DeepSeek is used above all as a benchmarking model." He added that new AI coding tools are reducing the time it takes to make software run smoothly on hardware "from months to weeks."

However, he analyzed that this move is likely part of a broader strategy by the Chinese government. He said it is "an attempt to put U.S.-made hardware and models at a disadvantage within China." In other words, it is seen as an effort to reduce reliance on U.S. technology and to strengthen China’s own semiconductor and AI industries.

In connection with this, a senior official from the Trump administration claimed that DeepSeek’s latest model may have been trained on Nvidia’s "Blackwell" chips, which are subject to U.S. export controls, using clusters located in China. The official further suggested that DeepSeek may be trying to remove technical indicators that could reveal its use of U.S.-made AI chips. It is also reportedly planning to publicly claim that it used Huawei’s chips to train the model.

This content was produced with the assistance of AI translation services.