July 7 (Reuters) – Chinese startup DeepSeek is developing its own AI chip, according to three people familiar with the matter, a move that could reduce its dependence on Nvidia and Huawei chips that it relies on to train and operate its globally popular models.
The chip is designed for inference — the phase of “AI computing” in which a trained model generates answers for users — rather than training new models, the sources said.
If successful, DeepSeek’s expansion into semiconductor development would “mark a major strategic shift for a company widely hailed in China as the country’s AI champion, and potentially exacerbate the challenges facing Chinese tech giant Huawei.”
Shares of U.S.-based Nvidia fell about 1.6% in premarket trading.
“Nvidia is at zero in China and staying there. DeepSeek has almost no chance of selling silicon outside China unless it gets access to cutting-edge manufacturing,” said Radio Free Mobile analyst Richard Windsor, adding that the development had no impact on the chipmaker.
DeepSeek rose to global fame more than a year ago after releasing two highly efficient AI models that went viral around the world, surprising many in Silicon Valley and Washington.
The company has long been known for placing an emphasis on breakthrough AI models rather than commercializing its technology.
Although Huawei’s offerings still lag well behind Nvidia’s most advanced chips, a U.S. ban on their “exports to China” has helped Huawei capture around half of the $50 billion domestic AI chip market, supplying DeepSeek and several other leading industry players.
However, Huawei’s influence on the market is already weakening as technology rivals Alibaba and Baidu develop their own AI chips and gain market share.
DeepSeek’s efforts to join this race are still in their early stages. The company is tapping outside partners and is in discussions with chip design, foundry and memory companies, the three sources said. The effort began about a year ago, one of the people said.
The Hangzhou-based company has also increased hiring of chip design engineers in recent months, but the recruitment was done privately and without job postings on public hiring platforms, two of the sources said.
All three people declined to be identified because the information is not public. Despite becoming a pioneer of China’s AI ambitions, DeepSeek has held back. The company did not respond to a request for comment.
FOLLOW GLOBAL TRENDS
With an in-house chip, DeepSeek would “join other global AI developers in seeking greater control over the hardware behind their models and reducing dependence on Nvidia.”
OpenAI last month unveiled Jalapeno, its first custom inference chip developed with Broadcom, while Anthropic is considering building its own AI chips, Reuters reported in April.
For DeepSeek, the operation has an additional strategic dimension. U.S. export controls ban Chinese companies from buying Nvidia’s most advanced chips, and Beijing is pushing its technology champions to develop domestic alternatives.
Liang Wenfeng, founder of DeepSeek, said in a rare interview with a Chinese media outlet in 2024 that chip export controls were a challenge for the company.
DeepSeek has used both Nvidia and Huawei chips. The company said the base model underlying R1, the reasoning model whose low-cost performance triggered a plunge in U.S. tech stocks in January 2025, was trained on Nvidia’s H800, a chip designed for the Chinese market that Washington banned in late 2023.
Since then, the company has increasingly oriented itself towards Huawei. In April, the company released its V4 model adapted for Huawei’s Ascend chips, and Huawei said its processors were used in part of the training of V4 Flash, a lighter version of the model. As Reuters reported, Chinese tech giants’ orders for Huawei’s Ascend 950 chips surged following its launch.
Tap into inferential demand
A DeepSeek inference chip would target the fastest-growing segment of AI computing demand. As AI applications proliferate, more of the industry’s computing work is shifting from training models to running them, relying on specialized chips that can be cheaper and less power-hungry than general-purpose GPUs.
However, there is no guarantee of success. Developing a competitive AI chip typically takes years and a lot of capital. Manufacturing poses another hurdle, with the U.S. banning Chinese designers from access to the most advanced foundries abroad, while separate U.S. restrictions have limited China’s access to high-bandwidth memory, a component crucial to AI inference chips.
DeepSeek’s chip push coincides with the company’s first raising of debt capital. As Reuters reported in June, the company was expected to raise $7 billion in an initial round of funding, valuing it at between $52 billion and $59 billion. The company is reversing its long-standing strategy of rejecting external investments.
(Reporting by Reuters Staff; Editing by Eduardo Baptista and Tomasz Janowski)
https://finance.yahoo.com/technology/ai/articles/exclusive-chinas-deepseek-developing-own-103437335.html
