Chinese AI startup DeepSeek is developing its own artificial intelligence chip, according to three people familiar with the matter, a move that could reduce its dependence on Nvidia and Huawei chips currently used to train and operate its widely popular AI models.
The chip is being designed for AI inference, the process where trained models generate responses for users, rather than for developing new models, the sources said.
If successful, DeepSeek’s move into semiconductor development would represent a major strategic shift for the company, which has emerged as a leading force in China’s AI sector and could intensify competition with technology giant Huawei.
Shares of U.S.-based Nvidia slipped about 2% in premarket trading.
DeepSeek gained global attention over a year ago after launching two highly efficient AI models that surprised Silicon Valley and Washington.
While Huawei has benefited from U.S. export restrictions to capture a large share of China’s AI chip market, competition is increasing as rivals such as Alibaba and Baidu develop their own chips.
DeepSeek’s chip effort remains in the early stages, with the company holding talks with chip-design firms, foundries and memory suppliers while quietly hiring chip engineers, sources said. The company has not commented on the development.
Following Global Trends
DeepSeek’s move into custom chips mirrors a broader push by AI companies to reduce reliance on Nvidia and gain more control over their hardware.
Companies such as OpenAI and Anthropic are also exploring or developing their own AI chips. For DeepSeek, the effort is driven partly by U.S. export restrictions on advanced Nvidia chips and China’s push for domestic alternatives.
DeepSeek has used both Nvidia and Huawei chips, including Nvidia’s China-specific H800 for training its R1 model. It has since increased reliance on Huawei, adapting newer models for Huawei’s Ascend chips.
Tapping Inference Demand
DeepSeek’s chip is aimed at AI inference, the fastest-growing segment of AI computing, where demand is shifting from training models to running them.
The project faces significant challenges, including high development costs and U.S. restrictions on advanced chip manufacturing and memory supplies.
The chip push also comes as DeepSeek prepares to raise $7 billion in its first external funding round, marking a shift from its long-standing policy of avoiding outside investment.
(With inputs from Reuters)
