DeepSeek is building its own inference chip to cut reliance on Nvidia and Huawei, Reuters reports. The move adds pressure on Nvidia's China business as export controls tighten.
China's DeepSeek is developing its own AI chip, a move that could reduce its reliance on Nvidia (NVDA) and Huawei hardware, Reuters reported, citing people familiar with the project. The chip is designed specifically for inference – the process of running trained models – rather than training, which requires more powerful and expensive processors.
The decision marks a strategic shift for DeepSeek, which has relied on Nvidia's GPUs and Huawei's Ascend chips for its large language models. By building in-house silicon, DeepSeek aims to secure supply and lower costs as export controls tighten on advanced semiconductors to China. Reuters said the company has hired engineers from chip design firms and is working with a domestic foundry.
For Nvidia, the development adds another headwind in China, a market that already faces restrictions on sales of its A100 and H100 chips. DeepSeek's custom chip, if successful, would remove a potential customer for Nvidia's lower-tier China-compliant products like the H20. Nvidia shares were little changed on the news, trading at $195.55, up 0.37% on the session. The stock carries an Alpha Score of 66 out of 100, a moderate rating.
The chip's focus on inference is telling. Training chips like Nvidia's H100 require massive memory bandwidth and interconnect speed. Inference chips can be simpler and cheaper, optimized for low latency and power efficiency. DeepSeek's design suggests it is betting that the bottleneck for AI deployment will shift from training to inference as models mature and usage scales.
DeepSeek is not alone. Chinese tech giants Alibaba and Baidu have developed their own inference chips. ByteDance is reportedly working on one as well. The trend reflects a broader push among Chinese firms to reduce dependence on foreign semiconductors, especially after U.S. export controls tightened in 2022 and 2023.
Reuters did not provide a timeline for DeepSeek's chip or details on its performance targets. The company declined to comment. Nvidia did not respond to a request for comment.
Drafted by a large language model from the source reporting linked above, then screened by automated publishing checks. It is not read by a journalist before publication. Some articles cite our Alpha Score. Verify prices and figures against the original source. Educational coverage, not personalized advice.