China's premier AI firm, DeepSeek, is reportedly working on its own AI chip, which will primarily be an inference accelerator.
Step Aside NVIDIA & Huawei, China's DeepSeek Prepares Its Own AI Inference Chip
DeepSeek, the leading AI firm which is based in China, has started developing its own AI chip. As per Reuters, multiple people have stated that the firm is now developing what could be the end of reliance on NVIDIA and Huawei.
We already know that China is focusing more on domestic solutions for AI. The country is banning the use of NVIDIA-based accelerators from being used, but despite rigorous restrictions, Chinese firms still rely heavily on NVIDIA GPUs and are also importing them through illegal means.
Chinese startup DeepSeek is developing its own AI chip, according to three people familiar with the matter, a push that could reduce its reliance on Nvidia and Huawei chips, which it has depended on to train and run its globally popular models.
via Reuters
Meanwhile, Huawei, China's crown jewel for chip development, has already shared its next-gen roadmap and is supplying domestic AI firms with the latest and greatest. DeepSeek has largely relied on the two (NVIDIA & Huawei) for training its AI models, but that is about to change as the company is now focused on developing an in-house chip that will be an inference accelerator.
As per Richard Windsor, Analyst of Radio Free Mobile, the development of DeepSeek's in-house inference chip will pose no serious threat to NVIDIA as the company will be unable to attract any major customers outside of China. That is in the case DeepSeek is at first successful in making the chip.
Just like global AI firms, Chinese AI firms such as Alibaba, Baidu, and now DeepSeek, have started development of their own in-house chips that are catered towards their latest AI models & work-loads.
In parallel, NVIDIA's Vera CPUs are said to be adopted by Chinese firms since there's no ban on CPUs, but GPUs, which are largely used as AI accelerators, are primarily banned. China only has access to NVIDIA's H200 and the cut-down H20 chip. And the latest Blackwell GPUs are entirely banned, leading many to smuggle them from bordering countries.
DeepSeek recently launched its latest V4 models, for which NVIDIA is offering out-of-the-box support on its latest Blackwell GPUs. Meanwhile, DeepSeek V4 was trained on Huawei's Ascend 950DT AI accelerator, fully supporting the 1.6T heavyweight model.
Looking ahead, DeepSeek making its own chip shows a fundamental shift in the strategy of major AI players. Custom silicon is rapidly becoming the leading choice for many AI firms, which poses a serious threat to chipmakers such as NVIDIA, Huawei, and the rest.
About the author: A Software Engineer by training and a PC enthusiast by passion, Hassan Mujtaba serves as Wccftech's Senior Editor for hardware section. With years of experience in the industry, he specializes in deep-dive technical analysis of next-generation CPU and GPU architectures, motherboards, and cooling solutions. His work involves not only breaking news on upcoming technologies but also extensive hands-on reviews and benchmarking.
Follow Wccftech on Google to get more of our news coverage in your feeds.
