DeepSeek developing AI chips to cut reliance on NVIDIA, Huawei
What's the story
DeepSeek, a leading Chinese artificial intelligence (AI) firm, is reportedly developing its own silicon chips. The move comes as part of the company's strategy to reduce dependence on third-party chipmakers such as Huawei and NVIDIA. The chip in question will be used for inference - a process where a pre-trained model generates responses for users - rather than training new models from scratch.
Progress report
DeepSeek's low profile and recent developments
DeepSeek is said to be in talks with manufacturing partners and has started hiring engineers for its chip development project. The company's move to develop its own chip aligns with a global trend among AI developers seeking more control over the hardware powering their models. OpenAI recently unveiled Jalapeno, its first custom inference chip developed with Broadcom. Meanwhile, Anthropic is considering building its own AI chips.
Industry shift
DeepSeek has used NVIDIA and Huawei chips in the past
DeepSeek has used both NVIDIA and Huawei chips in the past but has increasingly leaned on Huawei due to US export controls barring Chinese companies from buying NVIDIA's most advanced chips. As the industry grows, more computing work is moving from training models to running them, which requires specialized chips that can be cheaper and less power-hungry than general-purpose GPUs. However, designing a competitive AI chip usually takes years and a lot of capital.