Huawei chips power DeepSeek AI
The AI wars are heating up with DeepSeek, a Chinese AI model that claims to handily beat US competitors in the form of ChatGPT and Gemini in terms of cost-effectiveness. Its open-source chatbot has propelled the app to the top position in the App Store in 51 countries, and now it has been revealed that it runs on a Huawei AI chip.
The DeepSeek R1 LLM (large language model) was trained on Nvidia H100 chips, but uses an Ascend 910C chip for inference, which is the action of using the trained model to generate answers. The information comes from @Dorialexander, who points out that Ascend chips don’t deal with training, so GPU power requirements aren’t as high.
However, the relatively lower performance of the Ascend 910C limits its suitability for training. Huawei plans to address this issue with the upcoming 920C chip and compete with Blackwell B200, the leading Nvidia chipset for AI operations.
