The AI race is heating up, and China’s DeepSeek is making waves. The company claims its AI model is more cost-efficient than US competitors. This advantage has helped its chatbot become the top app in 51 countries. A key factor behind this success is its use of Huawei’s Ascend 910C chip.
China’s DeepSeek AI Challenges US Rivals with Cost Efficiency
DeepSeek’s large-language model (LLM), DeepSeek R1, was trained using Nvidia’s H100 GPUs. However, for inference—the process of generating AI responses—it relies on Huawei’s Ascend 910C chip. This move cuts costs while reducing dependency on US technology.
Huawei’s Ascend 910C Chip
AI researcher @Dorialexander noted that the Ascend 910C is used only for inference, not training. Training AI models demands significant computing power, which Nvidia GPUs handle better. Inference requires less power, making Huawei’s chip a viable alternative. This approach keeps performance high while saving costs.
Despite its strengths, the 910C has limitations. It lacks the power needed for training large-scale AI models. As a result, Chinese AI companies still rely on Nvidia for model development.
Huawei’s Next Move
To overcome this challenge, Huawei is developing the Ascend 920C chip. This new processor aims to compete with Nvidia’s Blackwell B200, the top AI chip in the industry. If successful, it could give Chinese firms a powerful domestic alternative.
The Future of AI Competition
DeepSeek’s rise shows China’s growing ambitions in AI. Its cost-efficient strategy and Huawei’s hardware advancements could reshape the industry. The battle between Chinese and US tech giants will define the future of AI. This competition will drive innovation, affordability, and global influence in the years ahead.