DeepSeek gives China’s chipmakers an edge in the race for cheap AI

Beijing. The rise of DeepSeek’s artificial intelligence (AI) models is giving some Chinese chipmakers like Huawei a better chance to compete against more powerful American processors in the domestic market. Huawei and its Chinese peers have struggled for years with Nvidia to create top-end chips that can compete with the American firm’s products for training models, a process in which data is fed into algorithms to help them make accurate decisions.

However, DeepSeek’s models, which focus on “inference”, or when an AI model draws conclusions, optimize computational efficiency rather than relying solely on raw processing power. That’s one reason why the model is expected to partially bridge the gap between Chinese-made AI processors and their more powerful American counterparts, analysts say. Huawei and other Chinese AI chip makers such as Haigon, Tencent-backed Enflame, TsingMicro and Moore Threads have released statements in recent weeks claiming products will support the DeepSeek model, although few details have been released.

Huawei declined to comment. Moore Threads, Hygon Enflame and TsingMicro did not respond to Reuters queries for further comment. Industry executives are now predicting that the open-source nature of DeepSeq and its low fees could boost AI adoption and the development of real-life applications for the technology, helping Chinese firms circumvent US export restrictions on their most powerful chips.

Even before DeepSeek made headlines this year, products like Huawei’s Ascend 910B were considered by customers like ByteDance to be better suited for less computationally intensive “inference” tasks, the post-training stage in which trained AI models make predictions or perform tasks, such as through a chatbot. In China, dozens of companies from automakers to telecom providers have announced plans to integrate DeepSeek’s models with their products and operations.

“This development is very much in line with the potential of Chinese AI chipset vendors,” said Lian Jae Su, chief analyst at tech research firm Omdia. “Chinese AI chipsets struggle to compete with Nvidia’s GPUs (graphics processing units) in AI training, but AI inference workloads are much more forgiving and require much more local and industry-specific understanding,” he said.