DeepSeek’s AI Models Give Chinese Chipmakers an Edge

China’s AI chip industry is receiving a significant boost from DeepSeek’s artificial intelligence (AI) models, which are helping domestic companies compete against U.S. chip giants like Nvidia.

While Huawei and other Chinese firms have long struggled to match Nvidia’s high-performance AI training chips, DeepSeek’s models focus on AI inference—the phase where trained models generate conclusions. This emphasis on computational efficiency rather than raw processing power is allowing Chinese-made AI processors to close the performance gap.

Chinese AI Chipmakers Embrace DeepSeek’s Technology

Several Chinese AI chipmakers have expressed support for DeepSeek’s models, including:
Huawei
Hygon
Tencent-backed EnFlame
Tsingmicro
Moore Threads

Despite their public endorsements, most of these companies have not disclosed specific technical details on how they plan to integrate DeepSeek’s models into their products. Huawei declined to comment, and other chipmakers have yet to respond to media inquiries.

Overcoming U.S. Export Curbs with AI Inference

Industry experts believe that DeepSeek’s open-source framework and low operational costs will drive AI adoption in China. This is particularly crucial as the country faces U.S. export restrictions on high-end AI chips.

Even before DeepSeek became widely recognized, Huawei’s Ascend 910B chip was already being used for AI inference by companies like ByteDance. Unlike AI training, inference requires less computational power, making it an area where Chinese chipmakers can thrive despite technological limitations.

China’s AI Expansion with DeepSeek

From automakers to telecom companies, dozens of Chinese firms are now incorporating DeepSeek’s AI models into their operations.

According to Lian Jye Su, a chief analyst at tech research firm Omdia, Chinese AI chipsets struggle with training but are well-positioned for inference tasks that demand localized, industry-specific expertise.