Chinese AI firms push beyond Nvidia as DeepSeek turns to Huawei
China’s DeepSeek AI startup announced its latest model is optimized to run on Huawei chips, marking a shift away from Nvidia and accelerating Beijing’s push for domestic AI infrastructure. The move follows U.S. export controls, with Huawei planning to release its own training chip this year, though it will take time to rival Nvidia’s performance." "article": "China’s DeepSeek AI startup made a significant move in its latest model release, optimizing it to run on Huawei chips instead of Nvidia’s hardware. This marks a rare break from Western technology dominance and underscores China’s efforts to develop self-reliant AI systems amid U.S. export restrictions. The announcement comes ahead of a delayed summit between U.S. President Donald Trump and Chinese leader Xi Jinping, signaling Beijing’s confidence in advancing its AI capabilities despite trade tensions. DeepSeek’s decision reflects a broader trend among Chinese AI firms adapting to U.S. restrictions on Nvidia chips. While the company still uses Nvidia chips for training—likely through remote access—its shift to Huawei for inference demonstrates a strategic pivot. Inference, the process of generating AI responses, requires less computing power than training, making it a key area for domestic chip alternatives. Huawei has announced plans to release its own AI training chip this year, though it expects another year to match Nvidia’s current performance. Analysts note that U.S. export controls are not halting China’s AI progress but instead forcing the development of alternative technologies. Wei Sun, a principal AI analyst at Counterpoint Research, stated that the restrictions are pushing China to build its own hardware ecosystem. Nvidia CEO Jensen Huang has previously warned that strict export controls could lead to a divided global AI market, with China developing its own systems independent of Western hardware. The U.S. has recently eased some restrictions, allowing Nvidia to sell its H200 chip to China, but the long-term impact remains uncertain. DeepSeek’s move highlights the growing competition between American and Chinese tech giants in the AI chip sector.
China’s DeepSeek AI startup made a significant move in its latest model release, optimizing it to run on Huawei chips instead of Nvidia’s hardware. This marks a rare break from Western technology dominance and underscores China’s efforts to develop self-reliant AI systems amid U.S. export restrictions. The announcement comes ahead of a delayed summit between U.S. President Donald Trump and Chinese leader Xi Jinping, signaling Beijing’s confidence in advancing its AI capabilities despite trade tensions. DeepSeek’s decision reflects a broader trend among Chinese AI firms adapting to U.S. restrictions on Nvidia chips. While the company still uses Nvidia chips for training—likely through remote access—its shift to Huawei for inference demonstrates a strategic pivot. Inference, the process of generating AI responses, requires less computing power than training, making it a key area for domestic chip alternatives. Huawei has announced plans to release its own AI training chip this year, though it expects another year to match Nvidia’s current performance. Analysts note that U.S. export controls are not halting China’s AI progress but instead forcing the development of alternative technologies. Wei Sun, a principal AI analyst at Counterpoint Research, stated that the restrictions are pushing China to build its own hardware ecosystem. Nvidia CEO Jensen Huang has previously warned that strict export controls could lead to a divided global AI market, with China developing its own systems independent of Western hardware. The U.S. has recently eased some restrictions, allowing Nvidia to sell its H200 chip to China, but the long-term impact remains uncertain. DeepSeek’s move highlights the growing competition between American and Chinese tech giants in the AI chip sector.
This content was automatically generated and/or translated by AI. It may contain inaccuracies. Please refer to the original sources for verification.