DeepSeek Switches to Nvidia AI Chips After Huawei R2 Failure
DeepSeek, the Chinese AI startup, has reverted to using Nvidia’s AI chips for its new R2 model after encountering persistent technical issues with Huawei’s Ascend chips. The challenges were significant enough to delay the model’s planned launch and force DeepSeek to return to Nvidia’s more reliable hardware for training the AI system.
The move highlights ongoing struggles with domestic chip technology in China despite government efforts to promote self-sufficiency. Huawei’s CEO recently admitted their chip technology is still a generation behind leading competitors, which impacted DeepSeek’s ability to deliver its AI model on schedule. Huawei engineers even collaborated closely with DeepSeek, but could not resolve the training problems.
Nvidia Chips Regain Favor Amid Technical Setbacks
DeepSeek’s decision underscores Nvidia’s continued dominance in AI hardware, especially for demanding model training tasks. While China pushes for localized technology, the superior performance and stability of Nvidia’s GPUs remain essential for complex AI development. DeepSeek is reportedly still working on using Huawei chips for less intensive AI inference tasks, but the fallback to Nvidia chips was necessary to maintain competitive progress.
This development suggests China’s AI ambitions face technical hurdles in chip development, even as domestic companies strive to reduce dependence on foreign technology. DeepSeek’s experience puts a spotlight on the balance between national policy priorities and practical performance demands in the AI chip industry.
The delay in DeepSeek’s R2 model release also reverberated in the AI market, where Nvidia’s chips remain critical despite increased competition and export restrictions.