DeepSeek’s R2 AI Model Delay: Chinese Authorities Recommend Huawei AI Chips as Beijing Seeks NVIDIA Alternatives

DeepSeek’s reliance on Huawei’s AI chips has not panned out as expected, leading the company to pivot towards NVIDIA for the training of its R2 model. This shift underscores the challenges faced by the Chinese AI sector in its quest for self-sufficiency amid escalating technological tensions.

DeepSeek Turns to NVIDIA’s AI Chips for R2 AI Model Training

The Chinese artificial intelligence landscape has predominantly leaned on NVIDIA and its advanced ecosystem since the onset of the AI hype. In light of recent U. S.export restrictions, Beijing has accelerated efforts to cultivate domestic alternatives. Companies like Huawei have endeavored to offer solutions, including the Ascend 910C AI chip, but initial adoption rates have been less than favorable.

According to a Financial Times report, the release of DeepSeek’s R2 has been postponed as a result of challenges attributed to Chinese AI chips, particularly in comparison to NVIDIA’s technology stack. Originally slated for a May launch, the project faced delays due to Huawei’s chips grappling with performance issues, including instability and slower interconnect speeds. The absence of a viable substitute for the CUDA programming model further complicates matters, necessitating a dependence on NVIDIA’s infrastructure for high-quality AI model training.

Huawei Preps Ascend 910C To Tackle NVIDIA's H100 In China's Domestic AI Market

Chinese firms are currently facing significant hurdles with Huawei’s AI chips, which compels many to seek NVIDIA’s H100 AI chips, inadvertently ramping up demand for NVIDIA in Asian markets. Compounding this situation is an investigation into potential security vulnerabilities linked to NVIDIA chips entering China, with reports suggesting the Chinese government has advised local firms against their use due to concerns over tracking capabilities.

As DeepSeek prepares for the anticipated release of its R2 AI model in the coming weeks, it will be intriguing to monitor its performance. This development highlights the substantial gap that remains in China’s efforts to create robust domestic alternatives to NVIDIA’s offerings.

Source & Images