the development of DeepSeek’s next-generation AI, DeepSeek R2, has been delayed because of US export restrictions on NVIDIA’s H20 chips. The H20 chip is only available in limited quantities because of tightened US trade policies. This is despite the fact that it was designed specifically for China’s market. DeepSeek used 50,000 NVIDIA GPUs, including H20s, for its R1 model. The H20 shortage has reportedly affected the deployment of DeepSeek’s R1 model. Many users are running the restricted chip. The Information reported that DeepSeek had been working on the R2 for months. However, CEO Wenfeng Liang is still not satisfied with the current performance of the model, and it isn’t yet ready for deployment. The export controls are part a broader US effort to limit China’s ability to access advanced AI hardware. This is putting pressure on Chinese AI companies that rely on GPUs imported from overseas.[The information]
