DeepSeek AI Model Development Delayed by US Chip Export Restrictions
The development of DeepSeek’s next-generation artificial intelligence model, DeepSeek-R2, has been significantly delayed due to United States export restrictions on NVIDIA’s H20 chips. According to a report by The Information, these specialized chips, designed specifically for the Chinese market, are currently available only in limited quantities because of tightened US trade policies. DeepSeek previously utilized 50,000 NVIDIA Hopper GPUs, including 30,000 H20 units, to train its earlier R1 model. The ongoing shortage of H20 chips is not only hindering the progress of the R2 model but is also affecting the deployment and operation of the existing DeepSeek-R1 model. Sources indicate that while DeepSeek has been aggressively working on the R2 model for several months, CEO Wenfeng Liang remains dissatisfied with its current performance, deeming it unready for public deployment. This situation highlights the mounting pressure on Chinese AI firms that rely heavily on high-performance GPUs from overseas. The export controls are part of broader US efforts to restrict China’s access to advanced AI hardware, creating significant operational challenges for leading technology companies in the region.
Wire timeline
DeepSeek AI Model Development Delayed by US Chip Export Restrictions
The development of DeepSeek’s next-generation artificial intelligence model, DeepSeek-R2, has been significantly delayed due to United States export restrictions on NVIDIA’s H20 chips. According to a report by The Information, these specialized chips, designed specifically for the Chinese market, are currently available only in limited quantities because of tightened US trade policies. DeepSeek previously utilized 50,000 NVIDIA Hopper GPUs, including 30,000 H20 units, to train its earlier R1 model. The ongoing shortage of H20 chips is not only hindering the progress of the R2 model but is also affecting the deployment and operation of the existing DeepSeek-R1 model. Sources indicate that while DeepSeek has been aggressively working on the R2 model for several months, CEO Wenfeng Liang remains dissatisfied with its current performance, deeming it unready for public deployment. This situation highlights the mounting pressure on Chinese AI firms that rely heavily on high-performance GPUs from overseas. The export controls are part of broader US efforts to restrict China’s access to advanced AI hardware, creating significant operational challenges for leading technology companies in the region.
TechNode