DeepSeek’s AI Model R2 Delayed Due to Huawei Chip Challenges

Chinese AI firm DeepSeek has postponed the launch of its R2 model after encountering technical hurdles with Huawei's Ascend processors. The company, pressured by Beijing to reduce reliance on U.S. technology, initially attempted to train R2 using domestic chips but ultimately reverted to Nvidia hardware for critical training phases.

The setback highlights China's ongoing struggle to match U.S. semiconductor capabilities in advanced AI applications. While Huawei chips proved adequate for inference tasks, their limitations in model training forced DeepSeek into a hybrid approach—using Nvidia for training while maintaining some domestic chip usage for deployment.

This delay comes amid heightened U.S.-China trade tensions over chip exports, with Washington recently allowing limited sales of Nvidia's H20 processors to Chinese firms. The technical difficulties underscore the competitive gap facing China's semiconductor industry as it races to achieve self-sufficiency in critical technologies.