According to reports, Leo Wang, Chief Financial Officer of Chinese autonomous driving company Pony.ai, stated that in anticipation of potential further tightening of NVIDIA chip export restrictions, Chinese autonomous driving firms have been preparing alternative solutions for several years. Currently, the demand for AI training compute power in autonomous driving exceeds supply, and Pony.ai is alleviating this pressure by improving model deployment efficiency, utilizing mid-tier compute resources in certain scenarios, and increasing adoption of domestic chips.

Wang noted that domestic chips have emerged as an alternative option for AI training, with the maturity of such solutions having improved significantly compared to a few years ago; however, they still lag behind NVIDIA chips in terms of cost-effectiveness. NVIDIA currently provides automotive processors, model training GPUs, and full-stack collaboration support to the autonomous driving industry, and any changes in chip supply could impact training and deployment plans for certain companies.