The Biden administration unveiled its “”, which aims to restrict the export of the most coveted GPUs for AI applications. Although it does not mention the nation by name, it is widely seen as a way to prevent China from overtaking the United States in AI development.
The rule provides three licensing levels. The first level is unlimited and includes the domestic market as well as 18 strategic allies. The majority of countries belong to a second tier, which will cap the amount of computing power they can import through the United States’ top GPUs. The third tier includes China, Russia, Iran and North Korea, and prohibits American companies from selling their most powerful GPUs there.
U.S.-based companies would also not be able to share many details of their AI software models with countries outside that first tier, and would have to seek federal government permission before building large data centers in a tier two country.
Many parties, including the Semiconductor Industry Association (SIA), have published condemning the decision, saying the restrictions will do more to push countries to work with China. “The new rule risks causing unintended and lasting damage to the U.S. economy and global competitiveness in semiconductors and AI by ceding strategic markets to our competitors,” SIA wrote.
NVIDIA also opposed it, with Ned Finkle, the company’s vice president of government affairs, saying the Biden administration “seeks to undermine American leadership with a 200-plus page regulatory quagmire, written in secret and without appropriate legislative review.”
The rule has a 120-day comment period, so whether it will survive the new Trump administration remains an open question.
