In an attempt to strengthen national security and foreign policy interests, the United States government recently ordered companies to stop selling advanced artificial intelligence (AI) chips to China and Russia.
Recently, Nvidia (NVDA) and Advanced Micro Devices (AMD) indicated that the U.S. government ordered them to halt exports of high-performance chips to China. The new limits affect high-end chips known as graphics processing units (GPUs). These products were originally developed for imaging in video games but, in the past decade, have been widely deployed in the largest supercomputers used by scientists and internet companies for applications such as recognizing speech and objects in photographs. Supercomputers can be used in weapons development and intelligence gathering, with larger systems in China linked to surveillance of the country’s minorities. The AI capabilities increasingly are being used for other purposes, such as identifying faces in videos.
The new restrictions implicate NVDA’s A100 chipset, used to power data centers for AI, data analytics, and high-performance computing, as well as its DGX Systems and forthcoming H100 chips. They also implicate AMD’s MI250 chip, used in data centers for high-performance computing and AI technology. Future chips will also be covered under the restrictions if they equal or surpass the current silicon in terms of peak performance and chip-to-chip I/O performance. Chinese customers of NVDA and AMD’s chips may apply for exemptions from the restrictions from the U.S. government; however, no assurances are given that they would be granted.
In light of these recent restrictions and efforts, it is likely other companies making similar advanced computing software may be implicated in the future. The restrictions will likely implicate the future advancement of AI, both domestically and abroad. For example, on September 1, NVDA’s stocks fell nearly 9% in trading following the announcement. The company also expects a $400 million hit to its revenue if Chinese companies do not purchase non-restricted products. In addition to monetary impact, NVDA’s forthcoming H100 chips are partially developed in China. While the government has indicated that NVDA may continue its development abroad, time will tell what impact the new restrictions or the Chinese government’s response will have. Across the globe, Asia Times reports that the “development of China’s artificial intelligence sector is expected to be slowed in the coming few years by the United States’ new ban on exports of several high-end chips made by Nvidia and AMD.” However, it’s also possible that Chinese companies will increase their own research and development into AI chips.
The government’s position is that the new restrictions relate to national security and foreign policy interests. The Department of Commerce (DOC) commented that the goal is to prevent “China’s acquisition and use of U.S. technology in the context of its military-civil fusion program to fuel its military modernization efforts, conduct human rights abuses, and enable other malign activities.” The DOC’s International Trade Administration has also requested public comments regarding “international AI policies, regulations, and other measures which may impact U.S. exports of AI technologies.” The goal is to solidify “U.S. leadership in emerging technologies, including AI, to drive U.S. innovation and global competitiveness.” A number of the questions seek input on the current state of trade export laws and their implications on AI products and services. Several questions also seek input on the challenges that companies may face when it comes to the protection of AI intellectual property rights when engaging in overseas dealings. Comments are due by October 17, 2022.
The Nixon Peabody team will monitor developments and assess public comments once available. Stay tuned for an update.