The US government proposed and immediately implemented a new rule restricting the export of computer chips used in supercomputers and artificial intelligence to Russia and China. NVIDIA revealed in an SEC filing that the government informed the company of the new license requirement affecting one of its current (A100) and upcoming (H100) GPUs designed to accelerate machine learning tasks.
According to the government, the new license requirement will address the risk that the affected products will be used for military purposes by either country. According to The New York Times, AI and machine learning are used for a variety of purposes, including weapon development and surveillance. According to a spokesperson for the US Department of Commerce,
“While we are unable to outline specific policy changes at this time, we are implementing additional actions related to technologies, end-uses, and end-users to protect US national security and foreign policy interests.”
This requirement follows another export control rule issued in mid-August that would deny China access to advanced chip design software required to produce next-generation processors.
In its filing, NVIDIA admits that the restriction may impede its ability to complete the development of the H100 chip. It does not sell its products in Russia, but for the third fiscal quarter, it expects to earn around $400 million from sales to Chinese companies. That is money that could be used to develop future products. All of those potential sales could be lost if the chipmaker’s Chinese customers refuse to buy its alternative offerings or if the government refuses to grant licenses to its biggest customers.
According to Reuters, while the new license requirement will prevent AMD from exporting its MI250 chips to China, the company does not believe it will have a significant impact on its business.