Nvidia and AMD: US restricts sale of AI training chips to Russia and China
Nvidia and AMD: US restricts sale of AI training chips to Russia and China
Competition rages between the world’s major economic powers. Recently, the US government introduced a new policy that limits the export of high-end computer chips used to conduct AI research. It seems that the main targets are China and Russia. According to big tech companies Nvidia and Advanced Micro Devices (AMD), the government has blocked them from selling certain chips to the Russian and Chinese markets.
These export restrictions are intended to curb the rise of rival powers. By way of reasons, the United States referred to possible military applications of the technology. These prohibitions will prevent US products from being used or diverted for dishonest purposes. The products affected by these regulations are Nvidia’s A100 and H100 GPUs and AMD’s MI250 accelerator chip.

Big financial impacts on Nvidia and AMD?
This decision had serious repercussions on the turnover of Nvidia which is the larger of the two companies. Shares of the latter fell more than 6% yesterday following the announcement of the restrictions. Which is not surprising because China would represent about 10% of its sales in the data center market. Note that this Californian company would have 400 million dollars in sales planned for this quarter which could go up in smoke because of this regulation.
Russia has some major tech companies like Yandex but that won’t affect Nvidia. Currently, it does not market any product in this country. As for AMD, it will be less impacted because in fact, all of the acceleration of calculations related to AI is done on hardware from the competing brand.
How were the chips affected by the restriction chosen?
So far, the US government has refrained from giving details of the criteria they use to target the chips. However, it is obvious that the high-end AI chips in the market are the main affected like the A100, H100 and MI250.
Note that these systems are mostly used to train a range of machine learning applications, from facial recognition to text generation. Moreover, Big Tech uses its chips to create internal superheaters for R&D.
source