China hits Nvidia with antitrust investigation in response to US sanctions

Transactions between large companies are traditionally approved by regional antimonopoly authorities, even if the headquarters of the participants are not located in a particular country. Chinese authorities are ready to reconsider the results of Nvidia’s deal to buy Mellanox in 2020, making a profile investigation a response to American sanctions.

Image source: NVIDIA

China’s antitrust watchdog SAMR has launched an investigation into Nvidia, looking into its recent behavior as well as the circumstances surrounding its 2020 purchase of Israel’s Mellanox Technologies, Bloomberg reports. Let us recall that at that time Chinese antimonopoly officials delayed their approval of the transaction between these companies the longest, but in the end they nevertheless began to oppose it. The approval came with the condition that Nvidia would not discriminate or harass Chinese companies.

In particular, the Chinese authorities at one time insisted that Mellanox would provide information about its new products within 90 days after the start of their deliveries for the needs of parent company Nvidia. Secondly, Nvidia agreed to adapt its solutions to work with the products of Chinese companies, at least at the software level.

Against the background of this not very pleasant news for Nvidia, the company’s stock price fell by more than 2% before the start of trading on Monday. It is noteworthy that even in the face of increasing US export restrictions, Nvidia continued to create computational accelerators that it could officially supply to China. American antitrust regulators also had questions about Nvidia because they were not sure that the company allowed customers to switch to products from other suppliers without any consequences. The company was suspected of imposing penalties against customers who switched to competitors’ products. French antitrust officials are also conducting their investigation into Nvidia and its policy in the distribution of chips for AI systems, and so far its results are not entirely clear.