Nvidia will adjust its export of chips to China, how big will the impact be?

On May 18, Reference News quoted a report from the Singapore United Morning News website on May 17, reporting that Nvidia CEO Jensen Huang stated that due to US government restrictions on the export of H20 chips based on the Hopper architecture to China, the company is reevaluating its market strategy in China. However, they will not launch the Hopper series of chips in the future. It is understood that the H20 is based on the Nvidia Hopper architecture, with two versions of memory. Compared to the H100, its performance is significantly reduced by about 80%, and tensor cores are disabled, limiting overclocking and cluster expansion. Morgan Stanley analysts previously lowered Nvidia's revenue expectations, citing concerns about the impact of the latest US chip export restrictions to China on Nvidia. They expect Nvidia's data center revenue to decline by 8% to 9% in the coming quarters. Some analysts point out that domestic AI chip manufacturers may have a huge market space to validate the performance, reliability, ecosystem compatibility, and supply chain stability of their products. Omdia's chief AI analyst Sullian Jie said that domestic companies' demand for AI is increasing, and the Chinese smart computing market is still growing, with high requirements for localization. Jensen Huang previously stated in an interview that China is a very large AI market, and the AI chip market in China may reach $50 billion in the next few years. Missing out on this market would be a huge loss; the United States must realize that it is not the only country in the AI competition. He also stated that export restrictions on AI chips would actually harm US national security. In the medium to long term, US export restrictions may provide opportunities for the development of Chinese domestic chips, accelerating the process of domestic substitution.