Figure According to US Securities and Exchange Commission official website
On August 31, local time in the United States, GPU giant Nvidia disclosed in a regulatory filing to the US Securities and Exchange Commission that the US government has introduced a new export license control that will Affecting NVIDIA's export of two high-performance GPU products to China and Russia.
Meanwhile, on September 1, according to media reports, a AMD spokesperson said that the company has received a new license requirement and will stop exporting MI250 artificial intelligence chips to China.
On September 1, Foreign Ministry spokesman Wang Wenbin said at a regular press conference that the U.S. approach is typical of technological hegemony. The United States has repeatedly generalized the concept of national security, abused national power, and attempted to use its own technological advantages to contain and suppress the development of emerging markets and developing countries. This move violates market economic rules, disrupts the international economic and trade order, and disrupts the stability of global industrial and supply chains. strongly oppose. The export restriction of
is aimed at the field of supercomputing and AI
"Similar to high-performance GPUs, it is generally used in the field of supercomputers or artificial intelligence ." Zhou Zhiping, a professor at the School of Information Science and Technology of Peking University, told Tianmu News.
In the disclosure document, Tianmu News reporter saw that Nvidia stated that the U.S. government notified Nvidia on August 26 that it had imposed future exports of the company’s A100 and H100 integrated circuits to China (including Hong Kong) and Russia in the future. new licensing requirements, effective immediately. The
license requirements also include any future NVIDIA integrated circuits with peak performance and chip-to-chip I/O performance equal to or greater than the A100 threshold, as well as any systems incorporating these circuits, exporting technology to support or develop covered products required license.
According to NVIDIA's official website,
A100 is a 7-nanometer GPU chip released by NVIDIA in September 2020, and H100 is a 4-nanometer GPU chip released by NVIDIA in March 2022 and is expected to be available in the third quarter. Both of these GPU products are among the highest performance GPU products available on the market.
can be seen in the introduction on NVIDIA's official website that the H100 has increased the speed of large language models by 30 times compared to the previous generation, and performed well in the fields of artificial intelligence training, real-time deep learning inference, exascale high-performance computing, and data analysis.
"It can be seen from the relevant information that the chips that are limited this time belong to the highest-end chips in terms of computing accuracy and speed." Song Changgeng, a senior analyst at Simmou Research, said in an interview with a Tianmu News reporter, such as some supercomputers Such GPUs need to be used for accelerated computing, and some companies doing AI research and development require such GPUs for model training.
Nvidia said in regulatory filings that the U.S. government's new rules could affect the company's ability to complete H100 product development on time or support existing A100 customers, and could require the company to relocate operations for certain businesses out of China. Nvidia is negotiating with the U.S. government to seek waivers for internal development and support. The company is reaching out to Chinese customers to try to provide data center products that are not affected by the new U.S. government regulations to meet customers' planned or future procurement needs. If customers need to purchase products affected by new U.S. government regulations, NVIDIA may apply for a license from the U.S. government, but there is no guarantee that a waiver or license from the U.S. government will be obtained or will be obtained in a timely manner.
"In fact, if the relevant GPU cannot be imported, it will not only affect the supercomputing itself, but also consider the impact on the upstream and downstream when the supply of technical points is cut off." Song Changgeng believes that the current application of AI is very extensive, such as various Face recognition is supported by artificial intelligence technology. If there is a lack of high-end GPU, model training will be affected, and the development process of related technologies will be slowed down.
intends to block China's cutting-edge technology
"The US's export license control has obvious intention to block China's development of cutting-edge technology." Song Changgeng said that in fact, this is not the first time the United States has done this.
On February 18, 2015, the US Department of Commerce included four Chinese technology centers related to Tianhe 2 into the "entity list" for export control, prohibiting US companies from exporting related chips to them. On June 21, 2019, the U.S. Department of Commerce announced that Chengdu Haiguang Integrated Circuit, Chengdu Haiguang Microelectronics Technology, Haiguang, Dawning andWuxi Jiangnan Institute of Computing Technology was included in the "Entity List". On April 8, 2021, the U.S. Department of Commerce announced that it has added seven Chinese supercomputer entities to the so-called "entity list". In an interview with the Economic Observer,
Deputy Secretary-General of the Beijing Semiconductor Industry Association, Zhu Jing, said that from the above process, it can be seen that for China's supercomputing, the United States has upgraded both in terms of attack methods and scope. If the news is true, this time the attack method will be upgraded from "supercomputing related units entering the entity list" to "direct ban on sales of related products that can provide services for supercomputing", resulting in the expansion of the affected area from the supercomputing field to the Internet field.
According to NVIDIA's official website
, NVIDIA expects that if the US government fails to approve or directly reject the company's application for shipment licenses to important Chinese customers, Chinese customers will also be reluctant to accept NVIDIA's alternative products. Potential sales revenue of $400 million was affected.
Correspondingly, China is extremely important to both NVIDIA and AMD. The
financial report shows that China (including mainland China and Hong Kong, China) is Nvidia’s second largest market, and its revenue scale is second only to Taiwan, China. In fiscal year 2022 (as of January 30, 2022), NVIDIA's revenue in China was US$7.111 billion, accounting for 26.42% of total revenue. As for AMD, the Chinese market is also its second largest market, and its revenue is only slightly smaller than the US market. In 2021, AMD's China revenue will be $4.096 billion, accounting for 24.92% of total revenue.
Song Changgeng believes that although the US export control will affect the development of related fields in China, it will help the localization of high-end GPUs, and it is also an opportunity for these startups to occupy the market.
On September 1, Ministry of Commerce spokesperson Shu Jueting said at a regular press conference that China has noticed the relevant situation. For some time, the United States has been abusing export control measures to restrict the export of semiconductor related items to China, which China firmly opposes. The relevant actions of the United States deviated from the principle of fair competition and violated international economic and trade rules, which not only harmed the legitimate rights and interests of Chinese companies, but also seriously affected the interests of American companies, hindered international scientific and technological exchanges and economic and trade cooperation, and affected the stability of global industrial and supply chains and the recovery of the world economy. shock. The U.S. side should immediately stop its wrongdoing, treat companies from all over the world including Chinese companies fairly, and do more things that are conducive to the stability of the world economy.
