Nvidia AI chip export ban expected to impact U.S., China AI race

Last week was filled with reports that U.S. government officials have ordered Nvidia to stop exporting its A100 and H100 GPUs to China, a signal of what Reuters called a “major escalation” of a U.S. campaign to slow China’s technological capabilities. 

Nvidia also said last Thursday that the U.S. government will allow it to continue developing its H100 artificial intelligence (AI) chip in China. 

The move immediately led to speculation about the impact on Chinese firms, including their ability to compete in areas of AI research such as image recognition. Questions also came quickly about the impact on Nvidia’s business in China. 

Nvidia’s upcoming H100 processor has been touted as “the next-generation flagship of Nvidia’s AI processor solution for the data center,” while the A100 is Nvidia’s mainstay enterprise GPU commonly deployed in supercomputers. According to Handel Jones, CEO of International Business Strategies, a chip consultancy, the cost to develop the H100 is in excess of $1 billion, and it took Nvidia, with its very strong R&D organization, over two years to implement the design. 

Short- and long-term impact on Nvidia

The short-term impact on Nvidia is potential reduction in revenues of $400 million in each of the next two quarters, Jones told VentureBeat by email. Last Wednesday, Nvidia said it had booked $400 million in Chinese sales of the affected chips this quarter, which could be lost if its customers decide not to buy alternative Nvidia products. 

The longer-term impact, however, is the loss of Nvidia’s China market, which “will result in Chinese companies producing their own AI accelerators,” said Jones, whose upcoming book, “When AI Rules the World: China, the U.S., and the Race to Control a Smart Planet,” details how China is progressing in AI. He also claims that by 2030, China will have a leadership position in many of the important areas of AI.

The A100 and H100 products are critical for training AI in data centers, he added, and companies in China that are impacted include Alibaba Group, Tencent, Inspur Group, H3C, Baidu and Lenovo Group.

Overall, IBS projects the semiconductor market to grow between 8% to 10% in 2022 and 2% to 8% in 2023, while the organization expects additional restrictions to be placed on the sales of semiconductor products to China. 

U.S. government concerns

According to Dan Pickard, international trade and national security expert and attorney at Buchanan Ingersoll & Rooney, it is widely believed that China’s state-led industrial policies are intended to advance both their commercial interests and their military development. “It is for these reasons that there is bipartisan support to ensure that U.S. technology is not transferred to China that ends up supporting the People’s Liberation Army,” Pickard told VentureBeat by email. 

The export ban came just one month after President Biden signed the CHIPS and Science Act into law, which includes $52.7 billion to increase domestic semiconductor production. Just two days ago, the U.S. Commerce Department detailed its plans to distribute the money, which includes $28 billion earmarked for underwriting expansions of existing semiconductor manufacturing facilities or the construction of new ones. 

It is “difficult to overstate” the concerns of the U.S. government regarding the semiconductor manufacturing shortages, said Pickard. 

Nvidia export ban impacts Chinese universities

On the Chinese side, a recent Reuters review showed that high-profile universities and state-run research institutes in China have been relying on a U.S. computing chip to power their AI technology, but its export to the country has now been restricted. 

For example, the review showed that Tsinghua University, China’s highest-ranked higher education institution globally, spent over $400,000 last October on two Nvidia AI supercomputers, each powered by four A100 chips.

Still, some Chinese companies have insisted U.S. restrictions would have a limited impact. Yesterday CNBC reported that Chinese electric car maker Nio said U.S. restrictions on Nvidia chip sales to China won’t affect the automaker’s business.

“Based on our estimations, our computing power is sufficient for our autonomous driving technology development in the aspect of AI training for now,” William Li, founder, chairman and CEO of Nio, said, according to the article. “And we have been working very closely with our partner Nvidia.”

More controls likely to come

Pickard pointed out that because of Nvidia’s significant position in the semiconductor industry, “it is front and center in the United States’ contest against China to see who will be the world leader in 21^st century technology.” This contest, he said, is “at its hottest in regard to artificial intelligence.” 

Chinese companies that need access to these types of GPUs should definitely be concerned, he added.

“One of the few areas in Washington where there is bipartisan consensus concerns controlling the export of key technologies to China,” he said. “It is reasonable to expect more, not less, controls on exports to China for the foreseeable future.” 

He also pointed out that U.S. companies are also going to need to be hypervigilant in regard to the increasing level of controls on the export of national security-sensitive items to China and Russia.

“The Department of Commerce has been clear that establishing export controls on certain new technologies are essential to the national security of the United States,” he emphasized, adding that the President has wide discretion in regard to prohibiting the export of U.S. goods, technology and services under the existing export controls and economic sanctions laws.

“These laws are powerful foreign policy tools and have been used more aggressively over the past several administrations,” he said.