On Tuesday, the U.S. Department of Commerce revealed its intentions to hinder the forthcoming sale of more advanced artificial intelligence chips to China. The U.S. government asserted that these new regulations are aimed at eliminating the loopholes that emerged after the implementation of AI chip export restrictions last year.
Following this announcement, shares of chip-related companies saw declines in Tuesday’s trading. Nvidia’s stock closed down approximately 5%, while Broadcom and Marvell experienced drops of around 2% and 1%, respectively. AMD shares fell by more than 1%, and Intel’s stock declined by about 1.4%.
The prior restrictions had prohibited the sale of Nvidia’s H100, a preferred processor for AI firms in the U.S., such as OpenAI. Instead, Chinese companies were able to acquire a slightly downgraded version known as the H800 or A800 to comply with U.S. restrictions, primarily achieved by reducing the on-device connection speed, referred to as an interconnect. The new rules will now extend to ban these alternative chips as well, according to senior administration officials.
These restrictions could also impact chips sold by Intel and AMD. Additional regulations are expected to impede the sale and export of semiconductor manufacturing equipment to China, including companies like Applied Materials, Lam, and KLA.
These restrictions effectively cut off a significant and expanding market for AI semiconductors, raising concerns about potential economic retaliation from the Chinese government against U.S. businesses operating in the country.
Nvidia seemed to have anticipated these restrictions and stated in August that they would not have an immediate material impact on earnings but could pose long-term challenges. The company emphasized its compliance with all applicable regulations and its commitment to delivering products that support various applications across industries, expressing confidence that there would be no immediate, substantial financial impact on their results.
Nvidia specified in an SEC filing on Tuesday that the restrictions would apply to several of their chip models and entire systems sold with these chips. They also noted potential delays in the development of new products.
The primary objective of these U.S. restrictions is to prevent Chinese access to advanced semiconductors that could potentially fuel advancements in artificial intelligence, particularly for military purposes. U.S. Commerce Secretary Gina Raimondo emphasized that these measures are not intended to hinder Chinese economic growth but rather to control access to computing power that could be leveraged in ways that pose threats to the U.S. and its allies, particularly in military and modernization applications.
The U.S. will restrict the export of data center chips that surpass performance thresholds set in October or a new performance density benchmark measured in flops per square millimeter. Companies wishing to export AI chips to China or embargoed regions will need to notify the U.S. government.
Additionally, the U.S. government plans to expand the list of semiconductor manufacturing equipment subject to these restrictions.
It’s important to note that chips for consumer products, such as game consoles or smartphones, will not be subject to export controls, although companies may need to inform the Commerce Department about their chip orders if they meet specific performance criteria.
The U.S. government is also addressing loopholes related to shipping chips to companies headquartered in China or other embargoed regions, like Macao, to prevent foreign subsidiaries from buying chips and then forwarding them to China.
Secretary Raimondo stressed that the new restrictions would affect only a small fraction of chip exports to China and emphasized that, even with these updated rules, China would still import a substantial number of semiconductors from the United States.
These rules will undergo a 30-day period of public notice before taking effect, according to U.S. officials.
