The U.S. Department of Commerce announced Tuesday that it plans to prevent the sale of more advanced artificial intelligence chips to China in the coming weeks.
The U.S. government says the new rules are intended to close loopholes that popped up after last year’s restrictions on AI chip exports went into effect.
Shares of chip stocks took a leg lower in premarket trading on the news. Nvidia was down about 4% while Broadcom, Marvell and AMD slipped about 2%. Intel fell about 1.7%.
Those earlier restrictions banned the sale of the Nvidia H100, which is the processor of choice for AI firms in the U.S. such as OpenAI. Instead, Chinese companies were able to buy a slightly slowed-down version called the H800 or A800 that complies with U.S. restrictions, primarily by slowing down an on-device connection speed, called an interconnect.
The new rules will ban those chips as well, senior administration officials said in a briefing with reporters.
The restrictions could also affect chips sold by Intel and AMD. Other rules will likely hamper the sale and export to China of semiconductor manufacturing equipment from companies such as Applied Materials, Lam and KLA.
The restrictions cut off a big and growing market for AI semiconductors, and could raise concerns that the Chinese government will retaliate economically against U.S. firms doing business in the country.
The goal of the U.S. restrictions is to prevent Chinese access to advanced semiconductors that could fuel breakthroughs in artificial intelligence, especially with military uses, U.S. Commerce Secretary Gina Raimondo said on a call with reporters. They’re not intended to hurt Chinese economic growth, U.S. officials said.
“The updates are specifically designed to control access to computing power, which will significantly slow the PRC’s development of next-generation frontier model, and could be leveraged in ways that threaten the U.S. and our allies, especially because they could be used for military uses and modernization,” Raimondo said.
Senior administration officials say the U.S. will simply restrict the export of data center chips if they exceed a performance threshold set last October, or exceed a new performance density threshold benchmark measured in flops per square millimeter.
Companies that want to export AI chips to China or other embargoed regions will have to notify the U.S. government.
Senior administration officials also said they plan to expand the list of semiconductor manufacturing equipment subject to U.S. restrictions.
Chips for consumer products, like game consoles or smartphones, will not be subject to the export controls, although companies may have to tell the Commerce Department about their orders if the chips are fast enough.
The U.S. government is also closing loopholes dealing with how to ship chips to companies that are headquartered in China or other embargoed regions such as Macao, to prevent a loophole where a foreign subsidiary buys chips and then ships them into China.
Raimondo said that the new restrictions will only affect a small fraction of chip exports to China.
“The fact is China, even after the update of this rule, will import hundreds of billions of dollars of semiconductors from the United States,” Raimondo said.
Nvidia seems to have anticipated the restrictions, and said in August that they would not have an immediate material effect on earnings, but might hurt over the long term.
“Given the strength of demand for our products worldwide, we do not anticipate that additional export restrictions on our data center GPUs, if adopted, would have an immediate material impact to our financial results,” Nvidia CFO Colette Kress said in August on an earnings call. “However, over the long term, restrictions prohibiting the sale of our data center GPUs to China, if implemented, will result in a permanent loss of an opportunity for the U.S.”
The rules will be available for public notice for 30 days, then will go into effect, U.S. officials said.
— CNBC’s Kristina Partsinevelos contributed reporting.