The Biden administration said on Tuesday that it will suspend sending more sophisticated AI processors made by Nvidia and other companies to China. This move is one of several steps Beijing is doing to prevent it from obtaining cutting-edge American technology that would enable it to bolster its armed forces.
In a press briefing on Monday night, top administration officials outlined the regulations, which blacklist Chinese chip designers Moore Thread and Biren and limit access to a wider range of sophisticated processors and chipmaking facilities to a bigger number of nations, including Iran and Russia.
“At least annually,” according to Commerce Department Secretary Gina Raimondo, the new laws are intended to impede China’s military growth by plugging gaps in regulations that were published in October of last year.
She emphasised that the administration was not trying to harm Beijing economically and that the objective was to restrict China’s access to “advanced semiconductors that could fuel breakthroughs in artificial intelligence and sophisticated computers that are critical to (Chinese) military applications.”
In addition to the years-long technological battle between the US and China, tensions between the two countries were further heightened by the extensive restrictions announced in October of last year.
Top AI chip designer Nvidia said it conforms with standards and does not anticipate “a near-term meaningful impact on our financial results” in a statement released after the guidelines were published.
In order to continue exporting to China, the business has produced chips like the A800 and H800 that went directly over the prior regulations. AMD, which was also hit by the legislation, has stated that it intends to take a similar approach.
Since the regulations were implemented last year, Nvidia’s company has flourished since its chips, available solely in China, continue to outperform competitors. The Silicon Valley company is now selling nearly every chip it can get since demand for chips is now higher globally than supply.
The new rules will affect Nvidia’s A800 and H800 processors because of a modification to the chip’s specifications meant to capture more chips.
However, the majority of consumer chips used in computers, cellphones, and games will be excluded from the regulations; however, some may be subject to licencing and reporting requirements by US authorities.
Raimondo emphasised that the restrictions were not intended to punish American firms, saying, “The fact is, China will import hundreds of billions of dollars of semiconductors from the United States even after the update of this rule.”
The prior regulations enforced a two-pronged test that evaluated a chip’s computational capacity as well as its inter-chip communication capabilities, an essential metric in artificial intelligence supercomputers that string thousands of chips together to process enormous volumes of data.
For the Chinese market, Nvidia and Intel developed unique CPUs that maintained strong computational capabilities but constrained transmission rates to adhere to prior regulations.
Both Biren and Moore Thread are Chinese businesses that were started by former Nvidia workers in China with the goal of competing with the American AI chip giant. Before sending equipment to them, U.S. vendors will now need to comply with a stringent licencing requirement.
According to a senior administration official, the regulations that were unveiled on Tuesday do away with the communication speed restrictions and place more emphasis on processing performance. As a result, Nvidia’s A800 and H800 processors will no longer be sold in China.
According to a senior administration official, American authorities on Tuesday introduced a new policy to limit chips that beyond a specific threshold of “performance density,” a metric that focuses on how much processing capacity can be crammed into a specific piece of silicon.
According to a senior administration official, the purpose of this regulation is to stop businesses from attempting to circumvent limitations on whole chips by employing a technique known as “chiplets,” which allows businesses to attempt to combine small bits known as chiplets into a huge chip that contravenes the
Chiplets are now a crucial component of China’s technological plan to develop its chip sector, according to Reuters story from July. Analysts have speculated that Chinese businesses may exploit this technology to get over US limitations.
According to the authorities, the new regulations further increase the number of nations to whom licences for advanced chip exports must be obtained. These include over 40 countries that are subject to U.S. arms embargoes and present dangers of diversion to China.
This action seems to be a follow-up to a letter Nvidia received in August stating that shipments of its A100 and H100 chips to other countries, including several in the Middle East, were restricted.
Chips will not be shipped to subsidiaries of corporations anywhere in the world if their parent company are based in China, Macau, or other nations that are subject to arms embargoes, according to a report confirmed by Reuters. The action is a part of an attempt to prevent the chips from being remotely or unlawfully carried into China.
Additionally, the Biden administration expanded the list of equipment prohibited from entering China and imposed licencing requirements on 21 non-Chinese countries for chipmaking instruments.
Confirming a previous Reuters report, officials went to great lengths to emphasise that National Security Advisor Jake Sullivan, Treasury Secretary Janet Yellen, and Commerce Secretary Gina Raimondo had forewarned Chinese officials that the changes were imminent.