The United States will take steps to prevent U.S. chipmakers from circumventing government restrictions on the sale of semiconductors to China, a U.S. official said, as part of the Biden administration's upcoming move to block more exports of artificial intelligence chips.

Details of the new rules, which will be added to sweeping U.S. restrictions on exports of advanced chips and chip-making equipment to China announced last October, were first reported by Reuters. Other people familiar with the matter said the update was expected to be released this week, though such timetables are often delayed.

The new rules will block some artificial intelligence chips that happen to meet current technical parameters, while requiring companies to report shipments of other chips, said the official, who asked not to be identified.

A spokesman for the U.S. Commerce Department, which oversees export controls, declined to comment.

The latest crackdown on Chinese tech exports coincides with U.S. efforts to thaw the difficult relationship between the world's two largest economies. Several senior Biden administration officials have met with their Chinese counterparts in recent months, and the latest round of rules could complicate diplomatic efforts.

The Biden administration has said it designed the export restrictions to prevent U.S. chips and equipment from strengthening China's military. Beijing accuses the U.S. of abusing export controls to suppress Chinese companies. The restrictions mark a historic shift in U.S.-China science and technology policy.

Chinese Foreign Ministry spokesman Mao Ning said Monday, "The U.S. side needs to stop politicizing and weaponizing trade and science and technology issues and stop destabilizing the global industrial chain supply chain." We will closely monitor developments and resolutely safeguard our rights and interests."

Last year, government restrictions prevented Nvidia, the world's most valuable chipmaker, from shipping two of its most advanced artificial intelligence chips, which have become the industry standard for developing chatbots and other AI systems, to customers in China.

But NVIDIA quickly released new versions for the Chinese market that are less complex and bypass U.S. export controls. One of them, called the H800, has the same computing power as the company's more powerful but hampered H100 chip in certain settings used in AI work. However, some key performance aspects are still limited, according to the spec sheet seen by Reuters.

New Restrictions

The U.S. now plans to introduce new AI chip guidelines that would restrict certain advanced data center AI chips that are not currently captured, the U.S. official said.

While the official declined to say which additional chips would be effectively banned, Nvidia's H800 is the one that semiconductor sources have hinted the government wants to block.

A spokesman for Santa Clara, California-based Nvidia declined to comment, and in June, the company's chief financial officer said that if the H800 and a related chip called the A800 were restricted, they did not expect "an immediate material impact on our financial results.

The official said chips for consumer products such as laptops would not be subject to the new restrictions. But the official said companies would need to inform the Commerce Department when filling orders for the most powerful consumer chips to ensure they are not used in ways that threaten national security.

In an effort to block Chinese AI chips that the US considers too powerful, the US plans to stop considering a "bandwidth parameter" limit on one of the parameters it has been using to determine which types of AI data center chips are available, the official said. The bandwidth or interconnect parameter is a measure of how fast chips can communicate with each other.

At current thresholds, chips must exceed both performance speed and bandwidth thresholds to be limited. By removing the bandwidth parameter, chips that individually exceed a certain performance speed-in other words, how powerful each chip is-will soon be limited.

It would also be limited if it exceeded another measure, a certain "performance density," to help prevent future workarounds, the official said, but declined to elaborate further.

Evolving Technology

The updated rules also aim to cover artificial intelligence chips as the technology evolves. The U.S. will require companies to notify the government before shipping semiconductors that perform slightly below guidelines to China, the official said. The government will decide on a case-by-case basis whether they pose a national security risk, but they can be shipped unless the chipmaker notifies otherwise.

The October 2022 update to the rules could also close a loophole for Chinese companies to obtain U.S. AI chips through Chinese subsidiaries located overseas, Reuters reported last week.

The rules are not expected to restrict access to cloud computing services of the U.S. or its allies, but the U.S. will seek comment on the risks of such access and how to address them, the official said.

The Biden administration told Beijing it planned to update the controversial rules this month as part of a policy aimed at stabilizing relations between the superpowers, Reuters reported in early October.