The United States plans to take measures to prevent American chip manufacturers from selling semiconductors to China that bypass government regulations. These steps are part of the Biden administration’s forthcoming efforts to tighten restrictions on AI chip exports.
These new regulations will build on the broader U.S. restrictions on advanced chip and chipmaking equipment shipments to China announced in October of the previous year. Although there is an expectation that these updates will be unveiled this week, it’s worth noting that such schedules can be subject to change.
The new rules will bar certain AI chips that just fall short of existing technical parameters while requiring companies to report shipments of others, as disclosed by an anonymous official. The U.S. Department of Commerce, responsible for export controls, has refrained from commenting.
This latest crackdown on tech exports to China occurs as the U.S. is working to improve relations with the world’s largest economies. While senior officials from both sides have engaged in discussions over the past few months, these new regulations may complicate these diplomatic efforts.
The Biden administration has asserted that these export restrictions were designed to prevent U.S. chips and equipment from enhancing China’s military capabilities. In response, Beijing has accused the U.S. of exploiting export controls to suppress Chinese companies, marking a significant shift in U.S.-China tech policy.
At the time, U.S. government restrictions prevented Nvidia, one of the world’s leading chip manufacturers, from shipping two of its advanced AI chips to Chinese customers. These chips have become industry standards for developing AI systems such as chatbots. To circumvent these export controls, Nvidia introduced new versions for the Chinese market with reduced sophistication. For instance, the H800 chip provides similar computing power to the blocked H100 but with some performance limitations.
The U.S. intends to introduce new guidelines for AI chips that restrict certain advanced datacenter AI chips not currently covered, according to the official. While the specific chips to be banned were not disclosed, there have been suggestions that Nvidia’s H800 was among them. Nvidia, headquartered in Santa Clara, California, has not yet commented on these reports. The company’s CFO previously stated that restricting the H800 and a related chip, the A800, would not have an immediate material impact on their financial results.
The new regulations will exclude chips intended for consumer products such as laptops, but companies will be required to notify the Commerce Department when fulfilling orders for the most powerful consumer chips to ensure they aren’t used in ways that threaten national security.
To prevent certain AI chips from reaching China, the U.S. will eliminate the “bandwidth parameter,” expanding the scope of restricted chips by triggering another guideline. This could reduce the speed at which AI chips communicate with each other, making AI development more challenging and costly.
The updated rules are designed to accommodate evolving technology. Companies will need to inform the government about semiconductors that nearly meet the guidelines but fall slightly below them. The government will determine, on a case-by-case basis, whether they pose a national security risk and will allow or prohibit the shipments accordingly.
These updates may also close a loophole that allows Chinese companies to access American AI chips through overseas Chinese units. The rules are not expected to include restrictions on access to U.S. cloud computing services or those of allies. However, the U.S. plans to seek input on the potential risks and their potential remedies regarding such access. The Biden administration informed Beijing of its plans to update these regulations earlier this month as part of an effort to stabilize U.S.-China relations.