More on News
In an effort to further tighten its export controls on advanced semiconductor technology, the United States will take steps to prevent American chipmakers from selling semiconductors to China that attempt to circumvent government-imposed restrictions. This move comes as part of the Biden administration’s ongoing efforts to block additional exports of AI chips to China.
The new regulations, revealed here for the first time, will be introduced as an extension of the comprehensive U.S. restrictions placed on shipments of advanced chips and chipmaking equipment to China, initially unveiled in October of the previous year. While the exact timing of the updates is anticipated this week, it’s important to note that such timelines are often subject to change.
The impending rules will block specific AI chips below the existing technical parameters. Furthermore, as an anonymous U.S. official outlined, these regulations will demand that companies report shipments of other AI chips.
This latest crackdown on tech exports to China comes amid efforts by the United States to improve its relationship with the world’s second-largest economy. Several senior members of the Biden administration have engaged in dialogues with their Chinese counterparts in recent months. However, the introduction of these new export rules has the potential to complicate diplomatic efforts.
The Biden administration’s stance has been to implement export restrictions to prevent U.S. chips and equipment from strengthening China’s military capabilities. In response, Beijing has accused the United States of abusing export controls to suppress Chinese companies. These restrictions have marked a significant shift in U.S.-China tech policy.
The Chinese embassy in Washington did not respond immediately to requests for comment.
Last year, existing government restrictions prohibited Nvidia, the world’s most valuable chipmaker, from shipping two of its most advanced AI chips to Chinese customers. These chips are widely recognized as industry standards for developing AI systems and chatbots.
However, Nvidia soon introduced new variants explicitly designed for the Chinese market. These variants are less advanced but circumvent U.S. export controls. One such chip, the “H800,” offers computing power comparable to Nvidia’s more powerful, blocked “H100” chip in certain AI work settings.
The U.S. intends to introduce fresh guidelines for AI chips, restricting specific advanced data center AI chips not currently encompassed by the existing controls. The restrictions will exempt chips meant for consumer products such as laptops. Nevertheless, companies must report to the Commerce Department when fulfilling orders for the most potent consumer chips, ensuring they are not used in ways that may threaten national security.
To deter AI chips deemed excessively powerful from reaching China, the U.S. will eliminate the “bandwidth parameter,” a factor employed to restrict exports of certain AI data center chips. This removal will trigger the activation of another guideline, broadening the scope of covered chips. Consequently, the communication speed between AI chips is expected to decrease, potentially complicating and increasing the cost of AI development. Training large AI models necessitates the use of multiple interconnected chips.
The United States also intends to introduce a “performance density” parameter to mitigate potential future workarounds, as confirmed by the official. However, further details regarding this parameter were not disclosed at this time.