U.S. chip export easing signals shift in strategy

U.S. President Donald Trump said on Dec. 8 that he would allow the export of Nvidia’s H200 AI chip to China. He added that he had informed Chinese President Xi Jinping and received a positive response. The United States has described the decision as a step toward improving South Korea-U.S.-China relations. Many analysts, however, view it as a reluctant move motivated by concerns that continued export controls on U.S.-made high-performance AI chips were hurting American companies and unintentionally accelerating China’s AI chip ambitions.

The H200 chip, now approved for export to China, ranks just below Nvidia’s latest Blackwell-based graphics processing units and delivers significantly higher performance than the previously exportable low-end H20 model. Since 2022, the United States has restricted the export of cutting-edge AI chips to China, but the latest decision marks a major relaxation of those rules. The Blackwell chips and the upcoming next-generation Rubin chips remain excluded from export approval.

U.S. restrictions on exporting AI chips to China were first imposed under the Biden administration to maintain a technological edge over China in a sector considered critical to future national competitiveness. The landscape has shifted dramatically, however, after China accelerated the domestic development of AI chips and succeeded in mass-producing them on a large scale. The Chinese government has ordered new data centers to adopt AI chips from domestic firms such as Cambricon and has also offered electricity bill discounts to companies that comply.

Urgency behind the decision stemmed from the fact that China once accounted for one quarter of Nvidia’s total sales. Nvidia CEO Jensen Huang persuaded U.S. regulators that banning American chips would only accelerate China’s domestic AI chip development and ultimately undermine U.S. national interests. That argument contributed to the rollback of restrictions. However, with China now pivoting to domestically produced AI chips, many analysts believe it will no longer purchase U.S.-made chips at previous levels.

Some analysts say the shift could benefit South Korean semiconductor makers that supply high-bandwidth memory to Nvidia. More significant, however, is the rapid rise of China’s AI capabilities, which have advanced to a point that forced Washington to loosen its regulations. China is also accelerating development of high-bandwidth memory and other advanced memory semiconductors, an area where South Korea currently holds a competitive edge. This has raised concerns over whether South Korea can maintain that lead while its semiconductor research workforce continues to struggle with issues such as strict compliance with the 52-hour workweek limit.