US forces NVIDIA, AMD to share 15% of China AI chip revenue

In a move that redefines the intersection of national security and international commerce, American semiconductor giants NVIDIA and AMD have reportedly agreed to a landmark arrangement with the United States government. The deal stipulates that the companies will remit 15% of their revenue from the sale of specific advanced artificial intelligence chips to China back to the U.S. This unprecedented condition was attached to the issuance of export licenses, allowing the firms to resume supplying a critical, albeit restricted, market. The agreement marks a significant escalation in the ongoing technology trade war between Washington and Beijing, introducing a novel mechanism that directly ties a company's foreign sales revenue to government coffers under the justification of national security.

The arrangement centers on specially designed processors created to comply with existing U.S. export controls.[1] NVIDIA's H20 and AMD's MI308 chips were developed as less powerful alternatives to their most advanced models, aiming to service Chinese customers without violating the stringent restrictions imposed by the U.S. to curb China's military and technological advancement.[2][3] These controls, first implemented in October 2022 and subsequently tightened, seek to limit China's access to high-end chips that power artificial intelligence and supercomputing, which Washington views as a direct threat.[4][5] The new revenue-sharing model emerged after the Trump administration, which had previously halted sales of even these modified chips in April, reversed its decision and began granting licenses.[6][7] The policy shift reportedly followed a meeting between NVIDIA CEO Jensen Huang and President Donald Trump.[3][8] While NVIDIA has not explicitly confirmed the 15% figure, the company stated it adheres to the rules set by the U.S. government for its participation in global markets.[9][7] AMD has not issued a public comment on the matter.[9]

This revenue-sharing agreement introduces a complex and controversial dynamic into U.S. trade policy. For decades, export controls have functioned as a clear-cut prohibition—a license to sell sensitive technology abroad was either granted or denied based on national security assessments.[1] This new policy, however, blurs that line by creating a system where market access is contingent upon a direct financial payment to the government. Analysts and legal experts have raised serious questions about the policy's constitutionality, with some arguing it effectively functions as an export tax, which is explicitly forbidden by the U.S. Constitution.[2][6][1] The arrangement is seen as a way for the U.S. government to mitigate perceived national security risks while also potentially generating significant funds, though the administration has not yet specified how this revenue would be used.[9][10] This quid pro quo has been criticized for potentially trading national security protections for financial gain, setting a worrisome precedent for future trade and technology controls.[11]

The implications for the global AI and semiconductor industries are profound. For NVIDIA and AMD, the deal offers a partial reprieve, allowing them to maintain a foothold in the lucrative Chinese market, which accounts for a significant portion of their revenue.[12][10] Before the recent halt, NVIDIA was projected to sell a substantial volume of H20 chips to major Chinese tech companies.[13] However, the 15% levy will undoubtedly impact their profit margins on these sales.[7] More broadly, the move could reshape competitive dynamics. While U.S. firms navigate these new financial hurdles, Chinese rivals are aggressively pursuing self-sufficiency in semiconductor technology, a national priority accelerated by U.S. restrictions.[5][14] NVIDIA's CEO has previously warned that being locked out of the Chinese market could endanger America's leadership in AI technology, as it would create a vacuum for domestic Chinese firms to fill.[8] The current arrangement, while allowing for continued sales, imposes a significant cost that could influence long-term strategic decisions for all players in the industry.

In conclusion, the decision to compel NVIDIA and AMD to share 15% of their China sales revenue with the U.S. government represents a significant and untested strategy in America's efforts to manage its technological rivalry with China. It moves beyond simple export bans to a more transactional form of control, raising legal, ethical, and strategic questions. While providing a conditional pathway for U.S. companies to continue operating in China, the policy also introduces financial burdens and sets a controversial precedent. The long-term consequences of this policy—on the profitability of American tech giants, the pace of China's indigenous chip development, and the very nature of U.S. export control policy—will be closely watched by the entire technology world. The arrangement signals a new, more financially entangled phase in the high-stakes competition for global technological supremacy.