Nvidia and AMD have agreed to allocate 15 percent of their chip sales revenue from China to the U.S. government, as disclosed by a U.S. official on Sunday. This move, considered unconventional, may unsettle American companies. Earlier in April, the Trump administration had suspended the sale of advanced computer chips to China citing national security concerns. However, in July, Nvidia and AMD announced that Washington would permit them to recommence sales of the H20 and MI308 chips essential for artificial intelligence development.
The unnamed official, who shared the information with The Associated Press, confirmed the revenue-sharing terms of the agreement, aligning with the details initially reported by The Financial Times. Both Nvidia and AMD apparently accepted this financial arrangement as a prerequisite for securing export licenses to resume sales to China. AMD stated that the U.S. government had sanctioned its license applications for exporting chips to China.
Although Nvidia refrained from commenting on the specifics of the deal or its reciprocal nature, the company emphasized its commitment to complying with the administration’s export regulations. In a statement, Nvidia emphasized the importance of adhering to U.S. government rules for global market participation. The company expressed optimism regarding export control rules enabling American competitiveness both in China and internationally, highlighting the significance of maintaining leadership in telecommunications and AI technology.
Amidst these developments, the top Democrat on a House panel focusing on competition with China raised apprehensions regarding the reported agreement, labeling it as a “dangerous misuse of export controls undermining national security.” Rep. Raja Krishnamoorthi, the ranking member of the House Select Committee on China, pledged to investigate the legal foundation of the arrangement and demanded transparency from the administration. He stressed the necessity of ensuring that export control policies prioritize genuine security concerns rather than resembling taxation strategies disguised as national security measures.
In a separate statement, Nvidia argued that stringent export controls surrounding chip sales could incur an additional $5.5 billion in costs for the company. The company contended that such restrictions impede U.S. competitiveness in a sector crucial for one of the world’s largest technology markets. Nvidia also cautioned that U.S. export controls might inadvertently drive other countries towards China’s AI technology.
This agreement to share sales revenue with the U.S. government from transactions in China stands out as an unusual move by a president and reflects Trump’s recent involvement in corporate decision-making. Trump has been urging company executives to invest in the United States to bolster domestic employment and manufacturing. Notably, Trump recently called for the immediate resignation of new Intel CEO Lip-Bu Tan, citing perceived conflicts of interest due to his connections with Chinese entities.
Geoff Gertz, a senior fellow at the Center for New American Security in Washington, D.C., remarked on the complex situation, questioning the rationale behind imposing an additional penalty on the sale of H20 chips to China. He highlighted the inconsistency in either deeming the transaction a national security risk or imposing a punitive measure despite the absence of such risks.
Commerce Secretary Howard Lutnick disclosed in July that the resumption of Nvidia’s chip sales in China was tied to a trade agreement between the U.S. and China concerning rare earth magnets. The restrictions on advanced chip sales to China have been pivotal in the AI competition between the world’s leading economic powers. While proponents argue that such controls are essential to maintain U.S. technological leadership, critics suggest that the restrictions could inadvertently foster innovation. The emergence of China’s DeepSeek AI chatbot earlier this year reignited concerns over China’s utilization of advanced chips to advance its AI capabilities.
