White House Offers NVIDIA and AMD a Pay-To-Play Scheme for Chinese Market Access

The U.S. government is going into business with America’s most potent adversary, China. On August 10, The New York Times reported that chipmakers NVIDIA and AMD will receive export licenses to sell high-end artificial intelligence (AI) chips in China in exchange for delivering 15 percent of the sales revenue to the U.S. federal government.

This deal undermines U.S. national security policy in favor of commercial interests. It further compromises already weakened U.S. export controls on advanced AI chips, which China uses to fuel its unprecedented military build-up. It also provides Chinese firms with crucial components needed to improve their AI models at the expense of American competitors.

NVIDIA and AMD Promise Share of Chip Sales in China

The reported agreement came after President Donald Trump announced in July that NVIDIA and AMD would be able to apply for export licenses to resume sales to China — both had previously been restricted by the Commerce Department due to its concerns over fueling Beijing’s military modernization. The deal is expected to generate billions of dollars in government revenue, though the legality of the unprecedented arrangement — given that the U.S. Constitution prevents the imposition of export taxes — remains unclear.

The agreement is expected to cover only NVIDIA’s H20 and AMD’s MI308 chips, both of which remain in high demand in the Chinese market. The H20 chip is valued for its substantial high-memory bandwidth, which allows AI models to process user queries faster and with a higher degree of accuracy. The MI308, which was specifically designed for the Chinese market, is also valued for its high-end AI performance.

China Relies on American Technology to Spur AI Development

China remains highly reliant on high-end American AI chips in developing its domestic AI industry. Despite deploying significant subsidies to leading foundries such as Huawei, Chinese AI firms continue to prefer American AI chips due to their superior performance in both training and inference tasks. This demand has further increased as Chinese models have transitioned from perfecting their capabilities to broader deployment — an issue noted by DeepSeek founder Liang Wenfeng in discussing his firm’s struggles to innovate due to U.S. export controls.

Aside from Chinese firms’ commercial reliance on foreign chips, the People’s Liberation Army (PLA) remains a major client of American AI firms. Tasked with integrating AI into its new warfighting concept of “intelligentized warfare,” the PLA has continued to purchase NVIDIA, Intel, and AMD chips to further its advances in autonomous weapons, surveillance platforms, and command-and-control systems.

Washington Should Reconsider Relaxing Export Controls

The decision to allow NVIDIA and AMD to resume sales within the Chinese market, while simultaneously signaling that export controls may be waived in exchange for business concessions, will irrevocably assist China’s burgeoning AI industry while harming U.S. national security.

Rather than relaxing export controls, the United States should tighten enforcement on existing restrictions while expanding bans on the sale of the software that is used to design chips and related manufacturing equipment. These efforts should be coupled with congressional action on screening outbound investment into China’s technology sector and support for the development of a strong domestic AI sector, including investments in energy generation, data center construction, and basic science and research.

Jack Burnham is a research analyst in the China Program at the Foundation for Defense of Democracies (FDD). For more analysis from the author and FDD, please subscribe HERE. Follow Jack on X @JackBurnham802. Follow FDD on X @FDD. FDD is a Washington, DC-based, nonpartisan research institute focusing on national security and foreign policy.