Nvidia, AMD to Cede 15% of China AI Chip Revenue to US Government
In an unprecedented move reshaping the landscape of global technology trade, American chip giants Nvidia and Advanced Micro Devices (AMD) have reportedly agreed to remit 15 percent of their revenue from artificial intelligence (AI) chip sales in China to the U.S. government. This highly unusual arrangement serves as a prerequisite for obtaining the crucial export licenses that permit them to continue selling their specialized AI semiconductors in the lucrative Chinese market.
The agreement, first reported by the Financial Times and subsequently confirmed by other major news outlets, follows a meeting last week between Nvidia CEO Jensen Huang and U.S. President Donald Trump at the White House. Under the terms, Nvidia will reportedly pay 15 percent of revenue from its H20 AI accelerator chip sales, while AMD will do the same for its MI308 chips. These chips are specifically designed to comply with existing U.S. export restrictions, which aim to curb China’s access to cutting-edge AI technology on national security grounds.
This revenue-sharing model marks a significant departure from traditional export controls, which typically involve outright bans or strict limitations on technology transfers. Experts and former U.S. officials have swiftly voiced concerns, with some, like Peter Harrell, the former White House senior director for international economics under the Biden administration, suggesting the deal may be unconstitutional, equating it to an export tax explicitly forbidden by the U.S. Constitution. Christopher Padilla, a former top export control official in the George W. Bush administration, described it as “unprecedented and dangerous,” warning it could undermine the U.S.'s credibility with allies when seeking to implement complementary controls on sensitive technologies.
The financial implications of this deal are substantial. Analysts estimate that Nvidia’s H20 chip sales in China alone could generate around $15 billion by the end of 2025, while AMD’s MI308 sales could add another $800 million. This could translate to over $2 billion in revenue for the U.S. government, though the Trump administration has yet to determine how these funds will be utilized. For the chipmakers, while the agreement reopens access to a vital market, it introduces a “political tariff” that will inevitably pressure profit margins. Nvidia has previously stated it was unable to ship $2.5 billion in H20 revenue to China due to prior restrictions.
The backdrop to this agreement is a highly volatile period in U.S.-China tech relations. The Trump administration had previously halted H20 chip sales to China in April, citing security risks, only to reverse that decision in July ahead of further trade talks with Beijing. This policy fluctuation has already drawn criticism from officials concerned about eroding international trust in U.S. export controls.
Meanwhile, China is actively pushing for a relaxation of U.S. export controls on high-bandwidth memory (HBM) chips, which are crucial for advanced AI applications, ahead of a potential summit between President Trump and Chinese President Xi Jinping. Beijing is reportedly concerned that these controls are hindering domestic companies like Huawei from developing their own competitive AI chips. Despite the restrictions, China remains a critical revenue source for American chip manufacturers, even as Chinese alternatives, such as Huawei’s Ascend chips, are beginning to capture a significant portion of domestic demand.
Nvidia, for its part, has affirmed its commitment to adhering to U.S. government rules for global market participation, expressing a desire for export control policies that enable America to compete worldwide and maintain its leadership in AI technology. This complex arrangement underscores a new, financially intertwined approach to national security and economic policy, with far-reaching implications for the global semiconductor industry and the ongoing tech rivalry between the U.S. and China.