As the Trump administration moved to resume certain AI chip sales to China, it did so under a novel arrangement: the U.S. government would capture 15% of the revenue from shipments of Nvidia and AMD chips. The policy, confirmed by President Trump and implemented through export licenses issued in August 2025, permits Nvidia and AMD to resume sales that had previously been halted on national-security grounds. The deal is being pitched as a pragmatic settlement—one that preserves leadership levers over Beijing’s AI ambitions while restoring a critical stream of commerce for U.S. chipmakers. Yet the 15% levy also reframes export controls as a revenue mechanism, raising questions about incentives, accountability, and the policy’s long-term purpose.

Under the arrangement, export licenses are granted with a condition that a direct 15% royalty flows to the U.S. Treasury on each qualifying sale to China. The policy aims to strike a balance between national-security restrictions and the realities of a global AI supply chain, where Nvidia and AMD rely on international markets for scale. Officials estimate the program could generate more than $2 billion for the U.S. government, money the administration frames as offsetting the cost of maintaining export controls and cybersecurity capabilities. Industry analysts have suggested that volumes to China could be substantial if licenses remain stable and demand holds, with Nvidia publicly linked to potential China shipments that could exceed $15 billion in value under the license regime.

The broader context is longstanding: in the years leading up to 2025, Washington restricted sales of advanced AI accelerators — especially those with the performance profiles of Nvidia’s Blackwell line — to China and a handful of other countries on national-security grounds. The policy shift that enables revenue collection from those sales represents an unusual fusion of diplomacy, trade policy, and fiscal strategy. The arrangement is presented by administration officials as a means to preserve strategic leverage while sustaining a crucial revenue-generating channel for the government, a move that would have been unthinkable in earlier decades when export controls operated largely as regulatory constraints rather than revenue devices.

From a policy and security standpoint, the settlement is controversial. Supporters argue that it keeps essential control over sensitive technology while providing a tangible fund to strengthen enforcement, intelligence, and cyber-defense capabilities. Critics warn that turning export licenses into revenue could dilute the intended purpose of the controls, inviting calls for broader licensing reform or sunset clauses that would reassert purely strategic rather than fiscal objectives. In comments cited by financial and policy outlets, a former National Security Council China director warned that Beijing could interpret the model as a tax on technology, potentially pressuring Chinese firms to accelerate alternative supply chains or domestic innovation to bypass U.S. licensing regimes. A former security official told the Financial Times that converting export licenses into revenue streams changes their underlying purpose and risks eroding export-control credibility.

The economic implications extend beyond the Treasury tally. Nvidia and AMD, already navigating volatile stock markets and shifting demand for AI infrastructure, now face a regulatory framework that ties licensing to government revenue. For Nvidia, the prospect of $15 billion in China-bound chip sales exists within the licensing envelope, but actual flows will hinge on license terms, enforcement consistency, and China’s own enforcement of foreign-made technology rules. AMD, too, confronts a new benchmark: how its product mix, pricing, and customer relationships adapt when a portion of proceeds is siphoned off by the U.S. government. Market observers caution that the policy could influence investment decisions, push for more diversified regional exposure, or accelerate alternative regional pipelines as multinationals hedge against policy uncertainty.

Geopolitically, the arrangement sits at the center of a broader U.S.–China tech rivalry. On one hand, it signals Washington’s willingness to monetize strategic controls even as it seeks to preserve access to crucial AI infrastructure that feeds domestic innovation and defense programs. On the other hand, it risks provoking a hardening in China’s stance toward U.S. suppliers, accelerating efforts to localize chip design and manufacturing or to accelerate non-Chinese suppliers’ practice areas that serve as substitutes. In parallel, a federal appeals court recently declined to block the administration’s export-control actions, underscoring ongoing legal battles and the possibility of new challenges that could shape the policy’s durability. The legal dynamics, coupled with congressional scrutiny and industry lobbying, will determine whether the 15% levy endures, evolves, or is scaled back.

Looking ahead, several strategic questions loom. Will Congress press for formal legislative oversight of the revenue-levy mechanism, or will policymakers embrace a more targeted licensing framework with explicit sunset provisions? How will license terms address sensitive subcomponents, such as neural-network accelerators with novel memory architectures, and what governance safeguards will prevent leakage or circumvention of controls? For Nvidia and AMD, the question is whether the revenue-sharing model improves resilience against policy shocks or merely enlarges the set of risks that accompany cross-border AI trade. For U.S. policymakers, the challenge will be balancing the fiscal incentive with the imperative to maintain credible export controls that others perceive as predictable and principled. And for the global AI ecosystem, the policy introduces a new governance variable that could influence the speed, direction, and geographic footprint of AI innovation for years to come.

In the near term, investors and industry watchers will monitor license issuance patterns, enforcement metrics, and any changes in Chinese procurement behavior as a barometer of how far the arrangement will alter incentives. The policy’s success will depend not only on the Treasury yield it generates but also on whether it preserves a coherent strategic objective: deterring sensitive technology from advancing in unintended directions while sustaining a robust, globally integrated AI supply chain that continues to fund innovation and growth. If the approach endures, it could set a precedent for how future high-tech exports are governed in a hyper-competitive, geopolitically tense environment. If it falters, lawmakers and industry alike will demand clearer boundaries and stronger assurances that national-security aims remain the guiding priority over fiscal expediency.