Why Is the US Government Taking a Cut of Nvidia's China Sales?
Sign up for ARPU: Stay ahead of the curve on tech business trends.
The US-China chip war has entered a strange and unprecedented new phase. In a highly unusual arrangement, the Trump administration has brokered a deal where American tech giants Nvidia and AMD will pay the US government 15% of their revenue from all advanced AI chip sales to China.
The move, which was a condition for lifting a costly export ban, effectively turns the US government into a direct financial stakeholder in the very international sales it had previously deemed a national security risk. This bizarre policy blurs the lines between regulation, taxation, and state-led commercial enterprise, creating what amounts to a new "national security tollbooth" on the road to the world's second-largest AI market.
What exactly is this new arrangement?
According to US officials, Nvidia and AMD have agreed to give the federal government 15% of the revenue they generate from selling specific high-end AI chips, like Nvidia's H20, to Chinese customers. This is not a tariff paid by importers. It is a direct payment from an American company's top-line revenue to its own government, required as a condition for receiving the export licenses needed to do business in China.
As one analyst quoted by Reuters noted, the arrangement is "unprecedented," and it has left policy experts questioning the logic. "Either selling H20 chips to China is a national security risk, in which case we shouldn't be doing it to begin with, or it's not a national security risk, in which case, why are we putting this extra penalty on the sale?" asked Geoff Gertz, a senior fellow at the Center for New American Security.
What is the government's strategic goal?
While the move seems contradictory, it aligns with a new strategic doctrine articulated by members of the Trump administration. The goal is no longer to completely block China from accessing US technology, but to ensure that China's AI ecosystem remains dependent on the "American tech stack."
As Commerce Secretary Howard Lutnick stated last month, it is in US interests for Chinese companies to use American technology, even if it's a downgraded version. This keeps Chinese developers locked into platforms like Nvidia's CUDA software, preventing them from fully migrating to domestic alternatives like Huawei's Ascend ecosystem. The 15% levy, in this view, is the price the US is extracting for allowing this controlled engagement. It's a way for the administration to look tough and generate revenue while pursuing a more nuanced, long-term competitive strategy.
Why would Nvidia and AMD agree to this?
For the chipmakers, the calculation is a simple, if painful, one. China is a massive and critical market. In the last fiscal year, Nvidia generated $17 billion in revenue from China (13% of its total), while AMD reported $6.2 billion (24% of its total). The export ban imposed in April was costing them billions.
Agreeing to give up 15% of future revenue is a significant hit to their gross margins, but it allows them to re-enter a market they cannot afford to lose. As one investment director put it to Reuters, 85% of a very large number is still better than 100% of zero. It is a pragmatic business decision made under extreme political duress.
What does this mean for the future of tech policy?
This move sets a startling new precedent. It signals a shift toward a more transactional and interventionist form of industrial policy, where the government acts not just as a regulator, but as a commercial partner that demands a cut of the profits.
For US tech companies, it introduces a new and unpredictable form of sovereign risk. The rules for engaging with global markets are no longer just about navigating export controls; they now may involve negotiating a direct revenue-sharing agreement with the White House. The government is no longer just setting the rules of the road; it's standing at a tollbooth, demanding payment to pass.