Nvidia and AMD to Give U.S. 15% of China Chip Revenue
The U.S. will let Nvidia and AMD sell certain AI chips to China in exchange for 15% of the revenue. The deal turns export licenses into a source of government revenue and raises legal and national security questions.
The Trump administration will allow Nvidia and AMD to resume certain sales of artificial intelligence chips to China, but it will take 15% of the revenue generated. The deal covers Nvidia’s H20 and AMD’s MI308, two processors adapted for the Chinese market after U.S. restrictions on advanced technology.
The move is neither a blanket tariff nor a fine for noncompliance. It is a condition attached to export licenses: the companies receive permission to sell, and in return hand over part of their sales revenue to the federal government. It is an unusual formula that blends trade policy, technology controls and revenue collection.
From April’s ban to a conditional sales permit
Washington halted exports of the H20 and other chips to China in April over national security concerns. Nvidia had designed the H20 specifically to comply with earlier U.S. restrictions: it offers significant capabilities for training and running AI systems, but falls short of the more powerful accelerators the company sells in other markets.
Even so, the Commerce Department required a license for that product as well. Nvidia then said it expected to take charges of up to $5.5 billion related to inventory, purchase commitments and reserves tied to the H20.
In July, the administration reversed course and said it would allow sales to resume. Commerce Secretary Howard Lutnick linked the change to trade negotiations with Beijing over rare earths, minerals essential to manufacturing everything from electric motors to defense systems and electronic components.
The new requirement puts a direct price on the permit. Donald Trump confirmed Monday that he initially wanted 20% of Nvidia’s revenue and that the figure ended up at 15% after negotiating with the company’s CEO, Jensen Huang. Trump described the H20 as essentially an old chip.
Nvidia told the Associated Press that it would comply with the rules set by the U.S. government. The company also argued that maintaining a commercial presence in China is important to ensure that U.S. technology and the AI ecosystem do not lose ground to local alternatives.
A delicate precedent for export controls
Export controls are generally used to prevent sensitive technologies from being put to military or strategic uses by geopolitical rivals. In this case, the license does more than determine what can be sold: it gives the government a share of the revenue from a private transaction.
The arrangement has drawn criticism in Congress from both Republicans and Democrats. John Moolenaar, the Republican chair of the House Select Committee on China, has questioned the deal’s legal basis and the precedent it could set. Democrat Raja Krishnamoorthi, the committee’s ranking member, has called for transparency around the agreement.
The central legal objection is serious. The U.S. Constitution prohibits taxing exports from the states, and several experts believe that calling the payment a “fee” does not necessarily resolve the issue if, in practice, it amounts to a tax on overseas sales. The administration would have to justify it as a legitimate administrative condition of a license, rather than a disguised tax.
China remains too important to ignore
For Nvidia and AMD, China is one of the world’s largest markets for AI computing. Giving it up means losing revenue, but it also leaves room for Chinese manufacturers seeking to replace U.S. chips. For Washington, allowing these sales means accepting that completely restricting China’s access could accelerate that technological substitution.
The H20 case captures that tension. The United States wants to preserve its lead in advanced chips, but it also wants U.S. companies to remain leading suppliers in the largest rival technology market. The 15% deal does not resolve that dilemma; it monetizes it.
The practical workings of the agreement still need to be clarified, including how the revenue subject to the payment will be calculated and where the money will go in the federal budget. It will also be important to see whether the formula is limited to Nvidia and AMD or extends to other export licenses. If the latter happens, selling strategic technology from the United States could begin to depend not only on a product’s technical limits, but also on the share of revenue the government is willing to demand.
