| FY 2023* | 37,943 |
| FY 2022 | 40.245 |
| FY 2021 | 40.567 |
“Chip giants Nvidia and AMD have agreed to pay the US government 15% of Chinese revenues as part of an ‘unprecedented’ deal to secure export licences to China, the BBC has been told. The US had previously banned the sale of powerful chips used in areas like artificial intelligence (AI) to China under export controls usually related to national security.”
* Using a journalistic source, as the administration hasn’t made an official statement as of our publication time.
This policy lurch is the most recent in a three-year back-and-forth, which began with an October 2022 ban on sales to Chinese customers of exports of high-end semiconductors meant for artificial intelligence programming. Nvidia, the California-based graphics processing unit and AI chip designer, followed up by designing a version of its “H100” and “H200” chips (designated “H20”) meant to be useful only for commercial markets. Then-Commerce Secretary Gina Raimondo approved the idea in December 2023. The Trump administration, having re-blocked the H20 chip in mid-April, has now apparently changed its mind, allowing Nvidia (and AMD as well) to proceed if they give the U.S. government 15% of the money they earn from these sales.
Outside the trade-and-security world, this sort of direct and apparently long-term government involvement in particular companies usually means trouble. (See below for some thoughts on the implications for taxation and market economics.) Taken strictly as export control policy, it’s worse. Decisions like “should advanced tech companies sell a particular type of computer chip to China?” are complex judgment calls, but their foundation ought to be simple: the best national security analysis available. Adulterating that with revenue concerns is a bad mistake. In specific cases it poses both the risks of ill-advised high-tech sales to potential adversaries, and the risk of lost exports of safe products. More generally, it opens an essential policy area to systemic danger of corruption.
To pull back: “Export control” policy attempts to ensure that military-related technologies — not only actual weapons but software, specialized ceramics and alloys, advanced chips and computers, biotechnology, etc. — developed in America and allied countries don’t go to adversaries, or spill out onto world markets from which they can then flow to unfriendly places. Using a base in American law and four international “regimes” meant to coordinate policies to the extent possible with allies and major powers (also see below), government experts centered in the Commerce Department’s Bureau of Industry and Security (“BIS”) try to categorize, track and when necessary ban exports of 538 classes of physical goods and software in 9 industry groups. (Nuclear technology and firearms; special materials, chemicals, toxins, and microorganisms; materials processing; electronics; computers; telecommunications and ‘information security’; sensors and lasers; navigation and avionics; marine; aerospace and propulsion.) BIS’s 600 staffers are busy; their most recent Annual Report records decisions on 37,943 license applications — about 100 a day — covering $26.7 billion in total exports, or about 1.5% of all U.S. goods sales abroad. To make these calls they need to:
Without passing judgment on the technical questions in this one — did the 2022 freeze slow Chinese tech development, or, contrariwise, accelerate Huawei’s own chip-design program? if the H20 ban were to stay on, would other European or Asian suppliers simply replace U.S. firms? how would lost export revenue affect U.S. firms’ research budgets and next-generation products? — it’s enough to say that U.S. officials need to base their decision on impartial analysis and objective national security criteria.
Adding a government revenue interest to this mix risks warping not only this particular decision, but future export control policy in general. When favored transactions will bring in money, after all, government will have an incentive to allow transactions that might not be harmless. Contrariwise, if it can collect money from one company, it will have an incentive to ask others for similar fees. That can mean a large incentive for corruption of government and business alike, with both sides aware that flows of money could ease approval of transactions that pose risk, and that government could withhold approval for useful and low-risk transactions when companies choose not to pay.
In sum: Taking money in exchange for approving export licenses is poor semiconductor policy, risky for national security, and bad precedent for future export control policy. Congress should reverse this ill-advised and dangerous call as soon as it returns to work in September.
PPI’s four principles for response to tariffs and economic isolationism:
A BBC look at the 15% Nvidia/AMD arrangement. (Using a journalistic source since, as of this writing, we’ve seen no formal statements from the administration or the companies.)
U.S. background:
Former Undersecretary of Commerce Alan Estevez on export control policy as of 2024.
BIS explains the Export Administration Regulations.
… publishes the list of controlled technologies.
… and documents its work (though only up to FY2023) in Annual Reports.
International background:
The Nuclear Suppliers Group: 1974, covers nuclear-power technology, uranium, heavy water, and transport.
The Australia Group: 1985, on chemical and biological weapons and related technologies.
The Missile Technology Control Regime: 1987, for ballistic missiles and associated technologies such as avionics, sophisticated ceramics and metals, rocketry, etc.
The Waassenar Arrangement: 1994, covering conventional weapons and technologies of the “powerful chip” sort.
And two other things:
Through an export control policy, the “15%” decision has big implications for broader and more abstract questions of governance. Here are two:
Taxation and separation of powers: The Constitution flatly bans taxes on exports. (Article I, Section 9: “No Tax or Duty shall be laid on Articles exported from any State.”). It’s not clear whether payouts from AMD and Nvidia under this arrangement would be considered a tax, a donation, or something else. But constitutionally, it’s a strange arrangement, and fits into an unwholesome pattern of attempts to create extra-legal “revenue streams”. See also the administration’s attempts to impose tariffs by decree and its (probably unsubstantiated) claims about the investment sections of the still-unpublished tariff “deals” with the European Union, Japan, and Korea.
State capitalism: Likewise, this arrangement is a second ill-judged move away from normal markets in which companies subject to impartial regulation compete with one another on the basis of quality and price, and towards “state capitalism,” . The first example, earlier this year, is the administration’s insistence on getting a “golden share” in U.S. Steel, with rights to participate in future investment and personnel decisions, as a condition of approving Nippon Steel’s acquisition of U.S. Steel last June. This makes the U.S. government a direct competitor to American steel companies as well as international metal suppliers. In much the same way, the Nvidia/AMD payout would make the government a direct beneficiary of exports to China from two American companies and implicitly a rival to others. To put it mildly, that’s not healthy for competing businesses and startups, and it probably, over time, isn’t good for favored “national champions” either.
Ed Gresser is Vice President and Director for Trade and Global Markets at PPI.
Ed returns to PPI after working for the think tank from 2001-2011. He most recently served as the Assistant U.S. Trade Representative for Trade Policy and Economics at the Office of the United States Trade Representative (USTR). In this position, he led USTR’s economic research unit from 2015-2021, and chaired the 21-agency Trade Policy Staff Committee.
Ed began his career on Capitol Hill before serving USTR as Policy Advisor to USTR Charlene Barshefsky from 1998 to 2001. He then led PPI’s Trade and Global Markets Project from 2001 to 2011. After PPI, he co-founded and directed the independent think tank ProgressiveEconomy until rejoining USTR in 2015. In 2013, the Washington International Trade Association presented him with its Lighthouse Award, awarded annually to an individual or group for significant contributions to trade policy.
Ed is the author of Freedom from Want: American Liberalism and the Global Economy (2007). He has published in a variety of journals and newspapers, and his research has been cited by leading academics and international organizations including the WTO, World Bank, and International Monetary Fund. He is a graduate of Stanford University and holds a Master’s Degree in International Affairs from Columbia Universities and a certificate from the Averell Harriman Institute for Advanced Study of the Soviet Union.
