Smart Decision or Big Shakedown?

The recent announcement that the Trump administration would license the export of Nvidia H20 chips and similar chips from Advanced Micro Devices, Inc. (AMD) to China, with the two companies agreeing to pay 15 percent of their revenue from the sales to the government, has raised many eyebrows. There appear to be two categories of argument—the national security implications and the payment issue.

On the national security issue, the development of these chips was the result of export controls the Biden administration put in place in 2022 and 2023. The chips were specifically designed to perform at a level that did not exceed the control limits that were imposed. This was entirely predictable and has occurred in the past. Essentially, when you draw a line and say manufacturers can’t export above the line, you create an incentive to manufacture products below the line. My first encounter with this was on an import-related matter—the Section 201 tariffs on Japanese motorcycles in 1983. The tariffs were imposed on motorcycles with an engine displacement of over 700 cubic centimeters (cc). Within a few months, motorcycles were being marketed with a displacement of 695 cc.

 In the current chip debate, one side is unhappy because they think the line is in the wrong place—that it permits the export of items that compromise U.S. national security, including the specific Nvidia and AMD chips in question. They point out that the Trump administration initially prohibited the export of these chips and then changed its mind and has hinted that the payments and/or Trump’s determination to get a one-on-one meeting with Xi Jinping are the reason for this softening of position.

The other side points out that the chips were below the existing line and therefore should not be considered a security threat, implicitly acknowledging that they think the line is in the right place. They also make the “flood the zone” argument I explained in a previous column. That argument is that these chips are essential to AI development and that if the United States wants to be the world’s AI leader, it needs to widely disseminate its products (and not just the chips) and attempt to make the U.S. tech stack the global standard. By that logic, the United States should not be exporting just chips but other parts of the AI tech stack as well. The opposing side also points out that permitting such sales will bring in substantial revenue for U.S. companies, which they can use to develop the next generation of products.

This replays a debate that has been going on for at least 40 years. The items are different, but the issue is the same—using my marathon metaphor, is it better to run faster or to trip the other guy? This time around, it’s more complicated because China is determined to run faster on its own and develop its own technology, including advanced chips, rather than relying on ours. The day will come—and may already have arrived—when they decide they neither want nor need our technology, which suggests both sides in this debate are wrong. Denying U.S. technology to China won’t change their plans, and neither will flooding the zone. The latter, however, might make the U.S. AI tech stack the global standard despite China’s efforts.

That brings us to the payment issue. There is no precedent for charging an export fee, and opponents argue that it violates both the Constitution and federal law. Article 1, Section 9, of the Constitution imposes a ban on taxing exports “from the states.” The latter phrase is important because many of the chips in question are manufactured in Taiwan and exported from there. They are subject to U.S. export controls but are not exported “from the states.” In addition, the tax is based on revenue derived from the sales and not directly on the volume of chips exported, although previous litigation on this clause suggests the courts will probably not buy that argument. Second, in a “belt and suspenders” moment, the authorizing statute, the Export Control and Reform Act, also explicitly prohibits charging a fee for an export license. It may be difficult to get around these two provisions, although it is uncertain who would have standing to sue, aside from the two companies, which have already agreed to pay.

The larger issue, of course, is not the legal one but the ethical one. The signal this sends—and administration officials have said it may be expanded to other items—is that U.S. national security is up for sale, and our economic regulators are now operating on a “pay to play” basis. If this becomes the norm, it will be enormously damaging to the U.S. economy, to our security, and to our standing in the world. Instead of making a clear, well-considered statement about which technology exports would endanger national security, the administration is saying, “Come visit with the president and bring your checkbook, or, preferably, your cryptocurrency.” In addition to creating great uncertainty about U.S. policy, it is also a shakedown. If big companies can buy the regulation they want from the government, then the door is open to both corruption and bad policy.

William A. Reinsch is senior adviser and Scholl Chair emeritus with the Economics Program and Scholl Chair at the Center for Strategic and International Studies in Washington, D.C.

Trade continues to be the hottest policy topic in Washington, which is why we’re bringing back our ⁠Crash Course: Trade Policy with the Trade Guys⁠ this fall. If you missed our spring course, now is the perfect time to register. The course runs October 8–9 at CSIS or via Zoom. Registration is open until October.