How Semiconductor Export Controls Work—and Why

Why Chips Are a National Security Issue

Advanced semiconductors power everything from smartphones to fighter jets, but their most consequential role may be in artificial intelligence. Training a frontier AI model requires thousands of cutting-edge processors—chips that only a handful of companies worldwide can design or manufacture. That concentration of capability has turned semiconductors into a geopolitical flashpoint, with governments racing to control who can buy, sell, and build the most powerful chips on Earth.

The Architecture of Control

The United States runs the world's most consequential semiconductor export control regime, administered by the Bureau of Industry and Security (BIS) within the Department of Commerce. BIS maintains the Export Administration Regulations (EAR), which govern what technology can leave the country—and to whom.

The system rests on three pillars:

• Performance thresholds — Chips are restricted based on measurable specs such as processing power, interconnect bandwidth, and transistor density. In October 2022, BIS banned exports to China of any AI chip equal to or more capable than the Nvidia A100, setting a benchmark that has been periodically updated since.
• The Entity List — A roster of foreign organizations deemed a national security risk. Any export of controlled items to an Entity List designee requires a specific license, which is usually denied. As of late 2024, BIS had added over 140 new entities in a single round, including Chinese chipmakers and companies linked to Huawei.
• End-use and end-user restrictions — Even items below the performance threshold can be blocked if BIS determines they will be used for military applications, weapons of mass destruction, or advanced computing by restricted parties.

The Foreign Direct Product Rule

America's most potent enforcement mechanism is the Foreign Direct Product (FDP) Rule. Because nearly every advanced chip in the world is designed with American software or fabricated using American-made equipment, the FDP Rule extends U.S. jurisdiction far beyond its borders. If a chip made in Taiwan or South Korea was produced using American technology, Washington can block its re-export to restricted destinations—even though the chip never touched U.S. soil.

Entities with a special "Footnote 5" designation on the Entity List face the strictest version of this rule. For these companies, any foreign-made product incorporating U.S. technology requires an export license, according to Congressional Research Service analysis.

How Enforcement Works—and Fails

On paper, the controls are formidable. In practice, enforcement faces persistent challenges. Smuggling networks use shell companies, fake end-user documentation, and transshipment through third countries to move restricted chips to prohibited buyers. A March 2026 federal indictment alleged that associates of a major U.S. server maker diverted roughly $2.5 billion worth of Nvidia AI servers to China through Southeast Asian intermediaries, reportedly using hair dryers to remove serial numbers and repackaging shipments to conceal their final destination.

Meanwhile, restricted countries invest heavily in domestic alternatives. China's Huawei has developed its own Ascend AI chips, and while experts at the Council on Foreign Relations note these remain generations behind Nvidia's best, Chinese AI labs have still managed to produce competitive models using workarounds and accumulated stockpiles.

The Policy Tug-of-War

Export controls carry economic costs. American chipmakers like Nvidia and AMD lose billions in potential revenue, while allies such as the Netherlands (home to lithography giant ASML) and Japan face pressure to align their own restrictions with Washington's. The Information Technology and Innovation Foundation has warned that overly broad controls risk undermining the very U.S. innovation base they aim to protect, by pushing foreign customers toward non-American alternatives.

Policy has oscillated accordingly. The Biden administration tightened controls three times between 2022 and 2024. The Trump administration subsequently eased some restrictions in early 2026, shifting from blanket denial to case-by-case licensing for certain AI chips—on the condition that exporters pay the government a percentage of proceeds and that aggregate shipments to China not exceed 50% of what a company sells to U.S. customers.

Why It Matters

Semiconductor export controls represent a new kind of economic statecraft—an attempt to use supply-chain dominance as a strategic weapon. Whether they succeed in maintaining a lasting technological advantage or merely delay the inevitable remains one of the defining questions of 21st-century geopolitics.