The geography of tech manufacturing is under pressure as Donald Trump imposes a 25% tariff on Nvidia AI chips and other high-performance semiconductors. The Wednesday order utilizes Section 232 authority to slap duties on imported devices that meet specific performance benchmarks, such as Nvidia’s H200 and AMD’s MI325X.
At the heart of the order is the statistic that the U.S. manufactures only 10 percent of the chips it requires. The administration views this as an unacceptable vulnerability. By making imported chips more expensive, they intend to make the U.S. a more attractive location for semiconductor fabrication plants, reducing reliance on the current hub of Taiwan.
To prevent self-inflicted economic wounds, the White House has carved out exceptions for the domestic market. Imports destined for U.S. datacenters—massive consumers of AI silicon—are exempt, as are imports for startups and the public sector. This ensures that the costs of building American AI models do not skyrocket overnight.
The tariffs do, however, bite hard on international transfers. Chips made in Taiwan and sold to China are now required to detour through the U.S. for inspection. This entry into the U.S. triggers the 25% tariff. This clever legal maneuvering allows the U.S. to impose costs on China’s tech sector under the guise of import duties.
This announcement adds to a volatile period for trade, following recent 100% duties on branded drugs and levies on heavy trucks. While share prices for major chipmakers dipped slightly, the long-term impact will depend on how Commerce Secretary Howard Lutnick manages the exemption process in the coming months.