Donald Trump has officially imposed a 25% tariff on select high-end artificial intelligence semiconductors, including the widely used Nvidia H200 and AMD MI325X. The Wednesday announcement is the culmination of a Section 232 investigation into the national security implications of semiconductor imports. The administration argues that heavy reliance on foreign manufacturing weakens America’s strategic independence.
The core motivation behind the tariff is to incentivize the “reshoring” of chip manufacturing. Currently, the vast majority of cutting-edge chips are designed in the U.S. by companies like Nvidia and Intel but manufactured abroad, largely by TSMC in Taiwan. The White House stated that the U.S. makes only 10 percent of its required chips, a figure they intend to change through these financial penalties on imports.
However, the administration has structured the order to avoid stifling the domestic AI boom. A fact sheet released by the White House clarified that U.S. datacenters, startups, and various civil industrial applications would be exempt from the duties. This suggests a “light touch” approach, aiming to penalize foreign dependence without handicapping American tech innovation.
The tariffs also serve as a geopolitical tool. Under new regulations, chips manufactured in Taiwan but sold to China must pass through the United States for testing. This detour triggers the 25% tariff, effectively placing a tax on Chinese access to advanced computing power. This follows previous threats to slap tariffs on Chinese legacy chips, a policy that has been delayed until 2027.
Industry experts are watching closely to see how Commerce Secretary Howard Lutnick utilizes his authority to issue further exemptions. While Nvidia did not immediately comment, AMD stated they comply with all U.S. laws. The move adds to a growing list of protectionist measures, including recent tariffs on branded drugs and trucks, as the administration seeks to fortify the U.S. industrial base.