Trump Imposes 25% Tariff on Nvidia AI Chips: Datacenters and Consumers Win Exemptions

While Donald Trump imposes a 25% tariff on Nvidia AI chips and similar high-performance hardware, the fine print of the order offers relief to major American buyers. The new duties, announced Wednesday, target specific high-performance benchmarks but notably exclude imports destined for U.S. datacenters, startups, and consumer electronics.
The tariff is grounded in a national security finding that the U.S. is dangerously over-reliant on foreign supply chains for critical technology. With only 10 percent of necessary chips manufactured domestically, the administration is using Section 232 of the Trade Expansion Act to force a shift in production geography. The goal is to reduce dependence on Taiwan and other foreign hubs.
The exemptions for datacenters are crucial, as these facilities are the largest purchasers of the Nvidia H200 and AMD MI325X chips targeted by the order. By shielding these entities, the White House aims to support the domestic AI infrastructure while still signaling a tough stance on trade. Non-datacenter civil applications and public sector uses are also spared from the cost hike.
Interestingly, the tariffs seem designed to hit chips in transit to other markets, specifically China. A new requirement mandates that China-bound chips travel through the U.S. for inspection. Upon entry, these chips are hit with the 25% levy. This mechanism allows the U.S. to collect revenue on technology sales to strategic rivals while protecting its own tech giants.
Investors reacted cautiously, with chip stocks dipping slightly in after-hours trading. The order also clarifies that these semiconductor tariffs will not be “stacked” on top of other Section 232 duties, such as those on steel or aluminum. This nuanced approach suggests the administration is trying to thread the needle between protectionism and economic stability.