Following months of speculation, Donald Trump imposes a 25% tariff on Nvidia AI chips, but the final order includes significant relief for U.S. businesses. The proclamation, signed Wednesday, targets powerful processors like the Nvidia H200 and AMD MI325X under the banner of national security. However, the administration has opted to exempt the biggest buyers of these chips: U.S. datacenters.
The investigation preceding this order concluded that the U.S. faces a security risk due to its reliance on foreign chip manufacturing, primarily from Taiwan. By imposing tariffs, the White House hopes to encourage domestic production. Yet, recognizing the critical role these chips play in the economy, exemptions were granted to startups, non-datacenter consumer applications, and civil industrial uses.
This “light touch” approach contrasts with tougher rhetoric heard in December regarding Chinese legacy chips. While those specific measures were delayed until 2027, the new tariffs strike an immediate blow to chips transiting through the U.S. en route to China. These units, forced to detour to America for testing, will face the full 25% levy.
Commerce Secretary Howard Lutnick has been given broad discretion to apply further exemptions, suggesting the list of protected industries could grow. This flexibility is likely intended to prevent supply chain bottlenecks while keeping pressure on manufacturers to build American factories. The order also clarifies that these tariffs will not be stacked on top of existing duties on raw materials like steel or aluminum.
The semiconductor industry’s response has been measured. AMD issued a statement confirming compliance, while Nvidia has not yet commented. With the tariff effectively shielding U.S. innovation while taxing foreign-bound tech, the administration aims to have its cake and eat it too—bolstering security without stalling the AI boom.