The US government has introduced new import tariffs on advanced AI chips from Nvidia and AMD, with the aim of channelling part of the proceeds from sales to China directly into the US treasury.
According to the Financial Times, the measure is part of President Donald Trump’s broader trade and industrial policy, which explicitly intertwines economic transactions and national security.
The core of the scheme is that the United States will receive approximately 25 percent of the revenue chipmakers generate from the sale of certain AI processors to Chinese customers. This will be done through specific tariffs on chips imported into the US and then shipped to customers elsewhere in the world. This effectively implements an earlier policy change that permitted the export of certain AI chips to China, subject to financial conditions.
Direct revenue for the US government
According to industry sources, tariffs were chosen to make the scheme legally robust. Direct payments of revenue to the government could be more easily challenged, while import duties are easier to defend under existing legislation. At the same time, the levies provide the US government with direct revenue from a market that was largely closed until recently.
The measure applies to, among others, Nvidia’s H200 and AMD’s MI325X, both intended for heavy AI applications. Chips used to build US AI infrastructure are excluded from the scheme to stimulate domestic investment. Nvidia and other chip designers mainly produce their chips through external manufacturers, particularly in Taiwan, after which assembly and integration take place elsewhere.
The new tariffs are part of a broader national security investigation into semiconductors that was launched last year. This investigation is looking at dependence on foreign production and the strategic risks involved. The government is explicitly leaving open the possibility of introducing broader tariffs on semiconductors and derivative products at a later stage.
Nvidia invests heavily in the US
At the same time, Washington is pressuring technology companies to move production to the US. Nvidia has pledged to invest hundreds of billions of dollars in US manufacturing capacity, while TSMC is building new factories in Arizona. Nevertheless, Taiwan remains the center for the most advanced chip production for the time being.
On the Chinese side, it is still unclear how the measure will work in practice. Although the US allows exports of certain chips, China seems reluctant to grant access. The Chinese government is actively encouraging the use of domestic chips to reduce dependence on foreign technology. Reports that customs authorities have temporarily delayed the clearance of specific AI chips underscore this tension.
For Nvidia and AMD, the new policy means more clarity but also higher costs and additional complexity. For the global semiconductor industry, it shows how geopolitics, trade, and technology are becoming increasingly intertwined.