The administration’s earlier semiconductor action, signed on January 14, placed a 25% tariff on a limited group of advanced chips. Nvidia’s H200 and AMD’s MI325X were among the products identified at the time, and the White House described the measure as the first phase of a broader effort. That initial policy included exemptions for several categories, including U.S. data centers, research and development, startups, repairs, consumer and industrial uses outside data centers, and public-sector applications. Commerce officials have privately indicated that some or all of those exemptions may not survive the next round, according to the report.
The possibility of extending tariffs to finished products has raised particular concern among technology companies. Trade groups have warned that duties covering smartphones, laptops, tablets, smartwatches and connected vehicles could increase costs and interfere with planned AI product rollouts.
The Computer and Communications Industry Association estimated in a June analysis that broad semiconductor tariffs could reduce U.S. GDP by $90 billion annually and delay or cancel about 20% of data center projects planned through 2030. The group has argued for a narrower policy that exempts chips used in AI servers and lowers the presumed tariff rate from 25% to 10%.
The industry’s concern is partly tied to the amount of semiconductor capacity currently available inside the United States. Taiwan produces more than 90% of the world’s leading-edge chips, while TSMC has committed $265 billion to expand its Arizona operations. Even after that buildout, POLITICO reported that TSMC expects roughly 30% of its most advanced capacity to be located there.
Taiwan’s January trade agreement with the U.S. already uses a production-linked quota system. Companies building new U.S. plants can import Taiwanese chips duty-free up to 2.5 times their existing American manufacturing capacity. That allowance falls to 1.5 times capacity once the facilities are completed.
Industry representatives have argued that a similar formula applied more broadly would not provide enough duty-free supply for current demand. “The volume they're talking about granting duty-free wouldn't cover the hyperscalers alone, let alone the rest of the industry,” one technology industry representative told POLITICO. “Those are chips we physically can't buy here, because the capacity doesn't exist yet.” The same person added: “The math literally just does not work.”
The dispute is especially relevant for companies building AI infrastructure. Nvidia and AMD depend heavily on overseas manufacturing, while cloud providers and other large technology companies are committing substantial capital to new data centers that require large volumes of advanced accelerators.
Jonathan McHale, digital policy chief at the Computer and Communications Industry Association, compared the data center expansion to “building the transcontinental railroad” and warned that additional costs and uncertainty could put those investments at risk.
Technology lobbyists have met increasingly often with Lutnick and Bureau of Industry and Security undersecretary Jeffrey Kessler since the beginning of the summer. Three people familiar with those discussions told POLITICO that the industry’s position has weakened in recent meetings.
One participant estimated that expanding domestic manufacturing capacity enough to meet demand would take more than five years, longer than the phase-in periods the administration has used for earlier tariff measures.
The administration has maintained that reshoring semiconductor production is a central policy goal. White House spokesperson Kush Desai said increasing domestic chip manufacturing remains a top priority for President Trump. The Commerce Department did not respond to POLITICO’s request for comment.
The next round could therefore determine whether semiconductor tariffs remain focused on a limited set of advanced chips or expand much more broadly across the technology supply chain. The biggest unresolved issue is whether existing exemptions for data centers and other users remain in place while domestic manufacturing capacity continues to grow.
This analysis is based on reporting from Tom's Hardware.
Image courtesy of Manufacturing Today.
This article was generated with AI assistance and reviewed for accuracy and quality.
