The Next Web reported on August 27, 2026, that the United States' semiconductor tariff remains narrower than its headline rate suggests. A 25% duty applies to a defined group of advanced computing chips, but several end uses — including US data centers — are currently exempt. The unresolved question is whether that treatment will survive the policy's next phase.
What the current rule actually says
The January 14 presidential proclamation imposed a 25% ad valorem duty on covered products from January 15. It also excluded imports used in US data centers, repairs and replacements, domestic research and development, startups, non-data-center consumer electronics, civil industrial applications and public-sector uses.
That distinction matters. The rule does not place a blanket 25% tariff on every semiconductor or every device containing one. The covered tariff classifications and technical thresholds narrow the first phase, while end-use exceptions further limit which imports pay the duty.
The proclamation directed the Commerce Secretary to provide an update on the market for semiconductors used in US data centers by July 1. The Next Web's report says that update has not been published publicly. In the absence of a new published measure, the existing exemption remains the clearest confirmed rule; claims about its removal are scenarios, not current policy.
Why data-center operators care
Advanced accelerators and servers are a large part of AI infrastructure costs. Applying a tariff to chips or to finished derivative products could change project budgets and procurement timing. But the size of any effect depends on the final scope, product classification, sourcing contracts and whether an importer qualifies for an exemption.
The proclamation also describes a possible second phase with broader semiconductor tariffs and a tariff-offset program for companies investing in US production. It does not provide the final rate, product list or implementation date for that phase. Those missing details prevent a responsible estimate of the cost for a specific operator.
The policy trade-off
The administration says the measure is intended to reduce dependence on foreign semiconductor supply chains and encourage domestic manufacturing. At the same time, exempting data-center imports supports near-term AI construction that cannot immediately rely on new domestic fabrication and packaging capacity.
Those goals can pull in different directions. A wider tariff may strengthen an incentive to localize supply over time while raising near-term costs for infrastructure built with imported equipment. That is analysis of the policy mechanism, not a forecast that projects will definitely move, stop or become more expensive by a particular amount.
What to watch
The decisive evidence will be a Commerce update, a Federal Register notice changing the tariff schedule or end-use certifications, or a new presidential action setting the second phase. Until one appears, procurement teams should distinguish the current narrow duty from broader proposals and model several scenarios rather than treating one as settled.
Sources and methodology
This analysis uses the original TNW report and the January 14 White House proclamation. The legal text is treated as the primary source; estimates and predictions from interested groups are not presented as confirmed outcomes.