A 25% US tariff on advanced semiconductors implemented in January includes a provisional exemption for semiconductors destined for American data centres, sparing the AI infrastructure buildout for now. However, a delayed Commerce Department report expected on July 1 has kept the future of this exemption uncertain, impacting investment decisions in US data centre capacity.
- 25% US tariff applies mainly to AI accelerator chips with seven specific carve-outs.
- Data centre exemption’s survival hinges on a delayed Commerce Department review.
- Removing exemption risks $90bn annual economic loss and 243,000 jobs at stake.
What happened
In January 2026, the US imposed a 25% tariff on select advanced semiconductors, particularly targeting logic integrated circuits used in AI accelerators. This tariff was designed to be precise, excluding broader categories like laptops to minimize collateral impact. Among seven carve-outs created to limit disruption was a crucial exemption for semiconductors imported for use in American data centres, intended to protect ongoing AI infrastructure development.
This data centre exemption was provisional and dependent on a Commerce Department report scheduled for release on July 1, 2026. However, the report remains unpublished, leaving operators uncertain about the future tariff status. The proclamation also authorized the potential expansion of tariffs to include semiconductor manufacturing equipment and assembled servers, which would significantly affect the economics of domestic data centre builds.
Why it matters
The exemption is a critical factor in balancing US national security goals of reshoring semiconductor manufacturing with the immediate need to support AI data centre growth. A detailed analysis by the Computer and Communications Industry Association estimates that if the data centre exemption is removed, the added 25% cost on computing equipment imports could effectively tax data centre development at around 15.6%. This would translate into an annual economic loss of approximately $90 billion and risk 243,000 American jobs.
Moreover, about 20% of planned US AI data centre capacity from 2026 to 2030—around $450 billion in capital expenditure—could be delayed, canceled, or moved overseas. This highlights the tension between the slow ramp-up of domestic semiconductor fabs, which take years to build, versus the urgent demand for advanced chips to power AI systems, largely supplied by Taiwan and other global sources.
What to watch next
The immediate focus is on the pending Commerce Department report that will determine whether the data centre exemption continues or ends. The outcome will significantly influence domestic AI infrastructure investments, potentially reshaping where and how US data centres are built.
Simultaneously, policymakers must clarify and potentially detail the promised tariff offset programs intended to support companies investing in US semiconductor manufacturing capacity. Meanwhile, global competition intensifies, with China drafting a $295 billion data centre plan explicitly excluding certain US semiconductor firms, underscoring the geopolitical stakes behind these tariff decisions.