Washington, D.C. — President Donald Trump has threatened to impose what he described as a “tremendous tax” on American companies that build AI data centres outside the United States, marking the first direct White House intervention into the geography of global compute infrastructure.
The statement, reported by The Verge AI, came during remarks in which the President framed overseas data centre construction as a form of economic leakage that undermines American technological leadership. The threat, while lacking legislative detail, represents a significant policy signal for hyperscalers, chip manufacturers, and sovereign AI programmes that have bet heavily on distributed compute footprints.
What Was Said
According to The Verge AI’s reporting, the President indicated that companies choosing to locate AI data centres in foreign jurisdictions could face punitive tax measures designed to incentivise domestic construction. No specific rate, mechanism, or timeline was provided, but the framing suggested tariffs or tax penalties tied to capital expenditure abroad.
The remarks align with the administration’s broader posture on trade and industrial policy, which has favoured tariffs as a primary lever for reshaping corporate behaviour. Unlike previous tariff threats targeting imported goods, this proposal would effectively tax the location of infrastructure investment — a novel application of trade policy to fixed capital assets.
Business Impact: Who Gains, Who Loses
The immediate beneficiaries of such a policy would be domestic data centre operators and construction firms with significant US capacity. Companies including QTS, Digital Realty, and CyrusOne, which hold substantial American real estate portfolios, could see accelerated leasing demand if hyperscalers are forced to repatriate compute workloads.
The clearest losers would be US-headquartered cloud providers with aggressive international expansion plans. Microsoft, Google, and Amazon have each committed tens of billions of dollars to overseas data centre construction in 2024 and 2025, targeting markets where energy costs are lower, permitting is faster, or proximity to customers is essential for latency-sensitive AI inference workloads.
Smaller economies that have positioned themselves as data centre hubs — Ireland, Singapore, and the Nordics among them — would face a chilling effect on foreign direct investment. Ireland alone hosts over 80 operational data centres, many serving US hyperscalers, and has become economically reliant on the associated construction and energy infrastructure spending.
The Energy Constraint
The threat arrives at a moment when US domestic data centre expansion faces significant bottlenecks. Grid interconnection queues for large power users in key markets like Northern Virginia and Silicon Valley now extend beyond five years in some cases. Power purchase agreements for renewable energy have become increasingly competitive, and natural gas peaker plants face permitting delays.
Industry analysts note that a punitive tax on foreign construction would not, by itself, resolve these domestic constraints. Without parallel reform of energy permitting and grid access, the policy could simply raise costs across the AI supply chain without materially shifting compute capacity onshore.
What to Watch Next
The critical question is whether the threat translates into legislative text or executive action. A formal tariff on foreign data centre construction would require either congressional authorisation or invocation of emergency trade powers under Section 301 of the Trade Act of 1974. Both paths face legal and political hurdles.
Corporate lobbying efforts from the cloud computing sector are expected to intensify in the coming weeks. The Information Technology Industry Council, which represents major hyperscalers, has previously warned that restrictions on global data flows and infrastructure investment would cede competitive ground to Chinese cloud providers expanding across Asia, Africa, and Latin America.
The next concrete signal will likely come in the form of draft language or a formal trade investigation notice. Until then, capital expenditure committees at major AI companies face a difficult planning environment: commit to overseas buildout and risk retroactive penalties, or pause expansion and risk losing ground in fast-growing international markets.
This policy direction, if pursued, would redraw the map of global AI infrastructure investment at a moment when compute capacity is the binding constraint on frontier model development.







