The Trump administration is preparing a significantly expanded semiconductor tariff regime that would cover not only chips themselves but also finished products that contain them — such as laptops, gaming consoles and data‑center servers. Politico reported the plan citing eight sources familiar with the proposals.
Under the proposal, companies that build U.S. chip manufacturing capacity would receive larger tariff‑free import quotas proportional to that capacity: the more a firm produces or invests in the United States, the more semiconductors it could bring in duty‑free. Howard Lutnick, the U.S. Secretary of Commerce, is described as a leading proponent of the construct. A key change compared with a system introduced in January is that chips bound for data centers would not necessarily be exempted.
The proposal has not been finalized; its phased implementation is under discussion, and neither the White House nor the Commerce Department commented on Politico’s reporting.
Background and rationale
In January, an executive proclamation imposed a targeted 25 percent tariff on a narrow set of advanced chips, citing national security, while at the time exempting U.S. data centers. The White House has argued that the United States consumes about a quarter of the world’s semiconductors but manufactures only roughly 10 percent of its needs entirely onshore.
The emerging second phase would impose broader and “significant” tariffs while offering exemptions or benefits to firms that invest in U.S. manufacturing. Lutnick has previously sketched a model that effectively pressures foreign manufacturers toward “make in America or pay tariffs.”
Cost risks for rapid AI infrastructure buildout
Experts warn the measure could have unintended consequences by increasing the cost of precisely the AI infrastructure the U.S. seeks to accelerate. The Center for Strategic and International Studies (CSIS) estimates that the United States could spend more than $2.7 trillion on data‑center infrastructure through 2030, and about $0.54 of every dollar spent would go directly to semiconductors.
Global manufacturing capacity is still concentrated outside the United States: about 12 percent of capacity is located in the U.S., and the most advanced AI processors are still mainly manufactured in Asian — particularly Taiwanese and South Korean — fabs. CSIS estimates an extreme, no‑exception 100 percent semiconductor tariff could add as much as $1.4 trillion to U.S. data‑center investment bills; even a much milder version would materially raise costs.
In short, Washington is using a tool intended to speed domestic chip production that could, in the near term, make the GPUs and servers relied on by OpenAI, Google, Amazon, Microsoft and the wider U.S. AI sector substantially more expensive.
Implications for Europe and EU policy responses
From a European perspective, the development is part of an accelerating global industrial policy competition. The European Union also faces significant dependence: in 2024 it consumed roughly €55 billion in chips, of which €43.6 billion were imported from third countries, covering only about 22 percent of consumption with domestic production.
Brussels therefore put forward a Chips Act 2.0 in June, focusing specifically on advanced chips needed for AI, European design capabilities and strategic autonomy. In parallel, the EU launched a call at the end of July to support up to seven AI “gigafactories,” offering up to €10 billion in public funds to mobilize at least €20 billion in private capital. The broader objective is to at least triple the bloc’s data‑center capacity within five to seven years: the Commission estimates installed capacity of about 12 gigawatts could grow to roughly 28 gigawatts by 2030.
Data centers already account for about 2.5 percent of the EU’s electricity consumption, so access to cheap and reliable energy is becoming an increasingly strategic issue for Europe’s AI competitiveness.
Limits and longer‑term outcomes
European exposure is somewhat mitigated by a 2025 U.S.–EU trade agreement stipulating that U.S. tariffs on semiconductors originating in the EU — counting both general tariffs and Section 232 duties — cannot exceed 15 percent. That provides only partial protection: the AI value chain is highly international, and many advanced chips used by European firms are manufactured in Taiwan or other Asian countries; servers and finished products contain parts from multiple jurisdictions.
If Washington extends tariffs to the broader product chain, it could raise costs for U.S. tech companies and lift AI‑infrastructure prices globally in the short term. Over the longer run, however, such a policy could create stronger incentives for TSMC, Samsung and other foundries to shift investment into the United States, drawing global manufacturing capacity westward.
For Europe, the Politico‑reported proposal is thus a twofold warning: it could temporarily improve the EU’s relative position in competition for data‑center investments, but it also risks accelerating an aggressive U.S. combination of subsidies and tariff protection that would attract manufacturing capacity to America. In that environment, the Chips Act 2.0 and the AI‑gigafactory efforts are not merely development programs but strategic attempts to build an independent European compute and semiconductor ecosystem amid U.S.–Asia technological rivalry.
