U.S. Ambassador to the European Union Andrew Puzder delivered a blunt message to Brussels last week: if Europe wants access to American AI hardware, models, and data centres, it needs to stop fining U.S. tech companies. "You can't over-regulate and move the goalposts on regulations and hit companies with huge fines," Puzder told CNBC, framing AI access as something Europe must earn through regulatory concessions. It was a remarkably explicit statement of what had previously been only implied: the Trump administration views AI technology as leverage in transatlantic trade negotiations.
The timing was pointed. Earlier this month, Bloomberg reported that the European Commission had placed the "entire" AI operations of major U.S. technology companies under antitrust scrutiny, widening existing investigations into how these companies bundle AI products with their dominant platforms. The EU's AI Act, meanwhile, is approaching its mid-2026 enforcement deadline for high-risk applications. From Washington's perspective, these two pressures, antitrust enforcement and compliance obligations, are converging into something that looks like a coordinated squeeze on American companies.
Puzder's comments land against a backdrop of escalating trade friction more broadly. A separate Bloomberg report from last week noted that the U.S. envoy had warned the EU to expect higher tariffs if trade deal negotiations fail. The message across both trade and technology is consistent: the administration is treating access to the American technology stack as a bargaining chip, not an entitlement. The implicit offer to Europe is transactional: pull back your regulators, and the data centres, the chips, and the foundation models will follow.
The EU's position is genuinely difficult. European governments have spent years building a regulatory framework premised on the idea that applying rules to powerful technology platforms is a matter of democratic governance, not protectionism. The AI Act emerged from that tradition: rules about transparency, risk classification, and prohibited use cases, applied irrespective of where the companies are headquartered. Accepting that those rules are now on the negotiating table, exchangeable for favourable chip allocations, would effectively concede that Brussels does not have the authority to regulate technology it does not build.
At the same time, Europe's strategic dependency on American AI infrastructure is real and growing. The continent has no hyperscale cloud provider with global reach, no foundation model at the frontier, and no domestic chip manufacturer capable of competing with Nvidia or AMD at the high end. The EU may regulate how AI is used, but it cannot regulate what it cannot build. That asymmetry is what Puzder is exploiting.
What makes this moment different from previous transatlantic tech disputes is the speed at which AI infrastructure has become geopolitically central. Five years ago, a fight about data localisation or platform liability was a trade dispute. Today, a fight about access to AI compute and foundation models touches military readiness, economic competitiveness, and the ability to conduct scientific research. The EU has been slow to fully absorb this shift. Washington has not.
The open question is whether the EU will hold its regulatory position or whether the combination of trade pressure and infrastructure dependency will produce meaningful concessions. Neither outcome is clearly better for consumers or for the long-term health of the technology sector. A world where AI regulation is determined by which country controls the chips is not obviously preferable to one where it is determined by democratic legislatures.