technology 7 min read

Trump's Data Centre Ban Could Fracture the AI Supply Chain

Trump's threat to ban US companies from building AI data centres overseas risks shattering the very supply chains powering global AI expansion—and may accelerate China's push toward technological self-sufficiency.

  • Artificial Intelligence
  • China Tech
  • US Politics
  • Tech Policy
  • Data Centres

A Policy Without a Policy

Donald Trump has done something unusual in the chaotic architecture of his second term: he floated a sweeping policy before the people supposedly responsible for writing it had even seen it.

Speaking to reporters after returning from a trip to Asia, the president said he would ban American companies from building AI data centres outside the United States. He offered no draft executive order. No statutory language. No timeline. Asked whether he had already directed his cabinet to begin work on such a measure, his answer was blunt: “I haven’t even said that to my people. But I’m telling them now, they’re here.”

He also threatened a “tremendous tax” on firms that ignore the rule—without specifying what form that penalty would take or how it would be enforced across multinational corporations whose construction projects span dozens of countries.

What is clear, though, is the direction of travel. Trump positioned the move as a direct counter to China, arguing that if American firms build AI infrastructure abroad, Beijing will eventually seize it. “Because China will take all of them,” Trump told reporters. “And we’re not allowing that to happen.”

It was a strikingly loose conception of geopolitics, one that glosses over how data centre capacity in Finland, Canada, or Southeast Asia isn’t owned by any single country and cannot simply be “taken” in the way territorial claims are. But the political instinct behind the rhetoric is unmistakable: keep AI infrastructure inside America, keep the economic benefits domestic, and force rivals to compete on turf that favours the United States.

The Numbers Trump Is Ignoring

America does lead in data centre construction, by a significant margin. According to Pew Research Centre, there were roughly 3,000 operational data centres across the US as of April this year, with another 1,500 in various stages of approval or construction. Many sit in rural communities where cheap land, low population density, and available energy grids make the economics work.

That rural advantage is precisely why Democrats targeted the issue during the November midterms. Environmental concerns—noise pollution, massive water consumption, and strain on local power grids—helped drive wins for candidates promising tougher regulation. In Finland, Google’s €13 billion AI infrastructure investment faced exactly this problem: construction on two of three planned data centres was halted because clearing more than 300 hectares of forest triggered widespread opposition.

Trump called out that Finnish project specifically, saying his administration was “not allowing that to happen.” The message was aimed squarely at Google, which has yet to respond publicly to the threat.

But banning foreign construction doesn’t create new domestic capacity overnight. It merely redirects demand onto a grid and supply chain already struggling to meet it. The question isn’t whether Trump’s rhetoric will slow overseas build-outs—it’s whether it can, and at what cost to the companies caught between Washington’s demands and their own commercial obligations.

Who Actually Builds Overseas (And Why)

The hypothesis that American data centres built abroad would fall into Chinese hands rests on a particularly aggressive reading of geopolitical risk. In practice, most of the recent wave of overseas construction is being driven by European and North American hyperscalers looking for space, power, and regulatory flexibility unavailable in saturated US markets.

Meta has already broken ground on a data centre in Alberta, Canada. Anthropic signed a long-term lease agreement for a proposed facility in Nova Scotia. Canada is positioning itself explicitly as an alternative to the US, marketing itself to hyperscalers with fewer environmental obstacles and a stable electricity supply.

Google’s Finland investment—its largest single AI infrastructure commitment—was an attempt to tap into northern Europe’s cool climate, renewable energy, and growing talent pool. The cancellation of two sites due to forest clearance disputes wasn’t a strategic retreat; it was a negotiation with local communities that ultimately tipped against the developer.

Southeast Asia has also become a major destination for data centre development, as regional governments court the same investment but face their own constraints around land, water, and grid capacity. Chinese companies have been renting computing power from these regions specifically to circumvent Washington’s export controls on advanced chips—a loophole that prompted Monday’s announcement requiring licences for purchases of advanced semiconductors.

So we now have a dual policy framework: the US is restricting who gets access to the chips needed to fill those data centres, while simultaneously threatening penalties for the overseas facilities that Chinese firms rely on to work around those restrictions. The goal is containment. The mechanics are untested and likely contradictory.

The Enforcement Problem

Let’s suppose Trump’s team develops a formal policy. What does enforcement look like?

A ban on US-based construction firms building overseas would require parsing corporate structure with surgical precision. Does it apply only to subsidiaries of US-headquartered companies, or to US citizens working on foreign projects regardless of employer? What about joint ventures where an American firm holds a minority stake? How do you distinguish between engineering design work (often handled by US firms internationally) and actual construction labour?

The “tremendous tax” Trump referenced adds another layer of complexity. Tax law requires congressional action for new levies—or a narrowly defined executive authority that would face immediate legal challenge. And penalties that target foreign construction activity could trigger trade disputes with allies, particularly Canada and EU members whose governments have actively courted the very investments Trump is trying to redirect.

Meta and Google have both been contacted for comment. As of publication, neither has responded publicly.

The Unintended Consequences

Here is what makes this policy proposal genuinely risky, beyond its enforcement difficulties: it may achieve the opposite of its stated goal.

If the United States effectively walls off its most advanced technology firms from building infrastructure abroad, it removes the primary mechanism through which those firms scale globally. AI development requires enormous computing capacity. US companies currently access it through a combination of domestic build-out and international partnerships. Cutting off the international side doesn’t shrink China’s ambition—it shrinks the competitive landscape in which American firms operate.

More importantly, it pushes every non-US actor toward self-sufficiency. China has already accelerated its domestic semiconductor programme precisely because US export controls made foreign sourcing unreliable. A ban on US firms building data centres overseas sends the same signal to Japan, South Korea, India, and the EU: if you want sovereignty over your own AI stack, you must build your own infrastructure independently of American supply chains.

This isn’t hypothetical. It’s the observed pattern each time Washington has extended its technology restrictions. The response hasn’t been capitulation—it’s been acceleration.

What Comes Next

Trump’s remarks came shortly after his administration announced that firms would need licences to buy advanced chips, so officials can track whether that hardware ends up serving Chinese companies or being smuggled into China. That licensing regime directly impacts Southeast Asia, where Chinese firms rent computing power from data centres that themselves depend on the same Western semiconductor supply chain Trump now wants to nationalise.

We are watching a policy framework collide with itself in real time: restrict chip exports to prevent Chinese AI advancement while simultaneously banning the foreign data centres Chinese firms depend on to bypass those very restrictions. The overlap is inevitable. The outcome is not guaranteed to favour Washington.

Whether this becomes formal policy remains unknown. Trump has rarely let a concrete regulatory architecture slow down a declarative moment. But the fact that he raised the idea publicly—before internal coordination, before legal review, before consultation with allies—suggests this isn’t an idle campaign gesture. It’s a directional nudge, signaling to his base and to markets alike that the administration views AI infrastructure geography as a strategic battleground.

The companies caught in the crossfire—Google, Meta, Anthropic, Amazon Web Services—will have to decide whether to comply, contest, or quietly continue their existing plans while the dust settles. The allies whose investments are affected will watch carefully to see whether this is rhetoric or law. And China’s policymakers will note yet another reason to reduce dependency on American technology, even as Washington frames the restriction as defensive.

The paradox of trying to nationalise AI infrastructure is that it may well accelerate the fragmentation of the global tech ecosystem into competing spheres of influence—each one less integrated, less interoperable, and ultimately less efficient than the system it replaces.