Data Centres: America Is Debating the Wrong Question

Data Centres: America Is Debating the Wrong Question

There is a debate happening in America right now about data centres. It is loud, it is political, and it is aimed at entirely the wrong target.

Communities are pushing back. Bernie Sanders is calling for a federal moratorium on all new construction. Lawmakers from both parties in multiple states have introduced similar measures. The mayor of Denver announced a local moratorium last month. An estimated 98 billion dollars in projects were blocked or stalled in just three months last year.

The frustration driving all of this is real and in many cases justified.

But the proposed solution is not just wrong. At this particular moment in history, it is dangerous.

———

The question America is not asking

The current debate is focused entirely on whether to build data centres.

That is the wrong question.

The right question is how to build them so that communities support rather than fight them. Because every month America spends arguing about whether, China is spending building more.

———

The last clear lead

The United States currently holds one decisive structural advantage over China in the artificial intelligence race.

Compute infrastructure.

The US hosts 54% of global hyperscale data centres and maintains roughly ten times more compute capacity than China. That advantage is not just a technology statistic. It is the engine behind every AI productivity gain, every drug discovery breakthrough, every efficiency improvement that American workers and companies are building on right now.

Everything else in this race, China is closing fast or already leading.

Chinese factories installed approximately 300,000 industrial robots in 2024. American factories installed 34,000. China accounts for over 80% of global humanoid robot installations. Chinese data centres pay less than half the electricity rates American ones do. Projects that take three years to build in the US take months in China. China added more power generation capacity in a single year than the US has built in its entire history.

DeepSeek demonstrated earlier this year that China can produce frontier AI models at a fraction of the cost American labs assumed was the floor.

The compute infrastructure gap is the last significant structural advantage America holds in this race. It is not permanent. It is not guaranteed. And it is currently being eroded not by Chinese investment alone but by American political paralysis.

A blanket moratorium does not protect American workers. It accelerates the erosion of the one advantage they still have.

———

The frustration is legitimate. The solution is not.

In Loudoun County, Virginia, residents near Amazon and Microsoft facilities describe a drone-like hum that hovers above their homes 24 hours a day. Over 9,000 diesel generators sit across Virginia. When the grid struggles and generators kick in, communities get noise and diesel fumes simultaneously. Utility bills have risen as grid demand surges. The financial benefits flow to shareholders in Seattle and Menlo Park. The costs sit permanently in the communities that host the buildings.

This is a broken bargain. Companies built what they wanted, where they wanted, to the minimum standard the law required, and moved on. That pattern breeds exactly the kind of backlash that is now threatening the entire industry.

But this is not inevitable. Meta proved it does not have to be this way.

———

What responsible development actually looks like

When Meta built its data centre in Odense, Denmark, it made a genuine deal with the community.

The facility recovers waste heat from its servers and uses it to heat approximately 11,000 local homes. It runs on 100% renewable energy. It has invested over 26 million dollars in local schools and community projects. The city is proud to have it.

When Meta built in Luleå, Sweden, it chose a location near the Arctic Circle to use natural cold air for cooling. The facility draws power almost entirely from hydroelectric sources and runs roughly 40% more efficiently than a traditional facility. The town now actively markets itself as a global data centre destination.

This was not Danish or Swedish ingenuity. This was Meta making a deliberate choice to build responsibly when the environment required it.

The same company. The same technology. A completely different relationship with the community.

The question for American policy is simple. Why are the same standards not being required at home?

———

The technology exists. The standards do not.

Modern liquid-cooled data centres can reduce water consumption by up to 91% compared to traditional facilities. Immersion cooling eliminates the constant fan noise driving most complaints. Waste heat recovery systems are proven at scale.

The barriers to better data centres are not technical. They are contractual.

Companies have not been required to offer communities a real deal because nobody has required them to. That is what policy should change. Not whether to build, but how.

Mandatory renewable power commitments as a condition of tax incentives. Noise standards that reflect what communities can actually live with. Waste heat reuse requirements where geography allows. Real transparency on water use, grid impact, and emissions from day one.

None of this slows the buildout. It shapes it.

———

What the moratorium argument actually costs

The 98 billion dollars stalled in the US last year is not an environmental win. It is AI capability that was not built. It is productivity gains that did not materialise. It is the compute advantage quietly narrowing while the political debate continues.

American workers stand to gain enormously from AI amplification over the next decade. But those gains require the infrastructure to exist.

If the buildout stalls, those gains go elsewhere. The companies that need compute scale their operations where the political environment is more predictable. And the communities that were supposed to benefit from American technological leadership end up with neither the data centres nor the prosperity.

Sanders says he wants to protect American workers. Stopping the infrastructure that could amplify them is not the way to do it.

———

The practical ask

For policymakers: mandatory standards, not moratoria. Require renewable power commitments as a condition of tax incentives. Set noise standards communities can actually live with. Mandate waste heat reuse where geography allows. Make the Meta model in Odense the minimum, not the exception.

For companies: responsible development is not charity. It is strategy. Communities that feel they got a fair deal do not become political movements. The 98 billion stalled last year is the price of failing to build trust.

For American leaders broadly: there are very few dimensions of the AI race where the US still holds a clear structural advantage over China. Compute infrastructure is one of them.

Allowing that advantage to erode through political paralysis, while China builds at full speed with no debate and no permission required, is not caution.

It is surrender.

The question was never whether to build data centres. The question is whether America builds them well enough that communities support them, or badly enough that communities stop them.

One of those outcomes strengthens American competitiveness.

The other hands China the race.

———

Daniel Kafer is a global futurist and strategic advisor helping senior leaders understand how artificial intelligence amplifies human capability, leadership, and organisational adaptability.

Want to go deeper on AI strategy and what is coming next?

Join the newsletter → daniel-kafer.kit.com/9f164e9023

Or join the free AI community → skool.com/the-strategic-edge-1049/about