Members of the Dutch direct-action group Geef Tegengas broke into the construction site of a Microsoft data center being built in Amsterdam's Westpoort district. The activists clashed with police; some climbed a crane, lit flares and tied helium balloons to the equipment. Similar actions took place at various points across Europe around the same time. The incident may seem minor, but it shows that physical protest against AI infrastructure has now spilled into the streets.

What happened?

The action took place simultaneously around and inside the construction site. The group aimed to temporarily halt construction and make the environmental burden of data centers visible. Police intervened and several arrests were made. There is no sign that Microsoft's construction plans have changed; the action remained symbolic. What matters is less the incident itself than the trend it represents.

Why does it matter? The hidden physical cost of AI

Most users experience AI as a "cloud": an abstract, everywhere, seemingly cost-free service. Yet behind every chatbot response lies concrete, copper, steel and an enormous amount of electricity. Data centers consume serious volumes of water and energy to cool GPU clusters running 24/7. The technology is abstract; what is concrete is a local town's grid, its water supply and its land.

Recently, Microsoft, Google, Amazon and Meta have announced record levels of data center investment. Their capital expenditures have grown exponentially compared with a few years ago. That is one side of the coin of this investment enthusiasm. The other side is the discontent building up in local communities: rising energy bills, disputes over water allocation and the question, "what does this facility actually bring us?" A data center creates very few jobs but places a permanent burden on a region's resources. For local residents, the equation often looks lopsided.

This is not a new dynamic. We have seen similar resistance before with wind turbine projects, mining sites and large logistics hubs. The difference is this: AI has been marketed to the public as a "clean" technology that belongs to the future. As its physical footprint clashes with that image, the backlash grows harsher.

Why did the first major backlash erupt in Europe?

The largest AI companies are US-based, but the first visible civil disobedience broke out in Amsterdam. That is no coincidence. In Europe land is scarcer, the population is denser and environmental regulations are stricter. Countries like the Netherlands have already been squeezed for years over nitrogen emissions and water management; adding a major new consumer is politically sensitive.

Europe also has a stronger tradition of direct action and a more organized climate movement. In the US, data centers are mostly built in sparsely populated states with cheap energy, entering local economies with the promise of tax revenue. In Europe, by contrast, a facility is closer to where people live and consent is harder to secure.

DimensionUS modelEuropean model
LandAbundant, cheap, sparsely populatedScarce, expensive, densely populated
RegulationState-based, incentive-drivenStrict environmental and water rules
EnergyCheap, sometimes fossil-heavyExpensive, grid under strain
Public backlashSoftened by tax revenueOrganized resistance is stronger

This table explains why the "social license" crisis became visible first in Europe. Social license refers to a community's de facto approval of a project, beyond its legal permits. The legal permit may be on the table; but if the public does not accept it, the project advances only with friction.

Who does it affect, and how?

For big tech companies, this means rising planning costs. Permitting processes may drag on, site selection may become harder and public relations budgets may balloon. On the investor side, this question is coming into focus: is record capital expenditure really sustainable, or are regulatory and social obstacles piling up ahead of it?

For local governments the equation is even sharper. Tax revenue and the prestige of being a "digital hub" on one side; water, energy and voter backlash on the other. In the coming period, it seems likely that some municipalities will impose conditions on new data center applications or reject them outright.

What's next?

There are a few things to watch. First, whether mandatory transparency requirements for energy and water use will be written into law. Second, whether companies will get ahead of the issue with concrete commitments such as waste-heat recovery, closed-loop cooling and renewable energy contracts. Third, whether such actions will remain one-off events or turn into an organized movement.

This action alone will not stop any construction. But it carries a signal: AI's growth story will now be constrained not only by capital and chip supply, but also by water, land and local consent. If the industry does not manage this cost early and transparently, resistance could turn into a political rather than a technical obstacle.