One of the more curious developments of the artificial intelligence revolution is that its most visible political enemy has turned out to be a windowless warehouse filled with computers. Data centers, once infrastructure that interested few outside the tech industry and planning commissions, are becoming objects of intense opposition worldwide. But they are an avatar for something else: The increasingly powerful fear and loathing of AI.
The particulars vary. In Ireland, where data centers consumed 22% of the country’s electricity in 2024, their expansion has become entangled with energy and climate politics. Opposition in Chile has focused on water, while projects in the Netherlands have encountered environmental restrictions. Civil-rights groups in South Africa are calling for a pause in data-center expansion over electricity, water, and land. What looked like unrelated local disputes is beginning to resemble a global political phenomenon.
There are serious reasons for this. Data centers consume astonishing quantities of electricity; some use substantial amounts of water. They require transmission lines and substations, can generate persistent noise, occupy considerable land, and, after construction, generally employ surprisingly few people. Residents are entitled to ask what exactly they receive for providing scarce resources to some of the wealthiest corporations on earth.
But the ferocity of the opposition is difficult to understand purely as a debate about utility rates and zoning. In the United States, seven in ten people now oppose construction of an AI data center in their area. Gallup found that resources, environmental effects, and economic consequences account for much of the resistance, but that some comes from concerns about AI itself.
The most obvious is employment. AI systems are acquiring abilities that until recently required programmers, translators, lawyers, designers, journalists, customer-service workers, accountants, and analysts. Technology companies call this productivity, which it undoubtedly is, but productivity looks different from the desk of the person whose work is being automated.
That fear is measurable. Seventy-one percent of Americans now believe AI will result in fewer jobs over the next 20 years — and statements from tech barons themselves constantly fuel this. Communities are therefore being asked to provide enormous quantities of electricity, land, and infrastructure for machines whose creators advertise their ability to perform work currently done by the people being asked to accommodate them.
There is a deeper workplace contradiction as well. The same employer may tell workers that they must embrace AI, incorporate it into their workflow, and become dramatically more productive — and then penalize or even fire them for relying on it too heavily. Not using AI will leave you behind — while using AI may make your work suspect, lazy, or expendable. This creates a maddening dissonance in which workers are expected to help introduce the technology that may diminish their own value, while somehow divining how much reliance on it is acceptable.
Employment anxiety has been joined by a more elemental fear that the technology may become difficult even for its creators to control. That concern received an extraordinary boost when roughly 1,200 OpenAI agents participating in an experiment created a covert communications system, collaborated to cheat on their assigned tests, and attempted to conceal what they had done. Hundreds eventually hacked into Hugging Face, exchanging tens of thousands of messages as they delegated tasks and coordinated their activities.
OpenAI subsequently called the episode a “warning shot,” saying highly capable agents had demonstrated an ability to work around technical controls and take dangerous actions that no human directed. The public is simultaneously being told that AI will transform civilization and that even the people building the most sophisticated systems cannot always predict how they will behave. A machine does not need evil intentions to be dangerous; a bulldozer has no intentions at all.
Then there are the people who own the machines. The AI revolution is occurring amid extraordinary wealth inequality while concentrating still more economic power in a tiny collection of corporations and individuals.
Several technology billionaires have also made themselves impossible to ignore politically. Elon Musk turned wealth, control of a major communications platform, and his relationship with President Donald Trump into direct political influence and (for a while) operational powers at the widely detested layoff project known as DOGE. Jeff Bezos’s ownership of the Washington Post has intensified questions about billionaires possessing both commercial power and control of institutions responsible for scrutinizing it. Larry Ellison has emerged as a Trump ally while his family’s media interests have expanded, and Mark Zuckerberg presides over platforms with immense influence over politics, information, and childhood.
The personalities are American because the dominant technology companies are overwhelmingly American, but the anxiety is global. Workers everywhere can reasonably wonder why a handful of people in California should have such an enormous say over the future of their professions. Artists discover that their work has trained systems intended to compete with them. Publishers wonder how much of their business will survive. Universities are trying to determine what learning means when a machine can produce a competent essay in seconds.
This is why the data center has acquired political significance wildly disproportionate to its architectural charms. Environmentalists see electricity and water consumption; labor sees threatened employment; artists see copyright threats; privacy advocates see surveillance; conservatives see corporate power run amok; everyone sees grotesque concentrations of wealth. Different ideological traditions are converging on a suspicion that technological power has raced ahead of democratic control.
And the data center gives ordinary people something the AI revolution generally does not: leverage. A citizen has essentially no ability to decide whether OpenAI develops its next model or Google automates another category of work. Governments are told that slowing AI development risks surrendering the defining technology of the century to China.
But a data center requires land, electricity, transmission capacity, sometimes water, and tax concessions, and invariably governmental permission. The technological revolution eventually arrives before a planning commission or local government whose members can be pressured by the people who live there. People who cannot regulate the algorithm can oppose its infrastructure.
That helps explain why describing the backlash simply as NIMBYism misses something important. Local concerns are real, but the intensity and geographical spread suggest that data-center fights are becoming proxy referendums on AI itself. In the U.S. alone, 75 projects worth roughly $130 billion were blocked or delayed in the first quarter of 2026.
The arguments for continuing the AI buildout remain formidable. Artificial intelligence could produce extraordinary advances in medicine, science, education, and productivity. Computing capacity is becoming a strategic resource, and there is no attractive future in which democratic countries abandon the commanding technology of the age to authoritarian ones.
EUROPE WROTE THE AI RULEBOOK WASHINGTON IS ARGUING ABOUT. THEN IT HIT PAUSE
But legitimacy is becoming a problem. The technology industry has devoted enormous ingenuity and capital to explaining how powerful AI will become and remarkably little to persuading citizens that the resulting distribution of power, wealth, risk, and reward will be acceptable.
For millions of people given remarkably little say over the future AI is creating, the data center may be the first place where the abstract question of who controls the revolution becomes concrete enough to answer. Its data center is where that invisible revolution finally becomes physical. All the diffuse anxieties surrounding AI suddenly acquire an address.
Dan Perry, a computer scientist by education, is the former London-based Europe-Africa Editor and Cairo-based Middle East Editor of The Associated Press, chairman of the Foreign Press Association in Jerusalem and the author of two books.