Frontier AI labs are betting trillions of dollars on scaling computation, yet more than 60% of Americans favour limiting new data centre construction in their communities.
Local opponents disrupted projects worth $68 billion in the second quarter of 2026, according to trackers at Data Center Watch. Sentiment against these facilities is growing fast, creating a friction point for AI companies trying to reconcile demand for their products with opposition to the infrastructure behind them.
Some industry explanations, such as the claim that opposition stems from secretive Chinese meddling, do not hold up to scrutiny. A new report from the nonprofit Data & Society offers a different perspective, based on 18 months of ethnographic fieldwork where researchers interviewed 44 Pennsylvanians between 2024 and 2026.
The authors were not trying to measure the empirical impact of data centres, but instead wanted to understand how people experience the debate over their construction. Livia Garofalo, a Data & Society researcher, told TechCrunch that doing research on the ground for an extended period helps pierce the bubble of AI hype.
People are making sense of the dissonances between AI being pitched as the next inevitable industrial revolution and the fears surrounding it. When the authors began their work, data centres were being sold as key to economic development. Governor Josh Shapiro headlined a July 2025 summit celebrating $90 billion in data centre and AI infrastructure investment.
A year later, Shapiro signed an order imposing new requirements on data centre projects and removing them from a regulatory fast-track initiative. The message from the industry that AI is inevitable has landed poorly with Pennsylvania residents. The state has a history of industrialisation, including coal mining, the early days of the oil industry, steel manufacturing, and fracking, which has left residents suspicious of utopian industrial schemes.
People in Pennsylvania especially have incredible experience with industrial change and deep lineages of industrial trauma, enough to sniff when things do not feel right to them, said Maia Woluchem, a program director at Data & Society who co-authored the report.
Trade unions remain perhaps the greatest source of support for data centre projects, thanks to the potential jobs they represent. Even acknowledging that these facilities will not be major sources of employment once built, unions see the projects as a lifeline to rebuild their ranks.
There’s an influence campaign, alright
In contrast to fears of foreign influence, the researchers describe the work of AI boosters as a kind of propaganda campaign. They point to the Data Center Coalition, a trade group that has set up websites like Pennsylvania Connects to boost projects and spent at least $19,000 lobbying the state government last year.
Despite the boosterism, resistance has grown in unlikely ways. The researchers were surprised by the diversity of the opposition, calling it not bipartisan but post-partisan. We do not want to romanticise that everyone is having this kumbaya moment, [but AI is] forcing people to be in conversation together about futures that they want and presents that they want … these fights are incredibly organic, Woluchem said.
Sometimes the opposition comes from folks who are against development, classic NIMBYs. The coalition grows from there to include people who dislike AI as a technology and the corporate interests behind it, environmental groups concerned about water usage and energy-related emissions, and those with more nebulous concerns about public health.
What is striking is how often material interests are top of mind. People are concerned that data centre electrical usage will affect their own power bills, worried that data centres will drive down property values, or are upset by the playbook that data centre builders use to get their way. Tactics include asking local officials to sign non-disclosure agreements.
There is a significant group of people who are just really clued into their town hall meetings and might notice that wow, a representative from Amazon, Meta, Microsoft, showed up here, or a data centre developer showed up at my community meeting …that’s a little odd, Woluchem notes. And my council person, who I have previously had a relationship with, now won’t talk to me about this sort of thing.
It is this patchwork quilt of motivations for anti-data centre activism that presents such a challenge for the AI industry. There is no simple way to alleviate this raft of concerns. The problem is trust, which has already deviled the industry.
It is an intimate and relational trust that is now being inserted in people’s lives by these industries, Garofalo said. I’m alone in my house, and I’m talking to a chatbot, and I’m feeling loved and cared for, and yet I know that this is a commodity, and I know I’m being extracted upon.
What it means
For people making things with AI, the practical reality is that the technology is being sold as a public good while the costs are localised. Communities are no longer asking if the technology works, but who gets to decide where the physical infrastructure sits and how the power bills are affected. The industry cannot simply deploy solutions; it must rebuild the trust that was broken when the pitch ignored the immediate material concerns of residents.
