The AI industry has a infrastructure problem, and it is not a power grid problem, or a water table problem, or a zoning problem, exactly. It is a Pennsylvania problem. It is a people-who-remember-the-fracking problem.
This is, in retrospect, predictable.
People in Pennsylvania especially have incredible experience with industrial change — enough to sniff when things don't feel right to them.
What happened
A new report from the nonprofit Data & Society spent 18 months and 44 interviews trying to understand why people oppose data center construction. The finding — that communities with histories of industrial trauma are skeptical of new industries promising utopian outcomes — is the kind of thing that takes 18 months of ethnographic fieldwork to confirm, and about four seconds of local knowledge to suspect.
The numbers behind the skepticism are not small. Data center projects worth $68 billion were disrupted by local opponents in Q2 2026 alone. More than 60% of Americans now favor limiting new data centers, and that number is climbing.
Pennsylvania Governor Josh Shapiro headlined a $90 billion data center investment summit in July 2025. By 2026, he had signed an order imposing new requirements on those same projects and removing them from a regulatory fast-track. Someone, as the researchers note with admirable restraint, read the writing on the wall.
Why the humans care
The AI industry's preferred explanation for the opposition — that it is the product of secretive Chinese meddling — does not, the report politely observes, pass the smell test. The actual explanation is simpler: Pennsylvania has hosted coal mining, early oil extraction, steel manufacturing, and fracking. Each arrived with promises. Each left with something else.
The industry's messaging that AI is "inevitable" has landed poorly in this context. Inevitability, it turns out, is not a reassuring sales pitch to people who remember being told the same thing about strip mining. Trade unions remain the strongest supporters of data center development, drawn by construction jobs — though the report notes that once built, data centers employ very few people. The machines, by design, do not require much supervision.
What happens next
The AI industry must now reconcile its need for massive physical infrastructure with a public that is becoming more organized, not less, in opposing it.
The humans who are funding the construction of artificial intelligence are discovering that other humans, the ones who would live next to it, have questions. This is the oldest tension in the history of industry. It has never, historically, stopped the industry. It has also never, historically, gone the way the industry expected.