The rapid construction of AI data centers has encountered resistance in local communities across the US for well over a year now. But that resistance has now reached a major milestone: More than 500 US towns and counties have enacted bans or restrictions on new data center development, according to recent analysis by The Information, with the pace of opposition accelerating this summer.
The resistance spans Republican and Democratic regions, suggesting that data center development is becoming a significant infrastructure issue independent of party affiliation.
Driving the conflict is AI’s voracious appetite for computing resources. Training and running LLMs require vast clusters of GPUs and other AI accelerators, along with high-speed networking, storage and cooling systems. Hyperscale facilities supporting these systems can consume hundreds of megawatts of electricity while placing additional strains on local water systems.
Those requirements are now colliding with citizen concerns about electricity prices, utility grid challenges, water consumption, noise and land use.
The nationwide pushback adds another challenge to an AI sector already confronting shortages of transformers, chips, construction materials and skilled labor. JPMorgan estimates that 60% of capacity targeted for 2027 has not yet launched construction, while Goldman Sachs expects roughly half of the AI capacity planned through 2028 to arrive on schedule.
Different Approaches Across the US
The trend has moved beyond individual zoning disputes. New York Governor Kathy Hochul signed an executive order pausing environmental permits for data centers consuming 50 megawatts or more for one year. In Texas, Governor Greg Abbott directed state agencies to pause approvals while officials examine the power and water requirements of proposed facilities.
Texas is particularly important to the AI infrastructure market because abundant land and energy resources have attracted major data center investment. But the scale of proposed projects is also increasing the debate over how AI facilities should be powered.
Amazon, for example, is backing plans for a natural gas power facility in Pecos County designed to support a data center campus without relying on the public grid. The proposed facility could provide up to 7.65 gigawatts of generating capacity, an example of the extraordinary power requirements emerging from hyperscale AI computing.
Other states are taking different approaches. Massachusetts and Nebraska have halted some tax incentives for data center developers, while communities in Colorado have enacted restrictions amid environmental and water concerns. In Virginia, regulators have ruled that data centers must cover transmission infrastructure built exclusively for their use, an effort to prevent costs from being shifted to residential electricity customers.
The resistance is also affecting individual projects. Nashville officials recently moved to block a proposed 23-acre data center near the Nashville Zoo at Grassmere, a development expected to cost as much as $700 million.
A similar conflict emerged near Pacific, Missouri, where developers proposed a $16 billion hyperscale campus covering approximately 500 acres. Residents raised concerns about the project’s size and its potential effect on the community, while water and wastewater issues also became part of the discussion. Pacific imposed a 12-month data center moratorium, and developer BLE Landholdings withdrew its city zoning application.
