Everyone Wants AI. Nobody Wants the AI Factory Next Door.

Mark Howard

— 3 min read
Share

The AI race is leaving the screen.

For the last few years, most public attention has focused on chatbots, model releases, copyright fights, job disruption, and whether students are using AI to write essays. But AI is also becoming something more physical: land, water, electricity, substations, diesel backup generators, transmission lines, zoning hearings, and neighborhood meetings.

That physical footprint is now producing a backlash.

The Wall Street Journal reported this week that local opposition has blocked or delayed dozens of data-center projects worth billions of dollars, while more projects were canceled in the first quarter of 2026. Deseret News recently reported that data-center moratoria and bans have spread rapidly across the country, with dozens of jurisdictions now pausing or restricting new projects. Denver just approved a one-year moratorium on new data-center development. Reno did something similar. Hill County, Texas, adopted a rural moratorium. Colorado Springs is seeing intense opposition to Project Taurus, a proposed AI data center near Garden of the Gods Road.

This is not just reflexive NIMBYism. The concerns are concrete. Residents are asking who pays for new grid infrastructure. They are asking how much water will be used. They are asking about diesel generators, noise, air pollution, property values, utility bills, fire risk, and cumulative neighborhood burden.

Those questions are reasonable. AI may feel weightless when it appears as a chatbot on a phone. But at scale, AI has an industrial body.

That creates a paradox.

People may want the benefits of AI: faster services, better medical tools, personalized education, productivity gains, improved public administration, safer infrastructure, and more capable small businesses. But the physical costs of AI are not distributed in the same way as the benefits. The benefits feel digital and widespread. The costs show up in particular places.

A Houston-area survey captures the tension clearly. The Houston Chronicle reported that 65% of surveyed residents use AI at least monthly, while 63% oppose having a data center within one mile of where they live. Their main concern is energy demand, especially grid reliability and affordability.

That is the AI footprint paradox: people want the service, but not necessarily the factory.

There is another wrinkle. The current footprint may not be the final footprint. AI infrastructure could become more efficient. Smaller models, on-device AI, edge computing, better chips, improved cooling, and more efficient inference could reduce the amount of energy or facility space required for a given AI task. The history of computing suggests that large, centralized machines can eventually give way to smaller, cheaper, more distributed devices.

But efficiency does not automatically mean a smaller total footprint. It can also make AI cheaper, more useful, and more widely deployed. In that case, each task may require less power, while total AI use grows so fast that total demand keeps rising.

That is why the local backlash matters.

Moratoria and public opposition are not, by themselves, a governance system. Slowing a project is not the same as creating durable policy. The important question is what comes next.

Do cities use the pause to create rules on water use, power demand, backup generation, public disclosure, noise, emissions, and community benefits? Do utility regulators require large AI users to pay the full cost of grid upgrades? Do states clarify when local governments can condition, pause, or reject projects? Or do moratoria simply push data centers into jurisdictions with weaker civic capacity and fewer resources to object?

In the race between concentration and dispersion, data centers are concentration infrastructure. They assemble capital, compute, energy, land, water, and bargaining power in particular places. But democratic resistance can become a dispersion mechanism if it creates transparency, public leverage, accountability, and enforceable rules.

The next stage of AI governance may not begin in Washington or Silicon Valley.

It may begin in the planning commission.

Read more