Local opposition is delaying AI data centers and forcing developers to price community support into every project.
In the first three months of 2026, local opposition blocked or delayed at least 75 data center projects worth roughly $130 billion, according to Data Center Watch. That’s the group’s count, not a government total, and it roughly matches all of 2025. Data centers like these help run AI for companies such as Meta, Amazon, and Microsoft, but a growing number of towns don’t want them nearby.
For two years, the AI buildout was sold as a race for chips, power, and capital. Now it has a fourth input nobody priced: permission to operate. Here, consent means the permits, political approval, and local support a project needs. It’s harder to buy than silicon, and it’s getting scarcer.
The Backlash Went National
On Saturday, July 18, opponents staged 142 protests across 42 states, the first coordinated national push against data center development. HumansFirst organized the effort. A June Reuters/Ipsos poll found only 14% of Americans would support an AI data center in their community. A Gallup survey fielded in March put local opposition at 71%, with 48% strongly opposed, higher than the opposition to a nearby nuclear plant.
Developers can’t treat opposition this broad and bipartisan as noise. It’s a planning assumption. With roughly seven in ten people opposed, resistance belongs in the schedule and budget from the start.
Why Moving an AI Data Center Still Costs
Developers have options. They can move to friendlier states, use old industrial sites, or build a private power supply. Each option may solve one problem, but it can add cost, delay, or another political fight.
Developers argue that these projects bring investment, tax revenue, construction work, and upgrades to roads and utilities. Gallup found that two-thirds of supporters pointed to local economic benefits, especially jobs. That case weakens when the benefits are vague but the concerns are easy to name: water use, diesel generators, and questions about utility rates. A community-benefit plan has to be specific enough to survive a hearing.
Karis Critical spent months planning a data center in Naperville, Illinois. It scaled back the facility and cut the number of diesel generators as residents objected, but the city council still rejected the plan 6-1 in January. Within months, Karis pursued a second proposal at Plum Farms in Hoffman Estates. In July, after the planning commission recommended against rezoning, the developer withdrew its request days before the village board vote.
Months were spent across two sites, and nothing was built. Any model that assumed a clean site-selection process was wrong twice.
Resistance can also harden into policy. In San Marcos, Texas, residents fought a roughly 200-acre data center over water use and strain on the power grid. The council denied the project’s zoning requests. In June, it went further and prohibited data centers in every zoning district.
The ban is now a test of local control. A state senator plans to challenge it. The Texas Tribune reports that the next legislative session could give cities more power to impose such bans or take some away.
Residents’ concerns are straightforward. Gallup found they worry about utility bills, water and energy use, noise, pollution, and taxpayer costs. Public hearings reveal another concern: many residents feel decisions were made before they had a say. Those disputes don’t follow a developer’s construction schedule.
Underwrite Consent Like Power
Here’s the operator lesson. In an industrial project, buying the equipment is rarely the hard part. The harder part is getting permission to run that equipment where it’s needed. You can speed up an equipment order, but you usually can’t speed up a zoning fight with an unwilling county.
That delay also changes the economics. Interest costs keep running, equipment prices move, and a site tied up for months can’t be used for another project.
Money can shorten the equipment timeline. It does little to the political one. Community consent belongs in the underwriting, not the cleanup. Treat it as a late-stage permit checkbox, and you’ve mispriced the project from day one.
Put local opposition, community benefits, and ratepayer protections into the financial model from the first investment meeting. Show who pays for power-grid upgrades, how much water the cooling system needs in a dry year, what noise reaches nearby homes, and which local benefits are guaranteed. Then add time for zoning reviews, appeals, and public hearings instead of using one best-case date.
AI infrastructure still depends on chips and capital, but local trust now sets the schedule. Before approving a site, ask whether residents can see a real benefit and whether ratepayers are protected. Then ask whether the project can survive organized opposition.
Answer those questions before you lock in the land and power. Chips can be ordered and capital can often be raised faster than local support can be earned. Permission has its own clock. The next scarce input in AI infrastructure may not be compute; it may be permission.
