In Sand Springs, Oklahoma, a city of twenty thousand people outside Tulsa, the entire City Council is facing a recall. Their offense: voting to rezone 827 acres of farmland so Google could build an AI data center on it.
A rancher whose fence line sits three hundred feet from the proposed site says his real estate agent told him to expect a seven-figure loss in property values. His horses will hear the hum. His view will be a concrete wall with controlled lighting. Nobody asked him.
This is not an isolated incident. At least five recall campaigns have targeted local officials over data center approvals since 2022. Augusta Township, Michigan. Port Washington, Wisconsin. The list is growing.
The Pitch
The hyperscaler pitch to small towns goes like this: jobs, tax revenue, economic development. Google's Project Spring website promises the buildings will occupy less than ten percent of the land. The power grid will not be strained. The water supply will be fine.
That pitch works in a conference room. It does not work when the bulldozers show up next to your pasture.
The uncomfortable math is simple. A hyperscale data center employs very few people relative to its footprint. A facility that covers hundreds of acres and consumes enough electricity to power a small city might employ fifty full-time staff. Maybe a hundred during construction. The tax revenue is real but modest compared to what the land could generate under mixed commercial development over decades.
What the facility actually needs from the community is enormous. Power. Water for cooling. Road access. Rezoning. Annexation into city limits to reach utility connections. What it gives back is a building you cannot enter, a hum you cannot escape, and a property tax check.
The Quiet Part
Seven hundred billion dollars in hyperscaler capital expenditure is looking for somewhere to land this year. That money does not care about your horses. It does not care about your property values. It does not care about the agricultural zoning that kept your corner of Oklahoma quiet for the last fifty years.
It cares about three things: power availability, land cost, and a permitting environment that will not slow it down.
Small towns check all three boxes. Rural land is cheap. Local grids often have spare capacity because there is no industrial base to consume it. And city councils with six members are easier to navigate than metropolitan planning commissions with environmental review boards and public comment periods that stretch for months.
That is not corruption. It is efficiency. The hyperscalers are optimizing for speed, and small towns are the path of least resistance.
What Nobody Says
Here is what the industry does not talk about at community information sessions.
Data centers are permanent. They do not close. They do not relocate. Once that concrete is poured and those fiber connections are lit, that facility will operate for twenty to thirty years minimum. The economics of abandoning a built-out site are so bad that it almost never happens. Whatever impact that building has on your community, you are living with it for a generation.
Data centers also grow. The initial footprint is the minimum. That ten percent land utilization number in the brochure is Phase 1. Phase 2 is already in the capital plan. By the time the facility reaches full build-out, the neighbor who was told the buildings would be set back from the property line is looking at something considerably larger than what was presented at the town hall.
And the power draw increases. Every generation of AI hardware consumes more watts per rack. A facility designed for today's GPU density will need grid upgrades within five years to support the next generation. That spare capacity on the local grid that made the site attractive in the first place will be consumed, and the town will be competing with the data center for electricity.
The Right Way
None of this means data centers should not be built. They should. The computational demand is real and it is growing. Somebody has to house the hardware.
But there is a difference between building infrastructure in a community and building it on a community.
Building in a community means genuine transparency about scale, timeline, and impact. It means buying the adjacent rancher's land at a fair price instead of letting his property values collapse. It means long-term power agreements that include grid improvements the whole community benefits from, not just the facility. It means hiring locally for operations, not just construction.
Building on a community means quiet annexation, fast rezoning, and a PR website that says all the right things while the lawyers handle the details.
The hyperscalers can afford to be good neighbors. Seven hundred billion dollars buys a lot of goodwill if you spend even a fraction of it correctly. The fact that communities are resorting to recall petitions suggests that fraction is not being spent.
A data center that poisons its relationship with the surrounding community on day one will spend the next thirty years dealing with the consequences. Hostile local governments. Permit delays on expansions. Bad press that follows them to the next site.
The cheapest investment in any infrastructure project is the one you make in the people who have to live next to it.