Out of dozens of data center votes on Tuesday, the industry won exactly one. It happened in North Dakota, according to Stateline’s election night roundup, and nowhere else did voters side with the builders.
Everywhere else, the answer was no. Ohio towns said no. Kansas towns said no. Communities in both red and blue territory wrote bans, caps and moratoriums into local law.
I spent the day after the election going through ballot trackers, exit polls and local reporting, because this is not just a zoning story. It is the first time so many communities answered the same question in the same week: do we want the physical machinery behind AI in our backyard?
Below are the seven reasons voters gave, backed by numbers and research. I also point out where the evidence is messier than a yard sign makes it look.
Table of Contents
- What Happened on Election Night: The Numbers at a Glance
- Reason 1: Their Power Bills Are Climbing, and They Suspect Data Centers
- Reason 2: Cooling Servers Takes Millions of Gallons of Water
- Reason 3: The 24/7 Hum, Noise and Pollution Near Homes
- Reason 4: Secret Deals and NDAs Kept Residents in the Dark
- Reason 5: Big Tax Breaks for a Small Number of Permanent Jobs
- Reason 6: Farmland and Neighborhoods Rezoned for Industry
- Reason 7: Trust in Big Tech and the Politicians Backing It Collapsed
- What Happens Next for AI Data Centers?
- Frequently Asked Questions (FAQ)
- Final Thoughts
What Happened on Election Night: The Numbers at a Glance?
Counts shifted through the night as late precincts reported, and outlets did not tally in exactly the same way. The direction never changed.
- Dozens of local votes: 404 Media counted data centers explicitly on the ballot in more than 40 races across 13 states. Voters sided against them in 38 of those races, and the count was still rising.
- Hard limits: Six cities banned data centers outright, and 12 put caps on how many megawatts a facility could draw, per the same count.
- Tracker totals: The Data Centers on the Ballot tracker showed 33 restrictive measures passing or leading late Tuesday. Stateline’s update put the number at 35 by Wednesday morning.
- Both parties: Residents in Republican and Democratic communities alike pushed back against tech companies’ building plans.
- It moved votes: In an NBC News exit poll, 56 percent of voters said AI and data centers were a crucial or major factor in their vote for Congress.
- Thin baseline support: A UMass Amherst poll taken earlier this year found only 11 percent of Americans would back an AI data center in their own community.
Not every measure did the same thing. Here is what voters actually approved:
| Type of Measure | What It Does |
|---|---|
| Outright ban | Prohibits data centers entirely (six cities passed these) |
| Temporary moratorium | Pauses new approvals while rules get written |
| Energy cap | Limits how many megawatts a facility can draw (12 places) |
| Voter approval rule | Requires a public vote before construction can begin |
| Advisory measure | Urges local officials to restrict data centers, but does not bind them |
Ohio was ground zero. NPR’s roundup of state ballot measures counted 21 Ohio measures that mentioned data centers, and candidates spent up to $40 million on ads about them this cycle.
Ohio’s campaign also showed how unusual this movement is. A group called Conserve Ohio helped neighbours in many towns draft ballot language, and one of its attorneys said people were setting party differences aside to work together. That does not happen often in a midterm year.
Also on the midterms: AI showed up in the campaign ads too, and we looked at AI-made campaign ads that carried no label before the vote.
Now to the reasons, starting with the one that hits closest to home.
Reason 1: Their Power Bills Are Climbing, and They Suspect Data Centers
Electricity prices almost never decide elections. This year they came close.
As Forbes reported days before the vote, soaring bills turned the AI buildout into a campaign issue in Ohio, Pennsylvania and Michigan. In those states, capacity prices lifted some monthly bills by as much as $18.
The wider picture is not small either. Residential prices rose 7.1 percent in 2025, more than double the rate of inflation, and topped 20 percent in some states, according to Consumer Reports’ investigation into data centers. The same report describes a Virginia homeowner whose bill jumped from about $100 to $281 in a single month.
Federal researchers expect demand to keep rising. A Lawrence Berkeley National Laboratory report put data centers at roughly 4.4 percent of US electricity use in 2023, with projections of 6.7 to 12 percent by 2028.
Now the part that rarely makes it into campaign speeches: the link between data centers and your bill is real in some regions and weak in others.
When Senator Elizabeth Warren said bills near data centers had risen as much as 267 percent, PolitiFact found the number described wholesale prices, not what households actually pay. A FactCheck.org review of the research found that most recent price hikes have other causes. It did flag the PJM grid region as a place where data centers helped push costs up through capacity auctions.
So is the fear fair? Partly. But voters did not need a spreadsheet to reach their verdict. They saw a bill climb, they saw a construction site go up, and they connected the two. That perception is now a political fact, whether or not every dollar can be traced.
Quick detour: We broke down how data center power demand reaches your electricity bill in a separate 10-fact guide, and it pairs well with this section.
Reason 2: Cooling Servers Takes Millions of Gallons of Water
Servers run hot, and keeping them cool can mean drawing enormous amounts of water from the same supply a town drinks from.
Figures reported by The Washington Times, citing the Environmental and Energy Study Institute, say a large facility can use up to 5 million gallons a day. That is roughly what a town of 10,000 to 50,000 people uses.
Numbers like that land hard in a dry county or a town with an aging water system. The question stops being abstract and becomes personal: who gets the water first?
To be fair, not every site is that thirsty. Berkeley Lab researchers found that water use depends on the cooling design, the local climate, how efficient the servers are and even where the electricity comes from. You can see that debate play out in a zoning fight in Adelanto, California, where residents argued the city should secure basic water needs before welcoming resource-hungry development.
Water problems have also reached the courtroom. According to a legal review of data center litigation, a data center subsidiary agreed in 2026 to a $20.5 million settlement over nitrate contamination of an Oregon groundwater basin. It denied that it caused the contamination.
That denial matters, and so does the settlement. Voters across the country read stories like this and drew their own conclusions about what a neighbour that size can do to a local aquifer.
Water was only one of the physical costs people kept raising at town halls. The next one is something you can hear from your own bedroom.
Reason 3: The 24/7 Hum, Noise and Pollution Near Homes
A data center never clocks out. Fans, chillers and generators run day and night, and the people next door hear every hour of it.
In Wisconsin, residents are suing Microsoft over noise at its Mount Pleasant data center. The suit blames diesel generators and cooling equipment. It also says much of the hum is infrasound that normal meters miss.
That gap is the real problem. Most local noise rules rely on dBA readings, which understate low-frequency sound. A site can pass its test and still keep a family awake.
The Environmental and Energy Study Institute reports that industrial diesel generators can reach about 105 decibels. Some facilities sit within 50 feet of homes.
Rules are catching up slowly. Virginia began requiring Tier 4 equivalent emissions standards for new data center generators on July 1, 2026. Loudoun County also admits that tonal and low-frequency noise is hard to regulate, and says more research is needed on long-term effects.
So the health science is not settled. But voters do not need a study to know what they hear through a bedroom wall.
The money behind this buildout is huge, and our breakdown of the AI bubble numbers for OpenAI, Anthropic and SpaceXAI shows why companies keep building so fast.
Reason 4: Secret Deals and NDAs Kept Residents in the Dark
For many voters, the secrecy hurt more than the building.
In Boyce, Louisiana, nondisclosure agreements shielded talks between local officials and the developer, Applied Digital. The deal used code names like “Project Lightning” and “Project Pixel.” Records requests found at least 54 NDAs signed by Louisiana elected officials across various projects.
The result is now a $3.6 billion campus called Delta Forge 1. Residents say they heard rumours for months and got no answers. At a town hall, officials admitted a third-party environmental study would not be made public, and the crowd jeered.
This is not one town’s problem. Researchers found NDAs in at least 25 of 31 data center projects they surveyed. Bills to ban them have been filed in Michigan, Oklahoma, Kentucky and Ohio, and Pennsylvania’s governor barred their use by agencies he oversees.
The defenders have a case too. Louisiana officials say companies will not negotiate openly without NDAs, and Applied Digital points to securities law. Even so, nearby West Feliciana Parish skipped NDAs on a $10 billion project and posted the key agreements online.
That showed voters that a big deal can happen in daylight.
Speed is often the excuse for skipping public input, and our look at the Trump AI race with China explains the pressure to build first.
Reason 5: Big Tax Breaks for a Small Number of Permanent Jobs
Developers arrive promising jobs and tax money. Voters checked the math.
Construction brings a burst of work, but running the building takes few people. In Louisiana, officials expect more than 1,000 construction jobs but only 200 permanent ones.
Now look at what states give up. Good Jobs First found that Georgia, Virginia and Texas each lose $1 billion or more a year to data center incentives. Another 14 states do not even disclose their losses.
The per-job numbers are sharper. One Microsoft data center in Illinois got over $38 million in sales tax exemptions and created just 20 permanent jobs, according to a state budget analysis. An earlier Good Jobs First study of 11 major projects put the average subsidy at about $1.95 million per job.
States that track returns report losing 52 to 91 cents per dollar of exemption, as one Governing magazine report shows.
Supporters argue these campuses bring billions in private investment and a base for other businesses. That is a fair point, and returns vary by state and deal. But a small town carrying the risk for a giant company is a hard sell.
The jobs promise looks even shakier when Meta, Amazon and Block cut thousands of jobs while pointing at AI, as we covered earlier.
Reason 6: Farmland and Neighborhoods Rezoned for Industry
Many fights started with a simple map change. Farmland or protected rural land was rezoned to industrial use, often over loud objections.
In Michigan, the story of Saline Township’s 21-million-square-foot project shows the pattern. Residents picketed and posted yard signs, and the project went ahead anyway.
In North Carolina, Stokes County commissioners rezoned roughly 1,800 acres of farmland by a 3-2 vote. The same margin had passed in January, but that vote was voided after a legal challenge.
In Georgia, residents sued Coweta County over rezoning about 829 acres from Rural Conservation to Industrial. The group behind the suit says it would support data centers in areas already zoned for industry.
That detail matters. Many residents are not against the technology. They are against where it lands.
Some counties are now writing limits into law. Mercer County, Kentucky, set a 1,500 acre cap on data center zoning and blocked certain farmland from being rezoned.
Reason 7: Trust in Big Tech and the Politicians Backing It Collapsed
Under every other reason sits one feeling. People no longer trust the people making the decisions.
A Fox News poll found voters now see Big Tech as a greater threat than Big Government, a 28 point reversal since 2019. A Reuters/Ipsos poll found that 77 percent of Americans worried AI would make electricity more expensive, with similar shares across both parties.
The politics cut across party lines. NPR reported that Republicans and Democrats both poured millions into ads attacking data centers. Axios obtained a Republican Senate committee memo warning AI companies that data centers had become tremendously unpopular.
Bloomberg Government reported that Trump’s strong support for the industry put him at odds with voters who increasingly oppose it. Democrats, meanwhile, have had their own tests, since the opposition is strongest where affordability is already the main worry.
Trust is hard to rebuild once lost. Secret deals, rising bills and tax breaks all point in the same direction.
Public trust also took a hit when tech leaders warned about AI job losses, which we tracked in our report on Big Tech CEOs and their AI job warnings.
What Happens Next for AI Data Centers?
The “no” votes are not the end. They are the start of a bigger fight.
The industry’s main answer so far is the Ratepayer Protection Pledge. Amazon, Google, Meta, Microsoft, OpenAI, Oracle and xAI signed it in March 2026, promising to pay for their own power. But Bloomberg Government notes the pledge has no enforcement behind it.
| What to Watch | Why It Matters |
|---|---|
| Federal action | The Ratepayer Protection Act passed the House but stalled in the Senate, with each side blaming the other |
| State tax breaks | Several states are rethinking their exemptions, though Ohio’s repeal effort has so far failed |
| Local bans and caps | Six cities banned data centers and 12 capped power use on election night |
| Advisory votes | Some measures only urge officials to act and do not bind them |
| Lawsuits | Developers and residents are both heading to court |
There is also a risk on the other side. Bans in one town can simply move projects to the next county that says yes.
If the spending slows, pressure on this whole industry changes, and our report on the IMF warning about an AI bubble explains that risk.
Frequently Asked Questions (FAQ)
How many data center measures passed on election night?
Counts differ by tracker. One tracker showed 33 restrictive measures passing or leading late Tuesday, and Stateline reached 35 by Wednesday morning. The direction never changed.
Do data centers really raise my electricity bill?
In some regions, yes. The link is strongest in the PJM grid area, and weaker elsewhere. Researchers say other factors also push prices up.
Do data centers create many jobs?
They create many construction jobs and far fewer permanent ones. Louisiana’s Boyce project, for example, is expected to bring about 200 permanent roles.
Does a local ban stop AI?
No. It stops one facility in one place. Projects can move to a friendlier county, and some measures are only advisory.
Can data centers be built responsibly?
Possibly. Open agreements, cleaner generators, closed-loop cooling and paying their own power costs all help. Voters want proof, not promises.
What can I do if one is planned near me?
Attend zoning meetings, ask to see the actual agreements, and check your state utility commission filings for rate changes.
Final Thoughts
Out of dozens of votes, the industry won one. That is hard to ignore.
The seven reasons are not random. Power bills, water, noise, secrecy, tax breaks, farmland and broken trust all feed the same message. People will accept new technology, but they want a say in where it goes and who pays for it.
The industry can still win this argument. It needs open deals, real job numbers and enforceable promises. Until then, expect more “no” votes, more lawsuits and more campaign ads.
Related reads: