Why Did 40 of 41 US Towns Just Vote to Limit Data Centers? AI's Body Problem, Explained

The AI industry spent the first week of November arguing about valuations, and while it argued, communities across the United States quietly held 41 little elections about whether the machine could move in next door. The machine went 1 for 41. In my two decades in the technical workforce, since 2006, I have never seen a technology this popular on the income statement and this unpopular at the county commission, at the same time, in the same country. Everyone writes about AI's mind, whether it will think, whether it will scheme, whether it dreams of us. Almost nobody writes about AI's body: the warehouses, the substations, the diesel generators, the water intake pipes. Your chatbot feels like a cloud. It is actually a building, and on November 3 the neighbors of that building voted, and it was a damn landslide.

Key Takeaways

• On November 3, 2026, voters favored limits in 40 of 41 local ballot measures about data centers, across 10 states. The wins included full bans, partial bans, moratoriums, voter-approval requirements and financial-guarantee rules. The industry's lone win was in Harwood, North Dakota.
• The backlash built fast, and it is measurable. Heatmap Pro polling found opposition to a nearby data center at 42% in August 2025, 51% by February 2026, about 70% by May, and 75% by mid-August 2026, before easing slightly to 71% in October. By spring 2026, 53% blamed data centers for rising electricity costs, up from 28% the previous August.
• The voters are not imagining the bill. PJM's regional capacity price went from $28.92 to $329.17 per megawatt-day in two delivery years, and the grid's own market monitor found that data center load accounted for 63.5% of one auction's revenues, about $9.33 billion, and about $6.50 billion of a later $16.41 billion auction.
• The jobs math reads worse out loud: thousands of construction jobs for a year or two, then a permanent staff of "only several dozen", often with generous tax abatements attached.
• This did not follow the usual national script; it went local. The same season brought local moratoriums and a proposed federal bill that would require operators to cover the added electricity and water costs they create.
• The lesson for everyone building with AI, my company included: the benefits of AI concentrate, its costs land locally, and any AI strategy that does not share gains with the people holding the costs will eventually meet a ballot box.

Disclosure: Agentic Trainers is paid to implement AI agents for businesses and to supply the human supervisors who check their work. We sit on the adoption side of this ledger. Weigh my opinion accordingly, and read the sources yourself.

Why Should You Listen to Me?

I started in 2006 as an outsourced call center agent in technical services, so the gap between a clean story on a slide and a messy bill in a real place is where my career began. Virtually all of my career since then has been on the technical side of outsourcing, including scaling strategy and the human-capital side of scaling for tech firms, later AI firms, and some semiconductor firms. I joined SuperStaff as CEO in 2019. SuperStaff was founded in 2009 and is now in its 18th year of business. Today I am CEO of SuperStaff and Managing Director of Agentic Trainers, the AI division of SuperStaff, where we build AI agent workforces for businesses and staff the human-in-the-loop teams that supervise them. I am not a utility regulator or a zoning lawyer. I am an operator watching AI's physical footprint lose friends faster than its software is gaining them.

What actually happened on November 3?

Voters favored limiting data center development in 40 of 41 local ballot measures across 10 states, a sweep of local elections about AI's physical footprint. The detail, from Ballotpedia's November 6, 2026 tally, is worth reading slowly, because this was not one angry town. It was a menu of different communities choosing different tools, all pointed the same way:

Ohio accounted for three of the five outright bans approved that night, with California and Florida adding one each. Voters in Kansas, Maine and Ohio approved twelve partial bans, some with only narrow exceptions and others targeting large facilities. Three six-month moratoriums passed in Maine.

Grove City, Ohio now requires voter approval for developments, including data center projects, using more than 50 acres, 20 megawatts of electricity or 500,000 gallons of water a day. Read those thresholds again: a town wrote the spec sheet of a hyperscale data center into its local rules as the definition of "ask us first."

Perry, Ohio now requires hyperscale developers seeking certain tax incentives to post financial guarantees to the community. Janesville, Wisconsin requires voter approval for a proposed data center at a GM/JATCO site if the cost exceeds $450 million.

Thirteen measures that would have favored development, rezonings, land-use additions, even non-binding statements of support, all failed. Voters would not even endorse data centers as a vibe.

Ohio alone accounted for 21 of the 41 contests, counted from Ballotpedia's own breakdown, which tells you where the buildout pressure is. The industry's single win: Harwood, North Dakota, where voters declined to add a review requirement for certain developments, including data centers.

It is not over. In December, Clarksville, Arkansas, voters will decide two measures on data-center bond ordinances (a December 8 special election that is not, itself, a yes-or-no on whether the facility can be built), and Mandan, North Dakota, voters will decide a December 15 measure that would prohibit large-scale data centers of 25 megawatts or more.

One honest caveat before anyone extrapolates to a national mandate: these are local measures, mostly in small jurisdictions, and the places angriest about data centers are the places most likely to put them on a ballot. This is a biased sample of America. It is also exactly how political eras begin: not with a national referendum, but with forty school-board-sized earthquakes on the same night.

Why are voters saying no? Follow the polling, then follow the bill

Because in a little over a year, data centers went from an abstraction to a line item on the electric bill, and the polling tracked the bill. Heatmap Pro's own account of its tracking polls, published August 19, 2026, is the cleanest timeline we have. In August 2025, Americans were about evenly split, with 42% opposed to a nearby data center and about 43% in support. By February 2026, opposition was 51%. By May, seven in 10 were opposed, and 55% were strongly opposed, a wave TechSpot reported on June 4, 2026 as 71% opposed. By mid-August 2026, in a Heatmap Pro poll of 2,045 registered voters conducted with Embold Research, opposition had reached 75%, and strong support had collapsed from 13% the previous August to 4%. A later Heatmap Pro poll published October 26, 2026 found opposition easing slightly to 71%, a small dip from that August peak, not a new high.

A SurveyMonkey poll for Decision Desk, reported by TechRadar on September 8, 2026, landed at 69% opposed, with 45% strongly opposed. When several pollsters using different methods find the same cliff, it is not noise. Gallup's March 2026 survey, published May 13, found 71% opposed, including 48% strongly opposed. In a follow-up April web survey of opponents, half mentioned excessive use of resources, including 18% each who specifically mentioned water and energy.

And a sharp number in the mid-year dataset: TechSpot, reporting Heatmap Pro on June 4, 2026, wrote that 53% of Americans blamed data centers for rising electricity costs, up from 28% nine months earlier. The industry's instinct is to call that a messaging problem. Let us check the meter before we blame the messaging.

Are the voters wrong about their electric bills?

No. The grid's own referee did the math. In PJM, the wholesale market serving 65 million people in 13 states and the District of Columbia, the regional capacity price rose from $28.92 per megawatt-day for the 2024/25 delivery year to $329.17 for 2026/27, more than eleven times as high in two delivery years. The second of those prices was the auction cap, and it cleared across the region.

PJM's independent market monitor, Monitoring Analytics, whose job is attribution, estimated that existing and forecast data center load increased revenues in the 2025/2026 capacity auction by $9.33 billion. Without that load, the monitor found, auction revenues would have been 63.5% lower. In the December 2025 auction, for the 2027/2028 delivery year, the same monitor found that existing and forecast data center load increased auction revenues by $6.50 billion, or about 40% of the $16.41 billion in total auction costs. In Washington, D.C., the Office of the People's Counsel warned in May 2025 that average Pepco residential bills would rise by about $20 a month beginning in June 2025, with roughly half of that increase driven by the capacity-price spike.

The fair caveats, stated plainly because this post does not do one-sided arithmetic: capacity auctions are one component of a bill, the monitor's figures are sensitivity estimates, and demand growth from electrification and ordinary economic activity shares the blame. But here is the part that should bother the industry's strategists: the voters' story ("the server farms are raising my bill") and the market monitor's story (63.5% of one auction's revenues, billions in costs) now rhyme. When the layman's grievance and the regulator's spreadsheet agree, you do not have a messaging problem. You have a math problem wearing a messaging problem's clothes.

What about the jobs? Read the fine print out loud

Thousands of construction jobs for a year or two, then "only several dozen" permanent staff, which is the quiet part every ribbon-cutting skips. That phrase is not mine; it is the data center industry's own trade press, in an August 28, 2025 Data Center Knowledge article, describing a typical facility's permanent headcount. A large facility might keep thousands of construction workers busy for a year or two. Once it is running, "the total data center staff number is only several dozen." Dedicated, single-company facilities generally create fewer permanent jobs than shared colocation sites, and critics point at the generous tax abatements that often ride along. A warehouse of GPUs is the opposite of a factory: enormous capital, almost no payroll. The acreage of a thousand jobs, staffed like a gas station.

Now hold that against the other side of the AI economy, the side I see every day. The U.S. Chamber of Commerce's 2026 Empowering Small Business report, based on a July 23 to August 6, 2026 survey of 3,732 small businesses, finds that 47% of AI-using small businesses say the technology is enabling job creation, while only 6% report headcount reductions, and that AI users are at least 7 points more likely than non-users to report growth in sales, profit and workforce. Put the two numbers in the same sentence and the whole problem snaps into focus. The job gains of AI show up diffusely, inside tens of thousands of adopting businesses. The costs show up concentrated: one town's substation, one county's water table, one zip code's electric bill. The benefits are everywhere and invisible; the costs are local and have an address. That is a classic case of concentrated costs and diffuse benefits, and on November 3 it ended that way 40 times.

Did the backlash follow the usual national lines? No. It went local.

The revolt is local, and it cuts across the usual national lines, which is exactly why the industry should take it seriously. The same season that national figures were cheerleading the buildout, local officials backed moratoriums, and a bill introduced in the U.S. Senate in 2026 would require data-center operators to cover the added electricity and water costs they create. Labor unions have generally stayed more favorable, mainly because of the construction jobs and wages tied to the facilities.

This is the detail the industry keeps missing while it watches Washington: the backlash did not go viral, it went municipal. National policy can be lobbied, but there is no lobbying firm on earth staffed for 3,000 county commissions, and the zoning power lives there. You cannot A/B test your way past a planning board, and you cannot patch a substation with a press release.

What does this mean for the AI industry?

The industry has been arguing about whether the money is a bubble while its physical footprint lost its social license, and the second problem does not care how the first one resolves. Every hyperscaler roadmap assumes sites, power and water on a schedule. Forty of forty-one is what it looks like when the schedule meets the neighbors. My read, labeled as my read: the industry is about to relearn a lesson that every extractive and industrial sector before it learned expensively, from pipelines to wind farms. Infrastructure is built on consent, consent is local, and consent is purchased with shared benefit, not with press releases about national AI leadership. That cost-internalization idea, a requirement that operators cover the electricity and water costs they add, is the shape of the future whether or not any one bill passes: internalize the costs you create, or watch them get internalized for you, ballot by ballot. The companies that get ahead of this will do boring, unglamorous things: pay for their own grid upgrades, guarantee ratepayer neutrality in writing, put real money behind local hiring and training, and accept community oversight as a cost of doing business rather than an insult to innovation. The ones that answer with slogans will keep going 1 for 41.

What does this mean for your business?

Your AI strategy lives on the good side of this ledger, and your electric bill lives on the bad side, so plan for both. Three practical takeaways for the operators and owners who read this blog:

1. Budget for power inflation. If you operate in PJM territory or anywhere the buildout is concentrated, the capacity-price pass-through is already in your cost base and compounding. Put an electricity line in your AI business case the way our posts on what an AI agent really costs and the missing line item in AI ROI tell you to put supervision there: honestly, before it surprises you.

1. Notice which side of this ledger you are on. A 40-person firm using supervised agents to grow is the sympathetic face of AI: the Chamber's data says you are the one creating the jobs. The industry's social license problem is not automatically yours, unless the backlash curdles into general anti-AI sentiment, which is precisely what the polling trend threatens. The whole economy of AI adopters has a stake in the hyperscalers behaving like good neighbors.

1. Adoption, not infrastructure, is where the gains are shared. The communities voted against AI's body, not its work: the same public telling pollsters the data center raised their bill is full of small businesses whose AI use correlates with growth and hiring. The way AI wins back the public is not a better ad campaign for server farms. It is millions of visible, local wins: the clinic that got faster, the broker that grew, the firm that promoted its people to supervise the machines instead of firing them, which, as the layoff-regret data shows, is also just better business.

FAQ: the data center backlash, answered straight

Why are communities rejecting data centers?

The most-cited reasons in Gallup's survey of opponents were excessive use of resources, especially water and energy. By spring 2026, 53% of Americans blamed data centers for rising electricity costs, up from 28% in August 2025. Typical facilities also create few permanent jobs, often a few dozen, while frequently receiving substantial tax abatements, which weakens the local economic case.

What happened in the November 2026 data center votes?

Across 41 local ballot measures in 10 states on November 3, 2026, the side favoring limits on data center development won 40, including full bans in California, Florida and Ohio, partial bans, moratoriums in Maine, and new voter-approval and financial-guarantee requirements. The industry's only win was in Harwood, North Dakota. Ohio alone accounted for 21 measures.

Do data centers really raise electricity bills?

In the PJM market, the regional capacity price rose from $28.92 to $329.17 per megawatt-day between the 2024/25 and 2026/27 delivery years. PJM's independent market monitor estimated that data center load accounted for 63.5% of revenues in the 2025/2026 auction, about $9.33 billion, and about $6.50 billion of a later $16.41 billion auction. Capacity costs are one component of bills and the monitor's figures are estimates, but the direction is documented.

How many jobs does a data center create?

Construction can employ thousands for a year or two, but a typical facility's permanent staff is only several dozen, mainly technicians, with single-tenant facilities creating fewer than shared ones. Indirect effects exist but are harder to measure.

Is the data center backlash anti-AI?

Not yet. The votes targeted AI's infrastructure footprint (power, water, land, bills), not AI use itself, and the backlash has been local, showing up in very different communities. A proposed federal bill would require operators to cover the added electricity and water costs they create. Whether the backlash hardens into general anti-AI sentiment depends partly on whether AI's gains become visible in local businesses and paychecks.

The part where I tell you what I do

Short plate, quickly passed. I run Agentic Trainers, the AI division of SuperStaff, and we sit on the adoption side of this ledger: training AI agents on real workflows and keeping trained humans supervising them, inside businesses whose growth is the only version of AI the public has ever actually voted for. If you want your AI story to be the kind a town hall applauds, measurable work, kept jobs, people promoted to run the machines, book a discovery call with us. If you are still choosing where AI fits, start with our AI agents for small business guide. The server farms can fight their own zoning wars. Your fight is making AI pay off visibly enough that nobody wants to vote against you.

---

Disclaimer: This article is commentary and opinion based on public sources available as of November 10, 2026. It is not legal, financial or political advice. Auction results, poll figures and pending local measures can be revised; read the sources and local notices themselves. The "body problem" framing, the benefits-diffuse-costs-concentrated analysis and the business takeaways are my own. Agentic Trainers is paid to implement and supervise AI agents, a conflict of interest I have tried to disclose.

Written by Matt Narciso with AI-assisted research and drafting. Every fact and source was reviewed by a human editor.

Sources

1. Voters favored limiting data center development in 40 of 41 local ballot measures on Nov. 3, Ballotpedia News (November 6, 2026)
2. Exclusive: 75% of Americans Now Oppose Local Data Center Development, Robinson Meyer, Heatmap News (August 19, 2026)
3. Exclusive: Public Support for Data Centers Has Slightly Recovered, Robinson Meyer, Heatmap News (October 26, 2026)
4. 71% of Americans now oppose data centers near their homes, up from 42% just nine months ago, Rob Thubron, TechSpot (June 4, 2026)
5. Americans are turning against AI data centers as local fears over water, power costs, and land use keep growing nationwide, Efosa Udinmwen, TechRadar (September 8, 2026)
6. Americans Oppose AI Data Centers in Their Area, Jeffrey M. Jones, Gallup (May 13, 2026)
7. Who is PJM?, PJM Learning Center (visited November 10, 2026)
8. 2024/2025 RPM Base Residual Auction Results, PJM
9. 2026/2027 Base Residual Auction Report, PJM
10. Analysis of the 2025/2026 RPM Base Residual Auction, Part G, The Independent Market Monitor for PJM / Monitoring Analytics (June 3, 2025)
11. Analysis of the 2027/2028 RPM Base Residual Auction, Part A, The Independent Market Monitor for PJM / Monitoring Analytics (January 5, 2026)
12. OPC Alerts DC Residents to Upcoming Significant Electricity Bill Increase Due to Unprecedented Wholesale Market Prices, Office of the People's Counsel for the District of Columbia (May 22, 2025)
13. How Many Jobs Do Data Centers Create? It Depends, Christopher Tozzi, Data Center Knowledge (August 28, 2025)
14. Empowering Small Business: The Impact of Technology on U.S. Small Business, U.S. Chamber of Commerce (October 2, 2026)
15. Clarksville voters to decide on data center bond ordinances in Dec. 8 special election, Taylor Howard, 5NEWS (September 28, 2026)
16. Special Election on Data Centers to be held Dec. 15, City of Mandan (October 29, 2026)

---