Americans have turned against the buildings that run the cloud. The Atlantic published “Americans Really Don’t Like Data Centers” on November 4, 2026, and the numbers it gathers are blunt: more than 70% of Americans oppose data-center construction in their area, and more than 80% say technology is moving too quickly. For anyone selling, building or renting cloud capacity, data center opposition is no longer a fringe zoning fight. It is now a measurable national mood.
The timing matters. The piece landed the day after the midterms. Our earlier look at how Big Tech eyed a $750B gain after the Democratic wins covered the investor side. This article covers the other side of that trade, the voters who live next to the servers and have lost patience with them.
Below, we line up every poll we can verify, explain why the answers vary by wording and distance, and map what the backlash means for hyperscalers, colocation firms, chip suppliers and the developers who depend on their regions. We also flag what the public record does not yet confirm. Where a detail is missing, we say so instead of filling the gap.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
The Numbers Behind Data Center Opposition
No single poll tells the story. Five separate surveys, run by different organizations in different months, all point the same direction. Gallup found 71% opposed to data centers in their area. A Heatmap Pro poll put local opposition to a new facility at 75%. The Atlantic’s own roundup cites more than 70%. Reuters and Ipsos asked in June and found opposition ahead of support by 45% to 16%. Marquette Law School asked about costs and benefits in May and got a 71% to 29% verdict against the industry.
| Source | When | Finding |
|---|---|---|
| The Atlantic (roundup) | Nov. 4, 2026 | More than 70% oppose construction in their area, more than 80% say tech moves too quickly |
| Gallup | 2026 | 71% oppose data centers in their area |
| Heatmap Pro | 2026 | 75% would oppose a new data center near where they live |
| Reuters-Ipsos | June 2026 | Opposition leads 45% to 16%, widening to 59% to 10% within 10 miles |
| Reuters-Ipsos | June 2026 | 61% call data centers necessary, 77% want them in remote areas |
| Marquette Law School | May 2026 | Costs outweigh benefits, 71% to 29% |
| Pew Research Center | Sept. 22, 2026 | 54% say bad for the environment, 50% for home energy costs, 49% for nearby quality of life |
| Data Center Watch | H1 2026 | At least 120 projects worth $198 billion blocked or delayed |
Note what the table does not contain. The public record we reviewed does not establish the sample sizes, geographic coverage or exact question wording for every poll. Comparing a 71% from Gallup with a 75% from Heatmap Pro as if they measured the same thing would overstate precision. The safer reading is the band: roughly seven in ten Americans, give or take a few points, say no to a data center near home.
Why Technology Anxiety Feeds the Backlash
The 80% figure deserves as much attention as the 70% one. The Atlantic reports that more than four in five Americans say technology is moving too quickly. That is a statement about pace, not about any single product. It suggests voters see data centers as the physical face of a change they never agreed to.
A chatbot lives on a phone screen and feels weightless. A data center sits at the edge of town with a substation, a fence and a construction schedule. When people want to push back on artificial intelligence, the building gives them a place to do it. A planning board meeting is a far easier target than a model release.
A share above 80% is too large to come from one party’s voters alone. A resident can enjoy the products and still resent being the neighbor of the infrastructure behind them, and the Reuters-Ipsos split between “necessary” and “not here” fits that pattern.
Reuters-Ipsos: Distance Changes the Answer
The June Reuters-Ipsos poll is the most useful dataset for operators because it tests geography. Across the general question, opposition to new data centers led support by 45% to 16%. Narrow the question to facilities within 10 miles of the respondent’s home and the gap blows open to 59% to 10%. Closer buildings produce angrier answers, and almost nobody on the fence.
The same survey exposes a contradiction the industry can use and a trap it should avoid. A 61% majority said data centers are necessary. Yet 77% wanted them built in remote areas away from communities. Voters accept the premise of the buildings and reject the address. That is a classic not-in-my-backyard split, and it is the sharpest argument for siting policy over persuasion campaigns.
The trap sits in the 61%. A company that reads “necessary” as consent will misjudge the room. Necessity is a statement about the national economy. Opposition is a statement about a specific parcel, a specific power bill and a specific commute. Executives who confuse the two tend to learn the difference at a public hearing.
Marquette and Pew: Costs Versus Benefits
Marquette Law School’s May poll framed the issue as a ledger. Respondents said the costs of data centers outweighed the benefits by 71% to 29%. That is a lopsided result for a technology that pays property taxes and promises construction work.
Pew Research Center’s survey published on September 22 broke the costs into three buckets. It found 54% of Americans consider data centers bad for the environment, 50% say bad for home energy costs, and 49% say bad for the quality of life of people living nearby. Energy bills are the item that hits a household every month, so it is the cost most likely to turn a shrug into a vote.
Pew also published an earlier short read on March 12 that asked how Americans view data centers’ impact across several areas. Its headline numbers read lower, and they include jobs and tax revenue, where concern was small. We quote them as published, because the contrast shows where opinion is soft and where it is hard:
- Pew Research Center, in its March 12 short read: “39% of U.S. adults said data centers were mostly bad for the environment.”
- Pew Research Center, same release: “38% of U.S. adults said data centers were mostly bad for home energy costs.”
- Pew Research Center, same release: “30% of U.S. adults said data centers were mostly bad for the quality of life of people living nearby.”
- Pew Research Center, same release: “15% of U.S. adults said data centers were mostly bad for local jobs.”
- Pew Research Center, same release: “12% of U.S. adults said data centers were mostly bad for local tax revenue.”
| Impact area | Pew, March 12 (“mostly bad”) | Pew, Sept. 22 (“bad”) |
|---|---|---|
| Environment | 39% | 54% |
| Home energy costs | 38% | 50% |
| Quality of life nearby | 30% | 49% |
| Local jobs | 15% | Not in our fact sheet |
| Local tax revenue | 12% | Not in our fact sheet |
We cannot say from the public record whether the two Pew releases used the same question wording or answer options, so we do not call the gap between them a trend. What the table does show is clear enough. Jobs and taxes, the two benefits developers lead with in every pitch, draw the least concern. The environment, energy costs and neighborhood quality draw the most. The industry’s selling points and the public’s worries sit on different lines of the ledger.
Data Center Watch: $198 Billion Blocked or Delayed
Opinion polls measure mood. Project trackers measure consequences. Data Center Watch counted at least 120 data-center projects, worth $198 billion, blocked or delayed during the first six months of 2026. That is a six-month figure, not an annual one, and it covers delays as well as outright cancellations.
The distinction between blocked and delayed matters for investors. A delayed project still might get built, but it ties up land, power-interconnection requests and financing in the meantime. In a market where chips and transformers already run short, a six-month slip can cost more than a rejected permit that frees the capital for a friendlier county. Our report on a Texas data center pause that sent Tesla and SpaceX shares down more than 4% showed how quickly a local policy move can ripple into equity prices.
A caution on scale. CNN Business ran a piece in August with the headline “Americans are rallying against data centers. Surprisingly few are actually getting built”. The headline points to a tension we return to below: loud opposition and a thin pipeline of finished buildings can coexist, and each may be feeding the other.
The Midterm Connection: Voters Take the Argument to the Ballot
Several outlets framed Election Night around the same theme. The Verge ran a story titled “Data centers lost the midterms.” Engadget’s headline read “Americans Overwhelmingly Voted Against Data Centers In The Midterms.” And CNBC’s November 4 report was titled “Voters rebuff data centers in every ballot measure except one.” We have not independently tallied the measures, so we pass along those headlines as the outlets’ own framing and leave the vote counts to their reporting.
What the headlines share is the verb. Voters rebuffed, lost, voted against. For years the industry treated data centers as a quiet economic-development win, the kind of project a county board approves in one session. That assumption has now been tested at the ballot box, and three major outlets describe the result in the same terms in nearly identical terms.
For cloud companies, the practical lesson is about sequencing. Winning a permit is no longer the end of the approval process. A project can clear staff review, survive a council vote and still face a petition, a lawsuit or a referendum. Each step adds months, and each month adds carrying cost on land and equipment that was ordered against a fixed delivery date.
Is the Backlash Real or Overstated?
Not everyone accepts the headline reading. Reason published a column in August titled “The data center backlash is bipartisan, nationwide, and wrong.” The title concedes the scale of the opposition while contesting its merit. That is a useful marker, because it shows even critics of the backlash treat it as broad and cross-partisan.
Other commentary reads the mood differently again. The University of Virginia’s National Security Institute titled a piece “The Data Center Backlash is Not About Data Centers,” and Derek Thompson’s newsletter called the moment “The Great American AI Rebellion.” Both titles point at the same idea: the buildings stand in for a larger argument about artificial intelligence, power and who pays for it. We link those pieces by title only, since we are summarizing their framing rather than their evidence.
Our own read is that both camps are partly right. The opposition is real, measured by multiple pollsters and visible in blocked projects. It is also true that polls on a hypothetical facility in your town tend to produce stronger answers than polls on a facility already built. Local resistance to other projects sometimes softens once a facility opens, but we found no data showing that pattern for data centers, and the 59% to 10% figure from Reuters-Ipsos suggests the industry should not count on it.
Market Impact for Cloud Providers and Suppliers
The money riding on this fight is enormous. Our report on the AI capex surge to $828B as the Nasdaq slid 2.9% framed the spending side. Public opposition attacks the other half of that equation: the ability to turn committed capital into energized floor space on schedule.
Three groups feel the pressure first. Hyperscalers hold the largest pipelines and face the most public hearings. Colocation and developer firms, which build on spec and lease later, carry the land risk. And equipment makers, from server vendors to electrical suppliers, depend on build schedules that a single delayed permit can scramble. A delayed $198 billion in projects does not vanish from the order book. It shifts quarters, and it shifts places.
Financing deals are adapting too. When a project’s timeline gets uncertain, lenders ask for stronger guarantees, and large tenants step in. Our coverage of Google’s credit backstop for $19B in Anthropic leases at TeraWulf is an example of balance-sheet strength substituting for site certainty. Expect more deals shaped by that logic, where a creditworthy tenant absorbs risk that local politics created.
Competitive Comparison: Siting Strategies Under Pressure
With 77% of Reuters-Ipsos respondents asking for remote locations, operators have a limited menu. The table below is our analysis, not a published ranking. It compares the common strategies against what the polling says voters want.
| Strategy | Fit with polling | Main trade-off |
|---|---|---|
| Remote, low-population sites | Strong (77% want remote siting) | Distance from users and from fiber and grid capacity |
| Established data-center hubs | Moderate (neighbors already live with them) | Power and land constraints, and fatigue in hub counties |
| Converted industrial or energy sites | Moderate to strong | Retrofit cost, and the lease or tenant structure |
| Community-benefit packages | Targets the energy-bill and quality-of-life worries | Cost, and credibility if benefits are not binding |
| Smaller, distributed facilities | Mixed (more neighbors, smaller footprint) | Higher cost per megawatt, more hearings |
Established hubs deserve a closer look. Our piece on Atlanta leading first-half 2026 data center construction and our look at Wisconsin’s AI data centers one year later, at $22B and 1,500 jobs show how builders chase markets with existing infrastructure and local buy-in. Those markets still have voters, though, and the polling suggests hub fatigue is a risk of its own.
Historical Context: From Invisible Utility to Unwelcome Neighbor
For most of the cloud era, data centers were out of sight. Companies placed them near cheap power and fiber, in counties that welcomed the tax base, and users never thought about them. The marketing helped. The word “cloud” implied something light and placeless, not a warehouse drawing electricity from the local grid.
Artificial intelligence changed the scale and the visibility. Training and serving large models demands dense clusters and large power connections, and the projects now announced are bigger than the ones residents saw a few years ago. Each announcement arrives with a headline number, and each number invites a comparison with the town’s own power bills. When a building is described in billions of dollars, neighbors reasonably ask what they get and what they pay.
The pattern is familiar from other infrastructure fights. Pipelines, power lines, wind farms and highways all moved from technocratic approvals to contested local politics once people understood who bore the costs. Data centers are following that arc faster than most, helped by social media and by a ready-made national story about AI. The industry’s problem is that it joined the debate late, after the first impressions formed.
What Developers and Cloud Customers Should Watch
If you buy cloud capacity rather than build it, the backlash reaches you through three channels. The first is regional availability. A delayed build can mean a region that launches later than the provider’s roadmap said, or that runs short of GPU capacity for longer. The second is price. Power costs and compliance requirements feed into rates, even if hyperscalers rarely pass the full burden through.
The third is concentration risk. If new capacity can only land in a handful of friendly places, more customer workloads end up in the same few regions. That raises the stakes of a single-region outage and of a single local policy change. Teams planning multi-year deployments should build region diversity into their architecture now, and should ask providers directly how siting delays affect committed capacity.
For procurement teams, a short checklist is enough to start:
- Ask your provider which of your target regions have capacity expansions that face local permitting disputes.
- Spread latency-tolerant batch and training jobs across at least two regions.
- Review reserved-capacity contracts for delivery-delay terms.
- Track local ballot measures and council agendas in the counties that host your primary region.
None of this calls for panic. Existing regions keep running, and existing contracts keep their terms. But the era of assuming that new capacity simply arrives on schedule is over.
Five Predictions for the Next 12 Months
These are our judgments, based on the polling and project data above, and not forecasts from any named analyst.
- Community-benefit agreements become standard. Expect binding commitments on power costs, water and local hiring to appear in most large proposals, because energy bills are the sharpest public worry.
- Remote and brownfield siting gains share. With 77% of Reuters-Ipsos respondents asking for remote sites, developers will favor land away from housing.
- Blocked-or-delayed tallies keep climbing. The first-half 2026 count of 120 projects and $198 billion looks like a floor, not a ceiling, while the political mood holds.
- Candidates run on data centers. The topic has crossed from zoning boards into statewide and national messaging, and the Election Night headlines make it hard to retreat.
- Tenants backstop more projects. Credit support from large customers will spread as lenders price in local-opposition risk.
What the Public Record Does Not Confirm
Several details remain open. The public sources we reviewed do not give the full methodology, sample size or question wording behind each poll. They also do not provide a complete, verified list of the companies and facilities named in The Atlantic’s article, or any on-record quotations from its pages that we can reproduce. We have therefore avoided naming specific operators or projects as targets of the backlash.
Readers should also treat headline comparisons across polls with care. A 71% from one pollster and a 75% from another answer slightly different questions, asked of different samples, in different months. The direction is unmistakable, and the exact level is not. We will update this analysis if the underlying polls publish fuller crosstabs.
The Takeaway for the Cloud Industry
The takeaway: Americans accept that data centers must exist, and reject having one next door. Gallup’s 71%, Heatmap Pro’s 75%, Reuters-Ipsos’s 59% to 10% within 10 miles and Data Center Watch’s $198 billion in stalled projects all describe the same gap between national need and local consent. The cloud industry can close it with siting discipline, enforceable benefits and honest accounting of energy and water use. It cannot close it with advertising. Follow our cloud computing coverage for updates as the next round of local votes and project filings arrives.
Frequently Asked Questions
What percentage of Americans oppose data centers in their area? Gallup found 71% opposed, Heatmap Pro found 75% would oppose a new facility near them, and The Atlantic reported more than 70% opposition in its November 4, 2026 article. The exact number varies by pollster and wording.
Do Americans think data centers are necessary? Yes, in the abstract. The June Reuters-Ipsos poll found 61% said data centers were necessary, while 77% wanted them built in remote areas away from communities.
How much does distance change opinion? A great deal. Reuters-Ipsos found opposition to new data centers ahead by 45% to 16% overall, and by 59% to 10% when the facility sat within 10 miles of the respondent’s home.
How many data-center projects has opposition stopped? Data Center Watch counted at least 120 projects worth $198 billion blocked or delayed in the first six months of 2026. The count includes delays, not only cancellations.
What are Americans’ biggest concerns about data centers? Pew Research Center’s September 22 survey found 54% consider them bad for the environment, 50% bad for home energy costs and 49% bad for nearby quality of life. Marquette Law School’s May poll found costs outweighed benefits by 71% to 29%.
Will the backlash slow AI and cloud growth? It will likely slow and relocate construction more than stop it. Delays shift capacity to later quarters and to friendlier locations. Customers should expect uneven regional availability and plan for multi-region deployments.
Is data center opposition a partisan issue? The polls we cite do not break results out by party in the sources we reviewed. Reason’s August column, whose title calls the backlash bipartisan and nationwide, argues the opposition is wrong, so even critics describe it as broad.
