AI Infrastructure’s Next Bottleneck May Be Public Acceptance

In just a few short years, data centers have moved from background infrastructure to contested industrial policy. Over the last 24 months, the rapid expansion of AI‑oriented campuses has turned what was once a largely technical siting discussion into a high‑stakes political issue across U.S. states and localities, where debates now center on power demand, water use, land intensity, tax treatment, and perceived community value.

That shift matters because the underlying demand is not discretionary. Hyperscalers are still planning enormous capital deployment for AI infrastructure, with leading firms forecast to spend roughly $710 billion¹ on data centers across North America by 2026, even as local resistance increasingly delays projects and narrows the set of politically viable sites. 

The result is a structural tension: digital infrastructure has become more economically indispensable at the same moment it has become more politically visible. The industry’s central challenge is no longer simply securing land, power, and fiber; it is securing durable social and political license to build at AI scale.

Meanwhile, as the AI‑driven data center build‑out is colliding with local politics, the backlash is still highly concentrated, unevenly grounded in facts, and absolutely manageable for operators that treat it as a political and infrastructure problem, and not a PR problem.

From Quiet Backbone to Political Flashpoint

The data center sector has entered a new phase in which community and political backlash must be treated as a first‑order infrastructure constraint, much like land, power, and connectivity have been in the past. Local opposition has evolved from episodic neighborhood resistance into a more organized and increasingly effective force expressed through moratoriums, zoning fights, utility restrictions, proposed statewide bans, and scrutiny of tax incentives.

Three converging dynamics explain why this issue now dominates headlines and boardroom risk discussions.

First, AI has changed the physical profile of the asset class. New campuses are larger, more power‑intensive, and more visible than legacy enterprise facilities, which makes their externalities harder for local officials and residents to ignore. In many jurisdictions, the question is no longer whether a data center can technically be connected to the grid, but whether the project can survive political scrutiny over electricity demand, water consumption, and land use.

Second, opposition has become more coordinated and institutionalized. According to CNN, citing nonpartisan research firm Data Watch, more than 140 local organizations across two dozen states² have organized specifically against data center projects, while separate reporting has documented at least $156 billion across 48 publicly disclosed projects³ that faced blocks or delays in 2025 because of local resistance. Michigan⁴ alone has seen a cascade of local moratoriums, with reporting in early 2026 showing at least 25 communities with active pauses while they rewrite zoning rules, alongside state legislation proposing a broader halt through April 2027.

Third, the issue has moved beyond local land‑use politics and into mainstream state and national policymaking. Maine’s legislature passed what would have been the first statewide moratorium on large‑scale data centers before Gov. Janet Mills vetoed it in April 2026, while federal proposals and state bills elsewhere have sought to pause new development or shift more infrastructure costs onto operators. Even where such measures do not pass, they alter the negotiating environment by legitimizing the premise that data center growth should be slowed, conditioned, or repriced.

The analytical conclusion is not that the sector faces an existential political ceiling, but the stronger conclusion is that data center growth is becoming more selective, more negotiated, and more dependent on governance quality than in the prior cloud build‑out cycle. Capacity will still be built because AI, cloud services, enterprise digitization, and national competitiveness all depend on it; the decisive variable is which jurisdictions can align infrastructure expansion with credible rules on energy, water, tax equity, and local benefits.

For operators, developers, and investors, the practical implication is straightforward: community backlash should be modeled as a core siting and execution risk, not as a downstream communications problem. The winning strategy is not retreat, but adaptation: better site selection, more transparent utility and water planning, stronger local benefit frameworks, and earlier political engagement before opposition consolidates.

Mapping the Backlash: Scope, Scale, and Trends

Before turning to specific drivers, it is useful to characterize the scale and structure of the backlash using the datasets now available. Recent project‑level tallies from Data Center Watch and subsequent analyses in major outlets document tens to hundreds of billions of dollars in U.S. data center investment that has been blocked, delayed, or materially constrained by community and political opposition.⁵ We are also witnessing clear geographic clustering and recurring political pathways—local moratoriums, state‑level legislation, and targeted changes in tax and utility rules—through which opposition converts into execution risk for new capacity.

Quantifying Disruption

Several independent analyses now put hard numbers on how much capital is being slowed or redirected by community and political opposition to data center projects.

The first systematic attempt to quantify the trend comes from Data Center Watch’s 2025 report, which estimates that approximately 18 billion USD of U.S. data center projects were fully blocked and a further 46 billion USD delayed over roughly the prior two years, for a total of 64 billion USD in projects materially impacted.⁶ This figure has since been echoed by trade press and mainstream coverage as an indicative order of magnitude for the emerging risk.

A second, broader compilation suggests the exposure is significantly higher for AI‑focused capacity. The New York Times, drawing again on Data Center Watch’s project‑level tracking, reports that in 2025 alone, at least 156 billion USD across 48 publicly disclosed AI data center projects encountered blocks or delays attributable to coordinated local opposition.^7,8 This framing has been repeated in derivative coverage summarizing the same underlying dataset, reinforcing the sense that the issue is not isolated to a handful of controversial sites.

Within that annual figure, the temporal clustering is striking. According to the same New York Times analysis, between April and June 2025 roughly two‑thirds of the AI‑oriented data center projects in one quarterly sample—20 projects representing approximately 98 billion USD in aggregate capital—were blocked, delayed, or materially re‑scoped due to community opposition, local permitting fights, or related political interventions. For an industry accustomed to treating entitlements as an execution detail, a quarter in which the majority of large projects face disruption is a meaningful regime change.

These numbers are directionally more important than point‑precise. The underlying methodologies differ (e.g., whether “delayed” is time‑bounded, how capital is attributed to multi‑phase campuses, how “opposition‑driven” is defined), and none captures the full universe of projects. However, taken together they establish that “permitting and politics risk” has graduated from a background nuisance to a first‑order variable in capacity planning and capital allocation, with tens to hundreds of billions of dollars of AI‑adjacent infrastructure now exposed to local veto points.

For investors and hyperscalers, this implies that traditional siting screens—power, land, fiber, incentives—need to be expanded to include quantified political risk metrics, including historic patterns of local activism, regulatory volatility, and the density of organized opposition groups.

Geographic and Political Pattern

The backlash is not evenly distributed. It clusters in specific geographies and political configurations, creating a patchwork of high‑risk and relatively stable jurisdictions.

Michigan is a prime case study. By early 2026, reporting from public broadcasters and state‑level outlets showed at least 19 municipalities with active or recently enacted moratoriums on data centers, often framed as temporary pauses to study impacts on the grid, water, and land use. These local actions sit alongside proposed statewide legislation that would impose a one‑year pause on large new data centers, underscoring how quickly local zoning debates can scale into state-wide policy.

Beyond Michigan, analysts and reporters have identified a growing number of states where lawmakers have introduced bills to pause or significantly constrain data center development. Coverage by NBC News and others highlights New York, Georgia, Maryland, Oklahoma, Vermont, Virginia, South Dakota, Washington, Colorado, Maine, and Florida among jurisdictions where legislators have sought to tighten rules, rethink tax incentives, or in some cases temporarily halt approvals.^9,10,11,12

At the local level, that legislative activity translates into a proliferating set of outright or de facto bans. Aggregators such as Data Center Watch, alongside environmental NGOs, count dozens of counties and municipalities across at least 14 states that have enacted either time‑bounded moratoriums, ongoing temporary bans repeatedly extended, or project‑specific vetoes.

Examples range from Prince George’s County pausing all data center activity to review its zoning framework, to individual townships blocking specific rezones or utility hookups.

Politically, the pattern is bipartisan but fragmented. Analyses by Harvard, NBC, and several advocacy groups converge on the idea that the coalition opposing data centers is unusual for its breadth:

• Left‑leaning actors (environmental organizations, climate‑oriented legislators, and some urban officials) tend to emphasize environmental externalities, including water use for cooling, compatibility with state climate targets, and the optics of offering tax incentives to large tech firms while basic services remain underfunded.
• Right‑leaning actors more often foreground tax policy, perceived corporate favoritism, and ratepayer risk. NBC reporting, for example, describes Republican‑backed efforts in South Dakota and Florida to claw back or condition tax exemptions and require data centers to shoulder higher shares of power and water costs, driven by concern that households and small businesses will subsidize AI campuses through higher utility bills.

Crucially, concern over grid strain and electricity prices is cross‑partisan. Both progressive critics and conservative ratepayer advocates now argue that large AI data centers could crowd out other uses or require costly grid upgrades that are not fully internalized by operators. Widely noted, residential rate payers are very concerned about rising electricity rates and skeptical that data centers deliver commensurate local economic benefits.

For operators, the implication is that this is not a simple blue‑state regulation vs. red‑state deregulation dichotomy. Democratic‑leaning jurisdictions may focus on climate and water, while Republican‑leaning jurisdictions frame their interventions as ratepayer and taxpayer protection, but the underlying structure is the same: a localized calculus in which visible impacts (land, noise, water, power) accrue to specific communities, while the benefits of AI and cloud infrastructure remain diffuse and abstract.

That asymmetry can surface almost anywhere, regardless of partisan alignment, and should therefore be treated as a core dimension of site selection and stakeholder strategy rather than as a niche regulatory issue.

Policy Landscape: From Symbolic Bills to Binding Moratoriums

Policy is where the backlash stops being rhetoric and starts shaping real projects. In the space of roughly two years, AI‑driven data centers have gone from a regulatory afterthought to a named target in federal bills, state statutes, and local ordinances, with lawmakers experimenting across a spectrum from symbolic resolutions to hard moratoriums and cost‑shifting rules. For operators and investors, the key fact is that there is no single policy front: national proposals define the narrative, statehouses write the most consequential rules on timing and cost allocation, and local governments increasingly wield veto power through zoning, water, and utility decisions.

Federal and National‑level Moves

National politics have caught up to the data center boom with surprising speed. In late March 2026, Sen. Bernie Sanders and Rep. Alexandria Ocasio‑Cortez unveiled the Artificial Intelligence Data Center Moratorium Act, a bill that would halt construction of new AI‑related data centers nationwide until Congress enacts a federal framework governing their power use, environmental impact, and social risks.

The proposal is explicit about its rationale: it links AI data centers to rising electricity bills, pollution from new generation, and broader fears about the social and economic consequences of unfettered AI deployment.

The bill is widely viewed, even by its supporters, as unlikely to pass current Congress. Leadership in both chambers has shown little appetite for a national construction freeze, and business groups have lined up against it.

But symbolically this matters: it normalizes the idea that AI data centers are an object of federal concern in their own right, not just a byproduct of tech growth, and it expands the Overton window for more targeted restrictions at the state and local level.

In parallel, other federal actors have begun to treat AI data centers as a direct lever on household energy costs. One example is the Protecting Families from AI Data Center Energy Costs Act, introduced in late 2025 by Reps. Greg Landsman and Don Beyer, which would require the Federal Energy Regulatory Commission (FERC) to convene utilities, regulators, and operators to develop safeguards so that increased costs associated with AI data centers are not passed through to residential ratepayers.

More recently, the White House has leaned on major tech firms to voluntarily commit to absorbing the incremental grid costs tied to AI build‑outs, framing this as a “Ratepayer Protection Pledge” designed to keep electricity prices down even as data center demand soars.

None of these moves yet amounts to a federal regulatory regime for AI data centers. But together they mark a shift in posture: regulators are not ignoring the infrastructure footprint of AI. Federal proposals and pledges establish narratives: that data centers can and should be constrained if they drive up bills or emissions, and state lawmakers and local activists are already using this ideology to justify more aggressive measures closer to the ground.

State‑Level Legislation and Vetoes

If federal efforts are mostly signaling for now, state politics are where those ideas turn into binding constraints. The recent Maine fight is the clearest example. Earlier this year, lawmakers in Augusta passed LD 307, a bill that would have imposed the country’s first statewide moratorium on large‑scale data centers—defined as facilities above 20 MW—through November 1, 2027, while a special commission studied impacts on energy, water, and land use. Governor Janet Mills vetoed the bill in April, warning that a blanket freeze risked sending a hostile signal not just to data center developers but to the wider ecosystem of advanced manufacturing and technology investment the state is trying to attract. The episode underscores a central tension: even where legislatures are willing to legislate moratoriums, governors may still prioritize competitiveness and fiscal growth over precaution.

Elsewhere, states are experimenting with more targeted mechanisms. In New Jersey, lawmakers have advanced S731/A796, a bill that would require “large load data centers” drawing 100 MW or more to commit to taking at least 85 percent of the service they request for a period of at least 10 years. The measure is explicitly framed as a way to protect ratepayers from footing the bill for new transmission and distribution infrastructure built to serve AI campuses that could downsize or walk away if market conditions change.

Other states are moving along the same spectrum. Bills in Vermont, Georgia, Oklahoma, Maryland, and New York have proposed temporary pauses, stricter environmental reviews, or revisions to incentive programs for very large facilities, with lawmakers in both parties citing a desire to get “ahead of” AI‑driven power demand before it reshapes state grids. It is still early days for many of these proposals, and not all will pass, but the range of tools under consideration—bans, time‑limited moratoriums, and cost‑shifting rules—illustrates how quickly the policy conversation has moved beyond simple tax abatements.

For operators and investors, state‑level risk is now multidimensional. A project can be killed outright by a moratorium, constrained by environmental or land‑use triggers, or rendered less attractive by utility rules that push a larger share of grid upgrade costs onto the balance sheet. The Maine veto shows that executive branches can still act as a counterweight when legislative politics tilt toward restriction; the New Jersey bills show how easily the financial assumptions under a project can change even without an outright ban.

Local Bans and Novel Constraints

Beneath the federal and state theater, local governments are developing their own playbook, and in many cases, it is more immediately consequential. One of the most important moves has come from utilities themselves. In April 2026, the Ypsilanti Community Utilities Authority (YCUA) in Michigan approved a 12‑month moratorium on providing new water and sewage connections to data centers and high‑performance computing facilities, effectively blocking new projects regardless of how they fare in zoning. The decision followed an earlier 12‑month zoning moratorium by the city, turning what began as a planning pause into a utilities‑side veto that developers cannot easily route around.

Similar tactics are emerging elsewhere. Across multiple states, counties and municipalities have enacted “temporary” bans on data centers that are nominally tied to impact studies but are written in ways that allow for extension or selective application to controversial projects. Some localities have used zoning overlays to exclude data centers from entire corridors; others have quietly tightened noise, height, or setback rules to make standard hyperscale designs non‑viable without a bespoke variance.

The net effect is that veto power is migrating beyond the planning commission. Water authorities, sewer boards, and public utility commissions can now function as parallel chokepoints, either by denying service outright or by conditioning it on requirements that materially alter the economics of a project.

For developers, the implication is straightforward: a permitting strategy that focuses only on land‑use approvals is incomplete. To de‑risk AI‑era builds, operators need an integrated view of local authority across zoning, water, sewer, and electric service—and a political strategy that treats each of those entities as a potential veto point, not an afterthought.

Drivers of Backlash: Stated Concerns and What Communities Think They Are Solving

Backlash only matters to operators if it is coherent enough to drive decisions. On the ground, it usually is. When residents show up to planning hearings or pressure state legislators, they tend to frame their objections in concrete, material terms—bills, water, taxes, noise—even if the underlying motivations are broader anxieties about AI, corporate power, or local identity.

The most useful way to parse this for analysis is to separate what people say they are worried about from the deeper political economy that shapes which projects they fight, and how hard.

Across reporting, hearing transcripts, and campaign materials, the same clusters of stated concerns show up repeatedly.

• Electricity prices and grid strain: Residents argue that large AI‑oriented campuses will drive up local electricity prices and consume grid capacity that could otherwise support housing, small business, or industrial electrification. They worry that utilities will overbuild wires and substations for a handful of hyperscale clients and then socialize the cost across everyone else’s bills. For many local officials, the central question is not whether the grid can technically serve a data center, but whether doing so crowds out other priorities or forces politically painful rate increases.
• Water use and environmental impact: In regions that rely on evaporative cooling and already‑stressed aquifers or rivers, opposition groups focus on water withdrawals and thermal discharge. They argue that round‑the‑clock cooling demand will compete with agriculture or residential supply, and that adding high‑load data centers to fossil‑heavy grids undermines state climate targets. For climate‑focused actors, the data center is often framed as the physical embodiment of AI’s carbon footprint, even when operators are signing clean‑power contracts.
• Tax fairness and public subsidies: Another recurring theme is fairness. Opponents highlight multi‑year property‑tax abatements, sales‑tax exemptions on equipment, and bespoke infrastructure deals, and contrast them with comparatively modest permanent headcounts. The critique is simple: local governments are “giving away the store” to wealthy technology companies in exchange for a few dozen high‑paid jobs, while schools, roads, and basic services remain underfunded. In some states this has become a core talking point for both progressive populists and fiscal conservatives.
• Noise, land use, and quality of life: At the neighborhood level, objections become more visceral. Residents complain about continuous mechanical noise, diesel backup testing, 24/7 lighting, and heavy construction traffic. They object to farmland or open space being converted into fenced‑off industrial campuses, and to the visual intrusion of new transmission lines and substations. In semi‑rural and exurban communities, that industrialization narrative is often as powerful as any spreadsheet about rates or jobs.

Taken at face value, none of these concerns is inherently unreasonable. When a project proposes to consume hundreds of megawatts, millions of gallons of water per day, and tens or hundreds of acres of land, questions about price impacts, resource allocation, and neighborhood change are exactly what any competent local government should be asking. And yet, they don't have the scientific skillset to understand our industry talking points.

Underlying Political Economy

Beneath those stated concerns, though, are deeper structural forces that explain why opposition is suddenly scaling and why AI data centers, in particular, have become lightning rods.

First, the build‑out has outrun public understanding. In many regions, data centers mushroomed from obscure facilities tucked into office parks to highly visible megaprojects with their own transmission lines and water infrastructure, without any intervening period in which residents were brought along on what this infrastructure is for. People see forests cleared, trucks on rural roads, and new lines across farmland; what they do not see is a tangible, local benefit from the additional compute. That asymmetry—visible local costs, abstract global benefits—primes communities to see new proposals as extraction rather than investment.

Second, the backlash is riding on broader unease about AI and automation. Polling consistently shows that U.S. residents are more skeptical of AI’s net impact on jobs, democracy, and social cohesion than many of their peers abroad. Hop on the Facebook group "Say No to Data Centers" if you want to get a more firm understanding about the greater fears. Skepticism does not always surface in the formal language of zoning, but it sits just below the surface. For activists, opposing the physical infrastructure of AI is a concrete way to express diffuse fears about everything from deepfakes to job displacement. The data center becomes a proxy: if you cannot vote directly on “AI,” you can vote on its housing and rental agreements.

Third, local fights are increasingly vehicles for general grievances about governance and corporate power. In case after case, opposition is intertwined with pre‑existing distrust of utility regulators, anger over prior broken promises on economic development, or resentment that decisions are being made in back‑rooms between state agencies and global companies. Communities that feel they have lost control over housing, traffic, or industrial siting see data centers as one more decision imposed from above, and they respond accordingly.

For all of us tech-enthusiasts, builders, and operators, the implication is not that communities are irrational or anti‑technology. It is that the sector’s value proposition has been articulated almost entirely in system‑level terms—AI competitiveness, cloud reliability, digital GDP—while the politics of land, water, and power are local and concrete. As long as the benefits of AI‑era data centers remain abstract in the places that host them, the stated concerns about bills, water, taxes, and noise will continue to be amplified by a deeper sense that the trade is one‑sided.

The Backlash is Still Structurally Manageable

The backlash is serious. But it is not terminal. The politics around AI data centers are tightening fast, but they are not yet closing the door on new capacity. They are redrawing the map and raising the bar for how and where projects are built.

Opposition is intense in specific places, not everywhere. Nationally, “data centers” in as a generalized concept do not draw uniform hostility; resistance spikes when a megawatt count is attached to a specific parcel identified as being in a community's "backyard." The organized infrastructure of protest is real but finite: a few hundred local groups and coalitions that can decisively affect projects in their footprint, and barely register elsewhere. It is the familiar pattern of other contested infrastructure classes: concentrated local costs, diffuse system‑level benefits.

Neighbors live with trucks, substation views, and uncertainty about their bills; the upside is expressed in terms of AI competitiveness and cloud resilience. That kind of problem is hard, but it is the kind of hard developers already know from highways, pipelines, and high‑voltage lines. Historically, the answer has been better siting, more credible benefit‑sharing, and more disciplined governance—not abandoning the underlying asset.

There is also real counter‑pressure at the top of the decision tree. The Maine moratorium fight is a clean example. Legislators were prepared to impose a statewide pause on large data centers; the governor vetoed it on economic‑development grounds, arguing that a blanket freeze would undercut the state’s ability to attract advanced industry more broadly. That is a governor absorbing short‑term political risk in order to preserve the long‑term upside of continued digital and industrial build‑out.

Market coverage points in the same direction. Wall Street notes the growing pile of delayed and cancelled projects, but it has not treated that as a reason to mark down AI or cloud demand. The language is about risk to build schedules and pressure on where capacity lands, not about a structural end to the build‑out. Hyperscale capex plans still assume substantial new infrastructure. That gap between local resistance and macro‑level demand is exactly what creates room for negotiated outcomes: governors, mayors, and development agencies are operating in the space between a loud minority of opponents and an economy that now expects AI‑grade compute as baseline.

Competition between jurisdictions keeps the system from locking up. When one state moves toward a freeze, a neighbor quietly sharpens its pitch: clearer tax policy, more predictable permitting, a public line about wanting to host AI‑ready campuses under defined rules. The same dynamic is visible internationally: countries facing their own noise about power and land are still cutting deals for AI‑heavy campuses, often packaged with new generation, storage, or research investments. Every jurisdiction that decides its politics cannot tolerate another hyperscale facility effectively hands an opportunity to one that believes it can. For operators and investors, that matters more than any single moratorium headline: it means the constraint has pushed us to be even more selective.

Now that the dynamics have shifted, our industry has to make the uncomfortable adjustment as a sector that was used to being the quiet backbone of digitization, but it is not entirely unfamiliar territory. Energy, transport, and heavy industry have operated in that regime for decades, balancing local opposition, political cycles, and long‑dated capital; data centers are now in the same club. The risk is real, but it is legible.

For a data center operator, the practical implication is straightforward: treat community and political risk as a design parameter alongside power, latency, and land, not as an afterthought or an existential verdict. Projects that build that reality into their site screening, their deals with utilities, and their local value proposition will still move. Those that assume the old invisible infrastructure model still applies will find out, late and expensively, that the ground rules have changed.

We Are Entering a Decisive Decade for AI Infrastructure Social License

AI infrastructure is heading into a hard decade, not a hype decade. The choices governments, utilities, and operators make in the next 5–10 years will set the ground rules for whether AI‑driven data center growth follows a stable, rules‑based path or keeps colliding with moratoriums, recalls and campaign‑season vetoes. Capacity will keep growing either way; what changes is which regions get it and on what terms.

The backlash is now a durable feature of the landscape, not a passing storm. It shows up most sharply in specific corridors, election cycles and regulatory environments, not as a universal rejection of data centers or AI. Some counties and states are building full toolkits of pauses, bans and cost‑shifting rules. Others, sometimes next door, are quietly refining frameworks to be restrictive enough to manage water, power and land, but predictable enough to win projects. The map is fragmenting.

At the same time, the macro story still runs in the opposite direction. National governments treat AI capacity as strategic infrastructure. Large platforms are structuring multi‑year capex plans on the assumption that training and inference demand will keep climbing. Capital markets have not repriced that thesis; they have repriced the execution risk. Underneath the noise, the basic signal remains: more compute, closer to more population and enterprise load, on tighter emissions and resilience constraints.

That combination means the data center sector’s central risk is distributional, not existential. Jurisdictions that keep relying on ad‑hoc deals and opaque utility arrangements will generate the next round of moratoriums and high‑profile project failures, and capital will move around them. Jurisdictions that build clear, credible frameworks on water use, power sourcing, emissions, and fiscal terms, will capture not only the next set of campuses, but also the manufacturing, software, and services ecosystems that follow dense compute.

Operators and investors have agency in how that plays out. The advantage sits with those who move early to align projects with grid realities, water constraints, and local budgets; who make their economics legible; and who treat host communities as enduring partners in national‑scale infrastructure rather than as the last box to tick before construction.

Over the next decade, the social license for AI infrastructure will not be granted once and filed away. It will be earned, site by site and deal by deal, by the players that accept the new terms and design around them.

References:

1.       https://www.jll.com/en-us/newsroom/jll-north-america-data-center-report-year-end-2025

2.       https://www.cnn.com/2026/04/12/climate/maine-data-center-ban-bill

3.       https://www.nytimes.com/2026/03/26/business/economy/ai-data-centers-construction-local-opposition.html

4.       https://www.wkar.org/wkar-news/2026-02-27/michigan-data-center-tracker-moratoria-legislation-and-grid-impact

5.       https://www.datacenterwatch.org/q3-q4-2025

6.       https://www.datacenterwatch.org/report

7.       https://www.nytimes.com/2026/03/26/business/economy/ai-data-centers-construction-local-opposition.html

8.       https://www.thecooldown.com/green-business/data-center-opposition-local-resistance-construction-challenges/

9.       https://www.nbcnews.com/politics/politics-news/reining-data-centers-sparks-rare-bipartisanship-statehouses-rcna262990

10.  https://www.axios.com/2026/04/05/data-centers-midterms-state-bans-bills-ai

11.  https://builtin.com/articles/state-data-center-moratoriums

12.  https://www.multistate.us/insider/2025/10/2/data-centers-confront-local-opposition-across-america