Data Center Siting / AI Power Politics

Data center siting refers to the location decisions hyperscalers and AI developers make when placing training and inference capacity. Those decisions turn on a narrow set of variables — cheap firm power, fiber routes, cooling water, low-humidity climate, and local tax abatements — but carry diffuse consequences for ratepayer bills, gas-turbine and nuclear restarts, water allocation, and the geographic distribution of AI capital across states. Siting has become a contested area of US energy, grid, and community politics, with local opposition, state legislation, and at least one documented case of political violence emerging by 2026.

This page covers the political economy. See Compute Governance for the broader policy frame, AI Environmental Impact for grid/water/emissions detail, and RAND — AI's Power Requirements Under Exponential Growth (2025) for the quantitative forecast.

The siting calculus

Hyperscalers optimize jointly over several factors:

• Firm power availability — interconnection queue position; stranded generation; willingness of utilities to sign 10–20 year take-or-pay PPAs.
• Power price — industrial rates, not residential; ERCOT and Southeast rates have risen as AI load bids up wholesale.
• Fiber backbone proximity — dark-fiber routes between training sites and inference regions.
• Water — evaporative cooling needs; drought-year curtailment risk.
• Tax treatment — sales-tax exemptions on servers (Virginia, Texas, Georgia), property-tax abatements, PILOT agreements.
• Permitting speed — NEPA reach for federal land/permits; state siting boards; local zoning.
• Political tolerance — whether local government will tolerate gas peakers, transmission expansion, and truck traffic.

Cost structure

Epoch AI researchers Amelia Michael and Ben Cottier modeled the total cost of ownership of a typical one-gigawatt US AI data center on May 15, 2026, putting annual total cost of ownership at $8.5 billion, of which $5 billion (60%) is servers alone, exceeding energy and all operating costs (Source: epochai.substack.com). On this accounting, server costs — essentially Nvidia GPU costs — are the dominant cost variable rather than energy or operations, making chip-supply policy (CHIPS Act, export controls, TSMC capacity) the largest cost lever and indicating that political pressure focused on power and water addresses only a minority of the underlying capital intensity.

A "neocloud" is defined in reporting as a facility class running more than 100 kW per rack with direct-to-chip or immersion cooling, distinct from hyperscale general-compute (Data centers are coming for rural America — Abigail Bassett (The Verge, May 13 2026)).

Public opinion

A Gallup poll released May 13, 2026 found that 70% of Americans oppose data-center construction in their communities, with nearly half strongly opposed, and that more respondents would rather live near a nuclear power plant than a data center (Source: washingtonpost.com). A Reuters/Ipsos poll of 4,531 adults conducted June 3–8, 2026 and published June 11 found 64% of Americans disagree that rapid data-center construction is mainly a good thing, 57% would oppose a data center in their community, and 77% — across parties — worry AI will raise electricity prices, ahead of the November 3 midterms (Source: reuters.com). The AI Backlash Could Get Very Ugly — Lila Shroff (The Atlantic, May 13 2026) and Data centers are coming for rural America — Abigail Bassett (The Verge, May 13 2026) had earlier described a rising trajectory of siting opposition.

The AP-NORC Center for Public Affairs Research and the Energy Policy Institute at the University of Chicago released on September 16, 2026 a survey of 3,424 U.S. adults conducted July 6–24, 2026. Sixty-three percent were extremely or very concerned about data centers' effect on electricity prices in the communities where they are built and 84% at least somewhat concerned; the at-least-somewhat figures were 79% for water supply and 78% for power outages. Seventy-three percent supported requiring companies that build data centers to pay for the electric grid upgrades needed to support them (79% of Democrats, 76% of Republicans), 65% requiring data centers to use clean energy sources (80% and 56%), 62% limiting the number of new data centers (70% and 57%) and 41% restricting new natural gas plants to supply them (50% and 33%). Concern did not differ among urban, suburban and rural residents (AP-NORC/EPIC 2026 Energy Survey: AI, Data Centers, Climate and Energy).

The same survey recorded regional and demographic differences. Residents of the Northeast were less concerned than residents of the South about electricity prices (58% against 66% extremely or very concerned) and water supply (50% against 60%), and residents of the West less concerned than those of the South about electricity prices (60% against 66%) and power outages (50% against 56%). Adults 45 and older supported the grid-upgrade requirement more than younger adults (80% against 66%), as did college graduates compared with adults without a degree (83% against 68%); college graduates were also more supportive of restricting new gas plants for data centers (47% against 37%). Republicans were more likely than Democrats to expect data centers to be at least somewhat beneficial for job creation (55% against 46%) and improved infrastructure (51% against 41%) (AP-NORC/EPIC 2026 Energy Survey: AI, Data Centers, Climate and Energy).

The Annenberg Public Policy Center's W38 national survey, fielded June 16 to July 19, 2026 among 1,320 U.S. adults with a margin of error of ±3.48 percentage points, found 61% opposing the construction of new data centers in their area — 44% strongly and 16% somewhat — against 14% in support. Opposition rose from 48.9% in the center's February–March wave, a wave-to-wave shift of 11.8 points (95% CI +6.9 to +16.6); support fell from 21.3%. Among the 1,241 respondents interviewed in both waves the within-person shift was +12.5 points. Majorities of Democrats (69%), Republicans (54%) and independents (53%) opposed, and opposition was highest among adults under 30 at 70%, declining to 57% among those 65 and older. It was the only item in the center's AI battery to move significantly: the share saying government has done "too little" to regulate AI (67.9%) and the share expecting AI's ten-year impact on the country to be negative (39.5%) both had confidence intervals spanning zero (Artificial Intelligence — Topline Report (APPC W38 National Survey); Source: annenbergpublicpolicycenter.org).

A Pew Research Center survey of 10,548 U.S. adults conducted July 20 to August 9, 2026 and published September 22 found 60% would be not too or not at all comfortable with a new data center operating in their area, against 19% somewhat and 7% extremely or very comfortable; the uncomfortable share was 54% among Republicans and Republican leaners, 68% among Democrats and Democratic leaners and 81% among liberal Democrats, with majorities in urban (57%), suburban (63%) and rural (60%) areas. Compared with January 2026, the share saying data centers are mostly bad for the environment rose from 39% to 54%, for home energy costs from 38% to 50%, for people's quality of life nearby from 30% to 49%, for local jobs from 15% to 24% and for local tax revenue from 12% to 21%, while 4% called them mostly good for each of the first three. Among rural Americans the share saying data centers are mostly bad for the environment rose from 32% to 50%; Pew notes that most planned data centers will be in rural areas (Americans' Views of Data Centers Have Turned More Negative (Pew Research Center, September 2026)). Unlike the AP-NORC/EPIC finding above that concern did not differ by community type, Pew records rural respondents moving further than urban ones on the environment question (+18 points against +10), though the two surveys ask different questions. The short read sets the results against the positions of elected officials: President Trump's support for construction, Texas Governor Greg Abbott's shift from supporting expansion to what Pew describes as "a statewide moratorium on new data centers in August," and moves by Pennsylvania Governor Josh Shapiro and Illinois Governor JB Pritzker toward additional rules and oversight on data-center construction and operation (Americans' Views of Data Centers Have Turned More Negative (Pew Research Center, September 2026)). Pew's "statewide moratorium" is a broader description than the halt to approvals for grid connection described under the Texas interconnection moratorium below.

The campaign acquired a presidential voice on August 31, 2026, nine weeks before the midterm elections, when President Donald Trump wrote on Truth Social that "The only reason that communities throughout the U.S.A. should not want Data Centers is if they want to end up being backwards and poor," that "If they want to be successful and rich, with far lower taxes and jobs all over the place, let Data Reign," and that "If we kill the Golden Goose, you will only have yourselves to blame." The post ran alongside an industry effort: Build American AI, an advocacy group linked to the pro-AI super PAC Leading the Future, is prepared to spend a $50 million war chest defending data centers in Kansas, Ohio and Wisconsin, and is launching a further super PAC for candidates aligned with its agenda. Investor Gavin Baker set out a six-part case for data centers on X covering water, taxes, jobs, power, environment and local economies, which Nvidia chief executive Jensen Huang amplified (Source: axios.com).

Hyperscalers have begun countering this opposition with public-opinion campaigns. Meta has spent more than $6 million on TV ads in state capitals and Washington since late 2025, with the message that data centers create jobs — "We're bringing jobs here. For us, and for our next generation." The ads run in markets adjacent to its Altoona, Iowa and other hyperscale campuses, and represent the most concrete documented case of a hyperscaler public-opinion campaign as a counter to siting opposition (Source: nytimes.com).

The integrity of siting comment processes became an issue in New Mexico, where residents of Las Cruces, Albuquerque, and Rio Rancho said their names appeared without permission on form letters supporting the $165 billion Project Jupiter data center — slated to support Oracle and OpenAI operations — on the state environment department's comment portal, per June 30, 2026 reporting. The comment period closed July 6; the department said it was looking into the issue, and Oracle denied wrongdoing (Source: route-fifty.com). By mid-July the New Mexico Attorney General was investigating the letter complaints, and the project itself had been revised: after its gas-turbine plan stalled on a state environmental permit, Oracle pivoted to 2.45 gigawatts of Bloom Energy fuel cells (an analyst-estimated ~$8 billion) for the proposed $165 billion, 2-plus-gigawatt Doña Ana County site, and the New Mexico Environment Department set an October 19 air-permit hearing (Source: kunr.org). The state also issued its second rejection of fuel-pipeline routes for the 1,400-acre site (Source: theinformation.com). Related cost pressures on Oracle's data-center program in Wisconsin are covered on the company page.

The foreign-influence framing

A second industry response has been to attribute the opposition itself to foreign actors. In a column published September 6, 2026, the Trump administration is described as increasingly joining technology leaders in suggesting that foreign powers, and China in particular, are behind the US populist movement against AI — a framing technology leaders had advanced for weeks. Only the column's opening passage could be retrieved and the claim is not verified beyond it (Source: wsj.com).

The evidentiary basis usually cited is OpenAI's disclosure of June 10, 2026, which describes a banned account cluster it named "Data Center Bandwagon" that generated social-media comments and images claiming AI data-centre buildouts were raising household electricity prices. What that disclosure claims is narrower than the framing it is used to support: OpenAI states that the operation's significance lies "not because the operation appears to have shifted public opinion, but because it shows PRC-origin influence operators testing narratives against AI infrastructure," and that it "found no evidence of meaningful breakout beyond its own activity." The disclosure also situates the behaviour in a longer pattern of foreign operations latching onto existing local issues rather than creating them — which, on its own terms, presupposes that the grievance predates the operation. The polling above documents opposition rising across the partisan divide over the same period. See AI-enabled influence operations.

On July 13, 2026, former FCC chairman Tom Wheeler warned that data-center opposition is becoming a proxy for broader anti-AI politics (Source: insideaipolicy.com). The same day, states pushed back on the EPA's proposed new-source-review "construction" definition sought by the AI sector, a permitting change that would affect how data-center power projects are reviewed under the Clean Air Act (Source: insideaipolicy.com). See AI Environmental Impact.

Polling released by the Information Technology and Innovation Foundation with Public First at an August 26, 2026 webinar, "Can the United States Win the AI Race if Americans Oppose Data Centers?", measured the opposition against other local developments. A new data center was the only one of ten tested to draw net opposition, at −20, against +43 for new housing, +30 for a road or highway, +25 for a solar farm, +21 each for a warehouse and a wind farm, +18 for a train station, +16 for a shopping mall, +10 for an office complex and +2 for a new power plant or other electric infrastructure. Drawing on a POLITICO poll by Public First, the deck reports net support for a data center within three miles of home falling from +9 in January 2026 to −17 in July 2026, with Harris 2024 voters moving from +3 to −30, Trump 2024 voters from +19 to 0 and 2024 non-voters from +1 to −21 — a decline across the partisan divide rather than within one side of it. Over the same period the share saying data centers deplete the water supply rose from 34% to 55%, the share saying they raise electricity bills rose from 43% to 58%, and the share saying they are bad for the planet rose from 32% to 49%; the belief that building more data centers enables more complicated AI fell slightly, from 62% to 59% (Public First / ITIF, \"The opposition to data centers\" (polling deck, August 2026)). In the underlying shares, 26% said they would support a new data center in their community and 46% said they would oppose one, and 49% said they would rather nearby land stay undeveloped than host a data center (Source: route-fifty.com).

On the cross-country question "Should your country build more or fewer data centers?", the United States scored +2 net (24% build fewer, 26% build more), the lowest of fifteen countries surveyed. The full ranking is Nigeria +68, Vietnam +59, India +58, Indonesia +48, Brazil +38, South Africa +36, Singapore +24, Japan +22, Mexico +22, Poland +21, Germany +20, Canada +18, the UK +17, France +15 and the US +2. The deck attributes the US position substantially to partisan divergence: the US has the widest partisan gap of the fifteen at 32 points (Democrats −10, Republicans +23), ahead of Japan at 25 and France and Mexico at 20, and the US Democratic figure is the only negative party score anywhere in the set (Public First / ITIF, \"The opposition to data centers\" (polling deck, August 2026)).

The deck also supplies a measurement bearing directly on Wheeler's warning above. Among respondents who said AI will make society worse, 73% opposed a local data center, against 32% opposing an average of nine other developments — a gap of 41 points, consistent with the data center serving as a proxy for a view of AI rather than as a discrete siting judgment. The comparison groups are what carry the inference: respondents who agree "I don't want new development in my area" oppose a data center at 61% and other developments at 35%, a gap of 26 points, smaller than the gap for AI pessimists; among all adults the figures are 46% and 23%. Measured as net scores, the gap between AI pessimists and others is 68 points for a data center (+5 against −63) against 39 for a power plant, 33 for a warehouse and 15 for a solar farm. ITIF president Daniel Castro opened the event by naming the same concern, saying there is a "risk that opposition to data centers becomes, in effect, opposition to AI itself" (Public First / ITIF, \"The opposition to data centers\" (polling deck, August 2026); Source: insideaipolicy.com; axios.com).

Distance and purpose moderate the response more than the presence of a data center as such. Asked about data centers "in the United States," respondents are net positive on one in a different state (+20), on data centers used for online banking, video streaming and photo storage (+18), on data centers anywhere in the world (+16), on data centers storing government data (+11) and on data centers within their own state (+7), and net negative on data centers used to train AI models (−2) and on one "within a 10 minute drive of where you live" (−19). Trust in technology companies produces a comparable spread: offered undeveloped land with no alternative use, respondents who trust technology companies completely or somewhat prefer a data center by +16, those who trust them a little are at −29, and those who do not trust them at all at −48 (Public First / ITIF, \"The opposition to data centers\" (polling deck, August 2026)).

The deck's policy testing, which no coverage carried, bears on the state legislative activity described below. Asked to support or oppose announcements from a hypothetical state candidate, respondents rank requiring data centers to pay the full cost of their own electricity and any new power lines highest at +57, followed by a binding local resident vote at +52, a temporary construction pause pending rules on power, water and bills at +48, and minimum distance requirements from homes and schools at +46. Repealing existing state tax breaks scores +34 and a permanent state ban on new data centers +20. The two measures that score negative are both pro-build: property tax discounts at −7 and preventing counties and towns from blocking data centers so the state decides at −9, a direct measurement of the preemption approach several states have taken. Fast-tracking planning permission scores 0 (Public First / ITIF, \"The opposition to data centers\" (polling deck, August 2026)).

Asked which developer commitments matter most, respondents rank resource guarantees above community benefits — guaranteeing local electricity bills will not rise (+19.3), generating all power on site (+16.0), near-waterless cooling (+12.7) and a household bill credit (+11.7) at the top, against a new park and sports fields (−18.7), setbacks and screening (−16.2) and demolition-and-restoration guarantees (−13.2) at the bottom. The ordering is sharper among respondents who oppose a data center in their community, for whom on-site power generation scores +24.9 and a new park −25.0. A conjoint experiment showing two proposals side by side with the option to leave the land undeveloped ranks water use as the widest-spanning attribute at 8.3 points — a proposal using as much water as 10,000 homes is chosen 26.4% of the time against 34.7% for one using almost no water and returning more than it takes — ahead of permanent jobs at 7.7 points, developer identity at 6.5 points and construction jobs at 2.1 points (Public First / ITIF, \"The opposition to data centers\" (polling deck, August 2026)).

Two cautions attach to the deck. It is commissioned research published by an organization with a stated pro-technology position, released at an event framed around whether opposition threatens US competitiveness. And it states no field dates, sample sizes, margins of error or fieldwork provider for either of the two instruments it draws on, so its levels cannot be reconciled with confidence against the Gallup, Reuters/Ipsos and Annenberg figures above. Those give 70%, 57% and 61% opposing a local data center on binary or near-binary items; the deck's 46% figure comes from a six-point scale offering "neither support nor oppose" and "don't know," and its headline measures are net scores rather than levels. The three should be read as measuring different things rather than as a trend (Public First / ITIF, \"The opposition to data centers\" (polling deck, August 2026)).

An Economist–YouGov poll fielded August 28–31, 2026 found that nearly two-thirds of Americans oppose building a data center in their local area (Source: yougov.com; fortune.com). Heatmap polling cited in a Wired column of September 6, 2026 puts the figure higher still, at as many as 75% saying they would oppose a local data center, against 42% a year earlier (Source: wired.com). The two are not directly comparable to each other or to the Gallup, Reuters/Ipsos and Annenberg figures above, since neither the question wording nor the response scale is stated in the coverage.

Republican pollster Tony Fabrizio of Fabrizio Lee sent a memorandum dated September 14, 2026 to "GOP Candidates & Policymakers" stating that "unqualified support for building more AI data centers is a losing position." It reports survey results in which voters opposed additional construction 65% to 24%, with 44% strongly opposed against 9% strongly in favor, and net support negative in every group reported: −4 among Republicans, −49 among independents and −73 among Democrats. Offered three options, 61% chose construction with conditions requiring the company to meet its own electricity needs, protect ratepayers from higher bills and protect nearby water and environmental resources, 28% chose a ban and 7% construction wherever local officials approve; conditional construction was the majority choice in every group, from 52% of generic-ballot undecided voters to 66% of Republicans. The memorandum states that the survey did not test effects on vote share and tested the three conditions only as a package, and it names no sponsor, field dates, sample size or question wording, so its levels cannot be reconciled with the public polls above (Fabrizio Lee — Neutralizing Political Attacks on AI Data-Center Development Through Enforceable Conditions (memorandum, September 14, 2026)). WIRED, which linked the memorandum, placed it after an August 2026 memorandum from the National Republican Senatorial Committee that WIRED described as sent to AI companies, warning that "campaign or party committees can not fix the toxic brand of an entire segment of the economy" (Source: wired.com); the Fabrizio memorandum does not refer to that document, which is itself addressed to interested parties (NRSC — Ohio Data Center Risk (memorandum, August 18, 2026)). A New York Times/Siena poll released the week of September 14 found more than 60% of Americans, including 47% of Republicans, opposed to data-center construction, while also finding that AI and data centers rank far down voters' priorities and that almost 20% do not know which party would handle the issue better. President Trump has continued to defend data centers, writing in late August 2026 that only communities wanting to be "backwards and poor" would refuse them (Source: wired.com).

Power procurement and the grid

Capacity markets and grid strain

PJM's 2025/2026 capacity auction cleared at roughly $269.92/MW-day, about 10× the prior year; the increase is widely attributed to AI and data-center load colliding with coal retirements, and ratepayer advocates across Ohio, Maryland, and Illinois have mobilized. In ERCOT, an isolated grid combined with light permitting and SB 6 (2025) load-interconnection reforms make Texas the highest-growth AI power market, with peak demand forecasts having doubled since 2023 (ERCOT LTLF 2024, 2025).

A separate question is how much of the requested load is real. Documents from ERCOT and Texas Governor Greg Abbott, described in a review published September 1, 2026, put requests from data centers and other large energy users to connect to the Texas grid at more than 474 gigawatts, up from about 48 gigawatts in 2023. Across portions of the Midwest, Mid-Atlantic and the South, electricity requests from very large power users, mostly data centers, exceed 700 gigawatts — more than ten times industry estimates of current U.S. data-center power use — and the account states that much of that demand may be an illusion, developers submitting the same project to several utilities or filing speculative requests. The cornerstone of Abbott's responding order is an audit requiring proposals to power up data centers to disclose who ultimately owns them, replacing earlier rules under which only an affiliate company had to be named (Source: reuters.com).

The cost-allocation question reached the federal regulator directly. On June 24, 2026 the Federal Energy Regulatory Commission pressed grid operators to justify or revise their market rules for connecting large power users such as data centers, with the stated goal of preventing the costs of serving that load from being shifted onto other customers (Source: insideaipolicy.com). The move put the central distributional question — whether data-center developers or existing ratepayers bear the cost of AI-driven grid upgrades — under active federal review.

The House passed the Ratepayer Protection Act (H.R. 9340) 417–3 on September 16, 2026. It adds to PURPA a standard, which state regulators must consider and may decline, under which rates for customers with 100 MW or more of information-technology load recover the full incremental cost of grid upgrades (Ratepayer Protection Act, H.R. 9340 — text as reported in the House (September 10, 2026); Source: clerk.house.gov).

Nuclear power-purchase agreements

Frontier labs and hyperscalers have signed direct power-purchase agreements with nuclear operators at a pace described as unseen since the 1970s:

• Constellation + Microsoft — Three Mile Island Unit 1 restart (announced Sept 2024, ~835 MW).
• Talen Energy + Amazon — Susquehanna nuclear campus co-location (FERC blocked the initial behind-the-meter structure in late 2024; restructured).
• Vistra — Comanche Peak PPA expansion; explored uprates.
• Small modular reactor (SMR) term sheets — Oklo, X-energy, Kairos with Google/Amazon/Microsoft.

Hyperscalers have paid premium prices for 24/7 carbon-free firm power to keep net-zero commitments alive. Whether this crowds out residential and industrial nuclear allocation, or instead finances new build that expands supply, is unresolved.

The regulatory side of nuclear-for-data-centers drew a legal challenge in mid-2026: Democratic attorneys general from nine states urged the Nuclear Regulatory Commission, in June 15, 2026 written comments, to withdraw its proposed streamlined licensing of microreactors intended to power data-center growth, arguing the plan violates federal statutes including the National Environmental Policy Act (Source: insideaipolicy.com).

Siting as a security question

A national-security reading of siting entered congressional testimony on September 2, 2026, when Hudson Institute senior fellow David Feith argued to the House Permanent Select Committee on Intelligence that large data centers should be treated as critical grid infrastructure and that the security perimeter of sensitive facilities "must now extend well beyond the fence line" (A New Counterintelligence Map for the AI Age (Feith, HPSCI, 2026-09-02)). Two threads in the statement bear on where facilities are placed and who owns the land around them.

The first is load behavior. Feith cites a routine transmission fault in Northern Virginia on July 22, 2026 that moved roughly 3,800 MW of data-center load from the grid to backup generation within seconds, producing voltage and frequency swings from Virginia to Chicago without a blackout and without any cyber component, and argues that the event shows the physics an attacker with synchronized access to inverters, batteries or uninterruptible power systems would exploit. That places the concentration of gigawatt-scale digitally controlled load in a single interconnection — the defining feature of the Northern Virginia and ERCOT clusters described above — inside the security frame rather than only the ratepayer frame.

The second is adjacency. Feith cites the 2024 CFIUS-ordered divestment of a PRC-owned cryptocurrency mine within a mile of Wyoming's Warren Air Force Base, along with a trailer park bought adjacent to Missouri's Whiteman Air Force Base through companies whose foreign owners press reporting linked to alleged Chinese-intelligence ties, and Ukraine's Operation Spiderweb of June 1, 2025, in which some 117 drones trucked into Russia struck bombers at four airfields. His conclusion is that "nearby farms, warehouses, and commercial properties can matter as much to base security as gates, guards, and guns," and that while Congress gave CFIUS authority over foreign real-estate transactions near sensitive installations and Treasury expanded the covered-installation list in 2024, implementation of the February 2025 America First Investment Policy memo "has been slow and limited." The mechanism is the mirror image of the local-opposition dynamics treated below: here the constraint on siting comes from federal review of who holds adjacent land, not from a county board.

Physical attack on operating facilities is on the record outside the United States. Drone strikes in March 2026, in the early days of the Iran war, directly hit two Amazon Web Services data centers in the United Arab Emirates and damaged a facility in Bahrain; Iran's Islamic Revolutionary Guard Corps claimed responsibility and said it targeted the Bahrain facility because of Amazon's support for the US military. AWS said on September 15, 2026 that it could not restore access to resources and data hosted exclusively in its Bahrain region, where damage at multiple sites exceeded what its services are designed to withstand, or to data in one of its UAE availability zones. The attacks have raised questions about the security of planned Gulf data-center developments, including the UAE Stargate campus, for which OpenAI, Nvidia, Oracle and Cisco signed on as core partners in May 2025; the UAE is considering dispersing that project's facilities across the country and adding protections including underground sites, blast-resistant concrete, and drone and missile interceptors, according to people familiar with the matter cited in reporting of September 11, 2026 (Source: cnbc.com).

Regional flashpoints

Memphis — xAI Colossus (2024–2026)

xAI's Memphis Colossus site drew sustained community opposition over on-site methane gas turbines installed without standard air permits. Local NAACP chapters, the Southern Environmental Law Center, and Memphis Community Against the Pollution sued and complained to EPA Region 4. The dispute has been characterized as a case of AI infrastructure outrunning Clean Air Act enforcement capacity. On June 22, 2026 the Department of Justice moved to dismiss the NAACP's Clean Air Act citizen suit, asserting a constitutional right to intervene in and override such citizen-enforcement actions — an escalation from its May 18, 2026 signal of a "possible intervention" on national-security grounds (Source: insideaipolicy.com). See xAI and Elon Musk for principals.

Louisiana — Meta hyperscale (2024–2026)

Meta's multi-billion-dollar Richland Parish campus, announced late 2024, will anchor a new Entergy generation buildout — including gas combined-cycle plants whose costs are proposed to be socialized across Louisiana ratepayers. Louisiana Public Service Commission hearings are a test case for whether state PUCs will let hyperscalers internalize their grid externalities.

Northern Virginia — load growth saturation

Data Center Alley (Loudoun/Prince William) consumes more power than many US states. Dominion Energy has warned of interconnection delays through the early 2030s; the state's 2024–2025 legislative session featured multiple bills on AI-load cost allocation (none passed).

Organized opposition has begun blocking Virginia projects outright: Blackstone's QTS abandoned plans to build its portion of the 2,100-acre Prince William Digital Gateway — billed as the world's largest planned data-center campus — after years of organized opposition and legal challenges, per July 2, 2026 reporting (Source: bloomberg.com). The end came on July 9, 2026, when QTS dropped its appeal of a court ruling that had voided the project's rezoning over a defective newspaper notice, ending years of litigation driven by local opposition (Source: washingtonpost.com; tomshardware.com). Americans for Responsible Innovation issued a "data center playbook" giving citizens, civil society groups, and local and state officials tools for engaging in data-center deliberations, per accounts published July 3, 2026 (Source: insideaipolicy.com).

Argentine Patagonia — proposed buildout (2026)

Investors have proposed a cluster of data centres across Argentine Patagonia, drawn by cool temperatures, renewable energy, shale gas and open land, in reporting published September 7, 2026. Pampa Energía is pushing for a facility of up to 500 megawatts beside its Loma de la Lata thermal plant in Neuquén and is supplying 30 MW for the first stage of a separate project in Bahía Blanca; Sur Energy has proposed a clean-energy site of up to 500 MW costing up to $25 billion; Green Capital is planning a Chubut project reaching 300 MW in its first stage at a cost of $3 billion; and FlexDomes has proposed 120 MW of IT load in Neuquén. President Javier Milei's RIGI regime supplies the tax breaks and regulatory stability underpinning the interest. Connectivity and electrical infrastructure remain hurdles, investors are seeking pilots of 20 MW to 40 MW before expanding, and some are waiting on the outcome of Milei's re-election before committing (Source: reuters.com).

Federal legislative proposals

Federal bills addressing data-center siting, measurement, resilience and cost allocation appeared as a package in September 2026. Rep. Suhas Subramanyam (D-Va.) introduced the National Data Center Plan on September 8, 2026, four bills covering different layers of the problem treated on this page (Source: nextgov.com):

The Data Center Fair Share Act states at federal level the cost-allocation principle the state measures below approach through tariffs and permitting conditions, and the measurement bill addresses the absence of a standard metric that makes the water and energy figures on this page difficult to compare across jurisdictions. Speaker Mike Johnson was separately reported on September 11, 2026 to be planning a vote the following week on a bill to prevent data centers from raising consumer energy costs (Source: axios.com). The bill was not named in that account, and whether it is one of the four above is not established.

State and local policy

State siting laws

• Virginia — HB 1601 (2024) commissioned a JLARC study on data-center impacts; the 2025 session saw failed bills on cost allocation. In March 2026, state senators backed a repeal of Virginia's data-center sales-tax exemption, worth about $1.6 billion annually (Source: datacenterwatch.substack.com). On September 18, 2026 Governor Abigail Spanberger signed Executive Order 22, which bars executive-branch nondisclosure agreements on data center projects and excludes new projects of 25 MW or more peak demand from state site-readiness and expedited-review assistance, and released a Data Center Accountability Framework proposing, for the 2027 session, local approval of any data center using more than 25 megawatts in place of by-right approval, an end to state site-development subsidies, and a larger share of transmission and generation costs allocated to large loads Virginia Executive Order 22 (2026) and the Data Center Accountability Framework — signed texts.
• Ohio — PUCO proceedings on AEP's AI-load tariff (proposed higher minimums and longer commitments).
• Texas — SB 6 (2025) imposes interconnection-study rigor on large loads and permits load-shedding orders during emergencies. Governor Greg Abbott has called for tighter oversight of data-center development and urged lawmakers to reconsider incentives amid growing local opposition to AI infrastructure projects across the state (Source: fortune.com).
• Georgia — Georgia Power's IRP updates have repeatedly raised load forecasts; the PSC approved new gas capacity partly on AI-load grounds.
• Pennsylvania — In March 2026, House lawmakers advanced bills providing a model municipal framework for data-center siting and requiring annual water/energy-use reporting (Source: datacenterwatch.substack.com).
• South Dakota — Gov. Larry Rhoden signed SB 135 and HB 1038 (March 2026) imposing infrastructure-cost-recovery requirements on large data centers (Source: datacenterwatch.substack.com).
• Arizona — Gov. Katie Hobbs asked the Legislature to repeal the state's data-center sales-tax incentive, a 2013 exemption costing about $38 million a year; the $18 billion budget she signed in June 2026 instead imposed a three-year moratorium on new data-center sales-tax breaks, Arizona joining Ohio and Illinois in pausing new exemptions (Source: news.bloombergtax.com).
• New York — Lawmakers, including Gov. Hochul, are working through whether to require new data centers to internalize their grid costs and bring their own power; no consensus had emerged, with the Democratic-led effort described as slow (Source: subscriber.politicopro.com). On June 8, 2026, environmental advocates pressed Gov. Kathy Hochul to sign legislation that would impose the nation's first statewide moratorium on data-center construction, framing her signature as "atonement" for easing the state's climate law in the budget bill the prior month (Source: insideaipolicy.com). The New York Legislature passed the Responsible Data Center Development Act, reported June 10, 2026 — a one-year moratorium on large data centers paired with new electric and water rate classes for facilities over 20 MW; Hochul's office said she would review the bill (Source: route-fifty.com). On July 14, 2026, Hochul acted by executive measure instead of signing that bill, imposing a one-year moratorium on new facilities drawing 50 MW or more — making New York the first U.S. state to halt construction of large new data centers — directing a Generic Environmental Impact Statement to set consistent standards, and saying she will pursue repeal of the state's data-center sales tax exemption; the legislature's 20 MW-threshold bill remains unsigned (Source: reuters.com). The order also directed Empire State Development to post a Community Investment Framework, which was released on September 15, 2026. The voluntary framework sets a recommended benchmark of $1 million per megawatt of utility demand for community investments that host localities may seek from data-center developers and operators, described in coverage as investments of at least that amount, so that host communities "share equitably in the long-term economic value created by these projects"; the accompanying guidance gives $50 million as the recommended level for a 50 MW project. The guidance states the benchmark is inclusive of grid-modernization investments required by the state, leaves it to municipalities with their counsel to decide whether to make a framework binding, and advises localities asked to sign non-disclosure agreements to limit them to material exempt from the state's Freedom of Information Law. Empire State Development President Hope Knight said that "because data centers are not major local job generators, their economic development profile differs from traditional manufacturing" (Source: reuters.com; New York State Guidance for Host Communities: Community Investment Framework for Data Center Development (ESD, 2026)).
• Ohio — Beyond the PUCO AEP-tariff proceeding (above), state Sen. Brian Chavez unveiled a data-center bill on June 9, 2026 that would halve Ohio's 100% sales-tax exemption (worth roughly $1.6 billion the prior year), cap local property-tax abatements at 50%, bar NDAs from overriding public-records law, and create a PUCO data-center rate class, with a Senate floor vote targeted for June 10 (Source: route-fifty.com).
• Massachusetts — Gov. Maura Healey signed Executive Order 658 on September 8, 2026, barring state agencies from permitting data centers with peak electricity demand above 25 MW unless developers meet state clean-energy standards and file a community benefits agreement. Facilities that do not procure their own qualifying clean power must fund new nearby generation or pay into a newly created Ratepayer Protection Fund; the Massachusetts Department of Environmental Protection must develop an alternative compliance payment mechanism by December 31, 2026, with those payments credited to the fund (Source: wbur.org; techcrunch.com). Where New York's measure pauses construction outright and Maine's proposed one would have, the Massachusetts order conditions permits on procurement and cost allocation rather than halting them, and it acts through executive order rather than statute. The order's large-load rate work continues a Department of Public Utilities investigation established six months earlier by Executive Order 654, an energy supply order that names neither data centers nor artificial intelligence and whose Section 3(4) directs the DPU to investigate grid reliability and affordability "with the growth of large commercial customers" (Massachusetts Executive Order 654 — To Secure Massachusetts' Energy Future (March 16, 2026)). That order disclaims any cause of action at Section 5, so the Commonwealth's requirements reach a developer only through the permitting condition in Executive Order 658 and through DPU tariffs (Massachusetts Executive Order 658 — Responsible Data Center Development and Operations).
• California — Gov. Gavin Newsom signed seven data-center bills on September 21, 2026, requiring operators to disclose electricity use, water consumption, land use and workforce needs, giving local communities more information to assess proposed projects, requiring compliance with the state's energy procurement requirements, and aiming to keep the cost of new generation and grid upgrades from being shifted onto other ratepayers. Khara Boender, a director of government affairs at the Data Center Coalition, said the laws "create significant uncertainty" and are "likely to further limit data center development in California — an already declining market" (Source: reuters.com). The seven chapters (Chapters 434–440, Statutes of 2026) are AB 1577, which requires facilities of 10 MW or more to report energy data to the California Energy Commission and permit applicants to give local agencies energy and sound-level estimates; AB 2383 and SB 886, which direct the Public Utilities Commission to adopt data-center tariffs by January 1, 2028 assigning transmission-upgrade, generation and societal costs to data centers, with 10-year early termination fees; SB 1168, which directs the Commission to assess data-center rate structures; AB 2469, which bars local approval until the applicant supplies a water supply assessment, a drought-stage water scarcity plan and workforce disclosures and pays for needed water infrastructure; AB 2619, which requires sworn water-use estimates at business licensing and annual water-use reporting at renewal; and SB 887, which applies CEQA to data-center entitlements, bars categorical exemptions, and makes data centers meeting eleven conditions eligible for environmental-leadership streamlining (California data-center bill package — chaptered texts (Chapters 434–440, Statutes of 2026)). Inside AI Policy described the package as including requirements that developers report expected air-quality and water-usage impacts (Source: insideaipolicy.com); the chaptered texts contain water-use reporting but no air-quality reporting requirement, the nearest provisions being AB 1577's reporting of fuel burned in onsite generators and of sound levels (California data-center bill package — chaptered texts (Chapters 434–440, Statutes of 2026)).

As of mid-2026 no state had passed a comprehensive "AI data-center siting" statute; most action has run through PUC ratemaking and existing zoning. California's September 2026 package, which its Governor's office called the "most comprehensive data center laws in the nation," combines statutory reporting, tariff mandates and local permit conditions without a moratorium (Source: gov.ca.gov; California data-center bill package — chaptered texts (Chapters 434–440, Statutes of 2026)).

Maine moratorium and the Jay case study

Maine's L.D. 307 would have made Maine the first US state to impose a state-wide data-center moratorium (≥20 MW, through Nov 2027). Gov. Janet Mills vetoed the bill on April 24, 2026, stating that she supports the principle but rejecting the bill for failing to exempt a $550 million brownfield-redevelopment project on the former Androscoggin Mill site in Jay; she committed to establishing an equivalent coordination commission by executive order. Reportage (Source: nytimes.com); primary text L.D. 307 Veto Message — Gov. Janet T. Mills (Maine).

The Jay redevelopment is documented in detail by Data centers are coming for rural America — Abigail Bassett (The Verge, May 13 2026) (May 13, 2026). The Androscoggin Mill is a 1.4-million-square-foot former paper mill. Its owner, McDonald (JGT2 Redevelopment), initially pursued a sale to Godfrey Forest Products for oriented strand board production with roughly 150 jobs, but federal tariffs killed the financing; McDonald then pivoted to Sentinel Data Centers, a New York-based developer specializing in healthcare, financial, and hyperscale clients. Per McDonald, many bidders were "data center cowboys" — fronts claiming hyperscaler clients without backing. The companion Maine moratorium bill, LD 713 (Rep. Melanie Sachs, D-Freeport), would have paused permits above 20 MW for 18 months and passed both the Maine House and Senate before Mills vetoed it April 24, citing 125-150 jobs at Jay; Sachs responded, "Even if it's 30 jobs, that means a lot to Jay, then, okay, but you've swept away protections for 1.4 million Mainers for 30 jobs." A separate Maine law (April 2026) excludes data centers from some state-level tax breaks but leaves the door open for local municipalities to offer their own, pushing the incentive-bargaining problem down to small towns hoping for windfalls. On sequencing, McDonald did not inform the Jay Select Board of the data-center plan until late February 2026, days before Sachs's moratorium bill was scheduled for a floor vote; the Select Board voted 4-0 in favor in March.

Texas interconnection moratorium (August 2026)

Texas Governor Greg Abbott halted approvals for data centers seeking connection to the state grid on August 3, 2026, directing the Public Utility Commission of Texas and the Electric Reliability Council of Texas to audit projects advancing through ERCOT's interconnection queue before any are approved. Abbott's letter asks the agencies to require developers to disclose tax breaks received, power use and generation, water use and cooling operations, efforts to reduce impacts on local communities, and facility ownership, and states that any project failing to comply "must be denied connection to the Texas grid" (Source: texastribune.org).

The scale of the queue is the stated reason: ERCOT is tracking more than 1,800 projects representing over 474 gigawatts, more than five times the grid's record peak demand, and Abbott said approximately 90% of new power requests are data centers. ERCOT said the same day it is pausing its "batch zero" review. Abbott cited non-compliance with a PUCT survey measuring water and power usage — of 377 companies PUCT staff said it notified, 28 submitted responses. Texas Agriculture Commissioner Sid Miller called the directive "all hat and no cattle — empty political rhetoric wrapped in meaningless fluff." The action is an interconnection-conditioned disclosure requirement rather than a statutory moratorium of the kind Maine considered, and it comes from a state whose posture on AI regulation is otherwise deregulatory (State-Level AI Regulation).

Annexation and tax-jurisdiction patterns

Local government boundaries can separate the jurisdiction that captures a data center's tax revenue from the community that borders the site and bears its externalities. In Texas, Infrakey, an AI data-center developer, is advancing a proposed $10 billion, roughly 1-gigawatt campus on a 520-acre site near Ross, Texas (reported June 16, 2026). Under a December 2025 memorandum of understanding, the larger neighboring city of Lacy Lakeview is moving to annex the unincorporated parcel and could collect up to $50 million a year in taxes, while Ross — population roughly 200, which borders the site — has no authority to stop or shape the project (Source: fortune.com). The arrangement is a variant of the incentive-bargaining problem visible in the Maine case, where a small bordering community absorbs the local effects of a project whose tax benefits accrue to a separate jurisdiction.

A working paper posted to SSRN on August 7, 2026 and dated July 29 argues that this displacement is measurable in credit markets. Alex Hsu and Hala Moussawi of the Georgia Institute of Technology's Scheller College of Business find, using high-power transmission-line proximity and the staggered adoption of state tax incentives, that the fiscal costs of data centers fall on municipalities neighbouring the host rather than on the host itself: neighbouring municipalities face wider bond spreads and issue more debt when data centers arrive, while hosts show little response. The authors report that state tax policy determines where the cost surfaces — without incentives, credit markets reprice neighbouring municipal debt; with incentives, foregone state revenue shrinks transfers to school districts, which see rising debt issuance in exposed municipalities and a statewide widening of education bond spreads after adoption. The 46-page paper had not been peer-reviewed, and only its abstract was retrievable (Source: papers.ssrn.com).

Jobs evidence

A Ball State causal analysis by Hicks of 254 Texas counties found net long-term data-center job creation of approximately zero, with whatever long-term jobs existed at data centers offset by losses elsewhere in the same sector (Data centers are coming for rural America — Abigail Bassett (The Verge, May 13 2026)). Whether this net-zero-jobs result replicates outside Texas is unresolved; if it does, it would change the political economy of data-center bargaining. A Pew survey (April 2026) found 67% of planned US data centers headed to rural areas, with 39% headed to counties currently without any (Data centers are coming for rural America — Abigail Bassett (The Verge, May 13 2026)).

Federal legislation

Federal proposals on data centers emerged alongside the state and local wave. Sen. Bernie Sanders and Rep. Alexandria Ocasio-Cortez introduced a National AI Data Center Moratorium Act, reported June 10, 2026 (Source: route-fifty.com). Sen. Ed Markey and Rep. Don Beyer reintroduced a bill, made public June 11, 2026, that would require AI data centers to report their energy and environmental impacts and direct NIST to convene a measurement-standards panel (Source: insideaipolicy.com). On June 26, 2026, Rep. Frank Pallone (D-NJ) said a House Energy and Commerce subcommittee's approval of bipartisan legislation shielding residential power consumers from AI-data-center-driven price increases could build momentum toward the national data-center moratorium that advocates seek over environmental risks, linking the ratepayer-protection track to the moratorium push (Source: insideaipolicy.com).

Federal bills on local control and disclosure (August 2026)

Representative Tom Barrett (R-Mich.) introduced two data-center bills on August 20, 2026. The Protecting Local Control of Data Centers Act would bar federal agencies from overriding state and local land-use zoning and permitting laws for data centers, and from conditioning federal data-center construction funds on state or local governments surrendering that authority. The No Data Center NDAs Act would prohibit sitting lawmakers from signing nondisclosure agreements barring discussion of possible or planned data-center locations and resource consumption, with the House Ethics Committee and the Senate Select Committee on Ethics tasked with implementation (Source: nextgov.com). The second addresses the nondisclosure agreements that recur in the siting disputes described below.

Public lands (2026)

The Bureau of Land Management asked its state directors in September 2026 to identify public land suitable for data center development, giving them three days to compile the lists, according to two people familiar with the request in an account published September 11, 2026. The request lands on a bureau that has shed nearly half its staff under the Trump administration; one former official said their state's land and realty divisions carried 50% vacancy rates. No environmental review for a data center on public lands had been completed at the time of reporting, and a judge halted the first approved project on BLM land, near Boulder City, Nevada, after developers sought to rely on an environmental study prepared earlier for a solar array at the same site (Source: washingtonsun.com).

The episode joins the permitting-speed variable in the siting calculus above to the federal-capacity question: the reach of environmental review is the stated obstacle, and the agency that would conduct it is the one being asked to move in three days. The Nevada ruling is the first judicial test of whether a prior review at the same site satisfies that requirement for a data center.

Community opposition

Opposition has clustered around several grievances: environmental justice, where siting falls in lower-income and majority-minority areas (Memphis being a prominent case); noise, including continuous fan and transformer hum and diesel-test events; water, from evaporative cooling in drought regions such as Arizona, Georgia, and Texas; ratepayer cost-shifting, where residential customers subsidize new transmission and generation; and truck and construction traffic during buildout. Opposition has extended beyond wealthy suburbs, with environmental-justice coalitions and ratepayer advocates forming cross-class alliances against hyperscaler siting. It reached a coordinated national scale on July 18, 2026, when the grassroots group HumansFirst organized protests in at least 125 locations across the United States — the first coordinated national effort against the data-center buildout — targeting what the group calls an "unaccountable" expansion that some towns approved under nondisclosure agreements between local officials and developers (Source: yahoo.com).

Opposition also appears at the individual landowner level, where the decision point is a purchase offer rather than a permit hearing. In an account published August 15, 2026, Delsia Bare and Ida Huddleston — a daughter and mother in Maysville, Kentucky — refused offers totalling $26.48 million for family farmland sought for a 2.2-gigawatt hyperscale data center: 463 acres from Bare at $48,000 an acre and 71 acres from Huddleston at $60,000 an acre. Both had initially agreed before learning what the land was to be used for. The developer's identity, the project's status and the town's response were not recoverable from the retrievable portion of the article, which is paywalled beyond its opening (Source: wsj.com).

Data centers as an electoral issue (2026 midterms)

By September 2026 the opposition had moved from permit hearings into candidate positioning. Texas Governor Greg Abbott, who called Texas the "epicenter of AI development" less than a year earlier, is campaigning for re-election on expansive new data-center regulations, repeal of construction tax incentives and rural development restrictions; the state audit he announced on August 3, 2026 has effectively paused approvals for new grid connections. An August 2026 Texas Politics Project poll found 30% of respondents would support a data center in their community. Republican candidates in at least five states have hardened their positions (Source: reuters.com).

Industry spending has followed. The same account puts big tech's midterm political spending at more than $300 million, including more than $2.6 million in Texas from two super PACs tied to Anthropic, more than $4.8 million from super PACs linked to OpenAI co-founder Greg Brockman, and about $1.4 million from Meta (Source: reuters.com). The alignment cuts across the usual partisan reading of AI policy: the tightening is coming from Republican candidates in states whose statewide posture on AI regulation is otherwise deregulatory, and the spending is directed at a siting and ratepayer fight rather than at the frontier-model rules the same firms lobby on federally (AI Industry Lobbying — The 2025-2026 Political Offensive, Public Opinion on AI).

The dispute reached the G20 innovation ministerial in Chapel Hill, North Carolina. On the morning of Wednesday, September 2, 2026, hundreds of protesters gathered outside the security perimeter at the Carolina Inn on the University of North Carolina campus to object to AI data centers. Inside, OpenAI chief executive Sam Altman told Commerce Secretary Howard Lutnick in a fireside chat that AI is "incredible magic of intelligence in a bottle" and that "even with great gains in efficiency, we are going to need to build much more infrastructure, unless we want this to be a highly priced model, which we don't"; Jensen Huang of Nvidia and Alex Karp of Palantir attended, and Elon Musk spoke by conference call. Melisande Timblin, a 2005 graduate of the university, said data-center developers "sell them a big lie" on long-term employment. Reuters counted 142 demonstrations against data centers across 42 states in July 2026, and the Associated Press puts more than 3,000 data centers as planned or under construction in addition to roughly 4,000 existing ones (Source: fortune.com).

The foreign-influence claim

A second line of argument attributes part of the opposition to a foreign influence operation. In late August 2026 X announced it had identified 200 accounts linked to China that were helping spread anti-data-center talking points on the platform, part of a bot farm in which the company ultimately identified 200,000 accounts; days afterward President Donald Trump wrote on Truth Social that "China could not be happier with this anti Data Center movement." OpenAI had banned a cluster of users in June 2026 that it said were linked to the Chinese government and "testing narratives against AI infrastructure," while acknowledging in the same report that the accounts were exploiting discontent already present. Tyler Williams, vice president of intelligence at the social-media analytics firm Graphika, said in the week to September 6, 2026 that there remains very little evidence of significant foreign intervention and that most of the opposition the firm has tracked is homegrown (Source: wired.com). The claim sits alongside the polling above, which measures opposition rising across parties and across the trust-in-technology gradient, and alongside the ITIF finding that the strongest predictor of opposing a local data center is a negative view of AI itself.

Pennsylvania Governor Josh Shapiro, previously an advocate, has shifted position as the midterm elections approach, as has Texas Governor Greg Abbott (Source: wired.com).

Local opposition wave (2026)

Data Center Watch (datacenterwatch.org) maintains a weekly tracker of local political opposition to US data centers, and on July 16, 2026 the legal blog Chat GPT Is Eating the World launched an AI Data Center Bans & Lawsuits Tracker cataloging litigation and local bans aimed at stopping AI data centers (Source: chatgptiseatingtheworld.com).

The Information reported on August 9, 2026 that local measures restricting data-center development had passed 500 nationwide, with New York and Texas joining the pushback, and that local resistance accelerated over the summer — its retrievable abstract stating that in July alone more than 150 towns and counties passed temporary or, at which point the available text is cut off. The abstract characterises the count as a warning sign for AI companies including Anthropic, Google and OpenAI that are pinning compute plans on data-center expansion. The article body sits behind a subscription and was not read, so neither the completed sentence nor the tracker's counting method is available (Source: theinformation.com).

Episodes from Data Center Watch's Q1–Q2 2026 briefs include:

• Denver, CO (Feb 2026). Mayor Mike Johnston and City Council pursued a moratorium on new data-center construction within city limits, citing community pushback over CoreSite's expansion in Globeville-Elyria-Swansea (an environmental-justice-burdened neighborhood); the GES Coalition petitioned for stronger protections (Source: datacenterwatch.substack.com).
• Sangamon County, IL (Feb 2026). County rejected a 6-month moratorium and advanced a $500M CyrusOne campus to the County Board (Source: datacenterwatch.substack.com).
• Pacific, MO (Feb 2026). Beltline Energy withdrew its $16B rezoning request minutes before a packed public hearing (Source: datacenterwatch.substack.com).
• Apex, NC (March 2026). Town Council voted to draft a 1-year moratorium after the proposed 300 MW New Hill Digital Campus was withdrawn (Source: datacenterwatch.substack.com).
• Florida — Project Tango (March 2026). Wellington's mayor urged state lawmakers to back restrictions on large data centers near homes and schools as pushback widened (Source: datacenterwatch.substack.com).
• Appomattox County, VA (March 2026). Pre-permitting opposition emerged against Avaio's $3B Project Hercules (452 acres, 50 MW initial scaling to 1 GW+) (Source: datacenterwatch.substack.com).
• Archbald Borough, PA (March 2026). 5–0 denial of the "Project Scott" data-center campus after organized opposition (Source: datacenterwatch.substack.com).
• Festus, MO (April 2026). Voters recalled all incumbent city council members in elections that became a referendum on a proposed $6B data center (Source: datacenterwatch.substack.com).
• Lansing, MI (April 2026). Deep Green withdrew its data-center proposal after early-stage opposition organized by the activist group Sunrise Spartans (Source: datacenterwatch.substack.com).
• Imperial County, CA (April 2026). Approved a major data-center complex despite strong local opposition, a counter-example to the moratorium pattern (Source: datacenterwatch.substack.com). In July 2026, developer Sebastian Rucci separately proposed a $10 billion data center in the Imperial Valley, which would be California's biggest (Source: wsj.com).
• Seattle, WA (June 2026). The City Council voted 9-0 on June 9, 2026 to impose a one-year emergency moratorium on new large data centers, halting applications for facilities above 20 megavolt-amperes of electrical capacity and committing the city to study effects on the power grid, water supply, utility rates, and economy ahead of permanent rules. The vote followed testimony from more than 50 speakers — none in favor — and more than 98,000 constituent emails; the ordinance exempts roughly 30 existing data centers, each of which may still expand by up to 20 megavolt-amperes. Councilmember Debora Juarez, who sponsored the accompanying policy resolution, framed it as "Seattle's position on AI and data centers." The action placed Seattle, home to Amazon Web Services and Microsoft Azure engineering operations, among the largest US cities to pause the buildout, alongside Minneapolis, Denver, Baltimore, and Indianapolis (Source: geekwire.com).
• Monterey Park, CA and others (June 2026). Monterey Park residents voted roughly 90% on June 2, 2026 to ban data centers, described as apparently the first such municipal ballot vote in the US; Socorro County, New Mexico adopted a moratorium on June 10, 2026; and NCSL counted moratorium bills under consideration in 14 states (Source: route-fifty.com).

The Data Center Watch Q2 2025 report estimated that $98 billion in data-center projects were blocked or delayed in Q2 2025 alone, more than the cumulative total for all previous quarters since 2023, and described data centers as becoming an electoral issue in some jurisdictions, with Festus as a leading case (Source: datacenterwatch.substack.com). Per Heatmap News, cited in The AI Backlash Could Get Very Ugly — Lila Shroff (The Atlantic, May 13 2026), a record number of proposed data-center projects were canceled in Q1 2026 following local pushback — a sign that the political-acceptability constraint was becoming binding faster than the energy or land constraint.

Federal campaign politics (2026)

Siting opposition reached federal campaign strategy in August 2026. A memorandum from the National Republican Senatorial Committee dated August 18, 2026, headed "RE: Ohio Data Center Risk" and addressed to interested parties, states that data centers are "the anchor hanging around Jon Husted's neck" in the Ohio Senate race and that his opponent Sherrod Brown "has made them his de facto opponent, and no one is correcting the record." It quantifies Brown's effort at three distinct television advertisements and "more than 6,000 points on television," which it calls more than a month's worth of messaging, and reports that the committee's own polling shows the attack outperforming conventional messaging because it introduces new information on an emerging subject — "a sleeper issue for the entire election cycle" (NRSC — Ohio Data Center Risk (memorandum, August 18, 2026)).

Two features distinguish the memorandum from the local disputes above. It disclaims the party's own capacity to respond — "Campaigns or party committees can not fix the toxic brand of an entire segment of the economy" — and assigns the work to the companies building the projects, specifying the case they must make as "who benefits, who pays, and why a community should want one." And it treats Ohio as precedent-setting: if the issue remains potent in November, "elected officials everywhere will treat Ohio as their reason to work against data centers in the future," and a loss attributed to data centers would mean politicians "will not go near the next one" (NRSC — Ohio Data Center Risk (memorandum, August 18, 2026)).

The position it advances is the two-condition standard that also appears in state instruments of the same period — construction only where a community approves by local vote, and full cost recovery for power, water, and other utilities — contrasted with what it characterizes as a Democratic position of halting construction, which it says "will cause America to lose the AI race to China." Pennsylvania Executive Order 2026-05 (Data Centers), signed the same day, carries both conditions.

The memorandum's characterization of the race is contested by public polling. It calls the contest "a dead heat" and "still a toss-up" on the basis of unreleased private polling, while a Fox News poll released the week of August 10, 2026 showed Brown leading Husted by 8 points. Axios, which published the memorandum on August 19, 2026, reported that more than 4,000 US data centers were online with over 3,000 more proposed or in progress (Source: axios.com).

A second Republican document of the same cycle addressed candidates directly. Tony Fabrizio's memorandum of September 14, 2026 identifies the effective attack as one about household cost rather than artificial intelligence, which arguments from innovation, investment, competitiveness and jobs "do not answer ... because none establishes who pays." It advises candidates never to state support for data centers without conditions and to avoid the phrase "build data centers," which it says "can be clipped and put in an attack ad," proposing instead: "Build the AI infrastructure America needs—but only when the companies building it pay their own way, provide the power they consume, protect families from higher utility bills, and safeguard local water and natural resources." It lists enforcement mechanisms it treats as the core of the position — independent demand estimates, public disclosure of projected infrastructure costs, binding agreements assigning those costs to developers, bonds, continuing compliance reporting, penalties and authority to suspend operation or expansion — and advises against treating local approval or voluntary commitments as sufficient, stating that "process is not a substitute for protection" (Fabrizio Lee — Neutralizing Political Attacks on AI Data-Center Development Through Enforceable Conditions (memorandum, September 14, 2026)). The two memoranda converge on cost recovery but differ on local consent: the NRSC position conditions construction on approval "by a local vote of the people" (NRSC — Ohio Data Center Risk (memorandum, August 18, 2026)), while the option Fabrizio's survey tested and his memorandum sets aside was approval by local officials.

President Trump defended data-center expansion in an interview with Michael Cohen recorded the prior week and aired in full on August 23, 2026, saying that communities which decline a data center are "making a mistake" because the facilities create "tremendous amounts of jobs and money." He said the United States was leading China in AI "by a lot" and that data centers were not drawing power from the grid, adding that "they're making their own power plants." Some operators are building behind-the-meter plants that supply their facilities directly, while others draw from the grid. The comments run against the direction taken by state officials of both parties over the preceding months, including Texas Governor Greg Abbott's statement that data-center companies "dug their own grave," Pennsylvania Governor Josh Shapiro's restrictions, and New York Governor Kathy Hochul's one-year moratorium on hyperscaler data centers (Source: axios.com).

Political violence

On April 6, 2026, Indianapolis City Councilman Ron Gibson, who had supported a data-center project in his district, awoke after midnight to find 13 rounds fired into his home, with a note reading "No Data Centers" placed under his doormat. After the shooting, the council introduced a measure allowing local officials to keep their home address private. Whether other US municipalities adopt the home-address-privacy precedent for local officials voting on data-center projects is unresolved. See AI-Driven Political Violence for the framework treatment (Beyond Misuse: Artificial Intelligence, Grievance, and the Future Landscape of Political Violence — Yannick Veilleux-Lepage (Combating Terrorism Center at West Point, April 2026)) and The AI Backlash Could Get Very Ugly — Lila Shroff (The Atlantic, May 13 2026) for the mainstream-media uptake.

Soufan Center reporting (cited in Shroff) states that the most common online threats now involve "physical sabotage of proposed or operational data centers," and that a "How to Stop a Data Center" guide circulated in Michigan recommends demonstrating outside local officials' homes as an organizing tactic.

In Defiance, Ohio, a city of 17,000, residents gathered signatures for a November 3, 2026 ballot measure to ban all but the smallest data centers, and city officials announced a six-month moratorium on new project approvals, in an account published September 19, 2026. State tax records released in summer 2026 showed Ohio forwent more than $2 billion in sales-tax revenue through data-center incentives in 2024 and 2025, and a Bowling Green State University poll dated August 31, 2026 found 71% of Ohioans would support a temporary ban on new data-center construction. Governor Mike DeWine has suspended new applications for data-center tax exemptions pending legislative reforms that lawmakers did not act on before the summer recess. The group Conserve Ohio is pursuing a 2027 state constitutional amendment barring data centers drawing more than 25 megawatts (Source: reuters.com).

Financing and new entrants

JPMorgan and other banks struggled to syndicate billions of dollars in loans tied to Oracle data-center leases in Texas and Wisconsin (April 24, 2026) (Source: wsj.com). See Circular Financing in AI.

In Europe, Helsinki-based Verda (formerly DataCrunch) raised €100M in debt and equity (April 24, 2026) to build out what it pitches as Europe's first AI cloud hyperscaler (Source: sifted.eu). SpaceX's proposed orbital data center drew "unproven" dismissals from multiple cloud-infra analysts (April 22, 2026) (Source: datacenterknowledge.com).

The buildout side continued to scale against the moratorium wave. On June 9, 2026, OpenAI entered advanced negotiations to lease a proposed 10-gigawatt data-center campus on Department of Energy land in southern Ohio — a project that could cost at least $500 billion to build at current prices, structured as a 20-year lease with SB Energy (a SoftBank unit) developing the facility and Nvidia expected to supply hardware and guarantee both OpenAI's lease and SB Energy's financing (first phase ~2028) (Source: theinformation.com; reuters.com). The same week, Broadcom, Apollo, and Blackstone launched a $35 billion "Big Sky" fund to finance more than 20 gigawatts of Broadcom-designed capacity through 2028, including projects tied to Anthropic and OpenAI (Source: ft.com). The use of federal DOE land for the Ohio campus marks a contrast with the local-government moratoria accumulating elsewhere. See Private Credit & AI Infrastructure.

Policy considerations

Several structural features shape the policy debate. Power is geographically fixed: unlike chips, data centers cannot be routed through shell companies, which makes siting a usable policy lever (see Compute Governance). Ratepayer cost allocation is an active fight, with the Louisiana and Ohio cases positioned to set precedent for whether hyperscalers internalize their grid externalities. NEPA and Clean Air Act enforcement capacity, rather than statutory authority, is described as the binding constraint, with the Memphis turbine case illustrating that on-the-ground enforcement is the weak point. Federal streamlining may also conflict with state and local authority: the America's AI Action Plan's NEPA push may run up against state PUC and local siting-board authority.

The federal executive branch has signaled a light-touch posture on the environmental dimension. EPA Administrator Lee Zeldin downplayed the agency's role in regulating data-center environmental harms in June 12, 2026 reporting, favoring best-practices guidance to state permit writers over direct federal regulation (Source: insideaipolicy.com). Industry water-use disclosure remained sparse: Amazon disclosed on June 11, 2026 — reportedly for the first time — that its data centers consumed 2.5 billion gallons of water in 2025 at 0.12 liters per kilowatt-hour, down 2% from 2024, claiming roughly seven times the industry-average water efficiency (Source: theverge.com). See AI Environmental Impact.

Relationships

• supports: Americans' Views of Data Centers Have Turned More Negative (Pew Research Center, September 2026) — Pew's January–August 2026 tracking of views of data centers and comfort with a new local facility.
• depends-on: Compute Governance, AI Environmental Impact
• supports: RAND — AI's Power Requirements Under Exponential Growth (2025) — power as the emerging binding chokepoint
• related: xAI, Microsoft, Nvidia & TSMC — AI Compute Infrastructure, America's AI Action Plan, Public Opinion on AI
• supports: Public First / ITIF, \"The opposition to data centers\" (polling deck, August 2026) — attitudes to AI as a predictor of siting opposition
• supports: AP-NORC/EPIC 2026 Energy Survey: AI, Data Centers, Climate and Energy — concern about local electricity prices, water and outages; bipartisan support for developer-pays grid upgrades
• related: A New Counterintelligence Map for the AI Age (Feith, HPSCI, 2026-09-02) — siting adjacency and load concentration as counterintelligence and grid-security questions
• supports: Fabrizio Lee — Neutralizing Political Attacks on AI Data-Center Development Through Enforceable Conditions (memorandum, September 14, 2026) — Republican pollster's advice that unconditional support is an electoral liability and conditional construction the majority position; related: Tony Fabrizio
• related: California data-center laws (2026), California data-center bill package — chaptered texts (Chapters 434–440, Statutes of 2026) — California's seven-chapter statutory package of September 2026 (reporting, tariffs, water and workforce permit conditions, CEQA)

Wiki Sources

• RAND — AI's Power Requirements Under Exponential Growth (2025) — quantitative forecast; grid capacity additions 3× historical.
• Epoch AI — How Much Power Will Frontier AI Training Demand in 2030? — per-run power trajectory.
• AI Environmental Impact — water, emissions, grid strain context.
• NRSC — Ohio Data Center Risk (memorandum, August 18, 2026) — party-committee primary text on siting opposition as a federal campaign issue.
• Public First / ITIF, \"The opposition to data centers\" (polling deck, August 2026) — fifteen-country comparison, January-to-July 2026 trend, policy and conjoint testing.
• AP-NORC/EPIC 2026 Energy Survey: AI, Data Centers, Climate and Energy — AP-NORC/EPIC national survey of July 2026: local-impact concern, expected benefits and data-center policy support by party, region, age and education.
• Fabrizio Lee — Neutralizing Political Attacks on AI Data-Center Development Through Enforceable Conditions (memorandum, September 14, 2026) — Fabrizio Lee messaging memorandum to GOP candidates (September 14, 2026): unattributed survey crosstabs and a conditions-first position.
• California data-center bill package — chaptered texts (Chapters 434–440, Statutes of 2026) — chaptered texts of California AB 1577, AB 2383, AB 2469, AB 2619, SB 886, SB 887 and SB 1168 (Chapters 434–440, Statutes of 2026).