Community Relations, Opposition & Social License

For most of the cloud era, community relations sat near the bottom of the development checklist: a few open houses, a chamber of commerce meeting, a press release about jobs. In 2026 it is a hard gate on the critical path, one of three things, with power and water, that can kill a project outright. The shift is measurable. Data Center Watch counted at least 75 projects, worth roughly $130 billion, blocked or delayed in the first three months of 2026. On its own, that quarter roughly equalled the count for all of 2025. The group's second-quarter count was at least 45 projects worth nearly $68 billion, and because the quarterly tallies are not deduplicated, the two cannot simply be added. By September 21, 2026 Bloomberg counted 843 active local opposition groups in 49 states, against 396 at the end of 2025 (Data Center Watch / 10a Labs; Bloomberg, 2026). This is no longer a handful of angry neighbors. It is a movement with a shared playbook, model ordinances and a grievance that crosses party lines.

The chapter identifies the approval pressure that each of five opposition vectors can create: ratepayer cost-shift, water, noise, property values and grid strain. It adds a sixth, air quality, that your own on-site gas creates. It then walks the rezoning battlefield and the moratorium wave that is rewriting by-right zoning out from under developers, and prices the toolkit that buys consent: community benefits agreements, host fees, PILOT structures and decommissioning bonds. It ends with the procedural work that decides whether an approval survives appeal: early disclosed engagement versus stealth assembly, and litigation-hardening of the record. → siting hierarchy in Chapter 3.1; permitting critical path in Chapter 3.9.

Electricity bills recruit opponents well beyond the site's neighbors. When a region's wholesale capacity price spikes and the cost flows through to residential rates, every household in the territory has a stake in your interconnection, whether or not they live near your site. PJM's capacity auctions are the canonical case. The December 2025 auction cleared at a record price, and data-center load accounted for roughly 40%, about $6.5 billion, of its capacity costs. Monitoring Analytics estimates a further ~$6.3 billion (38%) of the July 2026 auction's ~$16.4 billion, the second consecutive auction near two-fifths (PJM market monitor; Utility Dive, 2026). The July auction hit its $325/MW-day cap for the 2028/29 delivery year and still fell 6,831 MW short of the reliability requirement, which the December auction had missed too (PJM, July 14, 2026). The retail effect is utility-specific. In Virginia, the State Corporation Commission approved Dominion base rates that put $11.24 a month on a typical residential bill in 2026 and $2.36 more in 2027 (SCC, November 25, 2025). That figure covers every cost the biennial review approved, and the order does not pin it on data centers alone.

Federal emergency orders add a cost no auction sets. Since May 2025 DOE has kept Eddystone Units 3 and 4 in Pennsylvania (760 MW) available past their retirement date through a chain of Federal Power Act §202(c) emergency orders, each about 90 days long. The sixth, Order 202-26-40 of August 21, 2026, runs from August 23 to November 20, 2026. DOE justifies the orders with a model of about 25 GW of new PJM load, 15 GW of it data centers, so the bill argument now reaches federal orders too (Data Center Knowledge, August 2026). Constellation, the plant's owner, says PJM charges Eddystone's costs to load the way it allocates capacity costs. The orders can also fall in court: on September 11, 2026 the D.C. Circuit vacated DOE's initial order keeping Michigan's Campbell coal plant open, and challenges to the later orders extending it are pending. Underwrite these orders as a charge on load, never as capacity a project can count on.

So cost allocation can build opposition far outside the host community, and across party lines. The polls agree. Roughly seven in ten Americans oppose an AI data center in their area, and 78% told Consumer Reports they worry their household energy bills will go up because of new data centers (Gallup, March 2026 fieldwork; Consumer Reports, November 2025 survey). A grievance like "my bill went up so a trillion-dollar company can run AI" does not belong to one party. Gallup found majorities of Democrats, independents and Republicans opposed, though strong opposition ran highest among Democrats. The legislative response has been a flood: more than 300 data-center bills filed across 30-plus states in the first six weeks of 2026, with the policy center of gravity shifting from incentives to ratepayer protection and cost allocation (MultiState; ArentFox Schiff, 2026). In Washington, the White House launched a voluntary Ratepayer Protection Pledge on March 4, 2026, and on July 23 it brought in more than 200 additional signers, from utilities, developers and cooperatives to states. The pledge asks companies to pay for their own energy and delivery infrastructure. The protection a ratepayer can actually enforce still sits in the tariff and service contract that govern each load (The White House, 2026). California put that protection into statute on September 21, 2026. SB 886 directs the CPUC to set data-center tariffs by January 1, 2028, with an early-termination fee for a customer that leaves within 10 years of interconnection. AB 2383 and SB 1168 add minimum payments, upfront collateral and transmission and distribution cost allocation (Chapter 3.3).

A generic outreach program loses to a specific objection. The six recurring vectors reach different distances, trigger different permits and cost different amounts to mitigate. Diagnose which vector is dispositive in this jurisdiction and spend the concession budget there, rather than scattering goodwill evenly across all six.

Two notes on reading the table. First, opposition vectors can compound. When a Gallup Panel web survey in April 2026 asked opponents why, the largest group, about half, objected to resource use, with water and energy named by 18% each. A well-built ratepayer agreement leaves those objections standing. Second, the water vector depends on geography. In a water-rich, cool climate it is a manageable permitting condition. In a stressed basin, such as the arid West or a drought-prone metro, the cooling-water draw can decide the approval. Two answers hold up. You can design the water out, with closed TCS and FWS loops and dry/air-cooled final heat rejection. Or you can contract for reclaimed water, provided its availability in drought, the treatment it needs and its continuity of service all fit what the cooling duty requires. Promising stewardship around an evaporative plant is not one of them. → water sourcing and the design-out decision in Chapter 3.7.

The table's grid-strain fix creates a sixth vector, air quality. On-site firming and behind-the-meter gas answer the blackout grievance by putting turbines or engines near homes and schools, and the opposition follows them to the air permit. xAI is the case to study. Aerial images released by the Southern Environmental Law Center in April 2025 showed about 35 gas turbines at xAI's Colossus site in Memphis, where the company had applied to permit 15. The county approved 15 in July 2025, and an appeal by the NAACP and local groups was dismissed as moot in December. Colossus 2 runs on a gas plant in Southaven, Mississippi, where the state permitted 41 turbines in March 2026. In April the NAACP appealed that permit and sued xAI under the Clean Air Act (Southern Environmental Law Center, May 2026).

An air permit for a gas plant usually goes out for public notice and comment, often draws a hearing and can be appealed, so treat it as an approval vote. File it before the turbines arrive, fit NOx and CO controls such as SCR and oxidation catalysts, accept enforceable run-hour limits and a dated end for any bridge plant, and publish the monitoring data. Skip those steps and an appeal of the permit can stop the plant that your first-power date rests on. None of these controls cut carbon: the plant's combustion stays the largest source in your Scope 1 inventory until it shuts down, another argument for the dated end. → air-permit regimes in Chapter 3.9; the carbon cost of on-site gas in Chapter 15.3.

The most consequential structural change of 2026 is the erosion of by-right zoning. For years the fastest sites were those where data centers were a permitted use in an industrial or technology district. You could pull a building permit without a discretionary public vote, which removed the biggest place for opposition to gather. Communities have closed that door. California closed it statewide on September 21, 2026: SB 887 ends categorical CEQA exemptions for data centers, so even a project on by-right land must prepare at least an initial study, a discretionary step that opponents can contest (Chapter 3.9). Loudoun County, the densest cluster in the world, now requires special-exception approval for data centers in its IP, GI and MR-HI districts (March 18, 2025 amendment). Applications accepted before February 12, 2025 that qualify under the county's grandfathering resolution can keep by-right processing. Phase 2, which covers substation, design and other standards, was still a proposal as of October 12, 2026, and the Board expects to act on it in July 2027. On September 15, 2026 the Board voted 7-1-1 to pause final votes on legislative data-center and substation applications for up to 12 months, with the formal resolution scheduled for October 20. Filing and Planning Commission processing can continue, and grandfathered and administrative applications are outside the pause. Cities are pausing too. On October 6, 2026, Raleigh, North Carolina, paused data centers of 10,000 sq ft or more that draw over 5 MW: for six months it will neither accept nor review their applications while it drafts new rules.

Where the door has not been re-zoned shut, it has often been frozen. Temporary moratoria, typically 6 to 18 months, buy a jurisdiction time to write a permanent ordinance, and a moratorium can outlast a project's option period and its interconnection-queue position. In July 2026 the ladder reached the state level. Under New York's Executive Order 62 (July 14, 2026), DEC holds every incomplete discretionary-permit application for a new or expanded data center that can draw 50 MW or more. The hold has no fixed end date. It ends only when the Department of Public Service has submitted the final report of the Generic Environmental Impact Statement (GEIS) together with the findings statement that accompanies it. It does not reach applications DEC had already deemed complete, local-government permits, or facilities used mainly for manufacturing, research, education or medical care. The PSC published its draft GEIS scope on September 17 and is taking comments until October 20, 2026. The governor's proposals for the 2027 session would also end hyperscale data-center tax subsidies, sales-tax exemption included.

Some states now make the state permit wait on the local vote. Pennsylvania's Executive Order 2026-05 (August 18, 2026) ties DEP permitting to local approval for data centers with peak demand above 25 MW whose applications were filed after the order. Developers who sign the state's GRID agreement can get rolling DEP review, but their qualifying permits issue only once local approvals are documented and the project fits the local comprehensive plan. Those who do not sign need the local approvals in hand before DEP review starts. The order also forbids agencies under the governor's jurisdiction to sign data-center NDAs. → Chapter 3.9 for the permit sequence.

A discretionary approval is a political event with a public comment period, a planning-commission recommendation and a governing-body vote. That is three places where an organized coalition can defeat you, and the final vote can be challenged in court (in Virginia, within 30 days). The fork is uncomfortable: chase the shrinking set of remaining by-right jurisdictions (faster, but increasingly scarce and itself a target for the next moratorium), or commit to the discretionary path and invest in winning the public process. In 2026 the by-right shortcut is a depreciating asset. The durable answer is to assume a discretionary process everywhere and build the competency to win it.

A benefits package can address the specific local costs a project imposes, but it does not by itself establish consent. The structure of the value determines whether commitments are enforceable, reported and remediable, or merely cash moving toward a hostile audience. Four instruments dominate, and they are not interchangeable. A community benefits agreement (CBA) is a binding, often legally enforceable contract between the developer and the host that funds specific community priorities, such as broadband, parks, workforce training or a fire-station upgrade, and increasingly carries clawback and enforcement teeth (FAS, 2026). A host fee is a recurring per-MW or per-square-foot payment to the local government for the privilege of operating, distinct from taxes. A PILOT (payment in lieu of taxes) substitutes a negotiated payment schedule for ordinary ad-valorem property tax, usually via an industrial-development authority that holds title and leases the asset back. It trades the host's near-term tax base for a longer, more certain payment stream. A decommissioning bond is a surety or cash reserve that guarantees cleanup and equipment removal at end of life, so the community is not left with a stranded windowless box.

The recurring error, and an expensive one, is to lead with the PILOT and call it a community benefit. An abatement reduces the revenue the host would collect if it taxed the same project in full. It may still raise revenue compared with an empty site, but opponents will use the first comparison. When a developer's pitch is "we will create jobs and pay reduced taxes for 30 years," the opposition reframes it as a subsidy, and the cost-shift coalition that already objects to your power draw now objects to your fiscal footprint too. The defensible structure pairs any abatement you need with a CBA and host fee that elected officials can point to as net-new value and that beneficiaries can see. Make the benefits binding and enforceable, because a promised benefit the community cannot compel is, accurately, just a press release. Abatement durability carries legislative risk. Texas had not repealed its data-center sales-tax exemption as of October 12, 2026. But the governor has asked the 2027 Legislature to repeal it, senators have questioned whether existing exemptions should continue, and the Comptroller is clawing back benefits from audited facilities that missed the program's requirements (Texas Tribune, July 27, 2026). Other states have already acted. North Carolina repealed its sales-tax exemption on electricity used at certified data centers, already-certified sites included, from the first billing period starting on or after August 6, 2026 (S.L. 2026-41). Illinois has paused new incentive agreements by executive action since July 1, 2026, without changing the statute. → fiscal structuring and incentive durability in Chapter 3.10.

The most consequential community decision is made before the first public hearing: how the project becomes known. The default developer instinct is stealth assembly: option land through anonymous LLCs and code-named projects ("Project Bluebird") to avoid tipping off speculators and to keep land prices down. The logic is sound on land cost and badly wrong on social license. Against organized, networked opposition, the discovery moment becomes the opposition's founding event. Residents learn at once that a multi-hundred-megawatt facility is coming, that a faceless shell company concealed it, and that their officials may have been negotiating an abatement in the background. The story that forms, secrecy plus subsidy plus a fait accompli, is nearly impossible to recover from.

The opposite posture, early disclosed engagement, costs more in land because you reveal demand and your name, but it front-runs the narrative. You arrive with the ratepayer-protection story, the CBA term sheet, the water and noise commitments and an accountable operator, and you give the community a seat before the design is frozen rather than after. Stealth protects your acquisition budget and surrenders the narrative; disclosure protects the narrative and costs you on land. Price the land premium that early disclosure will cost you. Then estimate how much disclosure cuts the expected loss from a failed or delayed approval, working from this site's acquisition terms and approval risks. When the premium is the smaller number, favor disclosure. In a quiet, business-friendly jurisdiction with by-right zoning still intact, stealth can still be the rational choice. Decide it deliberately, not by default.

Assume the appeal. In a contested 2026 approval, the governing-body vote is the midpoint, not the finish line. Organized opponents litigate, and the question becomes whether your approval survives judicial review. That is decided in the record you build along the way, not on the night of the vote. Litigation-hardening means constructing an administrative record so complete and procedurally clean that a reviewing court has no defensible basis to overturn it.

The components are unglamorous and decisive. Complete, conforming applications leave no procedural gap an opponent can exploit. The governing body must write express findings of fact into the approval resolution, explaining on the record how the project meets each ordinance criterion. A one-word "approved" is far easier to attack in court than a reasoned approval. Conditions you can actually meet matter because every condition is a future compliance obligation and a future place to be sued for breach. A condition written to placate the room but impossible to satisfy is a self-inflicted wound. A clean procedural trail, with proper notice, honored comment periods and no ex-parte irregularities, matters because the cheapest way to lose an appeal is a process foul that never touches the merits. The concessions you negotiate to win the vote and the conditions you accept to harden the record are the same instruments: a binding CBA, a frequency-domain noise spec, a sized decommissioning bond and a ratepayer-protective tariff both buy consent and make the approval defensible. → permitting record and environmental review in Chapter 3.9.