The new NIMBY is real. The energy industry already has the playbook.

Two recent pieces on this site have argued that the binding constraints on AI-infrastructure deployment have moved. Capital is no longer scarce — institutional dry powder is chasing a finite pool of closeable deals. Power is no longer strictly the technical problem — the site-selection edge has shifted to interior US markets with behind-the-meter generation and to what the market is now calling powered land. Those two shifts are structurally correct and are the frame most operators are using to explain the vintage.

There is a third constraint. It is not new. It has been strengthening quietly for two years. It has now become impossible to ignore.

Community opposition — the coalition of homeowners, farmers, environmental councils, ratepayer advocates, tribal and iwi bodies, and municipal officials who decide whether a specific parcel gets rezoned, permitted, and connected — is now stopping data-centre projects at a pace that changes the underwriting maths. And it is doing so most sharply in the specific markets that the industry considered mature, because those are the markets where the cumulative footprint has crossed a visibility threshold. The reason this looks like a surprise to the data-centre sector is that the sector arrived at this fight without the playbook the energy industry has been refining since the 1980s.

The state of play

The most visible pushback is in Northern Virginia, the sector's largest single cluster. In March 2025 Loudoun County amended its zoning ordinance to remove by-right approval for new data-centre applications; every project now runs through public hearings and discretionary votes. In Prince William County, the proposed Digital Gateway — pitched as the world's largest data-centre campus, adjacent to Manassas National Battlefield — was voided by the Virginia courts on procedural grounds. Compass Datacenters withdrew from Prince William entirely. QTS is pursuing a last-minute Virginia Supreme Court appeal. The Piedmont Environmental Council and allied groups are pushing for a statewide moratorium.

The rate-impact fight is running in parallel. Dominion Energy's initial 2026 request would have raised residential rates by roughly 25 per cent this year and approximately 50 per cent next year; regulators set aside the ask. Whatever the eventual settled rate schedule, the political salience of "AI for billionaires while my bills go up" is now a permanent feature of the Virginia landscape.

The pattern is global. Ireland's Commission for the Regulation of Utilities ended four years of ambiguity in December 2025 by publishing a Large Energy User Connection Policy that now requires any scheme above 10 MVA to build behind-the-meter generation sized to 100 per cent of its grid connection, site in unconstrained parts of the grid, and match 80 per cent of annual electricity demand via Irish renewable investments. Data centres already consume roughly 22 per cent of Ireland's electricity, forecast to rise to 31 per cent by 2034. A joint Friends of the Earth Ireland and Beyond Fossil Fuels analysis estimated the average Irish household paid approximately €360 in additional electricity costs between 2015 and 2023 as a result of grid pressure attributed in part to data-centre load.

The Netherlands has kept its restrictions on hyperscale sites over 70 MW in force since the original Amsterdam moratorium. Microsoft's Hollands Kroon campus has been in a running dispute with local farmers over agricultural-land encroachment and water consumption through recent Dutch droughts. Denmark's grid is now flagged by CNBC and by domestic operators as approaching a data-centre-driven reckoning.

None of this cancels the sector's real economic contribution. Data centres create construction jobs, meaningful long-term operations employment for a small number of high-skill workers, materially expand host-jurisdiction tax bases, and — increasingly — anchor local digital sovereignty. Those benefits are real. What has changed is that the communities living next to the infrastructure have started asking, with more organisation and more legal support than the sector expected, whether the benefits reach them at all.

What the sector didn't plan for

The uncomfortable truth for the data-centre industry is that community engagement is not a new capability the sector needs to invent. It is a mature discipline the adjacent industry has been practising for four decades.

I have been in the trenches of this work myself. Early in my energy career I worked for a 100 per cent renewable New Zealand generator and retailer, at the company responsible for the country's first commercial wind farms. Getting those projects consented was never an engineering question and was rarely a finance question. It was an honest-conversation question, held over months of planning and meetings with the local communities and the mana whenua whose land bordered or fell within the project boundary. The team got each of the projects across the line. None of them were quick, and none of them were done by stealth. All of them were done by turning up, telling the truth about what the project would do, and negotiating around what the community and iwi could live with — because that was the process the sector already understood was the price of building at all.

That pattern is not New Zealand-specific. Energy developers globally — wind, solar, transmission, gas pipelines, mining, geothermal, hydro — have been losing local referendums, being invited to and disinvited from council meetings, negotiating host community benefit agreements, running open-days at community halls, funding independent rate-impact studies, and building formal liaison committees since the 1980s. In New Zealand and Australia, statutory mana whenua and Traditional Owner engagement processes make early consultation a legal precondition of project consent, not an optional courtesy. In the UK and continental Europe, statutory community benefit funds and rate mitigation are standard project deliverables. In the US, transmission and pipeline developers have spent thirty years learning that the only way to get a linear infrastructure project across a hostile jurisdiction is to arrive early, share benefits explicitly, and design the project around what the community can live with.

Data-centre developers arrived at exactly the same problem from a completely different heritage. Most of them are commercial real estate operators by muscle memory. The commercial real estate playbook is quiet land aggregation through shell entities, sudden public announcement, permit chase, and construction inside the twelve months before organised opposition can form. That playbook worked for the first decade of hyperscale deployment because the projects were small enough, discretionary enough, and remote enough that no coalition formed against them. It does not work in 2026. The projects are now visibly large, cumulatively visible even where individually justified, and often located in jurisdictions where the opposition coalition is well-funded and legally sophisticated. Stealth is now the reason a project dies at zoning, not the reason it survives.

The developers who are succeeding — Comstock and Jericho in Oklahoma, Prometheus Hyperscale in Wyoming, the interior-US operators who pair site selection with municipal partnership and behind-the-meter generation — are the ones who arrived with the energy-sector playbook or hired it in. The ones who are being blocked are the ones still operating as commercial real estate.

What actually works

The winners look structurally similar across markets.

They engage early and publicly, before the first zoning application. They fund independent studies on rate impact, water withdrawal, traffic, noise, and grid load, and release the results whether they flatter the project or not. They negotiate host community benefit agreements that include rate stabilisation contributions, workforce guarantees for local hiring including apprenticeship pipelines, infrastructure upgrades that the host municipality was going to have to fund anyway, and where relevant, direct partnership with local educational institutions on training programmes tied to the campus. They site preferentially on brownfield and industrial-zoned land where the trade-off with residential and agricultural uses is already resolved, and they use behind-the-meter generation not just as a time-to-power tool but as a structural way to keep the load off the ratepayer grid. Where the offtaker permits, they announce with a named counterparty already in place, so the community can see the actual economic proposition, not a speculative one.

None of these moves are cheap and none of them are optional. Together they cost between 3 and 10 per cent of project capex depending on jurisdiction — a materially non-zero number that, if not built into the underwriting model, makes a marginal project unbuildable. Priced in from the start, they are the difference between a project that opens on schedule and a project that runs three years late in litigation.

Regulatory navigation is now a core operator competency, not a legal-department exercise. The developers, sponsors and operators who understand this and staff for it will build. The ones who continue to treat community and regulatory work as friction to be minimised will underwrite projects that do not close.

What it means for capital

For an infrastructure allocator, the third constraint raises the qualification bar again — precisely where the qualification bar was already the binding variable on returns this vintage.

A deal is now closeable only if all four sides are aligned: the technology is bankable, the power path is credible, the capital structure is fit for purpose, and the community and regulatory position has been mapped, engaged, and de-risked. Mid-market infrastructure funds without local execution muscle — without on-the-ground staff who can navigate a specific county's planning politics, without long relationships with municipal officials, without the operational capability to design and deliver a host community benefit agreement — are structurally further disadvantaged. Sub-deployment risk concentrates further at exactly the funds that were already exposed to it.

The best returns will accrue to the platforms that pre-qualify for community viability at the sourcing stage, before any capital is committed and before a fund is asked to bid. The vintage is not going to be won by the fund with the sharpest term sheet on a marginal community-adverse project. It will be won by the funds seeing a curated flow of projects that have already cleared the community qualification.

What we built to solve this

This is the work we designed the SitePower matching layer to do — and community and regulatory de-risking is now an explicit part of the qualification gate, alongside the technology, site and capital pre-flights that were there from the start. We do not lobby, we do not represent communities, and we do not place capital or take fees from allocators; SitePower's economics come from sponsor-side carry, co-investment, and operating roles inside the deals we help shape. For financiers, developers and tech companies, what we deliver is a filter that surfaces deals where the community and regulatory position has been designed in from the start — the kind that clear zoning, close on time, and open. For anyone building or funding at the scale this vintage requires, the qualification gates are at sitepower.ai/apply.