Eagle Mountain, Utah is a city of roughly 60,000 people about 30 miles southwest of Salt Lake City. It's also home to one of the largest hyperscale datacenter campuses in the western United States: Meta's Eagle Mountain Data Center, announced in 2018 and continuously expanded since. The campus has become a focal point for a national conversation that's no longer hypothetical: communities are increasingly organized, vocal, and effective in pushing back on new datacenter construction.
What's happening in Utah is happening — in different flavors — in Northern Virginia, central Ohio, rural Texas, southern Arizona, and a growing number of US towns where hyperscalers have announced billion-dollar campuses. For vendors, developers, and the operators themselves, understanding the substance of the pushback is now a strategic concern. Permit timelines, site selection, and community-relations costs are being reshaped in real time.
This post lays out the eight specific concerns driving the backlash, why AI infrastructure is amplifying all of them, and what it means for where datacenters can realistically be built going forward.
The Eagle Mountain Story in Brief
Meta's Eagle Mountain campus was first approved in 2018 with public commitments to local jobs, sales-tax revenue, and renewable-energy investment. The site has expanded multiple times since, with the most recent Deseret News coverage of the 2024 expansion drawing organized community opposition that didn't exist when the campus was first approved.
The concerns voiced by Eagle Mountain residents — and by groups like the Utah Rivers Council and Utah Taxpayers Association — track closely with concerns heard at hyperscale sites across the country. Water consumption during a multi-year drought. Power-grid impact on residential rates. Promised jobs that materialized as far fewer permanent positions than projected. Property tax abatements that shifted the cost of local services onto homeowners. Noise from cooling chillers running 24/7.
Utah is not the worst case — but it is one of the most public. Similar fights are ongoing in Loudoun County VA, Prince William County VA, central Ohio, and rural Georgia.
The 8 Concerns Driving Community Pushback
1. Water consumption (the loudest issue in arid regions)
Datacenter cooling systems — particularly traditional evaporative cooling and certain hybrid systems — can consume millions of gallons of water per day at full load. In drought-affected regions like Utah, Arizona, Nevada, and parts of Texas, the math is politically explosive. A single hyperscale campus can consume more water than a small city.
Operators counter that water consumption per compute-unit has improved dramatically, that newer designs use closed-loop or air-cooled systems with near-zero water, and that hyperscale water use is small relative to agriculture. All true — but communities have largely rejected the framing. When a small town's reservoir levels drop while a Fortune 500 company's water permit is being expanded, the optics outweigh the engineering.
2. Power demand and grid strain
A single AI training campus can require 500MW to 2GW of grid capacity — equivalent to a small city's entire residential load. Utilities serving these facilities face genuine planning challenges, and the cost of grid upgrades has been shifting onto residential ratepayers in several states.
Ohio's PUCO recently approved a tariff structure that ties datacenter cost responsibility more directly to the new infrastructure they require — a model other states are studying. Virginia is in a similar conversation, with Dominion Energy's SCC filings documenting load growth projections that have been politically charged for two years running.
3. Noise from chillers, generators, and transformers
A datacenter at full operation produces a low-frequency hum from cooling infrastructure that doesn't quiet down at night. Monthly generator load-testing is louder still. Residents within a half-mile to a mile of campus boundaries have filed complaints in nearly every major metro — Loudoun, Eagle Mountain, Mona, Mecklenburg County VA, and elsewhere. Operators have responded with sound-attenuation walls and acoustic treatment in newer designs, but retrofit options for older campuses are limited.
4. Tax incentives without commensurate local benefit
States and counties have used aggressive property-tax abatements, sales-tax exemptions, and direct cash incentives to attract datacenter construction. The economic-development logic was simple: massive capex, premium-wage jobs, and downstream services would more than offset the foregone tax revenue.
The pushback argues the math doesn't actually work. A typical hyperscale campus may employ 50-300 permanent staff after construction — far fewer than the construction-phase numbers used in pro-incentive presentations. Construction work is often performed by out-of-state crews. And property tax abatements lasting 10-20 years mean that local schools, roads, and emergency services rely on residential property taxes to make up the difference.
The Utah Taxpayers Association and similar groups in other states have begun publishing analyses comparing promised vs. delivered economic impact. The numbers rarely match.
5. Property values and viewshed
Hyperscale campuses are visually massive — multiple buildings totaling a million square feet, with cooling infrastructure, generator yards, fuel storage, and security perimeters extending across hundreds of acres. Residents who bought homes with rural or suburban views before approval have, in some cases, watched a horizon transform into a 60-foot-tall fortress.
Property-value impact studies are mixed but trending toward documented losses for homes within close proximity. Class-action and individual lawsuits have begun appearing in 2025-2026.
6. Transparency about ownership and operations
Many datacenter projects are filed under LLC names that don't immediately reveal the operator — sometimes deliberately, to avoid public attention during the permitting phase. Communities have learned to push back on this, with some jurisdictions now requiring beneficial-ownership disclosure as a condition of approval.
The pattern is well documented: a project files under 'Project Lonewolf LLC' or a similar codename, secures permits and tax incentives, and the actual operator (Meta, Microsoft, Google, or a hyperscaler shell company) is revealed only after approval. By that point, reversing course is politically and contractually expensive.
7. AI density compounds every other concern
All of the above issues are getting worse, fast, because AI training workloads are dramatically more resource-intensive than the cloud workloads datacenters were originally designed for. AI clusters require 4-10x more power per rack, generate more waste heat (so more cooling), and cycle equipment harder (more frequent battery replacement, more generator testing).
Communities approving a 200MW facility in 2019 are now seeing operators request expansions to 500MW-1GW driven by AI demand that didn't exist at the original approval. The political contract — what was approved — feels broken to residents even when the legal contract was honored.
8. Environmental justice concerns
A growing body of analysis shows that datacenter sites are disproportionately located in lower-income, rural, and predominantly non-white communities — partly because cheaper land and less-organized political opposition make permitting faster. Environmental justice groups have begun framing the issue in civil-rights terms, particularly in rural Georgia, Mississippi, and parts of Texas.
How Operators Are Responding
The serious operators are responding, with varying degrees of authenticity:
Microsoft has published detailed water-and-power use figures by region and committed to specific net-positive water targets by 2030. Implementation is mixed but the disclosure framework is real.
Meta has invested heavily in air-cooled and adiabatic-cooled designs at newer campuses, dramatically reducing water consumption per facility. Eagle Mountain specifically has retrofitted some cooling infrastructure.
Google has committed to 24/7 carbon-free energy matching by 2030 and is investing in geothermal and small modular nuclear partnerships specifically to offload datacenter load from local grids.
AWS has been the slowest to publicly engage but is increasingly using nuclear PPAs (the Talen Energy deal in Pennsylvania being the most prominent) to avoid stressing local grids.
Smaller colo operators are mostly silent on the topic — partly because they lack the resources of the hyperscalers, partly because the political costs of engaging are higher than ignoring.
What This Means for the Industry
Several structural changes are now visible:
Site-selection is harder. The pipeline of low-friction sites is shrinking. New projects are routinely facing 12-24 month additional permitting delays even after technical approvals.
Costs are rising. Sound walls, larger setbacks, water-efficient cooling, and beneficial-ownership disclosures all add cost. Five years ago these were optional; today they're table stakes.
Tax-incentive structures are getting renegotiated. Several states are revisiting datacenter incentive programs and tightening eligibility. Ohio, Virginia, and Texas all have legislation pending or recently passed.
Site-selection now requires community-engagement budgets. What used to be a back-office function is now a meaningful line item in project planning — comparable to legal or geotechnical work.
What It Means for Vendors
Counterintuitively, the backlash creates opportunities for some vendor categories: closed-loop liquid cooling vendors, sound-attenuation specialists, beneficial-ownership and compliance consultants, water-efficient cooling-tower vendors, and operators capable of building in the 50-150MW range (smaller, harder to oppose) rather than 500MW+ behemoths. Our Site Selection briefings now routinely include community-opposition risk assessments alongside the traditional power, capacity, and competitive-density data.
If you're a developer or enterprise planning a new owned facility and want a candid analysis of community-opposition risk for your candidate sites — alongside the standard market data — that's increasingly the most valuable section of the brief. Founding member pilots include this analysis for any covered territory.
The Bigger Picture
Communities are not anti-technology, anti-AI, or anti-investment. The pushback is grounded in something more specific: a sense that the costs of hyperscale datacenter construction — water, power, noise, fiscal burden, visual impact — are being borne by local communities while the benefits flow to coastal corporate headquarters and global cloud customers who will never visit.
Operators that figure out how to credibly address that imbalance — through real, verifiable, locally-visible commitments — will keep building. Operators that don't will find their pipelines increasingly stuck. The next five years of datacenter construction will be shaped at least as much by community politics as by technology.
Sources & Further Reading
Deseret News: Meta's Eagle Mountain Datacenter Expansion Coverage — Local reporting on the campus expansions and community response.
Utah Rivers Council — Advocacy organization that has tracked datacenter water-use concerns in Utah.
Utah Taxpayers Association — Analyses of datacenter tax-incentive structures and projected vs. delivered economic impact.
Washington Post: Data Centers' Hidden Costs — National investigative coverage of the water and power impacts of US hyperscale infrastructure.
Data Center Dynamics — Community Opposition Coverage — Ongoing industry reporting on permitting fights, community engagement strategies, and regulatory responses.