Case Studies: How This Has Gone Elsewhere

How we chose these — and a note on sourcing. We included only claims we could trace to a reputable source: major news outlets, local papers of record, court and government records, or academic and state-legislature studies. We avoided advocacy blogs and pure industry marketing. One honest caveat to keep in mind while reading: news coverage skews toward conflict. A data center that quietly pays taxes for fifteen years makes no headlines, while a lawsuit or a dry well makes many — so cautionary tales are easier to find than success stories, which understates how often these simply work out. Where a figure comes from a company-funded study, we say so.

Where it worked

Mecklenburg County, Virginia — a generational investment in schools

In rural Southside Virginia, tax revenue from a Microsoft data center operating for 14 years helped fund a $154 million renovation of a combined high school and middle school. It's the clearest example of a small, rural community turning data-center revenue into a lasting public asset. And on the "is it a good deal?" question, an independent 2025 Virginia legislative study (JLARC) found the data-center tax break returns $6.10 of labor income for every $1 of exemption — its second-best incentive of any kind. In fairness, the same study found the state recovers only about 48 cents of that dollar as direct tax revenue, so the case rests on jobs and economic activity, not the direct tax math. (Cardinal News, Cardinal News (state study))

Los Lunas, New Mexico — revenue and a growth magnet

The village of Los Lunas (population ~20,000) has taken in about $145 million in cumulative local tax revenue since 2016 from a Meta campus, grew its population 15% in four years, and saw the campus attract an Amazon fulfillment center (600+ jobs) nearby. Honest caveat: it's a desert region, and the facility used roughly 75 million gallons of water in 2023 along with large amounts of electricity — the resource questions are real, which is exactly why our draft rules focus on water and cooling. (Santa Fe New Mexican)

Loudoun County, Virginia — where homeowners' taxes actually went down

Loudoun County is far larger than Linn Valley — it's the heart of the country's "Data Center Alley," not a size match — but it is the clearest real-world proof of the mechanism behind our financial case: a large non-residential taxpayer shifting the burden off homeowners. With data centers carrying a big share of the tax base, the county has lowered its real property tax rate every single year — from $1.145 per $100 of value in 2016 to $0.805 in 2025, which the county government itself says has "largely resulted in a reduction of the tax burden on residential taxpayers." That is the difference between a tax base that leans on families and one that doesn't. (Loudoun County Government)

New Albany, Ohio — structuring the deal so the community wins

New Albany (a Columbus suburb — again, not a size match) is worth studying for how it wrote its deals. Rather than simply handing out tax breaks, the city requires minimum annual payments in exchange for any partial abatement, structured so that — in the city's own words — "a data center campus with 200 to 250 employees has the same impact on the city's bottom line as an office complex or production facility with 2,000 workers." That is the Mecklenburg lesson in action: the benefit comes from how the deal is written, not from the technology itself. Honest caveat: Ohio still grants large statewide sales-tax breaks for data centers — a real public cost worth weighing. (News 5 Cleveland)

Cheyenne, Wyoming — paying full freight, funding schools

Cheyenne is bigger than Linn Valley, but this Plains city is worth noting because it answers the abatement worry head-on. Microsoft is the city's largest taxpayer at about $11 million a year, supporting schools and public services — and crucially, its latest expansion uses no local property-tax incentives at all. The company employs about 500 local residents full-time, ran a workforce certificate program that graduated over 1,000 students with the community college, and uses mostly air cooling (evaporative only about 37 days a year). Honest caveat: some residents have still raised concerns about oversight, air quality, and water — good questions don't vanish just because the deal is a good one. It's the clean counterpoint to Prineville: full taxes paid, schools funded. (Cap City News)

A note for balance: another often-cited success — Forest City, North Carolina, a struggling rural textile town revived by a Facebook data center — is real, but the largest local jobs figure comes from a company-commissioned study, so we don't lean on it here.

Where the results were mixed

Prineville, Oregon — a revival, but the schools paid for it

Prineville (population ~10,000) is the instructive companion to Mecklenburg. It was a sawmill town with around 20% unemployment when Facebook, then Apple, built large data centers there, and the revival is real — hundreds of good-paying permanent jobs, and the city itself collects millions a year in electricity franchise fees. But the deal was structured very differently from Mecklenburg's. Under Oregon's enterprise-zone programs the companies pay essentially no property tax, and as a result Crook County's schools gave up about $29 million in 2024 alone to those exemptions — up from $10 million in 2019. (Meta has donated more than $2 million to the school district over the decade — real, but a fraction of what the exemptions cost the schools.) Same technology as Mecklenburg, nearly the opposite outcome for the schools — because of how the deal was written. (Central Oregon Daily, WRAL / New York Times)

Quincy, Washington — a new school, and a years-long fight

In this farm town of ~7,500, data-center taxes helped fund a new high school — but the arrival also brought a years-long dispute over roughly 40 diesel backup generators permitted near an elementary school (a state board ordered a cumulative health-risk review), utility fines, and, by 2022, data centers consuming about 40% of the county's electricity, pushing the county off clean hydropower toward gas. Real benefits, real trade-offs. (Washington Dept. of Ecology, ProPublica / Seattle Times)

The Dalles, Oregon — a rescued tax base, and a secrecy fight

After its aluminum plant closed, this town of ~16,000 leaned on a Google data center that pays roughly $5.4 million a year in property tax — but at a 92% tax abatement for only about 200 jobs. And when the local newspaper sought records of Google's water use, the city sued on Google's behalf, claiming the data was a "trade secret," and fought for 13 months before settling and releasing it (Google had been using nearly a third of all the city's water). The lesson we took: never accept secrecy on resource use. (Fortune, Reporters Committee for Freedom of the Press)

DeKalb, Illinois — a real boost, with strings

A $1 billion Meta campus added roughly $103 million to the county's tax base and let the county lower its tax rate — but a 20-year, 55% abatement makes much of that boost temporary, and the local park district reported missing more than $1 million in revenue it would otherwise have collected. A genuine benefit, with genuine strings — the kind of deal that has to be negotiated carefully. (Shaw Local / Daily Chronicle, NPR Illinois)

Where it went badly — and why

Peculiar, Missouri — the closest cautionary tale to us

A small town of about 6,000 near the Kansas line faced a $1.5 billion data center. Roughly a sixth of the town organized against it over water, noise, light, and traffic — but the deeper grievance was that city staff had misrepresented what a data center is (one account compared it to "software/computer equipment manufacturing"), and the mayor publicly called critics "ill-informed." The board ultimately voted unanimously to remove "data center" from its zoning ordinance, killing the project; reporting indicates the controversy cost the mayor and the city administrator their jobs. The lesson: define the use precisely and disclose it honestly from the start — which is why this site explains in plain language exactly what these facilities are and aren't. (KSHB / NBC Kansas City)

Festus, Missouri — approving over residents, and the fallout

This town of ~12,700 approved a $6 billion data center 6–2 over a high-school gym packed with opponents. An open-records request later showed officials had privately called the opposition "uneducated." Days later, voters ousted all four incumbent council members who were up for reelection, and recall efforts and an open-meetings lawsuit followed. The lesson: don't disparage residents, and don't force a vote without building consensus first — which is why we're doing this in the open, before any project is even on the table. (KCUR / NPR, St. Louis Public Radio)

Newton County, Georgia — when siting goes wrong

Neighbors living about 1,000 feet from a Meta data center had their well water run dry, with sediment they attribute to construction; the facility uses an estimated 500,000 gallons of water a day, and the dispute reached a congressional hearing. The lesson: require baseline well testing for neighbors, developer-funded monitoring, a binding remediation guarantee, and real setbacks from homes and wells. (New York Times via The Spokesman-Review)

Prince William County, Virginia — when process is rushed

An approval for up to 37 data centers (~23 million square feet) sited beside homes and a Civil War battlefield was voided by the Virginia Court of Appeals in 2026 because the board failed to follow the state's public-notice law. The lesson: follow notice and procedure rigorously — a shortcut can unravel the entire approval years later. (Prince William Times)

The constructive counter-example: Chandler, Arizona

Not every cautionary story ends badly. Chandler, Arizona regulated proactively — adopting a 2015 water policy limiting data centers to 115 gallons per day per 1,000 square feet (the first U.S. ordinance of its kind) and a separate 2022 noise ordinance requiring pre-construction baseline studies, sound caps, and five years of monitoring. It's a model of writing clear rules before a developer arrives — which is exactly what our draft ordinance aims to do. (City of Chandler)

The bigger picture: what the broad evidence shows

Individual stories only go so far — and as Prineville shows, a town that looks like a win on one measure can be a loss on another. So it's worth stepping back to the wider, independent evidence:

• The revenue is real and large in aggregate. Government revenue tied to the data-center sector rose from about $66 billion in 2017 to more than $162 billion in 2023 nationally (World Resources Institute), and where a community captures a meaningful share, the effect on homeowners can be real — Loudoun County's documented rate cuts above are the clearest example.
• An independent state study found a strong return — on jobs and income. Virginia's legislative audit commission (JLARC, 2025) ranked the data-center tax exemption the state's second-best incentive, returning $6.10 of labor income per $1, while candidly noting the state recovers only about 48 cents of that dollar in direct tax. The value is in economic activity, not the direct tax math.
• But the jobs are modest, and the incentives are large and contested. Watchdog groups such as Good Jobs First have argued that in some states data-center tax breaks can cost more than the revenue they bring. In Virginia, the abatements now total about $1.6 billion a year, and a 2025 legislative audit found 53% of all the state's incentive spending went to data centers — fueling an open "giveaway or bargain?" debate (Cardinal News). The lesson isn't that incentives are always bad — it's that their size and terms have to be justified.
• Outcomes hinge on deal design, not the technology. Brookings' review of rural data-center hosts reaches the same conclusion this page does: the question "is not simply whether data centers are good or bad, but how benefits and risks get allocated" — through local rules, transparency, and how the deal is written (Brookings).

In short: the upside is real but not automatic. It shows up when a community captures enough revenue and writes the deal well — and it disappoints when a town gives away too much. That is a conclusion we can act on.

What separates a success from a disaster

Read across all of these and a pattern emerges that's worth stating plainly: the failures weren't caused by the technology. They were caused by governance, siting, and the quality of the deal. Secrecy sank The Dalles' trust. Bad siting next to homes and wells caused the harm in Georgia. A rushed process voided the approval in Prince William. Misrepresenting the use doomed it in Peculiar. Dismissing residents cost the council in Festus. The communities that did best did the opposite — they were transparent, sited carefully, and set clear rules first.

Two honest truths to carry out of this:

• The jobs are modest. These are not big employers — often a few hundred or fewer — though the permanent jobs pay well. The real prize is a broader tax base, not headcount, and we don't claim otherwise.
• The deal terms decide everything. The clearest proof is two towns with the same technology and opposite results: in Mecklenburg County, Virginia the deal helped fund a $154 million school, while in Prineville, Oregon the abatements left the local schools giving up about $29 million in a single year. Tax breaks can be generous — one watchdog (Good Jobs First) found that in some states they cost more than they bring, while Virginia's legislature found a strong return on jobs and income. The difference is how well a community negotiates, and how the abatement is written — which is why our approach front-loads the rules and the leverage.

That's why the draft ordinance is built the way it is: every use requires a public Conditional Use Permit, large facilities sit well back from homes and the lake, water and cooling are scrutinized, the developer funds the infrastructure its load requires, and large projects must enter a Community Benefits Agreement. Each of those provisions targets a specific failure mode above — we studied how this goes wrong so we could write the rules to prevent it.