Data center projects don’t usually get denied. They disappear.
If you are evaluating a county for a data center, someone has probably handed you its approval history. Ten industrial rezonings, ten approvals. Unanimous votes. No moratorium on the books. A board that has never turned anything down.
That record is real. It is also close to meaningless, and reading it as a forecast is how experienced developers lose eighteen months.
An approval rate measures one thing: what happened to the projects that reached a recorded vote. Everything else is missing — projects paused indefinitely, projects withdrawn the week of a hearing, projects approved and then unwound in court. In data center siting right now, those categories hold most of the useful information.
The clearest case: Pittsylvania County, Virginia
In October 2024, Balico LLC filed a rezoning application near Chatham for a complex of 84 data centers and a 3,500-megawatt gas-fired power plant — a plant that would have been the largest in Virginia — on roughly 2,200 acres. Facing immediate opposition, the company withdrew it.
That first withdrawal produced no denial and no record of a failure. It appears in no approval statistic anywhere.
Balico refiled in November with a smaller footprint: 750 acres, 12 buildings. In January 2025 the planning commission unanimously recommended denial. The board granted the applicant postponements in February and again in March, extending the process for residents living under the proposal.
On Monday, April 14, 2025 — one day before the scheduled vote — Balico announced it was withdrawing again.
The board refused the withdrawal 5-2 and held the hearing. Its stated reason was procedural: the meeting had already been advertised and relocated to Chatham High School to hold the expected crowd, and the chairman said the county had already spent money on the venue change and security and could not pull the item so late. Hundreds attended. Forty-seven speakers signed up. The board then denied the application 6-1.
What Balico had asked for was a withdrawal without penalty. Under county rules, the denial barred it from resubmitting a substantially similar proposal for twelve months; a granted withdrawal would have carried no such bar.
Why the project actually failed
The reasons the board gave are the more useful part for anyone siting a project.
Supervisors voted against it citing the layout, the height and profile of the buildings, expected emissions, limited local job benefit, and a lack of clarity in the proffers and plans. The chairman said the company had not been forthcoming with the board or the community about what it wanted, and described having to ask repeatedly for conditions and proffers, with information arriving late. His summary of the episode: you don’t get a second chance to make a first impression.
And the board was not opposed to data centers. The chairman told residents not to be disheartened by the vote, because the county still had the resources and a good place to put one.
The lesson: Pittsylvania was open to the land use. It rejected a developer. Six months of controversy, two postponements, a refused withdrawal and a 6-1 denial — in a county that was willing to say yes to someone.
The exit works when boards allow it
Citrus County, Florida ran a similar sequence to a different conclusion.
Deltona Corporation applied to rezone more than 1,300 acres for an industrial park with a large data center component. On June 19, 2026, the planning commission held a nearly nine-hour hearing with more than 200 residents attending, most opposed, and voted unanimously to recommend denial.
The binding vote was set for July 14. The developer asked to move it to November. The board refused the continuance. Five days later Deltona withdrew the applications and the July hearing was canceled.
No vote was ever taken. Read only the county’s approval record and you would never learn what nine hours of testimony revealed about how that community sees this land use — or that a planning commission had already found the project wanting.
Approval is not the end of the story either
Prince William County, Virginia shows the opposite failure mode, and it is the more expensive one.
The PW Digital Gateway was approved by the Board of Supervisors in 2023 — up to 37 buildings and more than 22 million square feet across roughly 2,000 acres along Pageland Lane, adjacent to Manassas National Battlefield Park. At full buildout it would have been the largest data center campus in the world.
Residents began fighting it in the weeks after approval. Two lawsuits followed — one from the Oak Valley Homeowners Association, one from the American Battlefield Trust. A circuit court invalidated the rezonings on the grounds that the county had failed to meet public notice requirements. The Virginia Court of Appeals upheld that ruling on March 31, 2026. The county and co-developer Compass dropped their appeals in the spring, leaving QTS alone; on July 2, 2026, QTS withdrew its final petition and terminated the project.
Roughly three years from approval to dead, on a notice defect.
The mechanism recurs, though it does not always end a project. In Stokes County, North Carolina, commissioners voted 3-2 in January 2026 to rezone roughly 1,845 acres near Walnut Cove for heavy manufacturing, overruling their own planning board after hundreds of residents packed the chamber and an overflow area. Residents and nonprofits sued in March. In April the board voided its own approval, acknowledging that the hearing notice had been published 31 days ahead of a statutory window of 10 to 25 days, and that the approvals had to be treated as though they never happened. The developer is pursuing the project again through a restarted process.
The lesson: an approval is a milestone. Where opposition is organized and funded, it is not a finish line.
Both failure modes cost a developer the same things: capital spent, schedule lost, a site no closer to being built. Neither leaves much behind. A project that withdraws before a vote never enters the record at all, and an approval that is later voided can be struck from it — in Stokes County, the board accepted that its own approvals had to be treated as though they never happened.
For scale, Data Center Watch counts at least 45 projects worth roughly $68 billion blocked or delayed in Q2 2026 alone — down from 75 projects and $130 billion in Q1 2026. The same research puts active opposition groups at 833 by March 2026, up from 396 at the close of 2025, across 49 states. Very little of that appears as a denial.
Why the record looks the way it does
The incentives point away from recorded outcomes.
A developer who expects to lose has concrete reasons to exit early. Pittsylvania shows one: a denial there triggered a twelve-month resubmission bar. Denials also harden positions and follow a site. A withdrawal or an indefinite pause preserves the option to return with a different design, a different parcel, or a different board after the next election.
Boards rarely have reason to object. Granting a withdrawal clears the calendar and spares elected officials a vote that will anger someone either way.
So the easy approvals land on consent agendas and the difficult projects tend to leave before a vote. What survives in the record is the narrow slice where someone kept a scheduled hearing, or where litigation forced the question afterward.
What this means for developers and capital
Treat approval rate as conditional, not predictive. It tells you how projects fared once they reached a vote. It says nothing about how many never got there, which is the number you actually need.
Track the disappearances. A project that appears, is continued twice, then goes quiet rarely comes with an explanation. Across the 400+ projects we’ve analyzed, when we trace a quiet withdrawal back, the answer is seldom the same twice. Sometimes it’s the neighbor whose property line touches the site. Sometimes it’s one person who gets the whole county behind them. Sometimes it’s a group that opposes every project that comes through, whatever the technology. Each of those means something different for the next developer. A neighbor’s objection may stay with that parcel. A group that fights everything will be there for the next project too, so you need a plan for them from day one. Telling them apart takes more than a vote count. You have to read the comment records, the meeting minutes and who keeps showing up.
Advisory votes inform, but they do not decide. A planning commission recommendation is often where a project’s trajectory becomes visible — Citrus County and Howell Township, Michigan, both saw withdrawals follow adverse advisory votes. But it cuts both ways, sometimes inside one county. Hanover County, Virginia saw both outcomes in five months. In January 2026 its planning commission voted against a 400-acre project off Ashland Road, and the developer, homebuilder HHHunt, later withdrew it. In April the same commission recommended approval of the 427-acre Mountain Road Technology Park 6-1, with a single dissent — and on May 27 the board of supervisors denied it 4-3, the chair casting the deciding vote. The two bodies answer to different constituencies and can reach opposite conclusions on the same file. What matters is not the institution but the people in it.
Watch what is organized, not what is recorded. Petition counts, how active the groups are, overflow crowds, turnout at advisory hearings, the means to litigate. All of it shows up before the decision, and none of it is in a typical siting dataset. Opposition isn’t a verdict. Most projects face some. What matters is how effective it is: whether it moves officials, holds together over months of hearings and reaches beyond the people who live next to the site. That’s what tells you what you’re up against and how to plan for it.
Where this information actually lives
It lives in the layer no GIS dataset holds — meeting minutes, voting records, local news, social media, and the history of relevant projects a community has approved or fought before. That is the layer we read. There are three things it produces that an approval rate cannot.
The people, not the institutions. We profile the individuals who will vote on, recommend, or shape a decision: how each has voted on comparable items, what they have said publicly, what they have asked developers to provide, and what has moved them before. Hanover County is the case in point. A developer tracking the planning commission would have seen a 6-1 recommendation to approve and read it as momentum. Six weeks later four supervisors denied the project — a different body, a different constituency, and positions that were on the public record beforehand. One of them said plainly that she could not ignore the number of residents who had come forward to object.
The disappearances, counted. Because we read vote records against agendas rather than in isolation, projects that never reached a vote become visible. In one Texas county we assessed, the report notes that the county’s clean approval record reflected only items that had reached a formal vote — contentious proposals were routinely withdrawn or tabled first. In a second Texas county, dissent on development items showed up not as a recorded “nay” but as absence from the room. A developer reading either county’s approval history alone would have seen an open door.
The first impression, before it is spent. Pittsylvania’s chairman was explicit that the project failed on how the developer arrived — not forthcoming about its intentions, information late, local buy-in never sought. By the time an application is filed, that impression is set and the record has begun forming around it. What a community’s actual concerns are, which officials need to hear what and from whom, and which past project shaped local expectations are all knowable months earlier.
None of this makes a hard site easy, and the goal is not to eliminate opposition. But the difference between a project that dies at a podium and one that gets approved often starts with what the town already knows. Very little of it sits where a search can reach. It’s in hours of meeting video, in a comment made at a hearing three years ago, in who sided with whom the last time a project came through. Finding it is the first step. The harder part is reading it right: who has real influence, which concerns can be addressed, and what it will take to get to a yes.
