Field notes · Public records

The Data Center Graveyard

Forensic autopsies of billion-dollar projects that local opposition killed or stalled. What killed them, what saved the survivors, and one quiet question for your own pipeline.

Every figure below comes from public court rulings, ballots, and filings. Sourced, no spin.

~$130B
of projects blocked or delayed in Q1 2026 alone
~100
local moratoriums in place by mid-2026
833
grassroots opposition groups, up more than 2x since late 2025

Power and land are no longer the only things that kill a data center. The community is. The projects below did not fail on economics. They failed on water, on a paperwork defect, on a tax break the public voted away.

The case files

The autopsies

Three billion-dollar projects, three ways a data center dies. Read one, then run your own site against it. Swipe or use the arrows.

Autopsy 01Dead

PW Digital Gateway

One of the largest campuses ever proposed. A $24.7B plan killed by a paperwork defect checkable from day one.

Location
Prince William Co., VA
Developers
QTS + Compass
Scale
2,100+ ac · 37 DCs · 23M sf
Value
$24.7B · $400.5M/yr tax
Approval
4-3-1 · Dec 2023
Status
Dead · Jul 2, 2026
Cause of death

A public-notice defect in the county's own rezoning. Not water, not the grid. It passed on a razor-thin 4-3-1 vote after a 29-hour hearing; opponents then sued on process, not merits, and won.

The 4 fatal mistakes
  1. Built on a cracked legal foundation. A $24.7B effort rested on a rezoning with notice defects. One flaw, one point of failure.
  2. Maximized opposition intensity. A 29-hour hostile hearing and a 4-3-1 split built the litigation-capable opposition and its record.
  3. No off-ramp, no phasing. One all-or-nothing rezoning across 2,100+ acres. When one thing broke, everything broke.
  4. Misread the opponent. A disciplined legal campaign. They brought a construction plan to a courtroom fight.
Signals in the public record beforehand
  • A 4-3-1 approval with an abstention: a fragile mandate.
  • A 29-hour hearing with opponents who sued almost immediately.
  • The notice-compliance record, auditable before the land spend.
What would have saved it
  • A procedural pre-mortem before capital goes in.
  • Phase and scale the ask to kill the all-or-nothing risk.
  • Concessions that make suing not worth it.
The question for your dealYour approval is only as strong as its weakest procedural link and its most motivated opponent. Do you know where both are, before the capital goes in?

Sources: Court of Appeals of Virginia; InsideNoVa; Prince William Times; Virginia Business; Data Center Frontier.

Read the full autopsy ›
Autopsy 02Rejected, then revived

Project Blue

Rejected 7-0 by the city, then revived one jurisdiction over. The comeback is the real lesson.

Location
near Tucson, Pima Co., AZ
Developer
Beale Infra. (Blue Owl)
Size
600 MW · ~$3.6B · 290 ac
Water
up to ~283M gal/yr
City vote
7-0 reject · Aug 2025
Status
Revived · county 3-2 · Dec 2025
Cause of the rejection

Water, secrecy, and a jurisdiction that could say no. A design needing City of Tucson water and annexation, up to ~283M gallons a year in a drought region, while the end user (AWS, reported) stayed undisclosed. The council had the votes, 7-0.

The 4 mistakes that lost the city
  1. Picked a water fight it did not need. City water in a drought region, the most inflammable objection possible.
  2. Secrecy backfired. Hiding the end user converted uncertainty into distrust.
  3. Chose a jurisdiction that could veto. A city council is a single winnable target.
  4. Led with the ask, not the concessions. No benefit story arrived before the opposition did.
Signals in the public record beforehand
  • A water-dependent design in a metro fighting over supply.
  • An undisclosed end user, read as something to hide.
  • Approval that required annexation and a council vote.
The comeback, which is the antidote
  • Removed the dependency: closed-loop cooling, no city water.
  • Changed the venue: Pima County, 3-2 approval, closed Dec 2025.
  • Locked power early via the Arizona Corporation Commission.
The question for your dealWhat single resource dependency, water, city power, or annexation, is your project's kill switch, and are you engineering it out before opponents find it?

Sources: AZ Luminaria; AZPM; KJZZ; Tucson Sentinel; Democracy Now!; DataCenterDynamics. AWS end user per reporting; sponsors did not confirm.

Read the full autopsy ›
Autopsy 03National template

Port Washington

Opponents did not fight the zoning. They put the developer's tax incentives to a public vote, and handed every town the template.

Location
Port Washington, WI
Developer
Vantage / Cloverleaf
Scale
up to 3.5 GW (reported)
Referendum
~66% yes · Apr 2026
New rule
incentives >$10M need a vote
First
anti-DC referendum in the U.S.
Cause of death

The financing assumption, not the zoning. Opponents made any incentive over $10M require a public vote, turning the economic model into a referendum they were nearly certain to win. Passed ~66%, pitched openly as a national model.

The 4 fatal mistakes
  1. Depended on discretionary incentives, a lever the resentful public can yank away.
  2. Fought zoning, lost the ballot. Opposition invented a new instrument while the developer played permitting.
  3. Lost the "costs vs. benefits" frame. At ~70% against, every ballot mechanism turns hostile.
  4. No economic narrative that survived a vote. Residents felt no gain worth approving the subsidy.
Signals in the public record beforehand
  • A model leaning on a revocable tax break.
  • Marquette polling moving against it: 55% to 70%.
  • A ballot mechanism available to organizers.
What would have saved it
  • Do not rely on incentives residents can revoke.
  • Win "costs vs. benefits" early with per-household benefits.
  • Assume the ballot box is a venue. Referendums spread.
The question for your dealIf your economics depend on a tax break, could a citizen vote take it away, and have you already lost the "costs vs. benefits" argument in that community?

Sources: NOTUS; Wisconsin Watch; WPR; FOX6 Milwaukee; Marquette Law School Poll. Campus scale reported up to 3.5 GW.

Read the full autopsy ›
The pattern

Three deaths, one anatomy.

Different states, different triggers, the same four failure modes. Every project that dies has at least one. Every project that survives designs them out.

MODE 1
A fatal dependency
Water, city power, annexation, or a tax break you do not control. The dependency is the vulnerability.
MODE 2
A veto point
An elected body that can say no, a rezoning that can be challenged, a ballot that can be run. A single winnable target.
MODE 3
The "costs vs. benefits" frame is lost
Once residents believe the project costs them more than it gives, every mechanism turns against you.
MODE 4
Concessions arrived late
Benefits offered after opposition organized read as damage control, not good faith. Timing is everything.
Apply it

How does your project compare?

Every fact in these autopsies sat in the public record before the project died: the vote counts, the notice filings, the polling, the water math. Assembling that for a single site is days of digging across court dockets, utility filings, council minutes, and local press. That is the pass this tool is built to run.

Drop your site on the map to see the nearest autopsied failure and what is happening in your state, then run the four failure modes against your own project.

Red is a project that died or stalled. Green is rejected but revived by re-routing.
Now run your own project against the four failure modes
Mode 1, fatal dependency. Does your power, water, or land access depend on a single source you do not control (a city, one utility, an annexation)?
This is what lost Project Blue its 7-0 city vote, and what its comeback engineered out.
Mode 2, veto point. Does approval run through one elected body, a challengeable rezoning, or a jurisdiction where a referendum could be called?
A procedural challenge killed PW Digital Gateway; a ballot measure gated Port Washington.
Mode 3, the frame. If residents were polled today, would a majority say your project costs them more than it gives?
Once that number crossed about 70% in Wisconsin, every ballot mechanism turned hostile.
Mode 4, timing. Have you put concrete community benefits on the table before organized opposition appeared?
In all three cases, concessions arrived late, or never.
Keep watch on this market

Most weeks, a project somewhere gets blocked. If a moratorium, an organized opposition, or a tariff change lands near this spot, we will flag it for you. Not a newsletter. Just the one alert that matters for the site you are weighing.

Set. If something moves near here, you will hear from us first.

Do not become an autopsy.

Each case has the full write-up: the dated timeline, the deal-level detail, every source, and the insight to carry away. Open one, and take the board-ready brief to your next investment committee.

PW Digital Gateway Project Blue Port Washington
Or screen a specific site's power path.