A $7,509 Newspaper Ad Killed the World's Largest Planned Data Center

Everyone was arguing about gigawatts. The judge was reading a calendar.

Here's the scale of what got erased. Prince William County, Virginia approved three rezonings in December 2023 covering 884 acres, 534 acres, and 342 acres along a two-lane road next to Manassas National Battlefield Park. Together they allowed about 22.2 million square feet of gross floor area — as many as 37 data centers, running 24/7. County finance staff pegged it at a potential $24.7 billion in investment and around $400 million a year in tax revenue at full buildout. Power demand was estimated at 2 to 3 gigawatts, which at the time was roughly equal to every data center already operating in Northern Virginia combined.

So the public fight was exactly what you'd expect. Noise proffers capped operations at 60 decibels daytime and 55 at night, measured at the edge of residential land — but those caps didn't apply to construction, or to the emergency generators when they test or run. Neighbors testified about well water, light pollution, smoke, transmission lines cutting through, traffic. The county's own staff recommended denying all three, partly because the actual building footprints hadn't been submitted for review yet.

None of that is what killed it. The court that finally ended this went out of its way to say the ruling was "not influenced by the vibrant public debate surrounding data-center development," because it isn't a court's job to decide whether data centers are good. It decided something much more boring, and much harder to argue with: whether the ads ran on the right days.

The 3:00 p.m. email nobody answered

Virginia law and the county's own zoning ordinance set out a small, rigid checklist before a rezoning hearing. Run the notice in a newspaper twice. First ad no more than 14 days before adoption. At least six days elapsing between the first and second publication. Say in the ad where the public can go read the actual proposed ordinance. And hold the hearing no less than five days after that second ad appears.

The county planned it correctly. First ad November 28 — exactly 14 days before the December 12 hearing. Second ad December 5 — more than six days later, and seven days before the hearing. That plan works. On Monday, November 20, the clerk emailed the newspaper with the ad and the run dates. Two hours later the paper wrote back quoting $7,509.48 and warning, in plain language, "Saved ads will not run unless they are scheduled and submitted." Two minutes after that, a second email: confirm by 3:00 p.m. tomorrow. Nobody responded by the deadline. The first ad never ran.

Then the county pressed ahead anyway and scrambled, running three ads instead: December 2, December 5, and December 9. Sit with that sequence, because it's a trap with no exit. December 2 to December 5 is three days — fails the six-day rule. Push the "second" ad out to December 9 and you clear six days from December 2, but December 9 to the December 12 hearing is only three days, and the ordinance demands at least five. There is no pairing of those three dates that satisfies both rules. On top of that, the proposed ordinances weren't actually available for anyone to read in the clerk's office until December 7 — after the first two ads had already told the public where to go read them.

The hearing went ahead on December 12. One supervisor moved to defer it "until such date as a proper public notice has issued" specifically to avoid the lawsuit everyone could see coming. That motion failed five to three. What followed was six hours of developer presentations and staff reports, then more than 17 hours of public comment. In the early morning of December 13, while people were still speaking, the developers handed staff revised proffers — which staff said they hadn't been able to thoroughly vet before the vote. A motion to deny all three failed four to four. A motion to approve all three passed four to three with one abstention. The meeting adjourned at 1:14 p.m. on December 13, twenty-seven hours after it started.

Every hour of that was legally worthless.

Why "close enough" wasn't good enough

This is the part I kept re-reading, because it's genuinely counterintuitive. The county's argument was reasonable on its face: thousands of people obviously knew about this hearing. Hundreds showed up and spoke for 17 hours. Nobody was actually kept in the dark by an ad running on a Saturday instead of a Tuesday. That's the "substantial compliance" argument — we hit the spirit of the rule, so let it stand.

Virginia's courts said no, and the appeals court explained why in a way I found hard to dismiss. The legislature could have written a substantial-compliance standard — it did exactly that for abandoning a secondary road, in a different statute. Here it didn't. It wrote a per se rule that specifies exactly how to advertise, against a long backdrop of cases holding that blowing the advertising makes the resulting ordinance void ab initio: void from the start, as if it never passed. The court acknowledged the harshness out loud, noting that strict compliance means "missing the deadlines by even one day renders the resulting ordinance invalid, potentially disrupting months of planning." Then it enforced it anyway, because reading a rule of reason into a per se rule would "invite other localities to be less vigilant."

The county also tried a safe harbor written into the statute for when a locality submits a correct and timely request and the newspaper fails to publish it. That failed on the facts. Newspaper staff testified they never send an ordered ad to print without client confirmation, and the trial judge found the fault was the county's, not the paper's. The county even asked the paper to sign an affidavit taking the blame. The paper declined.

A circuit court voided all three ordinances in August 2025. A three-judge appeals panel affirmed on March 31, 2026. The last developer withdrew its petitions to the state supreme court on July 2, 2026, and the largest planned data center campus in the world simply stopped existing. Worth knowing: Virginia has since amended the statute — a 2024 change moved the window to first ad no more than 28 days before and second ad no less than seven days before the hearing, and a 2025 change shortened that seven to five. The rule that killed this project is not the rule in force today. Which is its own kind of gut punch.

Two gigawatts don't disappear — they move, and the grid bill follows

Here's what I actually can't stop thinking about, sitting on my side of this as a person who just pays a power bill. That 2 to 3 gigawatts of demand didn't evaporate when the rezoning did. It went looking somewhere else. Nationally the load is still stacking up: data center power requests in Virginia alone now total around 70,000 megawatts in filings — roughly triple the utility's current peak. Meanwhile, heavy-duty gas turbine order books are effectively sold out into 2029 and 2030, large power transformers are quoted in the 128-to-160-week range, and the average wait just to interconnect to the grid runs about five years, longer in data-center-dense regions.

And the fight over who pays has already moved from zoning hearings into rate cases. Virginia regulators approved a new rate class for customers drawing 25 megawatts or more, starting January 1, 2027, requiring them to cover at least 85% of contracted transmission and distribution demand and 60% of generation demand. Regulators projected that shifts a typical residential rate down about 3.4% — roughly $5.52 a month — while the big-load rate goes up about 15.8%. That's the actual mechanism behind the comment I see under every single video on this topic: the grid upgrades get socialized and the profits stay private. Somebody decided to change the math.

Meanwhile the local opposition playbook is getting sharper and more procedural, which is exactly the lesson of this case. One tracker counted at least 75 US projects worth about $130 billion blocked or delayed in the first quarter of 2026 alone, with active opposition groups more than doubling to 833 across 49 states. Nobody in that movement won this one by proving harm from noise or water. They won it by reading a publication schedule.

I'm Robert. I'm not a lawyer and I'm not a grid engineer — I read the opinion, the statute, and the filings and tried to put them in plain English, and I'd rather you check me than take my word for it. If you work in county planning, run a substation, or live close enough to hear a generator test, you know things I don't. New one every week: subscribe on YouTube and set me straight in the comments.

Common questions

Keep reading