The biggest data-center campus ever planned just died. The pipeline didn't notice.

On Thursday afternoon, July 2, lawyers for QTSQTSOne of the largest US data-center developers, owned by the private-equity firm Blackstone. filed a short notice with the Supreme Court of Virginia. With it, the Prince William Digital Gateway ceased to exist: a 2,100-acre corridor along Pageland Lane in Northern Virginia, 37 planned buildings, more than 22 million square feet of server halls, a county-estimated $24.7 billion of investment and a projected $400.5 million a year in tax revenue, both figures produced during the approval fight, and best read as promotional projections rather than audited numbers. 'After careful consideration, QTS has made the decision to terminate the Digital Gateway project and withdraw its associated filings,' the developer said in a statement reported by InsideNoVa, promising a 'responsible wind-down.'

Our decision to fight an enormous and inappropriate data center project threatening one of America's most hallowed historic places has been completely vindicated.

Be clear about what the opponents were fighting for, because on their own terms they won completely. The American Battlefield TrustAmerican Battlefield TrustA nonprofit that preserves Civil War battlefields; it was a lead plaintiff against the Digital Gateway, which bordered the Manassas battlefield. and the Oak Valley Homeowners Association never claimed they could shrink Virginia's data-center buildout. Their goal was locational: to keep 22 million square feet of industrial buildings off the doorstep of the Manassas National Battlefield, where Stonewall Jackson earned his nickname. That land will now not be a data-center campus. Full stop, full victory.

But the case has been read, by both sides, as something bigger: a template for whether courts and county boards can check the largest infrastructure buildout in modern American history. So this article asks a narrower, more answerable question: did the cancellation of roughly 2 to 3 gigawatts of planned capacity actually change the regional supply picture? Or was the pipeline around the Gateway already so large that the biggest data-center cancellation on record rounds to zero?

The Gateway did not die on the merits. Virginia law required Prince William County to publish proper newspaper notice before the December 2023 rezoningrezoningA local government decision that changes what can legally be built on a piece of land. vote; the ad that ran was defective, and the county's own attorney warned supervisors before the 4-3 vote that proceeding carried 'significant risk.' They proceeded anyway. In August 2025, Circuit Judge Kimberly Irving voided the rezoning, writing that the defective notice was 'solely caused by the County.' The Virginia Court of Appeals affirmed unanimously on March 31, 2026.

That sequence matters for how you assign agency. No court ever ruled that the Gateway was too big, too loud, or too close to the battlefield; the rulings were about procedure. What the litigation did was impose three years of delay and uncertainty, and in this industry, delay is death by other means. The county quit defending its own vote in April after spending $1.72 million on the effort. Compass DatacentersCompass DatacentersA Dallas-based data-center developer that was QTS's partner on the Digital Gateway until it exited in April 2026., QTS's partner, exited the same month citing 'compounding regulatory hurdles.' By July 2, QTS was the last party standing, and it chose to stop. The court created the conditions; the developer made the call.

Here is how we sized the question, step by step, so you can check the work. First, the numerator: how big was the Gateway? The developers' plans combined Compass's share of up to 1,700 megawatts with QTS's share, reported at roughly 60 percent of Compass's, a derived total of about 2,700 megawatts. Contemporaneous local reporting bracketed the project at '2 to 3 gigawatts.' We use the range wherever it changes a conclusion.

Second, the denominators. We compared the Gateway against three yardsticks: the announced-or-approved capacity in the four counties one ring further from Washington (paper plans against paper plans, the fairest comparison); Dominion EnergyDominion EnergyThe investor-owned utility that supplies most of Virginia, including the world's densest data-center cluster.'s contracted data-center queue (the utility's order book); and the built colocationcolocationData-center space a landlord builds and rents to many tenants, rather than a tech company building for its own use. market's vacancy (the tightest measure of current scarcity, used only as context).

Drive 30 to 50 miles southwest from Pageland Lane and you cross into a belt of counties, Culpeper, Caroline, Fauquier, Stafford, that were rural a decade ago. By the time QTS filed its termination notice, those four counties had announced or approved roughly 3,900 megawatts of data-center capacity: about 1.4 times the Gateway's derived plan, or somewhere between 1.3x and 2x on the honest range. Culpeper County alone approved seven campuses totaling about 2,402 megawatts by August 2024, with an eighth 900-megawatt proposal pending that would push it past 3,300.

Two caveats keep this comparison honest. First, the timing cuts against the most dramatic reading. Spotsylvania's AWS rezonings were approved in October 2023, Culpeper's seven campuses by August 2024, before the first court ruling, and Caroline's and Stafford's projects were announced in November 2025, before anyone knew whether the Gateway's appeal would succeed. Nothing here shows the Gateway's demand relocating. It shows something quieter and arguably more telling: the same demand wave was building the ring regardless, at a scale that made the Gateway's fate a local matter, not a regional one.

Second, paper is not concrete on either side of the ledger. The comparison sets a never-built plan against a mix of approved, under-construction and merely announced projects, and announced megawatts in this industry attrite heavily, some of those 3,900 will never be built either. Nor are the megawatts perfectly standardized: developers variously report utility feeds, IT load, or ambitions. This is where the comparison breaks, and why we treat it as an order-of-magnitude test, not a precise offset. Only Caroline's CleanArc campus, under construction since November with its first 300 megawatts due in early 2027, has moved from paper to ground among the big ring entries.

The second yardstick is the utility's order book. When the Gateway's rezoning passed in December 2023, Dominion Energy had disclosed about 16.5 gigawatts of contracted data-center capacity (as of that July). By March 2026, the figure was roughly 51 gigawatts, it tripled during the lifetime of the litigation, of which about 10.4 gigawatts sit under final Electric Service Agreements, the binding last stage before power flows.

One arithmetic point matters here, and the first draft of this analysis got it wrong: the Gateway was never in that queue. It died before reaching the contracting stage, so its termination subtracts nothing from the 51 gigawatts. The right framing is counterfactual, had it been contracted at its derived 2.7 gigawatts, it would have equaled about 5 percent of the current queue, roughly what the queue added in the single most recent quarter disclosed (48.5 to 51 gigawatts between December and March). The largest data-center cancellation on record is, to Virginia's dominant utility, about one quarter's worth of new orders.

Northern Virginia's colocation vacancy fell to 0.5 percent in the second half of 2025, 21.5 megawatts sitting empty in a 4,040-megawatt built market, the lowest of any primary US market, with 96 percent of 2026's incoming supply committed before it even delivers. It is tempting to read that as the market shrugging off the Gateway, and an earlier version of this analysis did. That inference is wrong, and we want to be explicit about why: vacancy measures built space today, while the Gateway would have delivered years from now. A cancelled 2029 building cannot show up in 2025 vacancy no matter how much it matters.

If anything, the number cuts the other way. A market this tight is the market least able to afford lost future supply, which makes it all the more striking that the pipeline math above still shows no material loss. The scarcity is real; the Gateway's irrelevance to it has to be established on pipeline evidence, not vacancy, and that is what the previous two sections did.

The evidence admits more than one explanation, and it is worth laying them side by side rather than picking a winner by assertion.

• Immateriality: the AI demand wave is so large that no single project, even the largest ever planned, is load-bearing. This is the reading best supported by the numbers here, the ring out-booked the Gateway before it died, and Dominion adds a Gateway-equivalent to its queue roughly every quarter.
• Displacement: the Gateway's specific demand slid outward to weaker jurisdictions. This is the dramatic reading, and the evidence does not currently support it, every ring project predates the termination, and neither QTS nor Compass has been reported acquiring ring-county sites since the August 2025 ruling. It would become supportable if either developer, or a known Gateway tenant, surfaces one county out in the coming months.
• Common cause: the same scarcity that made the Gateway worth $24.7 billion on paper was independently driving the ring approvals all along. On this reading the Gateway and the ring are siblings, not substitutes, the buildout was always going to happen in both places, and the court fight determined only which one it skipped.

Note what all three readings share: none of them says the court victory reduced Virginia's data-center buildout. The preservationists won the battlefield's viewline, fully and legitimately. What nobody won, because nobody was fighting for it, was a smaller total.

The costs of this outcome land on identifiable groups, and they are not the ones who spent three years in court. Start with the Pageland Lane landowners, farm families who had contracted to sell into the project at data-center prices and were among its loudest supporters. Their life-changing payouts died with the termination; they keep their land and lose the windfall. Prince William County, meanwhile, loses whatever fraction of the projected $400.5 million a year in tax revenue would have materialized, a county projection from the approval fight, so treat the precision skeptically, but even a fraction of it is real money for schools and services, plus the $1.72 million the county spent defending a vote its own attorney warned against.

Then there are the ring counties' residents, who now live beside the buildout that exurbanexurbanThe ring of semi-rural counties beyond the built-up suburbs. politics invited in. Their county governments are not victims, data centers thicken a rural tax base dramatically, which is exactly why Culpeper and Caroline said yes. The strain shows up somewhere else: on the grid infrastructure. What is thin in the ring is not the tax base but the utilities.

Rappahannock Electric Cooperative, the customer-owned utility serving 183,000 accounts across much of this territory, has an all-time system peak of about 1,150 megawatts, meaning a single large modern campus can request roughly as much power as everything REC has ever served at once, a comparison its president and CEO, John Hewa, has drawn publicly. Statewide, nearly 17 gigawatts of data-center load is projected across Virginia cooperative territory by 2040. One important caution: Virginia service territories are a patchwork, Dominion serves swaths of all five ring counties and the Town of Culpeper runs its own municipal utility, and we could not confirm project by project which utility serves each ring campus. The scale mismatch is documented; the precise allocation of it is not, and we flag that squarely in the limitations.

Each of the open questions above comes with observable evidence that would settle it, mostly within a year.

• Displacement confirmed: QTS, Compass, or Blackstone-linked entities acquire or option land in Culpeper, Caroline, Fauquier, Stafford or Spotsylvania. Land records and county planning dockets would show it within months; that would revive the 'megawatts moved' story this article declines to tell.
• Ring attrition: Culpeper's pending eighth campus stalls, or CleanArc misses its early-2027 first-300-megawatt delivery in Caroline. That would shrink the 3,900-megawatt figure and weaken the immateriality finding at the margin.
• The land's second act: Prince William's board, now chaired by the data-center critic who unseated the project's champion, entertains new proposals for Pageland Lane, or the preservationists move to lock in conservation easements. Either would reveal whether July 2 was an ending or an intermission.
• The co-op squeeze becomes measurable: REC or Virginia's cooperative association discloses data-center interconnection requests, or a rate case attributes transmission costs to campus-scale load. That would convert the scale-mismatch warning here into documented ratepayer impact.