A field in Pittsylvania County, Virginia, has just become one of the largest data center plans in America, and the most important number in it is not the largest one. Stack Infrastructure plans up to 21 data centers at the Southern Virginia Megasite at Berry Hill, needing 3,000 MW at full build-out, with the company talking about roughly $100 billion of investment over three decades. The land deal underneath that headline is more modest and more binding: a performance agreement commits Stack to at least $73.5 billion over 30 years and at least 2,050 jobs, and the purchase agreement for up to 2,990 acres of the 3,528-acre megasite, at $238,000 per acre for a total of about $737.8 million, can be walked away from if one condition fails. That condition is first power.
The condition, as reported from the purchase agreement, is that Appalachian Power must confirm it can provide at least 299 MW to the project by the end of 2028. The first phase of the land purchase, at least 1,000 acres, must close by June 2027, with all phases closing by June 2031. But the megawatts come first in the logic of the deal: without a utility document saying 299 MW will be there, the rest of the transaction does not have to happen. That inverts the usual order of a megasite announcement. The land is not the constraint, nor is water — Eden has agreed to supply up to 6 million gallons a day against a projected peak need of 7 million — nor gas, with a Williams Transco pipeline gate station on site. The constraint is the first block of grid power, a tenth of the finished campus.
That ratio is the story. A 299-MW commitment by the end of 2028 is a near-term test with a deadline and a walk-away clause attached. It is also the only part of the power plan that depends entirely on someone else’s system. Everything after it — the two on-site natural gas generation campuses, the later buildings, the long ramp to full size — is Stack’s to build, permit, and finance. The first 299 MW is Appalachian Power’s to deliver, through a substation that does not yet exist at the scale the campus needs, on a system that is already fielding far more large-load interest than it has signed.
The gap between a megasite and a substation
Start with what is actually on the ground at Berry Hill today. The project summary filed October 8 with the Virginia Department of Environmental Quality and the U.S. Army Corps of Engineers — a 92-page document describing what Stack calls Project Crossroads — says the site has an existing substation with a 30 MVA transformer and 138 kV lines. A 30 MVA transformer is, in rough terms, about a hundredth of what the full campus needs. The megasite’s own marketing material has described the substation as offering 100 MW of immediately available power, a figure that sits uneasily next to a 30 MVA nameplate. Marketing “available power” and deliverable capacity under a utility’s planning criteria are different claims, and only the second counts when a tenant wants to switch on servers.
To use the first block of grid power, the campus needs a 150 MW substation by mid-2028, sited on the 80-acre southwest energy campus, according to the filing as reported. That is a substation to be designed, equipped, connected, and tested inside roughly twenty months from the filing date, and every step has its own lead time. Large transformers and breakers are not shelf items; they are ordered years ahead in the current market, and a substation due in the second quarter of 2028 needed its equipment conversation to start well before the permit summary was filed. The filing describes a clear path to the substation — the site already has 138 kV lines and an existing substation footprint to build from — but a clear path is not a completed substation, and the distinction is where first-power schedules usually slip.
Meanwhile the site work is already moving, which is exactly what a developer should do when the power clock is the long pole. On October 16, the Danville-Pittsylvania Regional Industrial Facility Authority voted to let Stack begin grading about 400 acres before completing the land purchase, for a $250,000 fee, with the county’s economic development director citing speed in meeting customer demand. Grading is not energization, and no tenant has been named. But graded pads are the visible half of readiness, and they create a particular pressure: once earth has been moved, every month without first power is a month of capital sitting dark.
A queue that dwarfs the promise
The reason that bet is not a safe one is visible in Appalachian Power’s own regulatory record. In its Virginia rate case filed this spring, case PUR-2026-00044 before the State Corporation Commission, the company told regulators it has contracted 540 MW of data center load in Virginia, with two customers under signed electric service agreements, worth about 3.8 million megawatt-hours a year. It also listed 24,881 MW of potential load from 36 potential customers in various stages of inquiry or negotiation, and it said it has no new large-load customers in service yet.
Against the request queue, Berry Hill at full size — 3,000 MW — is about 12 percent of everything Appalachian Power is being asked for in Virginia, and more than five times what the company has actually signed. The 299-MW first block alone is more than half of the company’s entire signed Virginia data center load. And the public record does not make clear whether Stack’s 299 MW sits inside the 540 MW already signed or comes on top of it. That ambiguity matters: it determines whether the first-power commitment reallocates an existing reservation or creates a new obligation the utility must resource. Until Appalachian Power’s confirmation document exists — the very document the purchase agreement requires — outsiders cannot tell which it is.
There is a structural reason the utility cannot simply absorb the request the way a larger system might. Appalachian Power is a Fixed Resource Requirement utility inside PJM, which means it does not buy capacity in PJM’s auction and must line up its own supply through its own plants or contracts. Every megawatt it promises a data center has to be sourced, not just delivered through wires. The company’s owned coal and gas plants total roughly 5,900 MW, according to the testimony as summarized in coverage of the case: Berry Hill at full build would equal about half that fleet. Nobody proposes serving the full 3,000 MW from those plants — that is precisely why Stack is building its own generation — but the comparison explains why the utility’s promise stops at 299 MW. The first block is sized to what the existing system can plausibly carry.
The terms on which that first block would be sold also explain why the utility can afford to be careful. Under the large-load terms the commission approved, a load of Stack’s size signs for at least 14 years, with up to four years to ramp up, and pays for at least 80 percent of its contracted demand every month whether it uses it or not, posting collateral equal to 24 months of its expected bill unless it holds a strong credit rating. Those terms protect other customers from stranded investment if a campus stalls, and they are the reason a developer reserves power only when it means it: reserving 299 MW means paying for most of it from the day service is available, full halls or not.
Why the rest is Stack’s to build
Beyond the first block, the plan stops being a utility story and becomes a generation-development story. The filing describes two on-site natural gas generation campuses: the 80-acre Southwest Energy Generation Campus, which also hosts the new substation, and a 125-acre Northeast Energy Generation Campus that would power the rest of the site. Their sizes in megawatts and their start dates are not disclosed in the filing. Until Stack files the size, technology, and schedule of its gas plants, the balance beyond 299 MW is an aspiration with acreage attached, not a plan with megawatts attached.
The gas itself is the least of the problems. The Transco gate station on site removes the pipeline question that kills many behind-the-meter proposals before they start. The machines that burn the gas are the harder part: gas turbine backlogs are long across the industry, with GE Vernova reported in early October to hold a 116 GW turbine backlog and slot reservations, and a campus building gigawatts of its own generation joins the same equipment queue as every utility and developer ordering turbines this decade. On-site generation also does not escape permitting; it moves the burden from the transmission system to air quality, water, and interconnection of the generation itself, under the same joint state-and-federal review already handling the campus filing, with a permit decision expected by mid-March 2027 and construction tied to the permit expected to start in the third quarter of 2027.
There is also a political logic to building the generation on site that has sharpened in the last month. The county’s economic development director has put the point plainly: one of the key ways of ensuring there is no impact on ratepayers is generating the electricity on site. That argument lands differently now, as Virginia’s debate over who pays for the data center build-out — existing customers or the new loads themselves — has moved to the center of the state’s regulatory politics. On October 19, the same evening the filing’s details were published, protesters in downtown Danville called for a statewide pause on new data center approvals, citing power bills, drought, and water, and pointing to Berry Hill by name. A campus that generates most of its own power is, in part, an answer to that protest before it is fully formed. Whether regulators and neighbors accept the answer will depend on what the gas plants actually emit, consume, and cost — details the filing does not yet supply.
What siting actually means now
Berry Hill is therefore a useful correction to the older idea of what a megasite is. The classic megasite pitch — thousands of cleared acres, a gas pipeline, water capacity, a rail spur, a highway — describes a site ready for industry in general. The Berry Hill filings describe something narrower: a site is ready for a data center only when a specific substation, a specific queue position, and a specific contract line up in the same two-year window. The megasite spent years being made construction-ready, including earlier Appalachian Power substation and transmission-line work under the state’s business-park program, and it is still, today, a site with a 30 MVA transformer waiting on a 150 MW substation. Construction-ready and power-ready are different states, and the gap between them is now the binding constraint on the whole sector in Virginia.
That gap is also why the siting question has migrated from economic development agencies to utility planning departments. Stack is reported to have examined up to ten other sites in Pittsylvania County, with none matching Berry Hill on land, water, gas, and transport. In the old framework, that was choosing a location. In the new framework, the developer is choosing a position in Appalachian Power’s queue, a substation delivery slot, and contract terms that make the first block of power bankable. When the purchase agreement makes 299 MW a walk-away condition, it simply writes the new framework into a contract: the site is worth $737.8 million only if the power arrives.
The honest way to watch this project, then, is to ignore the 3,000-MW headline for a while and track the smaller numbers in order. First, the permit decision expected by mid-March 2027. Second, the first-phase land closing by June 2027, which shows whether Stack will own the ground before the power is confirmed. Third, the 150 MW substation by mid-2028 — transformers delivered, lines connected, testing complete. Fourth, the confirmation document itself: Appalachian Power stating, in writing, that at least 299 MW will be available by the end of 2028. And finally, the disclosure that would turn the rest of the plan from acreage into megawatts: the size and schedule of the two gas campuses.
If those markers land in sequence, Berry Hill becomes a template — the case study for how a rural megasite, a mid-size utility, and a developer’s own generation combine to deliver a gigawatt-scale campus without waiting for the bulk grid to catch up. If the substation slips or the confirmation does not arrive, the project does not collapse; it pauses, with graded pads and a performance agreement and a walk-away clause, waiting for the power system to do the one thing it has not yet done. Either way, the lesson generalizes well beyond Pittsylvania County: in the current build-out, the scarce thing is not land, not water, not gas, and not even money. It is the first megawatts, delivered on a date certain, through a substation that exists. First power is the part that has to be true first.