Virginia draws its data center line at 25 megawatts

Gov. Abigail Spanberger's Sept. 18 executive order excludes data centers with anticipated peak demand of 25 megawatts or more from Virginia's site development assistance and expedited permitting programs, bars state agencies from using nondisclosure agreements to keep proposed projects and their incentives confidential, and opens a rulemaking on noise, emissions, water use, electricity demand and other impacts, with agency recommendations due at 120, 180 and 240 days. Spanberger's office has said she will pursue additional data center rules in Virginia's 2027 General Assembly session.

That would amount to routine statehouse posturing almost anywhere else, but Virginia has been the default answer to where American data centers go for so long that the question stopped being interesting; the Construction Dive report on the order calls it the data center capital of the United States before noting that it has become more selective about which projects it helps. Within a week of Spanberger's signature, Maryland Gov. Wes Moore signed an executive order of his own on data center guardrails, and opposition to data center construction is rising in Virginia and other parts of the country.

The instinct is to score this as an incentives story because the first clause reads like one, and the incentives are the cheapest thing on the page: state site development assistance is a line item within reach of a sponsor financing a 25-megawatt campus. Expedited permitting is a schedule, and schedule is what digital infrastructure underwriters are actually buying. Virginia took a little money and a material amount of time, and only the second is difficult to replace.

The two programs the order touches are the state's tools for attracting the projects it wants, and removing them for the largest projects changes Virginia's posture from competitor to gatekeeper without changing the underlying approvals those projects need. It does not make a 25-megawatt data center illegal in the commonwealth; it declines to help build one and declines to hurry one, which is a cheaper political move than a prohibition.

The near-term construction effect looks close to nil. Ked Whitmore, who chairs Venable's land use and zoning practice, told Construction Dive that most immediate impacts run to incentive agreements and nondisclosure agreements rather than to construction, landing far earlier in the development process than a site crew would; the consequences builders care about will arrive with the rulemaking, whose scope—noise, emissions, water, cooling, and where these facilities can be built at all—reaches the building rather than the paperwork around it.

Twenty-five megawatts of anticipated peak demand is a load figure, not a land figure, and load is the quantity the interconnection queue already governs; by writing its gate in the grid's unit, Virginia has attached a second set of consequences to a number that a utility ultimately determines. A project's eligibility for state assistance now moves with the same variable that decides when it can energize, so a sponsor's state-approval risk and its power risk are correlated rather than independent. Portfolios prefer independent risks.

The order's trigger is also a projection rather than a measurement: anticipated peak demand is an estimate made before the building exists, so the line between an assisted project and an unassisted one rests on a forward number rather than a fact and asks whoever signed the application to stand behind it. That is a workable administrative test and an awkward one to litigate, and it means the state's siting regime and its grid are reading from the same spreadsheet.

The rulemaking reaches past siting into design: Venable's takeaways document points to the agency recommendations as the point where the next set of changes will be specified, and Whitmore expects those rules to shape how these buildings are designed, not only where they are placed. What the reporting does not say is whether projects already holding expedited approvals keep them; that is the first question any pipeline under construction needs answered.

What the order does not reach matters too: the levers it names are state assistance and state permitting, and the agencies it binds are state agencies through their use of confidentiality agreements. Local land-use approval is a different currency, and the coverage does not describe changes to it.

The nondisclosure ban will matter more than the assistance cutoff, because confidentiality has been the mechanism by which a locality and a sponsor negotiate incentives without building a public record; removing it moves the first public account of a project to an earlier point in its life, when the applicant has less to show and critics have more time to organize. That is an inference about how the politics will run rather than a fact in the order, and Virginia appears willing to accept near-term noise in exchange for fewer fights after a site is under control—a trade a state with thinning patience for hearings might reasonably make.

Put the two halves together and the order reads more like a conversion than a halt: Virginia has not stopped 25-megawatt projects, it has declined to subsidize their schedule, declined to keep their negotiations private, and promised a rulebook in 2027. The executive order functions as scaffolding for legislation, letting the visible steps arrive now and the durable requirements arrive with a vote attached.

That Venable produced a client takeaways document at all says something about how the market is treating the order: as a diligence item rather than a political story. Sponsors with Virginia land under option have to price both the review and the chance that the 2027 session makes the rules permanent.

Executive orders carry a property that statutes do not—rules that arrive by order can be withdrawn by order, and rules that arrive by statute cannot—and that cuts both ways for a sponsor underwriting a long-dated asset; the answer is not obvious while the political mood is still moving.

For a developer, the operative fact is the length of the fuse: rules land at 120, 180 and 240 days, so a sponsor optioning land in Virginia this autumn is committing capital ahead of a rulebook it cannot read, and in most cases site control will be signed before the recommendations are written. The backdrop is a pipeline where planned project counts nearly tripled over five months while power emerged as the bottleneck, and a state gate in front of an interconnection queue does not remove the queue; it adds a second approval sequence to the same critical path.

As this publication has argued, consent is the product in every data center trade and grid permission is the underwriting asset, with queue positions and interconnection contracts pricing before electrons do. Virginia's order cuts both ways against that argument: in its favor, the state has moved public consent out of the county hearing room and into the permitting file, the same maneuver Texas made in August, when Abbott's halt folded community and ratepayer support into the same diligence binder as the interconnection agreement. Against it, nothing in the order decides what gets built this year, and Whitmore's assessment that current construction activity is largely untouched makes the order a price on consent rather than an enforcement of it—a compliance cost attached to a schedule instead of a veto on a project.

The distinction holds until the 120-day recommendations arrive: if they come back with siting limits rather than procedural standards, the cost moves off the calendar and onto the map, and a Virginia site stops being a Virginia site with slower paperwork.

The capital consequence of the order is a repricing of Virginia option value rather than a change in Virginia construction. A sponsor weighing a Virginia site against alternatives now carries two things it did not carry a month ago: a state approval sequence measured in months, and the possibility that the 2027 session writes 25 megawatts into statute or lowers it. Neither is fatal, but together they push the marginal project toward jurisdictions where the state's job is still to shorten the schedule, and the marginal project is where the next increment of capacity gets decided.

The number to watch is whether 25 megawatts survives contact with the process, because Maryland's guardrails order gives the region a chance to converge on a single threshold: convergence would turn the figure into a term every sponsor prices the same way, divergence would make it a reason to site on one side of a border and not the other. Either way, the permanent version waits for the 2027 session, and the first draft is due in 120 days.