Virginia Data Center Order Puts AI Growth on a Meter

The Virginia Data Center Order is an attempt to make AI infrastructure growth legible before it becomes harder to govern. Governor Abigail Spanberger signed Executive Order 22 on 18 September, launching immediate agency work on data-centre accountability and creating an AI task force. The framework targets transparency, local approval, energy costs, environmental safeguards and workforce planning.

The order matters because Virginia is a major global data-centre hub. New AI capacity can bring investment and tax revenue, but it also concentrates electricity demand, water use, construction impacts and grid costs. The framework tries to put those trade-offs into public decisions instead of treating a server campus as an ordinary commercial building.

The order starts agency action now, while several broader restrictions still depend on legislation or later rulemaking.

What the Virginia Data Center Order changes

The governor’s announcement says the framework will restrict secrecy around projects, strengthen community participation, examine how large users pay for energy infrastructure and raise environmental expectations. Executive action can direct state agencies and change administrative practice. Other ideas—especially broad permitting rules or statutory limits—will need legislation, formal rulemaking or both.

That distinction prevents the announcement from being overstated. Cardinal News reported that the plan stops short of a moratorium. Virginia Mercury described a mix of directives and proposals, while FFXnow detailed the 25-megawatt threshold discussed for stronger local review. Across the three independent reports, the common picture is a policy framework beginning now, not a completed regulatory code.

The AI task force connects software risk to physical capacity

The new task force is charged with concerns including workforce displacement, privacy and cybersecurity. Placing it beside the data-centre framework recognises that AI policy is not only about model behaviour. Training and serving models require land, power, cooling, networks and specialised labour. Decisions about physical capacity can shape which systems are built and who bears their costs.

The harder question is measurement. Policymakers need comparable disclosures on peak demand, backup generation, water use, noise, construction schedules and grid upgrades. Communities need enough information to evaluate a project without exposing genuine security details. Utilities need rules that keep speculative interconnection requests from shifting costs to households.

Other jurisdictions are exploring the same connection between compute and electricity. Lapaas Voice’s analysis of the AI Energy Management Alliance showed why flexible demand is becoming part of data-centre strategy. Our report on Google and Stegra’s green-steel certificates illustrates how infrastructure buyers can also influence upstream emissions.

What operators should do now

Developers planning Virginia capacity should map which elements of a proposal rely on state discretion, local approval or future legislation. They should prepare consistent figures for energy, water, noise and employment, then explain how those estimates change at full build-out. A promise about an initial phase is less useful when later phases multiply the load.

Enterprise buyers should ask where their cloud capacity sits and whether regulatory changes could affect price or delivery timing. Investors should separate immediate administrative actions from proposals that must survive a legislative session. Community groups should focus on enforceable thresholds and reporting rights rather than headline claims about being the strongest framework.

Immediate directives and later law are different layers

The official release presents a single accountability framework, but its implementation routes are not identical. An executive order can instruct agencies to change how they evaluate state-supported sites, assemble information and coordinate policy work. It cannot, by itself, enact every statewide land-use or commercial restriction described in the wider platform. Cardinal News and Virginia Mercury both emphasise this divide between actions placed in motion and proposals intended for the 2027 General Assembly session.

Reading the order requires separating measures agencies can start from proposals that still need legislation or formal rules.

That sequencing affects businesses now. A developer interacting with a state agency may encounter new disclosure expectations before the legislature finishes debating broader rules. Localities, meanwhile, should not assume that every headline proposal has already displaced their existing authority. The defensible approach is to identify the legal route for each requirement, the agency responsible and the date on which it becomes enforceable.

Transparency is a good example. The governor’s release calls for ending secrecy around commercial data-centre projects. FFXnow reports that the order’s immediate ban on nondisclosure agreements applies to state agencies or employees, while a wider ban would require additional action. That narrower scope still matters: state-controlled information can shape negotiations and public understanding. But it should not be described as a complete ban covering every private developer and local official today.

The 25-megawatt threshold creates a policy dividing line

The framework discusses stronger local approval for projects using more than 25 megawatts. A numerical threshold can make oversight more predictable, yet it also creates boundary questions. Policymakers will need to define whether the test applies to a single building, a campus, phased developments or related projects under common control. They will also need a consistent measure of expected demand rather than a figure selected only from the first construction phase.

For developers, the consequence is a need for a full-build energy narrative. That should include expected peak load, ramp-up timing, onsite generation and material changes that trigger another review. For communities, the threshold is useful only if the underlying demand estimate is visible and can be challenged. Otherwise, the number risks becoming a formal gate without a dependable measurement behind it.

Comparable project-level evidence is the bridge between a broad accountability promise and an enforceable approval decision.

Energy costs are the central commercial question

The governor’s release says the framework should protect households from higher energy bills and require data centres to carry a fairer share of related costs. The reporting reviewed for this package does not establish a finished tariff or a single cost-allocation formula. That absence is important. Grid upgrades can include generation, transmission, substations and local distribution, and the party paying for each element may differ.

A durable policy will have to distinguish infrastructure built specifically for a large customer from broader investments that benefit the system. It will also need rules for projects that reserve capacity and later shrink, delay or cancel. Without those details, a commitment to fairness remains a direction rather than a bill that utilities can calculate. Operators should therefore model several cost scenarios instead of assuming today’s connection terms will persist unchanged.

Environmental and workforce promises need comparable records

The framework connects data-centre approvals with water, air, land, noise, clean energy and local employment. These categories can be measured, but only if projects report them on a consistent basis. Water figures should distinguish withdrawals from net consumption and identify the cooling assumptions behind both. Noise analysis should cover normal operations and backup generation. Employment claims should separate temporary construction jobs from permanent technical and facilities roles.

Comparable records also make trade-offs visible. A design that lowers water use may consume more electricity; a site with fewer nearby homes may require more transmission. The order does not eliminate those choices. It creates a basis for agencies and communities to see them together. The AI task force can add a second layer by examining whether workforce disruption, privacy and cybersecurity concerns require safeguards beyond the physical permit.

The Virginia Data Center Order does not settle the state’s argument over AI growth. It creates a common ledger for that argument: what a project consumes, who approves it, who pays and what safeguards apply. Its success will depend on turning that ledger into public, comparable evidence and then acting when the numbers fail the promised standard.