Virginia Data Center Rules Put Local Control Ahead of Faster AI Expansion

Virginia Governor Abigail Spanberger signed one executive order that confronts two connected pressures: rapid data center construction and the risks created by artificial intelligence. The new Virginia data center rules immediately restrict secrecy inside the executive branch. They also begin a broader effort to give communities more influence over future projects.

Spanberger’s order does not freeze construction or instantly rewrite local zoning law. Instead, it uses state agencies to increase transparency, examine environmental effects, and create planning tools for local governments. Her larger framework proposes stricter approval requirements for facilities consuming more than 25 megawatts.

That distinction creates the central tension. Virginia wants to remain a major home for cloud computing and AI infrastructure. However, its government now argues that faster development cannot come at any cost to residents, electricity customers, or local planning.

The state’s new AI task force widens the issue beyond physical infrastructure. It will examine workforce displacement, privacy, cybersecurity, and other risks associated with expanding AI use. Yet the most immediate policy actions concern the buildings, generators, transmission lines, and utility costs supporting that technology.

Virginia is therefore testing a harder question than whether AI investment should continue. It is asking who gets to approve that investment, who receives its benefits, and who pays for the infrastructure behind it.

Virginia Data Center Rules Start With Secrecy and State Agencies

The executive order changes state government practice immediately, while leaving the most consequential restrictions for legislators.

Spanberger signed Executive Order 22 on September 18, 2026. The order establishes data center accountability initiatives and creates the Virginia Artificial Intelligence Task Force.

The most direct rule bars executive branch agencies and officials from entering into or requiring nondisclosure agreements for commercial data center projects. These agreements can prevent public scrutiny while officials and developers discuss land, incentives, infrastructure, or other project details.

The prohibition applies to agencies, departments, boards, commissions, authorities, and employees supervised by the governor. It does not automatically prohibit local governments from signing similar agreements.

That boundary matters because counties and cities oversee many land-use decisions. A statewide ban affecting local agreements would require legislation rather than an executive directive.

Spanberger’s administration plans to seek that legislation during the 2027 General Assembly session. Until then, the order ends one channel of secrecy without closing every channel.

The signed executive order also directs agencies to develop a Data Center Planning and Community Engagement Toolkit. This resource will help local governments evaluate proposals and negotiate agreements with developers.

The toolkit matters because localities do not enter negotiations with equal technical resources. A developer may employ specialists in energy procurement, land use, environmental permitting, and tax policy. A smaller county may confront the same issues with a limited planning staff.

Model disclosures and community benefit frameworks can narrow that gap. They can help officials ask for comparable information before approving projects with decades-long effects.

The administration also wants to end by-right approval for data centers using more than 25 megawatts. By-right development can proceed when a proposal meets existing zoning conditions, without requiring a discretionary public vote.

Removing that route would push large projects toward public hearings and specific local approval. Communities would gain a visible decision point before a qualifying facility could move forward.

That proposed change is not yet law. It belongs to Spanberger’s broader accountability framework, which includes both executive actions and requests for future legislation.

The difference between those categories is essential. The order initiates reviews, planning work, and executive branch restrictions. The General Assembly must enact several measures that would directly reshape private development.

Spanberger also proposes removing future large data centers from Virginia’s expedited permitting process. That policy would slow the route available to projects on sites prepared for industrial development.

Taken together, the transparency provisions challenge the assumption that speed should dominate data center policy. Virginia’s earlier economic strategy emphasized predictable approvals and infrastructure access. The new approach places public participation and disclosed costs closer to the center.

That is a meaningful change, even before lawmakers decide whether to adopt every proposal. Developers now know the governor wants large projects treated as public choices, not routine industrial applications.

Why Virginia Is Tightening Its Data Center Policy Now

The policy shift follows measurable growth in electricity demand, not a theoretical debate about AI’s future needs.

Virginia hosts the largest concentration of data centers in the world. Northern Virginia developed that position through dense fiber connectivity, access to government and business customers, available infrastructure, and a long history of cloud investment.

Generative AI has intensified demand for computing capacity. Training and operating large models require servers, cooling systems, backup generation, and dependable electricity around the clock.

The strain appears clearly in electricity data. According to the U.S. Energy Information Administration, commercial electricity sales in Virginia increased by nearly 30 million megawatt-hours between 2019 and 2025.

That increase was faster than in every state except Texas, which has a much larger population and economy. The EIA identified the concentration of data centers as a leading driver.

Peak demand has also moved upward. Summer peak load in the Dominion zone reached 23,905 megawatts in 2025, a 23 percent increase from 2019. Winter peak load reached 25,413 megawatts during 2025 and 2026, up 45 percent from six winters earlier.

PJM Interconnection operates the regional grid covering Virginia and twelve other states, plus Washington, DC. It expects the Dominion zone to record the region’s largest absolute summer peak-demand increase from 2026 through 2030.

PJM’s forecast attributes much of that growth to data center load. Its projections therefore connect AI infrastructure decisions with generation, transmission, reliability, and long-term customer costs.

The federal electricity analysis also shows why small policy adjustments may be insufficient. Virginia is not accommodating a few isolated industrial sites. It is managing sustained growth across one of the world’s most concentrated computing markets.

Electricity is only one concern. Data centers can require water for cooling, occupy large land parcels, produce continuous equipment noise, and rely on diesel backup generators.

Those effects differ by location and facility design. They can also accumulate when several campuses operate near the same neighborhoods or depend on the same watershed and grid infrastructure.

Spanberger’s framework directs agencies toward standards covering energy efficiency, water consumption, land effects, noise, and backup generation. It also calls for closer examination of cumulative emissions from diesel generators.

The administration wants to accelerate state noise rules, although the final limits and implementation schedule remain unsettled. It also seeks clearer water-use standards and stronger enforcement when operators violate them.

The timing reflects political pressure as much as engineering pressure. Residents have demanded greater visibility into proposed developments and the transmission infrastructure serving them.

Local governments face a difficult bargain. Data centers can generate substantial tax revenue while demanding fewer public services than large residential developments. However, nearby communities may bear noise, visual, environmental, and infrastructure effects.

The facilities themselves also create relatively few permanent jobs compared with their construction cost and physical scale. Construction employment can be significant, but the operational workforce is much smaller than a factory of comparable size.

Spanberger’s workforce proposals respond to that imbalance. They call for local hiring, apprenticeship programs, responsible contractor standards, and greater use of Virginia businesses.

The state is no longer treating tax revenue as the only measure of a successful project. Its emerging test includes cost allocation, environmental performance, job quality, and community participation.

This broader accounting explains why the action arrived now. Virginia has reached a scale where every new campus interacts with choices made for the entire electric system.

The Central Tradeoff Is Local Control Versus Infrastructure Speed

Virginia’s plan gives communities more leverage, but that leverage can make development slower, less predictable, and more expensive.

Data center companies value predictable permitting because computing demand moves quickly. Cloud providers and AI developers often plan capacity across multiple regions, then place workloads where power and facilities become available.

Local hearings introduce scrutiny and uncertainty into that process. Residents can challenge traffic plans, generator use, water demand, noise levels, transmission routes, and the terms of community agreements.

That scrutiny is the point of the new policy. Spanberger’s framework treats large facilities as developments with regional consequences rather than ordinary occupants of industrial land.

The proposed 25-megawatt threshold illustrates this approach. A facility above that level would need explicit local approval instead of relying solely on existing by-right zoning.

Twenty-five megawatts is substantial continuous demand. However, it is also well below the potential consumption of a modern hyperscale campus containing several buildings.

The threshold would therefore reach many proposals connected to cloud services and AI. It would not merely capture a handful of exceptional megaprojects.

Industry groups are likely to focus on whether the rules remain clear and consistent across localities. A fragmented process can expose similar projects to different disclosure requirements, hearing schedules, and negotiated conditions.

The Data Center Coalition said it was waiting for more detail. Its representative, Nicole Riley, said the organization wanted workable policies that keep Virginia competitive and honor earlier state and local commitments.

That response identifies the industry’s strongest argument. Companies made long-term plans under policies that actively encouraged development. Abrupt changes can affect land investments, power agreements, and construction schedules.

Yet predictability works in both directions. Residents and utility customers also need predictable information about energy costs, public subsidies, environmental limits, and enforcement.

Nondisclosure agreements weaken that predictability for the public. A community cannot evaluate a project’s tradeoffs when crucial terms remain hidden until approval becomes difficult to reverse.

The new Virginia data center rules attempt to rebalance information before changing the market more directly. State agencies must disclose more, study impacts, and provide negotiation tools.

The larger framework goes further by proposing an end to state subsidies for data centers in site-development programs. It also seeks to remove large projects from expedited permitting.

Those ideas reject a long-standing development model centered on rapid site preparation. They replace it with a model where the developer must demonstrate local and system-wide value.

The conflict is not simply government versus technology companies. It is a contest between two definitions of responsible growth.

The industry emphasizes capacity, reliability, investment certainty, and Virginia’s position in global computing. The administration emphasizes disclosed impacts, accountable cost allocation, and consent from affected communities.

Both sides can point to real risks. Slower approvals could redirect projects and tax revenue elsewhere. Faster approvals could lock residents into costly infrastructure and environmental consequences.

Virginia’s importance makes this experiment nationally relevant. A less established market can impose restrictions without threatening an entrenched cluster. Virginia must govern the cluster while depending on it.

If the state can preserve investment while increasing public control, other data center markets may copy its approach. If projects rapidly shift elsewhere, opponents will argue that fragmented state policy cannot manage national AI infrastructure.

Electricity Costs Put Data Centers and Ratepayers on Opposite Sides

The hardest part of the framework is deciding how much infrastructure cost belongs to data centers rather than households and ordinary businesses.

Data centers require generation and transmission investments that can remain in service for decades. Utilities must plan before projected demand fully materializes.

That creates two related risks. If utilities build too little, Virginia may face delays, reliability problems, or higher market costs. If they build too much, other customers can inherit stranded infrastructure expenses.

A 2024 review by Virginia’s Joint Legislative Audit and Review Commission examined those risks. Its modeling found that data center growth could raise generation and transmission charges for a typical Dominion residential customer.

Under a constrained growth scenario, the estimated increase reached approximately $14 per month by 2040. Under higher demand scenarios, the estimated increase ranged from $33 to $37.

Those figures were presented in constant 2024 dollars and did not cover every component of a customer’s bill. They were scenario estimates rather than guaranteed future charges.

The legislative analysis nevertheless showed a structural weakness. Existing rates were not designed for sudden infrastructure growth serving a small number of exceptionally large customers.

Spanberger’s framework calls for utilities to allocate a more equitable share of generation and transmission costs to data centers and other large-load customers. Large load describes customers whose electricity needs can reshape utility planning.

Dominion has already created a separate GS-5 rate class covering major customers, including many data centers. The structure includes long contracts and specific treatment of generation, transmission, and distribution costs.

Long contracts reduce the risk that a customer reserves capacity and then uses far less than expected. They can also protect other customers when a proposed campus never reaches its projected size.

The State Corporation Commission has taken another step. It directed Dominion to design a tariff assigning certain high-voltage transmission costs directly to the customers causing them.

Microsoft has preserved its ability to challenge that decision in Virginia’s courts. The company said rates should remain transparent, fair, and connected to costs created by each customer.

That dispute shows why the framework’s promises will be difficult to implement. Everyone can support fairness while disagreeing sharply about which facilities caused a particular transmission project.

Electric networks serve interconnected groups of customers. A new line may accommodate a data center while also improving reliability or capacity elsewhere.

Cost allocation must separate those shared benefits from infrastructure built mainly for one concentrated load. Regulators will need technical evidence, enforceable contracts, and consistent accounting methods.

Spanberger also wants stronger upfront financial commitments from developers. The goal is to prevent speculative projects from reserving land, grid connections, and public planning resources without a credible path to construction.

This measure addresses the problem of duplicate or inflated power requests. A developer considering several locations may seek access in each region before choosing where to build.

Utilities must distinguish real demand from optional demand while planning years ahead. Poor forecasts can create shortages, unnecessary construction, or both.

The framework also asks that data centers absorb costs arising from PJM processes responding to their growing demand. That request reaches beyond Virginia because PJM operates a regional market.

Even aggressive Virginia data center rules cannot completely isolate state residents from regional prices. A project rejected in Virginia could move into another PJM territory while still affecting shared capacity and transmission costs.

The policy can still change who bears directly attributable expenses. It cannot turn an interconnected grid into a closed state system.

This is the sharpest test of Spanberger’s accountability claim. Transparency can arrive through executive action. Meaningful bill protection requires regulators, utilities, lawmakers, and courts to agree on complex cost assignments.

The AI Task Force Connects Physical Growth With Digital Risk

Virginia is treating AI governance and data center policy as one problem, although the task force’s authority remains less defined than the infrastructure directives.

Executive Order 22 creates the Virginia Artificial Intelligence Task Force to examine workforce displacement, cybersecurity, data privacy, and responsible AI development.

This places software risk beside electricity, water, noise, and land use. The connection reflects a practical reality: AI services depend on physical facilities, while facility demand increasingly depends on AI investment.

The task force gives Spanberger’s administration a mechanism for coordinating agencies. That can prevent privacy, labor, procurement, education, and security questions from being handled independently.

However, the immediate announcement offers fewer details about the task force’s deliverables than about data center policy. Its membership, reporting cadence, and priority recommendations will determine whether it becomes influential.

Task forces can create useful common standards. They can also delay difficult decisions by converting them into studies and meetings.

Virginia must therefore clarify which risks require administrative changes and which require legislation. Workforce training, government procurement, consumer privacy, and automated decision systems involve different laws and institutions.

The governor framed the task force partly as a response to limited federal action. That choice puts Virginia among states attempting to shape AI policy through their own authority.

State-level action can move faster and address local needs. It can also produce incompatible rules for companies operating nationwide.

Virginia’s position gives its decisions unusual weight. The state combines a large public-sector presence, dense technology infrastructure, major cloud operations, and extensive data center development.

It can observe AI from both sides of the market. Virginia’s organizations use automated systems, while its electric grid and localities support the computing facilities behind them.

Still, policymakers should avoid assuming every data center is an AI facility. These buildings also support storage, traditional cloud applications, streaming, government services, and ordinary business computing.

Likewise, not every AI risk can be solved through infrastructure policy. Privacy failures and biased automated decisions concern data practices and institutional controls, not only server capacity.

The task force will add value if it maintains those distinctions. It should connect related risks without treating electricity policy as a substitute for AI governance.

One useful area is state procurement. Virginia can define disclosure, security, testing, and human-review requirements for AI systems purchased by public agencies.

Another is workforce measurement. General predictions about job loss offer little guidance without identifying affected occupations, employers, regions, and available retraining paths.

Cybersecurity also requires defined responsibilities. AI can help attackers create convincing fraud or automate reconnaissance, but agencies already have established security obligations.

The task force must determine where AI changes those obligations instead of producing a broad list of familiar dangers.

Its first reports will reveal whether the administration intends to recommend enforceable safeguards or provide voluntary guidance. That distinction will shape how technology vendors and public agencies respond.

The task force also creates a political bridge between visible data center opposition and less visible concerns about AI systems. Residents may encounter facility noise or transmission construction before encountering a disputed automated decision.

Combining these concerns can build a broader constituency for action. It can also overload one initiative with separate problems that require different solutions.

For now, infrastructure policy remains the more concrete half of the order. The task force is a commitment to develop policy, not a completed set of AI rules.

What the Framework Still Does Not Resolve

Spanberger’s plan is more restrictive than Virginia’s previous posture, but it is not the construction pause demanded by some community groups.

The order does not impose a statewide moratorium. Projects can continue while agencies develop guidance and lawmakers consider the governor’s proposals.

That omission drew criticism from environmental advocates. The Piedmont Environmental Council argued that the framework does not stop new grid connections while power, water, air-quality, and community concerns remain unresolved.

The Chesapeake Climate Action Network called for an immediate pause on approvals until stronger protections take effect. That position treats continued construction as an irreversible risk.

Spanberger defended local decision-making instead of ordering a statewide stop. Some localities are already considering temporary pauses or more demanding review processes.

Her approach preserves development while making future approval more contested. It avoids the economic shock of a blanket moratorium but leaves critics worried about projects advancing during the transition.

The framework’s prospective design creates another limitation. Many proposals focus on future applications rather than facilities already operating or moving through construction.

Virginia reportedly has more than 600 data centers operating or under development. Standards aimed only at new projects cannot quickly change the cumulative effects of that existing base.

The administration says it will continue raising expectations for operating facilities. Yet environmental permits, utility agreements, and local approvals can limit how quickly those expectations become enforceable requirements.

Noise regulation offers a concrete example. The framework directs agencies to accelerate standards, but the final thresholds and compliance dates remain unknown.

Backup generation creates a similar challenge. Diesel generators provide emergency power, which is essential for facilities promising continuous service.

Regulators must distinguish limited reliability testing from prolonged operation during grid emergencies or market stress. They must also measure cumulative emissions across clusters containing many generators.

The framework proposes restricting on-site natural gas generation while making clean power investments easier. That creates another tradeoff between local emissions and near-term grid capacity.

A facility using behind-the-meter generation produces some electricity outside normal utility delivery. Gas turbines can reduce pressure on grid connections, but they can also undermine state emissions goals.

Clean energy alternatives may require transmission access, storage, or new generation that cannot be built quickly. Rules must therefore align environmental standards with realistic interconnection schedules.

The political path is equally uncertain. Major provisions need General Assembly approval during the 2027 session.

Legislators may support transparency while disagreeing about thresholds, subsidies, local authority, or utility costs. Industry lobbying will focus on language that affects active projects and earlier commitments.

Courts may also shape the outcome. Microsoft’s potential challenge to direct transmission cost assignment shows that large customers will contest rules they consider unfair.

The executive order therefore begins a negotiation rather than settling it. Its lasting effect depends on legislation, agency rules, commission decisions, and local implementation.

This uncertainty does not make the order symbolic. Ending state-agency nondisclosure agreements and beginning agency work are tangible steps.

However, the administration’s broadest promises concern lower bills, stronger environmental protection, and greater local control. Those outcomes require institutions outside the governor’s direct command.

Readers should judge the framework by implemented rules, not its announced ambitions. The crucial question is whether Virginia converts its accountability language into binding decisions before another major wave of projects wins approval.

Three Signals Will Show Whether Virginia’s Approach Works

The next few months will reveal whether the plan changes approval economics or mainly changes the political language surrounding data centers.

The first signal is the content and timing of the local planning toolkit. It should define useful disclosures, community benefit terms, environmental questions, and negotiating practices.

A vague guide would leave smaller communities facing the same information disadvantage. Detailed templates could standardize expectations before developers enter local hearings.

The toolkit should also distinguish negotiable community benefits from basic legal obligations. A company should not receive approval merely for meeting requirements that already apply.

The second signal is the legislation prepared for the 2027 General Assembly session. The most important provisions concern local nondisclosure agreements, the 25-megawatt approval threshold, expedited permitting, and cost allocation.

Precise statutory language will determine which projects qualify, whether active applications are protected, and how local governments exercise their authority.

Broad exemptions could preserve the earlier approval system under new terminology. Clear coverage would turn the framework into a meaningful constraint on future campuses.

The legislative debate will also show whether Spanberger has assembled a durable coalition. Environmental groups, ratepayer advocates, labor organizations, local officials, utilities, and technology companies want different outcomes.

A successful package will not satisfy every group. It must still produce rules that agencies and localities can administer without years of unresolved litigation.

The third signal is the first substantive work from the AI task force. Readers should look for a membership list, public meeting process, deadlines, and defined recommendations.

An early focus on measurable state responsibilities would strengthen the initiative. Procurement standards, agency privacy rules, cybersecurity controls, and workforce data all fall within identifiable government functions.

A broad report describing AI’s benefits and risks would weaken the case that the task force adds anything new. Virginia needs specific decisions, responsible agencies, and implementation dates.

Electricity data will provide a longer-term test. PJM load forecasts, utility capital plans, and State Corporation Commission rulings will show whether policy changes reduce risks for ordinary customers.

Approval counts will matter too. A slower process is not automatically a successful one, and a continued construction boom is not automatically a failure.

The relevant measures are better disclosure, credible demand forecasts, enforceable developer commitments, lower exposure for ratepayers, and clear environmental limits.

Virginia’s choice reaches beyond one state. AI companies need physical capacity, and communities increasingly want control over where that capacity appears.

The state is now trying to prove those interests can coexist without returning to opaque deals or adopting an absolute construction ban.

For developers, the practical response is to prepare projects that survive public scrutiny before seeking approval. That means disclosing energy demand, water use, backup generation, local benefits, and long-term commitments early.

For local officials and residents, the priority is to track which parts of the framework are already binding and which remain proposals. Public pressure should focus on the decisions still open.

Will the Virginia data center rules make AI infrastructure more accountable without pushing investment elsewhere? The answer will emerge through the toolkit, the 2027 legislation, and the task force’s first concrete recommendations.