Virginia's Data-Center Demand Grew ~46% in Five Years — Utility Capex Is Rising Faster, but the Constraint Is Grid Build Time, Not Money

Summary

In Northern Virginia — home to the world's largest concentration of data centers — commercial electricity sales grew 45.8% from 2019 to 2024 (a 7.8%/yr compound rate), roughly 3x the growth rate of Texas (+15.7%) and Georgia (+16.4%) over the same window, and far above the low single digits typical of states with little AI-infrastructure buildout. PJM/Dominion's own peak-load data confirms this is real physical load, not just billed revenue: Dominion's Virginia zone summer peak rose 23% and winter peak rose 45% between 2019 and 2025. Utility capital spending IS scaling to match — Dominion Energy raised its 2026-2030 capital plan to $64.7 billion (up 29% from its prior $50.1B plan), and utilities nationwide are on pace for roughly $1.3 trillion in capex from 2026-2030 (about $240B in 2026 alone) — but multiple independent analyses agree the binding constraint is no longer capital, it is physical grid build time: new large loads in Northern Virginia can face multi-year queues for transmission upgrades regardless of how much money is committed, because interconnection studies, permitting, and construction of new lines and substations run on a multi-year clock that capital spending cannot compress.

How much has demand actually grown, and where

Using EIA Form 861 state-level commercial electricity sales (askamerica energy.eia_electricity_prices, filtered to sector='commercial') as the measurable proxy for data-center-heavy demand — commercial sales are the sector data centers are billed under, and are far more sensitive to hyperscale buildout than the residential/industrial totals:

Virginia's growth accelerates sharply after 2021 (2022-2024 alone: +14.8%), consistent with the AI-driven hyperscale buildout timeline. EIA's own Today in Energy analysis (Sept 2026) independently confirms this using PJM peak-demand data: Dominion's Virginia zone posted a 23% rise in summer peak load and a 45% rise in winter peak load between the 2019 and 2025/26 seasons, and PJM projects the Dominion zone will see the largest absolute peak-demand increase of any PJM zone through 2030, expanding at an average 5.4%/year, 'largely driven by a concentration of data centers.' The Belfer Center's April 2026 case study on Virginia and Texas similarly identifies these two states as the two leading data-center demand markets nationally, with Virginia's data-center share of total state electricity consumption estimated at 26% in 2023 and projected by EPRI to reach 41-59% by 2030. Nationally, US data center power demand is forecast to roughly double, from about 31 GW in 2025 to 66 GW by 2027 (Goldman Sachs), and Lawrence Berkeley National Laboratory projects data centers could reach 6.7-12.0% of total US electricity consumption by 2028, up from 4.4% in 2023.

Is utility capital spending keeping pace?

On dollars committed, spending is scaling up sharply and arguably ahead of demand growth in percentage terms. Dominion Energy — the utility serving Northern Virginia's data-center corridor — raised its 5-year (2026-2030) capital plan to $64.7 billion, a 29% increase over its prior $50.1B five-year plan, with $54.8B of that allocated to its Virginia utility and roughly $17B earmarked specifically for data-center-driven and clean-energy-transition grid work. Dominion reports having contracted nearly 48.5 GW of data-center capacity as of December 2025, with a further ~70 GW in new demand requests in its interconnection queue. Industry-wide, US utilities are projected to spend approximately $1.3 trillion in aggregate capex from 2026-2030 (S&P Global Market Intelligence, April 2026), including roughly $240 billion in 2026 alone, and utilities collectively planned $208 billion in grid capex for 2025 — spending levels multiple independent trade analyses describe as historically unprecedented and directly attributable to data-center/AI load.

However, capital committed and capacity actually delivered are two different things. The consistent finding across independent sources (Belfer Center, utility-sector trade press, Fitch Ratings) is that the bottleneck has shifted from money to physical build time: new large loads seeking interconnection in Northern Virginia can face waits measured in years even after a project clears queue review, because new transmission lines, substations, and generation additions take years to permit and construct regardless of capital availability — 'data centers demand power in months; grid upgrades take years.' PJM's own 2026 forecast work shows roughly 11 GW of expected 2026 Virginia-area data-center capacity at risk of delay from grid bottlenecks even with committed capital. Separately, the 'who pays' question is becoming its own constraint: Fitch Ratings shifted its North American utility-sector outlook from neutral to deteriorating in June 2026, citing political and regulatory resistance to the rate increases utilities need to recover this capex from ratepayers, and PJM's own congestion costs rose 64% in 2024. Virginia's regulators have responded with a new 'GS-5' large-load tariff requiring big data-center customers to prepay 85% of transmission capacity costs during ramp-up, shifting cost-recovery risk away from residential ratepayers — a sign that the capital IS being raised, but increasingly through customer-specific contracts rather than traditional socialized rate-basing.

Bottom line: in the leading AI-infrastructure region (Northern Virginia/Dominion), demand growth (~46% over five years, accelerating to ~15% in the last two alone) and utility capex growth (a 29% jump in a single five-year plan revision, to $64.7B) are both real and both large — capital is not obviously lagging demand in dollar terms. What is lagging is physical delivery: interconnection queues, transmission build time, and now political tolerance for the rate increases that pay for it all, are the actual constraints slowing how fast that capital turns into usable grid capacity.

Every query behind this report

1 warehouse call ran in this session, in order. Each is reproducible against the same snapshot.

SELECT state_abbr, "year", sector, SUM(sales_million_kwh) AS sales_mkwh
FROM energy.eia_electricity_prices
WHERE state_abbr IN ('VA','TX','GA','OH','AZ') AND sector = 'commercial'
GROUP BY state_abbr, "year", sector
ORDER BY state_abbr, "year"