Skip to content
DataCentersExposed
County data-center profile

Henrico County, Virginia

DataCentersExposed tracks 27 AI data centers in Henrico County, Virginia — 8 operating and 2 in the pipeline, drawing 210 MW of reported power demand, led by Meta Platforms, Inc.. Its transparent data-center risk score is 0/100.

Sources: Methodology, Coverage scorecard

Facilities
27
Operating
8
Pipeline
2

Reporting by · Updated May 29, 2026

Risk breakdown
19Low
  • Project exposure · 33
  • Power demand · 0
  • Water draw · n/a
  • Land footprint · n/a
Overview

What's happening in Henrico County

Facilities here with disclosed capacity draw 210 MW from the local grid. Data-center load on this scale is the single biggest driver of recent capacity-market price increases — costs that flow through to every ratepayer in the service territory, not just the operators.

Meta Platforms, Inc. has the largest presence in Henrico County. We resolve project codenames and shell LLCs back to the corporate parent wherever the public record allows, so the table below names who is actually behind each site.

Below: every tracked facility in the county, the active state legislation that names it or its county, and local news. Commissioner vote records arrive in a later release; the placeholders show where that accountability data will land.

Modeled scenario · not announced

2035 Buildout Outlook

via PNNL IM3 (CC BY 4.0)

Under the model's moderate-growth scenario (5%/yr annual load growth), PNNL's IM3 model sites about 1 standardized 36 MW campus in Henrico County by 2035 — roughly 36 MW of new electricity demand and ~35 M gal/yr of cooling water. For scale, we currently track 27 real data centers in Henrico County (210 MW of reported power).

Modeled new campuses by demand-growth scenario
  • Higher growth1
    36 MW · ~35 M gal/yr cooling water

The model assigns ~100% of these campuses' cooling to evaporative water cooling in Henrico County (the rest to mechanical/air cooling), a split it derives from local water stress and wet-bulb temperature.

These are modeled candidate sites, not announced or permitted facilities. PNNL's IM3 model places identical 36 MW unit-campuses at feasible locations under each scenario — it shows where demand could concentrate, never a specific parcel. Figures use a market-gravity weight of 50. How the model works.

27 facilities

Every tracked facility

FacilityOperatorStatusMW
Azalea
Richmond
BWS EnterprisesProposed10
Atlantic Crossing
Sandston
ATLANTIC CROSSING LLCProposed—
Sandston Data Center Campus
Sandston
—Operating200
Bank of America—Operating—
Cavalier Innsbrook Data Center
Glen Allen
HIGHWOODS PROPOperating—
Dc4
Sandston
—Mapped (unverified)—
Dc5
Sandston
—Mapped (unverified)—
Dc6
Sandston
—Mapped (unverified)—
Flexential - Richmond
Richmond
Flexential Corp.Operating—
Infotel Systems Data Center
Richmond
InfotelOperating—
Lumen Richmond 2
Henrico
Level3 CommunicationsOperating—
Meta Henrico Data CenterMeta Platforms, Inc.Mapped (unverified)—
Ric1 Dc1—Mapped (unverified)—
Ric1 Dc2—Mapped (unverified)—
Ric1 Dc3—Operating—
Ric2 Dc1
Sandston
—Mapped (unverified)—
Ric2 Dc2
Sandston
—Mapped (unverified)—
Ric2 Dc3
Sandston
—Mapped (unverified)—
Ric2 Dc5
Sandston
—Mapped (unverified)—
Ric2 Dc6
Sandston
—Mapped (unverified)—
Ric3 Dc1
Sandston
—Mapped (unverified)—
Ric3 Dc2
Sandston
—Mapped (unverified)—
Ric3 Dc3
Sandston
—Mapped (unverified)—
Ric3 Dc4
Sandston
—Mapped (unverified)—
Ric3 Dc5
Sandston
—Mapped (unverified)—
Richmond Data Center
Richmond
—Operating—
Wagner Urban Logistics LLC Data Center
Richmond
—Blocked—
Accountability

County commissioners

How the Henrico County board votes on each data-center matter — the rezonings, the abatements, the water permits — is built from public-meeting minutes and roll-call records. Know of a vote we should be tracking? Tell us.

Methodology & sources

The risk score weights project exposure (40%), power demand (30%), water draw (15%), and land footprint (15%). Water and land are not yet measured for our facilities, so those arcs render as "not yet measured" and the weights are renormalized over the factors we can source today — when the data lands, the score updates.

Every row carries a confidence level (high / medium / low) and a source URL. Spot an error? Tell us.