Data-centre electricity demand in Northern Virginia, shown in kettle-boils per second
The US Cloud Kettle Index translates Dominion Energy Virginia's published data-centre billing-demand figures into kettle-boils per second. It is framed for DC + Mid-Atlantic readers because Northern Virginia is the region's dominant cloud-infrastructure cluster.
Northern Virginia cloud-load proxy - 2026 projection
Dominion 4,753 MW billing demand equivalent ÷ 0.36 MW/kbs · projection
The published demand data is public. Measured load is not. This page uses the best public proxy sources. - Full methodology
2026 projection: Dominion Energy Virginia publishes 4,750 MW of data-centre billing demand in its Virginia service territory. At 0.36 MW per kettle-boil/sec, that is ~13,200 kbs.
Scope and average-load context: Dominion Energy Virginia service-territory data-centre billing demand. Framed as DC + Mid-Atlantic because Northern Virginia, which dominates Dominion’s data-centre load, is the cloud-infrastructure cluster most relevant to DC readers. Dominion reports an industry load factor of approximately 90%, implying roughly 4,280 MW average continuous draw (~11,900 kbs). The headline uses billing demand because that is the public series Dominion provides.
Source trail
The US page mixes several public sources. These cards show what each source does, and what it does not prove.
Dominion forecast
Year-by-year data-centre billing demand used for the headline and trajectory.
PJM deck
Dominion large-load request materials that corroborate the forecast series.
EIA PJM demand
Latest-reported PJM balancing-authority demand for grid-share context.
Census boundaries
County boundary geometry for the Northern Virginia locator map.
Dominion profiles
Generic large-customer settlement/load-research profiles, not data-centre profiles.
Service territory
Dominion and Virginia SCC service-area references; no open polygon is used here.
Billing demand, not metered load
4,750 MW is Dominion's 2026 projection data-centre billing demand - used for tariff calculations, not measured grid draw. Dominion separately derives coincident (metered) demand, which is generally lower. At ~90% load factor, average continuous draw is roughly 4,280 MW (~11,900 kbs).
What this is not
- -Data centres physically inside DC.
- -Total PJM data-centre load (13 states + DC).
- -A full Northeast model (PJM + NYISO + ISO-NE).
- -Live grid demand - modelled estimates only.
Published Dominion data-centre billing demand
Dominion's public figures show data-centre billing demand rising from 2024 actuals through the 2030 planning forecast. Select a year to inspect the conversion.
Published Dominion data-centre billing demand - 2024-2030
2026 projection: ~13,200 kettle-boils/sec
Selected year
2026 projection
Billing demand
4,753 MW
Average-load context
4,278 MW
View source table
| Year | Kind | Billing demand (MW) | Kettle-boils/sec |
|---|---|---|---|
| 2024 | actual | 3,584 | ~10,000 |
| 2025 | forecast | 4,149 | ~11,500 |
| 2026used above | projection | 4,753 | ~13,200 |
| 2027 | projection | 5,322 | ~14,800 |
| 2028 | projection | 5,863 | ~16,300 |
| 2029 | projection | 6,419 | ~17,800 |
| 2030 | projection | 6,992 | ~19,400 |
Select a year to inspect the conversion. The GB page can show a live 24-hour share because GB grid demand is public through Elexon. Here, the public Dominion series is a forecast trajectory, not a live data-centre meter.
Primary source: Dominion Energy Virginia 2025 data-centre forecasting presentation (Indiana IURC filing).
Also: PJM Dominion data-centre large load request (Load Analysis Subcommittee, September 2025).
Generic Dominion large-customer load shapes
Dominion publishes generic hourly profiles for customer classes. These are not data-centre profiles, but they are useful local context. Each line is scaled to its own average, so 1.0 means a normal hour for that class. Dominion has not published an equivalent hourly profile for data centres.
Median large-customer class: Median of GS3, GS4, LGEMLP, and MS normalized profiles.. This profile varies by 38.4% peak-to-trough on a weekday.
These profiles are settlement profiles, not measured data-centre profiles. COMM6VA is omitted because Dominion publishes it as a weather-sensitive formula rather than a simple hourly shape. Source.
What are these Dominion classes?
- GS3
- Large general-service customers connected at secondary voltage; a useful commercial/industrial comparator, but not data centres.
- GS4
- Large general-service customers connected at primary voltage; flatter than GS3 in this dataset and closer to very large continuous loads.
- LGEMLP
- Large miscellaneous light-and-power customers; broad, but still a large-load class in Dominion's published profiles.
- MS
- Military service profile; included as another large institutional load shape in Dominion's published profiles.
The median option takes the hour-by-hour median of GS3, GS4, LGEMLP, and MS after each class has been scaled to its own annual mean.
Latest-reported PJM grid context
Here the Dominion data-centre estimate stays fixed while total PJM demand moves through the day. It is regional context, not a Dominion-zone meter.
Grid context: Dominion proxy as share of PJM demand
5.9% of PJM hourly demand (4,753 MW ÷ ~80000 MW PJM)
Dominion 2026 projection data-centre billing demand is fixed here. PJM demand comes from EIA and changes hour by hour.
The line moves because PJM demand moves: usually higher on weekday afternoons, lower overnight and on weekends. The Dominion data-centre estimate is fixed, not a live meter. Cache generated Jun 1, 8:11 AM EDT.
Data request scoreboard
The estimate is only as good as the public record underneath it. These are the datasets that would make the DC + Mid-Atlantic index much stronger.
| Missing or partial data | Likely holder | Status | Next request |
|---|---|---|---|
| Measured Dominion-zone data-centre load | Dominion Energy Virginia / Virginia SCC filings | Missing | Ask for hourly or monthly data-centre load by tariff/customer class, aggregated enough to protect customers. |
| Dominion service-territory boundary as GIS | Dominion Energy Virginia / Virginia SCC | Missing | Ask for a public polygon or shapefile suitable for non-operational mapping. |
| Northern Virginia county-level data-centre load | Counties, PJM interconnection queues, utilities | Missing | Ask Loudoun, Fairfax, Prince William, and Arlington for planning assumptions and utility-impact evidence. |
| Measured regional data-centre load shapes | Dominion, PJM, operators, research partners | Missing | Ask for anonymised hourly profiles analogous to the UKPN data-centre profile release. |
| Latest PJM demand comparison | EIA Balancing Authority Areas hourly data | Wired | Let the scheduled EIA refresh populate the static cache; keep it labelled as reported data, not live data. |
Future expansion
A fuller Northeast version would need separate treatment for PJM, NYISO, and ISO-NE. A future PJM comparison could use Dominion-zone demand, if a reliable public series is available, instead of PJM total.
Great Britain page
The original Cloud Kettle Index: GB data-centre load based on NESO and DSIT figures, with live Elexon grid demand and UKPN load shapes.
Go to GB page →Request US records
Ask for better data using Virginia FOIA, federal FOIA, DC FOIA, or utility commission filings. Ready-to-send templates for DC-area data-centre electricity records.
US public records templates →Methodology
The formula, data sources, assumptions, and what this site does not claim.
Read the methodology →vetch
Open-source LLM inference energy measurement and circuit breakers for runaway cost and carbon.
GitHub →