Methodology and sources

What each sub-index measures

AI Electricity Demand Index

Electricity consumed by UK data-centre and AI activity: annual consumption, share of national demand, AI-specific consumption where it can be isolated, quarterly and annual growth, published 2030 and 2035 forecasts, and estimated peak-power requirements.

Data Centre Growth Index

Physical infrastructure: operational facilities, facilities under construction, approved and proposed developments, new capacity in megawatts, regional concentration, hyperscale developments, and planning approvals and refusals.

Grid Pressure Index

Network capability: connection requests, average connection waiting time, projects delayed by insufficient capacity, available regional capacity, substation and transmission constraints, planned network investment, flexible connection agreements, and the risk of local congestion.

How data is classified

Every value carries one of the following classifications, shown next to the figure itself:

• Official figure — published by the originating organisation and checked against the primary document.
• Calculated from official figures — derived from official figures. The inputs, units, sources and formula are retained with the record, and the figure cannot be published without them.
• Industry estimate — an estimate published by a third party, reproduced with its source.
• AI Energy Intelligence estimate — our own estimate, with the method, assumptions and limitations stated.
• Forecast — a projection of a future period, kept separate from observations.
• Not disclosed — the figure exists but has not been made public.
• Insufficient evidence — the available evidence does not support a defensible figure or rating.
• Data not yet available — no figure has been entered yet. The cell stays empty rather than being filled, and zero is never used in its place.

Estimates and forecasts

Forecasts are never mixed into current observations. On the AI Electricity Demand Index they appear in a separate table, and charts plot only recorded readings for periods that have already occurred. No sub-index carries a headline score until enough sourced readings exist to support one; until then each reads “Baseline in development”.

Preventing duplicate developments

Each development is held once, as a unique record with its own project name, operator, location, status, capacity, planning authority and planning reference. The database enforces a single record per planning authority and reference, so a project reported by several outlets cannot be counted twice. Where a project appears under different names, the earlier record is amended rather than a second one created.

Regional definitions

Both the growth and grid-pressure sub-indices use the same nine regions. Slough and the Thames Valley is separated from the wider South East because of its concentration of existing capacity.

• London
• Slough and the Thames Valley
• South East England
• Midlands
• North West England
• North East and Yorkshire
• Scotland
• Wales
• Northern Ireland

How grid-pressure ratings are assigned

A region receives one of five ratings: Low pressure, Moderate pressure, High pressure, Severe pressure, Insufficient evidence. A rating is only recorded when there is a written explanation and a stated evidence-confidence level. A rating of low, moderate, high or severe cannot be published unless at least two published evidence records are attached to the assessment; the database rejects the attempt and the region stays at “Insufficient evidence”. A rating is never assigned automatically from a single data point. Ratings reflect recorded evidence about connection demand, delays, constraints and planned investment — not opinion, and not the number of projects alone.

Review frequency and corrections

Every indicator carries its own update schedule and a “last reviewed” date, and the index as a whole is reviewed quarterly. Corrections supersede rather than erase: the previous record is retained and the change is logged. To request a correction, use our corrections policy or contact us with the indicator, the figure in question and the source you believe is correct.

Limitations of UK data

There is no single official register of UK data centres, their capacity or their electricity consumption. Operators are not required to disclose site-level load, planning records are held by individual authorities in inconsistent formats, and connection-queue data is published at network rather than facility level. AI-specific consumption is rarely separable from general compute. The index reflects those gaps openly instead of closing them with estimates.

Geographic definitions

Every figure records the geography it covers, because UK, Great Britain and single-nation statistics are not interchangeable. The available coverages are: United Kingdom, Great Britain, England, Scotland, Wales, Northern Ireland, UK region, Local authority, Individual facility, Other. Northern Ireland is assessed separately from Great Britain throughout the Grid Pressure Index.

Confidence ratings

Each figure and each regional assessment carries a confidence rating of High, Medium, Low, Not assessed. Confidence describes the strength of the underlying evidence, not the precision of the number.

Counting projects, campuses and capacity

One record represents one facility or development. Where several buildings share a campus, each building is recorded separately and the campus name links them, so campus capacity is never added to the individual building figures. Totals are reported by development status and are never combined: proposed developments are not added to operational facilities.

Capacity figures always state what they describe: IT load, Total facility load, Grid-connection capacity, Campus capacity, Capacity definition unclear, Not disclosed. IT load, total facility load and grid-connection capacity measure different things and are never treated as equivalent or summed together.

Forecasting rules

Forecasts are recorded against the year they describe, are labelled as forecasts wherever they appear, and are shown in their own table rather than alongside recorded consumption. A chart is only drawn when at least two genuine reporting periods have been entered and approved. We do not extrapolate a trend from a single observation.

Update process

• Sources are checked monthly; the whole index is reviewed quarterly.
• Every record stores the date it was last checked and the date it is next due.
• A replacement value never overwrites history: the earlier reporting period is kept, the superseded figure is marked as such, and the change is logged with who made it, when, why and against which source.
• Superseded sources are retained and marked superseded rather than deleted.

Source categories monitored

The index draws on publications from the organisations below. Listing an organisation does not imply a figure has been taken from it: a source only appears in the register once its document title, URL and publication date have been entered and verified.

• Department for Energy Security and Net Zero
• Department for Science, Innovation and Technology
• National Energy System Operator
• Ofgem
• National Grid Electricity Transmission
• UK Power Networks
• National Grid Electricity Distribution
• Scottish and Southern Electricity Networks
• Electricity North West
• Northern Powergrid
• SP Energy Networks
• SONI
• NIE Networks
• Utility Regulator Northern Ireland
• Local-authority planning portals
• Planning Data service
• House of Commons Library

Change history

• indicator — value

  The official DESNZ share for 2024 replaces an internal calculation whose reporting periods did not match.

  Previously: 1.9 per cent (internally calculated) · Now: Superseded by DESNZ 2024 official share of 1.8 per cent

  15 Aug 2026 · AI Energy Intelligence editorial

• subindex — status_label

  Published the first evidence-based baseline for the AI Electricity Demand Index using the DESNZ Energy Trends June 2026 special feature article and its accompanying workbook.

  Previously: Baseline in development · Now: 2024 baseline

  15 Aug 2026 · AI Energy Intelligence editorial

• indicator — value

  The NESO Clean Power 2030 modelling input is no longer presented as the principal baseline. It is retained as historical context with its original classification.

  Previously: 5 TWh per year (NESO modelling input) · Now: Superseded by DESNZ 2024 estimate of 4.5 TWh

  15 Aug 2026 · AI Energy Intelligence editorial

• dc_project — record

  New record added following primary-source verification on the London Borough of Hillingdon planning register: application 38421/APP/2021/4045, status Approval, decided 26 April 2022.

  Previously: — · Now: published

  12 Aug 2026 · Editorial team

• dc_project — record

  New record added on the basis of Environment Agency permit EPR/PP3225SX, issued 13 September 2023 and recorded as effective on the public register.

  Previously: — · Now: published

  12 Aug 2026 · Editorial team

• dc_project — record

  New record added on the basis of Environment Agency permit EPR/TP3530DV and its decision document, issued 27 February 2018. The 214MWth figure is generator thermal input and is excluded from capacity reporting.

  Previously: — · Now: published

  12 Aug 2026 · Editorial team

• dc_project — record

  New record added on the basis of Environment Agency permit EPR/QP3533ED and its decision document for Carlton Park Data Centre 1 & 2, operated by Santander UK Plc.

  Previously: — · Now: published

  12 Aug 2026 · Editorial team

• dc_project — record

  New record added following primary-source verification on the London Borough of Hillingdon planning register: hybrid application 78343/APP/2025/719, status Approval, decided 29 October 2025.

  Previously: — · Now: published

  12 Aug 2026 · Editorial team

• dc_project — record

  New record added following primary-source verification on the Cardiff Council planning register: application 24/00624/FUL, permission granted, decision issued 15 April 2025.

  Previously: — · Now: published

  12 Aug 2026 · Editorial team

• dc_project — record

  New record added following primary-source verification on the Salford City Council planning register: application PA/2024/1810, approved with conditions, decision date 23 July 2025.

  Previously: — · Now: published

  12 Aug 2026 · Editorial team

• dc_project — status_publication

  Primary-source verification completed on the South Lanarkshire Council planning register. Recorded as proposed: P/25/0490 is an EIA screening opinion (EIA required, issued 25 June 2025), not a planning permission for the campus.

  Previously: draft · Now: published

  12 Aug 2026 · Editorial team

• dc_project — status_publication

  Primary-source verification completed on the Northumberland County Council planning register: outline application 24/04112/OUTES permitted, decision issued 8 May 2025, for up to ten data centre buildings totalling 540,000 sqm GIA.

  Previously: draft · Now: published

  12 Aug 2026 · Editorial team

• dc_project — status_publication

  Primary-source verification completed on the Northern Ireland Planning Portal: reserved matters application LA11/2023/1729/RM determined and granted on 9 October 2024. Application reference, decision, decision date, applicant and site address corrected against the statutory register.

  Previously: draft · Now: published

  12 Aug 2026 · Editorial team

• dc_project — duplicate_reviewed

  Assessed as the same scheme as the Atlantic Hub record at Maydown Road, Derry. No separate planning application exists under the GreenScale name. Held unpublished to prevent double-counting.

  Previously: false · Now: true

  12 Aug 2026 · Editorial team

• dc_project — record

  New project record added as a private draft; evidence insufficient for publication.

  Previously: — · Now: Northtree conversion, Hemel Hempstead - approved (draft)

  12 Aug 2026 · Editorial team · Methodology v1.0

• dc_project — record

  New project record added as a private draft; evidence insufficient for publication.

  Previously: — · Now: Ratcliffe-on-Soar power station site - proposed (draft)

  12 Aug 2026 · Editorial team · Methodology v1.0

• dc_project — record

  New project record added as a private draft; evidence insufficient for publication.

  Previously: — · Now: Kao Data Kenwood Point, Stockport - approved (draft)

  12 Aug 2026 · Editorial team · Methodology v1.0

• dc_project — record

  New project record added as a private draft; evidence insufficient for publication.

  Previously: — · Now: Cobalt Data Centre Campus (Stellium) - approved (draft)

  12 Aug 2026 · Editorial team · Methodology v1.0

• dc_project — record

  New project record added as a private draft; evidence insufficient for publication.

  Previously: — · Now: Cambois data centre campus - proposed (draft)

  12 Aug 2026 · Editorial team · Methodology v1.0

• dc_project — record

  New project record added as a private draft; evidence insufficient for publication.

  Previously: — · Now: Skelton Grange, Leeds - proposed (draft)

  12 Aug 2026 · Editorial team · Methodology v1.0

• dc_project — record

  New project record added as a private draft; evidence insufficient for publication.

  Previously: — · Now: Deep Green Bradford, Listerhills Road - approved (draft)

  12 Aug 2026 · Editorial team · Methodology v1.0

• dc_project — record

  New project record added as a private draft; evidence insufficient for publication.

  Previously: — · Now: Teledata Wythenshawe - approved (draft)

  12 Aug 2026 · Editorial team · Methodology v1.0

• dc_project — record

  New project record added as a private draft; evidence insufficient for publication.

  Previously: — · Now: Next Generation Data Newport campus - operational (draft)

  12 Aug 2026 · Editorial team · Methodology v1.0

• dc_project — record

  New project record added as a private draft; evidence insufficient for publication.

  Previously: — · Now: Haspielaw Farm, Hamilton - proposed (draft)

  12 Aug 2026 · Editorial team · Methodology v1.0

• dc_project — record

  New project record added as a private draft; evidence insufficient for publication.

  Previously: — · Now: GreenScale, Foyle Port - proposed (draft)

  12 Aug 2026 · Editorial team · Methodology v1.0

• dc_project — record

  New project record added as a private draft; evidence insufficient for publication.

  Previously: — · Now: Atlantic Hub, Foyle Port - approved (draft)

  12 Aug 2026 · Editorial team · Methodology v1.0

• dc_project — record

  New project record added as a private draft; evidence insufficient for publication.

  Previously: — · Now: AtlasEdge Colombus Way, Salford - approved (draft)

  12 Aug 2026 · Editorial team · Methodology v1.0

• dc_project — record

  New project record added and published; primary source from the planning or permitting authority verified.

  Previously: — · Now: Digital Realty Crawley - operational (published)

  12 Aug 2026 · Editorial team · Methodology v1.0

• dc_project — record

  New project record added and published; primary source from the planning or permitting authority verified.

  Previously: — · Now: Global Switch London - operational (published)

  12 Aug 2026 · Editorial team · Methodology v1.0

• dc_project — record

  New project record added and published; primary source from the planning or permitting authority verified.

  Previously: — · Now: Equinix LD8 - operational (published)

  12 Aug 2026 · Editorial team · Methodology v1.0

• dc_project — record

  New project record added and published; primary source from the planning or permitting authority verified.

  Previously: — · Now: VIRTUS Stockley Park campus - operational (published)

  12 Aug 2026 · Editorial team · Methodology v1.0

• dc_project — record

  New project record added and published; primary source from the planning or permitting authority verified.

  Previously: — · Now: Manor Farm and land north of Wraysbury Reservoir - approved (published)

  12 Aug 2026 · Editorial team · Methodology v1.0

• dc_project — record

  New project record added and published; primary source from the planning or permitting authority verified.

  Previously: — · Now: DC01UK, South Mimms - approved (published)

  12 Aug 2026 · Editorial team · Methodology v1.0

• dc_project — record

  New project record added and published; primary source from the planning or permitting authority verified.

  Previously: — · Now: Vantage Bridgend campus (former Ford engine plant) - proposed (published)

  12 Aug 2026 · Editorial team · Methodology v1.0

• dc_project — record

  New project record added and published; primary source from the planning or permitting authority verified.

  Previously: — · Now: 1 Redheughs Avenue data centre, South Gyle - refused (published)

  12 Aug 2026 · Editorial team · Methodology v1.0

• dc_project — record

  New project record added as a private draft; evidence insufficient for publication.

  Previously: — · Now: Equinix Wexham Road (former AkzoNobel site) - approved (draft)

  12 Aug 2026 · Editorial team · Methodology v1.0

• dc_project — record

  New project record added as a private draft; evidence insufficient for publication.

  Previously: — · Now: Data Centre Campus, Ampthill Road, Bedford - proposed (draft)

  12 Aug 2026 · Editorial team · Methodology v1.0

• dc_project — record

  New project record added as a private draft; evidence insufficient for publication.

  Previously: — · Now: Kao Data Harlow, Plot F (KLON-07 and KLON-08) - proposed (draft)

  12 Aug 2026 · Editorial team · Methodology v1.0

• dc_project — record

  New project record added as a private draft; evidence insufficient for publication.

  Previously: — · Now: Premier Park data centre, Park Royal - approved (draft)

  12 Aug 2026 · Editorial team · Methodology v1.0

• dc_project — record

  New project record added as a private draft; evidence insufficient for publication.

  Previously: — · Now: Amazon Web Services, Plot 3 Maylands Avenue - proposed (draft)

  12 Aug 2026 · Editorial team · Methodology v1.0

• datapoint — gb-colocation-data-centre-capacity.publication_date

  Phase 1 audit: the DSIT research and analysis page was published on 1 May 2025; 11 March 2025 is the date the underlying Parliamentary Question answer was first published. Both dates are now recorded.

  Previously: 2025-03-11 · Now: 2025-05-01

  12 Aug 2026 · AI Energy Intelligence editorial · Methodology 1.0

• datapoint — Grid Pressure Index

  Phase 4 expansion from the NESO Grid Code Development Forum pack.

  Previously: No record · Now: Added total contracted connection demand in the GB queue: 125 GW (November 2025), recorded as an all-technology figure.

  12 Aug 2026 · AI Energy Intelligence editorial · Methodology 1.0

• datapoint — Data Centre Growth Index

  Phase 3 expansion from the DSIT regional capacity table.

  Previously: No record · Now: Added colocation data-centre IT capacity for eleven Great Britain regions (autumn 2024), totalling 1.6 GW.

  12 Aug 2026 · AI Energy Intelligence editorial · Methodology 1.0

• datapoint — Electricity Demand Index

  Phase 2 expansion from NESO primary documents opened and read in full.

  Previously: No record · Now: Added: GB data-centre demand 7.6 TWh (FES 2025); connected capacity 2.4 GW; GB total demand 263 TWh (2023); GB peak demand 58 GW (2023); GB total demand 287 TWh (2030 forecast); calculated data-centre share 1.9%; 2035 demand 30-41 TWh; 2050 demand 51-71 TWh plus the earlier 62 TWh figure.

  12 Aug 2026 · AI Energy Intelligence editorial · Methodology 1.0

• indicator — ai-attributable-electricity-demand.caveats

  Phase 1 audit: data-gap note required by the methodology.

  Previously: No caveat recorded · Now: Recorded why this indicator remains "Insufficient evidence": no reviewed UK source separates AI workloads from wider data-centre consumption.

  12 Aug 2026 · AI Energy Intelligence editorial · Methodology 1.0

• datapoint — uk-ai-capable-capacity-requirement-2030.value_text

  Phase 1 audit: the UK Compute Roadmap says "at least 6GW". The published wording was more precise than the source.

  Previously: 6 GW · Now: At least 6 GW

  12 Aug 2026 · AI Energy Intelligence editorial · Methodology 1.0

• datapoint — gb-data-centre-electricity-consumption-2030.limitations

  Phase 1 audit: FES 2025 replaces the Clean Power 2030 demand trajectory. The historical forecast is retained.

  Previously: No limitation recorded · Now: Flagged as superseded for forecasting purposes by FES 2025 (30-41 TWh by 2035).

  12 Aug 2026 · AI Energy Intelligence editorial · Methodology 1.0

• datapoint — gb-data-centre-electricity-consumption.limitations

  Phase 1 audit: a newer NESO estimate exists. The historical record is retained, not overwritten.

  Previously: No limitation recorded · Now: Flagged as superseded by NESO FES 2025 (7.6 TWh); retained as the published November 2024 position.

  12 Aug 2026 · AI Energy Intelligence editorial · Methodology 1.0

• datapoint — value

  First evidence-based baseline entry, verified against the original source document and approved for publication.

  Previously: — · Now: 22

  11 Aug 2026 · Editorial review

• datapoint — value

  First evidence-based baseline entry, verified against the original source document and approved for publication.

  Previously: — · Now: 6

  11 Aug 2026 · Editorial review

• datapoint — value

  First evidence-based baseline entry, verified against the original source document and approved for publication.

  Previously: — · Now: 1.6

  11 Aug 2026 · Editorial review

• datapoint — value

  First evidence-based baseline entry, verified against the original source document and approved for publication.

  Previously: — · Now: 5

  11 Aug 2026 · Editorial review

Submit a correction

If you believe a figure, rating or record here is wrong, tell us. Submissions enter an administrator review queue and never change published data automatically.