AI Electricity Demand Index

This sub-index tracks the electricity required to support AI and data-centre activity in Great Britain: how much is consumed, what share of grid electricity that represents, how quickly it is changing, and what modelled estimates and forecasts say about the years ahead.

Headline baseline

2024 baseline

4.5 TWh

Estimated electricity consumed by data centres in Great Britain in 2024.

This covers data centres generally and is not an estimate of electricity used specifically by artificial intelligence.

Change, 2020 to 2024
41% increase
An increase of 1.3 TWh. A DESNZ historical estimate, not a real-time reading.
Share of GB grid electricity, 2024
1.8%
DESNZ reports this as approximately 2 per cent of the 249.2 TWh consumed from the grid in Great Britain.
Source and review
Department for Energy Security and Net Zero (DESNZ)
Published 30 Jun 2026 · Last reviewed 15 Aug 2026
Official figure

Historical observations

Annual figures published by DESNZ for periods that have already happened. Every reading below comes from the official workbook accompanying the Energy Trends June 2026 special feature article. Nothing here is modelled by us and no demonstration values are plotted.

Great Britain data-centre electricity consumption, 2020 to 2024, in TWh per year. Vertical axis starts at zero. Source: Department for Energy Security and Net Zero (DESNZ), published 30 Jun 2026. Full figures are given in the table below.
Great Britain data-centre electricity consumption by reporting year, in TWh per year. Source: Department for Energy Security and Net Zero (DESNZ).
Reporting yearConsumption (TWh)Change on previous year (TWh)Change on previous year (%)Change from 2020 (%)Share of GB grid (%)Classification
20203.2Baseline year1.2%Official figure
20213.4+0.29+9.1%+9.1%1.3%Official figure
20223.8+0.4+11.6%+21.7%1.5%Official figure
20234.1+0.28+7.4%+30.6%1.7%Official figure
20244.5+0.33+8.1%+41.2%1.8%Official figure

Values are shown to the precision DESNZ uses. Annual and cumulative changes are calculated from the unrounded workbook values, so they will not always match a subtraction of the rounded figures above.

Geographic concentration, 2024

Where data-centre electricity is consumed in Great Britain. These figures describe total data-centre consumption. They do not indicate where AI workloads run.

Indicators, current values, classification and sources
IndicatorCurrent valueReporting periodPreviousChangeClassificationSourceReviewed
Data-centre electricity consumption, LondonElectricity consumed from the grid by data centres in London (Inner and Outer London combined).Describes total data-centre consumption in the region, not the location of AI workloads.1.7 TWh per year2024calculatedDepartment for Energy Security and Net Zero (DESNZ)15 Aug 2026
Data-centre electricity consumption, South EastElectricity consumed from the grid by data centres in the South East of England.Describes total data-centre consumption in the region, not the location of AI workloads.1.8 TWh per year2024officialDepartment for Energy Security and Net Zero (DESNZ)15 Aug 2026
London and South East combined share of GB data-centre consumptionShare of all Great Britain data-centre electricity consumption accounted for by London and the South East together.A concentration measure for data centres generally, not AI workloads.77 Percentage2024calculatedDepartment for Energy Security and Net Zero (DESNZ)15 Aug 2026
Data-centre electricity consumption, SloughElectricity consumed from the grid by data centres in the Slough local authority area.A local-authority figure, shown separately from the regional and Great Britain totals. Not AI-specific.1.3 TWh per year2024officialDepartment for Energy Security and Net Zero (DESNZ)15 Aug 2026
Slough share of GB data-centre electricity consumptionShare of all Great Britain data-centre electricity consumption accounted for by Slough.A concentration measure for data centres generally, not AI workloads.29 Percentage2024calculatedDepartment for Energy Security and Net Zero (DESNZ)15 Aug 2026
Data-centre share of Slough grid electricity consumptionShare of all electricity consumed from the grid in Slough that is consumed by data centres.A local grid measure for Slough only. It says nothing about the share elsewhere in Great Britain and is not AI-specific.65 Percentage2024officialDepartment for Energy Security and Net Zero (DESNZ)15 Aug 2026

Latest modelled estimate

NESO’s modelled estimate as at November 2025 is shown separately from the DESNZ historical series. It is a modelled estimate, it covers data centres generally, it is not AI-specific, and its methodology differs from the DESNZ series. It is not the next comparable annual observation after 2024, so no growth rate is calculated between the two.

Indicators, current values, classification and sources
IndicatorCurrent valueReporting periodPreviousChangeClassificationSourceReviewed
GB data-centre electricity demand - NESO modelled estimateNESO current estimate of annual electricity demand from data centres connected in Great Britain.A modelled estimate covering data centres generally, not AI workloads. Its methodology differs from the DESNZ historical series, so it is not the next comparable annual observation after DESNZ 2024 and no growth rate is calculated between the two.7.6 TWh per yearNESO modelled estimate as at November 2025Industry estimateNational Energy System Operator (NESO)15 Aug 2026
GB connected data-centre capacityElectrical capacity of data-centre facilities already connected to the Great Britain electricity system.Connected capacity, not IT load and not contracted future capacity. NESO does not state the capacity definition.2.4 GWAs at November 2025Industry estimateNational Energy System Operator (NESO)12 Aug 2026

Forecasts

Projections published by third parties. These are not observations and are kept separate from the figures above.

Indicators, current values, classification and sources
IndicatorCurrent valueReporting periodPreviousChangeClassificationSourceReviewed
GB data-centre electricity consumption (2030 projection)Projected annual electricity consumption of all data centres in Great Britain in 2030 under NESO's Clean Power 2030 pathway.Scenario-dependent forecast derived from planning and grid connection registers, which typically overstate eventual build-out. Not AI-specific.22 TWh per year2030 (Clean Power 2030 pathway)ForecastNational Energy System Operator (NESO)12 Aug 2026
GB data-centre electricity demand, 2035FES 2025 pathway projections of Great Britain data-centre electricity demand in 2035.Scenario projections, not predictions.30 to 41 TWh (Ten Year Forecast 33; Hydrogen Evolution 30; Electric Engagement 41)By 2035 - FES 2025 pathwaysForecastNational Energy System Operator (NESO)12 Aug 2026
GB data-centre electricity demand, 2050Published long-range projections of Great Britain data-centre electricity demand in 2050.Scenario projections, not predictions.62 TWh per yearBy 2050 - Clean Power 2030 (November 2024)ForecastNational Energy System Operator (NESO)12 Aug 2026
GB total annual electricity demand, 2030NESO Clean Power 2030 projection of total Great Britain electricity demand in 2030.Forecast, not an observation.287 TWh per year2030 (Clean Power 2030 pathway)ForecastNational Energy System Operator (NESO)12 Aug 2026

Other tracked indicators

Context indicators and placeholders retained from earlier editions of this sub-index.

Indicators, current values, classification and sources
IndicatorCurrent valueReporting periodPreviousChangeClassificationSourceReviewed
GB total annual electricity demandTotal annual electricity demand across Great Britain, used as the denominator for share calculations.Great Britain, not the United Kingdom.263 TWh per year2023VerifiedNational Energy System Operator (NESO)12 Aug 2026
GB peak electricity demandPeak electricity demand recorded across Great Britain.A capacity measure in gigawatts; not comparable with annual consumption in TWh.58 GW2023VerifiedNational Energy System Operator (NESO)12 Aug 2026
GB data-centre electricity consumption (superseded NESO modelling input)Annual electricity consumption of all data centres in Great Britain, used as the present-day baseline in NESO's Clean Power 2030 analysis. This measures consumption, not capacity, and covers all data centres rather than AI workloads specifically.A modelling input rather than a measured statistic. NESO did not state a precise base year. Great Britain, not AI-specific. Superseded as the principal baseline by the DESNZ 2024 official estimate and retained here as historical context only.Data not yet availableIndustry estimateNational Energy System Operator (NESO)15 Aug 2026
Data-centre share of GB electricity demand (superseded internal calculation)Data-centre electricity demand expressed as a share of total Great Britain electricity demand.Calculated by AI Energy Intelligence from NESO figures with mismatched reporting periods. Superseded by the official DESNZ 2024 share and retained only as history.Data not yet availableAI Energy Intelligence estimateNational Energy System Operator (NESO)15 Aug 2026
Data centre energy efficiency (PUE)Reported power usage effectiveness for UK data centre estates — how much total power is drawn per unit of IT power.PUE is self-reported, defined inconsistently, and best-performing operators are the most likely to publish.Data not yet availableData not yet availableIEA — Energy and AI

AI-specific evidence gap

Reliable AI-specific electricity-consumption data remains limited. General data-centre electricity use must not automatically be treated as electricity consumed by artificial intelligence. No authoritative UK or Great Britain source reviewed to date separates electricity consumed by AI workloads from electricity consumed by other data-centre workloads. We do not estimate an AI share by applying a global percentage to Great Britain consumption, and we do not treat processor shipments, compute capacity, investment announcements or projected AI-capable capacity as a substitute for measured electricity consumption.

  • AI-attributable electricity demand

    Insufficient evidence

    No authoritative UK or Great Britain source reviewed to date separates electricity consumed by AI workloads from electricity consumed by other data-centre workloads.

  • Quarterly AI electricity-demand growth

    Insufficient evidence

    No authoritative UK or Great Britain source reviewed to date separates electricity consumed by AI workloads from electricity consumed by other data-centre workloads.

  • AI-specific peak-power requirement

    Insufficient evidence

    No authoritative UK or Great Britain source reviewed to date separates electricity consumed by AI workloads from electricity consumed by other data-centre workloads.

Superseded readings

Earlier figures are retained rather than deleted, with their original classification and the reason they were replaced.

  • Data-centre share of GB electricity demand (superseded internal calculation)1.9 Percentage

    2023 total demand against the Clean Power 2030 data-centre baseline

    Replaced by the official DESNZ 2024 share of 1.8 per cent. The earlier figure was calculated by AI Energy Intelligence by comparing the NESO Clean Power 2030 data-centre modelling input with 2023 total demand, so its reporting periods did not match.

    AI Energy Intelligence estimate

  • GB data-centre electricity consumption (superseded NESO modelling input)5 TWh per year

    Baseline used in Clean Power 2030 (published November 2024)

    Replaced as the principal baseline by the DESNZ 2024 official estimate of 4.5 TWh. Retained as historical context: NESO used 5 TWh a year as a modelling input in its Clean Power 2030 analysis, did not state a precise base year, covered Great Britain, and the figure was not AI-specific. It is not directly interchangeable with the DESNZ annual historical series.

    Industry estimate

Data gaps and limitations

  • Data-centre electricity consumption, Great Britain

    A DESNZ statistical estimate of electricity taken from the public grid by data centres. It covers data centres generally and does not separate electricity used by artificial-intelligence workloads. Great Britain only: Northern Ireland is excluded.

    Assumptions: Displayed to one decimal place, matching the rounding DESNZ uses. Changes are calculated from the unrounded workbook values.

    Confidence: High · Last reviewed: 15 Aug 2026

  • Data-centre share of Great Britain grid electricity consumption

    The share of grid electricity consumed by data centres generally. It is not a measure of electricity used by artificial-intelligence workloads. Great Britain only.

    Assumptions: Displayed to one decimal place from the workbook value. DESNZ rounds the 2024 figure to 2 per cent in the article text.

    Confidence: High · Last reviewed: 15 Aug 2026

  • Change in GB data-centre electricity consumption, 2020 to 2024 (absolute)

    A change between two DESNZ statistical estimates for data centres generally. Not AI-specific.

    Confidence: High · Last reviewed: 15 Aug 2026

  • Growth in GB data-centre electricity consumption, 2020 to 2024

    A historical estimate of change over four years. It is not a current or real-time growth rate and is not AI-specific.

    Confidence: High · Last reviewed: 15 Aug 2026

  • Data-centre electricity consumption, London

    Describes where data-centre electricity is consumed. It does not indicate where AI workloads run.

    Confidence: High · Last reviewed: 15 Aug 2026

  • Data-centre electricity consumption, South East

    Describes where data-centre electricity is consumed. It does not indicate where AI workloads run.

    Confidence: High · Last reviewed: 15 Aug 2026

  • London and South East combined share of GB data-centre consumption

    A concentration measure for data centres generally, not AI workloads.

    Confidence: High · Last reviewed: 15 Aug 2026

  • Data-centre electricity consumption, Slough

    A single local-authority area. Not comparable with regional or Great Britain totals without care, and not AI-specific.

    Confidence: High · Last reviewed: 15 Aug 2026

  • Slough share of GB data-centre electricity consumption

    A concentration measure for data centres generally, not AI workloads.

    Confidence: High · Last reviewed: 15 Aug 2026

  • AI-attributable electricity demand

    No authoritative UK or Great Britain source reviewed to date separates electricity consumed by AI workloads from electricity consumed by other data-centre workloads.

    Confidence: Not assessed · Last reviewed: 15 Aug 2026

  • Data-centre share of Slough grid electricity consumption

    Applies to Slough only and says nothing about the share of grid electricity used by data centres elsewhere. Not AI-specific.

    Confidence: High · Last reviewed: 15 Aug 2026

  • GB data-centre electricity consumption (2030 projection)

    Forecast, not an observation. Great Britain only. NESO Clean Power 2030 pathway (November 2024). Superseded for current forecasting by FES 2025, which models 30-41 TWh by 2035; retained here as the published November 2024 position.

    Assumptions: Derived by NESO from registers of planning and grid connection applications, under the Clean Power 2030 pathway based on the Holistic Transition pathway from FES 2024.

    Confidence: Medium · Last reviewed: 12 Aug 2026

  • GB connected data-centre capacity

    Definition unclear: NESO does not state whether this is IT load, total facility load or grid-connection capacity. Excluded from any aggregate capacity total on this site. Great Britain only.

    Assumptions: NESO describes these as connected facilities; the capacity definition is not stated in the source.

    Confidence: Medium · Last reviewed: 12 Aug 2026

  • GB total annual electricity demand

    Great Britain only; excludes Northern Ireland.

    Confidence: Medium · Last reviewed: 12 Aug 2026

  • Quarterly AI electricity-demand growth

    No authoritative UK or Great Britain source reviewed to date separates electricity consumed by AI workloads from electricity consumed by other data-centre workloads.

    Confidence: Not assessed · Last reviewed: 15 Aug 2026

  • AI-specific peak-power requirement

    No authoritative UK or Great Britain source reviewed to date separates electricity consumed by AI workloads from electricity consumed by other data-centre workloads.

    Confidence: Not assessed · Last reviewed: 15 Aug 2026

  • GB peak electricity demand

    Great Britain only. Peak demand in gigawatts is a capacity measure and must not be compared with annual consumption in TWh.

    Confidence: Medium · Last reviewed: 12 Aug 2026

  • GB data-centre electricity demand, 2035

    Scenario projections, not predictions. Great Britain only. Reported individually because averaging pathways is not authorised by the methodology. Aggregate data-centre demand, not AI-specific.

    Assumptions: Each figure is a separate FES 2025 pathway. Pathways are not equally likely and no average is published.

    Confidence: Medium · Last reviewed: 12 Aug 2026

  • GB data-centre electricity demand, 2050

    Forecast, not an observation. Great Britain only. NESO states demand reaches up to 62 TWh in 2050 - an upper figure, not a central estimate. Retained as the published November 2024 position alongside the newer FES 2025 range.

    Confidence: Medium · Last reviewed: 12 Aug 2026

  • GB total annual electricity demand, 2030

    Forecast, not an observation. Great Britain only. Published November 2024.

    Assumptions: Assumes electrification proceeds at a pace consistent with the 2030 Nationally Determined Contribution and Climate Change Committee carbon budgets.

    Confidence: Medium · Last reviewed: 12 Aug 2026

  • GB data-centre electricity demand - NESO modelled estimate

    Modelled estimate, not metered outturn. Great Britain only. Aggregate data-centre demand - it is not AI-specific and must not be presented as AI electricity consumption. Supersedes the 5 TWh figure published in Clean Power 2030 (November 2024).

    Assumptions: NESO states this demand arises from 2.4 GW of connected facilities and is mainly for traditional services.

    Confidence: Medium · Last reviewed: 15 Aug 2026

  • GB data-centre electricity consumption (superseded NESO modelling input)

    A modelling input rather than a measured statistic. NESO did not state a precise base year. Great Britain, not AI-specific. Superseded as the principal baseline by the DESNZ 2024 official estimate and retained here as historical context only.

    Confidence: Not assessed · Last reviewed: 15 Aug 2026

  • Data-centre share of GB electricity demand (superseded internal calculation)

    Calculated by AI Energy Intelligence from NESO figures with mismatched reporting periods. Superseded by the official DESNZ 2024 share and retained only as history.

    Confidence: Not assessed · Last reviewed: 15 Aug 2026

  • Data centre energy efficiency (PUE)

    PUE is self-reported, defined inconsistently, and best-performing operators are the most likely to publish.

    Confidence: Not assessed · Last reviewed:

Continue through the index

Cite or share this index

AI Energy Intelligence, AI Electricity Demand Index, accessed 27 August 2026. https://aienergyintelligence.co.uk/uk-ai-energy-index/electricity-demand

Share on XShare on LinkedIn

Get UK AI Energy Index updates

Emails cover new index editions, material indicator changes, significant data-centre and grid developments, and updates to the methodology or source register.

We use your email to send your report and (if you've opted in) the weekly briefing. Unsubscribe any time. See our privacy policy.