AI & Energy·Free
UK AI Electricity Demand Outlook 2026
Bottom-up model of AI-driven electricity growth across UK regions to 2030, with sensitivity to model efficiency and adoption.
Published 14 May 202668 pagesVersion v1.2
Executive summary
What this report covers
AI workloads are on track to add 14–22 TWh to UK annual electricity demand by 2030 under our central scenario, concentrated in a handful of regional clusters. Model efficiency gains offset roughly a third of raw compute growth, but do not change the trajectory.
Key findings
Headline conclusions
- AI adds an estimated 14–22 TWh/year of UK electricity demand by 2030 (central case).
- 80% of new load concentrates in six clusters: West London, Slough, Manchester, Cardiff, Edinburgh, Newport.
- Efficiency improvements save 6–9 TWh/year but do not reverse growth.
- Inference — not training — is the dominant long-run driver.
Who should read this
Intended audience
- Energy analysts and grid planners
- UK policymakers and regulators
- Data centre operators and hyperscaler strategy teams
- Institutional investors and infrastructure funds
Table of contents
Inside the 68-page report
- 011. Methodology and scenario definitions
- 022. UK AI compute baseline (2025)
- 033. Adoption curves by sector
- 044. Regional cluster analysis
- 055. Efficiency and Jevons effects
- 066. Grid impact and connection queue
- 077. Policy recommendations
- 08Appendix: Model assumptions and sensitivities
Preview
Sample pages
Frequently asked questions
What scenarios do you model?+
Low, central and high adoption cases, each with three efficiency trajectories, giving nine total pathways to 2030.
Where does the data come from?+
Ofgem, ESO, DSIT, DESNZ, IEA, and our own bottom-up model of UK AI compute deployments.
How often is this updated?+
Twice a year, aligned with ESO Future Energy Scenarios releases.
Download the full report
Enter your email to get the full PDF plus the weekly UK AI energy briefing.