Transparency data

AI Energy Council minutes: Monday 30 June 2025 (HTML)

Updated 11 September 2025

Attendees

Ministerial attendance

• Ed Miliband MP, Secretary of State for Energy Security and Net Zero
• Peter Kyle MP, Secretary of State for Science, Innovation and Technology

External attendees

• Jonathan Brearley, CEO, Ofgem
• Fintan Slye, CEO, National Energy System Operator (NESO)
• Alice Delahunty, President, National Grid Transmission UK
• Lawrence Slade, CEO, Energy Networks Association (ENA)
• Josh Buckland, Director of Strategy and Policy, EDF Energy
• Tom Williamson, Head of Innovation, Scottish Power
• Tom Greatrex, CEO, Nuclear Industry Association (NIA)
• Maud Texier, Director EMEA Energy, Google
• Alison Kay Vice President UK and Ireland, Amazon Web Services (AWS)
• Christoph Mazur, Director - Energy and Sustainability, Microsoft
• Peter Stephens, Government Partnerships, ARM
• James Tyler, Managing Director, UK Operations, Equinix
• Suraj Bramhavar, Director, Advanced Research and Innovation Agency (ARIA) 
• Thomas Spencer, Analyst, International Energy Agency
• Sanjeet Sanghera, Head of Strategic Futures, NESO

Officials

(Officials from the secretariat and Private Office were also present but have not been included).

• Matt Clifford, Prime Minister’s Adviser on AI Opportunities
• Jonathan Mills, Director General of Energy Markets and Supply, DESNZ
• Henry Shennan, Deputy Director, Strategy Development, DESNZ
• Tim Bliss, DESNZ Special Adviser
• Sam Cannicott, Deputy Director of AI Infrastructure, International and Sovereign AI, DSIT
• Nicola Bartlett, DSIT Special Adviser (SpAds)

Summary

• Participants broadly agreed that inference will dominate AI-related energy demand.  While there was broad agreement that some AI compute will shift from data centres to the edge, there were differing views on the timescale and extent of that shift.  

• The group were supportive of a Demand Forecasting Working group, specifically focused on the energy demand of AI inference and using scenario-based assumptions to improve existing demand models. 

• Understanding location, latency, and price trade-offs is key, with the understanding that tech companies and investors are looking internationally, e.g. north of Scotland is competing with international locations, not just rest of UK.  

• There was consensus on the need for a credible, coordinated approach to forecasting and investment planning. 

• The group also supported a sprint looking at grid connections and regional coordination.

Item 1: Presentations 

IEA

• The IEA shared analysis on global AI energy demand, highlighting that model inference – not training – will be the dominant driver of energy use going forward. 

• Data centres are the key vector for this growth, with electricity demand projected to double by 2030 and nearly triple by 2035, becoming equivalent to the total electricity demand of Japan. 

• The rise of AI-optimised accelerated servers is a critical factor, these being significantly more power-intensive than conventional ones.  

• It was noted that electricity system integration, especially managing local grid impacts due to data centre clustering, will be as important as understanding total demand.

NESO 

• NESO shared insights from its Future Energy Scenarios, emphasising that data centre demand could triple by 2030, representing around 7% of domestic demand. 

• A heatmap of grid capacity showed significant regional imbalances, with the north of Scotland more favourable to new grid connections and the south-east of England heavily congested. 
• NESO noted the likely shift from training to inference workloads over the decade, and noted that the FES scenarios don’t focus on location and grid connection dynamics and that these would be more fully explored in the upcoming Strategic Spatial Energy Plan (SSEP) and then, in even more granular detail, in the Centralised Strategic Network Plan (CSNP).  
• Building a shared industry view on grid constraints, latency trade-offs, and siting challenges was identified as a critical next step.

Item 2: Discussion 

• Attendees discussed the balance between training and inference demand, noting that inference is expected to dominate AI energy use and typically requires proximity to end-users due to latency sensitivity.
• It was noted that high wholesale electricity prices, long lead times for grid connections, and energy pricing are challenges to data centre investment across the UK.
• Latency requirements were raised as a constraint on geographic flexibility for inference workloads.
• Temporary on-site generation, including natural gas fuel cells, was raised as an interim measure to meet power needs during grid connection delays. Limitations of batteries for backup were noted, particularly in relation to reliability and duration.
• Pre-2030 demand is expected to rely heavily on reprioritising existing connections. Beyond 2035, infrastructure will need to be planned and delivered through more structured mechanisms, such as spatial planning frameworks.
• The group supported the need for scenario-based forecasting to improve clarity of current projections and inform decision-making.
• Opportunities to develop data centres in less congested regions, emphasising the need for improved planning processes, streamlined grid connections, and reduced connection costs. The discussion also highlighted broader infrastructure challenges, including the impact of lengthy water connection timelines across the UK.
• Coordination between government, energy system operators and the technology sector was identified as necessary to align forecasting, investment planning, and infrastructure delivery.

Next steps

• The group agreed to contribute to work exploring how to unlock grid access in low-congestion areas, including assessing the broader offer in regions such as Scotland – particularly through the AI Growth Zones.
• The conversation emphasised the importance of refining demand forecasts using scenario-based methods and testing assumptions, especially around inference and latency and location.
• The group agreed that improved collaboration between the government, energy sector, and tech firms is key to developing workable solutions.
• There was shared recognition of the urgency to act within the next three years to align energy infrastructure with AI growth and avoid future mismatches.

Item 3: Closing remarks

• Chairs closed the meeting by highlighting a strong collective will to deliver solutions. The AI Energy Council was acknowledged as a crucial platform for unlocking investment, guiding infrastructure decisions, and ensuring the UK can lead in the AI era.