Why the UK's AI Boom Is Hitting an Energy Wall

The UK hosts one of Europe's largest data centre markets, with facilities clustered around London, and emerging hubs in Manchester, Leeds, Wales, and Scotland. The UK’s biggest data centre today runs at 120 MW. The AI centres now being planned will start at 500 MW and could reach a gigawatt: enough to power a city the size of Liverpool. The UK is the world's third-largest AI market, but before it can train the models, it must first find the electricity, and that search is proving far harder than anyone expected.

The UK data centre sector is entering a period of structural transformation driven by AI compute intensity. In 2023, UK data centres consumed approximately 5 TWh of electricity, representing roughly 2% of total national demand. According to analysis by Oxford Economics, electricity consumption from UK data centres is forecast to rise at least fivefold to more than 26 TWh by 2030. The scale of this growth highlights how the energy demand of data centres is evolving from a marginal load category to a fundamental component of the industrial power system.

Crucially, AI workloads differ from traditional enterprise computing. Training and inference tasks are significantly more power-dense. Whereas pre-AI facilities typically experienced steady 10–15% daily demand variation, AI-optimised sites can see 50% swings within minutes.

In order to establish digital infrastructure as a key national capability, the UK government has set a target to expand sovereign compute capacity by at least 20-fold by 2030. Over 450 operating locations and more than 100 additional data centre projects are presently under development, indicating a nearly 20% increase in installed capacity.

However, the UK electricity demand connection queue has expanded at a pace that now constitutes a structural barrier to the expansion of digital infrastructure. According to data published by Ofgem, demand connection requests surged from 41 GW in November 2024 to 125 GW by June 2025, more than double the UK’s current peak demand. 

A substantial proportion of this queue is attributable to data centre demand: an estimated 75 GW of that queue is data centre capacity. Around London alone, outstanding connection requests total roughly 400 GW, though Ofgem estimates 60–70% will never be built due to speculative applications, land banking strategies and developers reserving capacity before receiving confirmed funding or planning approval.

A timing mismatch has emerged between grid delivery and data centre build cycles. While facilities can be constructed within 18-24 months, grid connections take 3 to 8 years. These delays significantly impair project economics.

In response, Ofgem and NESO announced major connection reform in December 2025. The structure eliminates speculative proposals and prioritises shovel-ready projects instead of "first-come, first-served." The government claims the change could release over 400 GW from the queue and unlock up to £40 billion per year.

Beyond grid access, the UK’s electricity prices make it one of the most expensive places in the developed world to run a data centre. The UK currently records the highest industrial electricity prices among IEA member states, with costs more than four times those in the US, Finland, Norway, and Sweden. For AI training facilities, where energy represents a dominant operating cost, this makes the UK less competitive. Large-scale AI clusters can be situated wherever connection and policy circumstances are favourable. Oxford Economics warns that hyperscale investors may redirect projects to countries with lower costs including the Nordics and the US.

The government’s response centres on AI Growth Zones, which are locations that provide simplified planning, 500 MW or more of electricity, and up to £5 million per zone to encourage local adoption. The first zone, announced in January 2025 at Culham in Oxfordshire, will rise from 100 MW to 500 MW. Further zones are expected across de-industrialised regions with available grid capacity and local government support.

The UK is not alone in facing this tension. Globally, data centres consumed approximately 460 TWh of electricity in 2022, about 2% of total demand with the International Energy Agency projecting growth to 650-1050 TWh by 2026, driven primarily by AI. In the EU, usage reached roughly 70 TWh in 2024 and could rise to 115 TWh by 2030.