One in five UK firms are moving AI workloads overseas as high electricity costs and grid constraints start to dictate where compute gets built.
New research shows 20% of British companies have already shifted AI operations out of the UK with power pricing emerging as a core barrier to scaling.
A further third say energy costs are actively limiting their ability to expand AI infrastructure highlighting how compute is becoming an energy problem as much as a technology one.
A survey of 700 global tech firms by CUDO Compute, shows almost half of British organisations say geopolitical risks are pushing them to keep AI domestic, yet 43% admit cost and performance still outweigh sovereignty when it comes to deployment decisions. In practice economics are winning.
The pressure is most acute for compute heavy businesses. Around 32% of AI first firms say they would consider relocating workloads due to power costs compared with 18% of larger enterprises showing how energy intensity is shaping decisions at the cutting edge of the sector.
Where workloads are going is telling. The US ranks as the most attractive destination for new AI capacity followed by India and Eastern Europe – all markets with either lower energy costs greater grid capacity or faster infrastructure buildout.
AI infrastructure is no longer just about chips and software it is about access to land power cooling and grid connections. When those inputs tighten, the cost of computing rises sharply and location becomes critical.
At the same time sovereignty concerns are rising. Around 45% of UK firms say regulation security and data control are shaping where they deploy AI while nearly a third are willing to prioritise domestic or regional compute even at higher cost.
But the gap between ambition and reality is widening, companies want to build in the UK, yet the underlying infrastructure is not keeping pace with demand.
Matt Hawkins chief executive of CUDO Compute said: “AI is not abstract software. It is physical infrastructure that depends on power land cooling and grid access. If it is cheaper or easier to run workloads elsewhere they will move regardless of sovereignty ambitions.”
The warning is there, AI policy cannot be separated from energy policy and without affordable power and grid capacity – the UK risks losing ground in the global race.
Copyright © 2026 Energy Live News LtdELN