One in five UK firms move AI workloads abroad over power costs

One in five British firms have moved AI workloads out of the UK because of high power costs, according to research commissioned by CUDO Compute, adding to concerns about the country’s ability to keep more AI activity on home soil.

The survey covered more than 700 senior AI decision-makers across the UK, US and Europe, including 200 in the UK. Among UK respondents, 33% said energy costs were limiting their ability to scale AI operations, while 43% said cost and performance outweighed sovereignty when deciding where to deploy AI.

The findings suggest the UK’s push for AI sovereignty is running up against infrastructure constraints. Businesses may want to keep workloads in domestic or regional markets, but power prices, available land and access to grid capacity are proving more decisive.

Geopolitics is also shaping deployment choices. Among UK respondents, 46% said geopolitical instability was pushing them to keep AI workloads within home markets, compared with 36% across the full sample.

Even so, commercial pressures remain strong. Almost a third of UK organisations (32%) said they were actively considering relocating workloads due to geopolitical pressures, while 45% said data sovereignty, regulatory compliance, or national security concerns were shaping their AI deployment strategy. At the same time, 31% said they were prioritising sovereign or regionally controlled compute even at a higher cost.

Where Workloads Go

When asked which markets looked most attractive for new AI cluster capacity, respondents ranked the US highest, with 72% viewing it positively. India followed at 62% and Eastern Europe at 58%.

Eastern Europe ranked ahead of Western Europe at 45% and the Nordics at 44%. China scored 55%, ahead of Latin America at 40%, the Middle East at 39%, Africa at 38% and APAC at 29%.

The pressure appears sharper for businesses that depend more heavily on compute. Among AI-first businesses, 32% said they would consider moving workloads overseas because of power costs, compared with 18% of enterprise organisations.

That gap suggests companies running the most demanding AI systems may be quickest to shift work to lower-cost locations when domestic operating conditions worsen. For policymakers, it highlights the challenge of matching AI ambitions with the industrial base needed to support them.

Infrastructure Strain

The findings reflect a broader issue in the AI market: infrastructure supply is constrained not only by access to chips and software, but also by physical requirements such as land, energy, cooling and grid access. In that context, electricity costs become a central part of the cost of compute.

CUDO Compute commissioned the research with Censuswide as part of its Land. Power. Compute report. Respondents included decision-makers responsible for AI workload and infrastructure decisions, budget input, vendor selection or active deployment planning. The sample included enterprise businesses with turnover above GBP £50 million and AI-first companies with turnover above GBP £1 million.

Matt Hawkins, chief executive of CUDO Compute, said the UK risked a widening gap between policy goals and operational reality if it did not address infrastructure constraints.

“AI sovereignty is being hotly discussed as a priority for UK organisations, but it only works if the infrastructure exists to support it,” Hawkins said. “What we are seeing is a growing tension between where businesses want to run AI and where they actually can.”

“AI is not abstract software. It is physical infrastructure that depends on power, land, cooling and grid access. When those constraints tighten, economics take over. If it is cheaper or easier to run workloads elsewhere, they will move, regardless of sovereignty ambitions.”

“Right now, every UK boardroom is talking about AI, but almost nobody is talking about the infrastructure needed to power it. Until we close that gap, there will continue to be a disconnect between policy, ambition and reality. The countries that solve this first will shape the future of AI, and the UK still has a window to lead, but it needs to act quickly.”