UK Firms Shift AI Infrastructure Abroad Over Energy Costs
Rising electricity prices and constrained grid capacity are pushing British AI teams to run models and training workloads outside the UK. A CUDO Compute study, using a Censuswide survey of 700 senior AI decision makers, finds one in five UK firms have already moved AI workloads overseas and 32% of AI-first companies are considering relocation because of power costs. At the same time about 30% of organizations prioritize sovereign or regionally controlled compute despite higher expense. The mismatch between policy ambitions for domestic AI leadership and the realities of local energy and grid constraints is acute, prompting government moves on electricity contracts and levy changes to blunt gas-price-driven spikes.
What happened
A new study from CUDO Compute, based on a Censuswide survey of 700 senior AI decision makers, shows UK firms are actively relocating AI workloads overseas because of high energy prices and limited grid capacity. One in five UK firms have already moved workloads abroad, 32% of AI-first companies are considering moving, and roughly 30% still prefer sovereign compute even at higher cost.
Technical details
The study highlights two operational limits that shape deployment choices. First, high wholesale electricity driven by gas pricing increases operational cost-per-FLOP for GPU clusters. Second, physical constraints slow new datacenter buildouts: planning delays and grid connection backlogs mean capacity cannot be turned into usable compute quickly. The report cites the example of Santa Clara where nearly 100 MW of new datacenter capacity is awaiting power hookups. Key implications for practitioners include:
- •Increased unit compute cost in high-energy markets, raising total cost of ownership for GPU-heavy training and inference
- •Lead times for colocated capacity driven by grid upgrades and permitting, not just server procurement
- •Tradeoffs between cheaper offshore capacity and the operational, regulatory, and data-residency benefits of onshore sovereign compute
Context and significance
This is a structural infrastructure story, not a transient cloud-price blip. The UK government is pushing for domestic AI capability, but energy economics and grid readiness are shifting where models will be trained and hosted. The dynamic amplifies the bargaining power of regions with cheaper, more reliable power and faster grid expansion, and hurts local startups and research groups that need predictable, affordable power for continuous training runs. Vendors and cloud providers that can offer integrated power solutions, or long-term fixed-price contracts, will gain competitive advantage.
What to watch
Track government interventions such as long-term fixed electricity contracts and changes to the Electricity Generator Levy, hyperscaler investments in UK grid connections, and procurement choices by AI-first firms balancing cost with sovereignty. These signals will determine whether compute returns to the UK or continues to move offshore.
Scoring Rationale
This story has material operational impact for ML teams and infrastructure planners because it affects where compute is built and how much it costs. It is not a frontier-model development, but the infrastructure constraints are significant for deployments and capacity planning, warranting a mid-high importance score.
Practice interview problems based on real data
1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problemsStep-by-step roadmaps from zero to job-ready — curated courses, salary data, and the exact learning order that gets you hired.
