UK's Largest AI Supercomputer Project Delayed to Mid-2030s Over Power Grid Bottlenecks
UK Power Networks has informed tech startup Nscale that its planned flagship AI supercomputer facility in Loughton, Essex, cannot receive sufficient power until the early-to-mid 2030s. As a result, the high-profile project faces severe delays past its original operational target as it struggles to secure a 90MW power allocation. This delay highlights how regional electrical grid capacity bottlenecks are becoming a primary obstacle to scaling national AI infrastructure. It illustrates the mounting friction between the exponential power demands of modern AI hardware and the long lead times required to upgrade electrical power transmission grids. The Loughton facility requires up to 90MW of power—equivalent to the consumption of roughly 315,000 homes—while UK energy regulator Ofgem reported 315 data center connection requests totaling 73GW in queue, far exceeding the UK's peak demand of 45GW. In response, Nscale is evaluating on-site power generation solutions and exploring methods to accelerate grid connection timelines.
## BACKGROUND
Nscale is a UK-headquartered technology firm providing sovereign GPU cloud infrastructure and data center capacity tailored for high-performance AI workloads. AI supercomputers require continuous mega-watt scale power to run dense clusters of graphics processing units (GPUs) for model training and inference. As AI deployment expands globally, regional power grids struggle to accommodate the sudden surge in electricity demand from power-dense data centers.