Plans to build Britain's largest artificial intelligence supercomputer have hit a major roadblock, with launch dates potentially pushed back from next year into the mid-2030s. The ambitious facility in Loughton, Essex, was announced in 2025 as a centerpiece of the UK's national AI strategy, but local electricity grid operator UK Power Networks (UKPN) has reportedly warned that it cannot supply enough power for years to come.
The project is being developed by Nscale, a UK-based startup preparing for a £26 billion ($35 billion) stock market flotation. The proposed data center requires up to 90 megawatts of power—an amount equivalent to the energy consumption of roughly 315,000 homes. Although Nscale has cleared the site and is investigating options to generate electricity on-site or accelerate connection times, grid capacity remains a severe challenge.
The delay highlights a growing national issue as data centers, which serve as the backbone for AI technologies like chatbots, struggle to secure connections to the power grid. Energy regulator Ofgem recently revealed that 315 data centers are currently queueing to connect to the National Grid. Together, these applications represent 73 gigawatts of demand, far exceeding the entire country's peak electricity demand of 45 gigawatts.
While Nscale stated that it remains committed to the project, network operators say large-scale facilities require extensive infrastructure upgrades across the wider national transmission network. UKPN noted that many data centers face similar wait times for wider grid improvements, while National Grid stated it is working with regulators and operators to find ways to accelerate connections where possible.