The planned launch of the United Kingdom's largest AI-focused supercomputer in Loughton, Essex, has been pushed back by several years due to unresolved power supply constraints.

Originally announced in 2025 as the country's biggest sovereign AI datacentre, the project faced mounting uncertainty after The Guardian reported in March that the site remained an active scaffolding yard. Further complications have emerged as UK Power Networks (UKPN) informed developer Nscale that the local electricity grid cannot deliver sufficient capacity until the early to mid-2030s.

The delay was first disclosed by AI industry newsletter Morning Intelligence. Nscale, a UK-based startup pursuing a £26bn valuation through a planned stock market listing this year, confirmed its commitment to the project but acknowledged that securing adequate power remains a critical bottleneck for datacentre development. The company has since cleared the scaffolding yard and is exploring alternative solutions, including on-site power generation and accelerated grid connection processes.

The Loughton facility requires up to 90 megawatts of power capacity — equivalent to the annual electricity use of approximately 315,000 households. This demand places it among 315 datacentres currently awaiting grid connections nationwide, collectively representing 73 gigawatts of proposed energy demand, according to Ofgem. National Grid Electricity System Operator noted ongoing efforts to streamline connections and support economic growth through infrastructure upgrades.

Government officials emphasized that reversing grid underinvestment remains a priority, with reforms underway to improve the efficiency of power connection applications. While the project was previously highlighted in the UK's AI strategy as a cornerstone of national technological advancement, its timeline now hinges on resolving systemic energy supply challenges rather than technical or financial hurdles alone.