The UK’s much-heralded ambition to house the nation’s largest artificial intelligence supercomputer has hit a formidable structural wall. Announced in 2025 by Prime Minister Keir Starmer’s government, the flagship datacentre project in Loughton, Essex, will miss its targeted 2027 debut and now faces potential delays stretching into the early to mid-2030s.
The Power Grid Bottleneck Threatening Britain’s AI Ambitions
At the heart of the setback lies a severe mismatch between government ambition and physical energy infrastructure. The sprawling facility, developed by UK-based startup Nscale—which is currently seeking a $35bn (£26bn) public flotation—requires up to 90 megawatts of electrical capacity. That immense draw equals the energy consumption of roughly 315,000 households, turning a standard tech build into an energy engineering crisis.
UK Power Networks (UKPN), the utility responsible for linking the Essex site to the local distribution grid, informed Nscale that the regional infrastructure cannot supply sufficient power until the early to mid-2030s. The timeline collapse was first surfaced by the AI industry newsletter Morning Intelligence.
While Nscale confirmed its commitment to the project and noted that the previously cluttered scaffolding yard on the site has now been cleared, the physical reality of the grid remains unyielding. The startup is investigating alternative workarounds, including onsite power generation and accelerated connection procedures, to bypass the decade-long wait.
Ofgem Warnings and National Grid Logjams
The Loughton delay is far from an isolated administrative hiccup. It exposes a systemic grid logjam threatening the entire technology sector.
Energy industry regulator Ofgem revealed a staggering queue for grid access. In June, regulatory figures showed 315 datacentres waiting to connect to the National Grid, representing a combined demand of 73 gigawatts. To put that figure in perspective, the peak energy demand for the entire United Kingdom stands at approximately 45 gigawatts.
UKPN emphasized that major energy consumers require substantial upgrades at the transmission network level before regional distribution networks can safely energize them. National Grid officials stated they are collaborating with NESO—the government-owned National Energy System Operator—alongside distribution companies to find acceleration opportunities where safety and capacity permit.
Sovereign AI Visions Meet Hard Physics
When Downing Street championed the Loughton facility as the cornerstone of the UK’s sovereign AI strategy, it positioned the site as the central nervous system for future domestic AI models and chatbots. Sovereignty in the digital age, ministers argued, required domestic silicon running on domestic power.
Yet, the site’s rocky trajectory—moving from an operational scaffolding yard revealed by investigations in March 2025 to a cleared plot now stranded by a decade-long power wait—illustrates the hazards of political aspiration outpacing engineering reality. As energy constraints tighten their grip on datacentre development, Britain’s race to lead the artificial intelligence revolution must first solve the very basic challenge of keeping the lights on.
