In the scrubby outskirts of Loughton, Essex, diggers still churn the earth for what was meant to be Britain's flagship AI data centre. Nscale's £2bn facility, backed by Nvidia and anchored by Microsoft, now faces fresh uncertainty after the company was told its promised 90 megawatt grid connection will not arrive in time for the planned 2027 opening.
The news, first surfaced by industry sources, lands after the project had already slipped from an original target of late 2026. Construction carries on regardless. Yet the delay lays bare a harder truth: the United Kingdom's creaking electricity grid is becoming a decisive brake on the very infrastructure ministers say the country needs to compete in artificial intelligence.
Limited grid capacity has emerged as the single biggest bottleneck for data centre developments nationwide. For an industry that consumes power on a scale once reserved for small cities, the wait times for new connections stretch into years. Nscale's experience is not isolated. It reflects a pattern repeated across multiple sites where ambition collides with the physical limits of an ageing network.
Developers are responding with pragmatism rather than complaint. Nscale is in discussions with Bloom Energy about deploying solid oxide fuel cells to generate power on site while the permanent grid link catches up. The company has reaffirmed it remains fully committed to the Essex project, which secured planning permission despite local reservations and has been positioned by some in government as critical national infrastructure.
This search for alternatives matters. Onsite generation, once viewed as a niche or emergency option, is shifting toward mainstream contingency planning. It demonstrates the private sector's willingness to solve problems that centralised systems have left unaddressed. Fuel cells of the type under consideration can run on natural gas or, in future configurations, hydrogen, offering a bridge until grid reinforcement arrives.
The human stakes sit far from the server racks. Every month of delay postpones the thousands of skilled jobs, supply-chain contracts and computing capacity that were supposed to anchor UK leadership in AI. Microsoft, as anchor tenant, had eyed the site as home to one of the country's largest AI supercomputers, capable of housing up to 45,000 advanced Nvidia GPUs. That compute power is not abstract. It underpins everything from drug discovery to industrial optimisation and defence modelling.
