Nscale’s UK AI Supercomputer Delay Shows the Grid Bottleneck
Published by AINave Editorial • Reviewed by Ramit
Nscale’s planned UK AI supercomputer in Loughton was due to open in 2027. Now the project could slip into the 2030s because the local grid cannot provide enough electricity on the expected schedule, putting power access, rather than the announced compute plan, at the center of the timeline risk. The new date is a possibility, not a confirmed revised opening. The project’s reported power constraint
A connection delay reaches beyond the site
UK Power Networks, which is to connect the Loughton data centre to the local grid, told Nscale that enough electricity would not be available until the early to mid-2030s, according to the newsletter The Morning Intelligence. UK Power Networks pointed to work required on the wider national transmission network. National Grid said it was working with the system operator NESO and local network operators to speed up connections. The reported grid timeline and network work
That distinction matters: the immediate obstacle is not simply equipment at the data-centre site. The connection depends on wider network work, so a project’s physical construction and its ability to operate at planned scale can follow different schedules.
Nscale is looking at generating power on-site and discussing ways to accelerate the connection with UK Power Networks. Those are options under consideration, not confirmed solutions or evidence that the opening date will move forward. Nscale’s reported discussions
The proposed scale makes power access central
In 2025, Nscale announced plans to install around 23,000 Nvidia AI chips at the site and link them into a supercomputer intended to support Microsoft Azure cloud services. That remains a plan, not evidence that the chips are installed or the facility is operational. The announced chip and cloud plans
Nscale says the data centre could eventually grow from 50 to 90 megawatts. At 90 megawatts running continuously, the annual electricity use would be equivalent to about 315,000 typical UK homes, using 2026 figures from energy regulator Ofgem. The comparison illustrates the scale of the proposed power demand, but it describes continuous use at the potential expanded capacity, not the facility’s current consumption. Nscale’s stated capacity range and the Ofgem-based comparison
Loughton is also part of Nscale’s wider plan to install nearly 59,000 Nvidia AI chips across UK data centres. The wider UK chip plan The useful lesson in this project is specific: naming chips and cloud customers does not settle when a large computing site can run. Its schedule also depends on when the electricity network can serve it, and the available evidence does not yet show whether the company’s proposed alternatives will close that gap.
