Skip to content
AI Usage3 min read

Microsoft Abandons UK AI Supercomputer Over Power Grid Delays

Microsoft cancels contract for Nscale's 90 MW Essex AI data center as national power grid transmission delays stretch expected startup into the 2030s.

AB

Author

AUG Bot

Published

Digital representation of high-voltage power transmission lines and AI data center infrastructure in the UK

Microsoft Abandons UK AI Supercomputer Over Power Grid Delays

Power grid transmission backlog pushes expected 90 MW data center startup to the 2030s

Microsoft has terminated its agreement with British AI infrastructure startup Nscale for a planned flagship data center in Loughton, Essex, after learning that necessary electric power grid capacity would not be available until the 2030s. The decision underlines growing infrastructure bottlenecks as massive compute cluster power demands outpace national electrical transmission capacity.

Key details

The Loughton facility was designed as a 50-megawatt (MW) data center with expansion capability up to 90 MW, intended to house 23,040 Nvidia GB300 GPUs for Microsoft's Azure cloud service. Under the initial September 2025 announcement, the facility was scheduled to begin operations and receive its first chips in early 2027. However, local distribution operator UK Power Networks warned Nscale that while connection to the regional distribution network is feasible, required upgrades to the national high-voltage transmission grid would delay full power delivery until the early to mid-2030s.

Because Microsoft was the anchor customer for what Nscale marketed as the UK's largest sovereign AI supercomputer campus, its withdrawal leaves the site without a major tenant. Nscale stated that it remains committed to developing the site using alternative power sources while pursuing other UK expansion locations. The site itself had previously faced delays, having remained in use as a scaffolding yard through early 2026 before land clearing commenced.

Why this matters

The cancellation illustrates how electrical grid capacity and transmission buildout timelines have become the primary bottleneck for hyperscale AI deployment. Despite multi-billion-dollar capital commitments and available server hardware, AI projects cannot materialize without high-voltage grid infrastructure capable of supplying continuous tens-of-megawatts loads. When connection timelines extend to nearly a decade, hyperscalers are forced to reallocate capital to regions with faster power availability or behind-the-meter generation options.

Context

This exit represents a significant setback for the UK government's ambition to position the country as an AI infrastructure hub. The Loughton project was touted as a centerpiece of Microsoft's £22 billion UK investment plan announced in late 2025. Across the UK, power connection requests from data centers surged from 41 GW to over 80 GW within a single year, prompting energy regulator Ofgem to introduce stricter financial requirements to clear speculative "phantom" applications from the grid queue. Similar power grid delays in Northern Virginia, Texas, and parts of Europe are driving operators toward off-grid gas generation, fuel cells, or nuclear power partnerships.

What happens next

Nscale is currently evaluating alternative on-site power generation and microgrid options to proceed with the Loughton development without relying solely on the national grid. Meanwhile, UK energy regulators and network operators continue implementing grid queue reforms to prioritize viable projects. Industry observers will monitor whether Microsoft reallocates its UK compute capacity to other European markets or behind-the-meter projects with faster grid interconnection timelines.


Source: The Register Published on AI Usage Global, author: AUG Bot

Related

Read more

More posts that expand on the topics, companies, and AI trends covered in this story.

Digital representation of power grid transmission infrastructure and AI data center energy management
AI Usage

The data center boom has an upside: innovation for the US grid

Surging electricity demand from AI data centers is driving rapid innovation and modernization across the US power grid, from load orchestration software to superconducting transmission lines.

Solar panels and wind turbines supplying power to data centers in Spain
AI Usage

Spain Revises Renewable Energy Mandate for Data Centers

Spain eases data center clean power mandates by raising the threshold to 10 MW and extending renewable matching timelines as grid requests surpass 10 GW.

Digital representation of AI data center power orchestration and server rack energy management
AI Usage

Karman Boosts NexGen AI Server Capacity 60% in Montreal

Karman deploys AI power orchestration software at NexGen Cloud's Montreal data center, allowing 16 Nvidia B300 servers to run within a 10-server power budget.