Microsoft has withdrawn as the anchor customer for Nscale’s proposed Loughton data centre, according to AI News’s October 26 report. The report describes an electricity-connection constraint stretching into the early-to-mid 2030s, putting a physical limit alongside the much more visible competition for AI chips and capital.
Nscale remains committed to the project, the article says, and is exploring alternative power arrangements and a faster connection. The distinction is important: a customer departure is not the same event as cancellation of the development. Neither is an alternative under investigation equivalent to an available electricity supply.
A connection timetable collides with an earlier plan
AI News reports that UK Power Networks told Nscale sufficient electricity would not be available until the early-to-mid 2030s. It also says Nscale’s work with Microsoft continues in other markets. Those are claims in the discovered report; they should not be read as an independently published network commitment or a new joint statement from the two companies.
The immediate question is therefore one of sequencing. A site can have a developer, an intended customer and a proposed technical design without having a delivery timetable that fits the customer’s requirements. Changing one part of that arrangement may preserve the project while materially changing the commercial proposition.
That makes the reported withdrawal significant without requiring a prediction about the site’s eventual fate. The information establishes a mismatch between an earlier ambition and the connection timetable now described, not proof that every possible route to operating the facility has been exhausted.
What Nscale originally announced
The contrast is clear in Nscale’s September 16, 2025 announcement. At that point, the company described Microsoft’s commitment at Loughton as a 50-megawatt project, scalable to 90 megawatts, with 23,040 Nvidia GB300 GPUs targeted for the first quarter of 2027. Those figures describe the older plan. They are not evidence that the equipment has been delivered or that the original timetable remains achievable.
The same announcement covered a broader UK infrastructure programme. Its larger aggregate GPU figure should not be assigned wholesale to Loughton, and the site’s expandable capacity should not be mistaken for power already secured. Reading those different numbers as interchangeable would overstate both the size and readiness of the specific development.
This is why announcements about AI infrastructure need more than a headline capacity total. A proposed scale, an initial operating phase and a confirmed grid connection answer different questions. The reported change at Loughton concerns the relationship between them, rather than a simple reduction in a single inventory of computing equipment.
The queue problem predates this withdrawal
There is relevant regulatory context, but it is not a new policy announcement. In its November 2025 demand-connections update, Ofgem described applications rising sharply from November 2024 and exceeding even ambitious demand forecasts. It warned that this could delay projects ready to deliver benefits and outlined work on a viable queue, regulatory reform and strategic demand connections.
That document confirms the broader problem was already recognised. It does not promise an accelerated connection for Loughton, identify this project’s position or establish which alternative power option would be accepted. A national reform programme and an individual site’s available supply remain separate pieces of evidence.
The distinction also limits what can be inferred from the existence of a queue. Removing speculative applications might improve access for viable projects, but it would not by itself demonstrate that a particular network can deliver the required capacity on a particular date. An assessment of this site needs that site-specific evidence. Neither the older company announcement nor Ofgem’s general update supplies it, so the revised timetable cannot responsibly be converted into a confident opening date.
For now, the development to watch is a documented route from the revised project to usable power. A replacement customer or another capacity announcement would answer only part of the question. Without a compatible connection schedule, the earlier hardware target remains an ambition rather than a demonstrated operating plan. The reported departure makes that gap harder to overlook.










