Microsoft terminates its Nscale data-centre deal at Loughton after the grid operator says power won't arrive until the 2030s

Start with what is actually confirmed. On Sunday, October 18, 2026, the Telegraph reported that Microsoft had terminated its data-centre agreement with UK cloud provider Nscale for a site in Loughton, Essex. Reuters relayed the report, noting that Microsoft did not immediately respond to a request for comment outside regular business hours and that the newspaper did not specify why Microsoft withdrew. The context the Telegraph gave: the project had faced repeated delays, and national grid officials told Nscale not to expect a power supply for the data centre until the 2030s.

Why does this matter more than one site won or lost? Because of who the customer is. Reuters puts it plainly: Microsoft is one of Nscale's two largest customers. When a major customer cannot hold even a single building inside a campus it committed to, the problem is not commercial terms but deliverability. Reuters also notes that Nscale provides infrastructure — compute, power and data centres — plus software for training and running AI models. In other words it sells a bundle of power, racks and software, and the piece that broke is the one you cannot ship.

Nscale's statement to Reuters deserves a line-by-line read. A spokesperson said: “Our relationship with Microsoft remains strong and long-term, and this unfortunate London-wide grid issue has no impact on our other contracts.” They added: “We remain committed to both the Loughton project and the UK, and we are exploring alternative power sources for the site, in addition to other expansion sites in the UK.” Two things in that are verifiable: the company acknowledges a grid problem and characterises it as London-wide rather than project-specific, and it has not abandoned the site, instead hunting for alternative power. As for the relationship being strong, that is standard corporate phrasing; readers can weigh it themselves.

The timeline is what makes the exit expensive. Reuters reports that Nscale filed for a US IPO last month, with its prospectus naming Microsoft and Anthropic as expected major customers. So at the very moment the company is pushing toward listing with Microsoft written in as a core customer, it loses that customer's flagship UK site. This is not a disclosure problem — a prospectus describes expectations — but a narrative problem: a document whose central selling point is “hyperScalers are locked in” now meets a specific project cancelled because the grid could not connect, and the first investor question becomes whether the model is replicable.

This is not an isolated case but a systemic bottleneck in UK AI infrastructure. In July, the energy regulator Ofgem opened a consultation on clearing speculative data-centre projects out of the demand-connection queue and reforming the process; the consultation referenced roughly 73GW of data-centre demand. In the same period, an industry survey found that 70% of infrastructure leaders at a London roundtable expect grid connections to take more than 18 months. The IEA's October report “Powering Artificial Intelligence” notes that Ofgem is considering stronger queue entry and permanence criteria, potentially including evidence of financial backing, planning permission and refundable deposits tied to milestones. Put together, Loughton's outcome is unsurprising: in a connection system whose queue dwarfs actual demand, the first project to fall out is the one with the latest power date.

Finally, the boundaries — being precise matters more than sounding dramatic. First, Microsoft has not publicly explained the exit; Reuters carries only the Telegraph's account of repeated delays and a power date in the 2030s, and this article does not invent a motive for Microsoft. Second, numbers circulating online — 23,000 rented NVIDIA GPUs at the site, inclusion in a £22 billion UK investment pledge from 2025, an original 2027 go-live — do not appear in the Reuters relay or the Telegraph reporting points I checked; they are secondary aggregation and are not treated as fact here. Third, Nscale's claim that other contracts are unaffected is the company's own statement with no third-party verification; equally, there is no evidence that the exit will spill over into its other UK projects. Fourth, Anthropic appearing in the prospectus as an expected major customer is unrelated to this event and should not be conflated with it.