Microsoft reportedly leaves Loughton as Nscale searches for power

A delayed data centre can lose more than time. When a major customer leaves, the problem becomes who will buy the computing capacity—and when the developer can start earning revenue. Securing electricity and securing orders have to work together.

Microsoft has pulled out of Nscale’s planned AI data centre in Loughton, Essex, the Guardian and BBC report. The Guardian’s October 18 account says Microsoft was expected to be the anchor tenant, the large customer around which the project’s business would be built. It credits the Telegraph with first reporting the end of the deal. The BBC says it understands Microsoft terminated its involvement; Microsoft declined to comment. [1] [2]

Nscale’s September 16, 2025 announcement tied the Loughton site to Microsoft’s Azure services. It promised 50 megawatts of AI capacity, scalable to 90 megawatts, and an initial 23,040 Nvidia GB300 processors delivered in the first quarter of 2027. Those were commitments for a future facility, rather than evidence that the capacity was already operating. [3]

The electricity problem predates the reported customer exit. On September 24, 2026, the Guardian reported that Nscale had been told sufficient grid power would not be available until the early to mid-2030s. The new development is the loss of the expected anchor customer, beyond the already-reported delay. [4]

That distinction matters for Nscale’s financing and prospective revenue. A long wait for power leaves a delivery problem; losing a customer adds the task of selling the capacity again. In a hypothetical procurement, a buyer with a fixed start date might choose another facility rather than wait. The developer would then need both a workable power schedule and a fresh commitment to buy its services.

Nscale says it remains committed to Loughton and is exploring alternative power sources. It told the BBC the grid issue has no impact on its other contracts. The BBC also reports that Nscale and Microsoft are still working together on projects in Norway, Portugal and the United States. [1] [2]

The strongest alternative reading is a site-specific power setback, rather than weakening demand for AI computing. Alternative power and a replacement anchor could restore a route to delivery. A funded supply plan paired with a new customer agreement would weaken the interpretation that this delay has become a lasting commercial impairment.

The reporting does not establish a lost-revenue amount, a termination payment or a write-down. Assigning a financial loss would require contract terms and project-level figures that these accounts do not provide. Nor does one withdrawal establish that the wider Microsoft–Nscale relationship has failed.

Watch electricity and orders together. A firm date for usable power, or a funded alternative-power plan, would address delivery. An anchor agreement with a stated service-start date would address demand. The economic test is whether those milestones converge on paying workloads; a revised opening promise alone would leave the customer question unresolved. Alternative power would also need to be assessed for its operating cost, rather than treating an earlier start as an automatic improvement in project economics.