Microsoft Exits Nscale’s UK Data Center Deal: Why Power Access Is Now the AI Bottleneck

Microsoft has reportedly ended an agreement for a data center project in Loughton, Essex, that was being developed by British cloud and infrastructure provider Nscale. The report puts a spotlight on a constraint that often receives less attention than graphics processors: a data center cannot serve AI workloads without a dependable supply of electricity, grid connections, cooling and a realistic construction timetable.

What Reuters reported

Reuters reported on October 18 that Microsoft had recently terminated its agreement for the Loughton site, citing a report by the Telegraph. Reuters said the newspaper described repeated project delays and reported that grid officials did not expect power supply for the site until the 2030s. Microsoft did not immediately respond to Reuters’ request for comment.

Nscale told Reuters its relationship with Microsoft remained strong and long-term, and said the London-wide grid issue did not affect its other contracts. The company said it remained committed to the Loughton project and the UK, while exploring alternative power sources and other expansion sites. Reuters did not report Microsoft’s reason for ending the specific agreement. It is important to keep that distinction clear: the timing and power context are reported, but Microsoft’s decision rationale has not been publicly confirmed.

Why power has become an AI infrastructure story

AI data centers concentrate large numbers of servers and accelerators in one location. They need a lot of electricity, and they must be connected to a grid capable of delivering it reliably. Local infrastructure can take years to plan and build. Even when a company has funding, customers and equipment, a delayed grid connection can postpone the point when a facility begins operating.

That creates a coordination problem. Developers want to secure power and land before demand grows; utilities need to plan transmission and generation; local authorities must weigh jobs and tax revenue against traffic, noise, water use and other community concerns. A project can look promising on a contract spreadsheet and still face long lead times before it can deliver capacity.

What this means for AI cloud customers

For customers, provider announcements about planned capacity are not identical to available capacity. Businesses should ask where services run, how much is live today, what redundancy exists, and whether a provider has a credible plan for power and cooling. A project delay does not prove that an entire provider is unreliable, but it is one reason to avoid treating future capacity as operational inventory.

For infrastructure firms, diversified sites and energy strategies can reduce dependence on one constrained grid connection. Those alternatives can bring their own permitting, engineering and cost challenges. The Reuters account does not say which options Nscale will use at Loughton, only that it is exploring alternatives.

A reality check for the AI build-out

Recent coverage of the AI infrastructure boom has focused on the scale of planned investment and financing risk. This project offers a more concrete lesson: even a major cloud buyer may adjust its commitments when schedules and utility constraints change. The next phase of AI expansion will depend not only on faster chips but on energy planning, local approvals and disciplined project execution.

See our related analysis of AI deployment risks in live systems and the global technology policy debate.

Source: Reuters, October 18, 2026. Reuters attributed the termination report to the Telegraph; Microsoft’s reason was not specified in the report.