Microsoft has pulled out of a planned lease for a data centre in Loughton, Essex, after delays to the site’s expected electricity supply. The project, developed by Nscale, was intended to support Microsoft’s cloud infrastructure. Reports said grid power might not be available until the 2030s, a timeline that undermined the site’s usefulness for near-term expansion.
Power is becoming a deployment bottleneck
The proposed facility was described as a 50-megawatt site with roughly 23,000 Nvidia GB300 chips. Those figures underline the scale of modern AI infrastructure, but hardware plans cannot compensate for a missing grid connection. Data centres need dependable electricity alongside cooling, fibre connectivity, land, permits and specialist equipment.
Microsoft’s withdrawal is a setback for the Loughton project, but it should not be read as the end of its wider relationship with Nscale. The developer has said it remains committed to the site and is exploring other power options and opportunities in the UK.
The wider infrastructure challenge
Grid delays can strand investment and complicate cloud capacity planning. Operators may consider dedicated generation, energy storage or alternative locations, although each option brings cost, engineering and regulatory trade-offs. New power arrangements also need to be reliable enough for workloads that customers expect to run continuously.
The episode is a reminder that the AI build-out is not only a race for accelerators. Electricity infrastructure and the time required to connect it can determine whether a data-centre project is viable on its original schedule.
Power is now part of site selection
Data-centre projects depend on more than land, buildings and financing. Large facilities require a reliable supply of electricity, grid connections, cooling capacity and a schedule that matches customer demand. If power cannot be provideed for years, a developer may be unable to meet the commercial timetable even when the physical site is progressing. For AI infrastructure, the gap between planned capacity and usable capacity can be particularly consequential.
Leaving a project does not necessarily mean the facility itself has been cancelled. The ownership structure, lease commitments and responsibilities for construction can be complicated, and another operator may continue with a revised plan. The important distinction is between a project being abandoned altogether and one participant withdrawing from a particular commercial arrangement because a key dependency has changed.
Electricity constraints are becoming a planning issue for cloud providers and local communities. Data centres can bring investment and jobs, but they also compete for grid capacity and require long-term infrastructure upgrades. Utilities may need to build substations, transmission lines or generation capacity before a site can operate at its intended scale. Those works can involve planning reviews and supply-chain delays beyond a tenant's direct control.
What the withdrawal means for capacity plans
Developers can respond by selecting sites with stronger grid access, phasing construction, improving energy efficiency or negotiating alternative supply arrangements where legally and technically feasible. None of these options is instantaneous. Backup generators are designed for resilience and are not a simple substitute for a permanent high-capacity grid connection, particularly for facilities that need continuous operation.
For customers, the episode is a reminder to distinguish announced capacity from operational capacity. Cloud road maps and AI infrastructure plans often refer to future sites, but service availability depends on facilities that are actually powered, networked and commissioned. Delays at one project can lead companies to shift demand elsewhere, potentially increasing pressure on other regions.
Power availability should now be considered alongside land price, fibre access and local planning when organisations evaluate new data-centre regions. A site with a large announced capacity may provide less practical value than a smaller location that can connect to the grid sooner. Customers should ask providers which capacity is operational, which remains under construction and what dependencies could delay providey. The withdrawal shows a real constraint on AI growth, but it does not by itself establish that every UK project is stalled. Each development has its own contracts, grid position and timeline.
Local planning, grid capacity and providey schedules should be treated as connected parts of a data-centre proposal. A project can be well funded and technically sound yet miss its target if the utility connection is delayed. For customers, the sensible response is to maintain flexibility across regions and ask providers for realistic service dates. For developers, early coordination with grid operators can be as important as securing land or signing a tenant.