Nvidia / 4 October 2026

Grid Delay Threatens Nscale’s 23,040-GPU Essex AI Factory

Nscale’s planned Loughton AI facility, designed around 23,040 Nvidia GB300 GPUs, faces a reported electricity-grid delay that could prevent the project from receiving its full planned power connection until the 2030s. The Essex site had been targeted for an initial 2027 opening and was planned to begin around 50 megawatts before scaling toward 90 megawatts. The Guardian reported that UK Power Networks could not provide the required capacity on the original timetable, potentially pushing full operation into the middle of the next decade. Nscale is exploring alternatives including on-site generation and accelerated connection options. The update turns grid availability into a specific execution risk for a project intended to convert large financing commitments into Nvidia-powered cloud capacity.

The planned accelerator count gives the power problem commercial scale. More than 23,000 GB300 GPUs represent a substantial rack-scale installation, requiring not only processors but networking, cooling, backup power and a continuous high-density electricity supply. The site was intended to support Microsoft Azure workloads and form part of the UK’s expanding domestic AI infrastructure. Clearing the land or constructing a building does not create usable computing capacity if the grid cannot energize the racks. Nscale may be able to start at a lower power level or use temporary generation, but those alternatives carry different cost, emissions, permitting and reliability characteristics from the original grid-connected design.

This is a distinct risk from Nscale’s ability to raise capital. The company has secured large financing commitments and describes a very substantial contracted customer book, while Nvidia itself has committed capital to the business. Financing can pay for equipment and construction; it cannot create transmission capacity on demand. The UK has hundreds of data-center projects waiting for grid connections, according to the reported regulator figures, making the Loughton problem part of a wider infrastructure queue rather than an isolated equipment delay. That matters to Nvidia because sales pipelines increasingly depend on megawatts being available at the same time as GPUs.

The timing mismatch can shift hardware demand even when customer contracts remain intact. If a site planned for GB300 cannot be energized until several product generations later, the final equipment configuration may change. Nscale could move workloads to another location, deploy less initial capacity, or install a later Nvidia architecture when power becomes available. Each response preserves some long-term demand while moving the revenue timing and product mix. Nvidia therefore has an incentive to support designs that can be commissioned incrementally and to work with partners on power flexibility, but it cannot eliminate regional grid constraints through semiconductor supply.