> ## Content Index
> Fetch the complete content index at: https://nextwith.ai/llms.txt
> Use this file to discover other available public pages before exploring further.

# Nscale's Loughton AI campus faces a grid delay that could push it into the 2030s
- URL: https://nextwith.ai/nscales-loughton-ai-campus-faces-a-grid-delay-that-could-push-it-into-the-2030s/
- Published: 2026-09-25T08:06:41.000Z
- Updated: 2026-09-25T08:06:41.000Z
- Description: Guardian reporting says Nscale’s Loughton AI campus could slip into the early-to-mid 2030s if the grid cannot deliver enough power in time. Ofgem’s queue data shows why this is part of a wider UK bottleneck.
- Author: NextWith.ai Editorial Desk
- Tags: Business, News

Nscale’s planned AI campus in Loughton, Essex, is running into the constraint that matters most for large-scale AI infrastructure: electricity. In reporting published by [The Guardian](https://www.theguardian.com/technology/2026/sep/24/construction-largest-supercomputer-delayed?ref=nextwith.ai) on 24 September 2026, the project was described as likely to miss its intended 2027 launch because the grid connection may not be able to deliver enough power until the early to mid-2030s.

That is a consequential delay because the site is not a small server room waiting for rack hardware. Nscale’s own [infrastructure page](https://www.nscale.com/ai-infrastructure?ref=nextwith.ai) describes Loughton as a high-density AI campus that can scale to 90MW, with direct-liquid cooling and fibre connectivity aimed at GPU-heavy workloads. In other words, this is the kind of project where power delivery is the product. Without energisation, the building can exist while the compute cannot.

## What changed

The project has not been reported as cancelled. The reported grid timetable casts doubt on its planned 2027 launch. The Guardian reported that UK Power Networks, which would connect the site to the local grid, told Nscale that the network will not be able to supply sufficient power on the original timeline. Nscale said it remains committed to the project and is examining whether it can generate power on site or accelerate the connection process.

That distinction matters. Land clearance, site preparation and public announcements can move quickly. Grid reinforcement does not. For an AI facility, the useful date is not when construction is visible from the road but when the site can actually draw the load needed for production training or inference.

## Why the grid is the bottleneck

The broader problem is visible in [Ofgem’s 29 July 2026 press release](https://www.ofgem.gov.uk/press-release/ofgem-acts-free-grid-capacity-tackling-speculative-data-centre-projects?ref=nextwith.ai) on data-centre connections. The regulator said electricity demand connection applications had risen from 41GW to 125GW in under a year, driven largely by data-centre projects. Ofgem also said data centres accounted for at least 80GW of that increase and that around 73GW of data-centre demand was referenced in its consultation.

Guardian reporting on Ofgem’s figures put the queue in more concrete terms: 315 data centres were waiting to connect, representing 73GW of demand, versus a peak national demand of 45GW. That comparison does not mean every queued project is equally viable or equally delayed, but it shows the scale of the mismatch. A single campus in Essex is competing with a much larger system problem.

The technical reason is straightforward. Large AI sites require not only a local distribution connection but, in many cases, work on the wider transmission network before they can begin operating. UKPN said in general that many data centres wait on that wider network work, before they can begin operating. For AI infrastructure, the limiting factor is increasingly not the server design but the grid timetable.

## Who should care

For AI developers, the lesson is that location strategy is now an energy strategy. Proximity to London and strong fibre are useful, but they do not solve a power deficit. For investors, a site can be technically impressive and still slip by years if the connection path is not secure. For enterprise buyers, the practical implication is that capacity claims and delivery dates should be treated as separate questions.

There is also a policy implication. The UK government has leaned on projects like Loughton to support its AI strategy, but Ofgem’s queue data suggests that industrial policy now depends on grid planning as much as on incentives or planning permission. If the queue stays clogged, “AI capacity” can remain a promise rather than an operating asset.

## The limitation

The exact length of the delay is not independently verified in the material reviewed here, and the contractual impact on Nscale’s schedule is not public. The power estimate comes through Guardian reporting on what UKPN reportedly told Nscale, not from a direct public statement by the grid operator. Nscale’s statements about commitment and workarounds are company claims, not proof that the problem is solved.

Even so, the underlying signal is clear enough for decision-makers: for this project and others waiting on a grid connection, the operational launch depends on when sufficient power arrives, not just when construction ends. Watch whether Ofgem turns its consultation into binding queue rules, because that will show whether AI sites with firm power plans can move ahead of speculative projects.