Executive Summary

• Charles Deacon, Technical Director of Grid at Eclipse Power, takes a look at the new data centre dilemma: the realistic chances of delivery of data centre projects and why flexible routes to power will be essential.
• Ofgem is looking to flush speculative projects out of the power queue, paving the way for viable projects to access grid connections. In July it launched a consultation on a new data centre commitment fee – a charge, payable when a large data centre accepts a connection offer, and refunded on energisation, but forfeited if the project drops out early.
• Reform is coming, but developers can’t afford to wait for it. The routes around the queue already exist: private networks, co-located generation, and connections built and owned by independents rather than left in the incumbent’s backlog.

High electricity costs and slow grid connections are shaping the outlook for data centre development in the UK. With demand accelerating, Charles Deacon, Technical Director of Grid at Eclipse Power, explains where and why flexible routes to power will be essential.

The UK has spent the last year positioning itself as an AI superpower. But there’s a difference between our ambitions and the realistic chances of delivery. The pause of OpenAI’s Stargate UK project was a high-profile reminder of the challenge that high energy prices and grid connection delays create – a challenge that’s now central to whether major projects can move forward at all, and even the UK’s ability to compete for data centre investment.

Is the UK losing the data centre race?

It’s worth keeping a cool head and being realistic about the UK’s attractiveness for developers. The UK retains real advantages for the data centre market. It’s geopolitically stable, has a favourable policy environment, has reliable – albeit expensive – power, a temperate climate, and genuine demand for the tech. In other words, the fundamentals are there.

With that said, the data centre market isn’t uniform. There is nuance in data centre types, and not all areas of the market are equally mobile. Cloud development is often quite strongly anchored to its fibre routes – for example, most UK-serving cloud data centres are in West London – but hyperscalers and big tech developers are internationally mobile. These organisations are most able to shift their investment to different regions, and they’ve already shown their willingness to do so.

Why should that trouble us in the UK? High industrial energy prices are the first issue. The average transmission demand residual charge has risen by around 64% in a single year (2025/26 to 2026/27), meaning that large users are facing steep bill increases on the transmission element of their charges. That’s a direct result of two price-control periods that failed to incentivise grid investment for a renewable grid, now being corrected all at once.

Though this certainly isn’t a dealbreaker for data centre developers, it is nonetheless a worrying trend that – without recourse – can continue to make the economics of new projects more difficult to manage.

The thornier problem is the connection timeline. The demand connection queue has exploded, with Ofgem’s figures showing an increase from 41GW to 125GW in the eight months from November 2024 to June 2025 alone. Within that, transmission-connected demand surged from 17GW to 97GW – well over fivefold. In my estimation, over 100GW of the total queue will be unable to connect before 2030; a fresh application today could take up to 15 years to get a connection.

This is in part due to the influx of data centre developer projects themselves, but also supply chain issues constraining National Grid’s ability to build at a pace sufficient to match demand, and the historical omission of serious distribution-level demand reform within the overall connections reform agenda.

Happily, Ofgem is alert to the problem. In July it launched a consultation on a new data centre commitment fee – a charge of between roughly £237,500 and £712,500 per megawatt. This would be payable when a large data centre accepts a connection offer, and refunded on energisation, but forfeited if the project drops out early. The intent is to flush speculative applications out of the queue so that viable, investment-ready projects can move faster. It’s a sensible step, but it’s a filter on the queue rather than a fix for the underlying shortage of timely access to power. However, it appears that the initial quantum of fee proposed is far too high and could swing the balance too far the other way.

AI Growth Zones aren’t a ready-made shortcut

The final issue data centre developers face is the design of AI Growth Zones (AIGZs) themselves. Government-backed AIGZs are intended to speed up investment by combining planning support with a clearer policy signal, and potentially also location-specific discounts on network charges, some by recycling constraint costs. In principle, that should accelerate projects, although we’re waiting to see how those discounts will help in practice.

In practice, however, these zones still depend on network capacity. Zones are being announced without a clear, industry-wide view of how much demand the grid in those areas can realistically support; the fundamental issue of timely access to power remains.

None of these challenges is insurmountable, and our response with developers to date has been to offer alternatives to the traditional connection process rather than wait for reform to catch up.

One option is co-location. Private networks and microgrids can be constructed to pair new data centres with new generation behind a single meter point. This gives generators a strong route to market with a customer to sell directly to, and it gives data centre developers a connection that can be delivered in years rather than decades. The ‘privacy’ of the network is core, enabling the electrons (preferably green) to be kept within the localised grid. By effectively minimising the need to use the public grid, data centres can minimise the transmission costs added to their energy bills.

There’s an additional benefit to architecting private networks that make smaller demands on the grid. It’s often easier to secure smaller grid connections, in part because they may not require as much upstream reinforcement work – if any. Suitable connections may also already be available, particularly at brownfield sites. Alongside this, connections can be optimised to reduce their size and designed to ramp up over time to match real demand use, rather than requesting full capacity from day one. Most projects don’t use all of their demand from the start, so understanding what that ramp-up looks like and planning accordingly can help bring forward the initial energisation.

Demand connection and transmission reform on the horizon

Reform is on the horizon, and for transmission-level demand connections this will soon likely help. The Connect pillar of Ofgem and the Department for Energy Security and Net Zero (DESNZ)’s connections reform programme is likely to introduce a self-build, own or transfer option for demand customers. This will allow developers to build their own connection assets – with the support of their partners – rather than wait for the incumbent to do it for them, and to choose whether to hand that infrastructure over to a network operator once it’s built. They are also looking at alternative and flexible connection options that may help facilitate the ramp ups and flexible grid use discussed.

The reforms may also bring with them the introduction of independent transmission operators (ITOs). At the distribution level, independent distribution network operators (IDNOs) like Eclipse already connect 80% of new homes in England, proof that independents can take on network delivery at scale. Ofgem and DESNZ are now seriously considering extending that model to the transmission network, allowing licensed independents to step in and support major projects including data centres and their developers.

Breaking ground with compute

The UK’s AI ambitions won’t be realised by regulation alone. Reform is coming, but developers can’t afford to wait for it. The routes around the queue already exist: private networks, co-located generation, and connections built and owned by independents rather than left in the incumbent’s backlog. Get this right and the prize is enormous – investment, jobs and skills that will otherwise go elsewhere.