Do data centres clog the grid connection queue?
What are data centres?
Data centres are specialised and secure facilities that house and operate IT infrastructure used to store, process, and distribute data. If a service uses the internet, it is almost certainly running a data centre somewhere. Together with communications networks, they form our digital infrastructure.
Data centres underpin all digital activity. This covers everyday uses, such as backing up photos on a phone, as well as complex tasks like processing large datasets for weather forecasting and scientific research.
There are several types of data centre:
Enterprise data centres are privately owned and run by a single organisation for its own IT needs, while colocation facilities (the dominant model in the UK and Europe) let an operator provide space, power and cooling to multiple customers who bring their own servers.
Hyperscale data centres are much bigger sites run by major tech firms like AWS, Microsoft and Google to power cloud and global digital services, alongside an emerging category of neocloud providers building GPU-based platforms specifically for AI training and inference.
Finally smaller edge data centres are sited close to end users to minimise latency – the time it takes for data to travel between a user and a data centre – for things like gaming or real-time systems.
Why is there demand for data centres?
Attention is turning to data centres amid concerns over the impact substantial new demand connecting to the grid could have on our energy system. However, before addressing that impact, it is worth understanding where this demand is actually coming from and what is driving it.
Data centres are fundamental to delivering the everyday services we rely on – from public services, scientific research and banking, through to email, web hosting and media streaming. The public sector, academic institutions, finance, and healthcare are among the biggest demand drivers for new data centres. Despite AI being the main driver of new demand for data centres, AI workloads do not constitute the majority of existing capacity: according to DCByte's 2025 Global Data Centre Index, AI accounts for just 11% of global data centre use, while public cloud services make up the largest share at 52%.
Within AI there are inference and training workloads. AI training – the computational process of teaching an AI model to recognise patterns and logic within data – is more energy intensive and can involve unexpected spikes in demand. However, unless our energy prices coming down, training is unlikely to be large scale in the UK. Instead, our demand will more likely be for AI inference, which includes chatbots, voice assistants, and autonomous systems, which require real-time responses.
Government policy is amplifying demand too, by aiming to increase the UK's public sector (sovereign) compute capacity through initiatives such as the AI Opportunities Action Plan and AI Growth Zones (AIGZ). The last administration had a target to deploy at least 6GW of AI-capable data centre capacity by 2030. The previous Chancellor also set out ambitions for the UK to achieve the fastest AI adoption in the G7.
On top of this genuine demand, there is a speculative element too. Developers and landowners are betting on sites, planning consent, and grid capacity in the hope of future returns, rather than in response to confirmed end users. This speculative behaviour does not just shape the property market and planning landscape, but has also contributed to the demand connection queue ballooning from 41GW in 2024 to 125GW in 2025. As a result, viable, well-funded projects are being held back by applications that will never be built.
To note: according to NESO and DC Byte, there is currently around 1.8 to 2.4 GW of data centre capacity in the UK currently.
The demand connections queue
Securing an electricity grid connection is now the single largest priority and challenge for data centre development. Whilst the boom in AI and the surge in data centre interest is one driver of the rise in demand sitting in the queue, the way the grid application process itself works has compounded the problem.
Data centres receive power like any other large energy user, by applying for a grid connection. Reforms are currently underway that will fundamentally change this process, but historically offers have been issued on a first-come, first-served basis: whoever submits an application first gets served first, regardless of whether their project will actually come to fruition. Since April 2025, that has changed for generators. Now, new battery, wind and solar projects, for example, have to prove not just that they're ready to build, but that they are needed to meet the UK's clean power goals. This has yet to be applied to demand, which has allowed speculative volume to build up unchecked.
On top of this, there is poor and limited early-stage visibility of where capacity actually exists on the grid – so businesses cannot see where the grid has capacity before they commit. As a result, developers will apply simply to find out if there is capacity, which lengthens the queue for everyone. They will also spend months submitting applications — checking network heatmaps, hiring grid consultants, or simply filing multiple grid applications at different sites, paying thousands of pounds and waiting to see which one sticks. Access to the data that would fix this is taking years.
As the queue has grown, so has the wait: some developers face delays of 8-10 years, with a survey finding 80% face major delays and some connection dates pushed from 2027 to 2038. These delays hold back viable projects that could support the UK's economic goals, by threatening investment timelines, inflating costs, eroding returns, and prompting some institutional investors to reconsider their exposure.
q
Growing demand for data centres and their services translates into more applications in the grid connection queue. But the scale of what is currently sitting in that queue is striking, and not entirely representative of real need.
According to pipeline data collected in 2025 by National Energy System Operator (NESO) from distribution and transmission network operators, there is 72.8GW of data centre connection requests in the system through to 2039. To put that in context, the UK's entire winter peak electricity demand was 47.4GW in 2024. techUK believes there is a significant mismatch between the 72.8GW figure cited by NESO and the real volume of viable data centre demand in the queue.
Industry estimates put viable projects at only 8-15GW. DC Byte has identified 14.7GW of viable data centre projects, of which 0.3GW is under construction, 2GW is committed, and the remaining 12.4GW is early-stage capacity that remains subject to factors such as power availability, planning approval, and phased delivery. Notably, they found just 10 projects account for over 50% of that early-stage figure, making up 6.9GW.
The over-inflated figure is likely the result of A) duplicated applications (the same developer applying at multiple sites for one project), B) speculative "capacity squatting" in the hope that a project might eventually be built, and C) schemes with no realistic prospect of being built.
NESO itself acknowledges this. Its 2026 ten-year outlook cites stakeholder feedback that close to 50% of pipeline capacity may come from “highly speculative” or duplicative applications, and finds that the pipeline is likely to “significantly overstate” real demand. Despite the queue growing, NESO’s latest forecast for data centre demand is lower than its previous one. Its 2025 forecast expected just over 20TWh of data centre demand by 2030; the 2026 outlook now projects 14.6TWh by 2031, reflecting new data on utilisation and project attrition.
NESO notes it is unable to state what the real demand is, due to uncertainty around project attrition (the proportion of pipeline projects that drop out and never reach operation), utilisation rates (the proportion of a data centre's connected capacity that is actually used over time), and ramp-up rates (how quickly demand increases over time once a data centre is connected).
What needs to be done
Continue the good work on demand connections reform, to allow projects to connect on time, investment can flow and ensure the UK can build the digital infrastructure its economy needs. While techUK is supportive of a commitment fee in principle, the extreme level being proposed does not risks ruling out viable projects that may not be able to secure such a high upfront fee.
To help achieve this, techUK is working with Government, NESO and Ofgem on reforming how projects connect to the grid, to weed out speculative applications from the outset.
This includes introducing a strong yet proportionate financial mechanism – either a deposit that is refunded upon energisation or a fee that is paid as the project progresses, to show commitment.
There should be more measurable readiness requirements, for both entering the grid queue and for different connection queue milestones, such as a creditworthiness assessment and a “look back” test on past behaviour, to check for previous non-delivery, holding capacity without progressing, and repeated failure to meet milestones.
NESO should ground its projections in viable, actual demand data, so as not to over inflate what the network needs to support; costs which, ultimately, are paid for by the consumer.
DNOs and TOs must increase transparency over the grid connection queue, as set out in Ofgem's End-to-End Review. Data on order, connection dates, and connection points should be published for both transmission and distribution queues, regardless of whether connections are for generation or demand.
Ofgem and network operators are currently developing a Demand Capacity Register (DCR) to increase the visibility of connection queues and grid capacities. It aims to accelerate grid connections by categorising separate transmission and distribution asset availabilities.
