UK Grid Power Limits Reshape Data Centre Siting in 2026

Why the grid is now a siting decision, not just a utility bill

For two decades, London and the M4 corridor were the default answer to 'where should we put our infrastructure?' — proximity to customers, fibre density and an established ecosystem did the work. That logic is breaking down. Ofgem data cited by Computer Weekly shows connection offers across the network rose from 41GW to 125GW between November 2024 and June 2025, with data centres accounting for at least 80GW of that demand. Wait times for a grid connection have stretched to as much as 15 years in parts of the network, which is an impossible timeline against an 18-month build cycle for a facility.

That mismatch means the question buyers now need answered before anything else is not 'what does this site cost?' but 'when can it actually be energised?'. Our own look at the UK data centre power and energy landscape points the same way: power, not land or fibre, is becoming the binding constraint on expansion.

The £1bn blueprint: what a grid upgrade actually involves

National Grid's North West London Upgrade, announced in August, is a useful reference point for how much physical work sits behind a single regional capacity unlock. The £1bn project will connect five new data centres across Bedfordshire, Hertfordshire and London, delivering 1GW of demand. Delivering that requires roughly 60km of new high-voltage cable, 35km of re-cabling between Sundon and Elstree, upgrades to more than 200km of overhead lines, an extension to the existing 400kV substation at Elstree, and a new 132kV substation at Letchmore Heath.

That scale of build is not a one-off. Ofgem's RIIO-3 price control allocates up to £10.7bn for electricity transmission between 2026 and 2031, with flexible mechanisms potentially pushing total investment to as much as £90bn as projects mature. Ember's Elisabeth Cremona frames the underlying problem starkly: a single data centre site can represent the load of 500,000 houses arriving almost overnight, whereas grid build-out traditionally takes five to 15 years to catch up.

Queue reform and the ‘Pay to Stay’ fee

Part of the backlog is self-inflicted. Developers, wary of how slow the process is, have submitted multiple speculative applications for the same load to see which clears first — creating so-called zombie applications that clog the pipeline. Ofgem has proposed reforms to strip out non-viable projects: large data centre developers would pay an upfront commitment fee of between £237,500 and £712,500 per megawatt when accepting a connection offer, refundable at energisation but forfeited if they drop out early.

Onnec's Matt Salter calls the move a sign of how tight the market has become, and warns it adds another pressure point for operators already being pushed to build faster. Onnec research found 92% of operators are compressing build timelines to keep up with AI demand, and 75% say speed-to-market pressure is forcing design decisions before infrastructure requirements are fully understood — a dynamic that risks bolting on grid capacity without coordinating it with cooling, cabling and compute, as explored in our note on AI power consumption as a bottleneck.

Follow the power: the geographic shift already underway

Barbour ABI data analysed for the Computer Weekly feature shows that of the UK's top 10 data centre projects, only one — East Havering in Essex — sits in the London area. The rest, totalling just under 4GW, are spread from Oxfordshire to Lincolnshire, North Wales, the north-east and Scotland. Kao Data's £450m investment in Greater Manchester is a live example: CEO Spencer Lamb says grid availability is now one of the most important factors in deciding when and where a facility gets built, because without power within a commercially viable timeframe, no other site advantage matters.

Lamb also cautions against treating every upgrade as purely AI-driven — some substation work has been in planning for years against cloud and enterprise demand — but agrees AI has sharply accelerated the pressure. Denmark offers a cautionary parallel: its grid operator suspended new connection requests in March after receiving around 60GW of applications against peak national demand of roughly 7GW, showing this isn't a uniquely British problem.

What this means for UK buyers planning 2026–27 capacity

The practical takeaway is that site selection now needs a power-first checklist, run earlier in the planning cycle than teams are used to. Before signing a lease or committing capital, buyers should confirm grid connection status (not just offered capacity), understand regional queue conditions, and stress-test whether a provider's energisation date is realistic given the five-to-15-year upgrade cycle described by Ember. Our guide to choosing a UK colocation data centre is worth revisiting with grid timelines specifically in mind, and our wider piece on UK data centre grid queue and power shortage conditions sets out the regional picture in more detail.

There's also a cost dimension. Ofgem's reforms mean developers face new upfront commitment fees per megawatt, costs that will inevitably work their way into colocation pricing and contract terms. Buyers evaluating total cost of ownership should factor this alongside server running costs due to electricity prices, and use tools like a server power consumption calculator or rack space and power density planner to model how density assumptions shift if a workload moves to a region with different power economics.

• •Verify actual grid connection status and energisation date, not just a quoted MW offer
• •Treat northern and regional sites as genuine contenders where power is available sooner
• •Model how Ofgem's commitment fees could feed through to colocation and build pricing
• •Revisit on-prem, colo or cloud strategy if preferred regions face multi-year queues

Weighing on-prem, colo and cloud against grid reality

Grid constraints are also a reason to re-run the on-prem-versus-colocation-versus-cloud calculation rather than assume last year's answer still holds. A facility planned for a congested corridor may need re-scoping, while cloud regions and colocation providers with secured power in less contested areas could now offer faster, lower-risk paths to capacity. Our framework on on-premise, colocation, or cloud compute strategy is a useful starting point for teams revisiting siting decisions for 2026–27 given how fast grid conditions are moving.