Barbour ABI counts 162 proposed UK data center schemes with an estimated build value of about £101bn. Twenty-one are actually on site, worth £4.9bn. Roughly £37.2bn already holds detailed or outline permission and still has not broken ground. That gap between consent and construction is the number worth planning around, because the headline figure is a planning total, not a workload.
Every few weeks a new pipeline number lands and somebody in a board meeting uses it to justify a hiring plan. The latest one is a good example of why that is risky. Analysis carried out by Barbour ABI for Construction News puts around £101bn of proposed UK data center build value across schemes at four stages of the planning process. It is an accurate number. It is also being read as something it is not.
Here is the arithmetic the press releases skip. Seventy data centers are complete, at a combined build value of £2.6bn. Twenty-one are currently on site, worth £4.9bn. Everything else, all 162 schemes and all £101bn of it, sits somewhere in planning. Some of those schemes have full consent. Some have not yet submitted an application. Treating the lot as a forward order book is how contractors end up carrying overhead against work that never arrives.
Which does not make the opportunity fake. UK data center construction is real, growing and badly under-resourced on the specialist side. The useful move is knowing which slice of £101bn is close enough to matter, and what is standing between the rest of it and a site start.
Where the £101bn actually sits
Four stages, and a very thin end
The pipeline breaks down into four distinct groups, and they are not equally real. Fifty-five schemes worth about £12.7bn hold detailed planning approval without having started on site. Another 22, worth £24.5bn, have outline approval only. Sixty-four projects valued at £40.6bn are in the system as submitted applications. A further 21, worth an estimated £23.2bn, sit at pre-application stage and may never file at all.
Add the first two groups and you get roughly £37.2bn with some form of permission in place. That is the tranche worth resourcing against, and even then outline approval is a long way from a start date. The £40.6bn in live applications is a watch list rather than a pipeline. The £23.2bn at pre-application is an intention, and intentions at this end of the market change with every shift in AI hardware specification.
- Detailed approval, not started. 55 schemes, roughly £12.7bn. The closest thing to real work in the numbers.
- Outline approval. 22 schemes, roughly £24.5bn. Consented in principle, still years from a slab.
- Submitted for planning. 64 schemes, roughly £40.6bn. The largest slice and the least certain.
- Pre-application. 21 schemes, roughly £23.2bn. Announced positions, not commitments.
- On site now. 21 schemes, £4.9bn. The whole of today’s actual workload.
Permission is not a start date
Why £37.2bn of consented work is sitting still
The standard explanation for a stalled data center is the grid, and the grid deserves most of the blame. But Barbour ABI’s own analysts are careful not to pin it on one thing, and that nuance matters if you are trying to work out which consented sites will move first.
Grid connection can take years to secure. Specialist labour, particularly high voltage electrical and commissioning staff, is scarce relative to demand. Transformers and cooling equipment carry lead times measured in years rather than months. And the pace of AI hardware development means a scheme’s power and cooling requirements can change materially between the planning submission and the construction start, sending developers back to redesign a building they already have permission for.
There is undoubtedly a hold-up in turning data centre plans into construction, but there is not one single cause.Steve Shelley, chief data analyst, Barbour ABI
That last cause is the one people underestimate. A consented shell designed around a 2024 rack density assumption may simply be the wrong building for a 2027 deployment. Permission does not expire when the technology moves, but its usefulness does.
The capacity picture tells a different story
Project counts are flat compared with the megawatts
Value is only one way to read a pipeline, and for anyone selling power infrastructure it is the wrong one. Separate Barbour ABI data analysed by Computer Weekly puts the committed UK pipeline, meaning schemes inside the planning system but not yet complete, at 173 projects and more than 14.6GW. Six months earlier the same measure showed 114 projects and around 10GW. A further 24 announced but unsubmitted schemes carry another 4.4GW behind them.
So the project count rose by around half while capacity rose by nearly the same proportion, and the geography shifted. London and the M4 corridor still hold 82 of the 173 committed projects and 5.2GW, but the typical project there runs at 77MW while the largest campuses are landing further north. That matters for anyone deciding where to base a commissioning team or an electrical prefabrication facility.
It also sharpens the grid problem. Roughly 19GW heading toward a network that Ofgem is currently trying to thin out through its commitment fee consultation is not a queue that resolves itself quietly. The two datasets are describing the same bottleneck from opposite ends.
How to read a pipeline number without getting burned
Four questions that separate workload from wishlist
None of this argues against building a data center capability. Annual UK spend is forecast to climb steeply through the rest of the decade, and if every scheme in the pipeline were delivered, Barbour ABI estimates around 9.3 million square meters of new space. The point is narrower. Match your commitment to the stage of the work, not the size of the headline.
Pipeline numbers make headlines. The arithmetic makes decisions.
What is actually getting built, who is winning it, and where the constraints really sit. One briefing a week.
Join 3000+ Readers
| Pipeline stage | Schemes | Estimated value | How to treat it |
|---|---|---|---|
| On site | 21 | £4.9bn | Live workload |
| Detailed approval, not started | 55 | £12.7bn | Resource against selectively |
| Outline approval | 22 | £24.5bn | Track, do not commit overhead |
| Submitted for planning | 64 | £40.6bn | Watch list |
| Pre-application | 21 | £23.2bn | Market intelligence only |
No, but it is widely misread. It is the combined estimated build value of 162 proposed schemes across four planning stages, from detailed approval down to pre-application. It does not represent awarded contracts, work ready to start, or a forward order book. Completed projects and the 21 schemes already on site sit outside that total. (Source)
Twenty-one data centers are on site with an estimated build value of £4.9bn. A further 70 schemes worth £2.6bn are already complete. So the total construction activity the pipeline has converted into so far, live and finished combined, is around £7.5bn against £101bn of proposals. (Source)
Barbour ABI points to several causes rather than one. Grid connections can take years to secure. Specialist labour is scarce. Transformers and cooling equipment carry long lead times. And rapid changes in AI computing requirements can alter power and cooling designs while a scheme is still moving through development, forcing developers back to redesign. (Source)
Barbour ABI data analysed by Computer Weekly shows 173 committed projects inside the planning system with combined capacity above 14.6GW, up from 114 projects and roughly 10GW six months earlier. Another 24 announced but unsubmitted schemes add about 4.4GW. London and the M4 corridor hold 82 projects and 5.2GW, though the typical project there runs at 77MW and the largest campuses are moving north. (Source)
If every scheme in the pipeline were delivered, Barbour ABI estimates around 9.3 million square meters of new UK data center space. Treat that as a ceiling rather than a forecast, since it assumes no cancellations across schemes that in many cases have not yet filed an application. (Source)
It could thin them considerably. Ofgem is consulting on a Data Centre Commitment Fee of £237,500 to £712,500 per megawatt for projects at or above 40MW, refunded on energisation and forfeited on early exit, alongside progression milestones. Schemes in the pipeline without funding or equipment orders behind them are exactly the ones those measures are designed to remove. (Source)
The £37.2bn holding detailed or outline permission is the realistic near-term tranche, and within that the 55 detailed-approval schemes are closest to a start. Test each one for a written energisation date, evidence of long-lead equipment orders, and a design vintage that matches current rack density assumptions. Schemes failing all three are not near-term work regardless of consent status. (Source)
Related Articles
Big Tech Promised $650 Billion in Data Centers. Most of It Isn’t Being Built
Nscale Has £2bn, Microsoft and Nvidia. It Still Can’t Get 90 Megawatts.
Power Is the New Lumber: The Supply Chain Bottleneck No One Saw Coming
Major Data Center CEO on Labour Demand, Power Scarcity, and the Real Work Behind AI Buildouts
Construction News: More than £100bn of data centre construction planned in UK
Construction Magazine UK: £100bn UK data centre pipeline faces power delivery test
Computer Weekly: UK committed datacentre pipeline grows to more than 14GW
Low Carbon Buildings: UK data centre pipeline approaches £100bn
Ofgem: Acting to free up grid capacity by tackling speculative data centre projects
