Prefer to listen?
This article is also available as a podcast episode
The UK data centre pipeline is approaching £100 billion in combined project value, covering around 9.3 million square metres of potential new space. Annual construction spend is projected to rise from roughly £1.75bn in 2024 to £10bn a year by 2029.
Every trade publication in the country has written about it. Very few have said anything useful to a firm of under fifty people, because the honest answer to “can we get data centre work” is: not the main contract, no.
But there is a layer underneath that is genuinely accessible, and there is a structural reason it is valuable.
The scale, briefly
477 data centres operating. 70 complete at £2.6bn. 21 under construction at £4.9bn. In the pipeline: 55 schemes with detailed consent worth £12.7bn, 22 at outline worth £24.5bn, 64 applications submitted worth £40.6bn, and 21 at pre-application worth £23.2bn.
Five AI Growth Zones were designated following the AI Opportunities Action Plan:
- Culham, Oxfordshire — £100m, 100MW scalable to 500MW
- North East, Cobalt Park / Cambois — £10bn, 720MW, ten buildings totalling 540,000 m², at outline
- North Wales, Trawsfynydd — £1bn, 238,000 m² plus office and R&D, outline granted
- South Wales, M4 corridor — Bro Tathan £285m/148MW approved; Microsoft Imperial Park £400m completing January 2027; Carbon3.AI Newport starting December 2026
- Scotland, North Lanarkshire — DataVita DV3, £320m, detailed approval
Named elsewhere: a Blackstone £13bn hyperscale facility in the North East, a 280,000 m² scheme near Auchtertool in Fife, and a £202m National Wealth Fund guarantee for a Lanarkshire facility.
The bottleneck is power, not planning
This is the fact that makes the whole sector behave differently from ordinary construction, and it is the one that creates the opportunity.
Around 140 proposed data centres are seeking 50GW of grid connections, and some projects face waits of up to fifteen years. Grid connection reform is moving — the National Energy System Operator reported in August 2026 that connection offers now cover more than 60% of the projects needed for 2030 ambitions — but a connection offer is not capacity. Projects still need finance, planning, equipment procurement, construction and energisation.
The consequence for delivery, in the words of Barbour ABI’s chief data analyst: power and cooling requirements can change significantly between planning and construction.
Sit with that for a moment, because it is the commercial argument.
Why change-heavy projects need design managers
A building whose fundamental servicing specification changes between consent and construction is a building where:
- The MEP design is revised repeatedly, at extreme services density, across multiple specialist packages
- Clash detection is not a one-off exercise but a continuous discipline
- Model federation has to hold together across a large number of specialist contributors, many of whom are working to different information standards
- The change register is the single most important document on the project
- Revision control failure is not an inconvenience; it means something gets built to a superseded specification
That is the work. It is not the main contract, and it is not glamorous. It is exactly what a competent design and information manager does, and it is in short supply because the volume of this work has arrived faster than the capability to service it.
What the accessible layer actually looks like
Four realistic entry points for a small consultancy or specialist practice:
MEP coordination at density. Data centre plant rooms and containment runs are among the most services-dense environments in construction. Coordination is a specialist discipline, and tier one contractors buy it in.
Clash and model management. Federating models from a large specialist supply chain, running the clash process, and holding the resolution record. Platform-agnostic capability matters here — the client’s CDE is whatever the client’s CDE is.
Change control and revision discipline. Given the specification volatility, the party that keeps a clean, queryable record of what changed, when, why and what it affected is doing something structurally valuable rather than administratively tidy.
Asset information and handover. Data centre operators are sophisticated buyers of asset data, and the handover requirement is usually more demanding than a typical commercial building. Structured asset information — IFC 4.3, information delivery specifications, a properly negotiated exchange requirement rather than a boilerplate COBie demand — is a genuine differentiator.
Plus the enabling works and civils packages, which are conventional construction and are procured conventionally.
How to actually get in
Be realistic about the route. The clients are hyperscalers and specialist developers; the main contractors are a small set of tier ones; the MEP is a smaller set of specialists. You are selling to the second and third of those, not the first.
Which means:
- Framework and supply chain registration with the tier ones active in the sector, done before there is an opportunity rather than in response to one.
- Demonstrable ISO 19650 capability, because these are information-heavy projects with sophisticated clients. A competence framework you can show, named individuals mapped to standards, not a capability statement.
- Security clearance and ISO 19650-5 literacy. Data centres are critical infrastructure and increasingly defence-adjacent. A properly conducted sensitivity assessment is not a formality here, and being able to run one is a differentiator. Note also that ransomware is now construction’s leading cyber threat, with an average of 24 days’ downtime per incident.
- Geography. The work clusters — the North East, South Wales, the M4 corridor, North Lanarkshire, Oxfordshire. Proximity matters more than it does on paper.
A note on discipline
One habit worth applying to every announcement in this sector, and to every other pipeline story you read: a spending target is weaker than a funded programme; a funded programme is weaker than a procurement notice; and a framework ceiling is permission to spend, not committed spend.
A £100bn pipeline that includes £23bn at pre-application and £40bn in submitted applications is not £100bn of work. It is a pipeline with a grid-connection queue in front of it and a fifteen-year worst case on some connections. Position for it, do not bet the business on it.
Where else the growth is
For balance, and because data centres are not the only positive story: surveyed workloads in Q2 2026 put energy at +39%, water and sewage at +23% and communications at +22%, against private housing at -12%. Education and offices were the only two sectors positive on construction starts. The two-speed market is real, and data centres sit on the fast side of it along with everything else that is publicly funded, regulated-asset-funded or infrastructure-adjacent.
If your practice has the technical capability for services-dense coordination but not the framework readiness or the information management evidence to be credible with a tier one, that gap is closable in a quarter.