White paper: Ready to build, but waiting to connect

On 18 October 2026, various outlets reported that Microsoft had walked away as the anchor tenant for Nscale's planned £2bn AI data centre in Loughton, Essex. The site had planning permission, ministerial backing, a funded developer and a hyperscale customer. What it lacked was a grid connection.
Nscale announced the scheme in January 2025 at 50MW, scalable to 90MW. Microsoft was named as the anchor customer that autumn. But in September 2026, a Guardian report suggested that full grid power might not arrive until the early to mid 2030s. A few weeks later, reports began to emerge that Microsoft had pulled out.
It’s important to note that neither Nscale nor Microsoft has publicly confirmed the reason for the withdrawal, and Nscale says its relationship with the tech giant remains strong, it is committed to the site and is exploring alternative power options. However, whether or not the predicted grid delay was a factor in the breakdown, it clearly presents a roadblock to the programme.
It’s an important lesson for anyone building data centres in the UK. While land, planning permission, capital and components can be secured according to a commercial timetable, the same isn’t true of grid power. For these projects, everything hinges on the date the power can be switched on. But most programmes aren’t planned that way.
It’s not the cost, plan or patronage holding things up
The planning and approval process is often blamed when complex programmes hit delays, but in the case of the Loughton site, this wasn’t an issue. In fact, there was considerable pressure put on the local district council by government ministers keen to ensure that the project, announced by the UK Prime Minister during a state visit by the US President, would go ahead on schedule.
But ensuring on-time construction is just one piece of the puzzle. A building can be designed, consented and procured in two to three years, perhaps faster with the right political support. A new bulk supply point, on the other hand, requiring transformers, cable routes and upstream reinforcement can take far longer. And it seems that increasingly, prospective tenants aren’t prepared (indeed, can’t afford) to wait that long.
The queue is moving, but developers are impatient
Loughton is not unique: it’s just a current example of a system-wide constraint. In the UK, the waiting list for grid connectivity stretches a decade into the future in some cases, and Ofgem's figures report that data centre demand in the connection queue stands at around 80GW (by comparison, peak electricity demand this year in Britain was just over 47 GW, measured on 5 January)
The regulator is responding in three ways:
- Clearing speculative projects. NESO's connections reform, confirmed in December 2025, kept 283GW of generation and storage and 99GW of transmission-connected demand.
- Charging for commitment. Ofgem has proposed a fee of £237,500 to £712,500 per megawatt when a developer accepts a connection offer, refunded at energisation and forfeited if the project leaves the queue early.
- Testing readiness. Developers would have to show tangible progress, through financial capability, commercial maturity and procurement activity, to keep their place in the queue
But even with these measures in place, developers are looking for alternatives. Ofgem reports about 117 active cases and enquiries from data centres wanting gas network connections for on-site generation.
It’s understandable. Wait times of several years put developers and tenants in a difficult position: the technology these projects will support is evolving so rapidly that a data centre designed today could be obsolete when it finally comes online in five or six years. But solutions built on local power generation shift the engineering problem from the grid to the site, and onto the construction team.
How can projects secure (and hold) their place in the queue?
Most data centre programmes are built forwards from site start, with energisation pencilled in near the end. Loughton shows the logic has to run the other way: fix the realistic power date first, then sequence the build to meet it. That changes what the programme has to show.
- Phased energisation. If the network can offer 20MW now and 90MW later, the programme must show which halls, cooling plant and switchrooms are commissioned against each tranche, and which stay as shell.
- Temporary power giving way to permanent. Where gas engines, fuel cells or batteries bridge the gap, the sequence must show plant compounds, fuel routes and exclusion zones, and the later strip-out, without blocking the permanent substation.
- Interfaces and commissioning. The customer substation, the network operator's works and the cable route run on different contracts and different clocks. Integrated systems testing needs real load, so a programme that assumes full grid power at month 24 is unworkable if the supply arrives in year six.
This is where 4D planning earns its fee. A Gantt chart can show that energisation slips. A linked 4D model shows what that slip does to cranes, laydown, cable pulls, and commissioning, and lets the team test a phased or off-grid strategy before committing funds.
Proving forward progress to the regulator
If a developer must prove financial capability, commercial maturity and procurement activity to keep a queue position, the quality of its records decides whether it keeps a connection offer worth hundreds of millions.
That is an information management requirement in all but name. A project run to ISO 19650 already holds what a regulator is likely to ask for: tender returns, appointments and long-lead orders for transformers, switchgear and chillers in a common data environment, with dates and approval history; models and drawings at defined status codes that show the scheme is more than a concept on a land parcel; and a programme linked to the model and to the network operator's milestones, so progress can be shown, not asserted.
The developers who keep their place will be those who can produce that evidence on demand, not those who have to hastily rebuild it from inboxes and shared folders when the request comes in.
Designing the services for a phased build
A site that energises in tranches, or runs partly on its own generation, is a harder services problem than one that switches on in full. The coordination model must cover every interim state, not just the finished one. The points most often missed:
- Space for plant that arrives later. Plinths, cable routes and riser capacity for future transformers, generators or fuel cells need to be safeguarded in the federated model now, and clash-checked against structure and drainage.
- Cooling sized for the final load. AI halls on liquid cooling bring secondary fluid networks, coolant distribution units and heat rejection plant. Pipework, valve access and leak detection must be coordinated against cable containment from day one, even if the full load is years away.
- Retrofit around live equipment. If a hall goes live on partial power, later phases must be installable around secure, temperature-critical kit. That means sleeves, tie-ins and isolation valves modelled and verified before first fix. Switchgear and UPS designed for grid supply may also need reconfiguring for an on-site primary source, so the single-line diagrams and the 3D model must change together.
If the plan doesn’t take these critical elements into account, then there’s a risk that transitioning to the next stage will require downtime, and that’s something tenants are unlikely to accept.
A pragmatic approach to an evolving constraint
Ultimately, the solution to issues like the one that emerged at Loughton isn’t better planning, faster approvals or accelerated development, it’s more grid capacity. And that’s already underway. RIIO-T3 and the Accelerated Strategic Transmission Investment programme are funding the largest expansion of the country’s transmission system in decades. We’ve discussed the challenges and the digital engineering disciplines that will help solve them in our white paper, Building Britain's energy backbone.
But that doesn't help programmes underway now, or in the immediate future. Those programmes need to navigate the connection queue as it stands, and meet Ofgem’s new requirements to join and keep moving forward. Digital engineering helps here too: robust information management, comprehensive 4D visualisation and model coordination that cover all phases of a build, including interim states.
If your connection date is still uncertain, the first useful step is to see how each power scenario affects the programme before you commit to one. Talk to Majenta to build phased and off-grid scenarios into a linked 4D and coordination model, so you can visualise the sequence, not assume it.
Key contact
Priya Flora, Partnership Manager, Majenta Solutions
Sources:
- Nscale announces UK AI infrastructure commitment in partnership with Microsoft, NVIDIA and OpenAI
- Launch of UK’s ‘largest AI supercomputer’ delayed by power supply problems | AI (artificial intelligence) | The Guardian
- Microsoft pulls out of UK data centre deal with Nscale, Telegraph reports
- Microsoft quits Nscale UK data centre deal ahead of planned $3 billion IPO
- UK government reportedly urged Nscale AI data centre approval - Capacity
- Ofgem acts to free up grid capacity by tackling speculative data centre projects
- Energy Insights | Grid Reform | Philippa Orr | Galbraith
- Ofgem proposes commitment fee to tackle speculative data center projects in UK grid connection queue
- The New Race For Power: What Ofgem’s Grid Reform Means For Data Centre Development In Great Britain - Utilities - United Kingdom
- RIIO-3 Final Determinations for the Electricity Transmission, Gas Distribution and Gas Transmission sectors | Ofgem
- Great Britain gains edge in race to secure sought-after components as Ofgem greenlight early construction funding for key energy projects

