The Choreography Crisis: You ordered your transformer too late. You just haven’t found out yet.
November 2026
TL;DR
Every energised site runs on three clocks: the developer's stage gates, the factory's slot windows and, if Ofgem's proposals go through, the regulator's milestones. Each one is run well by the people who own it. Nobody owns the place where the three meet, and that's where energisation dates go missing.
Start at the end.
Here's a data centre. It's illustrative, a composite of the projects we see rather than any single one: 100 MW of IT load in Great Britain, with a grid connection date of June 2030.
The usual way to plan a project like this runs forwards. Find the land, win planning, accept the connection offer, design it, raise the money, appoint a contractor, build, switch on. We're going to run it backwards from June 2030, because the backwards version is the one that tells you when the order needs to be placed.
On the day of energisation, the transformer has to be on site, installed and tested. Before that, it has to travel from the factory. Before that, it has to pass its factory acceptance test. Before that, it has to be built, from drawings the buyer has approved. And before anyone can approve a drawing, someone has to decide what voltage levels the transformer needs to manage. To do this, a connection date and placement need to be offered by the grid operator.
Each of those steps sits with a different organisation, keeping a different timetable. We count three. Call them clocks. Each keeps good time on its own. The trouble starts when you hang them on the same wall.
Clock one: the developer's stage gates.
The developer's clock runs on gates. Secure the land. Win planning consent. Accept a connection offer. Design the site. Reach Final Investment Decision (FID). Appoint the engineering, procurement and construction (EPC) contractor. The order shifts from project to project, and the sequence here is typical rather than universal.
This clock has a peculiar property: it doesn't really tick. Each gate opens when its conditions are met, and a gate that takes longer simply takes longer. Planning can run six months over and the next gate waits patiently.
For the electrical equipment, two moments on this clock matter more than the rest.
The first is the specification freeze, when the design is settled enough to tell a factory what to build: the rating, the voltages, the cooling, the tap changer, how many units. Before that point, any order is an order for an educated estimation.
The second is the commitment decision, when someone is prepared to sign a contract and pay a deposit. For most developers that means FID, and it's the gate most procurement plans are built around, for good reason. Before FID the money isn't raised, the project might not happen and the design is still moving. Signing for a transformer then means spending money you don't have yet on a unit you haven't finished specifying.
Clock one's instruction is perfectly sensible: reach FID, finalise and freeze the design, then order. Hold that thought.
Clock two: the factory's slot windows.
A factory that builds large power transformers doesn't keep them on a shelf. Each one is built to order for a particular site, from drawings the buyer has approved.
What the factory holds is a calendar of production capacity, and what a buyer is really purchasing is a position in it: a slot.
How long until your slot comes round depends on whose calendar you're reading. For incumbent OEMs, lead times run to 4+ years (Wood Mackenzie, Q2 2025). We asked a panel of twenty manufacturers to quote for the same two specifications. Nine of twenty gave us a lead time, and those nine ranged from 14 weeks to 20 months (Opna supplier panel, 2026). Both sets of numbers are real. They come from different factories with different order books, and they don't all measure the same stretch of time.
That last point needs a moment. A quoted lead time answers two questions it rarely states: from when, and to where?
Three of our twenty start the clock at drawing approval. Drawing approval comes after the contract is signed: the factory draws the unit, the buyer's engineers review it, comments go back and forth, and only then does the factory's clock start. Eighteen months from drawing approval is longer than eighteen months from signature, by however long the back-and-forth takes.
The other end stretches too. One supplier on our panel put a long-lead unit at around 16 months to factory acceptance testing, then two to three months in transit to Europe. That gets the transformer to site. Installation, testing and energisation are still to come.
For a procurement team, that makes any lead time quoted before the design is frozen an estimate. The manufacturers on our panel who gave us a date attached conditions to it, most often about which tap changer or bushing brand the buyer would accept, and those choices are only settled when the specification is. A lead time becomes something you can plan around once two things are true: you hold a confirmed slot, and the manufacturer has set out what the time covers, from drawing approval through factory testing to delivery.
Clock two ticks whether you're ready or not. A slot is a date in somebody else's diary.
Clock three: the regulator's milestones.
The third clock is the newest, and for now it's a proposal. In its Curate consultation of 29 July 2026, Ofgem set out milestones that data centre projects in Great Britain would have to meet to keep their place in the connection queue.
The consultation closed on 16 September and Ofgem has said it will decide later in 2026, so the detail could still move.
The reason for it is easy to see. Ofgem counts around 73 GW of data centre demand in the queue, across about 315 projects, against a GB peak demand of 45 GW (Curate, para 2.8). It wants a way to recognise the projects that will be built from the ones holding an option.
Three of the proposed milestones matter here:
M2.Dc: evidence of a committed order for long-lead electrical equipment, such as transformers and switchgear. Ofgem lists a purchase invoice, an executed procurement contract or a binding supplier commitment, naming the equipment, its voltage and rating, the supplier, the commitment date and the expected delivery date (paras 6.43 to 6.49).
M5: submission of the design for the contestable works, the parts of the connection a developer can have built by someone other than the network operator.
M7: project commitment.
Each falls a set number of months before the connection date, in four bands (Curate, Appendix 1). M2.Dc lands at 12, 18, 24 or 30 months. M5 at 12, 15, 18 or 21. M7 at 6, 9, 12 or 15.
Read across those numbers and something odd appears. In three of the four bands, Ofgem would want evidence of a committed equipment order before you've submitted your design. In every band, it wants that evidence 6 to 15 months before project commitment.
The draft code text adds a consequence. If the order is cancelled, the developer would have 20 business days to say so and six months to provide replacement evidence, or the project is categorised for termination (draft CUSC paras 16.4.9.1A.1 and 16.4.9.2).
The first piece in this series did the arithmetic against a 4+ year lead time. The date you'd have to prove you ordered by lands 18 to 36 months after the date you needed to order.
Three clocks, one axis.
Put all three clocks on one axis, counting down to June 2030, and three collisions appear.
A. The order date arrives before the design does. On an incumbent lead time of 4+ years, the latest order date for a June 2030 connection is mid-2026. Design submission at M5 comes more than two years after that, whichever band the project sits in. The factory needs to know what it's building long before the developer has finished deciding.
B. The regulator asks for proof before the money is committed. In most bands, M2.Dc falls before the point most developers would call FID. Clock three wants evidence of an order that the developers clock says not to place yet.
C. There's a line after which the order can't move. Somewhere on the factory's lane, the manufacturer starts spending money on parts made for your unit. Before that line, the order bends. After it, it doesn't.
In the second frame, the developer's gates all shift six months to the right. The regulator's milestones stay exactly where they were, because they're pinned to the connection date. And the factory's slot, being a date in somebody else's diary, doesn't wait.
Where interchangeability ends.
Early on, a transformer order is surprisingly forgiving. The specification can still change. Slots can, very occasionally, be swapped. Walking away costs relatively little, because the factory hasn't spent much on you yet.
That changes at a specific moment: when the factory commits to the long-lead parts for your unit. Core steel, copper, the tap changer, the bushings, the tank. From then on, materials are being bought and shaped for one transformer, for one site.
Tailors know this moment well. Change your mind before the cloth is cut and they shrug. Change it after, and you've paid for a suit that fits nobody else.
The manufacturers on our panel who shared cancellation terms describe a ladder, with exit costs stepping up as the unit moves from design, to materials, to build. The rungs are set by what the factory has done, not by dates on a calendar.
Leaving is cheapest before the drawings are approved. Once the long-lead parts are committed, it costs most of the contract value. Once the core and coil are being assembled, it costs all of it. One supplier on our panel doesn't allow cancellation at all.
This is the moment the chapter is named for. Before it, your transformer is one of many possible transformers. After it, it's yours: your voltages, your rating, your site on the drawings, whether or not your project still needs it on that date.
Notice where Ofgem's proposed evidence sits. M2.Dc asks for "a genuine and material commitment" (para 6.43). The more material the commitment, the closer it sits to this line.
How one slip travels.
Give our June 2030 project a perfectly ordinary setback: six months lost in planning or design. Nobody has done anything wrong. Assume the developer did what Ofgem proposes and committed to equipment early. Here's the route the delay takes.
The design changes, so drawings that were approved have to be revised and resubmitted.
Drawing approval moves. For factories that start their clock there, the build date moves with it.
The slot window passes. A factory can't hold a production line idle for a customer who isn't ready, so the slot goes to the next order in the book.
The next free slot is wherever the calendar has a gap. On a full order book, that can be many months away.
If the order has to be restructured to fit a new slot or a revised design, the cancellation ladder applies, at whichever rung the unit has reached.
If the original order is cancelled, the evidence lodged for M2.Dc may no longer satisfy the milestone.
Under the draft code, the developer then has 20 business days to notify and six months to replace the evidence, or the project is categorised for termination.
Through all of this, the regulator's milestones haven't moved an inch.
The costs that follow are familiar to anyone who has watched a project slip. The equipment gets repriced. The finance gets reworked, because the lenders' model was built on the old date. Consents may need revisiting if the design change is big enough. Under Curate there's a fourth cost, and it's the hardest one to buy back: the project's place in the queue.
How far each step bites varies by region and network operator.
Why nobody sees it coming.
Each party reads its own clock well. Developers know their stage gates.
Manufacturers know their order books. The system operator knows its milestones.
When a factory asks for approved drawings and a deposit before it commits a slot, it's doing what a factory with a full order book should do. When a developer waits for FID before ordering, it's doing what anyone spending other people's money should do. Each decision is sound on the information in front of it.
What nobody holds is all three clocks at once. Energy UK's July 2026 report, "Industrial Strategy: Where next?", describes procurement across the sector as "inconsistent, stop-start" and short of long-term volume visibility. Factories can only plan for demand they can see. Developers can only commit to slots they can see. At the moment, neither can see much of the other.
The bill arrives early.
The collision has a price, and someone starts paying it long before the first transformer ships. Next in the series, we follow the money: the commitment fee Ofgem has proposed, the payments factories ask for before a unit reaches site, and what happens to all of it when the date moves.
If you'd like to see your own project's three clocks on one page, talk to us.
We'll start from your energisation date and work backwards.
Keep reading: