Key takeaways
- The Question: How long does a data centre take to build, what controls that timeline, and why do schedules slip when owners build at scale?
- The Answer: A large data centre building typically needs around one to two years of construction after a power and consent stage that can be just as long. The critical path runs through power and cooling equipment, not the frame. Schedules become predictable when each hall is planned from a measured template, long-lead equipment is tracked to factory test and planned output is checked against what crews have actually delivered.
- The YelloSKYE Angle: Construct IQ gives owners one dated record of every hall across every contractor, from drone, fixed-camera and 360° capture, so achieved production can be measured rather than reported.
A data centre construction schedule usually spans one to two years from groundbreaking to the first hall entering service, preceded by a power and permitting stage that can last a year or longer, with further halls handed over in phases every few months after that. Unlike most buildings, its critical path runs through electrical and cooling equipment: the grid connection, transformers, switchgear, generators and the permanent power needed to commission them. Owners who build many facilities make those timelines predictable by planning from proven templates and by measuring real production on site, which is where Construct IQ contributes.
This guide focuses on the timeline itself. The stages of a build and what happens in each are described in our guide to the data centre construction process, and the measurement routine for a live build is set out in data centre construction progress tracking.
A typical timeline by phase
The durations below are indicative planning ranges for a single large building with several data halls. They are not research figures, and your own will depend on capacity, design maturity, the local grid, lead times and prefabrication.
| Phase | Indicative duration | Overlaps with | What usually drives it |
|---|---|---|---|
| Power agreement, site and consents | 6 to 24 months or more | Design and early procurement | Grid capacity, utility programme, permitting process |
| Design development | 4 to 9 months | Procurement and enabling works | Tenant requirements, cooling strategy, rack density |
| Long-lead equipment procurement | Orders placed early; deliveries timed to fit-out | Everything from design to commissioning | Manufacturing slots and factory testing |
| Enabling works and groundworks | 2 to 4 months | Design completion | Ground conditions, platform size, drainage |
| Structure and envelope | 4 to 8 months | Fit-out of the first electrical rooms | Steel or precast supply, roof plant loads |
| MEP fit-out of the first hall | 5 to 10 months | Envelope and later halls | Plant deliveries, skilled labour, energisation date |
| Commissioning of the first hall | 2 to 4 months | Fit-out of following halls | Permanent power, defects found in testing |
| Each following hall | One every 2 to 4 months | Live operation of earlier halls | A repeatable sequence and crew continuity |
Two features stand out. First, almost every phase overlaps the next, which is how data centres are delivered fast and also why a slip in one phase spreads into several others. Second, the stage before construction is often as long as construction itself, and its outcome, the date permanent power is available, sets a hard limit on everything after it.
Why power and cooling set the critical path
On an office or residential building, the structure and envelope usually drive the programme. On a data centre, the building is closer to a protective shell around a power and cooling system, and that system has its own chain of dependencies.
| Link in the chain | What it needs before it can proceed | Why it is critical |
|---|---|---|
| Utility connection and substation | Grid capacity agreement, utility works, high-voltage equipment | Nothing can be tested at full load without it |
| Transformers | Plinths, cable routes, factory testing, delivery | Long manufacturing queues and heavy lifts |
| Medium-voltage switchgear | A dry, clean electrical room | Cannot be set in an open or dusty building |
| UPS and low-voltage distribution | Switchgear in place, rooms conditioned | Feeds every hall; installed and tested in sequence |
| Backup generators and fuel systems | Yard, plinths, controls, emissions approvals | Required for resilience testing |
| Cooling plant and distribution | Power to run it, controls, pipework or coolant systems | Halls cannot carry heat load until it works |
| Integrated testing of each hall | All of the above working together | The final gate before the hall goes live |
The structural milestone that matters most is getting the electrical rooms weathertight and clean, because switchgear and UPS equipment cannot be installed until they are. A late roof over the electrical rooms delays every link in the chain, while a late roof over a later hall may not delay anything yet.
Commissioning deserves its own line on the programme rather than being treated as the tail end of fit-out. It depends on permanent power, often needs temporary load banks to simulate IT load, and is where installation problems are found. If the schedule leaves only the time remaining after fit-out, commissioning absorbs every earlier delay.
Tracking long-lead equipment as schedule activities
Many programmes show long-lead equipment as a single "delivery" milestone. That hides most of the risk. A better approach is to schedule each major item as a short chain of activities, each with a date that someone can confirm.
| Milestone | What it confirms | Typical evidence |
|---|---|---|
| Manufacturing slot reserved | The item is in the supplier's production plan | Supplier confirmation |
| Submittal drawings approved | Design and supplier agree the item | Approved drawing register |
| Factory acceptance test | The item works before it ships | Witnessed test report |
| Shipped | The item is on its way | Shipping notice |
| Delivered to site | The item has arrived undamaged | Delivery inspection record, dated photographs |
| Set in place | The item is on its plinth or in its room | Dated site imagery |
| Energised | The item is connected and live | Energisation record |
Linking these milestones to the installation activities that depend on them turns procurement from a separate spreadsheet into part of the critical path calculation. When a factory test date slips, the programme then shows immediately which hall's ready-for-service date moves, and by how much.
Planning phased hall handover
Most large facilities are not handed over all at once. The first hall, or a block of halls with its supporting plant, goes live as early as possible so the owner or tenant can begin installing IT equipment, and the rest follow on a planned rhythm.
Plan every hall from one template
Halls in the same building are usually close copies of each other. The schedule should reflect that: one hall sequence, built and agreed in detail, then repeated for each hall with an offset. This is essentially line of balance thinking, where each trade moves from hall to hall at a steady rate and the offset between halls is set by the slowest trade. It makes the plan easier to read and, more importantly, makes each hall's performance comparable with the last.
Separate "ready for equipment" from "ready for service"
Two hand-over points are worth showing separately. Ready for equipment is when a hall is clean, secure and safe enough for the tenant's equipment to arrive. Ready for service is when it has passed integrated testing and can carry live load. Owners often need the first date for their own fit-out programme, and contractors are measured on the second.
Build the live side into the logic
Once the first hall is live, the rest of the work happens next to a secure, operating facility. Access becomes controlled, some activities need permits or shutdown windows, and change control tightens. These are real constraints with real durations and should appear as activities and lags in the schedule rather than as an assumed allowance.
Why data centre schedules slip at scale
Owners building several facilities a year face a different problem from a contractor building one. Their schedules rarely fail because of a single event; they fail because the planning method does not hold up when repeated.
- The plan asks for output nobody has delivered. A programme may require three halls of electrical fit-out in parallel when crews have never sustained more than one. On paper the logic works; on site the labour and supervision do not exist. Planned output spikes are one of the clearest warning signs.
- Each project is planned from scratch. When every building gets a new programme, durations are estimated rather than measured, and lessons from the last hall never reach the next.
- Procurement and construction schedules are separate. Equipment dates sit in one system and installation logic in another, so nobody sees a late factory test until the item fails to arrive.
- Commissioning is used as the buffer. Delays upstream are absorbed by shortening testing, which pushes problems into the most expensive point to find them.
- Contractors report in different ways. On a campus with several packages, each reports progress in its own format and level of detail, so the owner cannot compare like with like.
How owners make schedules predictable
Predictability comes from replacing estimates with measured performance, and from testing the plan before the site does.
Measure production on every hall
Keep a small set of production benchmarks and record them on every hall of every project.
| Benchmark | What to record | How it is used |
|---|---|---|
| Hall fit-out duration | Days from hall release to mechanical and electrical completion | Sets the template duration for the next hall |
| Set-to-energise interval | Days from major plant being set to being energised | Shows how much time connection and testing really take |
| Commissioning duration per hall | Days from first functional test to ready for service | Prevents commissioning being squeezed |
| Hall-to-hall offset achieved | Actual gap between successive ready-for-service dates | Tests whether the planned rhythm is realistic |
| Peak parallel output | Most halls in active fit-out at once, sustained over several weeks | Caps what the next plan can assume |
The figures must come from what was actually built and when, not from contractor returns, otherwise the benchmarks simply repeat the original optimism.
Check the plan for impossible peaks
Before a programme is accepted, plot the work it requires per week, for example halls in fit-out or plant items to be set, and compare it with the best sustained output on earlier projects. Any week that asks for much more than has ever been achieved is a risk to resolve now, by re-phasing, adding prefabrication or securing crews, not a target to hope for.
Test delay scenarios months ahead
Ask specific questions of the schedule early. If the second transformer's factory test slips six weeks, which halls move? If the utility energises a month late, how much of the commissioning sequence can still proceed on temporary power? Asking while there is still time to act is what turns a crisis into a managed slip.
Hold readiness gates per hall
Set a short list of conditions each hall must meet before the next stage starts, such as electrical room dry and clean, plant set, containment complete, and record each one with dated evidence. Gates stop work starting before the area is ready, which is the source of much of the rework that later surfaces in testing.
Where YelloSKYE fits
YelloSKYE does not have a data centre case study, and it does not provide scheduling software. What Construct IQ provides is the independent record against which a data centre schedule can be checked.
Drone flights capture the whole campus, including generator yards, plinths and external plant, so set-in-place milestones and hall-to-hall progress can be read from dated orthomosaics. Fixed cameras watch priority zones continuously; owners using our fixed-camera monitoring have reduced physical site visits by up to 80 per cent, and DEX can notify the team when a critical milestone is reached or work in a priority zone stops. Inside the halls, 360° walks record the services before they are closed in.
On the DEX platform, Progress vs. Plan compares actual execution with the planned 4D schedule, and Time Travel Playback steps back through any hall's history to show when each stage was really completed, which is the raw material for the production benchmarks above. For owners with several campuses, the portfolio monitoring use case brings ten or more sites into one dashboard in a standard format, and the construction progress capture use case shows how regular drone capture replaces hours of manual photo reporting.
A data centre schedule checklist
- Treat the power connection date as the controlling milestone and give it an owner and its own reporting.
- Show long-lead equipment as a chain from slot reservation to energisation, linked to installation activities.
- Make weathertight, clean electrical rooms the priority structural milestone.
- Plan halls from one detailed template with a deliberate offset between them.
- Show ready for equipment and ready for service as separate milestones for every hall.
- Give commissioning its own duration based on measured performance, not the time left over.
- Model live-side constraints such as access, permits and shutdowns as activities.
- Compare planned weekly output with the best sustained output from earlier halls and projects.
- Run delay scenarios on key equipment and the grid connection at least quarterly.
- Record production benchmarks on every hall from dated site evidence.
