Hyperscale data centre programmes used to be described as a race to enclose the building. That story is outdated. Clients care when IT load can land, when power and cooling are stable, and when commissioning gates are passed. Those dates now pull design freezes, equipment manufacture, installation sequences and even civil temporary works. If the schedule still starts with foundations and hopes commissioning will fit at the end, the critical path is already wrong.
The practical change for planners and directors is simple. Build the programme from the commissioning and energisation logic first. Then hang structure, envelope and fit-out from the dates those systems need.
Why commissioning became the driver
Hyperscale campuses stack long-lead electrical plant, cooling plant, controls, and staged hall handovers. Revenue and contract remedies attach to megawatt readiness, not to weathertight dates. A building that looks complete but cannot complete integrated systems testing is still late.
At the same time, vendors and utility interfaces have hardened. Transformer, switchgear and chiller slots do not move because the slab pour slipped. The commissioning window is often fixed by client IT waves, grid availability or a lease obligation. Construction then has to arrive in the right order for that window, not the other way round.
What “backwards planning” means in the file
Backwards planning is not a slogan. It is a network that starts from defined commissioning milestones—energisation, integrated systems test, level of redundancy proven, hall ready for IT—and works upstream to installation complete, factory acceptance, design freeze and enabling works.
- Name the commissioning gates in the WBS with clear done criteria, not a single “commissioning” bar.
- Link each gate to the systems and rooms that must be complete, not to a percentage of the building.
- Pull long-lead FATs and shipping from those same gates so procurement float is real.
- Only then schedule structure and envelope as predecessors where they truly unlock plant installation.
If civil works still sit as the only critical path in the early baseline, the model is describing the wrong project.
The dates that usually own the path
| Gate | Why it drives upstream work | Common failure if ignored |
|---|---|---|
| Utility / grid availability | Sets when primary power can be live | Building ready, campus dark |
| HV / MV energisation | Unlocks most downstream testing | Fit-out complete against dead boards |
| Cooling plant ready | Needed before meaningful IT load | Halls finished, thermal rejection missing |
| BMS / controls stable | Required for integrated tests | Equipment installed, sequences unproven |
| Hall / suite turnover | Client IT wave dates attach here | Shell progress reported; revenue date slips |
Design and procurement have to follow the gates
Once commissioning dates are fixed, design release plans must feed the equipment that those gates need. A late electrical design freeze is not a paper issue. It moves the transformer order, the FAT, the install and the energisation rehearsal.
Procurement schedules that sit as separate trackers are a hyperscale failure mode. Factory slots and shipping must be activities in the same network as installation and testing. A dashboard that shows “procurement green” while the P6 path to commissioning is red is noise.
Construction sequencing for turnover, not for optics
Trades still want continuous areas. Commissioning wants systems complete across the right rooms. Those aims conflict on every fast hyperscale job. The schedule has to prefer system completeness where a gate depends on it, even when that looks less efficient on a floor-plate percent-complete chart.
Temporary power, temporary cooling and phased fire strategy are not optional footnotes. They are often the only way an early hall can commission while later halls are still shells. Model them as real activities with owners. If they are missing, early revenue dates are hope.
One schedule owner still matters more
Hyperscale campuses multiply package contractors: civils, steel, envelope, electrical, mechanical, controls, commissioning management. Each can hold a local programme that looks healthy. Only an integrated owner can see whether the commissioning gate is still intact.
That owner must refuse package updates that improve a trade finish while breaking a predecessor to energisation. They must also refuse remaining-duration cuts on commissioning tasks made only to protect a client slide. Testing durations are evidence-based or they are fiction.
What directors should see every period
Stop leading the pack with concrete poured and cladding percent. Lead with the next commissioning gate, float to that gate, the driving system path, long-lead status against that path, and open design or change that threatens it. Shell metrics can sit second.
Ask whether the IT or capacity date moved. If it did not move on the slide but float to energisation collapsed, the programme moved. Report that.
Where 4D and commissioning meet
4D helps when it tests whether plant install and access routes exist in the right rooms before a commissioning window. It does not help when it is only a fly-through of the structure. Tie the model to the same gate milestones as the P6 network, or leave it as communication, not controls.
A working rule for hyperscale baselines
If the commissioning and energisation network cannot be walked without gaps, the baseline is not ready—however tidy the civil path looks. Fix the turnover logic first. Then set structure and fit-out to serve it. Then freeze the dates you will sell to the client.
EVM Solutions plans and controls data centre programmes with that order of work: gates, systems, long leads, then the building. If your hyperscale finish is still described mainly by shell progress, send the commissioning milestone list and the current P6 file. The path that pays is the one that turns systems on.
