On a hyperscale data centre build, MEP is not a phase — it is the project. The civil work and the shell are logistics problems that solve themselves once you’ve done a few. The MEP coordination is the discipline that determines whether the 18-month schedule is achievable or whether the go-live date slips into the next quarter. Every senior superintendent and MEP coordinator working these projects knows this. What’s less consistently understood is the specific shape of the coordination problem and the schedule artefacts that actually make it work.
This article is the MEP-specific Gantt, with the sequencing decisions, coordination cadence, and long-lead tracking that separate projects that hit their dates from projects that don’t. The reader we have in mind is an MEP coordinator, MEP PM, or the GC’s superintendent running MEP on a data centre build. This is specialist content — we’re going to assume you know what an STS is, why selective coordination matters, and why a 4000A switchboard isn’t interchangeable with a 3000A one even though they look similar on the drawing.
Why MEP is the critical path on data centre builds
The short version: on a typical commercial build, structural and envelope drive the critical path. On a data centre, they don’t. Four reasons.
Equipment is manufactured, not built. Switchgear, transformers, UPS modules, chillers — all arrive from the factory as complete systems. Their delivery dates are determined months to years in advance. You can’t accelerate a transformer that’s scheduled to ship in week 85 by putting more crews on it. The schedule logic is tied to manufacturing, not field productivity.
Selective coordination and integrated systems don’t forgive sequencing errors. If one piece of gear is installed in the wrong sequence, downstream work gets blocked. There’s no catching up by throwing labour at it. This is unlike drywall or interior finishes where parallelism and additional crews recover schedule.
Commissioning compresses the back end. L3 through L5 commissioning (covered in Data Center Commissioning Timeline) takes roughly 8 weeks and cannot start until MEP installation is far enough complete to support it. Every day of MEP slip is a day of either commissioning compression or back-end schedule slip. Compression is dangerous; back-end slip is just slip.
Failure modes are expensive. A mis-sequenced structural pour can usually be corrected. A mis-sequenced MEP installation often requires removing installed work. The cost curve bends sharply around the MEP phase, which is why the coordination cadence escalates rather than relaxes.
The practical implication: whoever owns the MEP coordination meeting is effectively running the project during Phases 5 and 6 of the schedule. That responsibility should sit with a named individual — not a committee, not a vendor — and the coordination meeting should happen often enough that decisions are made in hours, not days.
The MEP-specific Gantt
The MEP portion of a typical 18-month hyperscale build runs from roughly month 12 through month 18 — a 16-week window of active MEP installation plus a 4–8 week commissioning tail. Longer than that in typical (non-accelerated) builds.
Within those weeks, four parallel tracks run simultaneously with tightly coordinated handoffs:
| Track | Weeks (of 16-week window) | Key handoffs |
|---|---|---|
| Power | Weeks 1–14 | Utility energisation, internal distribution, UPS integration |
| Cooling | Weeks 2–13 | Chilled water loop, CRAH/CDU install, air/liquid balance |
| Fire protection | Weeks 3–12 | Detection, suppression, BMS integration |
| Low-voltage / BMS | Weeks 4–14 | DCIM, access control, controls integration |
Interleaved with those tracks: commissioning activities that start in week 1 (L2 receipt and placement verification) and ramp through the phase.
Power dominates. Power sets the critical path on roughly 85% of hyperscale MEP schedules, and the power sequence is what the coordination cadence should be built around.
Power delivery sequence
The sequence from utility to rack is long and every stage has dependencies both upstream and downstream.
Stage 1 — Utility substation energisation. Typically before week 1 of the MEP window, often 2–4 weeks before. If it’s late, everything downstream stalls. Utility-owned scope means the GC has limited ability to accelerate.
Stage 2 — On-site substation and MV switchgear. Some builds have on-site substation in the GC’s scope. MV switchgear arrives on site, typically 2–4 lineups depending on topology. Installation takes 1–2 weeks per lineup. Energisation happens after utility provides MV feed and protection studies are signed off.
Stage 3 — Transformers (utility-to-MV or MV-to-LV). Dry-type or liquid-filled, typically step-down from 13.8kV or 25kV to 480V or 415V. Large step-down transformers have been at 128-week lead times into Q2 2025, improving to around 120 weeks — still over two years from order to delivery. Install and test over 1–2 weeks per unit.
Stage 4 — LV switchgear. 4000A–8000A switchboards, main breakers, feeder breakers, tie breakers between redundant sources. Arrive as assemblies, often in sections for larger lineups. Install 2–3 weeks, commissioning overlaps heavily.
Stage 5 — UPS systems. Static UPS modules with battery strings, flywheel UPS on some designs. Modular UPS increasingly common as it speeds installation and commissioning. Install 1–2 weeks per module; parallelisation with multiple modules is normal.
Stage 6 — PDUs (Power Distribution Units). Row-level or rack-level distribution depending on topology. Installed after the UPS systems are ready to feed them but before or during rack installation.
Stage 7 — Busway / whip distribution and rack power. The last mile from the PDU to the rack. Fast once everything upstream is ready; the pain is usually in coordination with mechanical and structural above-rack work.
The critical-path logic here is that each stage cannot be energised until the upstream stage is complete and commissioned. Downstream crews can rough-in but cannot complete. This means stage-gated milestones, not continuous progress, dominate the schedule shape.
Cooling delivery sequence
Cooling is the second critical path, and on liquid-cooled halls for high-density AI workloads, it is increasingly competing with power as the binding constraint.
Stage 1 — Chiller plant. Air-cooled chillers on rooftops or ground-level, or water-cooled with cooling towers. Large chillers have 9–14 month lead times, shorter than transformers but still long. Install typically 2 weeks per unit.
Stage 2 — Chilled water piping. Primary piping loops from chillers to each hall. Large-diameter steel or grooved-mechanical piping. Installation is weeks-to-months depending on run length. Hydrostatic pressure testing is a required hold point before downstream work.
Stage 3 — Air handling or CDUs. For air-cooled halls: CRAH or CRAC units, typically 15–30 per hall depending on size. For liquid-cooled halls: CDUs (Cooling Distribution Units) that transfer heat from the rack-level coolant loop to the facility chilled water.
Stage 4 — Distribution. For air: under-floor plenum, overhead duct, or hot/cold aisle containment with in-row cooling. For liquid: manifolds, quick-connects, rack-level distribution.
Stage 5 — Containment and airflow balance. Only meaningful after racks are at least partially installed. This is the final tuning step before full thermal commissioning.
Cooling coordination with power matters because both feed the same halls and the installation sequences conflict if not managed carefully. Under-floor cable tray installation (low-voltage) needs to happen before raised floor panels are finalised; if cooling distribution goes in the under-floor space it needs its own window. In liquid-cooled halls with CDUs under or beside the racks, the conflict is more acute.
Long-lead equipment tracking
The most important artefact on an MEP-intensive schedule is not the Gantt — it is the long-lead equipment tracker. Every major item, its order date, its confirmed delivery date, its critical-path dependency, and the slack (or lack of it) on that date.
Typical 2026 lead times:
| Equipment | Typical lead time (Q2 2025 data) | Critical-path role |
|---|---|---|
| Large power transformers (MV step-down) | 120–128 weeks | CP on almost every build |
| Medium voltage switchgear | 40–52 weeks | CP on most builds |
| Low voltage switchgear | 30–44 weeks | CP on fast-tracked builds |
| Static UPS modules | 30–40 weeks | Usually not CP but close |
| Generators (>2 MW diesel) | 40–60 weeks | CP on some designs |
| Large chillers (>1000 tons) | 38–56 weeks | CP on liquid-cooled or high-density |
| CRAH / CRAC units | 16–28 weeks | Rarely CP |
| PDUs / busway | 12–20 weeks | Rarely CP |
The numbers shift quarter to quarter. Wood Mackenzie reported transformer lead times eased by about 10 weeks in Q2 2025; manufacturing capacity announcements from Hitachi Energy, Siemens Energy, Eaton, and others suggest further easing through 2026–2027. But the directional reality stands: the MEP schedule for a 2026 build is locked in by procurement decisions made in 2023–2024. If your equipment is not already on order, the 18-month schedule is not achievable regardless of how well you coordinate.
The practical discipline: a weekly long-lead equipment tracker, reviewed by project leadership, with every confirmed delivery date challenged against the manufacturer’s most recent status update. Dates that haven’t been re-confirmed within two weeks should be treated as uncertain. Slip any confirmed delivery date and all downstream scheduling needs immediate re-evaluation — the day you learn about a two-week transformer delay is the day you should already be adjusting the installation sequence, not the day the equipment arrives.
Coordination meeting cadence
The coordination cadence on a hyperscale MEP job is noticeably heavier than on typical commercial construction. This is proportionate to the cost of errors.
Daily stand-ups (15 minutes). Foremen from each active trade, superintendent running. What’s happening today, what’s blocked, what’s needed for tomorrow. No PowerPoint, no scope discussions — those belong elsewhere. Start-of-shift, same time every day, same location.
Weekly MEP coordination meeting (90 minutes). MEP subcontractor leads, GC’s MEP superintendent, commissioning agent representative, key equipment vendors on critical-path items. Review of schedule against plan, long-lead equipment tracker update, coordination conflict resolution, upcoming three-week lookahead.
Bi-weekly OAC (Owner-Architect-Contractor) meeting (2 hours). Owner’s representative, GC leadership, designer, commissioning authority. Decisions that require owner input or design clarification. This is where change management lives.
Monthly executive review (2 hours). Project leadership from all parties. Schedule health, budget health, risk log, forward-looking 90-day view. This is where bad news goes upstairs.
Some teams add a daily MEP pull-planning session during peak MEP installation (roughly weeks 8–14 of the 16-week window). Strongly recommended. The cost of 30 minutes of daily pull-planning is minuscule compared to the cost of two trades showing up at the same location on the same day unprepared for each other.
The coordination cadence should increase rather than decrease as commissioning approaches. The last four weeks of MEP installation are when small sequencing errors are most expensive and most frequent. Teams that relax the cadence here routinely pay for it in commissioning rework.
BIM-based coordination
Most 2026 hyperscale builds run MEP coordination through a BIM model — typically Revit for design, Navisworks for clash detection, often with a federated model linked to the project’s document control platform (Procore, Autodesk Construction Cloud, or similar).
The discipline that matters:
Clash detection cadence. Weekly at minimum during design and pre-construction, ideally continuous during MEP rough-in. The model should be updated with installed-as-built conditions at each major milestone.
Owner of model integrity. One named person. Usually the MEP coordinator or a dedicated BIM coordinator. Whoever it is, the model does not get updated without their approval — this sounds bureaucratic and is actually essential.
Handoff to operations. The as-built model should be commissioning-ready and handed over at project closeout. A lot of hyperscale builds that nail the construction schedule fail to nail the model handoff, which hurts operations long after construction leaves site.
BIM coordination is not a substitute for the weekly MEP coordination meeting. It is a tool that makes the meeting more productive — clashes identified ahead of the meeting, resolutions documented in the model. Teams that think the BIM workflow replaces the coordination meeting consistently miss field-level sequencing issues that the model doesn’t capture.
Common MEP coordination failures
Five patterns that show up repeatedly on hyperscale builds, every one of them avoidable.
Late transformer or switchgear delivery without early mitigation. The vendor tells you in week 10 that the transformer will be four weeks late. That news should have been in the tracker in week 8 at the latest. If your vendor communications aren’t surfacing delays early, the communication structure is broken.
Sequence errors in switchgear energisation. Missing a commissioning hold-point — energising gear before protection study sign-off, or connecting a downstream feeder before upstream breakers are tested — can require tearing out completed work. Every energisation should have a named sign-off authority and a signed hold-point release.
Under-floor and overhead coordination failures. Cable tray, piping, fire main, and ductwork competing for the same overhead or under-floor space. A BIM-coordinated design should prevent this, but field deviations from the model during installation often reintroduce the conflict. This is why installed-as-built model updates matter.
Commissioning agent under-capacity. Scheduling L3–L5 commissioning assuming commissioning agents are available when needed, then discovering the Cx team you assumed was yours is committed to three other projects. Secure Cx capacity at the same time you secure long-lead equipment orders, not later.
Electrical trade shortage absorbed as “standard schedule slip.” Skilled electrician availability in Northern Virginia, Phoenix, Dallas, and other boom markets has been the binding labour constraint on 2025–2026 builds. Treating it as a generic schedule risk rather than a specific MEP-critical-path risk produces exactly the kind of two-week-here, two-week-there slippage that compounds into a quarter-lost go-live date.
For the broader labour context, see Best PM Software for MEP Subcontractors.
FAQ
Q: What’s the typical MEP duration on a hyperscale build?
MEP installation runs roughly 12–16 weeks on an accelerated hyperscale schedule (18-month total build). That window plus a 4–8 week commissioning tail means MEP-plus-Cx is typically months 12–18 of the 18-month schedule. On longer builds (24–30 months) the MEP phase is often 20–26 weeks with a more relaxed commissioning window.
Q: Why are electricians the binding labour constraint?
Precision MV wiring for data centres is specialised work, and the population of electricians qualified to do it is not growing fast enough to meet 2025–2026 demand. Approximately one-fifth of US electricians are over 55 and heading toward retirement. Hyperscale boom markets (Northern Virginia, Phoenix, Dallas, Columbus) have seen qualified-electrician wage rates rise significantly. Some data centre GCs now pre-contract electrical labour allocations a year ahead.
Q: Can modular electrical rooms cut the MEP schedule?
Yes, meaningfully. Prefabricated electrical rooms delivered to site with switchgear, UPS, PDUs, batteries, and controls already installed and internally pre-commissioned cut 4–8 weeks off the on-site installation sequence. The trade-off is up-front commitment — modular designs are locked in earlier and changes late in design are expensive.
Q: What’s the most common single-point failure?
Switchgear lineup not ready for the scheduled energisation date. Whether that’s delivery delay, FAT failure requiring rework, or installation running behind, a single switchgear slip cascades into everything downstream of it. Every hyperscale MEP schedule should treat switchgear readiness as the single point of highest schedule risk.
Q: How does liquid cooling change the MEP coordination?
Significantly. CDU installation, quick-disconnect plumbing at rack level, and leak-detection systems add scope that didn’t exist on air-cooled halls five years ago. The mechanical trade takes on a larger coordination role. Plumbing tolerances are tighter. Commissioning complexity increases. Expect liquid-cooled halls to add 10–20% to total MEP duration versus equivalent air-cooled halls, shrinking as industry experience grows.
Q: Where does the commissioning team fit into MEP coordination?
The commissioning agent should be at the weekly MEP coordination meeting from the start of MEP rough-in, not just at the L3+ phase. Cx observations during installation catch problems that would otherwise surface during L4 functional testing — at which point fixing them is substantially more expensive. See Data Center Commissioning Timeline: The Final Eight Weeks for detail on Cx sequencing.
Q: What project scheduling software handles MEP coordination best?
Primavera P6 for the master schedule. For MEP-specific coordination, a combination of P6 for scheduling and Fieldwire or Procore for task-level field coordination is the typical pattern. BIM clash detection through Navisworks or similar. See Best Construction Scheduling Software for US General Contractors 2026 for the full landscape.