Commissioning is where a data centre construction project either earns its go-live date or loses it. The construction work gets the attention and the budget; the commissioning process gets the schedule pressure. On most 2026 hyperscale builds, the physical construction can be on time and still miss the owner’s operational date because commissioning ran long. This happens often enough that “how much Cx schedule do I need” is arguably the single most under-asked question in data centre project planning.
This article is the 8-week L1-through-L5 commissioning schedule with realistic sequencing, the handover documentation that actually gets demanded at closeout, and the specific places where commissioning schedules slip. The reader is a commissioning agent, an MEP commissioning manager, or the GC’s closeout PM running the final weeks of a build. If you’ve done this before, much of what follows will be familiar; the value is in the specific sequencing patterns and the failure modes.
The 8-week figure in the title is the active on-site commissioning window. L1 factory testing happens months earlier during equipment manufacturing, not in those eight weeks. And the 8 weeks assumes a well-run project — troubled projects can stretch commissioning to 12 or even 16 weeks as L4 and L5 failures cascade through rework cycles.
Data center commissioning: the five levels explained
The L1–L5 nomenclature is industry standard, though the specific definitions can vary slightly between commissioning firms. The following is the common definition used on most US hyperscale projects, aligned with ASHRAE commissioning guidance and the Uptime Institute Tier framework.
L1 — Factory Witness Testing (FWT). White-tag. Equipment is tested at the manufacturer’s factory before shipment. Witnessed either in-person by the commissioning agent or remotely via documented factory testing. Confirms the equipment meets specification before it leaves the factory. Typically happens during construction Phase 4 or early Phase 5 — months before the 8-week window.
L2 — Site Acceptance / Delivery Inspection. Yellow-tag. Equipment is received at the site, inspected for shipping damage, placed in its final location, and verified against the design. Internal inspection confirms no damage occurred in transit and that the unit matches what was ordered. Happens during MEP rough-in, typically weeks 1–3 of the 8-week commissioning window.
L3 — Pre-Functional Testing (PFT). Green-tag. Each individual system (a single chiller, a single switchgear lineup, a single UPS module) is tested standalone. Verifies that the system operates correctly on its own, before it is integrated with other systems. Weeks 2–5.
L4 — Functional Performance Testing (FPT). Blue-tag. Systems are tested under operational scenarios, including partial load, full load, and individual failure modes. Each system is proven to respond correctly to normal operating conditions and to the specific failure conditions in the sequence of operations. Weeks 4–7.
L5 — Integrated System Testing (IST). White-tag (confusingly, same colour as L1 at some firms; other firms use a distinct colour). All systems operate together under full load including worst-case failure scenarios. This is where utility loss, generator start, load transfer, cooling response, and all interdependencies are proven in combination. Weeks 6–8.
The phases overlap because they have to. Running them strictly sequentially would add 4–6 weeks to the overall duration. Overlap works because different systems are at different stages at any given time — while one switchgear lineup is undergoing L4, another is still at L3, and a third is still at L2.
The 8-week commissioning Gantt
A typical 8-week on-site commissioning schedule, showing the overlapping levels:
| Week | Dominant activities |
|---|---|
| 1 | L2 on latest-arriving equipment; L3 starts on earliest-installed systems |
| 2 | L2 continues; L3 expands across systems; first L4 test procedures drafted |
| 3 | L2 completes; L3 in full swing; L4 begins on first systems to complete L3 |
| 4 | L3 completes on most systems; L4 expands; L5 planning and procedures finalised |
| 5 | L4 expands; punch list from L3 addressed; L5 prep activities |
| 6 | L4 nearing completion; L5 begins; integration testing across subsystems |
| 7 | L4 completes; L5 full load testing; fault scenario testing |
| 8 | L5 completes; final punch list; handover documentation compilation |
Within those eight weeks, commissioning activity runs continuously — typically two-shift operations during weeks 4–8, and 24-hour operations during the L5 testing at weeks 7–8 when full-facility load testing can’t be paused for normal working hours.
The schedule is aggressive but achievable on a well-prepared build. The factors that make it achievable are covered below.
L1: Factory acceptance testing (pre-construction window)
L1 happens months before the on-site commissioning window. It’s on this timeline because its outcomes define whether L2 and L3 proceed smoothly on-site.
Typical L1 scope:
- Switchgear FAT at manufacturer: protection device testing, breaker operation, control scheme verification, internal arc flash testing where applicable.
- UPS FAT: battery discharge test, transfer testing, full-load operation.
- Chiller FAT: performance test against design conditions, efficiency validation.
- Generator FAT: load bank testing, fuel consumption validation, control interface verification.
Witnessing model: Most hyperscalers have moved toward virtual FWT where possible — video-documented factory testing with the commissioning agent reviewing remotely rather than travelling to the factory. This was accelerated during the pandemic and has stuck because it saves weeks and cuts travel cost.
Common L1 findings that matter: Protection setting discrepancies, control scheme errors in BMS integration, and performance failures against specification. L1 findings that require equipment return to factory are the ones that cascade into the 8-week window — they should be flagged and resolved during manufacturing if possible.
L2: Site acceptance testing (weeks 1–3)
Equipment arrives on site and L2 begins immediately. On-site verification that the unit that arrived matches the unit that was tested at factory.
Typical L2 scope:
- Visual inspection for shipping damage.
- Verification of nameplate data against purchase order and design.
- Placement verification — is the equipment where the drawings say it should be, with the right clearances, access, and service requirements?
- Connections inspection — terminations torqued to spec, cables correctly landed, labelling complete.
- Documentation completeness check — O&M manuals, as-built drawings, warranty paperwork all received.
Sign-off: Yellow-tag sign-off authorises proceed to L3 for that specific piece of equipment. The commissioning agent signs; the contractor countersigns.
Common L2 findings: Shipping damage not caught at offload (usually minor, occasionally significant), configuration mismatches where the shipped unit differs from the design intent (usually a factory or procurement error), and placement issues where equipment physically doesn’t fit or access is compromised.
L3: Pre-functional testing (weeks 2–5)
Standalone testing of each individual system. The goal is to confirm that every system, in isolation, operates correctly per the sequence of operations.
Typical L3 scope by system:
- Switchgear: Individual breaker operation, protection device testing, interlock verification, control scheme operation without external integration.
- UPS: Battery integrity, transfer to bypass, transfer from bypass, internal alarming.
- Generator: Start command, voltage regulation, frequency regulation, load acceptance (often with temporary load banks), shutdown.
- Chillers: Start-up sequence, setpoint control, individual control loop tuning.
- CRAH/CRAC/CDU: Airflow or flow balancing, individual unit operation at setpoint.
- BMS: Point-by-point verification that every input and output in the controls system reads/writes correctly.
Sign-off: Green-tag authorises proceed to L4 for that specific system.
Common L3 findings: BMS point mismatches (the most common L3 issue by volume — DDC points that don’t match the drawings, address conflicts, naming inconsistencies), control loop tuning issues, and individual equipment faults that didn’t surface at factory.
Most L3 findings are fixable within hours to days. L3 is where the bulk of the “punch list” is generated, and the punch list drives the tail-end of the commissioning schedule.
L4: Functional performance testing (weeks 4–7)
Where commissioning becomes meaningful. L4 tests each system under operational scenarios — the scenarios in the sequence of operations that the system must handle during normal operation and under specific failure conditions.
Typical L4 scope:
- Utility loss / generator start: Simulated utility loss, UPS carries load, generator starts, transfer occurs, load is carried on generator power.
- Utility restoration: Generator cooldown sequence, retransfer to utility, return to normal operating state.
- Individual component failure: Loss of one chiller in a redundant set, loss of one UPS module, loss of one PDU — system responds per design.
- Partial load and full load operation: Facility operating at 25%, 50%, 75%, 100% of design load with load banks providing the load.
- BMS integrated response: Controls correctly orchestrating the facility’s response to each test scenario.
Sign-off: Blue-tag authorises proceed to L5 for that system.
Common L4 findings: Controls logic errors (the system responds incorrectly to the fault scenario — often a programming issue rather than hardware), timing issues in transfer sequences (generator doesn’t start quickly enough, UPS drops load before transfer completes), and sequence-of-operations mismatches between the design intent and the installed controls.
L4 findings are expensive to fix. Hardware re-work, controls re-programming, retest — a significant L4 finding can cost 1–2 weeks on the schedule. L4 is also where the commissioning agent’s experience matters most; novice Cx agents routinely accept borderline L4 results that come back to bite during L5.
L5: Integrated system testing (weeks 6–8)
The culmination. All systems operate together under full facility load with coordinated failure scenarios. This is where the facility proves it can do what the design says it can do.
Typical L5 scope:
- Full facility load test: Load banks applying full design IT load, cooling systems carrying the thermal load, electrical systems delivering power with the target redundancy.
- Concurrent failure scenarios: Utility loss combined with one chiller failure, or UPS failure combined with specific branch circuit failures — worst-case scenarios defined in the owner’s project requirements.
- Maintenance mode validation: Systems correctly support maintenance on individual components without service interruption.
- Extended operation: Typically 24–96 hours of continuous operation at full load, verifying thermal stability, control stability, and absence of intermittent faults.
Sign-off: White-tag (or owner’s final acceptance signature) authorises operational handover.
Common L5 findings: Integration issues that weren’t visible at L4 — one system interfering with another, timing conflicts in controls orchestration, and thermal issues that only emerge at full sustained load. L5 findings are the most expensive to resolve because they typically involve more than one system and require coordinated re-work.
Handover checklist structure
Commissioning completion is not the same as handover. The handover package — the documentation that transfers the facility from construction to operations — is typically assembled in parallel with L5 testing.
Required artefacts typically include:
| Document type | Contents |
|---|---|
| Test reports | Full L1–L5 results, including pass/fail records, sign-offs, retest documentation |
| O&M manuals | Manufacturer operation and maintenance documentation for every installed system |
| As-built drawings | Updated construction drawings reflecting installed-as-built conditions |
| As-built BIM model | Federated model reflecting installed conditions, typically handed over on media |
| Sequence of Operations | Final SOO matching installed controls programming |
| Training records | Operations staff training completion records |
| Spare parts inventory | Physical spares on site, parts lists, recommended stocking levels |
| Warranty documentation | Manufacturer warranty start dates, service contact information |
| Certificates of occupancy / energisation | AHJ approvals, utility final sign-offs |
A complete handover package runs to thousands of pages in a hyperscale facility. The time to assemble it is not the 8-week commissioning window — it’s the 3–6 months preceding, with final assembly during commissioning. Projects that leave handover documentation to the last week consistently ship incomplete packages and pay for it in operations handoff friction.
Where commissioning schedules actually slip
Five patterns that show up repeatedly.
Late MEP installation cascading into compressed Cx window. When MEP runs 2 weeks late, commissioning either starts 2 weeks late (pushing go-live) or compresses (increasing risk of missed findings). Neither option is good. The answer is aggressive long-lead tracking during MEP, not commissioning heroics at the end.
L4 findings requiring hardware rework. Typically controls integration issues that weren’t caught at L3 because L3 tested standalone. Can be mitigated by running more integrated scenarios during L3 rather than waiting for L4 to expose them.
L5 discovering system-level integration failures. Rarer but more expensive. Best mitigated by pre-L5 integration walkthroughs before formal L5 testing begins — essentially a dry run that catches integration issues without triggering official fail documentation.
Commissioning agent availability. The most under-discussed constraint in 2026. Qualified Cx teams are booked out 12+ months in every major US data centre metro. A build that doesn’t secure Cx capacity at the same time it secures long-lead equipment will discover during MEP that its assumed Cx team isn’t actually available. The fix is to book Cx when you book transformers.
Documentation shortfall at handover. Commissioning passes technically but the handover package is incomplete. Operations refuses sign-off. The project technically isn’t complete until documentation is. Budget 2–3 weeks for documentation finalisation after L5 completion, not 2–3 days.
FAQ
Q: Can the 8-week commissioning window be compressed to 6 weeks?
In theory yes, on a well-prepared facility with extensive pre-commissioning integration testing. In practice, most attempts to compress below 8 weeks end up at 8–10 weeks anyway because the hidden work surfaces. The more reliable schedule compression lever is to start L2 earlier — if equipment can be received and placed during MEP rough-in rather than after, L2 overlaps with construction rather than extending commissioning.
Q: What’s the difference between commissioning and testing?
Testing verifies that a system does what it’s supposed to do. Commissioning verifies the same thing plus the documentation, the training, the handover artefacts, and the integrated-facility behaviour. Testing is a subset of commissioning. This distinction matters because many project teams underestimate commissioning as “testing plus paperwork” when it’s a substantially larger scope.
Q: Who signs off on L5?
The commissioning authority (often a third-party firm) signs technical sign-off. The owner’s representative signs operational acceptance. The AHJ signs occupancy. The utility signs final energisation. All four are required for full go-live, and any one can block the others. Owner-representative sign-off is typically the last and most political — operations teams often add requirements during L5 that weren’t in the original OPR.
Q: How does the Uptime Institute Tier system affect commissioning?
Tier certification (Tier III or Tier IV for hyperscale) imposes additional requirements on redundancy, concurrent maintainability, and fault tolerance that commissioning must verify. Uptime’s own Tier Certification of Constructed Facility is a separate commissioning-adjacent process that, where required, adds 2–4 weeks to the overall Cx timeline and requires Uptime-accredited agents.
Q: Why do commissioning agents get booked out so far in advance?
The supply is small. Qualified data centre commissioning engineers require years of experience and are produced slowly. The demand has risen sharply with the 2024–2026 AI infrastructure boom. The net result is that Cx capacity is harder to secure than most other MEP resources in the current market. Book 12+ months ahead for major projects.
Q: What’s the relationship between commissioning and the overall construction schedule?
Commissioning is typically phases 5–6 of the construction schedule, running concurrent with the back half of MEP installation. See Data Center Construction Schedule: A Realistic 18-Month Gantt for the full build-phase view and Hyperscale Data Center MEP Coordination for the MEP-Cx interface specifically.
Q: Is commissioning different for liquid-cooled versus air-cooled halls?
Yes. Liquid-cooled systems add leak detection testing, CDU-level flow verification, and rack-level quick-disconnect validation to the L3–L5 scope. Thermal testing at L5 is more complex because the facility must prove it can remove heat at the target density without hot-spot formation. Expect roughly 20% additional Cx duration on heavily liquid-cooled halls versus equivalent air-cooled halls, shrinking as liquid-cooling commissioning practices mature across the industry.