Will faster installation bring handover forward? A data-centre delivery guide
A faster task does not automatically bring handover forward. The saving must survive design coordination, manufacture, delivery, installation, inspection and commissioning. Data-centre robotics and prefabrication offer useful examples, but their headline rates measure different things. Before including a saving in a tender or programme, establish the baseline, test the full work package and recalculate the project’s critical path.
Part 3 of 4: What the data-centre build-out is teaching construction.
Repeated drilling offers a testable automation case
DEWALT’s January 2026 launch described a downward-drilling robot developed with August Robotics for repeated holes in data-centre slabs. Its pilot claims included:
- Reported hole-location and depth accuracy of 99.97% across more than 90,000 holes; the acceptance tolerance still needs to be specified.
- Peak drilling speed up to ten times the traditional method; this is a task-rate comparison.
- An aggregate 80-week saving across ten projects; this does not establish the handover benefit for an individual project.
DEWALT announced commercial availability in July 2026. Availability, productivity and project completion are separate claims. For a local procurement, obtain the current specification, Australian supply and support arrangements, and the conditions behind the performance comparison. Agree hole size, depth, reinforcement conditions, scanning, set-out tolerance and acceptance before the trial.
Compare accepted holes per crew-hour and elapsed package duration, including mobilisation, scanning, setup, supervision, stoppages and rework. A faster drill can reduce labour even where handover is unchanged. Keep those benefits separate so the commercial case remains useful.
Prefabricated overhead services can shorten site work but bring decisions forward
In March 2026, Compass Datacenters and Vertiv described SmartRun, a factory-built overhead system that combines power, cooling, containment and networking. Vertiv lists power busbars, stainless steel liquid-cooling pipework with isolation valves, hot-aisle containment and network infrastructure, built in blocks of about 2.4 metres and lifted into place.
- Compass reports a reduction of up to 85% in the data hall’s fit-out timeline. That is a reported scope-specific maximum, not an 85% reduction in the building programme.
- Vertiv reports installation capacity above 1 MW per day by one crew. Installed capacity is not the same as tested, commissioned and available computing capacity.
Neither cited page supplies a complete baseline programme or crew definition for reproducing the claimed saving. Prefabrication moves coordination and assembly earlier. Reserve time for interface approval, factory testing, transport, lifting and final site tests. If the layout remains unsettled, earlier manufacture can turn late changes into remanufacture.
An Australian example: prefabricated cooling towers at NEXTDC S3, as reported in March 2025
KUSCH’s March 2025 project account describes structural engineering for the NEXTDC S3 cooling-tower modules with Multiplex and A.G. Coombs. Its account records the delivery method rather than a controlled comparison of programme savings:
- Repeated plant: 54 towers, each nearly nine metres tall. Use the supplier’s certified weights and centres of gravity for actual lift planning.
- Structural design: the engineer reports Importance Level 4 wind and seismic requirements for this project. This is not a default classification for every data centre.
- Sequence: each tower arrived in three parts, was assembled in a designated area and was lifted onto pre-installed skids.
- Commissioning: the account distinguished operating towers from those still being commissioned. Installation and operation were separate milestones.
The lesson for other campuses is to resolve the interfaces before manufacture: support tolerances, connection positions, lifting access, laydown area, inspection access and the route for future replacement. Repetition helps only when the repeated detail is coordinated and accepted.
An 800 VDC proposal needs an architecture brief
NVIDIA’s October 2025 technical article proposed higher-voltage DC distribution for increasingly dense racks. Three distinctions matter to the building team:
- Rack power is a local concentration of demand. The distribution and cooling design must support that concentration, not only the total building load.
- For a given DC power, higher voltage reduces current. This explains the interest in higher-voltage distribution; it does not by itself determine conductor size or system efficiency.
- Conversion losses depend on the complete architecture and operating load. A voltage label alone does not establish an energy saving.
Vertiv’s July 2026 account distinguishes rack-adjacent DC sidecars, which retain much of the upstream AC infrastructure, from a later centralised DC architecture. It described validation and demonstrations, with commercialisation and deployments at later stages. For the project brief, request a tenant-approved single-line concept, equipment availability, protection approach, maintainability and system commissioning plan. Do not redesign switchrooms or risers around a roadmap alone.
Trades are a shared constraint
Google’s November 2025 Texas training announcement illustrates another dependency: new equipment still needs people able to install and operate it. Its apprentice numbers are programme targets, not a labour commitment to your project. Check the contractor’s actual trade capacity, training, supervision and vendor support against the proposed dates.
Worked check: faster drilling, unchanged handover
Hypothetical programme: drilling takes ten working days and a separate equipment-delivery path takes fourteen, both starting on day zero. Three days of testing can begin only after both are complete. Completion is max(10, 14) + 3 = 17 working days. Cutting drilling to two days still gives max(2, 14) + 3 = 17. This simplified example assumes no other dependencies or resource conflicts. It shows why a task saving must be tested in the linked programme before it becomes a handover promise.
The evidence pack before accepting a programme saving
Request | Acceptance test |
|---|---|
Comparable baseline | Same installed scope, tolerances, working hours, access and quality. State what the published comparison excludes. |
Complete package duration | Include design freeze, factory work, logistics, setup, installation, inspection and rework. Do not count parallel time twice. |
Measured trial output | Use accepted work per crew-hour and elapsed duration under representative site conditions, not peak machine speed. |
Design and interface freeze | Identify the drawings, dimensions and connections to be approved before manufacture; record the last responsible decision date. |
Logistics and replacement plan | Confirm transport, lifting, laydown, temporary works and access to maintain or replace the assembled equipment. |
Commissioning and support | Separate installed, tested and operational milestones. Name the integrator, acceptance tests and local support. |
Recalculated project programme | Show dependencies, resources and the critical path before and after the change. Identify the constraint that now controls handover. |
Makao’s recommendation: approve task productivity, package cost and handover improvement as three separate findings. GAO’s Schedule Assessment Guide explains why an integrated schedule is needed to evaluate change. If switchgear delivery or integrated testing still controls completion, report the labour or package saving without promising an earlier handover. Record the assumptions, decision owner and trigger for reassessment.
Before counting an installation gain as a handover gain, define the commissioning and integration evidence needed to accept the complete system.
Record cross-trade responsibilities in the subcontractor scope map before prefabrication freezes the interfaces.
Sources
Dated supplier and project accounts. Performance figures remain attributed claims. The acceptance framework is Makao’s synthesis, supported by GAO’s scheduling guidance.
- DEWALT commercial launch, 9 July 2026
- GAO, Schedule Assessment Guide (2015): integrated schedules and evaluation of change.
- Vertiv, 17 July 2026
- DEWALT, 20 January 2026
- Compass Datacenters, 3 March 2026 (updated 7 April 2026)
- Vertiv SmartRun case study (undated)
- KUSCH, 28 March 2025
- NVIDIA Technical Blog, 13 October 2025
- Vertiv, 13 October 2025
- Google, 14 November 2025
Questions or corrections: contact Makao.
