Base build

Base build testingLB500LB600LB700ACLB 200Standard or solid-state switching

End-of-line

End-of-line testingLCLB 160LCLB 240ACLB 60

Applications

Cooling

Chillers and cooling plantStatic and dynamic load testing

Power

UPS systemsGeneratorsSwitchgear and power qualityProtection settings

Explained

Liquid coolingGPU load testingWater qualityLoop monitoringCommissioning

Equipment

Commissioning equipment

Manifold trolleyFlushing & refill cart

Tools

The full rangeSizing calculatorDatasheets

Services

Commissioning supportLoad bank hireHigh load simulationBespoke load banksSupply and installationMaintenance and repairDeployment and managementBuy or hireProjects

Company

AboutPartnersProjectsAcademyNewsEventsFAQ Speak to the team
Home/News/Data Centre World 2026

PowerLoad at Data Centre World 2026

Three days at the Excel Centre with the liquid-cooled banks, the GPU emulator and the commissioning manifolds. Here is what we took, and the two questions that came up on the stand more than any others.

6 March 2026 · Event

Data Centre World brought the whole commissioning conversation into one hall, and it was noticeable how much of it had moved on from capacity to behaviour. The questions were less about how much load we can apply and more about whether it moves the way real AI compute moves.

On the stand at the Excel Centre: what we brought and why. 37 sec.

What was on the stand

We took the full commissioning set rather than a single headline unit: the 500kW liquid-cooled load banks through to the 160kW GPU emulator, alongside the commissioning manifolds.

The manifolds drew more interest than we expected. That tracks with what we see on site, where the difficulty is rarely generating load and usually understanding what each individual circuit is doing while the load is applied.

Replicating the workload, not just the wattage

The part that got the most attention on the stand was the behaviour rather than the numbers. A GPU under an AI workload does not sit still: it idles around a quarter of its capacity and ramps to full on instruction, repeatedly.

Our emulation reproduces that profile both electrically and thermally, so the building is proven against the load pattern it will actually carry rather than a flat figure it will never see. There is more on that in replicating GPU load.

The commercial questions: hyperscale support, agreements, and unit counts. 35 sec.

The question behind most of the questions

A lot of the conversations were commercial rather than technical: how PowerLoad supports hyperscale environments, what the agreements look like, and whether we can meet the demand on a programme of that size. Several visitors sent NDAs and programme documentation over for proper review against their commissioning plan.

That review is where the most useful thing tends to happen, because the number people arrive with is usually the wrong one.

Sometimes people say to us, I need 500 units. That is not necessarily true. When you break the programme down properly, it turns out you only need 70, across a number of tranches. PowerLoad, on the stand at Data Centre World 2026

Unit count is a function of how the programme is phased, not of how big the hall is. Testing in tranches, with equipment that moves and reconfigures as areas come available, is almost always cheaper and faster than sizing for the whole site at once. Working that out properly is part of the commissioning conversation, not something to settle on a purchase order.

On the floor

Building a liquid-cooled hall?

Send us the programme and we will tell you what testing it properly actually takes, including how many units you genuinely need.

Speak to the team →