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/Explained/Replicating GPU load/Testing for AI loads
Replicating GPU load

Why is testing for AI loads essential rather than optional?

A static load test can pass a hall that would fail in service.

Published 26 August 2026

0:55
Testing the whole system, not just the rack
Both waysLoad rising and falling
SurgesTransient electrical spikes
Whole systemUPS, generators and cooling
End-of-lineRack-level replication

A static load test can pass a hall that would fail in service. AI load changes behaviour, not just size: it swings up and down quickly, and contactor-based load banks step far more slowly. Testing properly means replicating those swings and surges, and it tests more than the rack. It shows how the UPS, the generators and the cooling respond when the load changes, in both directions.

The full answer

AI load behaves differently

The pattern of an AI load is unlike anything data centre infrastructure was designed around. It is not just larger, it changes quickly and in both directions.

Contactor-based load banks step the load far more slowly, so a conventional test may never show how the building behaves under AI load.

Beyond a static load

A static load proves the building can carry the power. It does not put the systems under the stress they will see in service.

That takes fast transitions, surges and precise load steps, applied in real time and replicating how the GPUs behave rather than approximating their peak draw.

It tests the whole system, not just the rack

Emulating GPU load tests more than the rack. It shows how the UPS, the generators and the cooling respond when the load changes rapidly.

A UPS that handles a steady draw perfectly may behave differently when the load collapses and returns. The only way to find out is to make it happen.

Where a static test stops

A bank that can only step slowly leaves a region of behaviour the test cannot reach, which is a blind spot in critical testing.

The base build range covers the static test, with solid-state switching as an option when the specification asks for transients. The end-of-line GPU emulators replicate the rack itself.

More in this topic

Replicating GPU load

FAQ

Answered

No. A static load holds the infrastructure at a steady state and never triggers the failure modes that matter. AI workloads change quickly, and it is the rate of change that stresses the UPS, generators and cooling.
It tests everything around the rack under rapid load change: the UPS, the generators, protection and the cooling system. Those components respond to transitions rather than to steady draw.
Where to next

Related reading

Testing for hyperscale?

We will talk through what your current test regime can and cannot prove.