Interactive lab · Nothing is saved

Plan a fictional data centre.

Help Rosie work through the design options for an AI data centre near a fictional Australian town, and the implications each option has for energy, water and the community. Then, have a go at building your own data centre.

Rosie investigates a felt data centre with visible server racks, cooling pipes and ventilation equipment. — AI-assisted illustration

This is a guided investigation, not an engineering calculator or an assessment of a real development. You are not deciding whether AI is good or bad. You are learning what to look at.

Data centre decision lab

What would you need to know?

Step 1 of 7

First step

The job is already chosen.

Use the Back and Continue buttons to first navigate the example. You do not need to choose a proposal for the first activity.

Your data centre build

Assembly · build in progress

Meet our guides

RosieAsks what we know, what we do not, and what to check next.

NeighbourNotices noise, construction and day-to-day local impacts.

Local politicianAsks about water, grid pressure and public responsibilities.

Independent community plannerChecks who benefits, who pays and what evidence is shared.

AI-assisted illustrations.

Signals update as you build.

Open day · Data centre decision lab

The centre is assembled. What is known?

A cooling label alone cannot establish PUE, WUE or annual resource use.

Data centre decision lab

Evidence trail

Where the ideas in this lab come from, and how to check them yourself. No exact figures are given; the numbers are for you to find.

What the words mean

PUE (Power Usage Effectiveness)

Compares all the electricity a centre uses with just the electricity its computers use. A lower overhead can improve it, but it does not measure total emissions or whether the computing work itself is worthwhile.

WUE (Water Usage Effectiveness)

Compares site water use with the energy used by the computers. What it means depends on where the measurement boundary is drawn, and it does not tell you whether the water is scarce, drinkable or recycled.

Potable water

Treated water that is safe to drink. A centre may instead use recycled or other non-drinking water for some jobs.

Operational data

Measured while the centre is actually running. It is stronger evidence than an estimate made before it was built.

Facility energy overhead

Extra energy a site uses for infrastructure such as cooling, power conversion and lighting, beyond the computers themselves.

Grid exposure

How much a build leans on the shared electricity network, rather than bringing its own power.

Reliability dependency

How much a build's day-to-day running depends on things outside its control, like grid stability or backup systems.

Evidence gap

The difference between what a company promises before building and what it actually reports once it's running.

Where this comes from · Australia

Australian Government: Expectations of data centres and AI infrastructure developersPublished March 2026

Water Services Association of Australia: Data Centres and Water in AustraliaDecember 2025

NABERS: energy ratings for data centres (water framework in development)National, government-backed

Infrastructure Australia: water supply analysisNational infrastructure body

AEMO (Australian Energy Market Operator): grid capacity and connection informationNational electricity market operator

If you don't live in Australia

Our town is Australian, so these are Australia's rules. But the heat and the water work the same everywhere. The US Department of Energy, for example, publishes very similar cooling-water guidance. The interesting part is finding out what your country or state expects.

Take it further · your own field investigation

Find the rules where you live.

Does your country or state have expectations for data centres? Who has to report what, and how often?

Track down one real centre.

Find a data centre near you. Can you discover how it is cooled, or what it publishes about water or energy? If you can't find it, that's an evidence gap, and spotting it is the whole point.

Match a signal to a story.

Pick one signal from your build. Find one real news story where a data centre ran into exactly that problem.

Ask about consultation.

Does your council or state require a company to consult residents before a project like this goes ahead? Find out what "consultation" actually means where you live; it isn't the same everywhere.

Stay safe: use trusted news and government websites, and it's completely fine to finish with “we don't know yet.” That is often the most honest answer.

This is a teaching build. It does not assess a real development, and a cooling label alone cannot establish PUE, WUE, emissions or annual resource use.