Special report · Energy & tech
The cloud has a smokestack
Every search, stream and AI prompt runs through a building that never sleeps. Australia is racing to build more of them than ever — and most still run on a grid powered by coal and gas.
Animated illustration: a data centre drawing coal-fired power, venting heat and running diesel backup.
Invisible infrastructure, very visible emissions
From the outside, a data centre is a windowless box on the edge of town. Inside, tens of thousands of servers draw power around the clock, and industrial chillers work just as hard to stop them from overheating. The pollution doesn’t come out of the building — it comes out of the power stations feeding it.
The AI boom is multiplying that demand. Operators are lodging proposals measured in gigawatts, not megawatts, and each new campus locks in decades of electricity and water use before a single renewable project is guaranteed to match it.
1,630 MW
running load across six clusters
9.6 Mt
CO₂e a year from the grid power they use
2.4 million
homes’ worth of electricity
4,220
Olympic pools of water evaporated for cooling each year
A data centre on a coal-heavy grid is a coal customer with a tech logo.
Climate Council analysis
Six clusters, one grid
Where a data centre plugs in matters as much as how big it is. Select a location to see its load, the pollution from the grid behind it, and the water it draws.
Interactive map of six data centre clusters.
It doesn’t have to be this way
South Australia shows the alternative: the same server load on a grid running mostly on wind and solar produces a fraction of the pollution. The difference isn’t the technology inside the building — it’s the rules around what it plugs into.
- New capacity, new renewables: every new data centre should be matched by new, additional clean energy on the same grid.
- Hour-by-hour, not annual offsets: 24/7 matching stops paper claims hiding night-time coal use.
- Public reporting: energy, water and backup-diesel use disclosed site by site.
How we calculated these figures
Running load × 8,760 hours × state grid emissions factor. Homes equivalent assumes 6 MWh per household per year. Water assumes evaporative cooling at each site’s reported or estimated rate.