Power gets the headlines in the data centre debate, but water is increasingly the input that decides whether a site can be built at all. Cooling is one of the largest non-IT resource draws in a facility, and the most common cooling methods spend water to shed heat. As campuses scale into the tens and hundreds of megawatts, the volume of water a site needs has become a planning constraint that sits alongside grid power, not below it.
This guide sets out how much water data centres actually use, where that water goes, why dry regions face the sharpest trade-off, and how water status now shapes site selection in Australia and other mature markets. It also explains why the public record on water filings and environmental referrals is one of the earliest signals that a site is real. Data Centre Axis tracks that record and takes no position in any transaction.
How much water does a data centre use?
Consumption varies enormously with cooling design and climate, from close to zero for closed-loop and air-cooled systems to several million litres a day for a large evaporatively cooled campus. As a reference point, hyperscale operators that report their figures have averaged roughly two million litres of water a day per large site, rising in hot weather when evaporative cooling works hardest. Put another way, a single megawatt of cooling load can consume up to around 25 million litres a year.
Across a whole market the totals are large. In the United States, data centres are estimated to draw on the order of 1.7 billion litres of water a day, although that remains a small share of national consumption. The harder problem is concentration: a cluster of large sites in one catchment can place real pressure on a local supply even when the national figure looks modest.
Direct and indirect water: two figures that matter
A site's water footprint splits in two, and both belong in any honest assessment. Direct water is what the facility draws on-site to cool its servers, mostly through cooling towers that evaporate water to reject heat. Indirect water is consumed upstream at the power stations that generate the site's electricity, since most thermal and hydroelectric generation uses water. The indirect figure is often larger than the on-site draw and is easy to overlook, because it does not appear on the operator's own meter.
This matters when comparing designs. An air-cooled site that uses almost no water on-site can still carry a large indirect footprint if it sits on a water-intensive grid, because it burns more electricity to move heat with air. Water usage effectiveness, the standard on-site efficiency ratio, only counts the direct draw, so it can flatter a site that simply pushes its water use upstream.
Why dry regions face the hardest trade-off
Evaporative cooling is most effective in hot, dry air, which is exactly where water is scarcest. That is the central tension in water-stressed regions. The engineering logic pulls an operator toward evaporative designs that cut power use and improve PUE, while the community's water security pulls the other way. A design choice made at approval stage tends to lock in a site's water profile for 20 to 30 years, because switching an operating facility from evaporative to closed-loop cooling is effectively a rebuild rather than an upgrade.
How water shapes where data centres get built
Water access now sits beside grid connection as a primary feasibility test. A site with abundant power but no secure water allocation may never reach approval. In Australia, water is governed at the state level and tied to catchment conditions that tighten in drought, so a proposed site's projected draw can become a live planning issue, particularly in regions already under allocation pressure.
New South Wales shows the dynamic clearly. It holds the largest concentration of data centres in the country, and Sydney Water has estimated the sector could account for 15 to 20 percent of the city's water supply by 2035. Federal expectations and a state consultation process are now pushing operators to evidence their cooling-water sourcing before consent, and planners increasingly scrutinise the cooling method rather than treating water as a footnote. The same pattern plays out in the constrained corridors of the United Kingdom.
Can operators cut their water use?
Yes, and several proven approaches are in use. Closed-loop liquid cooling recirculates coolant instead of evaporating it, cutting on-site freshwater draw sharply. Air-assisted and free-air designs use little or no make-up water, at the cost of higher energy use. Some operators draw on recycled or non-potable water, using treated wastewater rather than drinking supply, and direct-to-chip cooling moves heat at the source with far less water than open evaporative towers.
Each option carries a catch. Air cooling raises electricity use and so the indirect water footprint. Recycled water needs nearby treatment and distribution infrastructure. Cooler climates that suit free-air cooling may sit far from the population centres that need the compute. There is no universal answer, only the right design for a given catchment and climate.
Following the public record
Because water is decided early and locked in for decades, the documents that reveal a site's water plan appear long before any sale. Water utility filings, state planning assessments and environmental referrals are public signals that, read alongside grid and planning data, show a site's full feasibility picture rather than the version in a brochure. A powered site without a credible water plan carries real planning risk, and that risk is usually visible in the record to anyone who reads it. This is why water filings deserve the same attention as grid connection status when assessing, buying or developing a data centre asset.
Frequently asked questions
How much water does a large data centre use?
A large evaporatively cooled site can use several million litres a day, with reporting hyperscalers averaging around two million litres daily per site and more in hot weather. Closed-loop and air-cooled designs use far less, sometimes almost none.
Why do data centres use water at all?
To cool the servers. Most common cooling methods reject heat by evaporating water in cooling towers, which is cheap and effective but consumes water that does not return to the local system. Water can be replaced by air-based or closed-loop cooling at the cost of higher electricity use.
What is the difference between direct and indirect water use?
Direct water is drawn on-site for cooling. Indirect water is consumed at the power stations generating the site's electricity. The indirect figure is often larger and is missed by on-site metrics such as water usage effectiveness.
Is data centre water use a problem in Australia?
It can be, in water-stressed catchments. Water is allocated at the state level and tightens in drought, and in New South Wales the sector could reach 15 to 20 percent of Sydney's supply by 2035. A site's projected draw is now assessed in planning and can decide whether it is approved.
Where can I find a site's water information?
In the public record. Water utility filings, state planning assessments and environmental referrals set out a proposed site's cooling method and projected draw, often years before construction. Read alongside grid and planning data, they reveal whether a site is genuinely feasible.