Author: Mark Deguara, Technology Innovation Leader
Australia’s data centre market is entering a new phase of growth. As operators build new facilities and retrofit existing ones to support increasingly dense GPU workloads, cooling has become one of the most important infrastructure decisions.
Energy demand often dominates the conversation, but water use is equally important. Utilities, regulators, and local communities are paying closer attention to how new data centres will affect water resources, particularly in fast-growing regions such as Western Sydney and Melbourne, where infrastructure is already under pressure. At the same time, operators need to expand AI capacity. Data centre water cooling is now a strategic planning issue.
The good news is that supporting higher-density AI workloads doesn’t have to mean consuming significantly more water. Advances in liquid cooling technology are making it possible to remove heat more efficiently while dramatically reducing water consumption.
Why water efficiency is becoming a design priority in Australia
Australia presents a unique set of challenges for data centre operators. The country has become one of the world’s most attractive destinations for data centre investment, yet many of its fastest-growing digital infrastructure hubs are located in regions where water availability is a sensitive issue.
State governments, water authorities, and infrastructure planners are examining how large-scale data centres can continue to grow while making more sustainable use of local water resources. Recent policy discussions have emphasised efficient water use, greater transparency around water consumption, and increased use of recycled or non-potable water where appropriate.
Industry figures show that data centres use 0.04 per cent of Australia’s potable water nationally. However, demand is highly concentrated in specific growth corridors, creating localized challenges that require careful planning rather than one-size-fits-all solutions.
Water stewardship is becoming another design criterion alongside power availability, rack density, cooling performance, and total cost of ownership. Decisions made early in the design process, from cooling architecture to water source selection, can have long-term implications for operational efficiency, sustainability reporting, and future expansion.
Four planning decisions that improve water efficiency
As AI workloads continue to push rack densities higher, improving water efficiency requires evaluating how cooling fits into the broader facility design, from the cooling architecture itself to future expansion plans. The following considerations can help operators build AI-ready infrastructure that supports both performance and responsible water use.
1. Design for closed-loop cooling wherever practical
One of the most effective ways to reduce water consumption is to minimise or eliminate evaporative cooling. Traditional evaporative cooling relies on water evaporation, resulting in substantial water loss to the atmosphere. By contrast, closed-loop liquid cooling can recirculate more than 95% of its coolant, with only minimal losses during maintenance or in the event of leaks.
This approach also provides greater predictability. Because the cooling loop is isolated, operators can better manage water use while maintaining consistent thermal performance for high-density workloads. As organisations evaluate new builds or retrofit existing facilities, understanding where closed-loop cooling can be incorporated should be one of the earliest design decisions.
2. Match the cooling architecture to rack density
Not every workload requires the same cooling strategy. Lower-density applications may continue to operate efficiently with air cooling. AI training, inference, and other GPU-intensive workloads are a different story. As rack densities climb beyond what air alone can support (between 40 and 50 kW/rack), removing heat directly from processors becomes necessary.
Comprehensive direct-to-chip liquid cooling architectures transfer heat away from CPUs and GPUs at the source, allowing operators to support higher-density AI deployments without relying solely on larger room-level cooling systems. Rather than replacing air cooling entirely, many facilities adopt hybrid cooling strategies that use liquid cooling to remove heat directly from the chip, while air cooling manages heat from the remaining IT equipment.
3. Consider where your cooling water comes from
Across Australia, governments and water providers are encouraging large infrastructure projects to prioritise recycled or non-potable water where appropriate, helping preserve drinking water supplies for communities. Several major data centre operators have already adopted this approach, integrating recycled water into their cooling strategies while continuing to support growing AI capacity. For organisations planning new facilities, evaluating available water sources early in the design process can improve long-term sustainability while helping align with evolving regulatory expectations.
4. Plan for tomorrow’s demands
Organisations that size cooling infrastructure only for current workloads may face expensive upgrades as rack densities increase or new generations of processors arrive. Instead, cooling strategies should provide flexibility to accommodate future expansion, higher heat loads, and changing technology requirements without requiring a complete redesign. Planning for scalability should extend beyond cooling equipment into power distribution, facility layout, monitoring, and operational processes.
Planning for AI growth requires looking beyond cooling alone
As organisations deploy higher-density infrastructure, decisions about cooling become closely connected with power distribution, facility capacity, monitoring, and long-term operational resilience. Optimising one system in isolation often limits the performance of the entire facility.
That’s why many operators are shifting from evaluating individual cooling products to planning integrated infrastructure strategies that consider power and thermal management together. Taking this broader view helps organizations accommodate higher rack densities, improve efficiency, and build flexibility for future AI growth while supporting sustainability objectives.
Ultimately, water-efficient cooling is most effective when it is considered as part of the overall infrastructure design. By making cooling decisions within the context of the complete data centre ecosystem, operators can better balance performance, resource efficiency, and long-term scalability as Australia’s AI market continues to grow.
Building a smart data centre water cooling strategy
As Australia’s AI infrastructure continues to grow, cooling strategies will play an increasingly important role in balancing performance, sustainability, and long-term operational resilience. While liquid cooling is becoming essential for supporting higher-density AI workloads, it does not inherently mean high data centre water use. Decisions around cooling architecture, water source, and facility design can significantly influence both water efficiency and future scalability. Planning for these factors early allows operators to support growing compute demands while aligning with evolving regulatory expectations and sustainability goals. For organisations evaluating their next AI data centre project or retrofit, explore Schneider Electric’s Liquid Cooling Hub for guidance on comprehensive direct-to-chip architectures, including coolant distribution units (CDUs), closed-loop cooling, and greenfield and retrofit planning. For a closer look at the available system options, download the Liquid Cooling System Architecture Guide.
Add a comment