Author: Mark Deguara, Technology Innovation Leader
Australia’s data centre sector is scaling as two infrastructure pressures converge. Industry estimates suggest its water demand could more than triple from roughly 5.5 GL to 17 GL over the next five years, while Sydney Water projects data centre demand could reach up to 250 megalitres per day by 2035.
At the same time, compute is becoming denser and more difficult to cool. NVIDIA’s Vera Rubin allows AI workloads to reach up to 227 kW per rack, increasing the amount of heat operators must remove.
This creates a critical design question: how do you support denser compute without increasing dependence on constrained water resources? For drought-exposed Australian sites, closed-loop cooling offers a way to reduce reliance on continuous evaporative water use while supporting high-density workloads.
Water dependency is becoming a data centre design risk
Data centres still account for a relatively small share of Australian water consumption. However, demand is heading upward, and everyone is watching how policymakers and utilities respond.
In March 2026, the NSW Government identified efficient and sustainable use of energy and water as one of its principles for data centre growth. Sydney Water’s stated servicing principles are also pushing data centres towards rainfall-independent sources such as recycled water and desalination, alongside water-efficient design.
Australia’s Millennium Drought provides an important precedent: prolonged scarcity can change how water is allocated, priced and managed. For data centre providers, water dependency can significantly influence site selection, infrastructure requirements, operating costs and water-servicing considerations.
For projects designed to operate across multiple generations of IT equipment, it must be determined how much ongoing water dependency to design into the facility for each additional megawatt of compute.
The advantages of closed-loop cooling
The water advantage of closed-loop cooling is rooted in recirculation. Rather than continuously consuming water through evaporation, coolant moves between IT equipment and the facility’s heat-rejection infrastructure in a controlled loop.
In a direct-to-chip architecture, cold plates capture heat from components such as CPUs and GPUs. A coolant distribution unit (CDU) manages the technology cooling loop and provides the interface with facility cooling infrastructure.
For drought-exposed sites, there’s a critical distinction between circulating water and consuming it. Open-loop evaporative cooling towers can lose significant amounts of water to evaporation, while closed-loop direct-to-chip systems recirculate coolant without continuous evaporative loss.
But “closed-loop” does not automatically mean zero facility water consumption. Heat captured by the liquid cooling loop must still ultimately be rejected, whether through dry coolers, air-cooled chillers, cooling towers, or hybrid heat-rejection systems. If the facility-side architecture relies on evaporative heat rejection, water will still be consumed. Project teams therefore need to evaluate the entire thermal path.
There is no single closed-loop cooling design for every Australian site
Australia’s variation in temperature, humidity, water availability and existing infrastructure makes cooling architecture inherently site-specific. For example, chillerless operation is often climate-dependent rather than universally appropriate.
Architects, engineers and specifiers should start with the site conditions and thermal requirements, then select the cooling architecture.
A high-density facility might combine direct-to-chip cooling with Uniflair chillers where mechanical cooling is required. Where chilled-water infrastructure is unavailable, heat dissipation units (HDUs) can reject heat to air, which is particularly relevant to water-constrained regional or remote sites.
Brownfield and colocation facilities may need a hybrid approach rather than converting entirely to liquid cooling, with rear door heat exchangers (RDHx) or direct-to-chip cooling supporting higher-density zones alongside conventional air-cooled racks.
Schneider Electric supports this site-specific approach with an end-to-end liquid cooling portfolio designed for different thermal architectures and deployment requirements. We can help configure the thermal architecture to suit the site’s climate, water constraints, existing infrastructure, and expected compute density.
Higher rack densities make water-resilient closed-loop cooling more urgent
While water scarcity is one aspect of the design challenge, rising compute density is another. NVIDIA’s Vera Rubin allows AI workloads to reach up to 227 kW per rack. At densities of this magnitude, operators need to remove substantially more heat without creating a corresponding increase in dependence on water-intensive cooling.
Direct-to-chip liquid cooling addresses the thermal challenge by capturing heat directly from high-power components rather than relying primarily on conditioned air. For colocation providers supporting mixed customer environments and neocloud operators planning for successive generations of accelerated computing, cooling planning must be viewed from a long-term capacity standpoint.
Experts recommend designing closed-loop cooling as a comprehensive system
Because each stage of the thermal system affects the next, closed-loop cooling should not be treated as a point-product procurement decision. With Schneider Electric, specialist liquid cooling capabilities are combined with an end-to-end liquid cooling portfolio and broader expertise across facility cooling, power, commissioning, and lifecycle services. Learn more about comprehensive direct-to-chip architectures at Schneider Electric’s Liquid Cooling Hub. For a closer look at how to select and design the right system for your facility, download the Liquid Cooling System Architecture Guide.
Add a comment