Rethinking cooling strategies for Australian data centres

It doesn't matter where you are in the world, data centres are operating in the background to power your everyday life. These buildings are essential to the modern world, but these facilities require large amounts of energy to power them. In regions like Australia, where climate and regulations differ greatly from Europe and the US, designing efficient cooling solutions is not just about the best practices, it's essential. 

With hyperscale and colocation providers bringing more data centres to varied regions, understanding a fundamental design question is a big requirement: Should we be cooling with power or water? 

How do we cool today? 

Cooling systems remain one of the largest sources of energy consumption in any data centre. Traditionally, operators have chosen either air-based or water-based systems depending on climate, energy prices, and sustainability targets. In recent years, direct and indirect air solutions have been used due to their ability to lower power usage effectiveness (PUE). 

In a direct air system, outdoor air is passed through an evaporative humidifier and is then sent directly into the data hall. Indirect systems are different and instead use heat exchangers that reject heat to the atmosphere, preventing external air from entering the IT environment. These approaches help avoid running chiller compressors, which reduces electricity demand and maximises the power available for IT equipment. However, with sustainability metrics changing, our designs need to change too.

What about water? 

The industry’s focus on lowering PUE has at times obscured the impact of water use in data centres. However, this has recently become a greater focus for those in the industry, and there is pressure to reduce water use. With utility suppliers such as Sydney Water claiming that data centres could use the equivalent of 25% of the city’s yearly drinking water supply by 2030, water usage will become a greater concern to operators and consumers.

In regions like Australia, where we have seen states face scarcity of this resource, using water to cool data centres that could instead be used for agriculture or domestic purposes is a tough sell. Moreover, with the rise of high-performance computing and AI, the cooling needs of data centres are becoming more intensive. High-density racks and greater thermal loads are pushing operators to adopt liquid cooling systems that rely on chilled water loops. While this meets operational demands, it can increase the electrical load and raise the PUE, undermining the efficiency gains. However, the use of these air-cooled chiller systems can be a better solution in areas where water is scarce. 

Location should be a determining factor

The question of water or power should really be: what does the local environment require? 

Every region has different conditions in which it needs to operate. In California, water scarcity has led operators to focus their designs on using their reliable access to low-carbon electricity for cooling. Whereas in countries like Poland, which have access to water but face challenges with fossil fuel use, water can be the better option.  
In Australia, there is a complexity.

Given the size of the country, regions vary from state to state, with some having milder temperatures and reliable grid access, while others are more remote or face much hotter weather. These local factors need to shape the cooling strategy.

Australia has eight climate zones, including: 

• Tropical (North): Hot and humid summers, dry winters (e.g. Darwin, Cairns)
• Subtropical (East Coast): Warm, humid summers and mild winters (e.g. Brisbane, Sydney)
• Desert (Interior): Very hot summers, cold nights in winter (e.g. Alice Springs)
• Mediterranean (Southwest): Hot, dry summers and cool, wet winters (e.g. Perth, Adelaide)
• Oceanic (Southeast): Mild summers and cool, wet winters (e.g. Melbourne, Tasmania)

Australia’s climate is heavily influenced by El Niño and La Niña events. This sees a swing in Australia’s climate every 3-7 years and can last anywhere from 6 months to 2 years. La Niña promotes above-average rainfall and cooler temperatures in eastern Australia, while El Niño increases the likelihood of hot and dry conditions in Australia.

Other options: Heat reuse and seawater

Alternative strategies that have grown in popularity are waste heat recovery systems and the use of seawater for cooling. We’ve seen this increasingly used in areas in Europe.

Waste heat recovery is one option. This involves capturing and repurposing heat from IT equipment and redirecting it to nearby facilities or district heating networks. In Denmark, a hyperscale operator partnered with an energy service company to feed 165,000MWh of heat annually to a public heating system using heat pumps. This blueprint is impressive and attractive to operators looking to reduce costs but only works when stakeholders and infrastructure are aligned.

In Australia, some sites are well-positioned to take advantage of natural advantages like access to seawater or river water cooling. These systems use natural cold bodies of water to reduce reliance on energy-intensive compressors. This offers sustainability benefits; however, long-term viability and seasonality need to be assessed during the design period through modelling.

After all, the iconic Sydney Opera House is cooled using seawater taken directly from the harbour. The system circulates cold water from the harbour through 35 kilometres of pipes to power both the heating and air conditioning in the building.

Rethinking metrics

PUE has been the default metric for data centres, determining efficiency. The trouble is, this has its own limitations, particularly when looking at different regions and facilities.

Water usage effectiveness (WUE) has been pushed to offer a different metric that measures the water consumed per unit of IT power. However, this too can be distorted by the local climate and system choices. In turn, this makes comparing facilities in different regions harder as it’s not apples to apples. WUE also fails to account for the water used in power generation, so there is a hidden cost to the consumption measurements.

The other option is carbon usage effectiveness (CUE), which looks at the carbon intensity of both water and electricity. By looking at both in the same equations, we can get a more holistic view of the data centre’s efficiency and better identify the trade-off between power and water for cooling.

There is no one fit solution

Global colocation providers often rely on reference designs. These are standardised templates that can be adapted to bring similar data centres to new regions. However, while this saves time, applying the same cooling systems can lead to inefficiencies, high costs or project failures.

Understanding the impact that local conditions like the carbon intensity of the grid or the accessibility of water should be the factors that drive the design choices. To do so effectively requires an engineering team with expertise in multi-disciplinary projects so that they can work together to find solutions.

The National Australian Built Environment Rating System (NABERS) for data centres measures their energy efficiency and environmental performance to provide a star rating from one to six. Unlike design-based assessments, NABERS ratings are based on operational data to offer an accurate reflection of how efficiently a facility runs.

Overall, when bringing a reference design to Australia, choosing whether to use water or power for cooling largely depends on the type and location of the facility. Australia is so large and varies across seasons that when bringing reference designs to the region, they need to be even more localised, both in overall and cooling design.