Data centers in Sydney could consume as much as 25% of the city’s annual drinking water supply within the next decade, according to Sydney Water, the region's primary water utility. Currently, these facilities use about 3.5 billion liters of drinking water each year, which accounts for less than 1% of the total demand. However, projections indicate that this figure could rise significantly by 2035.

For perspective, Sydney's desalination plant provides 15% of the city's drinking water, equating to approximately 91.25 gigaliters annually. As Australia aims to become a global hub for data centers, both industry and government are exploring alternative water sources to reduce reliance on drinking water.

The rise of artificial intelligence (AI) is a key factor driving the demand for new data centers. These facilities house the servers that store, process, and distribute vast amounts of data. While AI can help address environmental issues, it also presents challenges, particularly in terms of energy and water consumption. Data centers are essential for various services, including internet access, healthcare, and national security. The increasing use of generative AI is expected to further escalate the need for data centers, which require substantial cooling to prevent server overheating.

Cooling systems in data centers traditionally rely on large amounts of water, often lost through evaporation. Currently, Australia has over 260 data centers, with about one-third located in Sydney. All 89 data centers in Sydney draw from the public drinking water supply, a practice that Sydney Water acknowledges cannot continue indefinitely.

To address this issue, many data centers are investing in more efficient cooling technologies, such as closed-loop and immersion cooling systems. Some companies, like CDC, have implemented fully recycled water systems, developed during the Millennium Drought, which began in the late 1990s. This technology allows CDC to save up to 5 billion liters of water annually.

James Harrington, a senior project manager at Sydney Water, emphasized the importance of finding alternative water sources for data centers. He noted that wastewater treatment technology could provide the high-quality water needed for cooling systems. "As data centers become more prevalent, they require a lot of water for their cooling, and we need to make sure that we're enabling that," Harrington said.

Paul Plowman, acting CEO of Sydney Water, explained that the anticipated increase in water demand from data centers is based on projected growth in processing power. He reassured residents that plans are in place to ensure drinking water supplies remain unaffected. "It's really important that people understand that there are many ways to provide water, and drinking water isn't always the answer," Plowman stated.

Wayne Rylands, CEO of Sydney's Ryde Council, expressed concerns about the rapid growth of data centers in the Macquarie Park area, where 11 facilities are either built or planned. He warned that the number could rise significantly, leading to increased water and power consumption. According to Sydney Water, data centers in Macquarie Park could use up to 30 megaliters of water daily within a decade, representing about 2% of the city’s current daily drinking water demand.

"We're not against data centers; it's about the location of the data centers," Rylands said. Plowman added that situating data centers near large sewer systems could help mitigate the impact on drinking water supplies.

Danielle Francis from the Water Services Association of Australia highlighted the potential for Australia to become a resource-efficient data center hub. "The growth of data centers creates a unique opportunity to help grow our national economy and productivity, but it will create extra demand for valuable water resources," she said.

The environmental impact of AI usage is complex. A recent Google report indicated that a median text prompt from its Gemini AI used about five drops of water and the energy equivalent of nine seconds of television. However, estimates vary widely, with some research suggesting that using AI models like ChatGPT-3 could consume up to 500 milliliters of water for multiple responses.

AI expert Professor Jon Whittle from CSIRO noted the importance of considering the entire lifecycle of AI technology. "When you look at the environmental impact of AI, you have to look at the full life cycle," he said. He pointed out that the training of AI models and the construction of data centers also contribute significantly to their environmental footprint.

The Pawsey Supercomputing Research Centre in Perth showcases innovative cooling solutions. Its supercomputer, Setonix, employs geothermal cooling, saving over 7 million liters of water annually. Executive director Mark Stickells explained that while this method is effective, it requires specific site conditions and a higher initial investment.

Despite being one of the most environmentally friendly systems, Setonix still consumes about 1.5 megawatts of power, comparable to a small suburb. Stickells emphasized the need for awareness regarding the infrastructure and resources required to support AI technologies.