Why circular water solutions are key to sustainable data centres

In an era where digital connectivity shapes every aspect of our daily life, the demand for data centres has reached unprecedented levels. These expansive facilities store and process immense amounts of information that fuel our modern world, making them indispensable to today's digital infrastructure.

Whether it’s browsing on social media platforms, streaming services or running complex artificial intelligence (AI) models, data centres are the hidden powerhouses enabling the seamless digital experiences we rely on every day.

However, data centre operations come with a significant environmental cost, particularly regarding materials, energy and, most notably, water usage. Even a small-scale data centre can have a substantial water footprint.

For instance, a 1 megawatt (MW) data centre can use up to 25.5 million litres of water annually just for cooling – equivalent to the daily water consumption of approximately 300,000 people. This water consumption exacerbates water stress, especially in vulnerable regions already facing shortages.

AI consumes a large amount of energy to train complex data models. When energy is used at those levels, water is needed to cool the machines doing the AI processing. The rapid growth of artificial intelligence and the expansion of associated data centres will only intensify this challenge. GPT-3, an AI model, is estimated to consume 500 ml of water per 10-50 responses. When multiplied with billions of its users, the total water footprint of AI becomes enormous.

Though data centre operators have started to understand and measure their water consumption, it remains an often-overlooked aspect of core data centre operations compared to carbon. This emphasises an important opportunity for scaling strategic circular water management solutions, such as water replenishment and optimization, in data centres.

Data centres consume water for cooling in systems such as chillers, cooling towers, and liquid cooling systems, and indirectly through electricity generation that often involves water-intensive processes like steam generation. Annually, a medium-sized data centre (15MW) consumes as much water as the yearly consumption of either three average-sized hospitals or more than two 18-hole golf courses.

A key reason for the high-water consumption is limited water reuse in cooling. During the cooling process, part of the freshwater evaporates, and the remaining water becomes wastewater. Wastewater is often contaminated with dust, chemicals and minerals, which hamper the efficiency of the cooling process if circulated back. Consequently, data centres are often unable to reuse wastewater to their maximum capacity.

The growing demand for digitalization and AI-enabled services has increased our dependency on data centres, driving high levels of computational power and energy consumption. Advanced AI models, such as GPT-3 and GPT-4, require immense computational resources that, in turn, drive up water consumption.

By 2027, global AI demand is expected to account for 1.1 to 1.7 trillion gallons (4.2 to 6.6 billion cubic metres) of water withdrawal, more than 4-6 times the total annual water withdrawal of Denmark.

Cooling dictates water consumption in data centres, but water conservation is still often ranked as a low priority. Less than a third of data centre operators actively track water usage metrics, and this lack of transparency and measurement significantly undermines efforts to understand the full environmental impact of data centre operations, especially in water-stressed regions.

Similar to how power usage effectiveness (PUE) has become the standard metric for measuring energy efficiency in data centres, water usage effectiveness (WUE) was introduced to measure water consumption efficiency.

However, WUE only accounts for on-site water use, ignoring the substantial indirect water consumption associated with electricity generation, which often relies on water-intensive processes like steam production in thermoelectric power plants.

By only looking at on-site usage, an operator is failing to capture their true water footprint. This highlights the urgent need for more comprehensive metrics like WUE that include both direct and indirect water use.

As AI models continue to drive demand for computational power, prioritizing water tracking and management is essential to ensure the sustainability of data centre operations. Data centres are making progress to adapt to advanced cooling technologies, which includes liquid cooling. This method uses a liquid coolant to efficiently dissipate heat from components, offering more efficient heat management compared to traditional air-cooling methods.

Water usage in data centres is not just a technical challenge; it is an urgent environmental and societal issue. The socio-economic well-being of a region is imperilled when data centres increase their reliance on scarce water resources.

For example, farmers face lower crop yields because water for irrigation becomes scarce and water prices often increase. Some cases lead to social unrest as seen in Holland, Chile and Uruguay. In early 2023, a hyperscaler’s plans to build a large data centre in Uruguay sparked protests. Residents, already suffering from severe droughts, opposed the development, fearing it would further threaten their limited access to safe drinking water and worsen agricultural losses.

While the growth of data centres will not slow down, the latest advancements in technology offer a hopeful outlook for the future.

Data centre design is intricately linked to the evolution of AI, processor technology and water management innovations. Data centre operators recognize water consumption as a critical challenge and are now setting ambitious water-positive targets, which drive them to incorporate water management as a key consideration in their standard operations.

This proactive approach ensures that water sustainability becomes an integral part of their business practices, helping mitigate environmental impacts and support long-term operational efficiency.

To further mitigate the broader environmental and societal impacts of data centres, operators can start adopting innovative circular water solutions. These solutions involve reusing and recycling water to create a sustainable and efficient water management system.

Practices such as closed-loop cooling systems, involving wastewater recycling or rainwater harvesting have already shown significant reductions in freshwater use, with potential savings of 50-70% where these systems are implemented.

The integration of circular water management principles represents a critical step towards ensuring data centres remain capable of supporting the next wave of technological advancements in a thirsty world.

Over the past three years, Accenture, The World Economic Forum, and UpLink have partnered to accelerate circular economy innovations, alongside partners such as Ecolab, Anglo American and AWS. To find out more about The Circulars Accelerator, visit UpLink.