Information recently obtained (by the Financial Times) has revealed that a huge spike in water consumption by dozens of facilities in Virginia’s “data-centre alley” likely means new initiatives to replenish or conserve water resources are urgently needed.
Usage Up By Two-Thirds
The county authority figures show that water consumption at the data-centres of hyperscalers which surround Ashburn, VA (which host a staggering 70 per cent of the world’s internet traffic daily) was up by nearly two-thirds between 2019 and 2023 – from 5 billion litres to 7 billion litres!
Hyperscalers
So-called hyperscalers are large-scale cloud service providers that offer massive computing resources and include companies like Amazon Web Services (AWS), Microsoft Azure, Google Cloud Platform (GCP), IBM Cloud, Oracle, and Facebook. These companies play a major role in the technology landscape, providing cloud infrastructure that supports a wide range of services and applications globally. However, due to the expansion of power and water-cooling-hungry AI-driven infrastructure and data-centres, the water consumption of these hyperscalers has significantly increased in recent years.
For example, water usage at Microsoft data-centres rose by 34 per cent between 2021 and 2022, driven by the need to cool denser AI server racks and, similarly, Google reported a 20 per cent increase in water consumption, using 19.5 million cubic metres of water in 2022.
Why Water?
AI workloads require highly efficient cooling systems to prevent overheating of servers, which run continuously and generate significant heat. The traditional cooling methods, therefore, often involve evaporative cooling, where water evaporates to absorb heat, lowering the temperature of data centre-equipment. This results in heavy water usage, especially those data-centres operating in warmer regions.
Water Demand Fuelled by AI Infrastructure Growth
As cloud computing and AI have expanded, the need for more water (and more efficient water usage) has grown, i.e. AI infrastructure growth has fuelled the spike in demand for water.
Issues
In addition to the sharp increase in water demand, there are in fact many other issues that need to be taken into account when looking at trying to tackle water usage. For Example:
– Water scarcity. Many data-centres are located in regions already facing water shortages or droughts. For example, in The Dalles, Oregon, a Google data-centre was criticised for using one-third of the city’s water supply in a drought-prone area. It’s easy to see how this places additional stress on freshwater supplies in regions where water is a finite resource.
– The environmental impact. The use of water for cooling in such large quantities, particularly in arid or drought-prone areas, can negatively affect local ecosystems and water availability for communities.
– An apparent lack of transparency. It seems that many companies are not transparent about their water usage, making it difficult to gauge the full impact on local water resources. Public reporting on water use, similar to energy use, remains inconsistent across the industry.
The Type of Water Used In Data Centres
One key issue that deserves special attention is what type of water is used for data centre cooling. For example:
– Freshwater. Most data centres rely on freshwater sources, which are used in cooling towers for evaporative cooling. However, freshwater is a limited resource, and overuse can stress local supplies.
– Recycled water. It is worth noting here, however, that some hyperscalers are now beginning to use recycled or reclaimed water to mitigate their environmental impact. For example, Amazon Web Services (AWS) uses recycled wastewater in its Virginia data-centres, helping conserve high-quality water for community use.
Research
Research by the British Standards Institution (BSI) and Waterwise’s “Thirst for Change” also makes some key points and recommendations that need to be considered when looking at the subject of data-centre use of massive amounts of water. It highlights the critical issues related to freshwater resource management, focusing on the growing urgency of water security in the context of global environmental challenges. Some of the key relevant points and conclusions from the research include:
– There is now a water security crisis. The research makes the point that freshwater is a finite resource, and the global water security crisis is just as urgent as climate change. Both population growth and increased demand for water, particularly in industrial sectors, are straining water supplies.
– The tech sector is still highly water-intensive, especially data-centres. With the rise of cloud computing, AI, and data-centres, the demand for water has skyrocketed, adding to the strain on limited freshwater resources.
– Water management and responsible water usage are now critical. The research emphasises the need for large-scale industries, including tech companies, to recognise their role in contributing to water scarcity and to adopt more sustainable water practices.
– There is a need for a circular economy in water usage, i.e. water recycling. One of the primary recommendations from the report is the need to transition towards a circular economy mindset in water use, particularly in sectors like tech. This involves recycling and reusing water wherever possible, reducing excessive freshwater extraction.
– Innovation in water efficiency is needed, i.e. water-efficient technologies, especially in data-centres. The research suggests that the wider tech sector needs to adopt innovative systems that support water reuse and reduce reliance on freshwater for cooling and other processes.
– Companies need to push beyond the environmental net gain of merely becoming water-efficient and to strive for a net positive environmental impact by replenishing water resources and engaging in water conservation initiatives.
Alternative Cooling Technologies
The recognition of the need for action in meeting the cooling requirements of a data-centre boom fuelled by the growth of AI, and for data-centres to reduce their reliance on water-based cooling systems has led to experimenting with several alternative technologies. The hope is that one or more of them could be viable ways to address both efficiency and environmental concerns. Examples of such innovations:
– Liquid cooling. This is increasingly being adopted to handle high heat loads generated by AI and high-performance computing. It includes two main methods, namely direct-to-chip cooling, e.g. circulating liquid directly over a system’s heat-generating components (e.g. CPUs and GPUs) using cold plates, and immersion cooling. This involves fully submerging servers in a dielectric (non-conductive) liquid that absorbs and dissipates heat. This technology can eliminate the need for air cooling entirely, offering higher efficiency, especially for dense computing environments.).
– Refrigerant-based cooling. This method involves using refrigerants instead of water. Refrigerant-based systems have excellent thermal conductivity, making them more efficient at transferring heat away from components. They are becoming popular for high-density racks and can be scaled to handle increasing workloads.
– Chilled water systems. Some data-centres continue to use chilled water, but advancements like rear door heat exchangers (RDHx) are improving efficiency. These systems use chilled water to cool the air before it enters the data-centre, but now take up less space and offer “room-neutral” cooling, meaning the air exiting the system is at near-ambient room temperature.
– Air-based free cooling. This method uses external ambient air, particularly in cooler climates, to reduce the need for mechanical cooling. This approach works best in regions with cold climates, and it’s already being used in data-centres in places like Sweden and Finland.
– AI-optimised cooling. Ironically, the AI that’s creating more heat can also be used to optimise cooling efficiency by predicting heat loads and managing energy use dynamically. AI can help balance the use of cooling resources more effectively, ensuring that the cooling system is only used when necessary.
Water Replenishment Programmes
It should be noted that one thing tech companies are increasingly investing in to help the situation is water replenishment programs. These are being used to offset their water usage, especially as data centres require significant cooling resources. As well as helping the tech companies to meet their sustainability goals and reduce water consumption, as the name suggests, these programmes are also designed to replenish water in communities, particularly in areas impacted by drought or water scarcity. Examples include:
– Amazon Web Services (AWS) which has implemented a range of water replenishment projects globally. For example, in 2023, its efforts returned 3.5 billion litres of water to local communities. AWS plans to expand this to over 7 billion litres annually across 21 projects, with initiatives in countries like the US, Brazil, Chile, and China. For instance, in Chile’s Maipo Basin, AWS is partnering with local farmers and using AI to improve irrigation efficiency, saving around 200 million litres of water annually. Similar AI-driven projects in Brazil are helping monitor water usage and soil quality.
– Microsoft is working towards becoming water-positive by 2030, aiming to replenish more water than it consumes. It has invested in over 49 replenishment projects worldwide, focusing on areas of high-water stress. These projects include restoring wetlands and repairing irrigation systems to improve water supply reliability. For example, in Mexico City, Microsoft is reviving traditional wetland agriculture, expected to replenish 3.8 million cubic metres of water over a decade.
– Google has committed to replenishing 120 per cent of the water it consumes by 2030. In 2023, its water stewardship projects have replenished over 1 billion gallons of water, addressing 18 per cent of its freshwater consumption. These projects focus on improving water quality and enhancing water efficiency across regions with high water scarcity.
All that said, critics might argue that water replenishment programmes often focus on offsetting usage rather than reducing consumption, making them more of a band-aid solution than a long-term fix for the growing water scarcity problem.
Energy-Hungry
In addition to their massive water demand for cooling, it should be acknowledged that data-centres are also known for their huge energy requirements, a situation that is also getting worse with the growing demand for AI infrastructure. For example, investment firm Carbon Collective estimates that the electricity currently used by data-centres could power around 6.5 million average (U.S.) homes!
What Does This Mean For Your Organisation?
As data-centres continue to expand and support the growing demand for cloud computing and AI infrastructure, their immense consumption of water presents a critical challenge that can no longer be overlooked. The surge in water usage, particularly in hyperscale facilities, means there’s now an urgent need for the tech industry to rethink its approach to sustainability. Relying heavily on water-intensive cooling systems is becoming increasingly untenable, especially as regions like Virginia and Oregon experience the strain of limited freshwater resources.
For businesses in the data-centre space, therefore, this trend highlights the necessity of embracing innovative cooling technologies, such as liquid cooling and AI-optimised systems, that reduce reliance on water while maintaining operational efficiency. Simultaneously, the shift toward using recycled water and investing in water replenishment programmes, as seen with Amazon, Microsoft, and Google, represents an important step toward more responsible resource management.
Ultimately, this evolving landscape presents an opportunity for tech companies to lead the way in sustainable water practices. By innovating and adopting circular water-use models, these businesses can mitigate their environmental impact, meet regulatory expectations, and build a more sustainable future for the industry. However, failure to act on this issue could not only jeopardise environmental sustainability but also risk operational and reputational challenges as resource scarcity intensifies.