The weight of the cloud: the cost of AI for water
When people talk about “the cloud”, it sounds weightless, something almost magical floating in the sky. Except that it isn’t. It lives on the ground, in data centres filled with servers that store, process and move staggering amounts of information every second. These data centres consume energy, occupy land, use metals and minerals from global supply chains – and they need vast amounts of water for cooling.
AI has a water footprint
By 2027, AI is projected to require between 4.2 and 6.6 billion cubic meters of water, roughly the equivalent of 1.7 to 2.6 million Olympic-size swimming pools. Many data centres are being built in places already facing drought, desertification or water stress, which raises a simple but uncomfortable question: how much water does AI need to power our digital lives?
In Uruguay, a planned data centre sparked public concern due to its projected water use: around 7.6 million litres per day, roughly equivalent to the daily needs of about 55 000 people. Access to information about the project reportedly required legal action invoking the constitutional right to water and international environmental agreements. This is why the debate on AI and digitalisation cannot be framed only around innovation, productivity or convenience. It must also include questions about how much water is used, under what conditions, and with what local implications.
The paradox: AI can help solve water challenges – but it can also aggravate them
To be clear, the problem is not technology in itself.
AI can deliver just as many benefits in water management as it can in other areas. It can help predict floods, detect leaks, improve irrigation efficiency, monitor pollution, and optimise water distribution. These are significant opportunities, especially in a world facing more climate stress and growing pressure on water systems. But, in producing those outputs, AI systems consume water. If the environmental and social costs of AI’s physical and digital infrastructure are ignored, any benefits they deliver in smarter water management may evaporate.
What is needed?
First, more efficient water management is necessary. Improving infrastructure performance through better insulation, more efficient servers, dry cooling, reclaimed water use, water recycling and closed-loop systems matters. Some places are already recognising this. In order to reduce pressure on an already constrained grid and align new development with energy security and decarbonisation goals, in Greater Dublin, Ireland, developers must demonstrate access to electricity capacity, integrate renewable energy and ensure demand flexibility before data centre development projects can be approved.
Second, rethinking digital infrastructure models at the territorial level can help reduce their environmental and social impacts, while improve their cost-efficiency. In practice, they should be planned and governed in relation to the needs, resources, and limits of a specific place, taking into account interactions with local energy systems, local communities, and local ecosystems. From this perspective, alternatives include smaller-scale and decentralised systems, local or household-based infrastructures, bioeconomy-based infrastructures and forms of public ownership that give territories more control over how digital systems are developed and used. In Sweden, for example, Jämtland’s new data centre reuses waste heat to warm nearby greenhouses.
The future is not about choosing between technology and no technology.
Used well, AI can help societies manage water more fairly and sustainably. But this has a cost. The cloud was never weightless. It heavily takes land, energy and a lot of water to run. Let’s make water part of the AI story.
Find out more about our work on Water governance and The OECD Principles on Water Governance and implementation strategy. Dive deeper into related dataThe OECD Water Governance Indicator Framework.
Ander Eizaguirre is a Policy Analyst at the Water Governance, Blue and Circular Economy Unit, within the Cities, Urban Policies and Sustainable Development of the Organisation for Economic Co-operation and Development (OECD). Since 2022, he has coordinated the OECD contribution to the Circular Cities and Regions Initiative (CCRI) of the European Commission, which includes 10 case studies and a synthesis report. Since Ander joined the OECD in 2019, he has taken part in over 15 policy dialogues on the circular economy and water management at national and subnational level in different countries. Before moving to the OECD, he worked on climate and energy issues at the Spanish Confederation of Business Organisation’s (CEOE) Permanent Delegation to the EU. Ander graduated in Economics from the University of the Basque Country and has a Master’s Degree in Economic Research from the UNED (National University of Distance Education) and an MBA from Deusto University.
Enol Nieto Jiménez has a degree in Law and Political Science and a Master's degree in Contemporary Latin American Studies. He has researched in the areas of Political Ecology, socio-environmental conflicts and ecosocial crisis. His latest works have focused on the human right to water, its ontologies and social movements, carrying them out through community and participatory methodologies. He has also worked in the legal field, getting involved in open litigation in human rights and environmental law, in different countries of the global south and in Spain. From that place he will provide an interdisciplinary approach in the TC BETA.
At the BETA Technology Center he is a senior technician in the area of Governance for Sustainability, managing European projects, which seek to establish the link between science and public policies, between international researchers and generate impact in the rural territory. All this in the areas of environmental and ecological transition.