Steaming intelligence – how artificial intelligence is devouring our water resources

In the 21st century, in an era of technological acceleration, artificial intelligence (AI) is becoming one of humanity’s most powerful tools in solving problems it has created for itself. AI promises increased efficiency, resource optimization, solutions in agriculture, medicine and climate crisis management. However, the technology, while intangible in its essence, has very physical requirements. One of these is water – a precious and increasingly scarce resource.

A paradox arises: AI can help combat water shortages, but at the same time it contributes to the shortage itself. This phenomenon requires our special attention, reflection and a strong response.

AI as a tool to combat the water crisis

Artificial intelligence is already playing a key role in combating the water crisis, offering advanced tools and solutions to help better manage water resources and minimize the impact of extreme weather events. One of the most important applications of AI is forecasting hydrological risks.

Advanced machine learning algorithms can analyze vast amounts of satellite and meteorological data, making it possible to predict floods, droughts and other weather extremes with increasing accuracy. An example of such a solution is Google Flood Hub, which supports flood warning systems in more than 80 countries around the world, giving residents and local authorities time to prepare for threats.

Another important area is the management of water supply infrastructure. AI-based intelligent systems can find even micro-leakage in pipelines, allowing immediate response and reduction of losses. A good example is the city of Dubai, which has implemented such solutions to quickly detect anomalies in water consumption and eliminate problems before they become serious deficits.

AI is also revolutionizing precision agriculture, where optimizing irrigation is crucial. Artificial intelligence systems monitor weather conditions and the actual needs of plants, allowing for precise watering. Such a method can reduce water use by up to 30 percent, which is invaluable, especially in deficit and drought-prone regions.

In addition, AI supports water quality monitoring by analyzing data from sensors placed in rivers, lakes or water supply networks, as well as from satellite images. This makes it possible to quickly detect pollution and identify its sources, enabling immediate corrective action and protecting ecosystems.

In theory, AI could become the foundation for digital transformation of water management, introducing a new quality of management and conservation. In practice, however, AI infrastructure’s growing demand for water means that the technology is also starting to become a co-conspirator in the problem. As the digital revolution proceeds and the number of data to be processed grows, so does the consumption of water for server cooling, which paradoxically could exacerbate the deficit.

The silent consumer: how does AI consume water?

Data centers are the backbone of the modern digital economy and artificial intelligence, but their operation involves huge energy and water consumption. Servers that run continuously require intensive cooling, as they generate a great deal of heat during operation. Water is the primary cooling medium in many such installations, making digital infrastructure one of the world’s largest consumers of it.

It is estimated that training large AI models such as GPT-3 requires the consumption of up to 700,000 liters of water over 2-4 weeks, equivalent to the annual water requirements of about 300 people. This huge appetite for water resources is due to the need to cool servers for weeks of intensive work on algorithms.

Global technology companies also consume huge amounts of water – Google, for example, needed more than 21 billion liters in 2022, and Microsoft 15.5 billion. With the rapid development of generative artificial intelligence, this demand is growing, as evidenced by the 2023 figures.

The problem is further exacerbated by the location of data centers. Many are located in water-stressed regions such as Arizona, Nevada and India. In these places, IT infrastructure competes with local communities for limited water resources, creating serious conflicts of interest and threats to the sustainability of these regions.

Another important aspect is that technology companies are not universally required to report on water use, and the available data typically appears in voluntary ESG standards reports. This limits transparency and makes it difficult to assess the true impact of digital infrastructure on water resources on a global and local level.

Sustainable standards for the digital revolution

The rapid development of artificial intelligence and digital infrastructure is forcing an urgent reflection on what standards should apply to their design and implementation. In the face of the water crisis, growing social and regulatory tensions, developing sustainability principles for the AI sector is becoming not only a matter of corporate responsibility, but also a condition for the survival of the data-driven growth model.

Green data centers are the first step toward responsible digitization. In practice, this means thoughtfully locating server rooms, preferably in regions with naturally lower temperatures – such as Iceland or Finland – which will reduce the need for water for cooling. Renewable energy is also increasingly being used: geothermal, hydroelectric or wind power.

Closed-loop cooling technologies – such as liquid loop cooling – reduce water loss to almost zero while increasing energy efficiency. Immersion cooling, which is gaining popularity among large IT infrastructure operators such as Equinix and Meta, works similarly.

Water offsets are an attempt to compensate for environmental damage. Microsoft has pledged to achieve water positive status by 2030, meaning it will return more water to the environment than it uses. It is accomplishing this by investing in wetland restoration, river restoration and funding retention projects. Google and AWS are undertaking similar initiatives, although these are still voluntary.

Transparency and reporting standards is an area where technology should catch up with regulation. Currently, companies only publish water consumption data when they want to – mostly as part of ESG. There is a lack of a mandatory measure like PUE (Power Usage Effectiveness) for energy. We need widespread introduction of WUE – Water Usage Effectiveness, which would allow real comparison of the ecological impact of different cloud providers and AI models.

The role of law and public policy

Technology is advancing faster than lawmakers’ ability to regulate it – a well-known paradox of the digital age. In the context of water consumption by AI and data centers, this gap can have real social and environmental consequences. That’s why the law should not only keep up, but even set the course toward sustainability.

The first step should be to include the digital water footprint in environmental decisions on new IT investments. Just as power plants or industrial facilities must provide environmental impact studies, data centers should demonstrate what their impact will be on local water resources.

Technical standards for digital infrastructure-especially for cooling-should include limits on maximum water consumption and guidance on water efficiency. These can be put in the form of national or EU standards, analogous to emission standards for cars.

Within the framework of European regulations, such as the CSRD (Corporate Sustainability Reporting Directive) or SFDR (Sustainable Finance Disclosure Regulation), it is possible to require the reporting of the WUE indicator and define so-called materiality thresholds for the impact on water resources. This will allow investors to better assess the actual environmental footprint of a given technology company.

The location of IT investments should also be regulated. In regions with high water stress – such as California, Spain and India – local governments should be able to restrict or condition data center construction based on water availability and community needs.

What’s more, public policy can play an inspirational role: by funding research and innovation in non-aqueous cooling, educating IT engineers about environmental impacts, or promoting public-private partnerships in sustainable digital infrastructure development.

Refrigeration innovation and a future without water?

Although today’s data centers consume millions of liters of water, technologies already exist that could change that. The future may belong to solutions that do away with water cooling systems altogether – or even redefine the way data is processed.

Closed-circuit liquid cooling, where servers are immersed in a special dielectric fluid, not only minimizes water consumption, but also increases computing density. Companies such as Meta, Google and Equinix are testing this solution in their busiest data centers.

Another innovation is locating data centers underwater to take advantage of natural deep-sea cooling. Microsoft’s project, which tested the operation of server rooms on the ocean floor, showed higher efficiency and lower failure rates than traditional solutions.

Air cooling in the polar regions is another strategy: in locations such as Sweden and Greenland, the outside temperature for most of the year allows the infrastructure to cool naturally, without the use of water and with minimal energy consumption.

And finally, AI that saves AI – algorithms that optimize themselves, reducing the number of computational operations without sacrificing the quality of the results, and thus reducing energy and water consumption. This approach is at the heart of the call for Green AI, promoted by academia and startups as an alternative to the current energy- and water-intensive model of artificial intelligence development.

The new ethics of technology

Digital transformation must not come at the expense of life’s basic resources. Water, though ignored by the AI sector for decades, is becoming its hidden currency. Every query to an AI model, every data analysis is not only energy, but also water.

We need a new ethic of technology, where innovation goes hand in hand with responsibility. AI can only help solve water crises if it stops driving them itself.

In the 21st century, the future of technology will play out not only in the cloud, but also…. using drops of water. That’s why we need to make sure that each of them matters.

Maja Czarzasty-Zybert, Ph.D. – Legal advisor and doctor of legal sciences in love with energy, yachts and motorsport, especially Formula 1. Graduate of postgraduate studies in nuclear energy at the Warsaw School of Economics. Member of the Governing Board of the Polish Committee of the World Energy Council and co-initiator of the “Energy is a Woman” program implemented by the Polish Committee of the World Energy Council, which aims to encourage women to work in the energy sector. She is a member of the Polish Nuclear Society and the European Nuclear Society.