Sustainable AI: Mitigating water risks in the data centre boom

This is the third article in our series in which Emily Owen, Associate Hydrologist examines the environmental impact of water usage in data centres.

The growth in data centres and AI and the impact and challenges for water

Everywhere you turn from the pub to LinkedIn, Artificial Intelligence (AI) is a hot topic. Just as the steam engine ignited the industrial revolution, generative AI is being touted as the catalyst for a transformational era of economic growth and improved living standards. The race has started for governments, businesses and individuals to become global leaders. But in the rush we must use old-fashioned human intelligence and learning from the past to ensure development of data centres and AI infrastructure does not destroy our precariously balanced natural capital. In this period of business focus on ESG, it would be convenient to assume that the impact of generative AI on our natural water capital will be subject to the same level of scrutiny as energy use and carbon footprinting.

However, water is complex, with risks and opportunities unique to each geographical location, and impacts throughout the value chain which are often unmonitored and not fully quantified. This is coupled with the economic challenge that water can require collective actions outside of individual operations, which does not always guarantee individual returns. There is also a common viewpoint that fixing water problems is a central or regional Government issue to solve.

Freshwater scarcity has become one of the most pressing challenges shared by all of us in the wake of a more extreme climate, rapidly growing population, depleting water resources, and aging water infrastructures. With the UK government releasing their AI Opportunities: Action Plan [1] and the launch of the international Coalition for Environmentally Sustainable Artificial Intelligence at the AI Action Summit in Paris, let’s dive into a few key considerations. For meaningful consideration of water management, it is necessary to consider impacts on regional scales, but it should be remembered that these are indicative of global trends with shared potential risks.

The challenge

The most commonly discussed element of water impact with data centres is the water used for cooling. In a thought piece [2] the World Economic Forum estimated that a 1MW data centre could use up to 25.5 million litres a year for cooling. There are emerging technologies such as liquid immersion cooling and free cooling (using outside air to ventilate IT rooms) that minimise the consumption of fresh water, however completely avoiding water use is difficult to achieve. Additionally, implementing new technologies to minimise the consumption of fresh water, can push up energy consumption. A deeper understanding of the AI data centre water-energy nexus is still the key ingredient in delivering on an array of operational objectives; from efficiency and sustainability to operational excellence and business continuity, to financial performance.

One of the most overlooked aspects of water consumption associated with data centres is indirect water use, which is also connected to their energy consumption. The estimated water footprint at the average thermal power plant globally is between 220 – 5,076 litres per megawatt produced [3]. Even as the world transitions to clean energy, water consumption for energy will remain high. Nuclear power, which could provide a significant baseload of electricity generation, requires slightly more water per megawatt than thermal power plants. Whilst solar and wind power have a more modest water footprint, it is unlikely that all parts of the world can move away entirely from large energy generation approaches within the near future.

The total water footprint from data centres including the indirect water use from energy was estimated in 2023 by an academic paper [4] as being about 500 litres per 20 to 25 queries answered by AI. However, given there is no environmental regulation reporting on water use for data centres in the UK and data is limited internationally, more recent research is indicating that this footprint could be massively underestimated. Whilst moves to net zero will reduce the reliance on thermal power plants in many parts of the world, the water consumption associated with the electricity for data centres will remain a significant burden for a large proportion of a data centre’s anticipated lifetime.

Looking out the window at another drizzly winter’s day in London, it's easy to forget the pressure already placed on water resources in the UK, particularly in the southeast. The UK government has already committed £1.6 billion of investment in water resource infrastructure aiming to mitigate the growing drought risk and deficit in water supply [5]. In a scenario where all water companies in England have incorporated ambitious demand reduction strategies into their water resource management plans, and with planning policies mandating water-neutral development alongside a new housing target of 1.5 million homes within five years, this investment in water resource infrastructure may only suffice to maintain current levels. Consequently, it may not be sufficient to support a significant rise in water consumption linked to the government's AI ambitions.

Whilst US federal spending is going through much review and change, the Biden administration had announced a US$242 million water infrastructure for western states including California, Colorado, Washington and Arizona [6]. It is unknown if this spending takes data centre consumption into account, however it helps us understand the cost of infrastructure to meet water demands.

Whilst licensing requirements vary by country, the overarching aims behind regulation and challenges associated with the cumulative impact of water users apply globally. New data centre developments therefore need to consider water an essential part of feasibility in choosing the location of any potential development.

Part of this conflict is driven by data centres' perceived reliance on potable water supplies. As with most production, the indirect water consumption through the supply chain remains an obstacle to understanding the full impact of construction that reaches across the globe, particularly in light of the rapidly escalating demand for rare earth minerals such as dysprosium, neodymium, and yttrium.

Flood risk zones and data centres

Following recent extensive flooding in parts of Europe it does feel strange to focus on the lack of water. So, turning the problem on its head a moment, let’s discuss flood risk. Space is becoming a leading issue for many countries, particularly within the more densely populated regions. Climate change is increasing the area of land at risk from flooding and the requirement for increased public spending to improve the standard of flood protection afforded by existing flood defences.

As an example, 6.3 million properties in England have been identified as being at risk of flooding from rivers, the sea, or surface water [7]. In the fight for space, utilising lower-risk flood zones is becoming more widely considered as a realistic development scenario. A number of data centre projects are now coming forward within flood plains and with suitable embedded design principles to mitigate any adverse impacts on flood risk, this can be appropriate. However, this needs careful consideration at the outset and early engagement with local authorities and statutory consultees to understand flood dynamics, quantify baseline flood risk, identify potential impacts and provide appropriate mitigation to prevent significant adverse effects to the development itself and everyone else.

Data centre siting and funding could bring benefits too

These constraints can give a negative perception of the data centre economy and rapid adoption of AI, however, life is never that simple. The reality is that the wave of AI can bring many positives for the water environment (and not just because as hydrologists we don’t have to do the mundane data processing ourselves).

As with all new industries, if the economic model set out takes into account the total value when considering the costs, additional benefits can be achieved. By siting data centres more intelligently based on potential synergies with adjacent commercial and industrial development, the opportunities to deliver enhanced benefits back to communities and the wider environment are considerable. For example, there are areas where upgrades to the foul water network are overdue and a data centre located appropriately can utilise treated effluent rather than potable water for cooling. Not all water is the same, and there are industrial sites that use and output water that is of a very high quality. Obvious restrictions for reuse in industries such as production of goods for human consumption may have prevented the facilities from recirculating their water, but with appropriate treatment there is the potential for data centres to make use of this wastewater.

Additionally, there are green funding opportunities. Where data centre developers and operators are expected to contribute back the total cost of their water consumption, there can be an injection of much-needed financial capital to address catchment wide water problems. This is an approach already being undertaken by the major data companies such as Google, Meta and Microsoft. It is crucial that stakeholders are engaged (and this cuts both ways) so that this funding reaches to where it delivers the greatest benefit. Too often, private companies seeking collaborative action are overwhelmed with where to start building partnerships and accessing these opportunities. All catchment stakeholders need to avoid escalating previous confrontational views and focus on how we can deliver water neutrality collectively.

Five key actions

So, what are the take-home points to deliver success for the water environment through this technology revolution?

1. Water needs to be a material consideration during feasibility and due diligence work for data centre projects, encompassing flood risk and alternative sources of water;
2. Catchment stakeholder engagement should be undertaken early to understand constraints and identify opportunities for synergies with adjacent commercial and industrial operators;
3. Assessment of capacity within the potable water network or licence availability should be a key consideration in locating sites;
4. The total economic value of water (not just the price paid to the water company) needs to be incorporated into cost-benefit analysis; and
5. Integrated design principles encompassing energy and water are required to ensure that substantial gains in one area do not come at the cost of the other.

How SLR supports data centre developers and operators

We support the sector by providing turnkey advisory solutions to the challenges of changing regulatory landscapes, carbon footprint reduction, energy supply and increasing energy efficient technologies, reducing water use, and the consumption of natural resources. ​

SLR is uniquely placed to provide solutions through the full life cycle of a data centre and our network of technical experts support clients in quickly bringing projects from concept to feasibility and planning to delivery.​ Our breadth of expertise and experience in this sector enables us to provide strategic advice and value which can unlock investment much more quickly and de-risk decision making.

• Site selection & feasibility studies
  • Evaluating geographical risks (such as; seismic, flooding, extreme weather)
  • Assessing grid reliability & energy supply
• Sustainability & ESG compliance
  • Implementing energy-efficient cooling & heat recovery solutions
  • Supporting carbon-neutral & renewable-powered data centre development
  • Conducting lifecycle emissions & environmental impact assessments (EIA)
• Risk, resilience & uptime optimisation
  • Designing for redundancy & disaster recovery planning
  • Optimising power & water usage efficiency (PUE & WUE)
• Planning, permitting & regulatory compliance
  • Successfully navigating planning & permitting regulations

The need for resilient, energy-efficient, and sustainable data centres has never been greater. By taking proactive steps today, data centre operators can future proof their infrastructure while supporting a greener, more connected world.

----------------------------------

References

1. Secretary of State Science, Innovation and Technology (UK Government), Independent Report, AI Opportunities Action Plan, https://www.gov.uk/government/publications/ai-opportunities-action-plan/ai-opportunities-action-plan, January 2025.
2. World Economic Forum, Climate Action – Why Circular Water Solutions are key to sustainable data centres, https://www.weforum.org/stories/2024/11/circular-water-solutions-sustainable-data-centres/ November 2024
3. Shinde, Rao, and Yogendra, Chapter 11 - Water Footprints and thermal power generation, Volume 8: Water Footprints and Sustainable Development, 2024
4. George, George and Martin, The Environmental Impact of AI: A Case Study of Water Consumption by Chat GPT, Partner Universal International Innovation Journal, Volume 01, Issue 02, April 2023
5. Department for Environment, Food & Rural Affairs, The Water Service Regulation Authority and Rebecca Pow, Press Release: £1.6 billion investment brought forward to speed up vital water infrastructure projects, April 2023
6. Belina Smart, US earmarks £189M for water infrastructure in western states affected by ‘severe drought’ New Civil Engineer, June 2024
7. Environment Agency, Press Release: Environment Agency publishes major update to national flood and coastal erosion risk assessment, December 2024