This article was first published in Environmental Finance.

The tension between soaring water consumption by data centres and finite water resources is rapidly growing. Left unabated, this dynamic increases risks for data centre operators, their clients and ultimately investors in both.

Better disclosure on water use by the industry is an important step towards managing these risks. We believe that investors should encourage more transparency as well as the adoption of existing and emerging technologies that can reduce data centres' dependency on water. The innovative companies behind these technological solutions stand to benefit from the long-term imperative to align the artificial intelligence (AI) revolution with the transition to a more sustainable economy.

The water intensity of data

The energy intensity of data centre operations is well understood: their share of global electricity consumption is expected to have tripled from roughly 1% in 2022 to 3% by the end of 2026.¹ This means the industry will roughly consume the equivalent of the electricity generated by the US' fleet of 50 or so nuclear power stations.²

Awareness surrounding the industry's water intensity is lower. Yet a 100MW hyperscale data centre can directly consume around 2.5bn litres of water each year to cool servers that generate a lot of heat. This is equivalent to the needs of about 80,000 people.³

The rapid adoption of data-intensive AI tools, from chatbots integrated into web browsers to industrial applications such as predictive maintenance, is driving higher water use by the data centres that they run on.

For example, it has been estimated that generating a 100-word email using Open AI's GPT-4 model - a task that involves thousands of calculations to determine the most appropriate set of words - uses the equivalent of a 500ml bottle of water.⁴

Based on projected demand growth, it is estimated that AI tools could indirectly lead to global water withdrawals of 4.2bn to 6.6bn cubic metres by 2027, equivalent to roughly half of the UK's water consumption.⁵

The risks of data centres' thirst

Surging water consumption is particularly concerning in regions that are grappling with water scarcity. Many data centre clusters are located in water-stressed areas, such as the southwestern US and parts of China.⁶

This not only exacerbates local water stress but can also expose data centre operators to significant operational risks. Assets in such regions face the possibility of costly shutdowns in periods of drought, given their need for constant cooling.

Operators may also face reputational risks. There has been growing opposition to new data centres on environmental grounds, from Chile to the US.⁷ Heavy corporate consumers of water can face a backlash from local communities as illustrated by opposition to Nestlé's bottled water operations in drought-stricken California.⁸

Despite these risks, and rising awareness surrounding the issue, only a trickle of information flows to investors.

Identifying opportunities to decouple data from water

Operational and reputational risks make effective management of limited water resources vital for data centre operators, especially as the AI boom drives rising demand for data services. As major water consumers, efficiencies can also deliver meaningful financial savings.

There are clearly long-term investment opportunities arising from solutions that can help the industry manage its water intensity over the coming decades.

Within data centres that use traditional air-based cooling, more efficient heating, ventilation and air conditioning (HVAC) systems, like those offered by US companies Carrier and Trane, can indirectly reduce water used in the power generation process.

In data centres that use evaporative cooling, water recycling solutions - which use media filtration, ion exchange and membrane filtration to return pure water for re-use - from the likes of US-listed Xylem can reduce water consumption by as much as 70%.⁹

Innovative closed-loop liquid cooling systems, like those developed by French-listed Schneider Electric and US-listed Vertiv, can go further: by continually circulating water between servers and chillers to dissipate heat, they can reduce or even remove the need for a fresh water supply.¹⁰ Indeed, the higher thermal density properties of the latest-generation power cabinets in data centres used to power AI mean they require liquid cooling.¹¹

Alongside investment in water-saving technologies, conservation measures taken by leading operators show how the industry can act to better manage its water-related risks. The likes of Microsoft have developed predictive models to anticipate water requirements, based on real-time weather and operational data, and more efficiently manage AI-related workloads (and so manage their cooling requirements.¹² Data centres can also harvest rainwater to reduce the pressure they place on local water infrastructure.

Where conservation measures are not being taken, we believe investors should engage with companies to monitor data centre water risk management and to encourage companies to manage these risks appropriately.

Transparency can help manage risks

Reporting on water consumption remains patchy and inconsistent across the industry. Only two-fifths of data centre operators actively track water usage metrics.¹³ This limits understanding of the issue and hinders effective risk management.

Given the materiality of these risks, we believe investors must engage with the industry to encourage more detailed water-related reporting as standard. Cloud service companies should also be encouraged to demand disclosures from the data centres they partner with, so their water-related dependencies and risks can be identified and evaluated.

An emerging metric for standardised disclosures exists in the shape of Water Usage Effectiveness (WUE), which quantifies how efficiently a data centre directly consumes water in its cooling operations. The metric mirrors Power Usage Effectiveness (PUE), an established sustainability-related performance indicator for the industry.

Encouragingly, some major data centre operators including Equinix and Microsoft have started disclosing WUE for certain regions.¹⁴ Several have also set goals to reduce their water intensity and even - as in the case of Microsoft and Google - to replenish more water than they consume by 2030.¹⁵

Given rising water demand in the context of limited resources, more emphasis should be placed on WUE in data centre building certifications. The incorporation of PUE in the widely used LEED (Leadership in Energy and Environmental Design) standards provides a template.¹⁶ Data centres with higher certification standards should command a rent premium, given corporate environmental targets, creating an incentive for better water management within the industry.

Less water, more data

While it remains unclear how AI will reshape the global economy, it looks certain that the future will involve more AI, and so more data. Unless the water intensity of data centres is addressed, both data centre operators and their customers will face growing risks arising from scarcity of a finite resource.

Proven technologies can quench the industry's thirst, reducing these risks and lowering the water footprint of data. We see growing opportunities for innovative companies that can meet the needs of a rapidly expanding sector.

As well as identifying these opportunities, we believe investors can help their portfolio companies navigate their water-related risks by advocating for better disclosure of water-related risks and their management.

¹ International Energy Agency, July 2024: Electricity mid-year update
² Impax analysis, based on figures from Ember, 2024: Global Electricity Review 2024 and the World Nuclear Association, January 2025: Nuclear Power in the World Today
³ Impax analysis, based on figures from World Economic Forum, November 2024: Why circular water solutions are key to sustainable data centres
⁴ Vermu, P. & Tan, S., 18 September 2024: A bottle of water per email: the hidden environmental costs of using AI chatbots. Washington Post
⁵ Li, P. et al, 2023: Making AI Less "Thirsty": Uncovering and Addressing the Secret Water Footprint of AI Models
⁶ University of Oxford, 2021: Case Study: Water-guzzling data centres / China Water Risk, 2024: China ICT running dry?
⁷ Woollacott, E., November 2024: The environmental campaigners fighting against data centres. BBC
⁸ California State Water Resources Control Board, 2024: BlueTriton Brands (formerly Nestlé) Spring Water Extractions in San Bernardino National Forest
⁹ Guardian Water & Air, March 2025: Keeping data centres cool and efficient
¹⁰ Schneider Electric, 2024: Navigating Liquid Cooling Architectures for Data Centers with AI Workloads
¹¹ Vertiv, 2025: Liquid Cooling Options for Data Centers
¹² Microsoft, 2024: Sustainable by design: Transforming datacenter water efficiency
¹³ Uptime, 2024: Global Data Center Survey
¹⁴ Equinix, 2024: What Is Water Usage Effectiveness (WUE) in Data Centers? / Microsoft, March 2025: Microsoft's Sustainability Targets
¹⁵ Microsoft, March 2025: Microsoft's Sustainability Targets / Google, 2021: Our commitment to water stewardship
¹⁶ LEED, 2024: Data Centers - Minimum energy performance

References to specific securities are for illustrative purposes only and should not be considered as a recommendation to buy or sell. Nothing presented herein is intended to constitute investment advice and no investment decision should be made solely based on this information. Nothing presented should be construed as a recommendation to purchase or sell a particular type of security or follow any investment technique or strategy. Information presented herein reflects Impax Asset Management's views at a particular time. Such views are subject to change at any point and Impax Asset Management shall not be obligated to provide any notice. Any forward-looking statements or forecasts are based on assumptions and actual results are expected to vary. While Impax Asset Management has used reasonable efforts to obtain information from reliable sources, we make no representations or warranties as to the accuracy, reliability or completeness of third-party information presented herein. No guarantee of investment performance is being provided and no inference to the contrary should be made.