Data centre water consumption could triple to 644 billion litres by 2030

Global data centre water consumption could nearly triple from 222 billion litres in 2025 to 644 billion litres a year by 2030 without significant efficiency gains, as rapid capacity expansion driven largely by artificial intelligence (AI) increases demand for cooling, according to Rystad Energy.

The energy research firm estimates that active water saving measures could reduce annual direct cooling water consumption to 543 billion litres under a moderate scenario and 388 billion litres under an aggressive water saving pathway by 2030.

“Forecasting the water impact of data centres is far from straightforward, with consumption varying significantly depending on cooling technology, climate and whether we measure water withdrawn or water actually consumed,” said Minh Khoi Le, global head of Data Center & Hydrogen Research, Rystad Energy. 

Based on Rystad Energy estimates, data centres consumed 222 billion litres of water directly for cooling in 2025, but that figure could nearly triple to just under 644 billion litres by 2030 in our risky central case. “More water-efficient pathways could bring this down to 543 billion litres in the moderate case or as low as 388 billion litres in an aggressive water-saving scenario. These figures only capture direct cooling-water consumption, while the less visible water footprint embedded in the electricity supply adds another layer of complexity,” Le said.

Cooling choices shape water demand

Data centres use water primarily to remove heat generated by computing equipment, but consumption varies widely depending on cooling technology and location.

AI servers generate substantially more heat. However, their growing deployment does not necessarily translate into a proportional rise in water consumption, as rack level liquid cooling can reduce the heat burden on facility level systems and enable greater use of dry cooling, according to Rystad Energy.

Technologies proposed for NVIDIA’s Vera Rubin platform could help limit water use, although the extent of savings will depend heavily on climate. Dry cooling is generally more effective in colder locations.

Water use effectiveness (WUE), which measures water used per unit of information technology energy, offers another way to assess the sector’s footprint. The metric can also account for water consumed in generating the electricity used by a data centre.

Some technologies reduce on site water consumption but increase electricity demand. Dry cooling, for instance, can save around 2.15 litres of water for every kilowatt hour (kWh) of information technology load but requires an additional 0.30 to 0.74 kWh of electricity.

The indirect water footprint can therefore be substantial and depends on the water intensity of the electricity supply. In the United States, the indirect portion of water consumption by data centres can be more than twice their direct water consumption, according to Rystad Energy.

Operators report wide differences

With no general requirement for WUE standards in most regions, water performance varies widely among operators, including those operating in similar climate zones.

Amazon Web Services (AWS) reported an average direct site WUE of 0.12 litres per kWh in 2025. Meta reported 0.19 litres per kWh for 2024, while Microsoft reported 0.27 litres per kWh for fiscal 2025. Digital Realty reported 0.59 litres per kWh and Equinix 0.91 litres per kWh in 2025.

Google’s disclosed data indicate a materially higher intensity, although the company does not publish an official WUE figure.

The figures are not directly comparable because of differences in the treatment of leased and colocation facilities and regional portfolios. AWS and Meta also define site WUE as water withdrawal per unit of information technology capacity rather than water consumption.

AWS’s regional figures for 2025 ranged from 0.02 litres per kWh in Stockholm to 2.85 litres per kWh in Jakarta, a more than 100-fold difference within the company’s own reporting framework, highlighting the influence of geography and cooling architecture.

AWS, Google, Microsoft and Meta have all set targets to become water positive by 2030, aiming to replenish more water than they consume. Their efforts include freshwater restoration and water efficiency projects, including in agriculture, with greater focus in some water stressed areas.

Regulation focuses on disclosure

Regulation on data centre water use is beginning to emerge, but reporting requirements continue to outpace binding performance standards.

The European Commission is planning a Data Center Energy Efficiency package that would introduce a labelling system and performance standards. Singapore’s Green Data Center Roadmap aims for water use of less than 2 cubic metres per megawatt hour (MWh), equivalent to 2 litres per kWh.

In the United States, regulations vary by state. Texas has recently imposed a moratorium on new approvals pending audits of tax breaks, power, water and cooling use associated with data centres.

Water stress could become an increasingly important factor in regulating the sector. Regions facing high and extremely high-water stress are projected to account for 34 per cent of global direct water consumption by data centres in 2030, according to Rystad Energy.

Requiring the least water intensive cooling technologies in these regions could reduce consumption by 45 per cent. Jamnagar and Thane in India, along with Reeves County in Texas, are among the areas identified as highly exposed, where projected data centre water consumption is high relative to local water stress.