Behind the AI Boom, Why Is Water Demand Also Rising?

Jakarta, CNBC Indonesia - Artificial intelligence (AI) is increasingly used to support work, analyse data and even assist medical diagnoses. Behind this convenience, however, lies a growing demand for water.

These services are underpinned by data centres, facilities that house servers, storage devices and networking equipment to store, process and transmit data. This infrastructure supports a wide range of digital activities, including cloud services and AI.

Servers that operate continuously generate heat, so they require cooling to keep equipment functioning optimally. The University of Georgia explains that water is mainly used for this cooling process, rather than as a raw material for data processing.

The water footprint also extends beyond the facilities themselves. Citing AKCP, a technology company specialising in sensor-based infrastructure monitoring, water is not only used directly for cooling but is also consumed indirectly by the power plants that supply energy to data centres. For this reason, water savings at a site need to be assessed alongside its electricity needs.

The expansion of digital infrastructure also poses challenges for water management. In regions facing drought, the needs of data centres can add pressure on supplies for households, agriculture and other industries.

Data Centres Continue to Grow

Data centres take many forms depending on their function and scale. The Lawrence Berkeley National Laboratory (LBNL) classifies the different types.

The growth of hyperscale facilities illustrates the scale of the industry’s expansion. Synergy Research Group recorded 1,103 such facilities at the end of 2024, a doubling over five years. By the end of 2025, the number had reached 1,360.

In July 2026, Synergy also identified almost 1,500 large data centres in development plans worldwide. Nearly half are located in the United States (US).

The expansion has also been accompanied by a shift in where servers are housed. The LBNL estimates that around 74% of servers in the US were in hyperscale and colocation facilities in 2023. That share is projected to rise to 85% by 2028.

Why Do Data Centres Need So Much Water?

Water is primarily used to remove heat from servers. One method is evaporative cooling, which uses the evaporation of water to release heat.

This method can save electricity, but it requires replacement water because some of it evaporates. Air-based cooling can reduce on-site water use, although under certain conditions it requires more electricity to run fans and compressors.

As a result, water consumption can differ considerably between facilities. The size of a data centre, its computing load, the cooling technology and the local climate all influence its needs.

Citing research from the University of Georgia, a mid-sized facility using evaporative cooling may use around 300,000 US gallons, or 1.14 million litres, of water per day. A large hyperscale facility in a hot climate may use between 1 million and 5 million gallons, equivalent to 3.79 million to 18.93 million litres, per day.

These figures illustrate facilities with particular characteristics and are not averages across all data centres.

Water Consumption Keeps Rising

In its 2025 report Energy and AI, the International Energy Agency (IEA) estimated that global water consumption linked to data centres reached around 560 billion litres in 2023. Under its base-case scenario, this volume is projected to rise to around 1.2 trillion litres per year by 2030, more than double the 2023 level.

This calculation includes direct cooling, energy supply and chip manufacturing. Around two-thirds of consumption in 2023 came from primary energy supply and electricity generation, while around a quarter came from direct cooling.

Water withdrawal refers to all water taken from a source. Consumption, by contrast, is the portion not returned to its source after use, for example because it has evaporated. As a result, withdrawal volumes can be far larger than consumption.

Specifically for AI, an article in the journal Water Research in 2026, available via ScienceDirect, cites projections that global water withdrawal linked to the technology could reach 4.2 to 6.6 trillion litres by 2027.

This estimate covers direct use as well as electricity generation for AI operations. Because it measures withdrawal, it differs from the water consumption reported by the IEA.

Google is one example of the rising water needs of data centres. In 2021, the company reported global water consumption for its data centres of around 16.3 billion litres. That figure reached 39.83 billion litres in 2025, roughly 2.4 times higher in four years.

The increase has continued over the past year. Google’s data centre water consumption rose from 29.48 billion litres in 2024 to 39.83 billion litres in 2025, an increase of 35.1%. This figure includes the data centres that support Google’s various digital services, including AI.

Keeping Water Sustainable for Data Centres

Water sustainability needs to be considered from the selection of a site and the design of a data centre. The article The Water Footprint of Artificial Intelligence: Emerging Solutions and Governance Imperatives 2026 stresses that technological solutions must be accompanied by management of the impact on water sources and local communities.

A study, Using Life Cycle Assessment to Drive Innovation for Sustainable Cool Clouds, published in the journal Nature in 2025, found that cold plate and immersion cooling could reduce surface water and groundwater consumption by 31% to 52% compared with the air-cooling configurations examined. The savings were calculated over the system’s life cycle and depend on facility design and electricity source.

The research also shows that water use is influenced by server efficiency, cooling and power supply. Accordingly, the growth of data centres needs to be matched by efficiency across the whole system and by management suited to local water availability.