Keyboard shortcuts

Press ← or → to navigate between chapters

Press S or / to search in the book

Press ? to show this help

Press Esc to hide this help

Water consumption estimation model

Model principle

First base principle to assess the water consumption of a Datacenter is to make the difference between water withdrawal and water rejection after use. The water consumed is the difference between water coming into the system and water going out, which is defined in EN 50600-4-9 on Water Usage Effectiveness indicator.

DCFootprint assesses water footprint over 3 main scopes: direct usage for cooling and maintenance purpose, indirect usage due to consumed electricity production, indirect usage due to materials manufacturing.

Direct usage, for cooling and maintenance purpose : “Scope 1”

This calculation is dimensionned mainly by the Water Usage Effectiveness parameter proposed by DCFootprint.

Scope1WaterConsumption(L) =  ITElectricityConsumption(kWh) * WaterUsageEffectiveness_site(kWh/L)

With:

  • ITElectricityConsumption(kWh): calculated using the energy consumption model or calculated from the Total (yearly) Electricity Consumption as the total electricity consumption divided by the Power Usage Effectiveness
  • WaterUsageEffectiveness_site(kWh/L): given as the Water Usage Effectiveness parameter

Indirect usage, due to electricity production : “Scope 2”

The indirect consumption of water, due to electricity production is computed with the formula:

Scope2WaterConsumption(L) = TotalElectricityConsumption(kWh) * LocalGridWaterIntensity(L/kWh)

With:

Indirect usage, due to materials manufacturing process: “Scope 3”

This part includes the water consumption that happened during the manufacturing process of the materials and hardware in the Datacenter.

In the current version of the methodology DCFootprint only accounts for the IT hardware manufacturing process water consumption, not for the building materials, nor for the non-IT technical environement gear and materials.

Water stress normalization

⚠️ In future version of the methodology, those volumes consumed will be weighted depending on the local zone water stress. This is not the case in the current version.

Litterature review

To explore

AI Data Centers and the Water Use Feedback Loop, Akinade et al. 2026

Data centre water consumption, Mytton, 2021

0.00–4.4 L/kWh (consumption) ; 0.31–533.7 L/kWh (withdrawal)

The carbon and water footprints of data centers and what this could mean for artificial intelligence, de Vries-Gao, 2025

Calculates:

  • an average WUE for Google datacenters of 1.04 L / kWh.
  • 3.40 L/kWh average for big tech’s datacenters in the USA
  • 4.80 L/kWh average for Apple and Meta datacenters in the USA

Includes tables of available reported data per USA and China big tech company.

Thus, based on available sustainability reporting, the IEA could be significantly underestimating the water footprints of data centers. the total water footprint of AI systems alone could reach between 312.5 and 764.6 billion L in 2025 the uncertainty surrounding these figures would also remain significant because the power grids supporting the specified United States data center locations of Apple, Google, and Meta can be associated with water intensities ranging from 0.68 to 11.98 L/kWh

2024 United States Data Center Energy Usage Report, Shehabi et al. 2024

Averages a direct consumption of 0.45–0.48 L/kWh (projection >2023).

Averages 4.52 L/kWh of indirect consumption (electricity production) for USA datacenters in 2023.

Myths vs. Reality: Data Centers And Water Usage, fwpcoa, 2026

data centers can evaporate 1–9 liters of water per kWh of energy used

Making AI Less “Thirsty”: Uncovering and Addressing the Secret Water Footprint of AI Models, Li et al. 2025

Averages 3.14 L/kWh for datacenters in the USA.

With good water quality, roughly 80% of water withdrawal is evaporated and considered “consump- tion”

citing Google reports.

Datacenter Water Usage 2026: How Much Water Hyperscalers Actually Use (And Where the Industry Is Going), Kova Stack, 2026

Typical ranges:

  • 1.8 L/kWh — traditional evaporative cooling, hot/humid climate. Most colocation and older hyperscale.
  • 0.7-1.2 L/kWh — well-tuned evaporative cooling, moderate climate.
  • 0.1-0.4 L/kWh — adiabatic / air-assisted cooling, modern design.
  • <0.05 L/kWh — fully air-cooled or closed-loop liquid cooling.

The Water Footprint of Data Center Workloads: A Review of Key Determinants, Lei et al. 2024

The environmental footprint of data centers in the United States, Sddik et al, 2021

China ICT running dry? The rise of AI & climate risks amplify existing water risks faced by thirsty data centres, CWR

Climate and technology-specific PUE and WUE estimations for U.S. data centers using a hybrid statistical and thermodynamics-based approach, Lei et al. 2022

For water consumption data by electricity generation technologies