Skip to content

Water methodology

AI inference consumes water in datacenter cooling and in electricity generation. This page specifies the modeled per-query water consumption reported for the public catalog’s How-Hungry rows. It complements the energy methodology. A water estimate requires compatible energy scope, PUE, and both WUE inputs; known energy alone is insufficient.

Consumption, not withdrawal

The figures here are water consumption (water that is evaporated or otherwise removed from the local water cycle), not water withdrawal (water drawn and largely returned). Consumption is the figure that matters for environmental impact and is the one Venturi reports.

The two-term formula

The water methodology distinguishes two components:

Text Only
water_site     = (energy / PUE) × WUE_site      (on-site cooling)
water_source   =  energy        × WUE_source    (off-site electricity generation)
water_combined = water_site + water_source

This is Equation 4 of How Hungry is AI? (Section 4.4). energy is the query’s total energy in kWh with PUE applied once, and both WUE values are in liters per kWh, so each term is in liters.

  • water_site is the water evaporated cooling the datacenter that runs the inference (an on-site, Scope-1-style term).
  • water_source is the water consumed off-site generating the electricity the inference drew (a Scope-2-style term).
  • water_combined is their sum, and is the headline per-query figure.

In public schema 1.1.0, the 25 modeled water values are combined consumption in milliliters per query, stored in water_ml_per_unit with measurement_unit = query. The legacy water_site_ml, water_source_ml, and water_ml_per_query fields are null on every row. The component formulas above explain the combined estimate; the public file does not store separate site and source results. The other 26 rows, including the five measured-energy rows, have no water value.

WUE is Water Usage Effectiveness. WUE_site is the liters of water consumed on site per kWh of IT energy, which is why the site term first removes PUE. WUE_source is the liters consumed in generating each kWh of electricity the facility draws, which is why the source term uses the full energy.

The PUE asymmetry: read it carefully

The two terms treat PUE differently, and mixing them up is the easy mistake this page exists to prevent:

  • The site term divides energy by PUE. The intent is to remove datacenter overhead and apply the cooling water-intensity to the IT energy itself.
  • The source term uses the full energy (the energy that was actually drawn from the grid, overhead included), because the grid had to generate all of it.

Applying PUE to the wrong term, or applying it twice, materially distorts the result. Venturi applies it exactly as written above, and a gate enforces the two-term shape so it cannot regress.

Null, never zero

Water follows the same null-not-zero discipline as energy and carbon. If any required term is unknown (energy is null, or a provider’s WUE_site or WUE_source is unknown), the corresponding water figure is reported as null, never 0. A zero would falsely imply no water was consumed.

Note that a single-term water model (cooling only, with the source term set to zero) is the degenerate case of the formula above. Venturi uses the full two-term model and does not silently treat an unknown source term as zero.

Provider water-intensity reference

The populated modeled water rows use versioned study assumptions from Table 1 of How Hungry is AI?. These environment and hardware descriptions define the study’s modeling assumptions. The site term reflects assumed datacenter cooling; the source term reflects assumed electricity generation. Values are in liters per kWh:

The final column identifies the sources cited in the study’s Table 1.

Environment WUE_site WUE_source Table 1 citations
Azure 0.30 4.35 WUE_site: Microsoft Cloud blog (December 2024); WUE_source: World Resources Institute
AWS 0.18 5.11 WUE_site: Amazon 2023 Sustainability Report; WUE_source: World Resources Institute
DeepSeek-native 1.20 6.016 WUE_site: Hequan Wu, China Communications Enterprise Association (April 2024); WUE_source: World Resources Institute

The source-term values come from the World Resources Institute’s Guidance for calculating water use embedded in purchased electricity (2024), applied by the paper’s authors to each provider’s datacenter locations. Licenses lists the full Table 1 citations, including PUE and carbon intensity.

Source-term water intensity varies widely by region, even within one provider, because it follows the local generation mix. Where a sourced region-level value exists, Venturi resolves it ahead of the provider-level value, mirroring the carbon resolution chain in the energy methodology. A provider or region with no sourced WUE value resolves to null, never 0.

These values are reference inputs reimplemented from published research and provider sustainability disclosures; see licenses for attribution.

Where to go next

Give feedback on this page

Sign in to send feedback. Opens in a new tab.

Copy page details 
Docs feedback: Water methodology
Page: https://docs.venturi.systems/energy/water-methodology/