Skip to content

ESRS E3-1 — water consumption disclosure

What ESRS E3-1 requires

ESRS E3-1 covers water-related impacts and dependencies. For data-centre-related water consumption, the relevant disclosures are:

  1. Total water consumption in m³ for the reporting period
  2. Water consumption in areas of water stress (AWARE-100 weighted)
  3. Cooling-method assumption (evaporative, dry, hybrid)
  4. Methodology and boundary

What we provide

Every receipt carries a water consumption estimate in millilitres, AWARE-100 weighted. Quarterly aggregates roll up to litres or m³ as appropriate.

Region Cooling Water intensity (mL/kWh) AWARE-100 stress factor
scaleway-par-1 Evaporative + free-cooling 1,800 0.7 (Paris)
atnorth-sto-1 Evaporative 1,400 0.4 (Stockholm)
atnorth-isl-1 Direct seawater + free-cooling 80 0.1 (Iceland)
ovh-gra Evaporative 1,500 0.6 (Gravelines)
ovh-rbx Evaporative 1,500 0.6 (Roubaix)
hetzner-fsn-1 Hybrid 1,200 0.5 (Falkenstein)

Water intensities are derived from provider disclosures (where published) and Carbone-4-style modelled defaults (where not). Methodology lineage is in the evidence pack.

Worked example

For the same Q1 2026 example as in E1-6:

Region Queries Energy (kWh) Water (L) AWARE-weighted (L)
scaleway-par-1 3,100,000 715 1,287 901
atnorth-sto-1 600,000 138 193 77
atnorth-isl-1 200,000 46 4 0.4
ovh-gra-1 300,000 69 104 62
Total Q1 4,200,000 968 1,588 L 1,040 L AWARE

Disclosure language:

AI inference water consumption, Q1 2026: 1,588 L direct withdrawal, 1,040 L AWARE-100 weighted. Cooling-method weighted: 89% evaporative, 11% direct/free-cooling. Methodology: Vetted Inference v0.4.2 with regional provider-disclosed water intensities; AWARE-100 weighting per WULCA factor table 2024.

Notes for assurance review

  • Water consumption is withdrawal, not consumption-net-of-return. Our reported number is consumption (water that does not return to the source as liquid; primarily evaporative loss). For some buyers, withdrawal-based disclosure is the requested form; we provide both numbers in the evidence pack.
  • Provider-disclosed water intensities vary in granularity. Where a provider does not publish per-DC intensity, we use a regional default and note this in the pedigree score.
  • AWARE-100 stress factors are updated periodically by WULCA; we pin to the version used at receipt time.

Where this is implemented

methodology/aggregation/csrd_e3_1.py