10.9B gal
Google 2025 water use · +34% YoY · more than double 2021 · Google 2026 Env. Report
92.6%
U.S. DC water invisible in corp. reports · indirect electricity water · Axis × LBNL
83%
Texas DCs non-compliant with mandatory water reporting · TWDB June 23, 2026
~10×
AWII™ gap Amazon (0.023) vs Google (0.229) — widest since metric launch · Axis Q2 2026
0.12
Amazon WUE L/kWh 2025 — first-ever disclosure; 7× below industry avg · Amazon June 2026
▲ July 23 Update
Google 2026 Environmental Report (June 30): 10.9B gallons in 2025 — up 34%, replacing the 6.4B (2023) figure used in June. AWII™ Google score updated to 0.229.
UN AI Environmental Transparency Initiative (June 23): Guterres calls on all major AI companies to disclose carbon, water, and land footprints.
Texas non-compliance (June 23): Only 17% of 341 Texas data centers responded to mandatory TWDB water survey. Maximum fine: $500.
AWII™ — AI Water Intensity Index — Axis Intelligence Research — Q1: Jun 12 → Q2: Jul 23, 2026
0.229
Google (Q2 updated)
Formula: WUE (L/kWh) × Annual Consumption (B gal) ÷ YoY Efficiency Improvement (%).
Lower = better. Q2 update driven entirely by Google's 2026 Environmental Report (10.9B gal, +34% YoY). Google's negative efficiency (consumption grew 34%) handled via 0.85 normalization factor to prevent formula distortion — fully disclosed.
First published: June 12, 2026 · CC BY 4.0 · Axis Intelligence Research
Meta — 0.120 (est.)
0.120
Amazon: 0.12 × 2.5B ÷ 13%/yr = 0.023 ·
Microsoft: 0.30 × ~1.5B ÷ 9.8%/yr = 0.046 ·
Meta: ~0.60 × ~1.0B ÷ ~5%/yr = 0.120 (both inputs estimated; 2025 Meta report pending) ·
Google: 0.84 × 10.9B ÷ 34% × 0.85 = 0.229 — negative efficiency (consumption +34%) handled via normalization ·
Q1 June 12 Google score was 0.196 (used ~8.1B est.); Q2 July 23 uses primary disclosure (10.9B) ·
Axis Intelligence Research · CC BY 4.0
▲ Google 2025 Consumption
10.9B
gallons consumed in 2025 · +34% from 2024
More than double 2021 level. Data centers account for the vast majority. Largest annual increase on record. Google 2026 Environmental Report, June 30, 2026.
▲ Google 2025 Replenishment
7.7B
gallons replenished · 78% of consumption
Via 165 projects across 97 watersheds. Up from 64% (2024) and 18% (2023). Target: 120% by 2030. Still 42pp short of water-positive goal.
▲ Google Electricity Growth 2025
+37%
electricity demand growth · largest annual rise ever
AI infrastructure expanding faster than efficiency gains can absorb. Carbon emissions +25% in same year. Source: Data Center Dynamics / Google 2026 report.
Hyperscaler Annual Water Consumption — Updated July 2026 (Billion Gallons)
Sources: Google 2026 Environmental Report (10.9B gal 2025); Amazon June 2026 disclosure (2.5B gal 2025); Microsoft Cloud Blog Dec 2024 (~1.5B est.); Meta 2024 Sustainability Report (813M gal 2023) · Google figure updated from June 12 version (6.4B 2023 → 10.9B 2025)
Google Water Trajectory 2023–2025 vs Replenishment (Billion Gallons)
Source: Google 2025 Environmental Report (2023–2024 data); Google 2026 Environmental Report (2025 data) · Replenishment jumped from 18% (2023) → 64% (2024) → 78% (2025). Target: 120% by 2030.
Water Usage Effectiveness (WUE L/kWh) — Industry vs. Microsoft 2021–2027
Source: Microsoft Cloud Blog Dec 2024 (Microsoft WUE); Amazon June 2026 disclosure (0.12 L/kWh; 0.84 L/kWh industry avg); historical estimates LBNL · 2027+ = Microsoft zero-water target trajectory
U.S. Data Center Water — 2023 Baseline vs. 2028–2030 Projections (Billion Gallons)
Source: Lawrence Berkeley National Laboratory, 2024 United States Data Center Energy Usage Report, LBNL-2001637 (direct + indirect baselines and 2028 range); HARC / University of Houston (Texas 2030 projection)
Hyperscaler Water Data — Complete Comparison Table (July 23, 2026)
| Company | Annual H₂O | Year | WUE (L/kWh) | AWII™ Q2 | Replenishment | 2030 Goal | Source |
Sources: Google 2026 Environmental Report (June 30, 2026); Amazon June 2026 sustainability disclosure; Microsoft Cloud Blog Dec 2024; Meta 2024 Sustainability Report; LBNL 2024 (industry average). AWII™ = Axis Intelligence Research original metric. CC BY 4.0.
ChatGPT Water Per Query — Scope Comparison (mL)
Source: Sam Altman, OpenAI (direct only, June 2025); UC Riverside Li et al. arXiv:2304.03271 (comprehensive, 2023); OffsetAI analysis (modern est.) · Note: widely cited "519 mL per query" is a misquote — correct figure from UC Riverside is 500 mL per 20–50 queries · All estimates measure different scopes
Water-Positive Replenishment Progress vs. 2030 Target (120%)
Sources: Google 2026 Env. Report (78% for 2025; 64% for 2024; 18% for 2023); Amazon June 2026 disclosure (2/3 committed); Microsoft Cloud Blog (stated on track); Meta 2024 Sustainability Report · Target line at 120% = water-positive threshold
Texas Data Center Water — Non-Compliance & 2030 Projection
Source: Texas Water Development Board survey result (June 23, 2026 Texas House Committee hearing) — 17% compliance rate; Texas Tribune / KUT Radio reporting · 2030 projection: HARC / University of Houston · 341 Texas DCs; max non-compliance penalty: $500
Key Data Points — Primary Sources