The bigger picture behind your estimate
Every chart updates from the calculator. Change your activity or infrastructure there, and this page follows.
Live from your calculator: Text generation · 1,000 interactions a month Change inputs
Monthly estimate
2.7 L
range 22 mL to 35 L
Direct share
25%
cooling water on site
Confidence
Low
based on the data you provided
Site water stress
Unknown
location not set
Next 12 months
About 44 L over the year at 5% monthly growth.
Central Low to high (log scale)
Where the water goes
2.7 L
per month
Cooling 25%
685 mL Electricity 75%
2.1 LThe power grid matters more than cooling here, so the electricity source is your bigger lever.
Infrastructure scenarios
The same usage under three infrastructure profiles.
Your assumptions
2.7 L
range 22 mL to 35 L
As set in the calculator
Greener profile
429 mL−84%
range 22 mL to 4.2 L
Low water coolingRenewablesHyperscale
Worst case
4.9 L+78%
range 580 mL to 38 L
Evaporative, hotThermal gridLegacy
Moving to the greener profile would cut your central estimate by about 84%.
Comparing AI workloads
Water per 1,000 items with your infrastructure. Marker is central, band is the range.
Text generation · 1,000 interactions2.7 L
Document processing · 1,000 documents14 L
Image generation · 1,000 images16 L
Regional water stress
The same liter matters more where water is scarce. Tap a region to set it.
Location unknown. Water stress cannot be assessed.
Only regional findings published by WRI (Aqueduct 4.0) are labeled. Stress varies strongly between river basins.
Sustainability indicators
Water Usage Effectiveness (WUE)0 to 2.5 L/kWh
Power Usage Effectiveness (PUE)1.1 to 1.8
Grid water intensity0.2 to 5 L/kWh
Regional water stressUnknown
Data availabilityLow confidence · high is 1568× the low
Potential improvementup to 84% with greener infrastructure