Efficiency And Total Use Answer Different Questions
A data center can use less water for each unit of computing while the total site uses more water because workload grows faster than efficiency improves. Both statements can be true. Communities, utilities, customers, and operators need the intensity measure and the absolute volume before they can understand the operating effect.
Industry updates increasingly highlight improved water usage effectiveness, closed-loop cooling, reclaimed water, and replenishment projects. Those are meaningful developments. A defensible claim must also show the boundary: which facilities, which water source, what period, whether the metric covers withdrawal or consumption, and how growth changed the total.
Metric Boundaries Create Confusion
Water withdrawal is the volume drawn from a source. Water consumption is the portion not returned, often because it evaporates. Discharge is water returned after use. Water usage effectiveness commonly expresses facility water per unit of IT energy. Replenishment estimates water restored or made available through projects, often in other places or periods.
These measures cannot be substituted for one another. A lower global WUE can coexist with stress at one local watershed. A replenishment project may provide real value without replacing water at the exact hour and location of a facility's demand. A zero-evaporation cooling design may still be connected to electricity generation with an indirect water footprint.
Reconcile Intensity With Growth
The basic relationship is absolute facility withdrawal equals IT energy multiplied by WUE, adjusted for the metric's coverage. If IT energy rises 50 percent while WUE falls 20 percent, modeled withdrawal becomes 1.5 multiplied by 0.8, or 1.2 times the original volume. Efficiency improved, yet absolute withdrawal rose 20 percent.
This is an illustrative calculation, not a forecast for a specific operator. Real sites require hourly weather, cooling mode, water source, power, workload, and discharge data. The formula's value is diagnostic: it prevents an intensity improvement from being presented as proof that total local demand declined.
Diagnose The Claim Before The Technology
Ask whether a reported number covers owned facilities, leased capacity, or both; whether it is global, regional, or site-specific; whether it measures withdrawal or consumption; and whether the denominator is IT energy, total facility energy, compute output, or revenue. Record the comparison year and whether historical figures were recalculated.
Warning signs include percentages without volumes, fleet averages without site ranges, replenishment presented as avoided withdrawal, annual totals that hide summer peaks, and “zero water” language that refers only to one cooling process. A claim can be technically accurate and still be inadequate for a local capacity decision.
Compare Cooling And Siting Options
Evaporative cooling can reduce electricity demand but consumes water. Air-cooled chillers and dry cooling reduce on-site evaporation but may use more electricity, particularly in hot conditions. Closed-loop direct-to-chip systems move heat efficiently and can avoid continuous water loss, while still requiring heat rejection and attention to indirect energy effects.
Reclaimed water can reduce competition for drinking water, but availability, treatment, pipes, and discharge limits vary. Siting in a cooler or less-stressed region may reduce water risk while changing latency, grid, land, and transmission constraints. The correct choice is a local resource tradeoff, not a universal cooling winner.
Build The Absolute Use Reconciliation
For each site, record monthly IT energy, total electricity, water withdrawn by source, water consumed, discharge, WUE, peak-day demand, cooling mode, workload capacity, and watershed stress. Separate operating water from construction and indirect estimates. Identify meters, calculation methods, missing periods, and independent assurance.
Then reconcile year-over-year change into workload growth, efficiency improvement, weather, equipment mix, and operating decisions. Show replenishment as a separate line with location, timing, expected benefit, verification method, and durability. The reconciliation should let a reader move from a percentage claim to actual volume and local consequence.
A Growing Campus Example
Consider an illustrative campus that used 100 units of water withdrawal for 100 units of IT energy, giving an index WUE of 1.0. The next year, cooling improvements lower WUE to 0.75, but IT energy rises to 160 units. Modeled withdrawal becomes 120 units, a 25 percent efficiency improvement alongside a 20 percent absolute increase.
The operator can explain both results and then show whether reclaimed water, seasonal controls, or replenishment changed local freshwater impact. A utility can plan around peak demand rather than the annual average. Customers can compare workload choices without assuming that a lower intensity number automatically means a smaller total footprint.
Measure Local And Fleet Performance
Track WUE, withdrawal, consumption, discharge, source mix, peak-day demand, water-stress weighting, percentage metered, reclaimed-water share, and verified replenishment. Report fleet totals and site distributions. Averages should not erase the highest-stress locations or facilities with incomplete measurement.
Set operating thresholds tied to drought stage, utility capacity, temperature, and workload scheduling. Review whether efficiency projects reduced the intended measure and whether rebound from growth consumed the gain. Good stewardship connects design claims with metered operation after the site scales.
Reconcile One Site This Quarter
Choose the facility or cloud region carrying the largest workload growth. Gather twelve months of water withdrawal, consumption or best available estimate, IT energy, weather, cooling mode, and peak demand. Recalculate the public or internal efficiency claim and show the absolute change beside it.
Document missing data rather than filling it with false precision. Ask the utility or operator which number matters for local planning and which seasonal constraint is binding. Use that answer to shape workload placement, cooling upgrades, contract requirements, and future reporting.
Sources, Method, And Limits
This article was prompted by current reporting on data-center water-efficiency claims. Primary context includes AWS reporting a 2025 global data-center WUE of 0.12 liters per kilowatt-hour of IT load and Microsoft's explanation of fleet WUE, cooling designs, and 2025 water intensity. These are company-reported measures with stated boundaries.
AWS separately explains the distinction among withdrawal, discharge, and consumption. The reconciliation framework and indexed example are SynHy analysis. Site evaluation requires local meter data, utility context, current watershed information, the operator's exact methodology, seasonal demand, water-source quality, discharge constraints, and a clear statement of any unmeasured indirect effects.