Does indoor cannabis cultivation use more water than a 1.5–5 MW AI node?
There is no honest, unconditional yes-or-no answer. The answer changes with the boundary, cooling topology, grid water intensity, cultivated area, and whether the two onsite numbers actually measure the same thing.
Three answers, not one
- Onsite / direct
- Irrigation makeup proxy versus modeled cooling-process water. Comparable meters are required before deciding which uses more.
- Electricity-linked / indirect
- Water used to generate electricity scales with the facility’s annual power use and may be consumed in a different watershed or balancing authority.
- Combined screening sum
- A transparent arithmetic screen—not a basin-aware water balance and not a substitute for engineering or provider review.
Screening disclaimer
This page is educational scenario analysis—not engineering, investment advice, water-rights advice, a utility commitment, a permit opinion, or property proof. It uses no private property data, sends no submissions, and stores nothing. A small adaptively reused 1.5–5 MW IT node is not comparable in scale to a 36 MW, 286 MW, or gigawatt campus merely because all are called “data centers.”
Interactive screening model
Set the boundary before reading the answer.
Choose a stress case or edit every input. The low, base, and high presets are deliberately different stress cases—not confidence intervals, forecasts, or a claim about a specific facility.
Select a representative NOAA station. Only Phoenix invokes the published hot-arid annual cooling anchor; all other regions retain the Standard annual anchor.
High Plains / semi-arid · Denver International Airport, Colorado. Uses the Standard annual PUE/WUE anchor; the NOAA series changes the monthly shape, not the annual assumption.
Scenario preset
Replaces every input and synchronizes the climate profile, cooling anchor, and comparison confirmation.
Check only after aligning withdrawal versus consumption, potable versus all water, meter boundaries, process inclusions, and time period. Until then, the direct verdict remains “Indeterminate.”
Direct water scenario verdict
Indeterminate
The model shows a mathematical direction, but this page withholds the categorical verdict until you confirm comparable onsite bases.
Cultivation footprint
60,000 ft²
Irrigation denominator
Canopy
60,000 ft²
Electricity denominator; tiers allowed
Onsite makeup proxy
2.91 Mgal · 8.94 acre-ft
Gross irrigation before recovery: 4.16 Mgal · 12.77 acre-ft
Annual electricity
7.68 GWh
Electricity-linked water
3.00 Mgal · 9.19 acre-ft
Screening sum
5.91 Mgal · 18.13 acre-ft
Onsite cooling water
0.000 Mgal · 0.00 acre-ft
Zero modeled evaporative/process water for cooling; other building water is excluded.
Annual electricity
30.91 GWh
25.75 GWh IT electricity · PUE 1.20
Electricity-linked water
12.05 Mgal · 36.99 acre-ft
Screening sum
12.05 Mgal · 36.99 acre-ft
Direct WUE crossover
0.428 L/IT kWh
Data-center site WUE at which onsite modeled water equals cultivation makeup.
Grid-factor crossover
0.125 gal/kWh
Electricity water factor at which the two screening sums are equal.
Combined screening direction
CalculatedAI reuse higher
This arithmetic sum is useful for sensitivity testing, but it can add water impacts that occur in different basins and under different accounting rules.
Seasonal sensitivity
Monthly heat and humidity change the cooling tradeoff.
Compare six representative U.S. climate references. NOAA monthly temperature normals and paired hourly dry-bulb/dew-point normals are used to build a transparent cooling sensitivity around the selected annual PUE and WUE anchors.
Curve controls
Dry cooling stays 100% dry-bulb-driven and keeps modeled process WUE at zero.
Denver International Airport, Colorado · Dry annual · Dry-bulb response
Equipment performance curveScreening sensitivity active. Open this panel to enter a certified vendor selection.Optional
Selection metadata
Use one model, capacity, fluid condition, altitude, and controls basis throughout the curve.
Use only when each point already represents total facility power divided by IT power.
Current IT load is about 853 thermal tons before redundancy or design margin.
Performance points
Enter 3–8 points copied from one certified selection. The model linearly interpolates between points and clamps outside the submitted range instead of extrapolating silently.
| # | Wet bulb °F | PUE | Site WUE L/IT kWh | Remove |
|---|---|---|---|---|
| 1 | ||||
| 2 | ||||
| 3 |
Curve cannot be applied yet.
- Enter the equipment manufacturer.
- Enter the model or certified selection identifier.
- Point 1: entering condition must be a finite number.
- Point 1: power value must be a finite number.
- Point 1: site WUE must be between 0 and 10 L/IT kWh.
- Point 2: entering condition must be a finite number.
- Point 2: power value must be a finite number.
- Point 2: site WUE must be between 0 and 10 L/IT kWh.
- Point 3: entering condition must be a finite number.
- Point 3: power value must be a finite number.
- Point 3: site WUE must be between 0 and 10 L/IT kWh.
- Add the vendor document or certified selection title.
- Add the document revision or selection date.
- Add the selected system capacity in refrigeration tons.
- Add the leaving-fluid temperature used for the selection.
- Document the fan, pump, economizer, and adiabatic control modes.
Evidence warnings
- Add a source URL or retain the dated submittal in the evidence room.
Session-only: this public tool does not upload or retain equipment data. Never combine points from different models, capacities, fluids, altitudes, control modes, or document revisions. A validated form is not an engineering certification.
Monthly outdoor temperature
Normal daily high and monthly mean · °F
Monthly moisture state
Derived wet bulb, dew point, and relative humidity
Monthly cooling efficiency
PUE and site WUE move on separate axes
Monthly operating impact
PUE-based facility overhead and modeled onsite cooling water
Compare all six climate references
Temperature means use NOAA monthly normals. Peak monthly wet bulb and July RH are derived from the paired NOAA hourly-normal layer. The annual cooling anchor identifies the model family; it is not measured performance for that city.
| Climate reference | Station | Jan mean °F | Jul mean °F | Annual mean °F | Peak wet bulb °F | Jul RH | Annual anchor | Action |
|---|---|---|---|---|---|---|---|---|
| High Plains / semi-arid | Denver International Airport, Colorado | 31.7 | 75.1 | 51.3 | 58.0 | 42.8% | Standard | |
| Hot-dry desert | Phoenix Sky Harbor International Airport, Arizona | 56.8 | 95.5 | 75.7 | 69.9 | 26.2% | Hot-arid | |
| Marine West Coast | Seattle–Tacoma International Airport, Washington | 42.8 | 67.1 | 53.7 | 58.3 | 64.3% | Standard | |
| Cold continental | Minneapolis–Saint Paul International Airport, Minnesota | 16.2 | 74.3 | 47.0 | 65.3 | 64.4% | Standard | |
| Mixed-humid Southeast | Hartsfield–Jackson Atlanta International Airport, Georgia | 44.8 | 80.9 | 63.7 | 71.7 | 70.0% | Standard | |
| Hot-humid subtropical | Miami International Airport, Florida | 68.6 | 84.1 | 77.5 | 76.8 | 73.0% | Standard |
View accessible monthly data table
| Month | High °F | Mean °F | Paired dry °F | Dew point °F | Wet bulb °F | RH | Driver °F | Curve range | PUE | WUE L/IT kWh | Overhead MWh | Onsite water gal |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| January | 44.6 | 31.7 | 31.4 | 15.0 | 24.9 | 47.5% | 31.7 | Sensitivity | 1.147 | 0.000 | 320 | 0 |
| February | 45.7 | 32.7 | 32.6 | 15.7 | 25.8 | 47.1% | 32.7 | Sensitivity | 1.149 | 0.000 | 295 | 0 |
| March | 55.7 | 41.6 | 41.2 | 20.3 | 32.0 | 42.4% | 41.6 | Sensitivity | 1.174 | 0.000 | 380 | 0 |
| April | 61.7 | 47.8 | 47.7 | 27.2 | 37.9 | 45.8% | 47.8 | Sensitivity | 1.191 | 0.000 | 403 | 0 |
| May | 71.2 | 57.4 | 56.7 | 37.3 | 46.2 | 50.6% | 57.4 | Sensitivity | 1.217 | 0.000 | 474 | 0 |
| June | 83.4 | 68.2 | 67.8 | 43.8 | 53.4 | 44.5% | 68.2 | Sensitivity | 1.246 | 0.000 | 521 | 0 |
| July | 89.9 | 75.1 | 74.4 | 48.6 | 58.0 | 42.8% | 75.1 | Sensitivity | 1.265 | 0.000 | 580 | 0 |
| August | 87.5 | 72.9 | 72.0 | 47.6 | 56.8 | 44.6% | 72.9 | Sensitivity | 1.259 | 0.000 | 567 | 0 |
| September | 79.6 | 64.8 | 63.8 | 38.9 | 49.8 | 42.5% | 64.8 | Sensitivity | 1.237 | 0.000 | 502 | 0 |
| October | 65.3 | 51.1 | 50.7 | 28.5 | 39.9 | 43.4% | 51.1 | Sensitivity | 1.200 | 0.000 | 436 | 0 |
| November | 52.9 | 39.4 | 39.6 | 20.2 | 31.1 | 44.1% | 39.4 | Sensitivity | 1.168 | 0.000 | 355 | 0 |
| December | 44.0 | 31.2 | 30.9 | 14.6 | 24.4 | 47.5% | 31.2 | Sensitivity | 1.145 | 0.000 | 317 | 0 |
Visual comparison
Direct water and grid water are shown separately, not collapsed into one number.
Each water bar is segmented into onsite and electricity-linked components. Read the segment colors before reading the total length.
Annual water screening components
Million U.S. gallons
Annual electricity
Facility GWh side by side
Cooling reality
Saving onsite water can cost electricity—especially at 110°F.
“Liquid cooling,” “closed loop,” and “dry cooling” describe parts of a thermal path, not a complete water claim. The final heat-rejection system, controls, weather, load, and water quality determine actual PUE and WUE.
Rejecting heat without routine evaporation can drive modeled cooling-process WUE toward zero. It may require more fan and compressor energy, larger equipment, or higher coolant temperatures in severe heat.
Using evaporation when weather and controls permit can reduce electrical cooling work but creates onsite consumption, blowdown, treatment, and water-quality obligations. “Hybrid” is a control strategy and equipment topology, not a fixed annual WUE.
Wet-bulb temperature—not dry-bulb temperature alone—constrains evaporative heat rejection. DOE guidance also focuses on cycles of concentration, filtration, treatment, controls, and heat-load reduction where cooling towers are used.
Hot, dry ambient conditions limit how well dry coolers can reject heat and can raise facility power. Evaporative assistance can improve energy performance while using local water; monsoon humidity narrows its wet-bulb advantage. An annual PUE/WUE pair cannot substitute for an 8,760-hour curve.
The Karimi study compared 54 MW and 34.5 MW Phoenix-area colocation facilities. Those reported anchors inform stress testing; they do not make those facilities analogs for a 1.5–5 MW adaptive reuse.
Fact-checked audit
Reading EcoPride precisely
EcoPride raises a legitimate policy question about measurement and accountability. Its public calculator, however, combines an extrapolation with statewide and household claims that need tighter qualification.
What 65.7–80.3 gal/ft²-year means
InferredIt extends the 0.18 and 0.22 gal/ft²-day rates reported for California greenhouse water application in August and September 2018 across 365 days. The Wilson et al. source is an online survey of growers, and the denominator is cultivation area. It is not a measured annual withdrawal or consumption value for indoor cultivation in Arizona.
The household equivalence does not reconcile
CalculatedEcoPride equates 650,000–800,000 gallons with 16–20 Arizona household-years while citing a Tucson benchmark of 10 CCF per month. At 748 gallons per CCF, that benchmark is 89,760 gallons per household-year. The stated volume therefore equals about 7.24–8.91 household-years, not 16–20.
Arizona statewide totals remain unsupported
InferredThe reviewed page does not supply a reproducible statewide inventory of active cultivation footprint, canopy, operating days, water sources, meters, irrigation rates, recovery, or facility-level observations sufficient to support its Arizona total. Treat those totals as advocacy estimates unless the underlying dataset and method are published.
This audit addresses claim support and arithmetic; it does not question motives and does not conclude that cannabis water use is immaterial.
Evidence plan
What turns this from a model into proof
A defensible comparison needs synchronized, property-specific evidence from both the former cultivation use and the final compute design.
- 0112–24 months of utility and submeter data, with interval detail where available
- 02A water balance covering irrigation, reverse-osmosis reject, condensate recovery, storage, leachate, and discharge
- 03Measured canopy, cultivation footprint, and gross building area—kept as separate denominators
- 04One vendor selection or rating export for the same configuration, with at least three supported ambient points, followed by an address-specific 8,760-hour plant model that covers the design extremes
- 05Current electric interconnection evidence and total facility MW—not nameplate anecdotes
- 06A location- and load-shape-appropriate LBNL Water IMPACT electricity factor
- 07Water-provider capacity and service confirmation plus discharge approval
- 08Separate post-conversion meters for cooling/process water and total facility electricity
Method boundaries
How to read the labels
The model is intentionally explicit about whether a statement comes from a source, an input choice, a reasoned interpretation, or arithmetic.
A source says it. Reporting does not itself establish transferability to Arizona or to a particular property.
A user input or preset represents a scenario assumption, including PUE, WUE, and recovery credit.
A conclusion follows from source context or reconciliation but is not directly measured in the cited record.
The shared model performs the arithmetic on current inputs. Precision is not proof of input accuracy.
Area denominators
GFA, cultivation footprint, and canopy are not interchangeable. Tiered canopy can exceed GFA while physical footprint cannot.
Water boundaries
This cultivation onsite value is a makeup proxy after recovery credit; the AI onsite value is modeled process/cooling consumption from site WUE.
The sum is a screen
Adding local and electricity-linked water highlights tradeoffs. It does not resolve timing, basin scarcity, source type, legal availability, or ecological impact.
Source register
Read the underlying record.
EcoPride Initiative
EcoPride / D2 Dispensaries
Campaign claims and calculator language audited on this page.
First known survey of cannabis production practices in California
Wilson et al. (2019), California Agriculture
Online grower survey; source of the August and September greenhouse application rates.
Electricity Consumption from Northwest Cannabis Production
Northwest Power and Conservation Council
Survey-based 128 kWh per canopy ft²-year indoor electricity anchor.
Water Circularity Best Practices Guide
Resource Innovation Institute
Controlled-environment agriculture water-balance and recirculation practices.
2024 United States Data Center Energy Usage Report
Lawrence Berkeley National Laboratory
National data-center energy and water context; not a design for this reuse scale.
Review of Operational Water Consumption and Withdrawal Factors
National Renewable Energy Laboratory
Operational generation-water factors and the consumption/withdrawal distinction.
Water IMPACT Tool
Lawrence Berkeley National Laboratory
Balancing-authority and time-sensitive electricity water-intensity factors.
Water-energy tradeoffs in data centers: a case study in hot-arid climates
Karimi et al. (2022)
Two Phoenix-area colocation facilities; hot-arid PUE/WUE tradeoff anchors.
U.S. Hourly Climate Normals (1991–2020)
NOAA National Centers for Environmental Information
Paired hourly dry-bulb and dew-point normals used to derive representative monthly RH and wet bulb.
PsychroLib psychrometric functions
ASHRAE-based open-source calculation library
Documents the ASHRAE 2017 equations used to derive RH and wet bulb from paired temperature, dew point, and pressure.
Energy and water dynamics in data center cooling: insights from a modeling study in hot-arid climates
Karimi et al. (2025), Applied Thermal Engineering
Peer-reviewed 8,760-hour Phoenix analysis showing why cooling architecture, weather, and economizer logic must be modeled together.
Cooling Water Efficiency Opportunities for Federal Data Centers
U.S. Department of Energy FEMP
Cooling-tower water efficiency, WUE, and operational practices.
Liebert AFC Data Sheet, SL-71060
Vertiv
Product-family reference and nominal rating basis; a project-specific rating export is still required for a multi-point curve.
Sintesis RTAF Product Catalog, RLC-PRC049K-EN
Trane
Product and operating-envelope reference; it does not publish a same-configuration multi-ambient curve.
YVAA Style B Engineering Guide, 201.47-EG1
Johnson Controls / YORK
Directs users to computerized job-specific ratings; public family claims are not used as curve points.
Oasis IEC Product Sheet, OASIS IEC-PS-EN-202409
Munters
Product-family reference; annualized pPUE and capacity ranges for other models are not treated as point-level curves.
TrilliumSeries Adiabatic Products Engineering Data, D037
Baltimore Aircoil Company
Nominal model ratings support selection requests; they are not a portable same-configuration ambient curve.
Industrial Program
Arizona Department of Water Resources
Arizona AMA industrial conservation categories and requirements.
Regulations for Large-Scale Data Centers
City of Tucson
Adopted August 5, 2026; effective September 4, 2026.
Use this screen with professional restraint.
Not engineering. Not investment advice. Not water-rights advice. Not a water availability determination, utility commitment, environmental review, permit opinion, or proof of any property's condition or performance. Inputs are public benchmark scenarios and user selections; there are no fake submissions, no persistence, and no private property data on this page.
Revisit assumptions