Green Hydrogen Water Use: Why Estimates Range 9–50+ L/kg
By Green Gas Turbines Team · Published January 4, 2026 · 18 min read
Electrolysis chemically consumes 8.93 litres of pure water for every kilogram of hydrogen. That is the minimum—not the quantity a project must withdraw from a river, aquifer, utility or ocean. Treatment losses, cooling, gas drying, blowdown and the water used to generate electricity can move a credible estimate from about 9 L/kg to 50 L/kg or more, depending on what is counted.
The first question for any water-use claim is therefore: Does the number mean stack make-up, plant withdrawal, net consumption or full life-cycle water use? Without that boundary, apparently precise comparisons are usually misleading.
Green Hydrogen Water Use: The Short Answer
| Number | What it measures | How to use it |
|---|---|---|
| 8.93 L/kg H2 | Stoichiometric water converted into hydrogen and oxygen. | A physical minimum for reaction water—not a source-water allocation. |
| About 9–10 L/kg | Demineralised make-up in current product examples. | Useful for sizing the stack make-up stream only. It excludes raw-water treatment reject and may exclude cooling. |
| About 9–30 L/kg | Direct onsite consumption in the U.S. DOE's 2025 PEM scenarios: 2.4–8.0 U.S. gal/kg. | Shows the effect of dry, hybrid or conventional cooling assumptions. It is not a universal project range. |
| 42–52.2 L/kg | Treated-water demand modeled for alkaline and PEM electrolysis in a UK planning study. | An illustrative planning case that includes specified cooling and treatment assumptions—not an OEM guarantee. |
| 1,000+ L/kg withdrawal | Possible in modeled once-through cooling cases. | Most is returned to the source warmer, so withdrawal can be very high while consumption is much lower. |
The wide spread is not a contradiction. Each number answers a different question.
Withdrawal, Consumption and Life-Cycle Use Are Different
- Withdrawal: all water removed from a source for the facility, whether it is consumed or later returned.
- Consumption: withdrawn water not returned to the same local watershed because it is converted, evaporated or otherwise removed.
- Discharge: water returned to a surface water body, sewer, injection system or other receiving route, with its flow and quality specified.
- Direct onsite use: water used at the hydrogen facility for reaction, treatment, cooling and auxiliaries.
- Indirect use: upstream water associated with electricity, equipment, materials and the delivered water supply.
A once-through cooling plant may have enormous withdrawals but return most of the flow. A cooling-tower plant withdraws less but consumes more through evaporation. A dry-cooled plant can approach the reaction-water minimum onsite, but normally pays with higher capital cost, fan power, plot area or reduced hot-weather performance.
The electricity source also changes the life-cycle answer. DOE's electrolysis assessment illustrates this clearly: thermal generation with wet cooling can add far more upstream water than wind or solar electricity. Report onsite and upstream water separately so a remote power-plant cooling load is not mistaken for the electrolyser's local demand.
What 100 MW of Electrolysis Means for Water Supply
The following example assumes a 100 MW plant, system electricity use of 52 kWh/kg H2 and 90% annual capacity factor. It is a sizing illustration, not a performance guarantee.
| Assumption | At full load | Annual at 90% capacity factor |
|---|---|---|
| Hydrogen output | About 1,923 kg/h | About 15.2 million kg |
| 8.93 L/kg reaction water | About 17.2 m3/h | About 135,000 m3 |
| 20 L/kg direct-use case | About 38.5 m3/h | About 303,000 m3 |
| 50 L/kg planning case | About 96.2 m3/h | About 758,000 m3 |
For a 1 GW plant under the same assumptions, multiply each result by ten. Capacity factor, degradation, turndown, planned outages and product pressure should be included in the project's hourly and annual mass balance.
Water Quality Is an OEM Specification, Not a PEM-vs-Alkaline Rule
PEM systems are often described as requiring “ultrapure” water while alkaline systems are said to be more tolerant. That shorthand is not sufficient for procurement. Current product documents show why:
- Accelera's 2025 HyLYZER 1000-30 PEM sheet specifies 9 L/kg of demineralised water meeting ASTM D1193 Type III Grade B.
- Nel's A485 alkaline sheet specifies 10 L/kg of deionised water at less than 1 µS/cm.
- Nel's current generator FAQ states a minimum of ASTM Type II, greater than 1 MΩ·cm, and recommends Type I, greater than 10 MΩ·cm, while also calling for silica, organic and carbon measurements.
These examples are not directly comparable product classes, but they disprove a universal conductivity limit for all PEM or all alkaline systems. The issued utility specification, operating manual and warranty agreement govern.
Conductivity alone cannot qualify the water
Conductivity or resistivity is a fast indicator of total ionic contamination. It does not reveal which ions are present and can miss non-ionic organics or particles. A complete acceptance specification may include:
- Conductivity or resistivity at a stated reference temperature
- Silica, sodium, chloride and selected metals
- Total organic carbon (TOC) and oil/organic contaminants
- Hardness, alkalinity and dissolved gases where relevant
- Microbiological limits and oxidants such as free chlorine or chloramine
- Sampling location, method, frequency and analyzer accuracy
At 25°C, 1 MΩ·cm corresponds to 1 µS/cm, while 10 MΩ·cm corresponds to 0.1 µS/cm. Always state the units and temperature because “DI water” alone is not a specification.
Where ASTM D1193 and ISO 3696 fit
ASTM D1193-24 defines reagent-water types and grades using several constituent tests; its own scope says users must determine whether a selected type is suitable for the intended application. ISO 3696 defines laboratory-water grades for analytical use. Both can provide a common vocabulary, but neither replaces the electrolyser OEM's complete utility specification.
Purification Train: Start With the Source Analysis
There is no mandatory “RO plus EDI” flow sheet. Treatment should be built from seasonal source-water data, the OEM limit and the permitted concentrate route.
| Source | Typical treatment building blocks | Issues to test |
|---|---|---|
| Potable / low-TDS utility water | Dechlorination or carbon where needed, cartridge filtration, RO and a final polisher if required. | Chlorine/chloramine, hardness, silica, utility outages and whether potable demand competes with community supply. |
| Surface water | Clarification or dissolved-air flotation, media filtration or UF, RO and polishing. | Seasonal turbidity, algae, natural organic matter, pathogens and drought reliability. |
| Groundwater / brackish water | Oxidation/filtration, softening or antiscalant, brackish-water RO and polishing. | Sustainable yield, subsidence, iron, manganese, hardness, silica, salinity and concentrate disposal. |
| Reclaimed municipal or industrial water | Source-specific biological treatment, UF/MBR, activated carbon or oxidation where justified, RO and polishing. | TOC variability, nutrients, biofouling, trace organics, regulated constituents and supply ownership. |
| Seawater | Intake screening, pretreatment/UF, SWRO—often with a second pass—and final demineralisation. | Intake impacts, biofouling, salinity, boron where relevant, marine outfall dispersion and final polishing. |
Mixed-bed ion exchange vs EDI/CEDI
Mixed-bed ion exchange can deliver very high purity, but resin regeneration requires acid/caustic systems or off-site exchange. EDI/CEDI continuously removes ions from suitable RO permeate using membranes, resin and electricity, avoiding routine chemical regeneration. That does not make the whole water plant chemical-free: pretreatment, RO cleaning and EDI maintenance can still require chemicals.
Selection depends on feed variability, product specification, minimum flow, turn-down, redundancy, cleaning strategy, waste route, operator capability and total lifecycle cost. EDI should not be chosen merely because the electrolyser operates continuously.
Treatment Recovery Changes Withdrawal, Not Reaction Chemistry
If the stack needs 10 L/kg of demineralised make-up, raw source-water withdrawal for that stream is approximately:
| Net treatment recovery | Raw water needed for 10 L demin water | Concentrate/reject generated |
|---|---|---|
| 90% | 11.1 L | 1.1 L |
| 75% | 13.3 L | 3.3 L |
| 50% | 20.0 L | 10.0 L |
| 33% | 30.3 L | 20.3 L |
Reject is not automatically consumed. It may be reused, discharged or disposed of, subject to chemistry and permits. Report withdrawal, recovered flow, discharge and consumption separately.
Choosing the Water Source
The cheapest treatment train is not necessarily the lowest-risk supply. Screen each option against:
- Current and future basin stress: include drought, seasonal variability and climate projections, not only annual average availability.
- Rights and priority: confirm abstraction rights, utility capacity, curtailment rules and competing community/agricultural demand.
- Quality variability: obtain a representative seasonal dataset and define upset cases.
- Conveyance and storage: pipelines, pumping, buffer capacity and redundancy can cost more than the polishing skid.
- Residuals route: determine where RO concentrate, filter backwash, spent media, cooling blowdown and cleaning waste will go.
- Social licence: even a modest volume can be contentious in a highly stressed basin or where potable access is constrained.
Reclaimed water can reduce pressure on freshwater supplies and offer a dependable industrial source, but it still needs treatment matched to the electrolyser. Seawater avoids freshwater abstraction but introduces intake, energy and concentrate-disposal impacts. Groundwater can look simple until sustainable-yield or salinity trends are considered. Use basin tools such as WRI Aqueduct for screening, then complete local hydrology and stakeholder studies.
Does Desalination Use Too Much Energy?
Usually not relative to electrolysis electricity, but the result depends on the stated boundary. The IEA reported in 2026 that seawater reverse osmosis typically uses about 2.5–6 kWh per cubic metre of product water for core desalination and associated steps.
- At 10 L/kg H2: about 0.025–0.060 kWh/kg H2
- At 20 L/kg H2: about 0.050–0.120 kWh/kg H2
- Against a 52 kWh/kg electrolyser system: about 0.05–0.23%
That calculation uses product-water volume, so no additional correction for SWRO recovery should be applied to the quoted kWh/m3. Conveyance, final polishing, storage and brine treatment can add energy.
The percentage is small, but it does not prove desalinated water is cheap or easy. Intake/outfall construction, marine studies, pipelines, pretreatment, redundancy and brine compliance can dominate water-system cost and schedule.
Can seawater go directly into an electrolyser?
Not as the bankable standard for commercial hydrogen projects in 2026. Direct seawater electrolysis continues to face chloride corrosion, competing chlorine chemistry, mineral precipitation and durability challenges. Conventional projects desalinate and polish water before it reaches the stack. Treat direct-seawater claims as emerging technology and request independently validated current density, operating hours, degradation, product purity and scale.
Recycling: What Can and Cannot Be Recovered
The 8.93 L/kg converted by the electrochemical reaction cannot be recovered inside the hydrogen plant unless the hydrogen is later recombined with oxygen at the same site. Other streams may be reusable:
- Hydrogen-side condensate: may be returned after separation and confirmation that it meets the stack-water specification.
- RO or EDI concentrate: may be used as cooling-tower make-up or in a lower-grade process if scaling, corrosion and permit limits allow.
- Cooling blowdown: can sometimes be treated for reuse, but recovery increases the concentration of salts and treatment residuals.
- Filter backwash: may be settled and recycled to the head of the plant where source-water and sludge management permit it.
- Recovered heat: does not save reaction water directly, but it can reduce cooling duty and therefore evaporative loss.
A claimed “closed loop” should identify its purge and final sinks. Evaporation, solid waste moisture, product gas moisture and the electrochemical reaction prevent a real plant from recycling every litre indefinitely.
Brine and ZLD: Choose the Permitted Route, Not the Acronym
RO concentrate composition depends on the source and pretreatment chemicals. Potential routes include discharge to a municipal or industrial wastewater system, permitted surface-water or marine discharge, evaporation ponds, deep-well injection where lawful, further concentration, or zero liquid discharge (ZLD).
ZLD is not an automatic requirement and is not impact-free. Conventional ZLD commonly uses brine concentration and crystallisation, which can add substantial energy, chemicals, equipment and solid-waste disposal. It can be justified where discharge is prohibited, water recovery is unusually valuable or the brine can be integrated with another process. Mineral recovery should be credited only after composition, purity, market, logistics and offtake have been demonstrated.
Current UK environmental guidance illustrates the correct decision framework: identify each effluent, minimise it, assess reuse, and then treat or discharge it to site-specific requirements. It does not prescribe ZLD for every electrolyser.
Water-Balance and Monitoring Requirements
A bankable water balance should cover normal, maximum, minimum and transient operation and include:
- Raw-water withdrawal by source and quality
- Demineralised stack make-up and electrolyser circulation
- RO/EDI/mixed-bed recovery and reject flows
- Cooling make-up, evaporation, drift and blowdown
- Hydrogen/oxygen condensate and gas-stream moisture
- Backwash, cleaning, drains, sampling and maintenance events
- Discharge flow, destination and contaminant load
- Startup, shutdown, outage and off-specification water volumes
Track at least raw-water withdrawal, direct consumption and discharge in L/kg of on-spec hydrogen. Also monitor conductivity/resistivity, flow, pressure, temperature, silica/TOC or other OEM contaminants, RO differential pressure and salt rejection, and polisher performance. Keep indirect electricity-related water use as a separate life-cycle KPI.
Project Due-Diligence Checklist
- Define the boundary. State whether each number is withdrawal, consumption, discharge, demin make-up or life-cycle use.
- Obtain the OEM specification. Use constituent limits and warranty conditions, not a generic PEM/alkaline conductivity table.
- Sample the source over time. Design for seasonal and upset chemistry, not one laboratory report.
- Build hourly and annual balances. Include treatment recovery, cooling, turndown, starts, cleaning and capacity factor.
- Test water availability locally. Verify rights, drought reliability, future basin stress and competing demand.
- Select cooling with water and energy together. Compare dry, hybrid, evaporative and once-through cases on the same environmental boundary.
- Define every residual stream. Characterise concentrate, blowdown, spent resin/media and cleaning waste before choosing a disposal route.
- Pilot difficult feeds. Reclaimed water, high-silica groundwater and variable surface water may justify a seasonal pilot.
- Design resilience. Include storage, duty/standby treatment capacity, analyzer redundancy and a response to off-spec water.
- Engage regulators and communities early. Do not wait for detailed design to discover a water-right, outfall or social-licence constraint.
Frequently Asked Questions
How much water is consumed per kilogram of green hydrogen?
The electrochemical reaction consumes 8.93 L/kg H2. Direct plant consumption is higher when water treatment, gas-stream losses or evaporative cooling are included. Current DOE PEM scenarios range from about 9 to 30 L/kg onsite, while other planning assumptions can exceed 50 L/kg.
Does a PEM electrolyser always need water below 0.1 µS/cm?
No. Requirements are product-specific. A current Accelera PEM sheet specifies ASTM D1193 Type III Grade B, while Nel documents other products using different water grades or resistivity. Obtain the complete OEM specification, including non-conductivity contaminants.
Is 15–25 L/kg a reliable project assumption?
It can be a screening range for some designs, but it is not universal. Dry-cooled projects can be lower; evaporatively cooled projects can be higher; and once-through cooling can withdraw orders of magnitude more while consuming less. Build a site-specific balance.
Can treated wastewater be used?
Yes, if it can be treated reliably to the electrolyser and cooling specifications. Evaluate source variability, trace contaminants, biofouling, ownership, outage risk, concentrate disposal and public-health regulations. “Reclaimed” describes the source, not the final quality.
Does desalination materially increase electrolyser electricity use?
The desalination electricity is usually a small fraction of stack electricity. At 2.5–6 kWh/m3, producing 10–20 L of desalinated water adds roughly 0.025–0.12 kWh/kg H2. Water-system capital, conveyance and permitting may still be material.
Is EDI chemical-free?
EDI avoids routine acid and caustic regeneration during normal polishing, but the overall treatment plant can still use pretreatment chemicals, antiscalant, membrane cleaners and maintenance chemicals. Describe it as continuous deionisation without routine chemical regeneration, not as a chemical-free plant.
Is zero liquid discharge required?
Only where the permit, water-recovery objective or project economics justify it. Many sites can use permitted reuse or discharge routes. ZLD transfers the remaining salts into a solid stream and can add substantial energy, complexity and disposal requirements.
The Bottom Line
The defensible answer is not “green hydrogen uses 9 litres of water.” It is: electrolysis consumes 8.93 L/kg in the reaction, then the plant's source, treatment recovery, cooling system and electricity supply determine withdrawal and total consumption. State the boundary, use the selected OEM's water specification, and design the residuals route at the same time as the purification train.
Official and Primary Sources
- U.S. DOE — Program Record 25003: Water Use in Hydrogen Production
- U.S. DOE — Hydrogen Shot: Water Electrolysis Technology Assessment
- UK DESNZ — Water Demand for Hydrogen Production
- UK Environment Agency — Hydrogen Production by Electrolysis: Environmental Guidance
- IRENA and Bluerisk — Water for Hydrogen Production
- Accelera by Cummins — HyLYZER 1000-30 Specification, Rev. 10/25
- Nel Hydrogen — A485 Alkaline Electrolyser Specification
- Nel Hydrogen — Current Water-Quality FAQ
- ASTM International — ASTM D1193-24 Reagent Water Specification
- ISO — ISO 3696:1987 Water for Analytical Laboratory Use
- IEA — Desalination Electricity and Water-Stress Context
- U.S. EPA — Industrial Water Reuse Resources
- U.S. EPA WaterSense — Water Purification and RO Recovery
- World Resources Institute — Aqueduct Water Risk Atlas
- European Commission JRC — Hydrogen Production via Direct Seawater Electrolysis