Why Water Is the Underestimated Constraint in Green Hydrogen Projects
Green hydrogen production requires roughly 9–22 liters of high-purity water per kilogram of H2, depending on electrolyzer technology (PEM ~9 L/kg, alkaline ~18 L/kg including balance-of-plant losses). A 100 MW PEM plant running at ~60% capacity factor consumes approximately 4–6 million cubic meters of treated water per year at the upper bound, with realistic inland designs landing closer to 60,000–80,000 m³/yr when cooling reuse is applied — making the water treatment train, typically RO plus mixed-bed polishing to <1 µS/cm, a critical CAPEX and permitting line item in 2026 project design.
The IEA's 2024–2025 pathway reports project tens of millions of cubic meters of incremental industrial water demand annually from electrolyzer capacity additions through 2030 (per IEA Global Hydrogen Review 2024), yet most hydrogen project developers still under-resource the water scope. Roughly 98% of current hydrogen production remains tied to fossil-feedstock routes (per Bramble Energy / industry context), so the scale-up of electrolytic capacity is a step-change in water demand rather than a marginal increment. Three water streams are typically scoped separately: process feed water (purified to electrolyzer spec), cooling and balance-of-plant water (often cooling-tower makeup), and demineralized rinse water for stack conditioning. Each has its own quality bar, its own unit cost, and its own permitting footprint — and they are rarely additive in a water balance.
The article frames the problem per kg H2 rather than per MWh, because that is the unit financial and LCA models use. Every kilogram of hydrogen produced requires a stoichiometric kilogram of water input plus balance-of-plant losses, and the engineering job is to minimize the multiplier on that kilogram while staying inside the conductivity, chloride, and silica envelope the stack will accept for 80,000 operating hours.
How Much Water Does an Electrolyzer Actually Need?
Per-kilogram water consumption for a green hydrogen plant ranges from ~9 L for a well-instrumented PEM system to ~18 L for an alkaline system once balance-of-plant losses are included, with SOEC units sitting between the two at 10–14 L/kg but with tighter purity constraints. These numbers are the ones a feasibility engineer should paste into a water balance, and they are stable across 2024–2026 vendor datasheets.
| Electrolyzer type | Stoichiometric feed water (L/kg H2) | Typical BoP losses (L/kg) | Total feed water (L/kg) | Cooling water adder (% of feed) |
|---|---|---|---|---|
| PEM | 9 | 1–3 | 10–12 | 30–50 |
| Alkaline | 9–11 | 5–7 | 14–18 | 40–60 |
| SOEC (R&D / pilot) | 9 | 1–5 | 10–14 | 30–40 |
Worked example for a 100 MW PEM plant at 60% capacity factor: annual hydrogen output is ~4,500 tonnes (assuming ~50 kWh/kg stack efficiency), feed-water demand is ~50,000 m³/yr at 11 L/kg, and cooling-tower makeup adds another ~25,000 m³/yr at 40% — totaling roughly 75,000 m³/yr of treated and makeup water. Scale that to 1 GW and the plant approaches 0.5–0.7 million m³/yr of treated feed water, equivalent to the daily supply of a town of 25,000–30,000 people. Three parameters move the result by ±30%: capacity factor (50% vs 75% shifts H2 output by 50%), stack specific consumption (45 vs 55 kWh/kg shifts kWh demand and indirectly water per kg), and the cooling-tower cycles-of-concentration setting, which controls blowdown. A project board looking at this needs to know that a 10% swing in capacity factor is roughly a 5,000 m³/yr swing in feed water at 100 MW scale.
Feed-Water Quality Requirements: What Electrolyzer Makers Actually Demand
PEM electrolyzer feed water must meet a <1 µS/cm conductivity, <50 ppb TOC, <10 ppb silica, and <0.1 ppm chloride specification in 2024–2025 manufacturer datasheets, with alkaline stacks accepting a wider envelope of up to ~5 µS/cm but still requiring chloride control to protect stainless components. SOEC units are the strictest on most parameters and operate at 700–800 °C, which amplifies silica fouling on heat exchangers but reduces chloride-driven membrane attack.
| Parameter | PEM spec (typical 2024–2025 datasheet) | Alkaline spec (typical) | SOEC spec (pilot) | Why it matters |
|---|---|---|---|---|
| Conductivity | <1 µS/cm | <5 µS/cm | <0.5 µS/cm | Ionic contamination degrades membrane and electrodes |
| TOC | <50 ppb | <1,000 ppb | <50 ppb | Organic carryover fouls catalysts and GDEs |
| Silica (SiO2) | <10 ppb | <500 ppb | <10 ppb | Silica scales on heat exchangers and RO membranes; very hard to remove downstream |
| Chloride | <0.1 ppm | <0.5 ppm | <0.05 ppm | Chloride pits 316L stainless and attacks PEM membranes irreversibly |
| Iron / manganese | <0.01 ppm | <0.1 ppm | <0.01 ppm | Catalyst poisoning and discoloration in alkaline loops |
| Residual free chlorine | Not detectable | Not detectable | Not detectable | Oxidizer attack on membrane |
Silica and chloride are the two silent killers. Silica fouling on the polishing-loop heat exchangers and on the final RO pass is the most common cause of progressive throughput loss, and chloride breakthrough during a polishing-resin exhaustion event can take out a PEM membrane in hours. Online conductivity at the polishing outlet is the single most useful sensor for protecting the stack — the Mettler-Toledo application note on green-hydrogen conductivity monitoring lays out the same logic. TOC and silica analyzers are slower but should be trended continuously for 2026 designs. If the source water is variable (a real risk for treated wastewater or surface water), redundant polishing trains with automatic diversion-to-drain on conductivity excursion are becoming standard rather than optional.
Source-Water Options: Tap, Groundwater, Treated Wastewater, or Seawater?
Source-water choice is the single most consequential decision in the project water plan: it drives pretreatment CAPEX, ongoing OPEX, permitting timeline, and ESG narrative, and it is decided long before the electrolyzer PO is signed. In 2025–2026 industrial tariffs across the US, EU, Middle East, and Australia, the unit cost spread between the four main options is roughly 5–8×, and the treatment intensity spread is even wider.
| Source | Typical conductivity (µS/cm) | Indicative 2025–2026 OPEX ($/m³ treated) | Treatment intensity | Best-fit project profile |
|---|---|---|---|---|
| Tap / municipal | 200–800 | $0.50–1.50 | Low (RO + polishing only) | Fast-track, small-scale, water-rich regions |
| Groundwater / borehole | 300–1,500 | $0.20–0.70 | Medium (iron/manganese removal + softening + RO) | Inland sites with stable aquifer permits |
| Treated municipal wastewater | 500–1,500 | $0.20–0.60 | High (MBR + RO + AOP + polishing) | ESG-led projects, water-stressed inland regions |
| Seawater (coastal) | 45,000–55,000 | $0.30–0.80 (SWRO finished) | Very high (intake + SWRO + boron-selective + polishing) | Coastal mega-projects in Middle East, Chile, Australia |
Tap water has the fastest permitting path and the lowest pretreatment CAPEX, but the highest unit OPEX and a social-license problem in any water-stressed basin — a 100 MW plant at $1.00/m³ is $60,000–80,000/yr of just water tariff, before any treatment. Groundwater is the cheapest by unit cost but increasingly difficult to permit in the US West, Spain, and northern China, and requires iron, manganese, and hardness removal before RO. Treated municipal wastewater has become the lowest-OPEX option in many jurisdictions and the strongest ESG story under CSRD reporting, but it needs MF/UF + RO + AOP to land inside the electrolyzer envelope, and the public-perception work is non-trivial. Seawater, via seawater reverse osmosis (SWRO) followed by boron-selective resin polishing, is now the lowest-risk option for coastal mega-projects in the Middle East, Chile, and Australia, where it competes on OPEX with treated wastewater once intake and discharge permits are secured. The decision logic in 2026 is driven less by chemistry than by three external variables: distance to coast, local water-stress index, and the realism of obtaining a discharge permit for RO concentrate.
The 2026 Treatment Train: From Raw Water to Electrolyzer Feed
The 2026 default treatment train for an inland green hydrogen project is multimedia filtration → ultrafiltration or DAF pretreatment → brackish-water RO → EDI or mixed-bed polishing → 0.2 µm final filter, with the polishing outlet conductivity verified online before it touches the stack. For seawater, the train becomes intake screens → DAF/MF → two-pass SWRO → boron-selective ion exchange → EDI. For reclaimed wastewater, the train is MBR system for reclaimed-wastewater feed to a hydrogen plant → RO → EDI, which cuts source-water cost by 40–60% versus municipal tap at typical 2025–2026 industrial tariffs.
Two engineering details to size correctly: the multi-media filter for source-water pretreatment upstream of RO determines how often the RO membranes clean and how long they last, and the industrial RO system for electrolyzer feed-water polishing sets the salt and silica load the polishing resins have to handle. A 30–70% recycle of polishing reject is feasible on most designs and should be designed in from day one — it lowers net water consumption and reduces RO concentrate volume, which is the single biggest discharge-permitting issue for inland sites. Online monitoring at the polishing outlet (conductivity, TOC, silica) supports fast diversion to drain on any out-of-spec event and is the cheapest insurance against a stack replacement. For broader context on how these trains fit into regional permitting and reuse frameworks, the industrial wastewater treatment in San Francisco engineering guide and the high-strength organic wastewater treatment engineering specs piece both walk through the design constraints in more detail.
Cost, Risk, and Permitting: The 2026 Project View
Water treatment CAPEX for a 2025–2026 green hydrogen project runs $0.5–2.0 million per 10 MW of electrolyzer capacity depending on source water, and OPEX runs $0.20–1.50 per cubic meter of treated water, dominated by RO pump energy and membrane and resin replacement. A 100 MW inland project on treated wastewater should budget $8–15 million of water-treatment CAPEX and ~$0.4–0.8 million/yr of OPEX; a 100 MW coastal project on seawater will run 20–40% higher on CAPEX for the SWRO pass and boron-selective resin but lower on raw-water tariff. Water treatment is typically 3–6% of total project CAPEX but it drives 30%+ of the permitting timeline in 2026, and that ratio is what makes it a board-level item rather than a procurement detail.
Permitting friction in 2026 is real and tightening. The EU's CSRD water-stress screening now requires disclosed water-withdrawal figures at the project level for any project above 1 MW, and large groundwater rights transfers in the US West are triggering NEPA review. China's dual-carbon policy is actively favoring treated-wastewater reuse, which is why inland Chinese projects in Inner Mongolia and Xinjiang are increasingly specced on MBR + RO rather than fresh groundwater. The single biggest non-technical risk in 2026 is RO concentrate discharge: inland mega-projects are now expected to design for zero-liquid-discharge (ZLD) using a thermal brine concentrator and crystallizer, which can double the water-treatment CAPEX. For a closer look at the membrane side of this trend, the membrane technology market growth 2026 analysis is a useful reference.
Frequently Asked Questions
How many liters of water to make 1 kg of green hydrogen? Approximately 9–12 L for a PEM system and 14–18 L for an alkaline system once balance-of-plant losses are included (per typical 2024–2025 electrolyzer manufacturer datasheets).
Can seawater be used for hydrogen electrolysis? Yes — seawater reverse osmosis (SWRO) followed by boron-selective resin polishing and EDI is the standard 2026 configuration for coastal mega-projects in the Middle East, Chile, and Australia.
What is the minimum water purity for a PEM electrolyzer? Conductivity below 1 µS/cm, TOC below 50 ppb, silica below 10 ppb, and chloride below 0.1 ppm per typical 2024–2025 PEM manufacturer feed-water specifications.
Can treated wastewater feed a hydrogen plant? Yes — MBR followed by RO and EDI hits the electrolyzer spec, and treated wastewater is the lowest-OPEX source option in 2025–2026 industrial tariffs at $0.20–0.60/m³.
How much water does a 100 MW hydrogen plant use per year? A 100 MW PEM plant at 60% capacity factor uses approximately 50,000 m³/yr of purified feed water plus 20,000–30,000 m³/yr of cooling-tower makeup — roughly 70,000–80,000 m³/yr total.
