Fresh water

The Cost of Fresh Water

Desalination has fallen from about sixteen kilowatt hours a cubic metre to under three, and the thermodynamic floor is about one. This is an account of a frontier that is nearly finished, and of why water is still expensive anyway.

Research through 20 September 2026
Figure 1

A curve that is running out of room

Unlike almost every other frontier in this series, this one has a hard floor that can be calculated exactly, and industry is already within sight of it.

Energy required to produce a cubic metre of fresh water, 1960 to 2025On a logarithmic scale, seawater reverse osmosis falls from about sixteen to under three kilowatt hours per cubic metre and approaches a shaded thermodynamic floor of about one kilowatt hour, leaving roughly a factor of three of remaining runway.Thermodynamic minimum for seawater at 50% recovery · 1.06 kWh/m³no seawater process can operate below this line, at any price3010310.41960197019801990200020102020kWh per cubic metre, logarithmicThermalSeawaterBrackishPotable
Seawater reverse osmosis · 2025

Best large plants operate near 2.5 to 3 kWh per cubic metre, within roughly three times the thermodynamic minimum at practical recovery.

Figure 1: Indicative specific energy consumption for producing a cubic metre of fresh water, on a logarithmic scale. Values are representative of good practice for each technology at each date, not fleet averages, and exclude intake works, conveyance, and distribution. The shaded band is the thermodynamic minimum work of separation for 35 g/L seawater at 50% recovery, which is about 1.06 kWh/m³. No process operating on seawater can go below it at any price. Brackish water and treated effluent have their own, far lower floors, which is why those two curves legitimately sit beneath the seawater line. Thermal figures are expressed as equivalent electrical energy and depend heavily on whether the heat is a by-product of power generation. Brackish and reuse curves flatten far above their own much lower floors because their energy is dominated by pumping rather than separation.
~6×Approximate fall in seawater reverse-osmosis energy since 1970, against eight orders of magnitude for DNA sequencing over a comparable period.
~2.5×Roughly what remains between the best plants today and a thermodynamic minimum that cannot be crossed.
Energy is not the billAt ordinary electricity prices the thermodynamic floor is worth a few cents per cubic metre. Delivered water costs far more, and the difference is not physics.

What the record shows

  • The separation problem is nearly solved. Best-practice seawater plants run within roughly a factor of three of a floor that is fixed by thermodynamics, so the remaining energy runway is small and well understood.
  • The single largest historical gain was not a better membrane. It was recovering the pressure from the brine stream, which roughly halved specific energy consumption in about a decade.
  • The cheapest new water is usually the least saline water: treated effluent and brackish groundwater need a fraction of the work that seawater does, and are often closer to demand.
  • Delivered cost is dominated by capital, utilisation, brine disposal, and the energy of moving water to people. A plant that runs at half capacity produces water at roughly double the capital cost per cubic metre.

This report separates the thermodynamic minimum work, the specific energy of a process, the production cost at a plant gate, the contracted price of water from a project, and the tariff a household pays. They are different quantities and should not be placed on one curve.

Figure 2: The supply trade space

Desalination is the last option, not the first

Sources are ranked by the work required to separate salt from water. The cheapest new supply is almost always the one with the least salt in it.

A supply that does not depend on rainfall anywhere

Process
High-pressure reverse osmosis with brine energy recovery
Energy
2.5 to 4 kWh/m³
Capital
High, and sensitive to utilisation
Reliability
Drought-proof and effectively unlimited in volume

Binding constraint: Energy, capital, coastal siting, intake ecology, and brine discharge, plus the cost of moving water inland and uphill

Energy ranges are indicative for the treatment process alone and exclude conveyance, lift, and distribution, which in inland cities can exceed the energy of desalination itself.
Part I: First principles

Separating salt from water has a price no one can negotiate

Dissolved salt lowers the free energy of water. Removing it means putting that energy back, and the amount required can be computed from the salinity alone, before any equipment exists.

Delivered costseparation work ÷ efficiency + capital ÷ utilisation+brine disposal + conveyance + losses
01

Find a feed

Every supply begins with a source and a salinity. The salt content of the feed sets the minimum work required before any equipment is chosen, and nothing downstream can undo that choice.

Measure
Salinity, turbidity, biology, distance, reliability
Failure boundary
Choosing the saltiest available feed when a less saline one exists nearby
What the record shows at this step
Reported evidence
Treated municipal effluent is roughly a thirtieth as saline as seawater, which is why potable reuse needs a fraction of the separation work.
Where it is moving
Treating source selection as the first engineering decision rather than a fixed constraint.
Principal risk
Choosing the saltiest available feed when a less saline one exists nearby

Evidence statements describe published engineering practice or operating plants. “Where it is moving” is an editorial reading of direction, not an announced capability.

02

Take it in without harm

Intakes must supply enormous volumes without entraining marine life or drawing in material that will foul the membranes days later.

Measure
Flow, impingement, entrainment, screening
Failure boundary
Ecological damage at the intake, and pretreatment failures that shorten membrane life
What the record shows at this step
Reported evidence
Subsurface and screened intakes reduce ecological impact but add cost and are not feasible at every site.
Where it is moving
Intake design as an environmental permitting requirement rather than a hydraulic afterthought.
Principal risk
Ecological damage at the intake, and pretreatment failures that shorten membrane life

Evidence statements describe published engineering practice or operating plants. “Where it is moving” is an editorial reading of direction, not an announced capability.

03

Pretreat the water

Particles, organics, and biology are removed before the membranes see them. This unglamorous step determines how long the expensive part of the plant survives.

Measure
Silt density index, chemical dose, backwash frequency
Failure boundary
Biofouling and scaling, which raise pressure, energy, and replacement rates simultaneously
What the record shows at this step
Reported evidence
Membrane replacement and chemical cleaning are recurring operating costs directly set by pretreatment quality.
Where it is moving
Fouling-resistant surfaces and better monitoring, which extend membrane life more than flux improvements save energy.
Principal risk
Biofouling and scaling, which raise pressure, energy, and replacement rates simultaneously

Evidence statements describe published engineering practice or operating plants. “Where it is moving” is an editorial reading of direction, not an announced capability.

04

Separate salt from water

Pressure above the osmotic pressure of the feed pushes water through a membrane and leaves salt behind. This is the only step with a calculable thermodynamic floor, and it is the step nearly won.

Measure
Specific energy, recovery ratio, salt rejection, flux
Failure boundary
Pushing recovery too high, which raises brine osmotic pressure and the energy floor with it
What the record shows at this step
Reported evidence
Best large plants operate within roughly a factor of three of the thermodynamic minimum at practical recovery.
Where it is moving
Better second-law efficiency at the margin. The remaining runway here is a factor of two or three, not ten.
Principal risk
Pushing recovery too high, which raises brine osmotic pressure and the energy floor with it

Evidence statements describe published engineering practice or operating plants. “Where it is moving” is an editorial reading of direction, not an announced capability.

05

Recover the pressure

Brine leaves the membranes still pressurised. Returning that energy to the incoming feed is the single largest efficiency gain in the history of the process.

Measure
Recovery device efficiency, parasitic losses
Failure boundary
Treating energy recovery as optional, which doubles specific energy consumption
What the record shows at this step
Reported evidence
Isobaric pressure exchangers recover brine energy at very high efficiency and account for much of the fall visible in Figure 1.
Where it is moving
Incremental. This lever has largely been pulled, which is why the curve flattens after 2010.
Principal risk
Treating energy recovery as optional, which doubles specific energy consumption

Evidence statements describe published engineering practice or operating plants. “Where it is moving” is an editorial reading of direction, not an announced capability.

06

Dispose of the brine

Concentrate must go somewhere. At the coast that is a diffuser and a permit; inland it is evaporation ponds, deep wells, or an unsolved problem.

Measure
Concentration factor, discharge salinity, disposal cost
Failure boundary
Inland plants with no viable disposal route at any recovery ratio
What the record shows at this step
Reported evidence
Brine management, not membrane performance, is the binding constraint on most inland desalination.
Where it is moving
Resource recovery from brine, which is technically possible and rarely economic on its own.
Principal risk
Inland plants with no viable disposal route at any recovery ratio

Evidence statements describe published engineering practice or operating plants. “Where it is moving” is an editorial reading of direction, not an announced capability.

07

Move it to people

Water is heavy and cities are frequently inland and uphill. Conveyance and lift can consume more energy than the desalination that produced the water.

Measure
Distance, elevation gain, pumping energy, network losses
Failure boundary
A cheap plant delivering expensive water because of where it had to be built
What the record shows at this step
Reported evidence
Lifting a cubic metre one hundred metres costs roughly 0.3 kWh before friction, comparable to a significant share of modern desalination energy.
Where it is moving
Siting supply near demand, and reducing the water lost in distribution before building new supply.
Principal risk
A cheap plant delivering expensive water because of where it had to be built

Evidence statements describe published engineering practice or operating plants. “Where it is moving” is an editorial reading of direction, not an announced capability.

The membrane was never the whole problem. The brine, the pipe, and the empty plant are the problem.
Figure 3 · The recovery wall

Every extra drop costs more than the last

Taking more fresh water out of a given feed leaves the remaining brine saltier, which raises its osmotic pressure, which raises the work needed for the next drop. The floor is not a constant. It climbs with ambition.

Feed:
Minimum separation work against recovery ratioThe thermodynamic minimum work rises slowly at low recovery and steeply above about seventy percent, so pushing a plant toward zero liquid discharge raises its energy floor sharply.012345610%25%50%75%90%Recovery ratio: share of the feed turned into product waterkWh per cubic metreTypical SWRO0.94 kWhHigh recovery1.06 kWhInland limit1.41 kWhNear-zero discharge1.95 kWhfloorat 40%

The solid line is the thermodynamic floor for seawater · 35 g/l; the second line is what a plant achieving 40% of that ideal would consume. Best large seawater plants operate somewhere around 25 to 40% of the ideal, which is why the remaining engineering runway is a factor of two or three, not a factor of ten.

Figure 3: Minimum work computed as (πfeed ÷ r) × ln(1 ÷ (1 − r)) for an ideal solution, where πfeed is expressed as 0.764 kWh/m³ for 35 g/L seawater at 25 °C. The model assumes perfect salt rejection, no pressure losses, complete mixing, and an ideal solution; real seawater departs from ideality and real plants add pretreatment, pumping, and losses. The steep rise above roughly 70% recovery is the reason brine management, not membrane performance, dominates inland and zero-discharge designs.

Five supplies, five different amounts of work

Because the minimum work is set by salinity, the ranking of supply options is fixed by physics before any engineering choice is made. What varies is availability, reliability, and where the water has to travel.

SourceTypical process energyWhat it depends onBinding constraint
Surface water0.2 to 0.5 kWh/m³Rainfall, catchment, upstream useThe resource is fixed and contested
Groundwater0.3 to 2 kWh/m³Aquifer depth and rechargeExtraction routinely exceeds recharge
Potable reuse1.2 to 2 kWh/m³Population and existing collectionAcceptance, monitoring, and redundancy
Brackish RO0.5 to 1.5 kWh/m³Inland saline aquifersNowhere to send the brine
Seawater RO2.5 to 4 kWh/m³Coastal siting and powerCapital, intake, brine, and distance to demand

Ranges describe the treatment process alone. Conveyance and lift are excluded and can exceed these figures where demand is inland or at elevation.

The scale of what is already built

1.06 kWhThe thermodynamic minimum work to produce a cubic metre of fresh water from seawater at 50% recovery. It is a fixed property of the feed, not an engineering target.
~0.3 kWhEnergy to lift that same cubic metre one hundred metres, before friction, comparable to a meaningful share of the desalination itself.
~2.5 kWhRoughly where the best large seawater plants now operate, which is a second-law efficiency in the region of 40%.

A technology that is within a factor of three of its physical limit is not a technology with a tenfold improvement ahead of it. The optimistic case for water has to be built out of something other than better membranes.

The discovery chain

Two inventions did most of the work: a membrane with a thin dense skin, and a device that gives the brine's pressure back.

  1. 1950s

    Thermal desalination at scale

    Multi-stage flash distillation made seawater drinkable industrially, at an energy cost that only fuel-rich regions could absorb.

  2. 1959-1960

    The asymmetric membrane

    Loeb and Sourirajan produced a cellulose acetate membrane with a thin dense skin, making reverse osmosis practical rather than theoretical.

  3. 1965

    First municipal RO plant

    A small plant in California demonstrated the process in continuous public service.

  4. 1977-1980

    Thin-film composite membranes

    Interfacial polymerisation produced polyamide membranes with far better flux and rejection, and they remain the industry standard.

  5. 1990s

    Energy recovery becomes standard

    Turbines and then isobaric pressure exchangers returned brine pressure to the feed, roughly halving specific energy consumption.

  6. 2005-2015

    Gigalitre-scale plants

    Very large coastal plants demonstrated seawater desalination near three kilowatt hours per cubic metre at municipal scale and contracted prices.

  7. 2008-2025

    Reuse becomes respectable

    Advanced treatment of municipal effluent moved from emergency measure to planned supply in several water-stressed cities.

Who is building what

Membranes, energy recovery, large project delivery, and reuse programmes are pursued by different organisations. Search the record, or filter by which part of the system a programme works on.

11 programmes
Energy Recovery IncPX Pressure ExchangerIsobaric transfer of brine pressure to the incoming feed
Reported evidence
Very high recovery efficiency, and the single component most responsible for the fall in specific energy visible in Figure 1.
Announced next step
Incremental efficiency and application to other high-pressure processes.
Unresolved risk
This lever has largely been pulled; the remaining gain from energy recovery is small.
DuPont Water SolutionsFilmTec membranesThin-film composite polyamide elements
Reported evidence
Interfacial polymerisation chemistry from around 1980 remains the industry standard for both seawater and brackish elements.
Announced next step
Higher flux, better rejection, and greater fouling resistance.
Unresolved risk
Membrane improvements now move energy consumption only modestly, because the process is close to its thermodynamic floor.
Toray and Nitto / HydranauticsRO element manufactureHigh-rejection seawater and low-energy brackish elements
Reported evidence
Competing element suppliers whose products define achievable flux and rejection at commercial scale.
Announced next step
Longer membrane life and lower cleaning frequency.
Unresolved risk
Life and fouling, rather than flux, now dominate the operating cost the membrane controls.
IDE TechnologiesSorek and Sorek 2Very large coastal seawater plants with vertical pressure vessels
Reported evidence
Among the largest seawater plants built, delivering municipal water at contracted prices far below earlier projects.
Announced next step
Scale and standardisation to lower capital cost per unit capacity.
Unresolved risk
Contracted prices reflect financing, offtake guarantees, and power arrangements, not production cost alone.
Acciona, Veolia, and SuezPlant delivery and operationDesign, build, and operate contracts for municipal supply
Reported evidence
A mature international contracting market where cost is driven by site works, financing, and offtake structure.
Announced next step
Lower delivered cost through standard designs and long operating contracts.
Unresolved risk
A plant built as drought insurance and run at low utilisation produces very expensive water.
Orange County Water DistrictGroundwater Replenishment SystemAdvanced treatment of municipal effluent with membranes and ultraviolet oxidation
Reported evidence
One of the largest potable reuse schemes in continuous operation, treating a feed far less saline than seawater.
Announced next step
Expansion and replication in other water-stressed regions.
Unresolved risk
Acceptance, trace contaminants, and the monitoring redundancy the public reasonably expects.
Saudi Water AuthorityNational desalination fleetLarge-scale conversion from thermal distillation to reverse osmosis
Reported evidence
The largest national desalination programme, shifting capacity from multi-stage flash toward membranes.
Announced next step
Lower specific energy and integration with low-cost generation.
Unresolved risk
Conveyance and lift to inland demand centres, and brine discharge into a confined sea.
NEOM and coupled renewable projectsRenewable-powered desalinationDesalination timed to abundant low-cost electricity
Reported evidence
Announced projects pair large membrane capacity with renewable generation and storage.
Announced next step
Water as a flexible electricity load, stored as water rather than as power.
Unresolved risk
Announced capacity and price targets, not demonstrated operating results.
Academic membrane researchUniversity programmesTransport theory, fouling, and alternatives to polyamide
Reported evidence
Published work establishes how little energy remains to be saved in separation and redirects attention to selectivity and fouling.
Announced next step
Selective removal of specific ions such as boron, and resource recovery from brine.
Unresolved risk
Laboratory flux gains rarely translate into system energy savings near the thermodynamic floor.
US Bureau of ReclamationDesalination and reuse researchPublic funding for inland desalination and brine management
Reported evidence
Sustained programmes targeting the inland brine problem that has no coastal equivalent.
Announced next step
Viable disposal or beneficial use of concentrate away from the coast.
Unresolved risk
Brine management remains the binding constraint on inland desalination at any membrane performance.
Utility network programmesLeak reduction and meteringDetecting and repairing losses in distribution networks
Reported evidence
Many systems lose a substantial share of treated water before it is billed, and repair often costs less per cubic metre than new supply.
Announced next step
Treating loss reduction as a supply option with a measurable cost per cubic metre.
Unresolved risk
Maintenance capital is politically less attractive than visible new infrastructure.

Plant capacities and energy figures are reproduced from the record below. A contracted water price is not a production cost: it reflects financing, offtake guarantees, land, power contracts, and local subsidy.

Part II: The economics of delivery

The physics is cheap. The infrastructure is not.

Priced at ordinary electricity rates, the thermodynamic floor is worth a few cents per cubic metre. Delivered water usually costs between ten and a hundred times that, and almost none of the difference is separation.

Figure 4 · Interactive model

What does a cubic metre actually cost?

The physics sets a floor of a few cents. Everything above that is capital, utilisation, maintenance, and the energy of moving water to where people are.

Large coastal SWRO$0.80/m³

Production cost at the plant gate

Capital recovery
$0.31 · 39%
Energy
$0.23 · 29%
Membranes, chemicals, labour
$0.26 · 32%
Thermodynamic floor
$0.07

30-year life at 6% real cost of capital; operating cost taken as 6% of capital per year covering membranes, chemicals, labour, and maintenance. Excludes distribution networks, non-revenue water losses, brine disposal permits, tariffs, and subsidies. The floor line prices the thermodynamic minimum at the selected electricity price, for comparison only.

Calculation & assumptions

Cost per cubic metre = capital cost × capital recovery factor ÷ annual output + (process energy + lift) × electricity price + annual operating cost ÷ annual output, with annual output expressed as cubic metres produced per cubic metre per day of installed capacity.

Move the utilisation slider before anything else. A desalination plant is a capital asset that must run to be cheap, and take-or-pay contracts exist precisely because an idle plant produces the most expensive water there is. Energy matters, but it is rarely the largest line at ordinary electricity prices.

Figure 4: An editorial model of production cost at the plant gate, not a tariff, a bid price, or a delivered retail price. Real projects differ on intake and outfall works, site conditions, financing structure, environmental mitigation, and contracted offtake. Published water tariffs frequently sit below production cost because water is politically priced, which is itself one of the constraints discussed below.
UtilisationCapital is recovered over the water actually produced. A plant held as drought insurance and run at a fraction of capacity makes the most expensive water on the system.
DistanceConveyance and lift are pure energy with no thermodynamic floor to approach. Geography sets this line, and no membrane improves it.
LeaksWater lost between the plant and the customer is paid for twice. In many networks, reducing losses is cheaper per cubic metre than any new supply.

What pushes cost down next

01

Use a less salty feed

Reuse and brackish sources start with a fraction of the separation work and are usually nearer to demand, which attacks both the energy and the conveyance terms at once.

Condition: public acceptance and monitoring that earns trust
02

Run the plant harder

Utilisation is the largest single lever in the cost model. Contracts and system planning that keep plants loaded matter more than incremental efficiency.

Condition: demand and tariffs that support baseload operation
03

Follow cheap electricity

Energy is a commodity input, so flexible operation timed to abundant low-cost power lowers the energy line without touching the process.

Condition: plants designed and financed for variable operation
04

Stop losing what is already treated

Leak detection and network renewal produce water at a cost per cubic metre that frequently beats new supply, with no permit and no intake.

Condition: capital for unglamorous maintenance rather than new assets
Part III: Where progress is stuck

A nearly finished technology inside an unfinished system

The separation problem has a floor in sight. Brine, capital, distance, losses, and pricing have no such floor, and they are where the remaining cost of water lives.

The floor is close

Seawater desalination now operates within roughly a factor of three of a thermodynamic minimum that no technology can beat. The energy story is largely over; a tenfold improvement is not available.

Brine has nowhere to go

Coastal plants dilute and discharge under permit. Inland plants face evaporation ponds, deep injection, or nothing, and the recovery wall makes concentrating further expensive.

Capital and utilisation

A desalination plant is a capital asset with modest operating cost, so an idle plant produces very expensive water. Plants built as drought insurance are often the least economic to run.

The pipe and the lift

Moving water inland and uphill can cost more energy than producing it. Geography, not membrane science, decides whether desalinated water is affordable at the tap.

Losses before the customer

Many networks lose a substantial share of treated water to leaks before it is billed. Repairing distribution is frequently cheaper per cubic metre than building new supply, and far less visible.

Price signals and governance

Water is commonly priced below its cost of production for reasons that are legitimate and political. That suppresses both investment and conservation, and no engineering advance resolves it.

An optimistic view, with conditions

Water becomes a manufactured commodity with a known price

The optimistic case is not a breakthrough membrane. It is that water supply becomes boring: a portfolio of reuse, brackish treatment, and coastal desalination, run at high utilisation on cheap electricity, priced close to what it costs, in a network that does not leak. Every element of that already exists somewhere.

Now to 2030

Reuse stops being exceptional

Advanced treatment of municipal effluent becomes a planned part of supply in water-stressed cities rather than a drought measure, because it is the least saline and nearest feed available.

2030s

Plants follow the power price

Desalination becomes a flexible electricity customer, produced when power is abundant and stored as water, which is far cheaper to store than electricity.

Longer horizon

Scarcity becomes a price, not a crisis

With a known production cost and a working portfolio, the question shifts from whether water exists to whether a given use is worth what it costs, including for agriculture, which uses most of it.

View the analyst probability ranges
MilestoneDateAnalyst probability
Best-practice seawater plants operate below 2.5 kWh/m³ at scale203050-70%
Potable reuse is a planned, permanent supply in most large water-stressed coastal cities203540-60%
A commercially viable inland brine solution is deployed beyond demonstration scale203520-35%
Large desalination plants routinely operate as flexible electricity loads203335-55%
Seawater desalination reaches 50% second-law efficiency in commercial operation204025-40%
Desalinated water becomes economic for broad-acre agriculture without subsidy20455-15%

Editorial judgements conditional on the record above, not published forecasts or project guidance.

There is more than one finish line

  1. Physically possibleSalt can be separated from the feed at all.
  2. Energetically sensibleConsumption is within a small factor of the thermodynamic minimum.
  3. Environmentally permittedIntake and brine discharge are acceptable at the site.
  4. Economically producedCapital is recovered over a plant that actually runs.
  5. Delivered affordablyConveyance, lift, and network losses do not exceed the cost of production.
  6. Priced honestlyTariffs support maintenance and investment without excluding people from a basic need.

Sources, method, and boundaries

Figure 1 is an editorial reconstruction of representative best practice, not a fleet average or a continuous series. Figure 3 is an ideal-solution thermodynamic model, exact within its assumptions and approximate for real seawater. Figure 4 models production cost at the plant gate and excludes distribution, tariffs, and subsidy; it is not a bid price or a procurement estimate. Energy figures for thermal processes are expressed as equivalent electrical energy and depend on the heat source.

Minimum work
The thermodynamic least energy to separate pure water from a feed at a given recovery, independent of technology.
Specific energy
Energy actually consumed per cubic metre of product by a real process, including inefficiency.
Recovery ratio
The share of the feed stream that leaves as product water rather than brine.
Second-law efficiency
Minimum work divided by actual energy consumed: how close a plant runs to the floor.