00Executive summary
The physics is real and elegant. The economics are not — by a factor you can hold in your head.
Pressure-retarded osmosis harvests the energy released when fresh water and seawater mix. That energy is genuine, weather-independent, and everywhere a river meets the sea. But two numbers decide whether it is a power plant or a science project, and on both, full-scale PRO misses by roughly an order of magnitude.
Statkraft — one of the most experienced hydropower operators on Earth — built the world's first osmotic prototype in 2009, studied it for five years, and walked away in 2014.[12] Seventeen years on, the entire global fleet is two co-located demonstration units (~100 kW each). No venture capital, no energy major, has funded a standalone plant. And the people who built the newest plant — the Fukuoka team — published the peer-reviewed math themselves showing standalone PRO sits at $0.20/kWh and 2.8 W/m².[4] The smartest money in the room has voted, and it voted no.
01The scorecard
Scored 0–10 across five gates
Economics is the binding constraint, so it scores lowest. The membrane IP is real (hence a 4 on moat), but seventeen years with zero standalone commercial plants and a departed pioneer caps maturity and timing.
Contents
02How it works
A hidden 270-metre waterfall at every river mouth
Place a semi-permeable membrane between seawater and fresh water. The membrane passes water molecules but blocks salt, and because salt pulls water toward it — osmosis, the same force that keeps your cells alive — fresh water crosses into the salty side. Keep that salty side under pressure and the water still crosses, because osmosis is strong enough to push against the pressure. Every litre that comes across arrives already pressurised; run that surplus high-pressure water through a hydro turbine and you get electricity. No fuel, no emissions, no weather dependence.[2]
The energy is real: mixing one cubic metre of river water into the sea releases about 1.4 megajoules[2] — roughly the electricity you'd get from dropping that cubic metre off a 40-storey building. Put differently, the salinity difference at a river mouth is equivalent to an osmotic pressure of ~27 bar — a ~270-metre column of water,[2] an invisible waterfall taller than the Hoover Dam, running day and night. Multiply by every estuary on Earth and you get the headline. This is the pitch, and it is a good one.
03The pioneer who quit
Seventeen years, two demos, zero standalone plants
This is not a new idea. The first chapter belongs to Statkraft, Norway's state-owned power giant — a century of hydropower expertise. In 2009 it opened the world's first osmotic prototype at Tofte (a ~2–4 kW pilot, ~2,000 m² of membrane), ran it for years, publicly assessed a 2 MW plant, and projected commercial osmotic power by 2020.[12] In 2013–2014 it halted and cancelled the program.[12] One of the most experienced water-power operators on the planet studied this for five years and quit.
| Year | Plant | Scale | Status | Note |
|---|---|---|---|---|
| 2009–2014 | Statkraft, Tofte (NO) | ~2–4 kW | Cancelled 2014 | World-first; pioneer exited [12] |
| 2023 | SaltPower, Mariager (DK) | ~100 kW | Demo | Hypersaline brine (~8× sea), Toyobo FO membranes [13] |
| 2025 | Mamizupia, Fukuoka (JP) | net ~110 kW | Demo | Bolted onto a seawater desalination plant [14][15] |
On the engineering readiness scale (1–9), PRO sits at TRL 4–5: demonstration plants only — zero commercial standalone facilities anywhere, seventeen years in. A 2026 peer-reviewed study puts it plainly: standalone PRO "remains elusive," and its real value is as an energy-recovery step in hybrid systems, "not stand-alone power generation."[7]
04The make-or-break number
Membrane power density: ~2.8 against the 56.4 it needs
One number decides whether osmotic power is a power plant or a science project: membrane power density — watts per square metre of membrane. The literature's commercial-viability floor is often cited at ~5 W/m² (though newer work argues even that is optimistic, putting feasibility nearer 6.5–50 W/m²).[8] To match the LCOE of ordinary solar — the real competitor, because a buyer just picks solar instead — you need 56.4 W/m².[4]
In a lab, on a small, flawless patch of membrane, researchers reach 12–16 W/m².[2] A real, full-scale module delivers about 2.8 W/m² net — and that figure comes from the Fukuoka builders' own peer-reviewed paper.[4] That is roughly one-twentieth of the solar-parity threshold, and below even the viability floor.
Membrane power density — real vs. what it takes (W/m²)
Real full-scale output (2.8) is ~1/20th of solar parity (56.4) and below the viability floor. Lab patches get partway, but only on small perfect membranes that don't survive full-scale pressure. [2][4][5]
Low power density is not a detail you optimise away — it is a money-burning machine. At ~2 W/m², the Fukuoka-class design needs roughly 45,000 m² of high-tech membrane to net on the order of 100 kW.[14] That is a basketball court of membrane to run about one microwave oven. And membranes foul, need pre-treatment, and the pumps consume ~10% of output before it leaves the plant, dragging whole-plant efficiency to 60–75% of gross.[7] Crucially, the most detailed plant-scale work finds that pushing membranes past ~10 W/m² buys almost nothing, because by then the cost is the pumps, pre-treatment and pressure vessels — not the membrane.[3] Better membranes don't save it; the whole system is expensive.
05Does it pencil?
$150–300/MWh standalone, against a ~$90 benchmark
Across the peer-reviewed techno-economic studies, the honest answer is a range — and the range is the whole story. For a standalone plant (the version in the big claim), agency and modern estimates run from about $150 up to $300–350 per MWh.[2][3] Pair it with hypersaline brine and bolt it onto existing infrastructure and the best current tech reaches ~$200/MWh — $0.20/kWh in the Fukuoka team's own calculation.[4] The absolute dream case, using membranes that exist only in a lab, touches $70–140/MWh.[3]
LCOE by scenario vs. the $90/MWh firm-power hurdle
Only the dream scenario (*needs membranes that don't exist at full scale) gets near the hurdle. The real standalone case is 2–4× over. The co-located niche (~$200) only pencils because the brine and the offset load are free. [2][3][4]
The cost is dominated by the membrane: 50–80% of total capital cost is membrane, exactly.[2] And membranes aren't bought once — fouling and pressure cycling make them a recurring bill. There is even peer-reviewed work whose entire conclusion is that this technology family is economically infeasible, because cheaper parts can't close a gap that the physics sets.[10] To be fair, PRO is the best of the salinity-gradient family — it captures ~37% of the theoretical maximum, versus under 10% for reverse electrodialysis and nanopore generators, whose costs exceed $0.60/kWh.[6] This is the family champion, and it still loses to the market by 2–4×.
Where the capital goes
06Where it actually works
Not a power plant — an efficiency bolt-on for desalination
The claim gets it backwards. Fukuoka isn't really a power plant; it's an efficiency gadget bolted onto a desalination plant — and as one of those, it's genuinely smart. A desal plant has a waste problem: it spits out brine twice as salty as the sea, and disposing of it costs money. Fukuoka feeds that free, already-concentrated brine into the osmotic unit on one side, treated wastewater on the other, and uses the electricity to shave the desal plant's own power bill.[15] Two waste streams in, a smaller utility bill out — a real, honest win.
But the math only works when a concentrated-brine source and a fresh-water source sit right next to each other. That single requirement shrinks the market from "every coastline on Earth" to "the back lot of a coastal desalination plant." And the output — ~880,000 kWh/year, about 100 kW continuous — is roughly one-thirtieth of a single modern wind turbine.[14] Useful and real; about a million times smaller than the headline that pulled everyone in.
07Follow the money
Who's in — and the louder fact of who isn't
Capital leaves fingerprints. In this space: membrane makers like Toyobo and Nitto, a few national labs, a handful of small ventures, and government demonstration money.[1][13] Who's not in it: venture capital and the energy majors. Seventeen years on, with one very public failed pioneer in plain sight, the smart money has looked at osmotic power and quietly kept its wallet shut. When nobody who profits from being right will fund it, that silence is data too.
08The 372-million number
The viral hook, and the figure that's actually sourced
The defensible "big number" is IRENA's technical potential of ~5,177 TWh/year — about 23% of global electricity[2] (the World Economic Forum's 2025 figure of "~20%" agrees[16]). That is a real, enormous theoretical ceiling. The gap between it and what's been deployed is the entire story: against a 5,177-TWh ceiling, the world's installed osmotic capacity in 2026 is two demos totalling ~0.2 MW. The physics permits an ocean of energy; the economics has delivered a puddle.
09The verdict
Overhyped for the grid; a narrow, real niche to watch
As a way to power the grid, osmotic power is overhyped — real power density is ~1/20th of solar-parity, the LCOE is 2–4× firm power, the pioneer quit, and it has missed its own commercialization date by six years. As an efficiency bolt-on for a coastal desalination plant with hypersaline brine, it is real, narrow, and worth watching — the Fukuoka and SaltPower model.
This is a "Not Yet," not a "Never" — there are concrete, monitorable triggers that would change it, and the sensitivity below shows exactly how far each lever has to move.
10Sensitivity & what would flip it
No achievable lever gets standalone PRO under the $90 line
A verdict is only as good as its sensitivity. We pushed each lever in the model to its best realistic value — and this is the thing the video can't show you: no combination of achievable, real-world inputs gets standalone osmotic power under the $90/MWh firm-power hurdle. Only the lab-membrane fantasy ties the market, and that membrane does not survive full-scale pressure.
| Lever | Base | Pushed to… | LCOE lands | Clears $90? |
|---|---|---|---|---|
| Membrane power density | 2.8 W/m² | 10 W/m² (where returns plateau) | ~$150/MWh | No |
| Membrane price | ~$3/m² | ~$1/m² | small move — the area drives cost, not the price | No |
| Discount rate | 8% | 5% | ~$210/MWh | No |
| All three + co-located brine | — | stacked, best case | ~$120–200/MWh | Niche only |
| Lab-only "unicorn" membrane | — | 14 W/m² + ideal | ~$70–105/MWh | Ties — doesn't exist at scale |
Directional, from the editable model (every input sourced — run your own in the spreadsheet). The pattern is robust: the membrane area forced by low power density is the cost driver, and it doesn't fall fast enough — exactly the conclusion of the peer-reviewed infeasibility work.[10][3]
The three triggers that would change this verdict
This is a "Not Yet," not a "Never." Three concrete, monitorable signals would move osmotic power from Overhyped to a genuine Watch:
- A full-scale module sustaining >10 W/m² net at survivable pressure (today: ~2.8). Above ~10 W/m² the economics finally start to bend; below it, better membranes barely help.
- A peer-reviewed standalone LCOE under ~$120/MWh that does not assume lab-only membranes. Today's honest dream case is $70–140 — but only with membranes that don't survive scale.
- The first venture or strategic capital into a standalone plant (not a co-located efficiency bolt-on). Capital validation is the market's own verdict — and it has been "no" for seventeen years.
None has moved since 2009. Watch those three; until one does, the niche is the bet, not the grid.
11Sources
16 sources — 10 peer-reviewed/agency (T1), 4 primary/company (T2, incl. Japanese & Norwegian originals), 2 reputable press (T3). Original-language titles preserved.
12Method & disclaimer
Every quantitative claim is traced to a source above. Two market benchmarks ($90/MWh firm power; $50–100/MWh solar+storage) are labelled as general market context, not osmotic-specific data. The ~45,000 m² membrane area is a single builder estimate, used order-of-magnitude. Numbers were re-verified 2026-06-13 against the latest available sources, including Japanese and Norwegian primaries. The companion financial model (editable, every input sourced) lets you run your own scenarios.