Ocean Thermal Energy Conversion

Ocean thermal energy conversion runs a heat engine on the vertical temperature difference in tropical ocean water: warm surface water at 25 to 28 °C against deep water at 4 to 5 °C drawn from roughly a kilometre down, giving a working temperature difference of about 20 to 24 K. In closed-cycle plants that difference boils and condenses a working fluid — ammonia, or R32 in the Korean barge — around a conventional Rankine loop. In open-cycle plants the warm seawater itself is flash-evaporated in a vacuum, driving a very large low-pressure turbine and yielding desalinated water as a by-product. Both configurations have been built. Both work. Neither has ever been built at a scale anyone would call a power station.

The idea is old enough that its age is part of the evidence. Jacques-Arsène d’Arsonval proposed tapping the ocean's thermal gradient in 1881; his student Georges Claude built the first plant at Matanzas Bay, Cuba, in September 1930. That is 145 years from proposal and 96 years from first demonstration, and the corpus of built plants across that span is short enough to list in full — which this brief does, in the engineering section, because a field with a complete enumerable deployment record should be argued about from the record rather than from the promise.

The discipline this brief imposes on itself, stated at the outset because it is the whole methodology. Every capacity figure here is marked gross or net, and where a source gives the split, the parasitic pumping load is stated with it. This is not pedantry. OTEC's parasitic load — the power consumed by the seawater and working-fluid pumps — has ranged from 52 to 80% of gross output in every plant whose outturn was published, so a gross figure quoted as though it were a net figure overstates delivered power by a factor of two to five. The field's promotional literature quotes gross. Government databases and the peer-reviewed literature quote net where they have it, and note where they do not. Nearly every dispute about OTEC's status dissolves once the two are kept apart.

Scope. This brief covers the thermodynamic cycle, the cold water pipe, the plant record, the resource ceiling and the environmental modelling. It also covers seawater air conditioning — SWAC — not as an aside but because the deep-cold-water pipe is the shared asset between the two, and cooling is the only product that has ever paid for one. Where OTEC's economics are argued from bundled cooling revenue, as the most favourable peer-reviewed cost study does, this brief says so rather than quoting the combined figure as an electricity price.

Established The all-time net power record for OTEC is about 103 kilowatts, it was set in Hawaii in the 1990s, and it has not been beaten in roughly thirty years. The Keahole Point open-cycle plant at the Natural Energy Laboratory of Hawaii Authority ran intermittently from 1993 to 1998 and produced up to 103 kW net from 255 kW gross on the United States government's own database — a parasitic share of 60%. The intergovernmental white paper gives slightly different figures for what appears to be the same plant, 100 kW net from 210 kW gross and a 1992 date, a parasitic share of 52%. This brief reports both and resolves neither; the correct way to quote it is 210 to 255 kW gross and about 100 to 103 kW net. Federal funding ended and oil prices collapsed, and the plant was decommissioned in 1998.

Established Nothing built since has published a net figure that beats it, and mostly nothing built since has published a net figure at all. Makai Ocean Engineering's plant at the same Hawaiian site, grid-connected in 2015 and the first closed-cycle OTEC ever connected to a United States grid, is rated 105 kW gross; its net output has never been published in eleven years, and the government database that lists it as active at technology readiness level 6 also records that it “does not operate on a continuous basis.” Okinawa Prefecture's Kumejima facility, running since 2013, is 100 kW gross across two 50 kW units; its net output has never been published in thirteen years, including in the prefecture's own final report. Korea's K-OTEC1000 barge achieved 338 kW gross off Busan in September 2019 at a temperature difference of only 18.7 °C — a genuine record for a seaborne plant, at off-design conditions, with no net figure published for that run against a design net of 877 kW at ΔT = 24 °C. So the honest statement is that the demonstrated frontier of OTEC is about a hundred kilowatts net, roughly the output of a mid-size diesel generator, and the record-holder was scrapped in 1998. The intergovernmental economics review of March 2024 states it plainly: there are currently no megawatt-scale plants in operation.

Established The cold water pipe is the binding physical problem, and the gap is a factor of seventy-three in cross-section. The largest deep cold-water pipe ever successfully deployed and operated anywhere is 1.40 m in diameter, reaching about 915 m depth, 3.2 km long, high-density polyethylene, installed at the HOST Park in Kailua-Kona, Hawaii, in October 2001 — and it is a seawater supply line for a research park, not a power plant intake. Two independent sources agree on it, the designer and the intergovernmental programme. A 100 MW-net plant, in the only detailed peer-reviewed design this brief could verify, needs a 12 m diameter pipe 800 m long in fibre-reinforced plastic with a syntactic foam core. The cross-sectional area ratio is (12 / 1.4)² ≈ 73; the arithmetic is this brief's, the two diameters are sourced. Even the 4 m pipe an intergovernmental source specifies for a modest 10 MW floating plant is eight times the cross-section of anything ever placed in deep water. And the pattern in the deployment record is sharper still: every pipe ever laid deeper than 600 m is 1.4 m or smaller, while the two 1.6 m-class pipes in service sit at 76 to 115 m in a lake.

Established The plants that actually run use pipes an order of magnitude smaller than the record. Kumejima draws 8 to 9 °C water from 612 m through two pipes of 28 cm diameter, more than 2 km long, 2.3 km offshore, moving 13,000 t per day. For the proposed 1 MW expansion the operator projects 180,000 t per day, and states that the large-scale intake pipe “would account for the majority of the equipment cost.”

Established The last serious commercial-scale attempt died on exactly this, and it is the most informative failure of the modern era. NEMO at Bellefontaine, Martinique — 16 MW gross, 10.7 MW net, a parasitic share of 33% by design — was suspended in April 2018 when the Martinique Assembly voted against it on 3–4 April. The stated objections were a cost above €450 million, low output per euro, and an on-board inventory of about 300 t of ammonia. The developer, Akuo Energy, cited “technical difficulties related to the main cold water intake pipe.” That is roughly €42,000 per net kilowatt, of order $45,000/kW; the arithmetic is this brief's on the developer's own figures. No large deep cold water pipe has been attempted anywhere since. The intergovernmental reassurance that a number of pipes were lost during deployment “although not during the last decade” is therefore true and close to vacuous.

Established The thermodynamics is brutal, and — this is the correction that most changes the shape of the argument — it is not the problem. Carnot efficiency for these temperature pairs is 6.71% at 25/5 °C, 7.35% at 26/4 °C and 7.97% at 28/4 °C; that arithmetic is this brief's and is independently confirmed at 7.96% for the 28/4 pair by a peer-reviewed source. But Carnot is the wrong ceiling, because the seawater streams are finite: they cool and warm as heat is exchanged, so the working fluid never sees the full ΔT. A finite-time thermodynamic analysis gives OTEC “a very low maximal theoretical thermal efficiency of 3%–5%”; the Curzon–Ahlborn endoreversible expression η = 1 − √(T_c/T_h) gives 4.07% at 28/4 and 3.41% at 25/5. Achieved net efficiencies are 2.5 to 3.0% on both peer-reviewed and government figures. OTEC cycles therefore run at 60 to 85% of their genuine thermodynamic limit. This is not an inefficient-cycle problem, and no heat-exchanger programme has a factor of two hidden in it. The problem is that 3% of a 24 K difference means moving enormous volumes of seawater per kilowatt-hour.

Frontier Which brings the argument to the single most important unresolved number in the subject: the parasitic gap. Every plant whose outturn was actually measured consumed between 52 and 80% of gross output in its own pumps. Every commercial-scale design assumes 28 to 40%. The 1 MW Japanese onshore reference case is 1,660 kW gross for 1,000 kW net, 39.8% parasitic. A major aerospace contractor's design is 69.35 MW gross for 49.8 MW net with 19.55 MW of pumping, 28.2%. The 100 MW-net peer-reviewed design is about 143 MW gross, roughly 30%. Every favourable levelised cost figure in the literature rests on the assumption that scale-up roughly halves the parasitic share, and nothing ever built has demonstrated it. The assumption is physically reasonable — pipe friction scales more favourably than flow with diameter, and large turbomachinery is more efficient — but it remains an assumption, flagged frontier here, and it is the hinge on which the entire commercial case turns.

Established And the flagship customer has written the technology out of its own energy plan. Global OTEC's Dominique project — 1.5 MW floating, expandable to 10 MW — has been “in development” in São Tomé & Príncipe since a 2022 memorandum, with its geotechnical, metocean and techno-economic studies explicitly aimed at a bankable proposal for that government's submission to the Green Climate Fund. As of August 2026 the Green Climate Fund's own São Tomé country page lists no OTEC project approved or in the pipeline, with $29.3 million of total financing across two World Bank projects and a food-system project in preparation. And the World Bank / Mission 300 National Energy Compact for the country, published in 2026, does not mention OTEC at all — “ocean energy” appears once, in a list of things promoters are presenting. The costed 2030 plan is 30 MWp of solar against 2 MWp today, 9 MW of hydro against 1.8 MW, 10 MW of bi-fuel thermal by 2028 and about 10 to 12 MWh of batteries, against roughly 38 MW of installed diesel supplying 95% of generation. Four years after the memorandum, a 1.5 MW OTEC platform is absent from the host country's own costed plan.

Three questions in this subject are genuinely open. Several more are routinely presented as open and are not, and separating the two is most of what a reader needs.

Frontier The first genuine frontier is whether the parasitic share falls with scale. This is the load-bearing uncertainty in OTEC and it is not a marginal one: measured outturn is 52 to 80% of gross, commercial designs assume 28 to 40%, and every levelised cost below $0.40/kWh in the literature lives inside that unbridged gap. The mechanism by which it might close is understood — frictional loss in a duct scales with the square of velocity while flow scales with the square of diameter, so a larger pipe moving the same volume per megawatt should cost less pumping power — but the largest plant ever operated was 255 kW gross, so the scaling law has never been tested more than about two orders of magnitude below where it must hold. No source consulted for this brief reports a measured parasitic share at any scale above a quarter of a megawatt.

Frontier The second is whether the cold water pipe can be built at all at commercial scale, and it is a marine civil engineering question rather than a physics one. The requirement schedule from the intergovernmental and peer-reviewed sources runs: about 2.5 m diameter in polyethylene for a 2.5 MW land-based plant; about 4 m for a 10 MW floating plant; eight pipes of 3.26 m outer diameter and 1,000 m length for a 50 MW plantship, costed at about $37.9 million in factory cost; and 12 m diameter for 100 MW net. Against a demonstrated maximum of 1.4 m at 915 m depth. Nobody knows whether a 4 m pipe can be deployed and survive, because nobody has tried in twenty-five years, and the last organisation that seriously proposed to try cancelled and cited the pipe. The intergovernmental programme's own framing is that the pipe is “a technological frontier” beyond 10 MW.

Frontier The third is the scale trap, and it is an economic frontier with no proposed solution. The peer-reviewed and intergovernmental cost work agrees that OTEC needs roughly 50 MW or more before its levelised cost gets under about $0.30/kWh. Its addressable customers need 1 to 10 MW. São Tomé & Príncipe has about 38 MW of installed generation. South Tarawa in Kiribati has 5.45 MW installed and a 3 to 4 MW peak. A single 50 MW OTEC plant — the smallest size at which the technology's own advocates project a tolerable cost — would be larger than the entire national grid of most of the customers it is designed for. No source read for this brief resolves this. It is not a technology-readiness problem and it does not get better with engineering; it is a structural mismatch between the size at which the machine is cheap and the size at which the market exists.

Frontier A fourth question is open in a different sense: the environmental effect of large-scale deployment is modelled, contested in framing, and unmeasured. The only large-scale peer-reviewed environmental modelling runs scenarios of 3, 5, 7, 10 and 15 TW ramped from 2030 to 2100 in a coupled climate model, with deep water drawn from about 1,100 m and mixed effluent discharged at about 20 m, at 314 m³/s per plant. The results are not small: global sea-surface cooling of 0.8 °C at 3 TW rising to 3.1 °C at 15 TW by 2500 against control, with maximum cooling of 3.3 °C in the West Pacific warm pool; subsurface warming centred at 18 °S between 500 and 1,000 m; surface phosphate rising sharply and staying elevated while nitrate is drawn down in the nitrogen-limited tropics; net primary production up by 0.06 to 0.14 nmol m⁻³ s⁻¹ of nitrogen globally with regional maxima of 1.7 to 3.0; an Atlantic overturning decline by 2500 reduced from 11.5 Sv in control to 3.3 Sv at 15 TW, with a transient peak of 22.6 Sv against a historical 17 to 18; surface pH rising from 7.4 to between 7.5 and 7.9; and polar warming of 0.7 to 1.0 °C above control persisting after OTEC ceases.

Frontier And the framing conflict inside that result is genuine and should not be smoothed. The United States government's environmental fact sheet warns that discharging cold, nutrient-rich water into warmer shallow layers risks “indirect and cumulative impacts to marine biota and the dynamics of the marine ecosystem.” The climate modelling deliberately does exactly that at planetary scale and counts the resulting mixing as a benefit: atmospheric carbon dioxide in 2500 of 623 ppm at 15 TW against 1,930 ppm with no OTEC, cumulative carbon reduction of 323 to 981 Pg C, and 60% of the year-2100 cooling coming from OTEC-induced ocean mixing rather than from emissions avoided. The same physical act is a hazard in the regulatory literature and a feature in the climate-modelling literature. This brief reports both and resolves neither, and notes that the modelling authors' own resource paper concedes that “caution is required, and the engineering challenges would be large.”

Handwave What is not open: whether better heat exchangers or better cycles make OTEC viable. Kalina and Uehara cycles simulate at 5.0 to 5.3% against a real ceiling of 3 to 5% and achieved 2.5 to 3.0%. Heat-exchanger selection is worth a great deal in relative terms — one analysis puts it at 158% in net power — and none of it changes the order of magnitude of water moved per kilowatt-hour. Framing OTEC's difficulty as cycle inefficiency invites the inference that engineering unlocks a factor of two. It does not.

Handwave Also not open: whether the resource is large enough. It is, and the section on the resource ceiling gives the numbers. Nobody has ever produced 0.0001 TW of net OTEC power against a sustainable global ceiling of 5 to 10 TW. The ceiling has never been the constraint and treating it as the interesting question is a category error.

Established The first bottleneck is the cold water pipe, and it binds because it is simultaneously the largest technical unknown and the largest or the smallest cost item depending on whom you ask. Eight to seventy-three times the demonstrated cross-section, in open ocean, with storms, currents, biofouling and a deployment operation that has failed in four of the historical attempts this brief can document: Cuba in 1930, La Tunisie in 1935, the OTEC-1 redeployment in 1983, and NEMO's cancellation in 2018 on the developer's own account of pipe difficulties.

Established The pipe's share of capital cost is contested by an order of magnitude, and the disagreement is unresolved. The intergovernmental economics review puts the cold water pipe at about 5% of total capital cost for a 50 MW plantship — $43.5 million of $886.9 million. A 2025 peer-reviewed Applied Energy study puts heat exchangers and cold water pipes together at 55.4% of total capital expenditure. The Japanese operator at Kumejima says the intake pipe would be “the majority of the equipment cost” at 1 MW onshore scale. These cannot all be right, and resolving them would require cost models that are not public. This brief reports the spread and takes no side.

Established The second bottleneck is the parasitic pumping load, and it is the reason the first one matters so much. Because 52 to 80% of gross output has historically gone into the pumps, a plant must be built roughly two to five times larger than its delivered output — more heat exchangers, more pipe, more platform, for the same saleable kilowatt. Every commercial design assumes this improves to 28 to 40% at scale, and if it does not, the capital cost per net kilowatt in every published estimate is understated by a factor approaching two.

Established The third is cost, and the spread in the published literature is a factor of thirty-four. This is not a range of uncertainty around a central estimate; it is a set of numbers for plants of different sizes, none of which exists.

Established Every low number in that table is for a plant of 50 MW or more, which requires a pipe nobody has built. The advocates' own conclusion is the honest one: first-generation levelised cost lacks competitiveness “without environmental credits or subsidies.”

Established The fourth bottleneck is what it must compete against, and here the comparison has moved decisively against OTEC in the last decade. The 2026 intergovernmental renewables cost report gives global weighted-average levelised costs of $44/MWh for solar photovoltaics, $33/MWh for onshore wind, $78/MWh for offshore wind, $89/MWh for geothermal and $115/MWh for concentrated solar. Wave energy sits at about $120/MWh and tidal stream at about $140/MWh — at 2 GW of cumulative deployment. The same report has no OTEC line item at all: not a cost, not even a projection. Wave and tidal are in; ocean thermal is not. Even the most favourable peer-reviewed OTEC figure, $120/MWh at the best Caribbean sites, is nearly three times unsubsidised solar, for a technology with no installed base against which to price risk.

Frontier The fifth is the scale trap, restated here as a financing bottleneck because that is what it becomes in practice. A 50 MW plant exceeds the entire grid of most target markets; a 1 to 10 MW plant sits in the cost range where diesel wins on capital even while losing on fuel. There is no size at which an island utility can both use the output and afford the plant, and no source consulted proposes one.

Established And a sixth, which is smaller but revealing: the sector has no functioning information infrastructure. The field's own news aggregator, otecnews.org, is a suspended hosting account. One developer's announcement of its “first commercial OTEC” is hosted there and is therefore unreadable. When a technology's trade press goes offline and nobody notices, that is a measurement of the sector's activity level.

Established OTEC depends on no unresolved physics and on no other brief in this corpus. The cycle is conventional Rankine engineering, the resource is measured, the materials are commodity polyethylene, titanium and fibre-reinforced plastic. What it waits on is marine civil engineering at a scale nobody has attempted and a financing route for a first-of-a-kind offshore structure whose customer is usually a small island state. The isolation is close to total: the corpus collision check found this slot touching nothing across a 135-brief map, and no typed dependency edge is recorded.

Established It depends, in an underappreciated way, on the cooling business. The deep cold-water pipe is a shared asset, and every economically successful deep or cold water intake in the world sells cooling — Toronto, Cornell, Bora Bora, Bahrain — or research-park seawater, as at the Hawaiian facility. None sells electricity. The most OTEC-favourable peer-reviewed cost study reaches $0.16/kWh only by bundling cooling revenue, and quantifies the bundle: seawater air conditioning raises capacity-factor-weighted revenues by 25 to 40% against standalone OTEC, and the authors conclude it “serves as a critical financial justification for onshore infrastructure investment.” If OTEC has a commercial future it probably arrives attached to a cooling asset rather than as a power station, which makes district cooling a dependency rather than a by-product.

Frontier It depends on development finance behaving in a way it has not yet behaved. The addressable market is small island states with acute energy costs, thin balance sheets and limited capacity to evaluate offshore engineering proposals. That makes concessional climate finance the only plausible capital source for a first plant, and the one project that pursued that route for four years does not appear in the fund's pipeline.

Established What depends on OTEC is, at present, nothing. No brief in this corpus, no national energy plan examined for this brief, and no industrial process. That is the accurate statement of a technology's position when its all-time output record is a hundred kilowatts, and it is stated here plainly rather than softened into potential.

Established The decisive experiments in this field have mostly already been run, and the sequence is the result. Twenty-two kilowatts gross in Cuba in 1930. Zero off Rio in 1935. Fifty kilowatts gross for ten to eighteen net in Hawaii in 1979. A heat-exchanger test loop with no turbine in 1981. A hundred and twenty kilowatts gross for about thirty net at Nauru in 1981. Two hundred and ten to two hundred and fifty-five gross for a hundred to a hundred and three net in Hawaii from 1993 to 1998. A hundred gross with net unpublished at Kumejima from 2013. A hundred and five gross with net unpublished in Hawaii from 2015. Three hundred and thirty-eight gross at off-design temperature difference on a Korean barge in 2019 that never reached its intended site. Ninety-six years of experiments, and the highest net output ever measured was achieved by a plant decommissioned in 1998.

Established The most recent object in the water is not a power plant, and its developer says so. The PLOTEC platform “Don” at the Canary Islands ocean platform, lifted in November 2024, deployed in October 2025 and with its riser connected and installation completed on 22–23 April 2026, is described on the developer's own project page as “a structural demonstration rather than a full power-generation system… designed to validate that an OTEC platform can survive a ‘100-year storm’ based on Caribbean data.” A trade report from October 2025 recorded the riser pipe as “not yet connected” and the platform undergoing structural testing with instrumentation. The experiment being run is a hull survivability test in a location that is not a tropical-cyclone environment, against storm criteria taken from one that is. That is a legitimate and useful experiment. It is not a generation experiment, and no power figure — gross or net — attaches to it.

Established The clearest negative result in the whole subject comes from the adjacent business that should have been easy. Honolulu Seawater Air Conditioning proposed a 4.7-mile pipeline drawing about 7 °C water from beyond 520 m (1,700 ft) to cool downtown Honolulu, cutting 77 GWh a year — the consumption of roughly 13,000 homes — and up to 75% of customers' air-conditioning cost. All major regulatory hurdles were cleared by June 2019. The cost estimate rose from $275 million to $400 million, the project was cancelled in December 2020 after $25 million spent over sixteen years, and operations formally ceased on 31 January 2021. A board member of the backing investor said it would not have been prudent to pursue it further. Cooling has a firm, high-value, round-the-clock product and no thermodynamic penalty whatever, and it could not finance a 520 m pipe. The case for a 1,000 m pipe financed by 3%-efficient electricity requires an explanation nobody has supplied.

Frontier The experiment that would actually settle the central question has a clean specification and nobody is running it. Instrument an existing plant — Kumejima or Kailua-Kona, both of which have run for over a decade — and publish the net output and the parasitic breakdown at a stated temperature difference over a stated period. That is not a research programme; it is a metering exercise on hardware that already exists. Its absence after eleven and thirteen years respectively is the most striking omission in the field's evidence base, and it is why this brief treats every unqualified capacity claim in OTEC as gross until shown otherwise.

Frontier Two scheduled experiments are worth watching for what they would actually establish. The Japanese 1 MW programme at Kumejima has funded fabrication of large parallel-type titanium heat exchangers equivalent to 200 kW under a Ministry of the Environment carbon-neutrality demonstration; if a 1 MW plant is built and publishes a net figure, it would be the first megawatt-scale OTEC result in history. Global OTEC's stated next step is an onshore demonstration of “up to 500 kW” in Hawai‘i, which if built would be the largest onshore OTEC plant ever attempted, against Makai's 105 kW — and for which no site, date, cost or financing has been published.

Established A negative result recorded as a result. No OTEC power plant of any size is under construction anywhere, as far as this brief can verify. What is physically in the water is one non-generating storm-test hull off Gran Canaria; what is being fabricated is titanium heat exchangers in Japan. The one United States company listed on a public market as an OTEC developer filed an annual report for its 2025 financial year in which its auditors state “substantial doubt about our ability to continue as a going concern,” and which states that it has not completed any projects.

Established The cycle is conventional and the flow rates are not. Closed-cycle plants boil ammonia against warm seawater and condense it against cold, around an ordinary Rankine loop; open-cycle plants flash-evaporate the seawater itself at low pressure, drive a very large low-pressure turbine, and condense the vapour to fresh water. What makes OTEC unusual is not the machinery but the ratio of water to power. The intergovernmental figure is 2.5 to 3 m³/s of seawater per megawatt. The detailed 100 MW-net design specifies 235 m³/s of cold water and 470 m³/s of warm, 705 m³/s in total, about 7 m³/s per net megawatt; the United States government fact sheet gives 10 to 20 billion US gallons per day for a 100 MW plant, roughly 438 to 876 m³/s. For a Canadian reader the calibration is unflattering: 705 m³/s is of the order of the mean discharge of the Ottawa River, pushed continuously through a power plant.

Established Where the parasitic load actually comes from — and the usual explanation is wrong. “Pumping cold water up a kilometre” is a misleading picture. The intake is not a lift against 1,000 m of head; it is a duct in which the pump fights friction and the density difference between the cold column and the ambient water, not gravity. Claude measured 3 m of head loss on a 2,000 m pipe of 1.6 m diameter at 0.2 m³/s in Cuba in 1930 — a figure this brief reports with a caution, since its only source is an enthusiast history of low standing. Consistent with it, the warm-water pump is a large share of the load too: a peer-reviewed analysis attributes about 70% of closed-cycle parasitic consumption to the two seawater pumps together and about 30% to the working-fluid pump. The finite-time analysis states the magnitude directly: pumping power to overcome pressure drop “is of the same order of magnitude as the gross power generated.” The killer is flow volume multiplied by friction and heat-exchanger pressure drop, not vertical lift. Keep the conclusion; fix the mechanism.

Established Every OTEC plant ever built, with outturn. The record is short enough to enumerate, so this brief enumerates it. Gross and net are stated separately; where net was never published, the table says so rather than repeating the gross figure.

Established The cold-water-pipe deployment record, from the one firm that has laid most of them. The pattern is the finding: the two large-diameter pipes are shallow, and everything deep is small.

Established Fabrication and materials, since the pipe is the plant. The 2001 Hawaiian pipe was fused into nine sections of about 300 m onshore at Kawaihae Harbour from high-density polyethylene supplied by KWH Pipe of Mississauga, Ontario, and installed by a marine contractor — the pipe that holds the world record for deep cold-water intake was made in Canada. Commercial-scale designs abandon polyethylene: the 100 MW design specifies fibre-reinforced plastic with a syntactic foam core at a 2 m/s inlet velocity, and the large-composite design literature that would settle whether such a pipe is manufacturable sits in a United States Department of Energy technical report on advanced composite cold water pipes that this brief could not retrieve.

Frontier What commercial scale actually requires, in one place. About 2.5 m diameter in polyethylene for 2.5 MW land-based; about 4 m for 10 MW floating; eight pipes of 3.26 m outer diameter by 1,000 m for a 50 MW plantship at about $37.9 million factory cost; 12 m by 800 m for 100 MW net; and the planetary-scale climate modelling assumes 10 m at 4 m/s and 314 m³/s per plant. Against 1.4 m demonstrated.

The closest working relative is not another power technology but seawater district cooling, and the relationship is asymmetric in a way that matters. Both need a cold-water pipe; only one has ever paid for one. Toronto's deep lake water cooling system serves more than a hundred buildings after an expansion commissioned between August 2024 and summer 2025, avoids over 60 MW of peak electrical demand, cuts cooling electricity by 75%, moves 704 ML a day, and drew a $600 million Canada Infrastructure Bank loan against a C$230 million expansion. Its intake is at 83 metres, in a lake. Cornell's is at 76 metres and cut air-conditioning electricity by 87%. Both are financed by mainstream infrastructure lenders. The deep-ocean systems that operate are resort-scale: the largest is 450 tons of cooling from a 900 m pipeline in French Polynesia, about one hundred and thirtieth of Toronto.

Ocean engineering is adjacent in the strong sense: OTEC's binding constraint is a marine structures problem, and if the field advances it will be because deep-water pipe deployment advanced for some other reason. Geothermal megaprojects is the closest analogue in kind — low-grade heat, large capital, a long history of being nearly ready — and the useful comparison is that geothermal has a real installed base at $89/MWh while OTEC has none at any price.

Planetary-scale energy systems is where the resource ceiling and the climate modelling belong, and it is also where OTEC becomes something other than a power technology: the modelling that treats 3 to 15 TW of OTEC as a climate intervention with 60% of its cooling coming from induced ocean mixing is a geoengineering result wearing an energy-technology label. Future ports and shipping is adjacent for offshore deployment practice. Energy storage is adjacent by contrast rather than by kinship: OTEC's one genuine advantage over solar in its own latitudes is that it is continuous, with capacity factors above 85% at good Caribbean sites, and cheap storage erodes exactly that advantage.

Established The defining institutional fact about OTEC is that it has no customer, and has never had one. Every plant in the record was built by a government, a national laboratory, a university, a prefecture or a research consortium, ran, and was decommissioned without a commercial successor. Not one was procured by a utility buying power. The intergovernmental economics review's own conclusion is that first-generation levelised cost lacks competitiveness “without environmental credits or subsidies,” which is an accurate description of a technology that has been demonstration-funded for a hundred years.

Established The clearest institutional signal available is an omission. The 2026 intergovernmental renewables cost report — the standard reference for what electricity costs worldwide — carries line items for solar, onshore and offshore wind, concentrated solar, hydro, geothermal, bioenergy, wave and tidal stream. Wave and tidal are included at roughly $120 and $140/MWh on 2 GW of cumulative deployment. OTEC has no line item at all: not a cost, not a projection, not a footnote. The institution whose job is to price every renewable technology does not price this one, because there is nothing to price.

Established The development-finance channel, which is the only plausible route to a first plant, has not opened. A 1.5 MW project in São Tomé & Príncipe has been in development since 2022 with studies aimed explicitly at a Green Climate Fund submission, and produced a certificate of approval for riser installation methodology and an approval in principle from a classification society in 2023. As of August 2026 the fund lists no OTEC project for that country, approved or in pipeline; its $29.3 million of financing there goes to two World Bank projects and a food-system project. The World Bank's own national energy compact for the country, published in 2026, does not mention OTEC. Approvals in principle from classification societies are engineering documents, not financing; four years of them have not produced a funded project.

Established The corporate layer is thinner than the press coverage suggests, and this is checkable. The United Kingdom developer has about eleven named people, disclosed funding of over £250,000 from 2018 plus its share of a €3.5 million consortium grant and an undisclosed pre-seed round, and by its own account no operational plants. The United States listed company's auditors record substantial doubt about its ability to continue as a going concern; it states it has not completed any projects; its non-affiliate market value was about $181,247 in March 2026 against roughly 190 million shares outstanding; it carries defaulted debt including a confessed judgment of about $2.5 million under forbearance; and its only live contract is a $3.6 million United States Army design study with no power-purchase assurance. A French supplier's OTEC pages name no project, capacity, location or date. The Japanese programme is the most substantial and is a shipping company with an undecided business model fabricating 200 kW-equivalent heat exchangers under an environment-ministry demonstration grant.

Established Where the institutional capability plainly does exist, it is in cooling. A Canadian district-energy operator commissioned a 60% expansion of its deep lake water cooling system between August 2024 and summer 2025, adding forty buildings to reach more than a hundred, financed by a C$230 million programme and a $600 million Canada Infrastructure Bank loan. That is what a bankable cold-water asset looks like: an infrastructure bank, a municipal water partner, a committed customer base, and an intake at 83 metres. The contrast with OTEC is not about technology readiness. It is that one product has customers who sign contracts before construction, and the other has research programmes.

Frontier The institution that does not exist is an OTEC buyer of any kind, and the open question is whether one can. The scale trap is institutional as much as economic: the plant is cheap at 50 MW and the customer's whole grid is 5 to 38 MW. That mismatch cannot be closed by a utility, because no utility in the addressable market is that size; it would require either a regional interconnection that does not exist, an industrial off-taker in a small island state, or a bundled cooling-and-power business model that no developer has yet published. Which of those, if any, appears is the institutional question this brief is actually about — and the field's own news aggregator being a suspended hosting account is a fair measure of how many people are currently working on it.

Established The evidence base in this field is dominated by interested parties, and the pattern of what they publish is itself the ethical finding. The most complete plant lists come from an intergovernmental programme authored largely by OTEC proponents; the most cited economics comes from a career advocate for the technology; the pipe deployment record comes from the firm that laid the pipes; the current project claims come from developers' own websites. These sources are individually reliable on facts they would be embarrassed to get wrong — dates, diameters, depths, delivered tonnages — and unreliable on framing. The specific asymmetry is that gross figures are published and net figures are not. Two plants have run for eleven and thirteen years respectively without publishing a net output. That is not a measurement difficulty; both are instrumented research facilities. It is a choice about what to disclose, and it is the reason this brief marks gross and net on every number it states.

Established Public money has carried this field for a century and the accounting is thin. Cumulative public spending in Hawaii alone to 1991 was at least $122 million — over $90 million federal and over $20 million state capital — established by independent investigative journalism rather than by any programme report. OTEC-1 alone cost $45 million to refit and about $50 million to operate, ran for four months, never generated electricity, and its hull was scrapped for $587,000. A 1983 attempt to recover and redeploy its pipe cost $754,000 and failed twice. Present-day funding is public too: Horizon Europe and a United Kingdom research council for the Gran Canaria platform, a Japanese environment ministry programme for the heat exchangers, a United States Army design contract for a company with substantial doubt about its going-concern status. None of this is improper. It does mean that a technology whose entire output record is about 103 kW net has been financed almost exclusively by taxpayers for a hundred years, with the disclosure practices of a private consortium.

Frontier The distinctive ethical question concerns small island states, and it is not abstract. The addressable market is jurisdictions with acute energy costs, small grids, limited technical capacity to evaluate offshore engineering proposals, and strong incentives to appear attractive to climate finance. A technology with a 145-year record of hundred-kilowatt demonstrations being advanced toward climate funds in those jurisdictions warrants asking who bears the cost of a failure. The most defensible position available is the one the host governments' own planning documents take: proven solar, hydro, thermal and batteries first, at known cost. That is not hostility to OTEC. It is a revealed preference by the customer the technology exists to serve, expressed in a costed national plan that does not mention it.

Established There is also a genuine knowledge gap with an ethical edge. There is no modern, quantitative, peer-reviewed environmental assessment of an operating OTEC plant. The regulatory baseline rests on studies from the 1980s. A 2010 conference position paper argued explicitly that those assessments are inadequate — they missed trace-element controls on primary production, inorganic carbon chemistry, and the observational and modelling tools since developed — and offered no quantitative results of its own. Sixteen years later that critique still stands unanswered by measurement, while the modelling literature has moved on to 15 TW planetary scenarios. Proposing large plants in tropical waters against an evidence base that its own critics called obsolete in 2010 is a defensible engineering decision and an uncomfortable environmental one, and the discomfort should be stated rather than managed.

What this brief could not establish, stated as an obligation rather than an embarrassment. Whether Claude's 1930 plant was net-positive. The net output of the Hawaiian 105 kW plant or the Okinawan 100 kW plant, or whether the former ran at all between 2024 and 2026. Whether the Korean barge ever reached Kiribati. What happened to India's 1 MW Tuticorin plant in 2000. The European grant amount for NEMO, widely cited at €72 million. Global OTEC's total capital raised, or any site, cost, timeline or financing for its claimed Hawaiian 500 kW demonstration. Whether the Curaçao deep-water district cooling scheme was ever built. And the cold-water-pipe cost share, which three credible sources put at 5%, at 55.4% jointly with heat exchangers, and at “the majority of the equipment cost” respectively. Several primary sources that would settle these — the 1981 and 1984 papers on Mini-OTEC and Nauru heat-exchanger performance, the 1983 offshore-technology paper on the OTEC-1 pipe salvage, and a United States Department of Energy final technical report on advanced composite cold water pipes — are behind paywalls or robots restrictions this brief could not pass, and are named here unverified rather than cited.

Frontier The civilisational case for OTEC is genuine and unusually specific, which is why it survives repeated failure. It is a continuous, non-intermittent, equatorial renewable resource with capacity factors above 85%, delivering fresh water and cooling as real by-products, in exactly the latitudes where electricity is most expensive and least reliably supplied. Nothing else offers that combination. The sustainable global ceiling of 5 to 10 TW is comparable to total present-day human primary energy consumption. If it worked, it would matter enormously.

Established Against that, the record is the finding, and the record is 145 years long and a hundred kilowatts wide. This is one of the cleanest available instances of a general pattern the corpus keeps encountering: a technology can be physically sound, thermodynamically near its limit, resource-rich and continuously funded, and still not exist, because the binding constraint is a civil-engineering artefact nobody will pay to build first. OTEC's constraint is a pipe. The pipe has not grown since 2001, and the reason is not that anyone tried and failed in that time — it is that nobody tried.

Frontier The reframing that has the most claim on a reader's attention is that the deep-cold-water business which demonstrably works is cooling, not electricity. Toronto avoids more than 60 MW of peak demand from an 83 m intake and attracted a $600 million public infrastructure loan; the deepest cooling systems that operate serve a resort. If ocean thermal has a civilisational future it likely arrives as a by-product of cooling and desalination infrastructure that someone builds for other reasons, rather than as a power station built for its own sake.

Frontier And there is a larger version of the question that is not an energy question at all. The only large-scale peer-reviewed modelling of OTEC finds that at 3 to 15 TW it would cool global sea-surface temperature by 0.8 to 3.1 °C by 2500, shift the Atlantic overturning circulation, raise surface pH, redistribute nutrients across the tropics, and leave polar warming 0.7 to 1.0 °C above control after operations cease — with 60% of the year-2100 cooling coming from induced ocean mixing rather than from emissions avoided. That is a description of deliberate climate intervention. Whether a technology with those effects should be evaluated as a power source or as geoengineering is a question the literature has not asked, and it is the most consequential unexamined issue in the subject.

Established What already happened, in one sequence, because the shape of it is the argument. 1881: d'Arsonval proposes the idea. 1930: Claude runs 22 kW gross for several days at Matanzas Bay after three pipe-deployment attempts, and a storm destroys the plant. 1935: Claude's 2.2 MW shipboard plant off Rio fails during pipe assembly and never produces net power. 1956: a 3 MW French state design for Abidjan is killed by cheap oil before construction. 1979: Mini-OTEC produces 10 to 18 kW net from about 50 kW gross. 1981: OTEC-1 tests heat exchangers with no turbine and is cancelled in April after four months; Nauru delivers the first OTEC power to a real grid at about 30 kW net from 120 kW gross. 1993–98: Keahole Point sets the net record at 100 to 103 kW and closes. 2001: the largest deep cold-water pipe in history, 1.40 m to 915 m, is laid in Hawaii for a research park. 2013: Kumejima begins operating at 100 kW gross. 2015: Makai's plant becomes the first grid-connected closed-cycle OTEC in the United States, at 105 kW gross. 2018: NEMO is cancelled. 2019: a Korean barge reaches 338 kW gross at 18.7 °C. 2026: a non-generating structural hull is installed off Gran Canaria.

Established 2026: what is actually scheduled. The PLOTEC project's completion is scheduled for mid-2026 and remains a structural validation with no generating cycle. Nothing else is under construction.

Handwave The 1 MW Japanese plant, “around 2027.” The operator's 2023 announcement targeted “the world's first commercialization of OTEC on a 1 MW scale around 2026.” Its own later publication says “around 2027.” The same publication states that the company “has not yet definitively decided on the specific direction it will take or the business model it will construct.” What is funded is 200 kW-equivalent heat-exchanger fabrication; no 1 MW plant is under construction. A target that has already slipped once and is attached to a business model that does not exist is recorded here as an intention.

Handwave Global OTEC's 500 kW onshore demonstration in Hawai‘i, and the 1.5 MW Dominique platform. Neither has a published site, date, cost or financing route. Dominique has been “in development” since 2022 and is absent from both the climate fund it was aimed at and the host country's national energy compact. No date is stated here because none has been published.

Handwave The KRISO barge's relocation to South Tarawa, Kiribati, planned for 2020. Six years overdue with no evidence either way. This brief records it as unresolved rather than as cancelled.

Frontier The one milestone worth watching for, with no date attached. A published net output above 103 kW from any plant anywhere. That number has not moved in about thirty years, it requires no new construction to establish for two existing plants, and it is the cleanest single test of whether this field has advanced.

Speculative Beyond 2040: 50 MW and 100 MW plants. These require a 3.26 m-class or 12 m pipe, a demonstrated parasitic share near 30%, and a customer whose grid can absorb the output. All three are individually plausible on decadal timescales; their conjunction is speculative, and no source consulted proposes a development path from the present state to any of them.

• Depends on Nothing on this map, and the isolation is nearly total — the collision check found this slot touching nothing in a 135-brief corpus. OTEC waits on no physics result. Every property it needs has been measured, and the thermodynamic ceiling of 3 to 5% is not a pending discovery but a settled consequence of finite heat sources. What it waits on is recorded below, and none of it is knowledge.
• Requires (not on this map) A cold water pipe of 4 to 12 m diameter deployed and survived in open ocean, against a demonstrated maximum of 1.40 m at 915 m — eight to seventy-three times the cross-section. A measured parasitic share near 30% of gross at any scale above a quarter of a megawatt, since every favourable cost estimate assumes one and no plant has ever shown one. A financing route for a first-of-a-kind offshore structure costing of order $45,000 per net kilowatt. Concessional climate finance willing to underwrite it in a small island state, which the one attempt to obtain it has not produced in four years. And a cooling revenue anchor, because bundled cooling is the only thing that brings any peer-reviewed cost estimate near competitiveness. All five are engineering, financial or institutional capabilities rather than discoveries.
• Enables Continuous, non-intermittent equatorial baseload with fresh water and cooling as genuine by-products, in the latitudes where electricity is most expensive. No typed enabling edge is claimed for a capability whose demonstrated maximum is about 103 kW net, set in 1998.
• Adjacent Ocean engineering, which owns the marine structural problem this brief is gated on; geothermal megaprojects, the closest analogue in kind, since both extract low-grade heat and both have been about to scale for fifty years; future ports and shipping for offshore deployment practice; and planetary-scale energy systems, where the 5 to 10 TW sustainable ceiling and the climate-model results actually belong.

“OTEC's problem is that the cycle is only 3% efficient.” Established Carnot for these temperature pairs is 6.7 to 8.0%, which invites the inference that a better engine could double the output. It could not. Because the seawater streams themselves change temperature during heat exchange, the operative ceiling is the finite-heat-source limit of 3 to 5%, and achieved net efficiency of 2.5 to 3.0% is therefore 60 to 85% of the real thermodynamic maximum. Advanced cycles simulate at 5.0 to 5.3%. OTEC's cycles are close to as good as they can get; the problem is that 3% of a 24 K difference means moving of order 7 m³ of seawater per second per net megawatt.

“The parasitic load is the killer because you have to pump cold water up a kilometre.” Established The conclusion is right and the mechanism is wrong, and the mechanism matters. The intake is a duct, not a hoist: the pump fights friction and the density difference between the cold column and ambient water, not gravity. Claude measured about 3 m of head loss on a 2,000 m pipe in 1930 — from a low-standing source, but consistent with the rest of the evidence. The warm-water pump is a comparable burden: the two seawater pumps together account for about 70% of closed-cycle parasitic consumption, the working-fluid pump for about 30%. The load is flow volume multiplied by friction and heat-exchanger pressure drop. A physics-literate reader who is told OTEC is defeated by a kilometre of lift will stop trusting the rest of the brief, and should.

“Plants are operating in Hawaii, Okinawa and Korea.” Established Their published figures are gross. Net output has never been published for the Hawaiian 105 kW plant in eleven years, nor for Okinawa's 100 kW plant in thirteen, including in the prefecture's own final report. The Korean 338 kW was gross, at an off-design 18.7 °C, on a barge whose planned relocation to Kiribati this brief cannot confirm ever happened. And the Hawaiian plant is recorded by the United States government database as not operating on a continuous basis — it is a heat-exchanger test bed with a turbine attached. The all-time net record of about 103 kW belongs to a plant decommissioned in 1998.

“The world's first purpose-built OTEC platform was installed in 2026 and generates electricity.” Handwave A €3.5 million storm-survivability hull with an instrumented riser was installed off Gran Canaria on 22–23 April 2026, with no generating cycle, outside the tropics. Its developer's own project page calls it “a structural demonstration rather than a full power-generation system.” The consortium's research-institute partner nevertheless announced that the prototype “is designed to deliver stable electricity by harnessing the ocean's thermal gradient,” and trade coverage described it as “the first offshore system designed to generate continuous electricity from ocean temperature differences at sea.” Neither statement is true of the object in the water. The developer's own description is the most conservative account available, and it is the citable one.

“Moving offshore cuts the pipe length by 80%.” Handwave A developer marketing claim. Moving the platform to deep water shortens the horizontal run; it does not reduce the depth the pipe must reach, which is what sets the thermal resource and most of the structural difficulty. The 1,000 m of vertical is the hard part and it is unchanged.

“The engineering is de-risked — no cold water pipes have been lost in a decade.” Established True, and close to vacuous. The careful intergovernmental wording is that a number of pipes were lost during deployment “although not during the last decade” — and no large deep cold water pipe has been attempted in the last decade. The last serious attempt, NEMO, was cancelled in April 2018 with its developer citing pipe difficulties and its assembly citing a €450 million price for 10.7 MW net.

“Georges Claude's second plant was at Abidjan.” Established It was not, and the two projects are routinely merged. Claude's second attempt was aboard the M.V. La Tunisie about 70 nautical miles south of Rio de Janeiro in 1935 — a 2.2 MW design of eight 275 kW turbines that failed during cold-water-pipe assembly and never produced net power. Abidjan was a separate, later French state project: 3 MW, designed in 1956, never built, killed by cheap petroleum. And on Cuba: 22 kW gross is documented, and the universally repeated claim that the plant consumed more than it produced is not documented in any source this brief could verify. It is probably true. It is not established, and this brief does not assert it.

“Nauru was a 100 kW plant.” Established One hundred kilowatts is the nameplate. The plant was 120 kW gross and about 30 kW net, with roughly 90 kW consumed by the plant itself — a 75% parasitic share, and the first OTEC power ever delivered to a real grid.

“OTEC is inefficient but the resource is so large it does not matter.” Established The resource is large and it is not unlimited, and four independent peer-reviewed treatments converge on the number. Extraction is cold-water upwelling; upwelling cools the surface, erodes the thermocline, and destroys the resource it feeds on. The sustainable global ceiling is 5 to 10 TW against a naive maximum of 30 to 35 TW, with about 6.5 TW accessible inside exclusive economic zones by the end of the century. That is genuinely enormous — comparable to total present-day global primary energy consumption. It is also irrelevant to the field's actual problem, because nobody has ever produced 0.0001 TW of net OTEC power.

“OTEC is zero-carbon.” Established Not quite, and the number is measured rather than assumed. Deep seawater carries about 65 µmol of carbon dioxide per kilogram, and the cold-water non-condensable gas stream runs up to 14% carbon dioxide. A peer-reviewed assessment puts outgassing at 38.5 g CO₂/kWh — 11.7 from the warm stream and 26.8 from the cold. That is about 8% of a gas turbine's roughly 450 g/kWh, so the conclusion is favourable; the point is that the figure is not zero and is rarely quoted. Note also a framing split this brief does not resolve: the intergovernmental programme calls open-cycle carbon dioxide “less than 1%” of an oil station and closed-cycle negligible, while the peer-reviewed authors treat the same gas stream as a harvestable carbon-capture feedstock at €15 to 35 per tonne. Neither treats it as an environmental debit.

“Seawater air conditioning proves the deep pipe works, so OTEC is next.” Established The evidence supports a sharper and less comfortable version: the pipe is the business, and cooling is the only product that has ever paid for one. But the systems that work at scale draw from 76 to 115 m in a lake or a stratified harbour, not from 900 m in open ocean. The deep-ocean systems that operate are resort-scale. And the one serious attempt at city-scale deep-ocean cooling — Honolulu, 520 m, sixteen years, $25 million spent, all permits in hand — was abandoned in December 2020 when the estimate went from $275 to $400 million. So it is not simply that cooling works and OTEC does not. It is that shallow cold water works, and the kilometre-deep pipe has not yet been paid for by any product tried.

“There is a commercial OTEC industry.” Established There are perhaps four organisations claiming commercial ambition. One has never generated electricity from OTEC and has installed a non-generating hull. One is a Japanese shipping company fabricating heat exchangers with no decided business model. One is a United States listed company whose auditors record substantial doubt about its ability to continue as a going concern, which states it has not completed any projects, whose non-affiliate market value was about $181,247 in March 2026, and whose only live contract is a $3.6 million Army design study. One publishes OTEC pages with no named project, no capacity, no location, no date and no deployment history. That is not an industry; it is a literature with a website.