Future Ports and Shipping

Established Shipping is roughly 3% of world CO2 emissions, and roughly 40% of the CO2 emitted by ships that visit ports is emitted during port activities. Those two figures split the subject into two problems that are usually run together and should not be: what a ship burns on a deep-sea voyage, and what a port supplies at the quay. They have different physics, different owners, different regulators and different answers.

Established The sea's decarbonisation is normally posed as a fuel-choice question — ammonia, methanol, hydrogen, LNG, biofuels — and is better posed as an electricity question wearing a fuel's clothes. Electrolysis takes 50–60 kWh per kilogram of hydrogen, and every candidate except LNG and crop-derived biofuel is downstream of that number. Established Green hydrogen is under 1 million tonnes a year against a world supply of about 100 million. The molecule is a way of moving electrons onto a ship.

Frontier The strongest measured answer to whether any of this pays is no. A 2026 total-cost-of-ownership study of a 15,000 TEU methanol dual-fuel container vessel on a named green corridor found 134.4% higher present-value cost than very-low-sulfur fuel oil at a carbon price of EUR 73.5 per tonne of CO2, and a Monte Carlo run of 5,000 iterations returned zero probability of a cumulative cost advantage at 15, 25 or 30 years. The authors say plainly that carbon pricing alone cannot close the differential. That is the spine of this brief.

Frontier On the quay, marketing and the independent literature disagree: a 20-port Mediterranean study concludes automation alone does not raise terminal efficiency, and a Ro–Ro simulation found automated tractors increasing unloading time. Speculative At the exotic end sit two proposals that keep returning because they dissolve constraints rather than easing them — nuclear merchant propulsion, built four times and abandoned, now reopened in the nuclear engineering literature, and uncrewed deep-sea shipping, whose obstacle is crewing law rather than autonomy.

Established One sourcing constraint governs this page. The research behind it could not read the primary regulatory instruments, so no IMO adoption date, article number or entry-into-force date appears anywhere here. Where regulatory targets appear they are attributed to a peer-reviewed secondary source that was read, and flagged as such.

Established Start with energy content, and state the heating-value basis, because the marine-fuels literature routinely does not. Methanol carries 20.0 MJ/kg, approximately 50% that of conventional marine gasoil, and 15,800 MJ per cubic metre — 15.8 MJ/L, the one row of the source table this brief would reprint as published, because it is internally consistent with its own gravimetric figure and density. Established Ammonia carries 19 MJ/kg, comparable to methanol and about 40% of conventional fuel. Established Hydrogen's gravimetric figure in the same table is 141,500 kJ/kg — its higher heating value set beside everyone else's lower heating value, a defect examined in section 14.

Frontier The volumetric penalty decides ship design, and it has to be computed rather than quoted. Taking marine gas oil at about 42.7 MJ/kg and 0.86 kg/L, so about 36.7 MJ/L: methanol at 15.8 MJ/L needs roughly 2.3 times the tank volume per unit of delivered energy; liquid ammonia at 682 kg per cubic metre and about 18.6 MJ/kg, about 12.7 MJ/L, needs roughly 2.9 times; liquid hydrogen at 70.8 kg per cubic metre and about 120 MJ/kg lower heating value, about 8.5 MJ/L, needs roughly 4.3 times. Frontier The hydrogen density is a fetched figure; the gas oil and ammonia heating values are not independently verified here, so read the ratios as working estimates. Established All three are fluid-volume ratios, excluding insulation, cofferdams, double-walling and hazardous-zone standoff. Frontier The installed-volume penalty is the operationally meaningful one, it is worse, and it is almost never published — itself a finding about the field.

Established Storage conditions separate the candidates more sharply than energy density does. Liquid hydrogen: minus 252.77 °C, 1.5–5 bar, 70.8 kg per cubic metre, daily boil-off 0.3–1.0%. Liquid ammonia: boiling point minus 33.3 °C, 1.5–2.5 bar at 15 °C, daily boil-off 0.04%. LNG: minus 161 °C, 421–470 kg per cubic metre, daily boil-off 0.10–0.15% and 0.15–0.30% on maritime vessels. Methanol: boiling point 64.5 °C, flash point 10.8–12.2 °C. Established The boil-off column is the decisive one for deep sea. Ammonia loses roughly an order of magnitude less cargo energy per day than liquid hydrogen, which over a forty-day voyage compounds into the fuel plan; more than any cost argument, that is why ammonia and not hydrogen is the deep-sea candidate. Established Methanol's whole commercial case is the fourth line: the only candidate needing neither a cryogenic nor a pressurised tank, and the only one whose bunkering resembles the industry's existing practice.

Established Production scale is where the transition stops being a shipping story. World hydrogen production is about 100 million tonnes annually and green hydrogen under 1 million — below 1% of supply, at 50–60 kWh per kilogram of electrolysis. Established World ammonia production is about 180 million tonnes annually, Haber-Bosch emitting 450–500 million tonnes of CO2 a year at about 2.7 tonnes per tonne of grey ammonia; methanol about 107 million tonnes at about 1.5 tonnes per tonne; LNG capacity above 400 million tonnes with over 700 vessels and more than 50 import terminals across more than 30 countries. Established The most arresting scale figure is liquefaction. World liquid-hydrogen liquefaction capacity is about 400 tonnes per day — roughly 146,000 tonnes a year — against a projected maritime hydrogen demand the same review puts at 50 million tonnes annually by 2050. Frontier That is a factor of about 340, arithmetic on the review's own numbers rather than a claim it makes.

Frontier The cost gap has now been measured properly, and it is the most important result in the topic. Bo, Cai and Zhang modelled a 15,000 TEU methanol dual-fuel container vessel on the Ningbo–Zhoushan–Valencia corridor at a carbon price of 73.5 EUR per tonne of CO2 and green methanol at USD 1,500 per tonne. The methanol case shows 134.4% higher present-value costs than VLSFO. A Monte Carlo across 5,000 iterations found zero probability of a cumulative cost advantage at 15, 25 and 30 years, and the authors state that carbon pricing alone cannot close the cost differential under the scenarios tested. Frontier This brief read the publisher-deposited abstract, not the body: the numbers are the authors' own summary claim. Frontier It is nevertheless a different kind of result from the comparative life-cycle assessments that dominate this literature, because it reports a probability rather than a point estimate, and zero is hard to argue with.

Frontier A second study points the same way with a smaller ship: for a 6,600 TEU vessel over five years, pure VLSFO operation raises total cost 69.90% against baseline and a 30% e-methanol blend raises it 178.15%. Frontier The same paper reports CO2-equivalent reductions of 12.18% and 24.72%, and those are not quotable — reduction against what is never stated, VLSFO appears as both baseline and case, and a 30% blend delivering 24.72% would need near-zero-intensity e-methanol and a blend fraction measured by energy rather than volume.

Frontier Even the most optimistic published green-fuel cost still loses on energy. A 2025 process-design study puts the optimal levelised cost of green ammonia at approximately 0.59 USD/kg by 2050 — a wind-driven system, minimal battery storage, 20% flexibility factor, 64 tonnes per day — competing with grey ammonia when carbon allowances reach USD 0.12 per kg of CO2, that is USD 120 per tonne. Frontier At an ammonia lower heating value near 18.6 MJ/kg that is about USD 31.7 per gigajoule, against VLSFO at USD 600 per tonne and 40.5 GJ per tonne, or about USD 14.8 per gigajoule. The best published case is still about 2.1 times fossil fuel per unit of delivered energy, before an ammonia engine's pilot-fuel requirement and before any efficiency penalty. Frontier The heating-value and price inputs are unverified here; read it as an order-of-magnitude comparison, not a quotation.

Established Well-to-wake accounting is where the emissions move rather than disappear, and LNG is the canonical case. LNG delivers approximately 20–25% lower CO2 than heavy fuel oil when methane slip is strictly controlled below 2%. Actual unburned slip runs 1–3%, methane's global warming potential is 28–34 times CO2 over a century, and slip can offset 30–50% of the climate benefit. Established The tank-to-wake advantage is real; the well-to-wake advantage is contingent on a combustion parameter that varies with engine type and load. Frontier Ethanol across four production pathways delivers well-to-wake reductions from 50.2% for corn to 76.9% for wheat straw, against 20.6% for LNG. Established Green ammonia is quoted at up to 90% lifecycle reduction, bio-methanol up to 80%, e-methanol potentially net-zero on renewable electricity. None is 100%, and the residue has moved upstream to the electrolyser's grid connection and the synthesis plant.

Established On the landward side, shore power is an electrical engineering problem with a frequency in it. Low-voltage shore connection runs 380–690 V; high-voltage typically 6.6–11 kV; nominal onboard voltage is 400/440 V AC with 6.6 kV and 11 kV also in use. Established Approximately 75% of ships are designed for 60 Hz and the remaining 25% for 50 Hz, so a European port on a 50 Hz grid must convert for three-quarters of arriving tonnage, and the converter is a large fraction of shore-power capital cost. Established Centralised AC architectures rate converter and transformer from a few megawatts up to 10 MW; distributed AC uses converters typically around 1 MW each, with containerised blocks from several hundred kVA to more than 10 MVA. Established Installed port-level capacity reaches up to 60 MW at Los Angeles; cruise terminals at Vancouver and Hamburg exceed 12 MW; early implementations at Gothenburg, Zeebrugge and Juneau were typically 1–10 MW. These are port and terminal aggregates serving many berths, not per-vessel hotel loads, and a verified per-vessel figure could not be obtained for this brief.

Established What shore power actually buys is smaller than its billing, and the reason is boilers. Current EU regulations are projected to achieve only a 24% reduction in annual CO2 emissions within EU ports, approximately 4.37 Mt. Established Boilers account for 44% of CO2 emissions in EU ports, and electrifying them would raise shore-power infrastructure demand by approximately 15–26%. Shore power as usually specified addresses auxiliary engines; the boiler share is nearly half the problem and sits outside most scopes. Established For calibration on money: an EUR 18.8 million Connecting Europe Facility project across four northern ports, an EUR 3.2 million grant at Port of Antwerp Bruges, and an EUR 570 million Italian scheme reducing electricity charges until 2033. Frontier HVDC cut transmission losses by approximately 33% against the AC alternative considered.

Frontier Terminal automation does not outperform manned operation in the independent literature, and this is the finding that most contradicts the sales material. A two-stage data envelopment analysis across 20 Mediterranean container ports for reference year 2023 found average efficiency 0.62 under the constant-returns model with three fully efficient ports, and 0.72 under variable returns with seven. The authors conclude that “automation alone does not necessarily lead to higher efficiency unless it is effectively integrated into operations accompanied by adequate staff training and supported by gradual investment strategies”. The stronger performance driver they identify is cargo intensity — TEUs per call. Frontier An independent second instance: a discrete-event simulation of Ro–Ro terminal tractors at the Port of Ravenna held the fleet at seven tractors across internal-combustion, battery-electric and electric automated guided configurations, with 253 semitrailers unloading and 184 loading, triangular payload of 15,000/25,000/35,000 kg, two charging stations, triangular charging time 20–30 minutes with mode 25, charging triggered at 15% state of charge, 20 repetitions per scenario. Frontier Emissions fell up to 40% for both electrified configurations — and the automated-tractor scenario showed increased unloading time and reduced fleet availability from charging constraints and routing limitations. Established The honest reading of both, and the one this brief adopts, is that the productivity case for terminal automation is unproven in the independent literature — not that automation makes terminals worse. Twenty ports, one region, one year and a conditional conclusion is strong evidence against a reliable effect and no proof of a negative one.

Frontier On regulation this brief is deliberately thin, and the thinness is the point. A 2026 review reports FuelEU Maritime phased GHG-intensity reductions from a baseline of 91.16 gCO2eq/MJ — 2% by 2025, 6% by 2030, 14.5% by 2035, 31% by 2040, 62% by 2045, 80% by 2050. That ladder is internally coherent and could not be checked against the regulation itself; treat it as a secondary-source reading. Frontier The same review attributes to IMO strategy a carbon-intensity cut of at least 40% by 2030 and annual GHG reductions of at least 20% aiming for 30% by 2030, alongside a long-term target of at least 50% by 2050 — and that last figure appears to belong to an earlier instrument than the checkpoints beside it. No adoption date, article number or entry-into-force date is stated on this page.

Established Wind assistance is the only abatement option on this page measured on working ships by a third party, and the number is smaller than the brochures and larger than the scepticism. The reference case is a product tanker fitted with two rotor sails and monitored across a year of ordinary trading, with the saving verified by a classification society rather than by the supplier: about 8% of fuel burned. The best-publicised rigid-wingsail installation, on a bulk carrier in 2023–24, was reported by its charterer at roughly three tonnes of fuel a day, with much higher figures on favourable routings. Frontier Read those together and the finding is not a percentage but a dependency: a wind-assist saving is not a property of the device. It is a property of the device, the route’s wind-angle distribution, the ship’s speed and the master’s willingness to sail a longer great circle for a better beam reach. Frontier The installed base is of order a hundred vessels on an industry register against a world fleet above a hundred thousand ships — about one in a thousand.

Established The well-to-wake rulebook now decides more of the answer than the chemistry does, and its most consequential parameter is a default slip factor. Lifecycle guidance for marine fuels works as the section above assumes — well-to-tank plus tank-to-wake, methane and nitrous oxide carried as carbon-dioxide equivalent — and the default tank-to-wake tables are resolved by engine family rather than by fuel. A low-pressure dual-fuel Otto engine of the medium-speed four-stroke type carries a default methane slip near 3%; the slow-speed two-stroke Otto variant about 1.7%; a high-pressure diesel-cycle gas injection engine about 0.2%. Two ships burning identical LNG from the same bunker barge are therefore credited with materially different climate performance because of what is bolted to their crankshafts. Frontier These are defaults, and that is the provision that matters: a ship measuring its own slip can generally be credited with the measured value, which turns a combustion-laboratory result into money and creates the first commercial demand for onboard slip measurement. The figures are quoted as defaults in the European fuel-intensity instrument and should be checked against the current table; no resolution number or adoption date is asserted here, consistent with this page’s treatment of the regulatory record.

Established The time horizon is the other lever, and it is chosen rather than measured. Methane’s global warming potential is roughly 30 times carbon dioxide over a hundred years and roughly 80 times over twenty; exact values move between assessment reports, the ratio does not. Every lifecycle result for LNG in this literature is computed on the hundred-year convention, and on the twenty-year convention the same slip rates consume most of the advantage. Nothing physical distinguishes them; the choice is a policy judgement made in a footnote. Frontier The equivalent arithmetic for ammonia is worse and can be done on this page’s own numbers. Nitrous oxide is about 270 times carbon dioxide over a century; ammonia is 82% nitrogen by mass at about 18.6 MJ/kg, so converting a fraction of the fuel nitrogen to nitrous oxide costs about 350 kg of carbon-dioxide equivalent per kilogram of ammonia per unit fraction, against roughly 1.4 kg of CO2 for the gas oil delivering the same energy. Converting about four parts in a thousand of the fuel nitrogen erases ammonia’s entire advantage over fossil fuel. That is arithmetic on stated assumptions, and it is why the unmeasured nitrous oxide term in the workback chain is not a rounding error but the whole case.

Frontier The unit of analysis is shifting from the fleet to the corridor, and the shift is not cosmetic: a named origin-destination pair with coordinated bunkering at both ends is the only structure in which a fuel with no global supply chain can actually be burned. Frontier Alongside it, total-cost-of-ownership modelling with Monte Carlo over corridor-specific fuel prices is methodologically newer than the comparative life-cycle assessments it displaces, because it prices emissions-trading exposure and compliance charges into a present value and reports a probability rather than a point estimate.

Frontier Ammonia's frontier is toxicity, not combustion. The immediately-dangerous-to-life-or-health concentration is 300 ppm, the lethal concentration 2,700 ppm, and the flammability range 16–25% by volume. The open question is not whether an engine will burn it — low-speed two-stroke retrofit work is proceeding, including at doctoral-thesis depth in 2025 — but whether a crewed vessel can carry it under any manning regime. Frontier A hazard-identification-led design study across hydrogen, battery and ammonia vessels put hydrogen flammability at 4–75% and its explosion range at 18–59% in air, driving double-walled super-insulated tanks and gastight fuel-cell enclosures, and assessed fuel-cell-room leakage and battery-room fire as the most severe hazards for those two variants. Frontier A parallel line treats the tank arrangement rather than the fuel as the cost driver. That is the right variable and almost nobody uses it.

Frontier Ethanol has re-entered the marine fuel conversation on well-to-wake numbers that beat LNG by a wide margin, and on the argument that it is a drop-in-adjacent ambient liquid needing neither cryogenics nor pressure. It is the least-discussed serious candidate, and the reason is inertia rather than evidence. Frontier At the other end, well-to-wake assessment of scrubbers against low-sulfur fuels is now being supported by measurements taken aboard an ocean-going vessel; measurement-supported rather than model-only LCA is this field's methodological frontier, and there is very little of it.

Frontier Shore power is being reclassified from an air-quality measure into a compliance instrument, with cold ironing folded into greenhouse-gas fuel-intensity accounting and net-zero fund contribution calculations — which changes who pays for it and why. Frontier In parallel, fleet-based compliance analysis asks which ships in an existing fleet can comply, rather than what a new ship should be; that is the question determining the 2040 fleet and the newbuild-centric literature has largely skipped it. Frontier On the electrical side, distributed containerised converter blocks are emerging as an alternative to a single centralised substation, reducing the step change in grid connection. Frontier And patent-landscape analysis has become a research method in port electrification, on the argument that the patent record shows where converter-topology work is actually happening, as distinct from where papers are published.

Speculative Nuclear merchant propulsion has been formally reopened. A reactor-selection methodology for merchant shipping appeared in Nuclear Technology in 2025, and that such a paper appears in that venue is itself the news. The historical programme — four civil nuclear merchant ships, built and withdrawn — could not be sourced for this brief: no reachable journal carried the build costs, reactor powers or withdrawal reasons. Those figures are named here as a lead and no number is printed for them. The retrospective that would close the gap is Khlopkin and Zotov's 1997 review, verified to exist and not read.

Speculative Fully autonomous deep-sea shipping is blocked by crewing law rather than by autonomy. There is no international competence standard for remote operators; a goal-based gap analysis against the seafarer training convention identifies the shortage of “ship sense” available to a remote operator, on a focus group of three veteran training-ship instructors, which is expert elicitation and not measurement. Frontier Collision regulations are written in natural-language seamanship terms — “narrow channel”, “restricted visibility”, “best aid to avoid collision” — with no machine-checkable definition, and a 2023 paper attempts to quantify them; if that succeeds it is the enabling step for deep-sea autonomy, and if it fails autonomy stays coastal permanently. Frontier Salvage law has no counterparty without a master: the 1989 convention does not cleanly accommodate uncrewed vessels, and rescuing a drifting hull presumes someone aboard who can accept a contract. Speculative The most interesting unexplored link in the whole area is the complementarity: removing the crew removes the accommodation block, the hotel load, the lifeboats and the freshwater plant, and frees exactly the volume an ammonia or hydrogen ship needs — while the ammonia hazard is fundamentally a hazard to crew. Nobody has run that calculation.

Frontier One economic result cuts against the pessimists and deserves its weight. A 2026 study of shipping costs and UK inflation finds higher shipping costs a significant driver of both the level and volatility of inflation, with the strongest effects from containerised freight — but that the inflationary effects are predominantly transitory. If that holds, the political-economy objection to a fuel mandate is weaker than the industry's own rhetoric assumes, and the affordability case has to be made on capital cost and fuel supply instead.

Frontier Wind assistance couples to speed, which makes it a commercial decision rather than an engineering one. Required propulsive power rises roughly as the cube of speed while the thrust a rotor or wingsail produces does not, so the wind share climbs steeply as the ship slows: the installation worth a few per cent at service speed is worth a much larger fraction at slow steaming. Frontier Routing is the second coupling — optimising for wind rather than for waves buys fuel at the price of voyage days — and no study located reports that joint optimum as a curve for a named route. That curve, not a headline percentage, is what a charterer would need.

Frontier The unresolved question under the fuel-supply problem is whether the molecule has to arrive in the ship that claims it. A physical-delivery rule means green fuel is bunkered at the ports the ship actually calls, which is exactly the corridor structure treated above as this field’s emerging unit of analysis. A mass-balance or book-and-claim rule lets an owner buy the attribute of fuel produced anywhere and burn conventional fuel on the voyage, decoupling compliance from bunkering geography and making the corridor unnecessary. The same production volume yields two completely different industries, and the choice sits in chain-of-custody provisions rather than in any headline target. This page has not read the operative text, so the observation is about what turns on the decision.

Established Here is the ordered chain a zero-carbon deep-sea fleet has to walk, with the measurement that would tell you each link had landed. The target state is deep-sea merchant shipping at well-to-wake near-zero greenhouse gas, at a freight rate world trade absorbs, at fleet scale, without a carbon price so high it is politically uncollectable.

Frontier L1 — a fuel whose installed volume penalty is bounded. Measure displaced cargo volume and deadweight for a reference hull at reference range, including insulation, cofferdams and hazardous-zone standoff. The published 2.3–4.3 times figures are fluid volume only. Frontier L2 — an engine burning it at deep-sea power and load profile, in service. Measure brake thermal efficiency and pilot-fuel fraction on a real voyage rather than a test bed, with nitrous oxide and unburned-fuel slip measured aboard; for ammonia the nitrous oxide term can erase the carbon saving and no measured marine figure was found. Frontier L3 — a safety case accepted for a crewed ship. Measure the class notation actually issued, the volume of the vessel inside the 300 ppm envelope, and whether crew are inside it during bunkering.

Frontier L4 — green fuel produced at price. Measure delivered USD per gigajoule at the bunker port, with the contracted electricity cost at the production site stated alongside it. Frontier L5 — additional zero-carbon generation. Measure TWh per year contracted to a marine-fuel offtake with additionality demonstrated. Frontier L6 — bunkering at both ends of real routes. Measure ports with certified bunkering, weighted by tonnage calling. Established L7 — fleet turnover. Measure the share of deadweight that is fuel-capable and the remaining asset life of the share that is not; twenty-five-year lives mean the 2050 fleet is largely ordered before 2035, which makes this link arithmetic rather than uncertain. Frontier L8 — an instrument that makes L4 rational for a private owner. Measure the realised carbon price or mandate stringency faced by a non-EU operator on a non-EU leg. Established L9 — port-side electrical capacity. Measure MW per berth installed including boiler load, plus grid carbon intensity at the connection point.

Frontier L5 binds. L1 to L3 are engineering with visible progress and no discontinuity in them. L4 is downstream of L5, because the fuel cost is the electricity cost plus conversion. L7 is slow but forecastable. L8 is political and could move quickly in either direction. L5 is the only link that requires physical generating capacity at national-grid scale, additional to decarbonising the grids that already exist, on a schedule set by somebody else's siting and permitting — and siting has no learning curve. Speculative The only route that removes L5 from the chain entirely is nuclear propulsion, and that is an argument about the structure of the workback plan rather than an argument about reactors. Frontier The secondary binding candidate is L9's boiler term: the cheapest link to fix, the most consistently descoped, and therefore the likeliest source of an unpleasant surprise in a port's capital plan.

Frontier L10 — conversion throughput, which the chain above leaves implicit. Measure dual-fuel retrofits completed per year, in yard-days per ship and berths available. The first large container-ship methanol conversion was reported completed in 2025 at a Chinese yard on a 15,000 TEU hull, and the instructive quantity is not the capital cost but the roughly three months out of service: at charter rates the off-hire is frequently the larger half of a retrofit’s true cost, and it is the half that cannot be subsidised away, because it is a physical queue. Frontier That changes the fleet-turnover argument. Twenty-five-year asset lives make the mid-century fleet an order-book question only if retrofit is negligible; if it is viable, the 2050 fleet is also a yard-capacity question, and conversion slots compete directly with newbuilding in a small number of Asian facilities. No conversion-throughput figure is published, so the most important sensitivity in the fleet forecast is a blank.

Established Six things this subject needs from other people, five of which are measurements nobody has published. First, an authoritative heating-value and density table for marine fuels, with the basis stated per row. The best available review contains at least three errors in that one table, and the correction had to be made by internal inconsistency rather than against an authority, because the general reference literature was unreachable.

Frontier Second, a measured nitrous oxide slip figure from marine ammonia combustion. It is potentially decisive for ammonia's well-to-wake case and it does not appear in anything read for this brief; a 2025 life-cycle assessment of ammonia and hydrogen for marine internal combustion engines is the likeliest place it lives. Frontier Third, a per-vessel shore-power hotel load for a large container ship, in megawatts. Only port-level aggregates were obtainable, and dividing an aggregate by a berth count is not a substitute; a Port of Gävle demand-forecasting study is the likeliest source of a real per-terminal profile.

Frontier Fourth, global marine bunker consumption in tonnes per year and the electricity volume an e-fuel transition implies from it. The statistical agencies that hold this were unreachable, and without it the electricity claim in section 2 stays qualitative. Speculative Fifth, container-terminal productivity in crane moves per hour, automated against manned, with a control group. This appears not to exist in the independent literature at all: it is a research gap rather than a retrieval gap, and it is the number the whole automation debate is missing. Speculative Sixth, the operating and capital record of the four civil nuclear merchant ships. That record exists and was not obtainable here; it is named as a lead and no figure from it is printed. Established The pattern across all six is worth naming: five are numbers somebody could measure this year for the cost of a stack test, a metering campaign or a spreadsheet, and the sixth is a library visit.

Speculative The highest-value unrun experiment in this subject is cheap, and it is a variance test. If terminal automation's gains are real but conditional on integration and training — which is what the Mediterranean study's own hedge says — then automated terminals should show higher between-terminal variance in productivity than manned ones, not a uniform advantage. Compute the between-terminal variance in productivity among automated terminals against manned ones. If automation's variance is higher, the technology is not the causal factor and the implementation is. Nobody appears to have run this, the data mostly exists, and it would settle a decade of marketing.

Frontier The second-cheapest test has already run and nobody has read it out. Carbon-intensity regulation has driven speed reductions across parts of the world fleet, which is a natural experiment in freight demand elasticity to transit time. Measure the realised freight-rate and volume response. If demand for deep-sea tonne-miles is more elastic than the industry assumes, the cheapest abatement is fewer tonne-miles rather than cleaner ones, and the entire fuel-centric literature is answering the wrong question. This is the least-explored branch in the subject.

Frontier Third: measure nitrous oxide and unburned ammonia slip aboard a vessel in service, at deep-sea load profile, not on a test bed. It is a stack measurement on a ship that already exists. Speculative Fourth: run the naval-architecture comparison nobody has run. For one reference hull, compare cargo capacity of uncrewed-plus-ammonia against crewed-plus-ammonia against crewed-plus-VLSFO. If the first beats the third, autonomy and decarbonisation become easier together than either is alone, and the strategic picture for deep-sea shipping changes shape. The closest existing work is a policy chapter, not a naval-architecture study.

Frontier Fifth: publish the lending spread between green-fuel newbuilds and conventional ones. That spread is the market's estimate of policy credibility, it is observable today, and it is a better instrument than any announced carbon price. Frontier Sixth: report corridor green-fuel tonnage as a share of world green-fuel production. If corridors are consuming most of the world's supply, they are a display case rather than a ramp, and the distinction is measurable now.

Established The ship-side engineering problem is volume, and it is worse than the fluid ratios suggest. Double-walled tanks, super-insulation, cofferdams, gastight enclosures and hazardous-zone standoff all consume hull volume that never appears in an energy-density table. For hydrogen the hazard-identification work drives double-walled super-insulated tanks and gastight fuel-cell enclosures on a flammability range of 4–75%; for ammonia the constraint is a 300 ppm exposure limit inside a crewed hull. Frontier Treating the storage arrangement as the cost driver rather than the fuel is the correct engineering framing and it is rare in the literature.

Established The port-side engineering problem is a substation and a frequency. Three-quarters of ships run at 60 Hz; most large ports outside the Americas sit on 50 Hz grids, so the converter is not an accessory but a structural cost, and neither newbuild convergence on one frequency nor a collapse in converter cost is visible in the sources read here. Established Centralised architectures need converter and transformer ratings from a few megawatts to 10 MW; the distributed alternative uses roughly 1 MW blocks, up to containerised units above 10 MVA, which lets a port build incrementally rather than committing to one grid-connection step change. Frontier HVDC reduced transmission losses by about 33% against the AC alternative in the review read here.

Established And the descoped item is the boiler. Boilers are 44% of in-port CO2 and electrifying them adds approximately 15–26% to shore-power infrastructure demand. A port that sizes its connection for auxiliary engines only has sized it for slightly more than half the problem. Frontier A port's honest capacity number is megawatts per berth including boiler load, and that is not the number most capital plans carry. Frontier The same applies to the grid behind it: a shore-power berth is a large new load on an urban distribution network that was never planned for it, and the connection study rather than the quayside equipment is usually the long pole. Speculative Nothing in this section is technically hard. All of it is slow, because it is civil and electrical works inside a working port that cannot stop working.

Frontier The wind-assist engineering problem is deck real estate and air draught, not aerodynamics. Unobstructed deck is scarce on a container ship stacked with boxes and on a bulker occupied by hatch covers and cranes, so the installed fleet is dominated by clear-deck hull types — a selection effect rather than a market preference — and a fixed device must clear the bridges and gantries of every port on the trade, which is why tilting mechanisms exist and why they carry the reliability risk. Frontier The unbuilt item is the instrument: continuous funnel measurement of methane and nitrous oxide slip, robust enough for unattended use, which the default-versus-measured provision has just made worth having.

Established Electrolyser manufacturing and renewable siting are the adjacent technologies that actually decide this. At 50–60 kWh per kilogram of hydrogen, every green marine fuel is a derivative of installed zero-carbon generation, and the relevant metric is gigawatts per year of shipped electrolyser capacity rather than announced capacity. Frontier An electrolyser capital cost of around USD 3,000 per kW circulates in the review literature without a technology, a scale, a date or a range attached to it; this brief treats that figure as unusable and does not carry it.

Frontier Transmission choice is genuinely open: electron or molecule. Levelised cost of transmission has been compared across pipelines, tankers and HVDC for green hydrogen and ammonia in new-build offshore infrastructure, and the answer determines whether marine fuel is synthesised where the wind is or where the port is. Frontier Downstream of that sits carbon capture, because e-methanol needs both green hydrogen and a CO2 source, and the provenance of that CO2 decides whether e-methanol is net-zero or merely delayed.

Speculative Small modular reactors are adjacent in a way that is not obvious. If merchant nuclear ever returns it will return through a design certifiable by a classification society, not only by a nuclear regulator — a different artefact from the design certifications the SMR sector is currently pursuing. Frontier On the autonomy side, the adjacent stack is collision-avoidance certification, AIS-derived scenario generation built from twelve months of real traffic data rather than from arbitrary or regulation-derived test cases, and cybersecurity now treated at chapter length as a first-class constraint on uncrewed vessels rather than as an afterthought. Frontier And ammonia's marine safety case is adjacent in the other direction: solving it is directly transferable to the ammonia-as-energy-carrier trade that the hydrogen economy assumes and has not demonstrated, which means marine ammonia is being asked to de-risk a much larger sector than shipping.

Frontier The central institutional question is price versus mandate, and the measured evidence leans away from price. If a carbon price at 73.5 EUR per tonne cannot produce a positive probability of cost advantage over thirty years, and the best published green-ammonia case needs about USD 120 per tonne merely to reach parity with grey ammonia, then the price required is well above anything imposed on international bunkers. Frontier The serious alternative is a fuel-intensity mandate, which bypasses price discovery by making non-compliance impossible rather than expensive; the ban-fossil-fuels position has been argued in the policy literature, and any brief concluding that decarbonisation is unaffordable has to engage the mandate route or it is not arguing at full strength.

Frontier Enforcement is the condition everyone skips. A price or a mandate is only real if it is collectable, which makes port state control the operative institution rather than the flag state, and the share of tonnage that reflags or reroutes in response to a first price step is the measurement that tells you whether the instrument is enforceable at any level. Frontier Compliance cost also has to land on somebody: the allocation between owner and charterer under time charterparties is an active legal question and it determines who has the incentive to invest.

Speculative For nuclear, the binding institutions are not nuclear ones. Port-state acceptance killed the historical programme substantially, and no technical advance touches it; the measurement is the number of major ports that would grant entry. Liability and insurance are the second: whether a protection-and-indemnity club writes the cover, and at what premium, under existing nuclear liability conventions. Speculative For uncrewed deep-sea shipping the equivalent list is a remote-operator certification standard analogous to the seafarer training convention, a classification society willing to certify a collision-avoidance algorithm rather than review a paper about one, a salvage framework with a counterparty who is not a master, and a flag state willing to register an uncrewed hull for international voyages. Speculative The fourth binds first, because a flag state can act unilaterally and quickly while everything else requires multilateral consensus. The testable prediction is that deep-sea autonomy, if it arrives, arrives through one permissive register rather than through a convention — a near-term indicator nobody is watching.

Established The clearest ethical fact in the subject is that port emissions fall on identifiable people. Roughly 40% of the CO2 from ships visiting ports is emitted during port activities, and the co-emitted pollutants land on populations who live beside working waterfronts and who did not choose the trade. That is why shore power is worth building even where its climate contribution is modest — and it also means the honest case for shore power is a public-health case, not a climate one.

Frontier Ammonia relocates risk onto crews. A fuel with a 300 ppm immediately-dangerous concentration and a 2,700 ppm lethal concentration is being proposed for vessels whose crews sleep aboard, bunker in port, and are often recruited from labour markets with weak bargaining power. The safety case is not an engineering formality; it is a question about who bears a hazard chosen by somebody else. Speculative The uncomfortable corollary is that the strongest argument for uncrewed ships is that they remove the person the hazard is dangerous to.

Frontier Cost incidence is the second distributional question. Freight cost increases propagate into consumer prices and into the trade costs of economies that are far from markets, and the one empirical study reached here finds the inflationary effect predominantly transitory — which weakens the standard objection but does not make the incidence even. Frontier Terminal automation carries its own: dock labour is one of the few remaining well-paid, unionised, non-graduate occupations in many port cities, and it is being displaced on a productivity argument that the independent literature does not support. Speculative A port that automates and does not gain efficiency has transferred income without producing anything, which is the worst available outcome and the one the current evidence cannot rule out. Frontier And there is a consent question at the end of the nuclear branch that no engineering answer touches: the historical merchant programme died substantially on port access, which is to say on cities declining to host a reactor at their waterfront. That refusal is a legitimate exercise of local authority, and any serious workback plan for nuclear shipping has to treat it as a position to be negotiated rather than an obstacle to be engineered around.

Established Sea freight is the substrate the rest of the economy sits on, and it is unusually cheap. That cheapness is why supply chains are shaped the way they are, and a durable increase in the cost of moving a tonne-mile would reshape them — not catastrophically, but structurally, by making distance matter again in decisions that currently ignore it.

Frontier The branch nobody presses is demand. Slow steaming, routing and reduced tonne-miles are the only measures with plausibly negative abatement cost, and they are nearly absent from a literature organised around fuels. If deep-sea freight demand is more elastic to transit time than the industry believes, the civilizational answer to shipping emissions is not a new molecule but a slower, shorter, more regional pattern of trade — which is a political outcome disguised as a technical one.

Speculative The opposite branch is nuclear, and it is the only one that makes the problem smaller rather than moving it. Nuclear propulsion solves the volumetric penalty and the electricity-supply constraint simultaneously, which is exactly why it keeps returning after each abandonment. It fails on port access and liability, and neither of those is a technology problem, which means the question of whether ships go nuclear is a question about what ports will consent to. Handwave A world in which a hundred ports change that answer at once looks very different from a world in which none do, and there is no gradual version.

Speculative The quieter civilizational point is about who does the deciding. Nothing on this page is settled by naval architects. It is settled by whoever sets the price of electricity at a production site, whichever flag state first registers an uncrewed hull, and whichever port authorities grant or refuse entry. Frontier Shipping is the sector where a global physical system is governed by consensus among sovereigns who can each defect at low cost, and it is therefore the best available test of whether that governance form can deliver a capital transition at all.

These horizons track what would have to be observable, not what is announced. Each is anchored to a measurement named earlier on this page.

• 10 yr: Frontier Methanol dual-fuel newbuilds accumulate on corridors with coordinated bunkering at both ends, and the honest test is whether corridor green-fuel tonnage grows faster than world green-fuel output rather than being reallocated from it. Frontier Ammonia enters service in retrofitted low-speed two-stroke engines and the first measured nitrous oxide slip figures appear. Frontier Shore power moves from air-quality programme to compliance instrument, and the first ports discover the boiler term.
• 25 yr: Frontier Fleet turnover decides the outcome: twenty-five-year asset lives mean the mid-century fleet is largely ordered before the mid-2030s, so the binding decisions in this horizon are order-book decisions taken under policy uncertainty rather than technology decisions. Speculative If additional zero-carbon generation has not been contracted at national-grid scale with demonstrated additionality, no fuel pathway matters, because the fuel cost is the electricity cost. Speculative A single permissive flag state registering an uncrewed hull for international voyages is the most likely form of the autonomy breakthrough.
• 50 yr: Speculative Either a fuel-intensity mandate has survived evasion and reflagging, or the transition has stalled into a two-tier fleet: compliant tonnage on regulated routes and conventional tonnage everywhere else. Speculative Nuclear merchant propulsion is decided in this window and it is decided by port authorities and insurers, not by reactor designers. Handwave A genuinely uncrewed, ammonia-fuelled deep-sea hull — the configuration where autonomy and decarbonisation pay for each other — is physically coherent and institutionally unprecedented, and nobody has yet published the naval architecture.
• 100 / 250+ yr: Handwave On this horizon the interesting scenario is not a better fuel but a different trade pattern: if the cheapest abatement turns out to be fewer tonne-miles, the sea gets quieter rather than cleaner, and the port becomes a regional rather than a global institution. Handwave The competing scenario is abundant cheap zero-carbon electricity making synthetic fuel a rounding error, at which point every constraint on this page dissolves at once and the whole debate reads as a transitional artefact.

Frontier Two cheap additions to the ten-year row. Whether wind-assist installations are reported with route and speed attached rather than as bare percentages; and whether onboard slip measurement becomes ordinary, since the lifecycle rules credit a measured value against a default and that is the first commercial reason to instrument a funnel. Handwave The counter-case is that owners whose engines already sit on the favourable line of the table simply accept the defaults, in which case nothing is measured and the parameter stays a negotiated number.

• Depends on This brief depends on results that are not produced by shipping research at all. It waits on electrolyser manufacturing scale-up and on renewable siting and grid connection, because a marine e-fuel is a derivative of installed zero-carbon generation and its cost is the delivered electricity price plus conversion. It waits on a trustworthy thermophysical properties table for marine fuels with the heating-value basis stated per row, because the best available review's table carries at least three errors. And it waits on measured nitrous oxide slip from ammonia combustion in service, which is the one unmeasured quantity that could invalidate ammonia's well-to-wake case after the fleet is ordered.
• Requires (not on this map) The first constraint is the one the cost evidence is really about: an adopted, enforceable greenhouse-gas framework covering international bunkers, not only voyages that touch one regulated bloc. A measured carbon price of 73.5 EUR per tonne produced zero probability of a methanol cost advantage over thirty years, so the instrument has to be either far stronger or of a different kind — a fuel-intensity mandate rather than a price — and it has to be collectable against reflagging. The second is industrial: electrolytic hydrogen at marine-fuel volumes, which today means moving from under 1 million tonnes a year of green hydrogen, and from about 400 tonnes a day of world liquefaction capacity, to something two to three orders of magnitude larger, with the additional generation contracted and additionality demonstrated. The third is the landward build. The token names a 2029 shore-power obligation; this brief carries it as the adjudicated constraint it is and states no instrument date, because the primary source could not be read. The obligation's shape is what matters here: a requirement that ports serving container and passenger vessels above a size threshold provide shore-side electricity at berth, converting a discretionary programme into a mandated one. Sizing it correctly means megawatts per berth including boiler load, since boilers are 44% of in-port CO2 and electrifying them adds roughly 15–26% to infrastructure demand, plus frequency conversion for the three-quarters of ships that run at 60 Hz. The fourth constraint is a queue rather than a technology: conversion yard throughput, measured in dual-fuel retrofits completed per year and in the months a hull spends out of service to receive one, because it decides whether the mid-century fleet is only an order-book question or also a retrofit question. No conversion-throughput figure is published anywhere located.
• Enables What this subject enables runs outward from the quay. A port with real electrical capacity is the precondition for battery-electric short-sea and harbour craft, for electrified terminal equipment, and for cold ironing as a compliance instrument rather than a courtesy. A certified collision-avoidance stack and a permissive register would enable uncrewed deep-sea shipping, which in turn frees the accommodation volume that low-density fuels need. And a solved marine-ammonia safety case is directly transferable to the ammonia-as-energy-carrier trade that the hydrogen economy assumes but has not demonstrated.
• Adjacent Adjacent work sits in three places: HVDC and pipeline transmission economics, which decide whether fuel is synthesised at the wind or at the port; carbon capture, which decides whether e-methanol is net-zero or merely deferred; and small modular reactors, whose certification pathway runs through nuclear regulators when merchant shipping would need a classification society notation instead.

Established Start with the largest single error found anywhere in this project: a 2026 peer-reviewed review prints liquid ammonia's density as 0.68 kg per cubic metre. The correct value at ammonia's boiling point is approximately 682 kg per cubic metre — 0.682 kg per litre. The printed figure is out by a factor of about 1,000, and it is close to the density of ammonia vapour at standard conditions, so the likely mechanism is a kilograms-per-litre to kilograms-per-cubic-metre conversion failure, or a gas-phase value pasted into a liquid-phase row. Established It is self-refuting from inside the same paper. That paper states ammonia at 19 MJ/kg, states its boiling point correctly two lines away, reports 0.04% per day boil-off from a liquid tank at 1.5–2.5 bar, and proposes ammonia as a deep-sea marine fuel. At 0.68 kg per cubic metre a cubic metre of tank would hold 0.68 kg and about 12.9 MJ — roughly a third of a litre of diesel per tonne of tank structure. Established No fuel with those properties could be proposed for deep-sea propulsion, and the same paper proposes it. Any tank-sizing calculation taking the figure at face value is wrong by three orders of magnitude, in the direction that makes ammonia look impossible.

Established The same table carries two more errors, and the second cuts the other way. Under liquid hydrogen it prints 1,170 kJ/L “at 100 bar” — but liquid and 100 bar are mutually exclusive for hydrogen in marine storage, and the same paper states two lines away that liquid hydrogen sits at 1.5–5 bar and 70.8 kg per cubic metre. Established Taking the paper's own density with a 120 MJ/kg lower heating value gives about 8,500 kJ/L; the printed 1,170 kJ/L is roughly consistent with compressed hydrogen gas at about 100 bar, so a compressed-gas figure appears to sit in the liquid row. Established Note the direction. At face value it implies a volumetric penalty against marine gas oil of roughly 31 times, where the paper's own density implies about 4.3 — it makes hydrogen look about seven times worse than the source's own data supports. A brief arguing hydrogen is volumetrically hopeless must not lean on it, and this one does not.

Established The third error is a basis mismatch. The table prints hydrogen at 141,500 kJ/kg against methanol at 20.0 MJ/kg and ammonia at 19 MJ/kg. 141.5 MJ/kg is hydrogen's higher heating value; its lower heating value is about 120, and methanol's 20.0 is an LHV figure. Established The table therefore compares hydrogen on an HHV basis against everything else on an LHV basis, inflating hydrogen's apparent gravimetric advantage by about 18%, and the error is invisible because there is no basis column. Marine engines and most fuel cells cannot recover the latent heat of the product water, so LHV is the appropriate basis throughout. Three errors in one table, in one peer-reviewed 2026 review, all detectable without leaving the paper.

Established So here is what to do with any fuel-comparison table, this page's included. Check three things before using a row. One: does every energy density state its heating-value basis? If not, assume HHV and LHV are mixed. Two: does each density row state a phase, a temperature and a pressure, and are the three mutually possible? “Liquid at 100 bar” is not a state hydrogen occupies in a ship. Three: multiply the gravimetric figure by the density and see whether the volumetric column comes back. Frontier Two of the three errors above fail that last test in one line of arithmetic. The review is not unusually bad — this brief relies on it for production scale and regulatory framing. A table is simply the part of a paper least checked by reviewers and most copied by everyone else.

Established “Green hydrogen is USD 5–7 per kilogram” and “green ammonia will be USD 0.59 per kilogram” are real published figures, and neither is a cost figure. Neither carries an electricity price, and at 50–60 kWh per kilogram electricity dominates: every USD 10 per MWh moves hydrogen by roughly USD 0.55 per kilogram. Frontier The 0.59 is an optimal 2050 case for one wind configuration at a 20% flexibility factor producing 64 tonnes a day. A green-fuel cost claim without a stated electricity price and capacity factor is not a cost claim. Frontier The discipline cuts both ways: an electrolyser capex of around USD 3,000 per kW, given as a projection with no technology, scale, date or range, sits at the far edge of the spread and makes green hydrogen look worse than it is.

Frontier “Automated terminals are more productive” does not survive the independent literature. Productivity claims for terminal automation overwhelmingly originate with equipment vendors and integrators, are stated as ranges, and carry no manned control group. Established But state the counter-caution honestly, because it applies to this brief's own strongest finding: twenty ports, one region, one year, a small sample relative to the model's dimensions, and a conclusion phrased as a conditional. That is strong evidence against a reliable automation effect — not proof that automation reduces efficiency, and it must not be used as one.

Established “LNG is a 20–25% CO2 reduction” is a tank-to-wake number quoted as a climate number: it holds only if methane slip stays below 2%, actual slip runs 1–3%, and slip can offset 30–50% of the benefit. Established “Ammonia is a zero-emission fuel” is wrong twice — it is zero-carbon, green ammonia is quoted at up to 90% lifecycle reduction rather than 100%, and its combustion produces nitrous oxide, for which no measured marine figure was obtainable here. Established “Hydrogen has the highest energy density” is true per kilogram and false per litre, where it is the worst of the five — and a ship is volume-constrained, not mass-constrained, on the fuels that matter.

Established “A container ship needs 60 MW of shore power.” That is port-level installed capacity at Los Angeles serving many berths, not a vessel hotel load; conflating them overstates per-berth substation requirements by a large, unquantified factor, and no per-vessel figure is supplied here from memory. Established “Shore power decarbonises ports” overstates a real but modest effect: 24% of annual port CO2 EU-wide, about 4.37 Mt, while boilers are 44% of in-port emissions and sit outside most scopes. Shore power on a coal grid moves emissions rather than removing them.

Frontier “The IMO has committed to net-zero by 2050” and “to 50% by 2050” both circulate, and they are not the same commitment. A widely-read 2026 review appears to conflate them, listing 2030 checkpoints from one instrument beside a long-term ambition belonging to an earlier one. This brief flags the inconsistency rather than asserting the correct text, because the instruments could not be read. Frontier “Green corridors show the transition is under way” is the enthusiast's version of the same slippage — a corridor shows scarce fuel can be concentrated on one route, not that it can be produced.

Frontier Finally the sceptic's overreach: “decarbonising shipping will wreck consumer prices.” The one empirical study reached here finds the inflation effect predominantly transitory. The affordability argument — which this brief takes seriously and states at full strength — belongs on capital cost, fuel supply and marginal abatement cost against other sectors, not on consumer prices. Established A closing note for the record: nearly every correction above was found by internal inconsistency inside a fetched paper, because the reference works and agency datasets that would normally settle such questions were unreachable. That is a real fact about how citations get checked in machine-assembled scholarship.