Hydrogen and E-Fuels: A Practical Reference

Alkaline electrolysis drives direct current through 25–30% potassium hydroxide between nickel electrodes separated by a porous diaphragm, at 60–80 °C. Hydroxide ions carry the charge, so the catalysts are nickel and its alloys rather than platinum-group metals, and the stack has the cheapest bill of materials of the four electrolyzer types. Industrial units have run since the 1920s, originally on Norwegian hydropower to make fertilizer, and alkaline is still roughly 70% of installed electrolyzer capacity worldwide (it was 60% in 2022, and the share rose as Chinese projects came online). Systems consume 50–55 kWh of electricity per kg of hydrogen and deliver it at 1–30 bar. The mechanism that sets the operating envelope is gas crossover: at low current the gas production rate falls but hydrogen diffusion through the diaphragm does not, so the hydrogen content of the oxygen stream climbs, and controls trip the plant when it reaches 2% by volume, half the flammability limit.

The strengths are the lowest capex per kW of any electrolyzer, the longest stack life (60,000–90,000 hours), no iridium or platinum anywhere in the cell, and a Chinese manufacturing base running at gigawatt scale. The weaknesses are dynamic: minimum load is typically 10–20% of rating, ramping takes minutes rather than seconds, output pressure is low enough that a compressor is needed downstream, and the KOH loop adds pumps, degassers, and a caustic inventory to maintain. The failure mode to plan around is the crossover trip. A stack chasing a variable wind profile does not gently turn down. It hits minimum load, accumulates hydrogen in the oxygen header, and shuts off, which is why alkaline projects usually buy firm power or add a buffer rather than bidding for the cheapest hours.

Default to alkaline for anything above a few MW where the power supply is firm or can be firmed. If you are buying tens or hundreds of MW and expect a capacity factor above roughly 50%, the capex advantage compounds and nothing else competes on cost per kilogram. Pick PEM instead if you need second-scale load following, 30 bar or higher delivery without a compressor, or a small footprint inside an existing plant. Pick solid oxide only if you have free steam on site. And if the business case depends on Western-priced equipment, price it against a landed Chinese system first, because the gap is often wider than the project's entire margin.

25–30% KOH at 60–80 °C · output 1–30 bar · 50–55 kWh/kg at the system level · stack life 60,000–90,000 hours · minimum load 10–20% with minute-scale ramping · roughly 70% of installed electrolyzer capacity · $200–500/kW installed in China against $750–1,300/kW in the West.

A kilogram of hydrogen holds 33.3 kWh on a lower-heating-value basis, so a system drawing 50–55 kWh/kg puts 61–67% of the electricity into the fuel (75–79% measured against the 39.4 kWh higher heating value, which is why vendor efficiency claims need a basis attached). The stack itself accounts for roughly 47–50 kWh/kg; rectifier losses, KOH circulation, gas drying, and water treatment take the remaining 5–10%. Output at 1–30 bar leaves compression still to be paid for downstream, roughly 3–4 kWh/kg to reach 700 bar. On an average grid at 0.35 kg CO2/kWh, that same 55 kWh/kg produces about 19 kg CO2e per kg of hydrogen, so the low-carbon tag on this entry assumes dedicated renewables, not a grid connection.

Sinopec's Kuqa plant in Xinjiang is the reference project: 260 MW of alkaline stacks on dedicated solar, designed for 20,000 t/yr of hydrogen piped to the nearby Tahe refinery. It ran near 20% of design output for its first two years, with Sinopec targeting full throughput only at the end of 2025. Baofeng's 150 MW plant in Ningxia has operated since late 2021. Chinese suppliers include Longi Hydrogen, PERIC, Sungrow Hydrogen, and Cockerill Jingli; Western ones include thyssenkrupp nucera (whose "scalum" alkaline module is rated at 20 MW), John Cockerill, Nel, Sunfire, and Green Hydrogen Systems. Hysata's capillary-fed cell is the most credible attempt to close the efficiency gap, with a claimed 41.5 kWh/kg at system level.

Two numbers drive everything. The first is equipment price: Chinese alkaline systems have sold at $200–500/kW, with recent domestic procurements reported near $200–230/kW, against $750–1,300/kW for Western systems. The second is electricity, which is 60–75% of the cost of electrolytic hydrogen; at 55 kWh/kg, every $10/MWh on the power price adds about $0.55/kg. Capacity factor matters as much as the power price, because a stack running 2,000 hours a year has to recover its capital over a quarter of the kilograms of one running 8,000 hours. Manufacturing alkaline stacks has been a hard business to make money in, since the design is old, widely copied, and priced from China. The durable positions have been in project development and in owning cheap firm power, not in the stack.

11 minLesson 3 Unit 2 Alkaline electrolysis technology Part 1ASSET - Hydrogen · 10k+ views

15 minGreen Hydrogen Systems Electrolyser TourEngineering with Rosie · 100k+ views

14 minBuild an Alkaline Water Electrolyser to Produce H2 | High OutputDominic Ó Gallachóir · 5k+ views

Hydrogen Production: Electrolysis (US Department of Energy) · Separators and Membranes for Advanced Alkaline Water Electrolysis (Chemical Reviews)

A PEM electrolyzer replaces the liquid electrolyte with a solid perfluorosulfonic-acid membrane. Protons cross the membrane and the two gas streams are separated by a sheet of polymer rather than a porous diaphragm, so the cell can hold a large pressure difference across itself. That is what lets the stack deliver hydrogen at 30–50 bar routinely and 70–80 bar in high-pressure designs while the oxygen side stays near ambient, and what lets it run anywhere from 0 to 100% of rated load in under a second without a crossover problem. The acidic environment rules out nickel: the oxygen electrode needs iridium oxide and the hydrogen electrode platinum, and the flow fields and porous transport layers have to be titanium. Systems consume 50–55 kWh/kg, close to alkaline, and stacks last 50,000–80,000 hours. PEM is about 20% of installed electrolyzer capacity, down from 30% in 2022 as Chinese alkaline built out around it.

PEM's advantages are all about behavior rather than cost: sub-second response, full turndown to zero, high current density and therefore a small footprint, and delivery pressure that removes a compression stage. The constraint that defines the technology is iridium. Stacks use roughly 0.3–0.7 g/kW, and world iridium production is about 7–8 tonnes a year, all of it a byproduct of platinum mining in South Africa, so 10 GW of PEM at today's loadings absorbs something like half of a year's supply. PEM also needs ultrapure water rather than the treated water alkaline tolerates. The failure mode is membrane thinning and pinholing under load cycling, which raises hydrogen crossover into the oxygen stream until the stack has to be rebuilt, and stack replacement is a scheduled capital event that lands two or three times in a 20-year project.

Pick PEM when the power is variable and you intend to follow it, when you need 30 bar or more at the fence without buying a compressor, or when the electrolyzer has to fit inside an existing plant. Grid-services and refueling applications are its natural home for the same reason. Do not pick it because it is "the modern one": at a high, steady capacity factor above a few MW, alkaline makes the same hydrogen for less capital, and the flexibility you paid for goes unused. If your gigawatt plan is PEM, do the iridium arithmetic before the power arithmetic, and ask the supplier what loading their current product actually uses rather than what their roadmap says. AEM is the technology aimed squarely at this weakness, but it is not a procurement option yet.

Output 30–50 bar routinely and up to 70–80 bar in high-pressure designs · sub-second response with 0–100% turndown · 50–55 kWh/kg at the system level · stack life 50,000–80,000 hours · iridium loading 0.3–0.7 g/kW against world supply near 7–8 t/yr · about 20% of installed electrolyzer capacity.

Constellation's 1.25 MW PEM unit at the Nine Mile Point nuclear station makes about 560 kg of hydrogen a day, which works out to roughly 54 kWh/kg at the plant boundary, and that is a good number to hold onto because it is measured on a fielded system rather than quoted from a datasheet. Against the fuel's 33.3 kWh/kg lower heating value that is about 62% efficient. Delivering at 30–80 bar saves the 1–2 kWh/kg that the first compression stage would otherwise cost, which is a real but small share of the total.

Nine Mile Point in New York was the first US nuclear plant to make its own hydrogen, using a 1.25 MW containerized PEM system from Nel that started up in February 2023; Davis-Besse in Ohio is the second low-temperature demonstration. Yara's 24 MW unit at Porsgrunn in Norway feeds an existing ammonia plant. Air Liquide's 20 MW plant at Bécancour in Quebec has run since 2021. The suppliers are Plug Power, Cummins (Accelera), ITM Power, Siemens Energy, Nel, and Electric Hydrogen in the West, with Chinese PEM prices falling faster than Chinese alkaline prices.

Iridium and platinum are only a few percent of stack cost today, so the metals are a scaling constraint rather than a cost problem, and every PEM maker's roadmap is built around cutting iridium loading by an order of magnitude. Western PEM systems have historically cost around EUR 1,400/kW against EUR 1,200/kW for alkaline, and the same Chinese cost pressure that hit alkaline is now arriving in PEM. The commercial question for a PEM developer is whether flexibility is worth paying for in the customer's specific power market. It usually is where hourly matching rules or curtailment discounts exist, and usually is not where the plant runs flat out against a firm PPA.

2 minHow do you make GREEN HYDROGEN? | The Secrets Behind PEM ElectrolysisITM Power · 10k+ views

4 minPEM electrolysis at Bosch | Scaling production of green hydrogenBosch Hydrogen Energy · 10k+ views

6 minUnboxing : We dismantled a PEM electrolyzer to better understand its internal structure.SENZA Hydrogen Energy And Environmental Technology · 50k+ views

Technical Targets for Proton Exchange Membrane Electrolysis (US Department of Energy) · Proton Exchange Membrane (PEM) Water Electrolysis: Cell-Level Considerations for Gigawatt-Scale Deployment (Chemical Reviews)

AEM electrolysis puts PEM's architecture on alkaline's bill of materials. A solid anion-exchange membrane carries hydroxide ions instead of protons, which makes the cell alkaline, and an alkaline cell can use nickel-based catalysts rather than iridium and platinum. The electrolyte is dilute KOH, typically around 1%, or in some designs pure water, so the caustic inventory and the pumps and degassers that go with a 30% KOH loop mostly disappear. Because the membrane is solid, the cell holds a pressure difference the way a PEM cell does: Enapter's modules deliver up to 35 barg with the oxygen side near ambient. The physics is settled and the cell performance is closing on PEM. What is not settled is how long the membrane lasts, because the polymer's cationic groups sit in hot alkali, which attacks them, and that chemical degradation is the reason this entry sits at pilot rather than commercial.

The strength is the materials list. No iridium, no platinum, no titanium flow fields, and no concentrated caustic, while keeping high delivery pressure and fast response. If AEM works at scale it removes the one constraint that caps PEM's growth. The weakness is durability, and it is not a small gap. NREL's baseline pure-water-fed AEM cells degraded roughly 0.67 mV per hour after break-in, while a 40,000-hour stack needs something closer to 2 microvolts per hour, three orders of magnitude better. The failure mode is chemical: hydroxide attacks the quaternary ammonium groups that make the membrane conductive, conductivity falls, cell voltage rises, and efficiency drifts down until the stack is not worth running. Thermal and mechanical degradation add to it, and running hotter to raise performance makes the chemical attack faster.

Use AEM today for pilots, for hydrogen at a few hundred kW where module-level redundancy is worth more than efficiency, and for research where you want PEM-like behavior without buying iridium. Do not specify it above a few MW for a project that has to run for 20 years, because the fleet hours to underwrite that do not exist yet. If you are evaluating a supplier, ask for a published multi-thousand-hour degradation curve at the current density and temperature you intend to run, not a cell-level efficiency claim measured at beginning of life, and ask what a stack module costs to replace and how often. For a firm-power project of any real size, alkaline is the cheaper answer; for a variable-power project that has to be bankable now, PEM is the one with the operating history.

Dilute KOH (roughly 1%) or pure water, nickel-based catalysts, no platinum-group metals · outlet pressure up to 35 barg · 51.3 kWh/kg at system level on Enapter's MW-class AEM Nexus 1000 · 2.4 kW stack modules, 420 of them in a 921.5 kW system · baseline cell degradation near 0.67 mV/h against roughly 2 µV/h for a 40,000-hour life.

Enapter quotes 51.3 kWh/kg for the AEM Nexus 1000 including all in-system utilities, which is 65% of the 33.3 kWh/kg lower heating value of hydrogen (77% on the 39.4 kWh higher heating value). That puts AEM roughly level with alkaline and PEM at 50–55 kWh/kg, so efficiency is not the reason to pick it. Delivering at up to 35 barg saves the 1–2 kWh/kg the first compression stage would cost, and the dilute electrolyte loop draws less parasitic power than a 30% KOH circuit. The number that moves over a stack's life is the degradation rate: at 0.67 mV/h on a cell running near 1.9 V, energy per kilogram climbs a few percent within a thousand hours.

Enapter is the only supplier shipping AEM in volume. Its approach is mass-produced 2.4 kW modules rather than one large stack: the EL 4 makes 500 NL/h at 99.9% purity, and the AEM Nexus 1000 packs 420 modules around a common balance of plant for 921.5 kW. Protium's installation in the UK is among the larger fielded arrays. On materials, Ionomr Innovations sells Aemion and Aemion+ membranes, Versogen sells its PiperION line, and Dioxide Materials' Sustainion is used in CO2 electrolysis as well as water. Alchemr is a US developer working on the stack itself.

The cost argument for AEM is a bill-of-materials argument: strip out iridium, platinum, and titanium, and the stack should eventually land nearer alkaline's cost than PEM's while behaving like PEM. Nothing on the market demonstrates that yet, because the volumes are small and modular manufacturing carries its own overhead. If you are building a business here, the value is in the membrane and ionomer rather than the system, since a durable membrane is the scarce part and every stack builder needs one. For a buyer, the honest position is that AEM is not yet a procurement option above a few MW, and the thing that would change that is a published multi-thousand-hour degradation curve from a fielded unit, not another announcement of a record current density.

2 minAEM Electrolysis Technology | EvonikEvonik Industries · 10k+ views

2 minEnapter AEM electrolyser EL 2.1Enapter · 10k+ views

4 minGreen Hydrogen using anion exchange membrane (AEM) electrolysis of waterSolFinder Research · 5k+ views

Recent Advances and Challenges in Anion Exchange Membranes Development/Application for Water Electrolysis: A Review (Membranes) · Elucidating the Degradation Behavior of a 25 cm2 Pure-Water-Fed Non-Precious Metal Anion Exchange Membrane Water Electrolyzer Cell (Small)

Solid oxide electrolysis splits steam rather than liquid water, at 700–850 °C, in a ceramic cell where oxide ions move through a zirconia electrolyte. Heat does part of the work that electricity does in a low-temperature cell. At 800 °C, the thermodynamics call for roughly 25 kWh/kg of electricity and 9 kWh/kg of heat, against 39.4 kWh/kg of electricity alone at room temperature, and real systems land at 37–42 kWh/kg of electricity where alkaline and PEM need 50–55. The same cell run backwards is a solid oxide fuel cell, and the same stack will co-electrolyze steam and CO2 into syngas in one step, which is why this technology shows up inside e-fuel plants rather than as a standalone hydrogen source. The catch is in the word "heat": the electricity saving is real only if the steam comes from somewhere that was going to reject heat anyway.

The strength is the electricity number, and electricity is 60–75% of the cost of electrolytic hydrogen, so 40 kWh/kg instead of 52 is worth roughly a quarter of the operating cost. Co-electrolysis is the second strength: going straight to syngas removes a separate reverse water-gas-shift step in a methanol or Fischer-Tropsch plant. The weaknesses all come from running a ceramic stack hot. Cell life is measured today in the 20,000–40,000 hour range against 60,000–90,000 for alkaline, degradation accelerates with temperature, and the stack needs to sit at steady load and steady temperature. The failure mode is thermal cycling: ceramic layers with mismatched expansion coefficients crack and seals leak when a stack is heated and cooled, so every unplanned shutdown costs life. A SOEC plant that follows a wind profile is being damaged by design.

Pick solid oxide when you have waste heat or steam at temperature next to the electrolyzer and a load you can run flat out. Ammonia synthesis, methanol synthesis, and Fischer-Tropsch loops all reject heat, and a nuclear plant has steam by definition, which is why the US demonstrations sit at power stations. If you want syngas rather than pure hydrogen, co-electrolysis is a strong reason to choose it even before the efficiency argument. Do not pick it for a project running on variable renewables with no heat source, because you will pay for the ceramic and get none of the benefit: raise the steam with an electric boiler and total consumption comes back to roughly 47–50 kWh/kg, which is what a cheaper alkaline plant does with a longer stack life. If flexibility is what you need, buy PEM.

700–850 °C operating temperature · 37–42 kWh/kg of electricity, plus roughly 9–10 kWh/kg delivered as heat · Sunfire's 720 kW GrInHy2.0 unit at 200 Nm³/h works out near 40 kWh/kg AC · stack life 20,000–40,000 hours against 60,000–90,000 for alkaline · Topsoe's Herning factory rated at 500 MW of stacks a year.

Splitting steam at 800 °C needs about 25 kWh/kg as electricity and 9 kWh/kg as heat, so a system that is handed steam consumes 37–42 kWh/kg of electricity, or 79–90% of hydrogen's 33.3 kWh/kg lower heating value counting electricity alone. GrInHy2.0 at Salzgitter is the fielded check on that: 720 kW AC for 200 Nm³/h is about 40 kWh/kg. If the steam has to be raised electrically, add roughly 9–10 kWh/kg and the system lands near 47–50 kWh/kg, which is most of the advantage over alkaline gone. Degradation eats the rest slowly, with the best published large-system tests holding under 0.2% efficiency loss per 1,000 hours.

Sunfire delivered the 720 kW GrInHy2.0 unit to Salzgitter Flachstahl, where it ran at 850 °C on steel-mill steam and made close to 100 t of hydrogen before the EU project closed, and has since installed a 2.6 MW SOEC at Neste's biorefinery in Rotterdam under MultiPLHY. Topsoe inaugurated Europe's largest SOEC stack factory at Herning in Denmark in 2025, rated at 500 MW a year, and sells the same cell for co-electrolysis into e-fuel plants. On the nuclear side, Bloom Energy and Idaho National Laboratory tested high-temperature steam electrolysis at INL, and Bloom's system is the one Xcel Energy is demonstrating at the Prairie Island station in Minnesota under a $12 million DOE grant, using plant steam that Xcel expects to be up to 30% more efficient than cold electrolysis.

Capital cost per kW is higher than alkaline and the stack lasts a third to a half as long, so SOEC only pays where the electricity saving is large and the plant runs at a high capacity factor. The arithmetic is straightforward: saving 13–18 kWh/kg at $40/MWh is $0.50–0.70/kg of operating cost, which has to cover a shorter stack life and more frequent replacement. That pencils at 8,000 hours a year next to free steam and does not pencil at 3,000 hours on curtailed power. Supply is concentrated in a few vendors, mainly Topsoe, Sunfire, and Bloom Energy, with FuelCell Energy in the US, and the manufacturing base is small enough that lead times matter as much as price. For anyone building on it, the durable position is the integrated plant, not the stack: SOEC is worth the most to whoever also owns the heat and the downstream synthesis loop.

3 minSOEC - THE green hydrogen solutionTopsoe · 10k+ views

3 minLecture 5: Green Hydrogen Production Process, What is Solid Oxide (SOEC) Electrolysis?allaboutsustainability · 10k+ views

5 minGrInHy2.0 - The World's Largest High-Temperature ElectrolyzerSunfire · 5k+ views

Technical Targets for High Temperature Electrolysis (US Department of Energy) · Degradation of solid oxide electrolysis cells: Phenomena, mechanisms, and emerging mitigation strategies—A review (OSTI.GOV)

Steam methane reforming passes natural gas and steam over a nickel catalyst in externally fired tubes at 700–1,000 °C and 15–30 bar, producing a syngas of hydrogen and carbon monoxide. The gas then goes through one or two water-gas shift reactors, where the carbon monoxide reacts with more steam to make additional hydrogen and CO2, and finally through pressure swing adsorption, which pulls off 99.9%+ pure hydrogen and sends the PSA tail gas back to the reformer furnace as fuel. Two carbon streams leave the plant: the process CO2 in the shifted syngas, which is concentrated and at pressure, and the furnace flue gas, which is dilute and at atmospheric pressure. That split is the single most important fact about the technology, because it is what determines the cost of capturing anything. Roughly two-thirds of the world's hydrogen comes from unabated natural gas this way, most of it made inside the refinery or ammonia plant that consumes it.

This is a hundred-year-old process built by a handful of engineering contractors on catalysts sold by a handful of suppliers, in single trains of up to 200,000 Nm³/h. It is reliable, well-understood, and cheap, and it is the reason low-carbon hydrogen has such a hard time selling. The weakness is carbon: about 9 kg of direct CO2 per kg of hydrogen, and 10–12 kg CO2e once upstream methane production and transport are counted. Perfect stoichiometry would give 5.5 kg, so roughly 40% of the emissions come from burning gas to heat the tubes and from incomplete conversion. The other weakness is exposure. Gas is 60–70% of the cost of the hydrogen, so the plant has no cost floor of its own; a European reformer and a US Gulf Coast reformer are the same equipment making hydrogen at very different prices.

If you need hydrogen at industrial scale today and carbon is not priced or regulated in your market, this is the answer, and every other entry in this sheet has to beat it at your gate. Build or buy SMR when gas is cheap and firm, when you need 99.9%+ purity in bulk, and when the plant can run near-continuously, which it must: reformers are slow to start and are not load-following equipment. Do not build unabated SMR into a project that has to sell into the EU after RFNBO and CBAM rules bite, or into a US project counting on 45V, because at 9–12 kg CO2e/kg it does not qualify for anything. If you are in that position, price entry 006 for the cheap partial retrofit and entry 007 for the high-capture new build, and expect to pay more per kilogram for both.

700–1,000 °C over nickel catalyst at 15–30 bar · PSA product at 99.9%+ purity · roughly 0.18 MMBtu of natural gas per kg of hydrogen, feedstock plus furnace fuel · about 9 kg direct CO2 per kg H2 and 10–12 kg CO2e including upstream methane · roughly two-thirds of world hydrogen supply · $1–2/kg at $3/MMBtu gas and $2.5–3.5/kg at $10/MMBtu.

A reformer consumes roughly 0.18 MMBtu, or about 53 kWh, of natural gas per kg of hydrogen, counting both the gas that becomes hydrogen and the gas burned in the furnace. Hydrogen's higher heating value is 39.4 kWh/kg, so about 75% of the gas energy ends up in the product on a like-for-like basis, or about 63% measured against the fuel's 33.3 kWh/kg lower heating value. Roughly 30–35% of the gas is burned as fuel rather than reformed, and that is the share whose carbon leaves in the dilute flue-gas stream. The PSA gives up 10–15% of the hydrogen to the tail gas, but that hydrogen is burned in the furnace rather than lost, which is why plant efficiency stays as high as it does.

Almost every refinery hydroprocessing unit and almost every ammonia plant has a reformer attached, and captive production of this kind dwarfs the merchant market. The merchant business is dominated by Air Liquide, Air Products, and Linde, which build reformers over the fence from a customer and sell hydrogen by pipeline under long-term contracts; Air Products' Gulf Coast network, more than 700 miles of pipeline from Galveston Bay to New Orleans fed by about 25 production plants, is the largest of these systems. The technology licensors and catalyst suppliers are Topsoe, Johnson Matthey, KBR, Technip Energies, and Linde Engineering. Small packaged reformers from HyGear and Mahler AGS serve on-site industrial users at a few hundred Nm³/h.

The cost structure is simple enough to do in your head. Gas is 60–70% of the total at typical prices, and at 0.18 MMBtu/kg every $1/MMBtu on the gas price adds about $0.18/kg of hydrogen. At $3/MMBtu that puts hydrogen near $1–2/kg; at $10/MMBtu it lands at $2.5–3.5/kg. Capital cost is a few hundred dollars per annual tonne of capacity and matters most at low gas prices, when it is the larger share. The consequence for anyone building a low-carbon hydrogen business is that the number to beat is not fixed: the same electrolytic hydrogen that looks expensive against a US reformer can look competitive against a European one, and a project's whole margin can turn on which gas market the customer sits in. Any business case that quotes "gray hydrogen costs $X" without naming a gas price is not a business case.

2 minH2 academy: hydrogen plant flowsheet overviewJohnson Matthey · 10k+ views

2 minHydrogen generation by steam reforming (Mahler AGS GmbH)Mahler AGS GmbH · 100k+ views

10 minHow to produce hydrogen efficiently? (PART 1) Discovering the fuel of the future - steam reformingJAES Company · 100k+ views

Hydrogen Production: Natural Gas Reforming (US Department of Energy) · Comparison of Commercial, State-of-the-Art, Fossil-Based Hydrogen Production Technologies (NETL)

Bolting carbon capture onto a steam methane reformer means choosing which of the plant's two CO2 streams to treat. The shifted syngas leaving the water-gas shift reactors is 15–20% CO2 at 20–30 bar, and the PSA tail gas is more concentrated still, so an amine or vacuum-swing unit on either stream recovers 90%+ of the carbon in it at low cost, because concentration and partial pressure are what set capture cost. The reformer furnace's flue gas is different: a few percent CO2, at atmospheric pressure, in a large volume of nitrogen. Capturing that too is what takes a plant from roughly half its carbon to 90%, and it roughly doubles the cost of capture per tonne. Every operating SMR retrofit in the world took the cheap half. That is why the technology's reputation and its published capture rates disagree so often: the equipment does hit 90%+ on the stream it treats, and the plant does not.

The strength is that it is a retrofit on proven equipment with a known cost, using amine or adsorption units that chemical plants have run for decades, and it is the cheapest low-carbon hydrogen available at scale today. The weakness is the ceiling. Because the furnace flue gas carries roughly a third of a reformer's CO2, a process-stream-only retrofit leaves 4–5 kg of direct CO2 per kg of hydrogen in the air, and once upstream methane production and transport are counted most operating plants land at 4–6 kg CO2e/kg. The failure mode is regulatory rather than mechanical: at 4–6 kg the hydrogen fails the EU's RFNBO ceiling of 3.38 and the US 45V cutoff of 4, so a plant built to capture "up to 90%" can be fully operational, meeting its own design spec, and still qualify for nothing. Capture also costs steam and power, which pushes gas consumption per kg up by a few percent.

Choose an SMR retrofit when you already own a reformer, you have a CO2 offtake or storage site within reach, and the objective is a real emissions reduction at the lowest dollars per tonne. It is the best value in that framing and nothing else comes close. Do not choose it when the objective is a certificate. If your project needs to clear 3.38 kg CO2e/kg in Europe or 4 kg for 45V in the US, run the numbers on the flue gas capture as well before you commit, and if the answer is that you need 90%+ plant-wide, build an autothermal reformer instead (entry 007), because a purpose-built ATR gets there with one capture unit rather than two. As a rule of thumb, retrofit for tonnes and build new for qualification.

Shifted syngas at 15–20% CO2 and 20–30 bar against a few percent in flue gas at 1 bar · 90%+ recovery from the process stream, roughly half to 60% of plant CO2 · most operating plants land at 4–6 kg CO2e/kg H2 · fails the EU's 3.38 ceiling and the US 45V cutoff of 4 · roughly $0.3–0.8/kg over unabated hydrogen.

Capture does not change the reforming itself: the plant still burns roughly 0.18 MMBtu of gas per kg of hydrogen. What it adds is regeneration steam and CO2 compression, typically a few percent on gas consumption and 0.1–0.2 kWh of electricity per kg of CO2 compressed to pipeline pressure, so the delivered hydrogen costs slightly more energy per kg than the unabated case. The carbon accounting is where the losses actually show. Starting from about 9 kg direct CO2 per kg of hydrogen, capturing the process stream removes 4–5 kg and leaves 4–5 kg going up the stack, and upstream methane emissions add another 1–2 kg on a typical gas supply, which is how a plant with a "90% capture" unit reports 4–6 kg CO2e/kg.

Shell's Quest at the Scotford upgrader in Alberta is the longest-running reference: an amine unit on the reformers' process stream, designed for at least 1 Mt CO2/yr, which is about 80% of the hydrogen units' emissions and about 35% of the whole upgrader's. Between 2015 and 2019 it captured 5 Mt of the 7.5 Mt those reformers produced, an average of 67%. Air Products' Port Arthur project retrofitted two SMRs with vacuum swing adsorption on the process gas in 2012–2013, capturing about 1 Mt/yr at 95% purity and over 90% recovery from that stream, with the CO2 delivered to Denbury for enhanced oil recovery; the furnace flue gas is not captured. Both are the standard the technology is judged against, and both are process-stream-only designs.

Capture adds roughly $0.3–0.8/kg to hydrogen made this way, and the spread inside that range is almost entirely about how much of the flue gas you take and what CO2 transport and storage costs in your basin. Published avoidance costs for SMR with conventional capture sit around $60–75 per tonne of CO2, which is why the 45Q credit and the EU carbon price are the variables projects are actually underwriting against rather than the hydrogen price. The commercial risk is that the qualification thresholds move faster than the plant can be modified: a retrofit is a 20-year asset priced against rules that have already tightened twice. If you are financing one, the question to answer first is what the plant does if the threshold drops to 2 kg, and the honest answer for a process-stream-only design is that it becomes an ordinary reformer with an expensive amine unit attached.

4 minHow is blue hydrogen produced?Future Energy & Technology · 5k+ views

2 minCarbon Capture & Storage - How It WorksShell · 100k+ views

2 minCarbon capture and storage research at the British Geological SurveyBritish Geological Survey · 50k+ views

Quest Carbon Capture and Storage project: annual report, 2024 (Government of Alberta) · Techno-Economic Evaluation of SMR Based Standalone (Merchant) Hydrogen Plant with CCS (IEAGHG)

Autothermal reforming is the reformer you build when the capture rate is the product. Instead of heating catalyst tubes from outside with a gas-fired furnace, an ATR burns part of the feed with pure oxygen inside a single refractory-lined vessel, and the heat released there drives the reforming reaction over a catalyst bed in the same vessel. Removing the furnace removes the dilute flue-gas stream that caps what an SMR retrofit can capture (entry 006). All the carbon leaves in one place, in the shifted syngas, concentrated and at 30 bar or more, where a single amine or physical-solvent unit can take 90–95% of it and licensors quote 98–99%. The price of that is an air separation unit: a typical design runs at an oxygen-to-carbon ratio near 0.61, and cryogenic oxygen costs about 0.35 kWh per kg, so the plant carries an electrical load an SMR does not.

The strength is architectural. One concentrated CO2 stream means one capture unit, high recovery, and a plant-level carbon intensity of 1–3 kg CO2e/kg where the gas supply is clean, against 4–6 for a retrofit SMR. The vessel itself is not exotic; ATRs are standard equipment in large methanol and gas-to-liquids plants. The weaknesses are the balance of plant and the supply chain. An ASU is a large capital item with a real parasitic load, roughly 1.2 kWh of electricity per kg of hydrogen against about 0.13 for an SMR, and CO2 transport and storage has to be contracted before the plant is worth building. The failure mode is not in the plant at all. Upstream methane leakage is counted in the hydrogen's carbon intensity, and at about 3.2 kg of natural gas per kg of hydrogen, 1% leakage adds roughly 0.9 kg CO2e/kg on a 100-year global warming potential and closer to 2.5 kg on a 20-year one. A 95%-capture plant on a leaky gas field can miss the 4 kg threshold on emissions it never touched.

Build an ATR when the hydrogen has to qualify, not merely be cleaner: a new-build plant aiming under 4 kg CO2e/kg for 45V, or under 3.38 for the EU's RFNBO rules, needs plant-wide capture in the 90s, and only a design with no fired furnace gets there with one capture unit. It also suits large single trains, since ATR scales better in one vessel than a tube furnace does. Do not choose it to decarbonize a reformer you already own, because the retrofit in entry 006 gets you tonnes for a fraction of the capital. Before committing, get measured methane intensity from the specific gas supply, not a national average, and get the CO2 storage contract signed, because those two items decide the carbon number and neither is inside the fence. If you cannot get both, an ATR is an expensive way to build an ordinary reformer.

Oxygen-to-carbon ratio near 0.61, steam-to-carbon near 1.37 · 90–95% plant-wide capture, with licensors quoting 98–99% · 1–3 kg CO2e/kg on a clean gas supply against 4–6 for a retrofit SMR · about 3.2 kg of natural gas per kg of hydrogen · ASU power near 0.35 kWh per kg of O2, giving roughly 1.2 kWh/kg of hydrogen against 0.13 for SMR.

Gas consumption is close to SMR and slightly better on a process basis, around 3.2 kg of natural gas per kg of hydrogen, because burning feed inside the vessel wastes less than firing a furnace from outside. The difference is electrical: the ASU takes roughly 1.2 kWh/kg of hydrogen against about 0.13 kWh/kg for a reformer, a ninefold increase, and CO2 compression adds roughly another 0.1 kWh per kg of CO2 sent to the pipeline. On carbon, a plant starting near 9 kg of CO2 per kg of hydrogen and capturing 95% keeps about 0.5 kg in the stack gas, so the delivered number is set almost entirely by the upstream gas supply rather than by the plant.

The technology is licensed by Topsoe as SynCOR, by Johnson Matthey as LCH (a gas-heated reformer feeding an ATR), and by Air Liquide and Technip Energies in their blue hydrogen packages. The project record is where this entry gets uncomfortable. ExxonMobil's Baytown plant, roughly 1 billion cubic feet a day of hydrogen with 7 Mt/yr of CO2 capture using SynCOR and a stated capture rate above 98%, went through FEED and was then paused indefinitely in 2026 for lack of offtake. bp's 1.2 GW H2Teesside, which had selected Johnson Matthey's LCH, was abandoned in December 2025 and the site redirected to a data center. Air Products canceled its $4.5 billion Louisiana complex in June 2026 and took a pre-tax charge of up to $2.9 billion. As of mid-2026 there is no large ATR-plus-capture hydrogen plant operating in the West.

On paper ATR is the cheaper way to buy deep decarbonization: capture cost per tonne is lower than an SMR retrofit that goes after the flue gas, because the stream is concentrated and pressurized, and the capture rate is roughly double. The problem is that the whole plant is capital, not a bolt-on, so it needs 15 to 20 years of contracted offtake at a premium over gray hydrogen, and that offtake has not appeared. Three flagship projects were canceled or paused between December 2025 and June 2026, in each case citing weak demand rather than technology risk. If you are building a business here, treat the binding constraint as the offtake contract and the CO2 storage agreement, and note that a project's carbon intensity is set by a gas supply chain the developer does not own. The technology works; the commercial case is the part that has not been demonstrated.

2 minATR: how does it work? - Air LiquideAir Liquide · 1k+ views

4 minHow an Air Separation Plant Works | Air ProductsAir Products · 50k+ views

Technological evolution of large-scale blue hydrogen production toward the U.S. Hydrogen Energy Earthshot (Nature Communications) · Guidelines to Determine Well-to-Gate Greenhouse Gas (GHG) Emissions of Hydrogen Production Pathways using 45VH2-GREET (US Department of Energy)

Gasification burns coal or petroleum coke with a controlled shortage of oxygen at roughly 1,300–1,500 °C, so the carbon comes out as carbon monoxide rather than CO2. The resulting syngas goes through the same water-gas shift and pressure swing adsorption steps a reformer uses, plus acid gas removal to strip out the sulfur compounds that coal brings with it and a slag handling system for the ash. Entrained-flow gasifiers, fed either as a water slurry or as dry pulverized coal, are the standard design at industrial scale, and all of them are oxygen-blown, so a coal-to-hydrogen plant carries an air separation unit the way an ATR does. About 20% of the world's hydrogen is made this way and almost all of that is in China, where roughly two-thirds of hydrogen comes from coal because domestic gas is scarce and coal is not. Petcoke gasification is the same process applied to a refinery's own bottom-of-the-barrel residue.

The strength is feedstock cost and feedstock security. Coal is cheap, it is priced locally, and a country with coal and no gas can make its own ammonia and methanol without importing anything, which is the entire logic of China's coal chemical industry and of India's gasification push. The plants are large, mature, and run at high capacity factors. The weakness is carbon, and it is not close: roughly 16–21 kg of direct CO2 per kg of hydrogen, and 18–24 kg CO2e once coal mining and transport are counted, against 9–12 for unabated reforming. The plants are also capital-heavy, water-hungry, and produce sulfur, slag, and wastewater that a reformer does not. The failure mode for anyone competing with this route is commercial rather than technical: a Chinese ammonia or methanol plant fed by coal sets a world price that green projects have to match, and it will not stop doing so because a European buyer prefers a lower carbon number.

You choose coal gasification when you have coal, no gas, and a domestic chemical industry to feed, and when carbon is not priced in the market you sell into. Petcoke gasification makes sense inside a refinery that is already stuck with the coke and needs both hydrogen and steam. For anyone outside those situations, this is not a route to build; it is the benchmark to price against. If you are modeling a green ammonia or green methanol project for export, put a coal-based Chinese plant on the other side of the comparison rather than a European reformer, because that is the marginal supplier your customer can actually buy from. The mechanism that decides your fate is border carbon pricing: CBAM and similar rules exist precisely to close the gap between an 18–24 kg CO2e product and a sub-1 kg one, and if they are weakened your project's margin goes with them.

Entrained-flow gasification at roughly 1,300–1,500 °C, oxygen-blown · about 20% of world hydrogen supply and roughly two-thirds of China's · 16–21 kg direct CO2 per kg H2, 18–24 kg CO2e life-cycle · roughly 8–13 kg of coal per kg of hydrogen depending on rank · $1.3–2/kg in China, near 10 RMB/kg on average and about 8.5 RMB/kg in the northwest.

Coal's lower heating value is around 25 MJ/kg and hydrogen's is 120 MJ/kg, so at the 50–60% thermal efficiency these plants achieve, roughly 8–13 kg of coal go in per kg of hydrogen out, with lower-rank coals at the top of that range. About two-thirds of that coal is carbon, which is where the 16–21 kg of CO2 per kg of hydrogen comes from; the number is set by the feedstock, not by how well the plant is run. On top of the coal, the air separation unit and the acid gas removal, shift, and PSA sections draw electricity and steam, and one published cogeneration case landed at 48.7% efficiency on a higher-heating-value basis, rising to 52.9% when the oxygen was bought over the fence instead of made on site.

China's coal chemical industry is the reference: Ningxia Baofeng's coal-to-olefins complexes and the large Shenhua coal-to-liquids plants run gasification trains at a scale nothing outside the country matches. Gasifier technology comes from Shell, Air Products (which acquired both the GE/Texaco water-slurry and Siemens dry-feed lines), thyssenkrupp Uhde's PRENFLO, ECUST's opposed multi-burner design, and China Aerospace's HTL. Petcoke gasification runs inside refineries, with the Texaco process the long-standing reference design. India is trying to build a version of the Chinese industry from scratch: the National Coal Gasification Mission targets 100 Mt of coal gasified by 2030, about 22.6 Mtpa is operating or under construction, and in May 2026 the cabinet approved a ₹37,500 crore incentive scheme paying up to 20% of plant and machinery cost, aimed at cutting imports of urea, ammonia, and methanol.

This is the cheapest hydrogen made at scale, at roughly $1.3–2/kg in China and near the bottom of that range where coal is cheapest, and the cost structure is coal price plus a large capital charge rather than the almost-pure feedstock exposure of a reformer. Adding CCUS raises Chinese coal hydrogen by roughly 14%, from about 10 to 11.5 RMB/kg, which is a much smaller premium than most Western capture projects carry because the syngas stream is concentrated and the plant already has acid gas removal. That is worth knowing: the cheapest large-scale route to genuinely low-carbon hydrogen may be a Chinese coal gasifier with capture, not a Western reformer. For anyone building a low-carbon business, the practical implication is that the floor price for ammonia, methanol, and urea is set by this route, and it moves with coal prices and Chinese policy rather than with anything a hydrogen developer controls.

3 minGasification AnimationNational Energy Technology Laboratory · 500k+ views

7 minTexaco Pet Coke Gasification Process Video by Planned ReactionPlanned Reaction · 10k+ views

Life-cycle analysis of greenhouse gas emissions of hydrogen, and recommendations for China (ICCT) · Deploying green hydrogen to decarbonize China's coal chemical sector (Nature Communications)

Biomass gasification is the same reactor physics as coal gasification with the opposite carbon accounting. Wood waste, agricultural residue, or sorted municipal waste is partially oxidized with oxygen or steam, the syngas is cleaned of tar and shifted, and PSA takes off the hydrogen. Because the carbon in the feedstock came out of the atmosphere within the last few decades, capturing the CO2 from the shift step and injecting it makes the hydrogen carbon-negative rather than merely low-carbon, and no other production route in this sheet can do that. Yield runs around 100 kg of hydrogen per tonne of dry biomass, with the wider literature spanning 40–190 kg/t across thermochemical routes and feedstocks, at 40–70% energy efficiency on a lower-heating-value basis. Fixed-bed, fluidized-bed, and two-stage torrefaction designs all appear in current projects, and the sub-processes are individually mature; the integrated hydrogen chain sits at TRL 5–7.

Carbon-negative hydrogen is the strength, and the numbers are real rather than notional: Mote's first-of-a-kind US plant is designed to process at least 300,000 tonnes of wood waste a year, make more than 20,000 tonnes of hydrogen, and sequester more than 450,000 tonnes of CO2, which is over 20 kg of CO2 removed per kg of hydrogen sold. The weakness is feedstock logistics, and it is a hard cap rather than a cost line. Biomass has low energy density, so it has to be collected from a radius around the plant, and the delivered price rises with that radius until the project stops working. That is why announced plants are small: Haffner Energy's European units are sized at 720–1,000 kg of hydrogen a day, and the largest developments run to about 100 MW of gasification input. The failure mode is contracted feedstock. A plant sized on regional biomass availability and financed on a 20-year basis competes for the same residues as pellet mills and biomass power stations, and a plant without a long-term feedstock contract is a plant that will eventually run at partial load.

Choose biomass gasification when you have a concentrated, contracted, low-value residue stream and a CO2 storage option, and when a buyer will pay for carbon removal on top of the hydrogen. Waste wood, sawmill residue, demolition timber, and pulp mill black liquor are the good cases, and black liquor is the best of them because it is already collected and pumped inside the mill, which removes the transport problem entirely. Sorted municipal waste works too and yields more hydrogen per tonne when it contains plastics, since those carry more hydrogen than wood. Do not plan on this route to supply an industrial hydrogen demand of any size, because the collection radius will not support it; electrolysis and reforming both scale in a way this does not. As a rule of thumb, size the plant from the feedstock you can contract at a known price, then see what hydrogen that buys, rather than the other way round.

Around 100 kg of hydrogen per tonne of dry biomass, with a 40–190 kg/t span across processes and feedstocks · 40–70% energy efficiency on an LHV basis · TRL 5–7 for the integrated chain · roughly €2.7–4/kg at a 200 MW plant with biomass at €20/MWh, and €2.2–3.5/kg with a €50/t CO2 credit · Mote's plant designed for over 20 kg of CO2 sequestered per kg of hydrogen.

A tonne of dry biomass holds roughly 18–19 GJ on a lower-heating-value basis and yields about 100 kg of hydrogen, which carries 12 GJ, so the plant converts around two-thirds of the feedstock energy into fuel in the best modeled cases and 40–50% in the weaker ones; drying, tar cracking, and the oxygen plant take most of the difference. Reported plant-level efficiencies bracket that widely, from 35% to 75%, mainly because studies define the product and the boundary differently. As a working feedstock ratio, one demonstration project reports 12–15 kg of wood per kg of pure hydrogen. Adding carbon capture costs energy but qualifies for the negative-emissions credit that makes the economics work, and the CO2 stream from the shift step is concentrated, so capturing it is cheap compared with capturing a reformer's flue gas.

Mote in the US is the largest project of its kind, an oxygen-blown fluidized-bed gasifier with integrated capture and geological sequestration, sized at 300,000 t/yr of wood waste and about 21,000 t/yr of carbon-negative hydrogen. Torrgas in the Netherlands licenses a two-stage torrefaction-plus-gasification design, with a 10 MW demonstration planned at Brightlands Chemelot and a 100 MWth biohydrogen plant under development by its HyCarb subsidiary on demolition wood and prunings, targeting about 15,000 t/yr of hydrogen. Haffner Energy is building three smaller European plants, at Gloveller in Switzerland, Alkmaar in the Netherlands, and Montbéliard in France, at 720 to over 1,000 kg of hydrogen a day. Cortus Energy in Sweden sells the WoodRoll process for low-grade biomass, and A.H.T. Syngas Technology and BtX energy have built a containerized waste-wood unit.

Feedstock price drives everything, the same way gas price drives a reformer. At €20/MWh biomass, a 200 MW plant lands near €4/kg of hydrogen, with the plausible range €2.7–4/kg; expected capital and efficiency improvements of 10–20% would take it under €3/kg, and a €50/t credit for the sequestered CO2 brings it to €2.2–3.5/kg. That last figure is the one to focus on, because it is competitive with European SMR, which averaged €3.8/kg in 2023 and spiked to €5.7/kg in 2022. In other words, this route is priced against European gas rather than against US gas, and it only clears the bar when the CO2 removal has a buyer. If you are building a business on it, the carbon removal credit is the product and the hydrogen is close to a co-product, so contract the removal first. Small scale also means small absolute volumes, so this is a niche supply business rather than a commodity one.

3 minThermochemical Conversion of Biomass to Biofuels via GasificationBioenergyKDFChannel · 500k+ views

3 minLecture 6: Green Hydrogen Production, Biomass Gasificationallaboutsustainability · 5k+ views

5 minHydrogen from biomassFuture Energy & Technology · 5k+ views

Biomass gasification for hydrogen production (IEA Bioenergy Task 33) · A Comprehensive Review on Hydrogen Production from Biomass Gasification (Molecules)

Methane pyrolysis heats natural gas without oxygen or steam, so the methane splits into hydrogen gas and solid carbon instead of hydrogen and CO2. There is no process CO2 at all, which is a different proposition from capturing it: nothing has to be compressed, transported, or injected, and there is no storage site to permit. The thermodynamics are the attraction. Splitting methane takes 37.4 kJ per mol of hydrogen, about 5.2 kWh/kg in theory, against 286 kJ/mol or 39.4 kWh/kg for splitting water. Three reactor families are being commercialized: plasma torches at 600–1,200 °C, molten metal and molten salt bubble columns at 900–1,400 °C, and thermo-catalytic fluidized or moving beds at 600–1,000 °C using iron, nickel, or carbon catalysts. The reaction also fixes the plant's second product, because every kg of hydrogen comes with 3 kg of solid carbon that has to go somewhere.

The strength is the energy number and the absence of a CO2 stream: a moving-bed pilot has run at 9.7–11 kWh/kg, roughly a fifth of what an electrolyzer needs, and the plant needs no pipeline or reservoir. The first weakness is that pyrolysis uses more gas than reforming, not less. The stoichiometry gives 2 mol of hydrogen per mol of methane where reforming gets 4, because reforming pulls the extra hydrogen out of steam, so pyrolysis burns roughly 4 kg of methane per kg of hydrogen against about 3.2 for a reformer. The second is the reactor: arc plasma designs report 25–39 kWh/kg, which throws away most of the thermodynamic advantage, and the choice of reactor decides whether the technology is interesting at all. The third weakness is the co-product, and it is the one that caps the whole route. Carbon has to be sold or landfilled, and solid carbon accumulating on reactor walls and catalyst surfaces is also the main mechanical failure mode, because it fouls heat transfer and plugs the bed until the reactor has to be cleaned out.

Consider methane pyrolysis where gas is cheap, electricity is expensive or scarce, and there is no CO2 storage within reach, which describes a lot of industrial sites. It also suits distributed generation, since a pyrolysis unit at a customer's meter avoids both a hydrogen pipeline and a CO2 pipeline. Before committing, do the carbon arithmetic first: work out how many tonnes of solid carbon the plant makes each year, find out what grade it is, and get a price for it, because at $600–900 per tonne the carbon pays for the hydrogen outright and at landfill prices it is a disposal cost. If your project needs hundreds of thousands of tonnes of hydrogen a year, this route cannot get there on carbon sales, and reforming with capture (entries 006 and 007) is the comparison to run: same feedstock, carbon as a solid you must sell or bury rather than a gas you must inject.

37.4 kJ/mol H2, about 5.2 kWh/kg in theory, against 39.4 kWh/kg for water splitting · reported 9.7–11 kWh/kg on pilot moving beds and 25–39 kWh/kg on arc plasma · 3 kg of solid carbon per kg of hydrogen · roughly 4 kg of methane per kg of hydrogen against 3.2 for a reformer · $1.7–4/kg before any carbon revenue.

The theoretical energy is 5.2 kWh/kg of hydrogen, and real reactors span a wide range: pilot-scale moving beds have reported 9.7–11 kWh/kg, microwave plasma about 8.7 kWh/kg in the laboratory, and arc plasma pilots 25–39 kWh/kg depending on whether power is generated on site. Even the worst of those is below electrolysis at 50–55 kWh/kg, but only the moving-bed numbers are a large enough advantage to matter. Against that energy saving, the plant consumes roughly 4 kg of methane per kg of hydrogen, about 25% more feedstock than a reformer, so the trade is less electricity for more gas. With no process CO2, the carbon intensity of the hydrogen is set almost entirely by upstream methane leakage and by how the reactor heat is produced.

Monolith Materials runs thermal plasma at Olive Creek in Nebraska, selling carbon black as the primary product with hydrogen as the co-product, which is the commercial model that works today. Hazer Group in Perth uses iron ore as a catalyst in a fluidized bed and produces graphite alongside hydrogen. BASF is developing a moving-bed reactor with induction heating at Ludwigshafen. HiiROC in Hull uses plasma torches with porous anodes, Ekona Power in British Columbia a pulsed intermittent-flow reactor, and Modern Hydrogen in Washington state an autocatalytic carbon-based process aimed at distributed units at the customer's gas meter. Molten Industries is one of several developers working on molten metal bubble columns, and Pacific Northwest National Laboratory has published on nickel-copper fluidized beds.

Published estimates put turquoise hydrogen at $1.80–4.00/kg before any carbon revenue, with molten metal and molten salt designs at the low end near $1.70–1.80/kg and iron-based fluidized beds at $3.20–3.50/kg, against $0.90–3.00/kg for gray hydrogen and $1.40–2.50/kg for reforming with capture. Carbon revenue is what moves it: at $300–800 per tonne the economics match unabated SMR, and at $600–900 per tonne the carbon covers the hydrogen cost entirely. The problem is that those prices belong to the carbon black market, which produces only about 14 million tonnes a year, roughly 70% of it into tires. At 3 kg of carbon per kg of hydrogen, fewer than 5 Mt of hydrogen a year would saturate the entire market, against world hydrogen demand near 100 Mt. So the honest framing is that methane pyrolysis is a carbon black business with a hydrogen co-product, and it stays a good business up to a few million tonnes of hydrogen a year. Beyond that the carbon price collapses and the carbon becomes something to bury.

3 minHazer Pilot PlantHazer Group Limited · 10k+ views

1 minMethane pyrolysis: How do we obtain clear hydrogen for the chemical industry?BASF · 10k+ views

Life Cycle Analysis of Hydrogen Production via Methane Pyrolysis Using Plasma Arc (OSTI.GOV) · Thermocatalytic Decomposition of Methane: A Review on Carbon-Based Catalysts (ACS Omega)

Byproduct hydrogen is hydrogen that falls out of a process run for something else. A chlor-alkali cell splits brine into chlorine and caustic soda and evolves hydrogen at the cathode, about 28 kg per tonne of chlorine, fixed by the reaction stoichiometry rather than by anyone's choice. Steam crackers making ethylene, catalytic reformers inside refineries, and coke ovens at steel mills all throw off hydrogen-rich offgas as well. Together these streams are over 15% of world hydrogen supply, which makes them the third-largest source after reforming and coal gasification. The molecule carries no production capex and no emissions of its own, because the host plant would run either way and the carbon belongs to the chlorine or the ethylene. What it does carry is a purification requirement and an owner who is usually already burning it.

The economics are hard to beat: $0.5–1/kg at the fence, no production capex, and 0 kg CO2e/kg under the accounting convention every scheme uses. Volume is the first weakness, because output is set by the host plant's production rate, so you cannot turn it up when demand grows and you lose it when the host takes an outage. Location is the second: the hydrogen appears where the chlor-alkali plant is, which is rarely where the fleet is. Purity is the third: chlor-alkali hydrogen is already about 99.9% but carries oxygen, water, and chlorine traces, and coke-oven and refinery offgas carries carbon monoxide and sulfur, so reaching ISO 14687 grade D (99.97%, CO under 0.2 ppm) means a PSA train and a deoxo bed rather than a filter. The failure mode nobody prices is the one that matters most: most of this hydrogen is currently burned on site for process heat, so diverting it makes the host buy natural gas instead, and the emissions reappear in the host's boiler.

Take byproduct hydrogen when you need modest steady volume within trucking distance of a chlor-alkali plant, a steam cracker, or a large refinery, and when the purification to fuel-cell grade is worth paying for. It is the cheapest molecule in this sheet, it needs no subsidy, and it is where most European and Korean refueling stations actually get their supply. Don't build a scaling business on it. World byproduct volume is set by chlorine and ethylene demand, and the accessible share (the part not already fueling a furnace) is much smaller than the headline 15%. Before signing, ask the host what the stream does today and what replacing it costs them; if the answer is a boiler, price the natural gas, because that is your real floor. If you need tens of thousands of tonnes a year at a location you choose, build an electrolyzer or a reformer instead.

Over 15% of world hydrogen supply · about 28 kg H2 per tonne of chlorine from a chlor-alkali cell · $0.5–1/kg at the plant fence · 0 kg CO2e/kg by accounting convention · membrane cells draw roughly 2,300–3,000 kWh per tonne of chlorine · ISO 14687 grade D needs 99.97% purity with CO under 0.2 ppm · PSA recovers 85–90% of the stream.

Nothing is spent making this hydrogen, which is the entire point. A membrane chlor-alkali cell draws roughly 2,300–3,000 kWh per tonne of chlorine, and the 28 kg of hydrogen that comes with it holds about 930 kWh on a lower-heating-value basis, so attributing the cell's power by energy content would load the hydrogen with roughly a third of it. Nobody does, because the plant was built for chlorine. What you actually pay for is cleanup and pressure: PSA to fuel-cell grade sends 10–15% of the stream out as tail gas, and getting from cell pressure to a 200 bar tube trailer costs a few kWh/kg, against 50–55 kWh/kg for electrolysis.

Hanwha Energy's 50 MW fuel-cell power plant at the Daesan Industrial Complex in Seosan, South Korea, commissioned in 2020 for about $212 million and running entirely on byproduct hydrogen piped from the Hanwha Total petrochemical plant next door, which makes up to 3 tonnes of hydrogen an hour. Air Liquide and Linde both recover and resell chlor-alkali hydrogen in Europe; Covestro, Nouryon, and Westlake are typical host producers. Korea's and Germany's refueling networks lean heavily on petrochemical and chlor-alkali streams rather than dedicated production.

There is no production capex, so the price is set by the host's opportunity cost plus purification and logistics. Where the stream is vented or flared the floor is near zero; where it is fueling a furnace the floor is the natural gas it displaces, roughly $1.1–1.6/kg equivalent at $8–12/MMBtu. The buyers are industrial gas companies that own the purification skids and the trailers, and that is where the margin sits. The cost curve barely moves, since neither chlorine demand nor cracker economics respond to hydrogen prices. If you are building on it, the asset is the contract and the location rather than the technology, and the supply is capped, so treat it as the bridge that gets a fleet or a station running while cheaper clean production arrives.

2 minChlor-Alkali Process AnimationJones Chang · 100k+ views

15 minThe Chloralkali Industry: Applications of ElectrolysisMindset · 50k+ views

Could hydrogen from chlor-alkali energise Europe's future? (Euro Chlor) · Hydrogen Recovery from Coke Oven Gas. Comparative Analysis of Technical Alternatives (Industrial & Engineering Chemistry Research)

Geologic hydrogen is hydrogen generated underground and trapped in rock, so the business is drilling for it rather than making it. Two mechanisms produce it. Serpentinization is the larger one: water reacting with iron-rich ultramafic rock oxidizes the iron and releases hydrogen, mostly at roughly 200–350 °C. Radiolysis is the slower one, in which radiation from potassium, uranium, and thorium splits water in pore space over geological time. Unlike oil and gas, generation is ongoing, so some reservoirs recharge while they are being produced. The reference case is Bourakébougou in Mali, where a water well drilled in 1987 hit gas that ignited from a cigarette, was plugged for 25 years, and on re-entry in 2011 tested at 98% hydrogen with 1% nitrogen and 1% methane. It has powered a converted engine supplying the village with electricity since 2012.

The attraction is that nothing is converted, so there is no conversion loss and the cost floor is a drilling and cleanup budget rather than an energy bill. Developers in Spain, Australia, and the US target around $1/kg, and Hydroma has claimed $0.5/kg from very shallow Malian wells; nothing else in this sheet can quote numbers like that. The weakness is that almost none of it is proven. The USGS's 2024 model puts about 5.6 trillion tonnes in place worldwide and says most of it is impractical to recover; the fraction that matters, on the order of 100 billion tonnes, would cover roughly 200 years of projected net-zero demand, but it is a modeled fraction of a modeled number. The failure mode is not a dry hole. It is a good well that flows for two years and stops, because the accumulation was static rather than recharging, and nobody yet knows how to tell the two apart before drilling.

Treat this as exploration, not technology. If you are underwriting a geologic hydrogen company, the two numbers that decide the outcome are sustained flow rate per well and evidence of recharge, and resource-in-place estimates are decoration. Ask for a well test measured in months rather than a peak rate, and ask what the gas analysis says, because purity at the wellhead is most of the difference between a cheap kilogram and an expensive one. The category suits investors who can price dry-hole risk across a portfolio on a five-to-ten-year horizon. It does not suit an offtaker who needs contracted volume in 2030: there is no commercial production anywhere, maturity sits around TRL 5, and over 40 companies were still exploring at the end of 2023. If you need hydrogen at a known price on a known date, buy it from an electrolyzer or a reformer and watch this from the sidelines.

Bourakébougou gas tests at 98% hydrogen, 1% nitrogen, 1% methane · producing village power since 2012 · developer cost targets around $1/kg, with $0.5/kg claimed from shallow wells · USGS 2024 estimate of about 5.6 trillion tonnes in place, most of it unrecoverable · a recoverable 100 billion tonnes would cover roughly 200 years of projected net-zero demand · over 40 companies exploring at the end of 2023, roughly TRL 5.

This is the only route in the sheet with no conversion loss, because nothing is converted. The parasitic load is a wellhead compressor plus separation of nitrogen, methane, and any helium, together on the order of a few kWh per kg, against 50–55 kWh/kg for electrolysis and 9–12 kg of CO2 per kg for a reformer. Wellhead purity does most of the work: gas at Bourakébougou's 98% needs very little cleanup, while a 70–80% hydrogen stream needs a PSA unit that leaves 10–15% of the molecules in the tail gas. That is why a discovery's gas analysis matters as much as its flow rate.

Hydroma's Bourakébougou field in Mali, where 24 further exploratory boreholes around the discovery well have found hydrogen. Koloma in the US has raised close to $400 million from Breakthrough Energy Ventures, Amazon's Climate Pledge Fund, United Airlines' Sustainable Flight Fund, and Mitsubishi Heavy Industries. Snowfox Discovery in the UK took a Series A led by BP Ventures; Mantle8 raised a EUR 3.4 million seed from Kiko Ventures and Breakthrough Energy Ventures Europe. Gold Hydrogen is drilling on the Yorke Peninsula in South Australia and Helios Aragón in northern Spain. The USGS has published a first prospectivity map for the United States.

The cost structure is an oil and gas exploration structure: geochemical and seismic survey, a wildcat well, then completion, gathering, and cleanup if it flows. There is no plant to build and no feedstock to buy, which is why the target prices sit below every production route in this sheet. The money so far is venture capital plus strategics buying option value (Mitsubishi Heavy, BP, United Airlines), and returns depend on discovery rather than on a cost curve. What compounds here is acreage: mineral rights over prospective ultramafic belts and cratonic basement, secured before the play is proven. If the category works the winners will look like early shale operators; if it does not, the losses land in exploration budgets rather than in stranded plant, which is a materially better risk shape than most of this sheet offers.

13 minWe went looking for a hidden gas worth MILLIONSDW Planet A · 50k+ views

12 minNatural Hydrogen - clean energy right under our feet | Viacheslav Zgonnik | TEDxManhattanBeachTEDx Talks · 10k+ views

The Potential for Geologic Hydrogen for Next-Generation Energy (US Geological Survey) · Geological Hydrogen: From Natural Occurrences to Anthropogenic Generation – A Critical Review of Potential, Challenges and Prospects (NETL)

Hydrogen leaves an electrolyzer at 1–30 bar and a reformer at 20–30 bar, and almost nobody uses it there, so nearly every delivered kilogram passes through a compressor, a liquefier, or both. Compression is mechanical and mature: multi-stage reciprocating or diaphragm machines take gas to 200 bar for a steel tube trailer, 500 bar for a composite one, or 900 bar for the cascade bank behind a 700 bar dispenser. Liquefaction is a cryogenic plant that cools hydrogen to -253 °C with nitrogen pre-cooling and a helium or hydrogen refrigeration cycle, and it is expensive for a reason peculiar to this molecule. Room-temperature hydrogen is 75% ortho and 25% para; the equilibrium at 20 K is essentially all para; and the ortho-to-para conversion releases 527 kJ/kg against a latent heat of vaporization of only 446 kJ/kg. Left alone, freshly made liquid hydrogen boils itself away, so liquefiers stage catalytic converters through the cold box to do the conversion where the refrigeration can absorb it. This entry also owns cryogenic storage: vacuum-jacketed dewars that lose 0.1–0.3% a day in large tanks and several times that in small ones, because the surface-to-volume ratio is worse.

Compression is cheap, fast, and boring: 3–4 kWh/kg from 30 to 700 bar, $0.3–1/kg, and nothing is lost while the gas sits. Its limit is density, since even at 700 bar hydrogen is only about 40 kg/m3, which caps a steel tube trailer at 300–400 kg of cargo. Liquefaction buys density (71 kg/m3, and a tanker carries 3,500–4,000 kg) and pays 10–13 kWh/kg for it, roughly 30% of the fuel's own energy, plus $1–2.5/kg. The theoretical minimum is under 4 kWh/kg, so there is real headroom, and most of the gap is plant scale: units typically run 5–30 t/day, small enough that fixed inefficiencies dominate. The two failure modes differ in kind. Compressors fail mechanically, and reciprocating and diaphragm machines are the leading maintenance item at refueling stations, so a station with one compressor is a station that goes dark. Liquid fails by evaporating: inventory shrinks every day and eventually vents, which is why liquid supply suits high-throughput sites and punishes anything parked.

If the customer takes gas below 500 bar and sits within a few hundred kilometers, compress. If you are moving more than roughly a tonne a day over long distances by road, or supplying a site that turns its inventory in days, liquid usually wins on delivered cost despite the 30% energy penalty, because one tanker replaces about ten steel tube trailers. The deciding question between them is throughput rather than distance: a station dispensing 100 kg/day should never take liquid, because boil-off eats the advantage before the fuel is sold. If a pipeline exists, use it and skip this decision, since pipe costs a fraction of either path. Watch electrochemical compression (a PEM stack run in reverse, no moving parts, and it purifies while it compresses) for small quiet duty, but it is not yet an answer at throughput.

Compression from 30 to 700 bar costs 3–4 kWh/kg and $0.3–1/kg · liquefaction 10–13 kWh/kg and $1–2.5/kg against a theoretical minimum under 4 kWh/kg · liquid hydrogen at -253 °C and 71 kg/m3, against 40 kg/m3 for 700 bar gas · ortho-to-para conversion releases 527 kJ/kg against a 446 kJ/kg latent heat · boil-off 0.1–0.3% a day in large tanks · liquefiers typically 5–30 t/day.

A kilogram of hydrogen holds 33.3 kWh on a lower-heating-value basis. Compression to 700 bar spends 3–4 kWh of electricity, about 10–12% of that, and nothing more. Liquefaction spends 10–13 kWh, about 30–40%, and then the fuel loses 0.1–0.3% of itself per day in the tank, so a month of storage costs another 3–9%. Chain it: electrolysis at 50–55 kWh/kg, then liquefaction, then a week in a dewar, and you have spent 60–68 kWh of electricity to deliver a kilogram worth 33.3.

Air Products, Linde, Air Liquide, and Plug Power own most of the world's liquefaction capacity; Linde's McIntosh plant in Alabama runs up to 30 t/day, and Plug Power operates roughly 45 t/day across its sites. Kawasaki Heavy Industries built the Suiso Frontier, the first liquid hydrogen carrier ship, for the Australia–Japan demonstration voyage. Compressor suppliers include Howden's Burton Corblin diaphragm machines, Burckhardt Compression, PDC Machines, and Linde's ionic compressor, which Linde quotes at 2.7 kWh/kg from 50 to 900 bar including auxiliaries. HyET Hydrogen and Skyre are the electrochemical compression names; Chart Industries and Linde supply the vacuum-jacketed cryogenic tanks.

Conditioning is capex-heavy and scale-driven. A liquefier's cost per tonne per day falls steeply with size, which is why the industry keeps proposing 30–100 t/day plants and keeps building 5–30 t/day ones: the volume is not there, and the plants that exist were sized for merchant industrial gas and aerospace rather than for fuel. Compression is a different business, with commoditized equipment, margin in service contracts, and operators buying availability rather than efficiency. For anyone modeling delivered cost, this is the step most often left out of a $/kg headline and it can double it. The curve bends with volume rather than with new physics: liquefier efficiency improves as plants get bigger, and compressor prices track the same industrial machinery market they always have.

2 minLinde - Your experts for hydrogen liquefactionLinde · 10k+ views

3 minDiaphragm Compressor Animation | Burton Corblin | HowdenHowden · 50k+ views

18 minIntroduction to CryogenicsChart Industries · 100k+ views

Hydrogen Delivery (US Department of Energy) · Hydrogen liquefaction: a review of the fundamental physics, engineering practice and future opportunities (UWA Research Repository)

Pipe is the cheapest way to move hydrogen, and there is almost none of it. Roughly 5,000 km exists worldwide, about 2,600 km of that in the United States, nearly all of it private industrial line along the Gulf Coast and in northwest Europe connecting refineries and chemical plants that run at steady load. Air Products' Gulf Coast system alone runs about 700 miles from the Houston Ship Channel into Louisiana. A dedicated line costs $0.1–1/kg per 1,000 km, roughly an order of magnitude below trucking, and it works because hydrogen's low density is offset by moving it fast in a large diameter. The alternative that gets proposed instead is blending hydrogen into the existing natural gas network, which needs no new pipe at all. That arithmetic is unkind: hydrogen holds about 10.8 MJ per normal cubic meter against methane's 35.9, so 20% hydrogen by volume is only about 7% by energy, and the CO2 cut is about 7% while every appliance, meter, and compressor on the network has to be checked.

Pipeline transport is cheap in both money and energy, it has no boil-off and no trailer fleet, and where it already exists it removes dollars per kg from delivered cost. Its weaknesses are all about getting one built. Hydrogen embrittles carbon steels, so a repurposed gas line usually needs inspection, fitness-for-service assessment, and a pressure derating, and hydrogen's low density means roughly three times the volumetric flow for the same energy, so existing compressor stations need rework rather than a fuel change. Recompression is needed every 125–250 km. Then there is the chicken-and-egg problem, which is the real one: a pipeline is only financeable with anchor volume at both ends, and neither the producers nor the buyers commit before the pipe exists. Blending has the opposite problem. It is easy to announce and it delivers very little, because a fifth of the network's throughput buys a 7% emissions cut while creating an embrittlement, metering, and appliance-certification program across the whole system.

If you have a producer and a consumer both above roughly 10,000 t/yr within a few hundred kilometers, and preferably a corridor with several of each, dedicated pipe is the right answer and it is worth spending years on. Below that, or with a single offtaker, trucking is cheaper than the pipe's fixed cost, because a hydrogen line is a fixed-cost asset that only pays back at high utilization. Repurposing existing gas line is worth studying wherever a corridor is being freed up, since roughly 60% of Germany's core network is repurposed pipe and that is what makes the budget work, but derate the pressure and price the inspection before assuming the saving. Treat blending as a demand-creation program rather than a decarbonization one: if the goal is CO2, spend the same hydrogen on a refinery or a fertilizer plant that wants pure hydrogen, where it does roughly three times as much work per kilogram.

About 5,000 km of hydrogen pipeline worldwide, roughly 2,600 km of it in the US · transport at $0.1–1/kg per 1,000 km, about a tenth of trucking · recompression every 125–250 km · Germany's core network: 9,040 km approved, EUR 18.9 billion, about 60% repurposed gas pipe, staged to 2032 · 101 GW of feed-in and 87 GW of feed-out capacity at completion · 20% hydrogen by volume is about 7% by energy.

Pipe is the cheapest step in this sheet on energy. European Hydrogen Backbone work puts compressor capacity at 190–330 MW per 1,000 km on a line carrying several gigawatts, which is on the order of 1–3% of the hydrogen's energy, against 10–13 kWh/kg (30% or more) to liquefy the same kilogram for a truck. Blending flips the accounting. A network taking 20% hydrogen by volume swallows a fifth of its throughput as hydrogen and cuts CO2 by about 7%, so each kilogram displaces roughly a third of the carbon it would displace at a customer that actually wanted hydrogen.

Air Products' Gulf Coast network, the largest hydrogen pipeline system in the world. Germany's Wasserstoff-Kernnetz, approved by the Bundesnetzagentur in October 2024 at 9,040 km and EUR 18.9 billion. The European Hydrogen Backbone initiative and Gasunie's HyNetwork in the Netherlands. Sinopec's approved 400 km line from its 1 GW Ulanqab project in Inner Mongolia to the Yanshan petrochemical plant in Beijing, China's first cross-provincial hydrogen pipeline, sized for 100 kt/yr in phase one. On blending, HyDeploy in the UK and the US Department of Energy's HyBlend program are the reference studies.

A hydrogen pipeline is a regulated-utility asset with utility economics: high fixed cost, near-zero marginal cost, and returns that depend entirely on utilization. That is why almost every line built so far is owned by an industrial gas company serving its own customers under long contracts, and why the public networks now being financed (Germany's core network carries a state-backed amortization account to bridge low early throughput) need a policy structure to survive the years when volume is thin. For anyone building a hydrogen business, the practical point is positional: access to a corridor, or a site on one, is worth dollars per kg against a trucked competitor, and that advantage compounds in a way a better electrolyzer does not. Blending's economics are mostly political, since the gas is more expensive per unit of energy than the methane it displaces and the emissions benefit is small.

15 minHydrogen in the Natural Gas NetworkEngineering with Rosie · 100k+ views

2 minCan natural gas pipes handle hydrogen too?Science Animated · 10k+ views

HyBlend: Opportunities for Hydrogen Blending in Natural Gas Pipelines (US Department of Energy) · A unified European hydrogen infrastructure planning to support the rapid scale-up of hydrogen production (Nature Communications)

This is how hydrogen reaches anyone without a pipeline, and it explains why retail hydrogen costs what it does. A 200 bar steel tube trailer carries only 300–400 kg; a 500 bar composite trailer carries roughly 560–1,100 kg; a liquid tanker carries 3,500–4,000 kg. At the far end, a 700 bar station takes delivery into low-pressure storage, compresses to a cascade bank at around 900 bar, chills the gas to -40 °C so the vehicle tank stays under 85 °C during a three-to-five minute fill, and dispenses under the SAE J2601 protocol. Compression and chilled cascade storage dominate the station's capital cost. About 1,160 stations were operating worldwide at the end of 2024, 748 of them in Asia (384 in China, 198 in South Korea, 161 in Japan) and 294 in Europe, of which 113 are in Germany.

Trucking works anywhere, needs no right-of-way, and scales one trailer at a time, which is the only reason hydrogen mobility exists at all outside a few industrial corridors. The weakness is payload. Hydrogen is so light that a full 200 bar steel trailer weighs tens of tonnes to carry 300–400 kg of cargo, so the truck is mostly hauling its own pressure vessels, and delivery adds $1–4/kg over a few hundred kilometers. Composite trailers and liquid tankers fix the payload problem and add their own: composite tubes are expensive, and liquid boils off. The station is the harder half. Capital equipment ran $1,200–3,000 per kg/day of capacity across 111 California station awards reviewed by the Department of Energy, so a 1 t/day site is $1–3 million of equipment before installation, and the failure mode is availability rather than cost. Stations are compressor-limited, single-compressor sites go offline when the machine does, and California drivers spent 2023 and 2024 finding pumps down and prices at records.

Truck compressed gas when a customer takes under about 500 kg/day and sits within a few hundred kilometers of supply. Switch to liquid when daily volume passes roughly a tonne and inventory turns in days, because one tanker replaces about ten steel trailers. Build a station only where the fleet is captive and the utilization is known in advance: a bus or truck depot with a single dispenser serving 20 or more vehicles a day amortizes the compressor, while a public retail site depends on a vehicle population that mostly has not shown up. If the customer's volume ever justifies a pipeline, take the pipeline, since it costs a tenth as much per kg. On-site electrolysis at the station is the other alternative, and it trades delivery cost for a much worse capacity factor on the stack, so it usually only wins where trucking is long and power is cheap.

200 bar steel tube trailer 300–400 kg, 500 bar composite trailer 560–1,100 kg, liquid tanker 3,500–4,000 kg · trucking adds $1–4/kg over a few hundred kilometers · station capital equipment $1,200–3,000 per kg/day of capacity, so $1–3 million for a 1 t/day site · dispensing at 700 bar with -40 °C pre-cooling under SAE J2601 · about 1,160 stations worldwide at the end of 2024 · California retail averaged $33.49/kg across 2024.

Dispensing costs more of the fuel's energy than most people expect. A 700 bar station compresses to a roughly 900 bar cascade at 3–4 kWh/kg and pre-cools the gas to -40 °C at about another 0.15 kWh/kg, so roughly 10–13% of the kilogram's 33.3 kWh goes into putting it in the vehicle. Trucking adds little energy and a lot of cost, because a 200 bar steel trailer moves 300–400 kg of hydrogen inside tens of tonnes of steel. Stack the chain and it shows: 50–55 kWh/kg to electrolyze, 3–4 to compress for the trailer, 3–4 more at the station, and the vehicle receives a kilogram worth 33.3 kWh.

TrueZero, California's largest operator with 37 stations, charged $36/kg through late 2024 and early 2025. Shell permanently closed all seven of its California light-duty stations in February 2024, citing supply problems and market conditions. Iwatani, Air Liquide, and Nel supply station equipment; Hexagon Purus and FIBA Technologies build composite and steel trailers; Burckhardt Compression and PDC Machines supply the compressors. South Korea's and China's networks, which together hold about half the world's stations, are built around bus and commercial-fleet depots rather than passenger retail.

Retail hydrogen is expensive for four stacked reasons, only one of which is the molecule: production, trucking at $1–4/kg, station capital amortized over utilization that rarely reaches design, and retail margin. That is how a kilogram produced in the low single digits reaches $33.49 on average at a California pump. The cost curve improves mainly through utilization, so the economics work first where a depot guarantees throughput and last in public retail, which is why the operators that quit (Shell) quit the light-duty side. For anyone building here, the durable positions are the depot contract and the composite-trailer supply chain, not the dispenser. And if a business case assumes $4–6/kg at the nozzle, ask which station has ever charged it.

5 minExplainer video - Hydrogen Refuelling Station - CMB.TECHCMB. TECH · 10k+ views

20 minTech Talk - Hydrogen Refueling Station - Hydrogen Fuel Cell Technology Explained - Hyfindr MariacherHyfindr - Hydrogen & Fuel Cell Technology · 5k+ views

Hydrogen Fueling Stations (Alternative Fuels Data Center) · 2025 Annual Evaluation of Fuel Cell Electric Vehicle Deployment and Hydrogen Fuel Station Network Development (California Air Resources Board)

Compressed hydrogen is held in one of five tank types, and the type decides the application. Type I is monolithic steel or aluminum, cheap and heavy, rated to 200–300 bar, and it is what industrial cylinders, tube trailers, and stationary buffers use. Type II adds a hoop wrap over a metal liner. Type III wraps a full composite over a thin aluminum liner. Type IV puts the full carbon fiber overwrap over a polymer liner, usually high-density polyethylene or polyamide, and it is the 700 bar automotive standard at roughly 5–6% hydrogen by system mass. Type V removes the liner entirely and is still developmental. The physics that drives all of it: hydrogen at 700 bar is only about 40 kg/m3 and at 350 bar about 23, against 71 for liquid hydrogen, so the tank is always large for what it holds, and the wall has to carry the pressure with the least mass possible.

Pressure vessels are the most boring and most reliable storage in this sheet. Nothing evaporates, nothing degrades, fill and release take minutes, and there is no thermal management beyond the fill itself. The certification base is deep (ISO 19881, UN GTR 13, EC 79) and the failure behavior of a wrapped composite is well characterized: it leaks before it bursts, and it is proof-tested and cycle-tested to prove it. The weakness is volume and the cost of buying it back. A 5 kg automotive tank is about 125 liters of internal volume before insulation and structure, which is why hydrogen cars have small trunks and hydrogen trucks give up frame space. Carbon fiber is the reason the cost does not fall on its own: the fiber layer is 62–65% of the cost of a 700 bar Type III dual-tank system in the Department of Energy's analysis, and coordinated DOE-industry work cut the fiber's share of a Type IV system from about 52% in 2019 to 33.5% in 2023. Tank prices therefore track aerospace-grade fiber supply more than they track hydrogen demand.

Use Type I for anything stationary or industrial, where mass does not matter and the price per kg stored is the only question. Use Type IV at 700 bar when the application is a passenger vehicle and range per unit of volume is the binding constraint. Use 350 bar Type III or Type IV for buses, trucks, forklifts, and marine, where the packaging space exists and halving the pressure roughly halves the compression energy and simplifies the station. If your vehicle needs more than about 60–80 kg on board, price liquid hydrogen against compressed before committing, because at that size the tank volume and the composite cost start to dominate. And if the equipment is heavy and slow-moving and cares more about pressure than mass, look at metal hydrides instead, which are denser by volume and far heavier.

700 bar hydrogen at about 40 kg/m3, 350 bar at about 23, liquid at 71 · a 5 kg automotive tank is roughly 125 liters of internal volume · Type IV systems store 5–6% hydrogen by system mass · Type I rated 200–300 bar · carbon fiber is 62–65% of a 700 bar Type III system's cost, and 33.5% of a 2023 Type IV system's · filling costs 3–4 kWh/kg from a 30 bar source.

A pressure vessel loses nothing while it holds: no boil-off, no chemistry, no heat leak that matters. The whole energy cost is the fill, 3–4 kWh/kg to reach 700 bar from a 30 bar source, about 10% of the kilogram's 33.3 kWh, and roughly half that for a 350 bar tank. What you pay instead is volume: 700 bar hydrogen holds about 4.7 MJ per liter against diesel's 36, so a 5 kg automotive tank runs to roughly 125 liters of internal volume before insulation, valves, and structure. That volumetric penalty, rather than the energy penalty, is what has kept compressed hydrogen out of most vehicle classes.

The Toyota Mirai carries 5.6 kg in three Type IV tanks; Hyundai's Nexo and its XCIENT fuel-cell truck use the same architecture at 700 and 350 bar respectively. Hexagon Purus, Forvia, Plastic Omnium, Luxfer Gas Cylinders, NPROXX, and Worthington Enterprises are the main vessel suppliers, with Toray, Hexcel, and Mitsubishi Chemical supplying fiber. Type I steel cylinders remain the workhorse of the industrial gas business and of every tube trailer built before composites got cheap.

Tank cost is a materials cost wearing a manufacturing cost's clothes. Carbon fiber dominates, filament winding is capital-intensive but well understood, and the learning curve runs through fiber loading and layup optimization rather than through process invention. The Department of Energy projects a 700 bar Type IV system holding 5.6 kg of usable hydrogen at $12.1–13.3/kWh in 2020 dollars at 100,000 units a year, and today's volumes are two or three orders of magnitude below that, so real prices are much higher and will stay there until a vehicle program creates the volume. The strategic consequence is uncomfortable for the sector: tank makers are exposed to a fiber market driven by aerospace and wind, and a fuel-cell vehicle program that assumes tank costs fall with hydrogen adoption has the causality backwards.

4 minH2 tanks type 1 - type 5, technological differences and advantagesVoith Group · 10k+ views

2 minHow it works: Filament WindingFarixltd · 100k+ views

Physical Hydrogen Storage (US Department of Energy) · Onboard Type IV Compressed Hydrogen Storage System Cost and Performance Status (DOE Hydrogen Program)

Underground storage is the only hydrogen reservoir cheap enough to hold months of industrial inventory, and whether you can have one depends entirely on geology. A salt cavern is solution-mined by circulating fresh water through a well into a salt dome or bedded salt over one to three years, leaving a void that then holds hydrogen at 100–200 bar. Salt suits hydrogen better than any other rock: it is effectively impermeable, chemically inert, and creeps closed around damage rather than fracturing. Four caverns have stored pure hydrogen commercially for decades, three on the US Gulf Coast (Clemens Dome since 1983, Moss Bluff since 2007, and Spindletop) and one site at Teesside in the UK since the 1970s. Clemens and Moss Bluff are each roughly 580,000 m3 at about 800 m depth; Teesside runs three caverns totaling 210,000 m3 at only 45 bar, holding about 25 GWh. Depleted gas fields and saline aquifers are far larger and cheaper per unit of capacity, and unproven for hydrogen.

Nothing else stores hydrogen this cheaply. Levelized storage in a salt cavern runs about $0.39–2.41/kg against $5–15/kg in above-ground steel tanks, hydrogen does not degrade or boil off while it sits, and a single cavern holds thousands of tonnes. The constraints are geographic and financial. Suitable salt is concentrated in the US Gulf Coast, northern Germany, the Netherlands, Poland, and the UK, so most of the world simply has no option. Cushion gas is the capital item people forget: 20–50% of a salt cavern's inventory has to stay in place permanently to hold pressure, and over half in a depleted field, and that hydrogen is bought once at production cost and never sold. Porous-rock storage adds risks salt does not have. Methanogens and sulfate-reducing bacteria eat hydrogen and return methane and hydrogen sulfide, withdrawal rates are unknown, and the returned gas needs more cleanup than a salt cavern's, which mostly needs drying and brine-mist removal.

Site industrial hydrogen production over salt if you can. A green ammonia or methanol plant running on variable renewables needs days of buffer to keep its synthesis loop at steady state, and a cavern supplies that at a cost that never shows up in the levelized number; without one, you pay instead by oversizing the electrolyzer or by curtailing the plant. A refinery or steel mill converting to low-carbon supply needs weeks of inventory against an outage, which is the same argument at a different time constant. Secure the cavern rights early, since the geology is a fixed asset and the permitting runs years ahead of the mining. If there is no salt, do not force it: take the buffer above ground and accept 10x the cost per kg for a small working volume, or move the storability into the supply contract by importing ammonia. Treat depleted-field storage as a pilot until someone publishes multi-cycle withdrawal data.

Salt caverns hold thousands of tonnes at 100–200 bar · levelized storage $0.39–2.41/kg against $5–15/kg in above-ground steel tanks · Clemens Dome and Moss Bluff are each about 580,000 m3 at roughly 800 m depth · Teesside stores about 25 GWh across three caverns at 45 bar · cushion gas is 20–50% of a salt cavern's inventory and over half of a depleted field's · solution mining takes one to three years.

A cavern is close to free on energy once the gas is in. Compressing from an electrolyzer's 30 bar to 100–200 bar costs on the order of 5–15% of the hydrogen's energy depending on inlet pressure, and after that salt loses essentially nothing to permeation or evaporation no matter how long the gas sits, which is the difference between this and a cryogenic tank. The loss that hurts is financial rather than thermodynamic: cushion gas is 20–50% of the inventory, bought once at whatever your production cost is, so a cavern with 3,000 tonnes of working gas needs another 750–3,000 tonnes of cushion, which at $3/kg is $2–9 million of hydrogen permanently underground before it sells a kilogram.

Chevron Phillips' Clemens Dome, Air Liquide's Spindletop, and Moss Bluff in Texas, plus Sabic's Teesside caverns in the UK, are the four long-running commercial stores. ACES Delta in Utah, a Mitsubishi Power and Magnum Development joint venture, is mining caverns sized to feed the Intermountain Power Project's hydrogen-capable turbines. Gasunie's HyStock in the Netherlands and EWE's Rüdersdorf test cavern in Germany are the European references, and Storengy is developing salt storage in France. RAG's Underground Sun Storage in Austria is the leading depleted-field trial.

The capital is solution mining priced per cubic meter of cavern, plus wells, brine disposal, compression, and the cushion gas inventory; operating cost is small and the asset lasts decades. Value therefore concentrates in the geology rights rather than in any equipment, and in a corridor where salt sits near a pipeline and a customer, the rights are worth more than a production technology edge. For an industrial project the storage has no revenue line of its own. It shows up as a smaller electrolyzer, a steadier ammonia loop, and a supply guarantee an offtaker will sign a longer contract against, all of which are worth more than the storage costs. The energy-storage sheet covers the separate question of caverns as seasonal grid storage; here the case is inventory and supply security for a plant that has to run.

5 minHow does underground gas storage work?Uniper · 10k+ views

2 minK+S’s Bethune Mine – How does solution mining work?K+S Agrar · 10k+ views

Subsurface Hydrogen and Natural Gas Storage: State of Knowledge and Research Recommendations (NETL) · Underground hydrogen storage: The techno-economic perspective (Open Research Europe)

Solid-state storage holds hydrogen inside a material instead of inside a pressure envelope. In a metal hydride, hydrogen atoms occupy interstitial sites in a metal alloy's lattice, and temperature and pressure decide which way the reaction runs: absorption is exothermic, desorption endothermic. The workhorse families are AB5 alloys such as LaNi5, which take up 1.2–1.5% hydrogen by mass, and AB2 Laves-phase alloys such as TiMn2 and ZrMn2 at 1.8–2.2%, both operating near room temperature below 100 bar. Magnesium hydride reaches about 7.6 wt%, but its bond is much stronger and it needs roughly 300 °C to give the hydrogen back. Physisorbents are the other branch: metal-organic frameworks and activated carbons hold hydrogen by van der Waals attraction on very high surface area, which works at 77 K and barely at all at room temperature. The number that makes the category interesting is volumetric: LaNi5 hydride holds about 115 kg of hydrogen per cubic meter, against 71 for liquid hydrogen and 40 for 700 bar gas.

A hydride tank is the safest and densest storage on this list by volume. It sits at low pressure, so a puncture releases very little, and it holds indefinitely with no boil-off. The disqualifying weakness is mass. At 1.5–2 wt% of the alloy alone, a full system lands closer to 1%, which is why no materials-based system has met the Department of Energy's 5.5 wt% and 40 g/L targets, and why the ultimate 7.5 wt% goal remains a research target rather than an engineering one. The second problem is thermal. Absorption releases heat, so a hydride bed has to be actively cooled during a fast fill or it stops taking hydrogen, and desorption needs heat back at a temperature the alloy dictates. That is the failure mode in the field: a system sized on material capacity that cannot fill in reasonable time or cannot hold discharge rate because the heat exchanger, not the alloy, is the limit. Sorbents have the mirror problem, since keeping a MOF at 77 K costs more than the hydrogen is worth outside a laboratory.

Choose a hydride when the platform is already heavy, cares about pressure or safety more than mass, and can supply or absorb heat. Submarines are the canonical case, and stationary buffers, ballast-carrying marine vessels, and backup power in occupied buildings are the same argument. Match the alloy to the heat you have: an AB5 or AB2 alloy releases at low enough temperature that a PEM fuel cell's own coolant can drive it, which is what makes an integrated system work at all, while magnesium hydride only makes sense next to an industrial heat source. Do not specify a hydride for anything that has to move its own storage on wheels or wings, because a 700 bar Type IV tank beats it by a factor of five on system mass. And be skeptical of any material claim quoted at the powder level, since the tank, the heat exchanger, and the balance of plant typically halve it.

AB5 alloys hold 1.2–1.5% hydrogen by mass and AB2 alloys 1.8–2.2% · LaNi5 hydride holds about 115 kg H2/m3, against 71 for liquid hydrogen and 40 for 700 bar gas · magnesium hydride reaches about 7.6 wt% but needs roughly 300 °C · desorption enthalpy about 31 kJ/mol H2 for AB5 and about 75 kJ/mol for magnesium hydride · DOE system targets of 5.5 wt% and 40 g/L remain unmet · Type 212A submarines carry 18 hydride tanks of 4.4 tonnes each.

Nothing is lost in storage, but the release costs heat, and the bond strength sets how much. An AB5 alloy gives its hydrogen back at about 31 kJ per mole, roughly 4 kWh/kg or 13% of the fuel's 33.3 kWh, and at a low enough temperature that a fuel cell's waste heat covers it, so in an integrated system the loss is close to free. Magnesium hydride holds five times as much per kg of alloy but binds it at about 75 kJ per mole, roughly 10 kWh/kg or 31%, and demands 300 °C to deliver it, which is why the high-capacity materials have never shipped in a vehicle. Filling is the mirror image: absorption releases the same energy as heat, and a tank that cannot reject it fills slowly.

German Type 212A submarines are the fielded case: 18 metal hydride tanks of about 4.4 tonnes and 1,200 liters each, feeding Siemens PEM fuel cells for roughly two weeks submerged, which buys air-independent propulsion at a mass penalty a submarine can carry because it needs ballast anyway. GKN Hydrogen sells containerized stationary hydride units in the HY2 range; Hydrexia works on magnesium-based storage. On the sorbent side, MOF-5 and NU-1501 are the reference materials, and the Department of Energy's HyMARC consortium coordinates most of the US research.

Hydride systems cost far more per kg stored than pressure vessels, and the cost is in the alloy: rare earths for AB5, titanium and zirconium for AB2, all bought by the tonne for a tank that holds tens of kilograms of hydrogen. There is no obvious learning curve, because the constraint is materials chemistry rather than manufacturing, so prices will not follow the trajectory carbon fiber or electrolyzer stacks are on. That leaves the addressable market to applications that will pay for safety, low pressure, and volumetric density rather than for cheap energy: naval propulsion, stationary buffers in occupied spaces, and equipment where a compressor is unwelcome. Anyone underwriting a solid-state storage company should be asking which of those markets is large enough to build on, not when the material target will be met.

3 minHydrogen Storage in Metal Hydrides [Reupload]Helmholtz-Zentrum Hereon · 10k+ views

2 minHy2green ‒ The safest storage system for renewable energyGKN Powder Metallurgy · 10k+ views

13 minSolid Hydrogen Explained (Again) - Is it the Future of Energy Storage?Undecided with Matt Ferrell · 1m+ views

Materials-Based Hydrogen Storage (US Department of Energy) · Thermal Management in Metal Hydride Hydrogen Storage Reactors: Mechanisms, Architectures, and Design Trade-Offs (Nanomaterials)

Ammonia is the one hydrogen carrier with a working global industry already behind it. NH3 is 17.8% hydrogen by mass, and liquid at -33 °C and atmospheric pressure it holds about 121 kg of hydrogen per cubic meter, more than liquid hydrogen's 71 and three times what a 700 bar tank manages. About 180 Mt a year is produced and roughly 20 Mt of that moves by sea, so the ships, terminals, refrigerated tanks, and handling codes exist and have existed for fifty years. There are two entirely different businesses inside that fact. One is selling ammonia as ammonia, to a fertilizer plant, a ship's engine, or a coal boiler, where the molecule is the product and no hydrogen is ever recovered. The other is shipping ammonia to an import terminal and cracking it back over a nickel or ruthenium catalyst at 400–600 °C, then purifying the output, which is the route that turns ammonia into a hydrogen delivery method rather than a chemical. Entry 025 covers the synthesis plant from the other side, as the 32 Mt/yr buyer of hydrogen that it is.

The strength is that none of the logistics has to be invented: gas carriers, refrigerated storage, port procedures, and insurance markets all exist, and -33 °C is a far easier temperature to hold than -253 °C, so boil-off is a manageable operating cost rather than a design constraint. The weaknesses are toxicity and the round-trip loss. Ammonia's odor threshold is a few ppm and its IDLH is 300 ppm, so leaks announce themselves early, but a large release near a port is a mass-casualty event and permitting reflects that. Combustion adds a second problem: nitrous oxide has a global warming potential of about 273, so a fraction of a percent of N2O in the exhaust can undo most of the carbon saving, and unburned ammonia slip has to be caught as well. The failure mode on the cracking side is a project underwritten on nameplate cracker capacity. A cracker's heat usually comes from burning part of the product, purification to ISO 14687 fuel-cell grade means getting residual ammonia under 0.1 ppm, and both of those are routinely left out of the delivered-cost model.

Ship ammonia when your buyer wants ammonia. Fertilizer offtake, marine bunkering, and coal co-firing all take the molecule as delivered, the value chain is bankable today, and the carbon accounting is straightforward because there is no reconversion step to argue about. If the customer at the far end wants hydrogen molecules, do the arithmetic before committing: cracking and purification add roughly $0.5–1.5/kg and destroy a fifth to a quarter of the energy, so imported-then-cracked hydrogen has to beat whatever the destination can make locally, and in most of Europe and Japan that comparison is close enough to be decided by power price rather than by the carrier. Prefer LOHC (entry 020) if toxicity is what blocks the permit, and liquid hydrogen (entry 013) if the volumes are small and the customer needs high purity with no cracker at all. Treat any import project whose economics depend on selling both the hydrogen and the nitrogen as speculative.

17.8% hydrogen by mass and about 121 kg H2 per cubic meter, against 71 for liquid hydrogen and 40 for 700 bar gas · liquid at -33 °C and atmospheric pressure · roughly 180 Mt/yr produced and 20 Mt/yr moved by sea · cracking at 400–600 °C, endothermic at 46 kJ per mole of ammonia · synthesis loses 10–20% of the hydrogen's energy and cracking another 15–25% · cracking adds roughly $0.5–1.5/kg and the full export chain $2–4/kg · fuel-cell grade needs residual ammonia below 0.1 ppm.

One kg of hydrogen makes 5.6 kg of ammonia, which carries 29 kWh of the hydrogen's original 33.3 kWh, so the chemistry alone loses about 13%; the air separation unit and the loop compressors add roughly 0.6–1.0 kWh per kg of ammonia and take the synthesis step to 80–85%. Cracking is endothermic at 46 kJ per mole, which is 4.2 kWh per kg of hydrogen recovered, or 13% of its heating value at the thermodynamic floor, and a real cracker running at 400–600 °C with PSA cleanup behind it consumes 15–25%. Multiply the two and about 70% of the hydrogen energy survives the round trip before shipping fuel, boil-off, and the compression at each end are counted, which is why ammonia only makes sense over distances where a pipeline is impossible.

JERA's Hekinan Unit 4 in Japan is the largest demonstration: 20% ammonia by heat on a 1 GW coal unit for two months, with NOx no higher than before, SOx down about 20%, and N2O below the detection threshold. JERA is targeting commercial 20% co-firing there in fiscal 2029, which needs roughly 500,000 t/yr of ammonia for that single unit. On the marine side, MAN Energy Solutions' ME-LGIA and WinGD's X-DF-A dual-fuel engines are in production, around 40 ammonia-capable ships are on order, and the tug Sakigake took the first truck-to-ship ammonia bunkering at Yokohama in 2024. Crackers are being commercialized by thyssenkrupp Uhde, Air Liquide, Topsoe, KBR, and Siemens Energy, with the first demonstration units at import terminals in Rotterdam and northern Germany.

Gray ammonia is a traded commodity with a published price, which makes this the easiest entry in the sheet to underwrite and the hardest to fool yourself about: CFR Tampa settled around $585–615/t through the first half of 2026, a multi-year high driven by gas costs. Work backwards and a tonne of ammonia contains 178 kg of hydrogen, so at $2/kg the hydrogen alone costs about $355/t and at $5/kg it costs about $890/t, before the air separation unit, the synthesis loop, the shipping, or any margin. That arithmetic is the whole business case. Green ammonia does not become competitive on price against a gas-based plant; it becomes competitive when a buyer is paying for compliance, which today means FuelEU Maritime and the IMO's fuel-intensity rules for shipping, Japanese and Korean co-firing mandates for power, and CBAM for fertilizer imported into Europe. If a project's revenue line assumes the fertilizer market pays a premium for low-carbon ammonia rather than a regulator forcing someone to buy it, the offtake is not real.

7 minHow to transport hydrogen using ammonia (NH3)Future Energy & Technology · 10k+ views

1 minHow does ammonia cracking work?thyssenkrupp Uhde · 10k+ views

15 minCan this magic fuel clean up the shipping industry?DW Planet A · 100k+ views

Innovation Outlook: Renewable Ammonia (IRENA) · Ammonia: Zero-Carbon Fertiliser, Fuel and Energy Store (Royal Society)

A liquid organic hydrogen carrier is an oil that you hydrogenate at one end of a trade route and dehydrogenate at the other, so the hydrogen travels chemically bound inside an ordinary liquid at ambient temperature and pressure. The two systems with real projects behind them are toluene and its hydrogenated form methylcyclohexane, used by Chiyoda, and benzyltoluene or dibenzyltoluene, used by Hydrogenious. Both hold 5–6% hydrogen by mass and about 50–57 kg per cubic meter, which is less than ammonia's 121 but is carried in a chemical tanker with no refrigeration, no pressure, and no toxic-release scenario. Hydrogenation runs at 150–200 °C and 30–50 bar over a nickel or ruthenium catalyst and releases heat; dehydrogenation runs at 250–320 °C over platinum on alumina and absorbs it back. The carrier itself is never consumed, so the return leg of every voyage carries the lean liquid home and the whole inventory in circulation is working capital. That closed loop, and the temperature gap between the two reactions, is what decides whether an LOHC project works.

The safety and logistics case is genuinely good. Dibenzyltoluene is a commercial heat transfer fluid, not flammable at ambient conditions and not acutely toxic, so a spill is an industrial cleanup rather than an evacuation, and the liquid can sit in an unheated tank for years without losing anything. The problem is the release step. Dehydrogenation absorbs about 65–71 kJ per mole of hydrogen, which is 9–10 kWh per kg, or 27–30% of the hydrogen's heating value, and it needs that energy as heat at 250–320 °C. Hydrogenation gave the same energy back at 150–200 °C, too cool to drive the release, so the two ends do not cancel and the destination has to find clean high-grade heat it usually does not have. The failure mode is the one that follows: a project that models the release step on waste heat, then discovers the only heat available at the terminal is made by burning a quarter of the delivered hydrogen. Cycle life is the second thing to check, since the carrier degrades slowly and the make-up rate sets a real operating cost.

Pick an LOHC when the reason ammonia is blocked is toxicity or permitting rather than cost, and when the destination has 250–320 °C of clean process heat sitting idle. A refinery or chemical site is the natural host, because it already has that heat, already handles aromatics by the tanker load, and already has a hydrogen user on the fence line. Also consider it for inventory rather than transport: dense urban sites, tunnels, and buildings that cannot permit compressed or cryogenic hydrogen can hold months of it in a diesel tank. Do not pick it for a long export route judged purely on delivered cost, because ammonia carries twice the hydrogen per cubic meter, has an existing fleet, and does not tie up a carrier inventory (entry 019 is the comparison to run). And do not pick it if the customer wants fuel-cell-grade hydrogen with no cleanup train, since the released gas carries traces of the carrier.

5–6% hydrogen by mass and about 50–57 kg H2 per cubic meter, against ammonia's 121 · liquid at ambient temperature and pressure · hydrogenation at 150–200 °C and 30–50 bar, dehydrogenation at 250–320 °C · release absorbs 65–71 kJ per mole of hydrogen, 9–10 kWh/kg, 27–30% of the fuel's energy · roughly 16 tonnes of carrier in circulation per tonne of hydrogen delivered · lean dibenzyltoluene runs about EUR 2–4/kg · Chiyoda's Brunei to Kawasaki demonstration moved 210 tonnes of hydrogen.

Hydrogenation loses almost no hydrogen, since every molecule ends up in the carrier, but it dumps 9–10 kWh per kg of hydrogen as heat at 150–200 °C, which is normally rejected because nothing on a hydrogenation plant can use heat at that grade. Dehydrogenation then demands the same 9–10 kWh back at 250–320 °C, 27–30% of the hydrogen's 33.3 kWh, and if that heat comes from burning product then roughly a quarter to a third of the delivered molecules never reach the customer. Add compression at both ends and cleanup after release and an LOHC chain typically delivers 65–70% of the hydrogen energy that entered it, comparable to ammonia's round trip but concentrated almost entirely in one step at the destination.

Chiyoda's SPERA system ran the first international hydrogen supply chain, moving 210 tonnes of hydrogen as methylcyclohexane from Brunei to Kawasaki in 2020 under the AHEAD consortium, dehydrogenating it at Toa Oil's Keihin refinery and burning the hydrogen in a gas turbine at the Mizue power plant. Chiyoda has since run an MCH dehydrogenation demonstration in Singapore. Hydrogenious LOHC Technologies is the benzyltoluene reference, with its plant at Erlangen, Project Hector at CHEMPARK Dormagen storing byproduct hydrogen, and a release hub sized at roughly 1,800 t H2/yr targeted for 2028. Ayrton Energy in Canada and Asemblon in the US are working on alternative carrier molecules. Johnson Matthey and Clariant supply the dehydrogenation catalysts.

Three cost lines matter and only one of them is the plant. The carrier inventory comes first: at 6 wt% you need about 16 tonnes of liquid per tonne of hydrogen, and at EUR 2–4/kg for lean dibenzyltoluene that is EUR 32,000–64,000 of working capital per tonne of hydrogen in circulation, plus enough extra to cover both legs of every voyage and the make-up for degradation. Second is the heat at the destination, which is the operating cost that decides the delivered price, and published techno-economic studies vary by more than a factor of two on exactly this assumption, so read any LOHC cost estimate by looking at its heat source first. Third is platinum on the dehydrogenation catalyst, a smaller line but one that scales with throughput. The result is that LOHC is not a cheap way to move hydrogen and is unlikely to become one; the business it can win is the one where a customer is paying for a safety or permitting property that compressed gas, liquid hydrogen, and ammonia cannot offer.

3 minChiyoda’s Hydrogen Supply Chain千代田化工建設 Chiyoda Corporation · 1k+ views

4 minMCH Dehydrogenation Demonstration Project in Singapore千代田化工建設 Chiyoda Corporation · under 1k views

2 minSafe and easy hydrogen handling with Hydrogenious' LOHC benzyltolueneHydrogenious LOHC · under 1k views

Large-Scale Stationary Hydrogen Storage via Liquid Organic Hydrogen Carriers (iScience) · Development of Liquid Organic Hydrogen Carriers for Hydrogen Storage and Transport (International Journal of Molecular Sciences)

E-methanol is hydrogen and carbon dioxide combined over a copper-zinc-alumina catalyst at 50–100 bar and 200–300 °C, the same catalyst chemistry that has made methanol from syngas for fifty years. The product is 12.6% hydrogen by mass and about 100 kg of hydrogen per cubic meter, liquid at ambient temperature and pressure, and already a global commodity with tanks, pipelines, barges, and terminals everywhere. That combination makes it the easiest synthetic liquid to produce and the only one with a buyer that has already ordered ships. The chemistry carries one penalty that shapes the economics: CO2 plus three hydrogen gives methanol and water, while a conventional syngas plant gets its carbon as CO and needs only two. Building methanol from CO2 therefore costs 50% more hydrogen per tonne than the incumbent route, which is why the electrolyzer, not the synthesis loop, dominates the plant cost.

Methanol is the cheapest place to put a kilogram of hydrogen into a shippable liquid, it needs no cryogenics or pressure vessels, and the handling risks are ordinary industrial ones rather than novel. The weaknesses are the price gap and the carbon supply. IRENA puts e-methanol at $800–1,600/t with biogenic CO2 and $1,200–2,400/t if the carbon comes from direct air capture, against $300–400/t for fossil methanol in the US Gulf, so nobody buys it as methanol. It also holds 15.8 MJ per liter against diesel's 36, so a methanol-fueled ship needs roughly 2.3 times the tank volume. The failure mode is the CO2 contract. EU RFNBO accounting credits CO2 captured from a power station only until 2036 and from other fossil industrial sources only until 2041, so a plant built around a nearby cement kiln or refinery has a dated asset, and one with no biogenic or direct-air source within pipeline distance has nowhere to go afterwards.

Pick e-methanol when your buyer is a methanol-capable ship or a chemical plant, and when you have biogenic CO2 under contract within pipeline distance. Under those conditions it is the cheapest synthetic hydrocarbon per unit of hydrogen delivered, and the logistics chain already exists. Don't pick it if the end use is aviation; you will pay for a methanol-to-jet step on top, and entry 022 covers whether that is worth it. Don't pick it if the customer only wants hydrogen back, because ammonia carries more hydrogen per cubic meter and needs no carbon at all. And treat any pro forma that prices the product against the methanol market as broken. E-methanol sells into a compliance market, so the number that matters is what FuelEU Maritime compliance costs the shipowner, not what Methanex charges.

0.19 kg of hydrogen and 1.4 kg of CO2 per kg of methanol · 12.6% hydrogen by mass and about 100 kg H2 per cubic meter · synthesis at 50–100 bar and 200–300 °C · $800–1,600/t with biogenic CO2 and $1,200–2,400/t on direct air capture, against $300–400/t for fossil methanol · world methanol demand roughly 110–120 Mt/yr, consuming about 16 Mt of hydrogen · Kassø, the first commercial plant, runs at 42,000 t/yr.

Stoichiometry sets the inputs at 0.19 kg of hydrogen and 1.4 kg of CO2 per kg of methanol, and the CO2 route needs three hydrogens where a syngas plant on CO needs two, so the carbon source alone costs 50% more hydrogen. At 55 kWh/kg that hydrogen represents about 10.4 kWh of electricity per kg of product, and methanol holds 5.5 kWh/kg, so roughly half the electricity ends up in the fuel before capture and compression are counted. The synthesis step itself is the efficient part, retaining about 85% of the hydrogen's energy, and it runs exothermically, so the heat is usually rejected rather than recovered.

European Energy's Kassø plant at Aabenraa, Denmark, inaugurated in May 2025 at 42,000 t/yr next to the 304 MW Kassø solar park, with Mitsui holding 49% and Maersk, LEGO, and Novo Nordisk as offtakers. HIF Global's Haru Oni in Chile, which runs e-methanol on to gasoline. Carbon Recycling International's George Olah plant in Iceland, the long-running geothermal-powered demonstration. On the demand side, Maersk's Laura Mærsk and a methanol dual-fuel orderbook of about 25 vessels. Methanex is the incumbent to price against, with single plants such as Geismar 3 at 1.8 Mt/yr, forty times Kassø's output.

Hydrogen is 50–70% of the cost and CO2 is the next largest line, so an e-methanol plant is an electrolyzer with a synthesis loop attached, and its cost curve is the electrolyzer's cost curve. That means the levers are power price, capacity factor, and CO2 price rather than anything in the methanol reactor. The demand is regulatory: FuelEU Maritime requires a 2% cut in fleet GHG intensity from 2025 and 6% from 2030 against a 91.16 gCO2e/MJ baseline, and a shipowner will meet that at the lowest cost available. E-methanol competes for that compliance dollar against biomethanol, LNG, and pooling arrangements, all of which are cheaper today. If a business case assumes shipping pays a fuel premium rather than a compliance premium, it is counting revenue that does not exist.

Innovation Outlook: Renewable Methanol (IRENA) · Green Methanol from CO2 Hydrogenation at Industrial Scale: Progress, Challenges, and Perspectives (Chem & Bio Engineering)

E-kerosene is synthetic jet fuel built from hydrogen and CO2 by one of two routes. The Fischer-Tropsch route converts CO2 to CO by reverse water-gas shift or by co-electrolysis of steam and CO2, chain-grows the resulting syngas over an iron or cobalt catalyst into a synthetic crude, then hydrocracks and fractionates that crude into jet, diesel, and naphtha cuts. The methanol-to-jet route makes entry 021's product first and converts it through olefins into the kerosene range. Either way the finished fuel is certified under ASTM D7566 at blends up to 50% and burns in an unmodified engine. What makes this the hardest fuel on the sheet is the step count: electrolysis, carbon conversion, synthesis, upgrading, and fractionation, each with its own loss, stacked in front of a product that airlines currently buy for about $800/t.

Aviation has no electric substitute for long-haul flight, the fuel is a genuine drop-in, and airports need no new equipment, so the demand is durable in a way that hydrogen trucking's is not. Against that, roughly 40–50% of the input electricity reaches the finished liquid, the fuel costs several times fossil jet, and only part of a Fischer-Tropsch syncrude falls in the jet range, so the naphtha and diesel cuts have to find buyers or the jet fraction carries their cost. The failure mode is not the chemistry and not the penalty regime. It is final investment decision. Europe needs about 552,000 t of e-SAF a year to meet the 1.2% sub-quota averaged over 2030 and 2031, Europe has around 41 e-SAF projects at pre-FID totalling roughly 2.8 Mt of paper capacity, and the plants actually expected to be running by then add up to roughly 100–120 kt.

Build e-kerosene only if you have three things at once: biogenic CO2 under contract, cheap power at a high capacity factor, and an offtake that is a priced contract rather than an MoU. If any one of them is missing, make e-methanol instead. The front end is identical, you drop two conversion steps, and the buyer exists today. If you are an airline or an obligated fuel supplier, buy HEFA and other bio-based SAF first for the headline SAF target, since it costs a fraction as much and is available now, and treat e-SAF as a separate procurement problem driven entirely by the sub-quota, which has no carry-over and cannot be met with bio-based fuel. When diligencing a developer, the questions in order are: is the CO2 contracted, what is the power price and capacity factor, and has anyone signed a price.

About 0.43 kg of hydrogen and 3.1 kg of CO2 per kg of hydrocarbon at Fischer-Tropsch stoichiometry · roughly 40–50% of input electricity into finished liquid fuel · certified to 50% blend under ASTM D7566 · ReFuelEU sub-quota 1.2% averaged over 2030–31, 2% over 2032–34, and 5% from 2035 · the 2030–31 quota works out to about 552,000 t/yr against roughly 100–120 kt expected to be operating · Infinium's Roadrunner is sized at about 23,000 t/yr of e-fuel products.

Building the CH2 unit of a hydrocarbon chain from CO2 takes three hydrogens (CO2 plus 3 H2 gives CH2 and 2 H2O), which is 0.43 kg of hydrogen and about 3.1 kg of CO2 per kg of fuel. At 55 kWh/kg that hydrogen alone is roughly 24 kWh of electricity, and jet fuel holds 11.9 kWh/kg, so the theoretical ceiling is near 50%. Reverse water-gas shift, hydrocracking, product separation, and CO2 capture pull the delivered figure to 40–50%, and because only part of a syncrude lands in the jet range, the electricity per kilogram of certified kerosene is higher still unless the other cuts are sold.

Infinium's Project Roadrunner in Texas, about 23,000 t/yr of e-fuels with production due in 2027, whose fuel flew an American Airlines service from Corpus Christi to Dallas-Fort Worth in August 2026. Twelve's AirPlant One at Moses Lake, Washington, opened in June 2026 at an initial 50,000 gallons a year, backed by Alaska Airlines and Microsoft, which is a useful reminder of how small a "commercial-scale" first plant is. Norsk e-Fuel's Mosjøen project in Norway and Arcadia eFuels' Vordingborg plant in Denmark, permitted at roughly 80,000 t/yr. Topsoe, Johnson Matthey, and Sunfire supply the reverse-shift, Fischer-Tropsch, and co-electrolysis technology.

Cost is dominated by hydrogen and therefore by power price and capacity factor, with CO2 second and the upgrading train third. Fossil jet runs roughly $700–800/t, and every published e-kerosene estimate lands at a multiple of that, which is why no volume moves without the mandate. The revenue question is who bears the gap: fuel suppliers are the obligated parties under ReFuelEU, airlines pay through ticket prices, and member states set the penalties, so this is a policy-constructed market with a legislated volume and an unlegislated price. That is workable for a developer, but only with a signed contract-for-difference or a fixed-price offtake, because a project financed against a spot compliance value has no floor. The Commission's first ReFuelEU review is due by January 2027, and whether it holds the 1.2% sub-quota is the single biggest variable in any e-SAF business case.

SAF Conversion Processes (ICAO) · Power-to-Liquids: A Scalable and Sustainable Fuel Supply Perspective for Aviation (German Environment Agency)

E-methane is hydrogen and carbon dioxide run over a nickel catalyst at 250–400 °C to make methane and water, the Sabatier reaction, which has been industrial chemistry for a century and is used today to clean trace CO2 out of ammonia syngas. The reaction is strongly exothermic at about 165 kJ per mole, so the engineering problem is heat removal and temperature control rather than energy input, and the catalysts are cheap and well understood. Stoichiometry sets the inputs at 0.50 kg of hydrogen and 2.75 kg of CO2 per kg of methane. What comes out is indistinguishable from natural gas, which is the entire argument for the product: LNG carriers, import terminals, transmission pipelines, city gas networks, industrial burners, and domestic boilers all work unchanged, with no embrittlement, no appliance conversion, no meter recalibration, and no new ships. A biological variant exists too, where methanogenic archaea do the same conversion at 60–65 °C in a bioreactor, which trades reaction rate for tolerance of impure CO2 and intermittent hydrogen.

The strength is that e-methane needs no new infrastructure anywhere in the chain, which makes it the only synthetic fuel a gas utility can sell into its existing customer base without touching a single appliance. It also substitutes energy for energy, so 1% e-methane in a city gas network cuts 1% of the CO2, where 20% hydrogen by volume cuts only about 7%. The weakness is that it is the least efficient way to spend a kilogram of hydrogen. Methanation itself keeps about 83% of the hydrogen's energy, and that loss lands on top of electrolysis, CO2 capture, compression, and liquefaction, so the finished LNG carries maybe a third of the electricity that started the chain. The failure mode is accounting rather than engineering. The CO2 is captured in one country and released at a burner tip in another, and until the two governments agree on who books the reduction, the product's whole value proposition is a bilateral negotiation. If it is settled the wrong way, a buyer pays a large premium for a fuel that still counts as fossil gas on its own books.

Pick e-methane when the customer is a gas utility, an LNG importer, or an industrial site with thousands of installed burners, and when the alternative on the table is converting all of that to hydrogen. Under those conditions the comparison is not e-methane against natural gas, which it loses badly, but e-methane against the capital cost of rebuilding a distribution network, and it can win. Also pick it where a mandate is written in energy terms and the network cannot physically take much hydrogen. Do not pick it when the end user is a single large site that can be converted, because feeding that site hydrogen directly skips a 17% loss and a CO2 supply contract. And do not treat it as a competitor to hydrogen blending (entry 014): blending is cheap and delivers little, e-methane is expensive and delivers in full, and the choice depends on whether you are buying a headline or a compliance obligation.

0.50 kg of hydrogen and 2.75 kg of CO2 per kg of methane · Sabatier reaction exothermic at about 165 kJ per mole, run at 250–400 °C over nickel · retains about 83% of the hydrogen's energy · 1% e-methane in a gas network cuts 1% of the CO2, against about 7% for 20% hydrogen by volume · hydrogen alone costs roughly $29/MMBtu at $3/kg and $48/MMBtu at $5/kg · Japanese city gas utilities target 1% e-methane by 2030 and 90% by 2050.

One kg of methane needs 0.50 kg of hydrogen carrying 16.8 kWh, and the methane holds 13.9 kWh, so methanation keeps about 83% and rejects the rest as heat at 250–400 °C, which is hot enough to raise steam if there is a use for it on site. That 17% comes after everything upstream: at 55 kWh/kg for electrolysis the hydrogen already cost 27.7 kWh of electricity per kg of methane, so the fuel returns roughly half the input electricity before CO2 capture, compression, or liquefaction are counted, and a chain that liquefies the product for shipping lands nearer a third. Every other end use of hydrogen in this sheet is more efficient than this one, which is why the case for e-methane has to be made on infrastructure cost rather than on energy.

Osaka Gas and Inpex began blending synthetic methane into Inpex's pipeline network in central Japan in February 2026, from a 400 Nm3/h methanation unit at the Koshijihara field. The ReaCH4 project pairs Tokyo Gas, Osaka Gas, Toho Gas, Mitsubishi, and Sempra Infrastructure on roughly 130,000 t/yr of e-methane near Cameron LNG in Louisiana for 2030. The Live Oak project in Nebraska brings TotalEnergies, TES, Osaka Gas, Toho Gas, and ITOCHU together on a similar model, with an investment decision expected in 2027. Audi's e-gas plant at Werlte in Germany was the long-running early demonstration at 6 MW. Electrochaea is the reference for the biological route.

The cost is the hydrogen, and the arithmetic leaves no room because methane is a cheap commodity sold by the MMBtu. At $3/kg the hydrogen inside a tonne of e-methane costs about $29/MMBtu and at $5/kg about $48/MMBtu, before the CO2, the methanation plant, liquefaction, or shipping, against LNG that has mostly delivered into Asia in the low tens of dollars per MMBtu. Nothing on the electrolyzer cost curve closes a gap that large this decade, so every project in the pipeline is underwritten by a Japanese utility buying compliance with a national target rather than by a customer buying gas. That makes the diligence questions unusual: who is contractually obliged to take the volume, at what price, for how long, and which country's inventory the emission reduction lands in. If the answer to the last one is unresolved, the offtake is not financeable regardless of what the plant costs.

5 minCatalytic Methanation Converts CO2 to CH4 (Methane)APMonitor.com · 10k+ views

2 minMaking Methane out of Hydrogen and Carbon Dioxide // Team ASAPTeam ASAP · 10k+ views

10 minHow we produce e-NG with methanation | Jens SchmidtTES · 1k+ views

Gas 2025: Analysis and Forecasts to 2030 (IEA) · Research Progress of Non-Noble Metal Catalysts for Carbon Dioxide Methanation (Molecules)

Refinery hydroprocessing is the largest single use of hydrogen in the world, taking roughly 43 Mt a year out of a total near 100 Mt. It covers two related families of unit. Hydrotreating passes a feed and hydrogen over a cobalt- or nickel-molybdenum catalyst at 300–400 °C and 30–100 bar, where hydrogen strips sulfur out as H2S and nitrogen as ammonia, and saturates olefins and aromatics; it is what makes a diesel meet a 10 ppm sulfur spec. Hydrocracking runs hotter and much higher, 350–450 °C and 100–200 bar over a bifunctional acid-and-metal catalyst, and breaks heavy vacuum gas oil into diesel and jet, consuming far more hydrogen in the process, on the order of 300–400 Nm3 per cubic meter of feed against 50–150 for a diesel hydrotreater. Refineries have run these units for decades and make about half their own hydrogen in dedicated reformers, taking most of the rest as a byproduct of catalytic naphtha reforming. That existing plumbing is why refining is the most realistic first market for low-carbon hydrogen: every compressor, pipe, and permit downstream of the hydrogen header already exists.

As a demand sink this is about as good as it gets. The volumes are enormous, the consumption is continuous, the purity requirement is modest compared to a fuel cell, and a refinery buying clean hydrogen changes nothing about its operation. The weakness is willingness to pay. A refiner's alternative is its own steam methane reformer, so the internal transfer price sits near $1–2/kg at typical US gas prices, and no clean production route in this sheet reaches that number without a subsidy. The failure mode for a supplier is treating a refinery's hydrogen appetite as a market when it is really a make-or-buy decision at a price the refinery already knows to three decimal places. The second weakness is direction of travel: refining hydrogen demand is not a growth market, because gasoline demand is flat to falling in most OECD countries and refineries close.

If you are siting a large electrolyzer or a blue hydrogen plant and want an offtaker that will take steady volume from day one, a refinery is the first place to look, particularly one with a hydrocracker rather than only hydrotreaters. Sell into it when a policy makes the arithmetic work: the EU's RED III industrial hydrogen targets, California's Low Carbon Fuel Standard, or the US 45V credit can each lift what the refinery will pay well above its reformer's cost. Do not build a project whose base case assumes a refinery pays a voluntary premium, because it will not. And check what the site is doing with renewable diesel and SAF co-processing, since those feeds consume several times more hydrogen per tonne than fossil diesel and are the one part of refining demand that is genuinely growing. The alternative sink to compare against is ammonia synthesis (entry 025), which takes similar volumes but is far less tolerant of a variable supply.

Roughly 43 Mt of hydrogen a year, the largest single sink · hydrotreating at 300–400 °C and 30–100 bar over CoMo or NiMo catalyst · hydrocracking at 350–450 °C and 100–200 bar · hydrocracking consumes 300–400 Nm3 of hydrogen per cubic meter of feed, against 50–150 for a diesel hydrotreater · refineries make about half their hydrogen on purpose and take much of the rest from catalytic reforming · willingness to pay $1–2/kg unless a fuel standard pays more · Shell's Holland Hydrogen 1 at 200 MW makes 60 t/day.

Hydrogen is a reagent here rather than a fuel, so its energy is not lost: most of it ends up chemically bound in the finished product, raising the fuel's hydrogen-to-carbon ratio, and the rest leaves as H2S, ammonia, and water. What a refiner buys is typically 15–30% more than the stoichiometric requirement, because hydrogen dissolves in the liquid product and because the recycle loop has to be purged to stop light ends from accumulating, which makes purge-gas recovery a standard piece of refinery hydrogen management. Hydrocracking is exothermic once the hydrogen reacts, so the units inject cold quench hydrogen between catalyst beds to hold temperature rather than firing a heater, and that quench flow is part of why hydrocrackers dominate a refinery's hydrogen balance.

Shell's Holland Hydrogen 1 at Tweede Maasvlakte in Rotterdam is the reference project: 200 MW of thyssenkrupp Uhde alkaline electrolysis making about 60 t/day, or 21,900 t/yr, feeding the 404,000 b/d Pernis refinery from late 2026. Shell's REFHYNE program at the Energy and Chemicals Park Rheinland ran a 10 MW PEM unit from 2021 and is finishing a 100 MW PEM unit. Air Products and Linde are the merchant suppliers who own hydrogen plants inside refinery fences along the US Gulf Coast and sell over the fence. On the demand-growth side, Neste's Singapore and Rotterdam renewable diesel plants and TotalEnergies' La Mède and Grandpuits conversions are where refining hydrogen consumption is actually rising.

The price of hydrogen inside a refinery is set by the refinery's own reformer, which means it tracks natural gas: near $1/kg at $3/MMBtu and closer to $3/kg at $10/MMBtu. That is the number every supplier is bidding against, and it is the reason refining shows up as a large but low-value sink. Policy is what moves it. A Low Carbon Fuel Standard credit or a production tax credit can add a dollar or two per kg to what the buyer will pay, and every announced refinery electrolyzer sits on top of one of those. Scale is the other thing worth keeping in view: Holland Hydrogen 1 is the largest electrolyzer in Europe and makes 60 t/day, while world refining consumes roughly 118,000 t/day, so it would take about 2,000 such plants to cover the sector. Refining is the easiest market for clean hydrogen to enter and the slowest to move volume.

6 minHydrocracking Unit - How it WorksIcarus-ORM Academy · 50k+ views

3 minHydrotreatment - Section Overview (Lecture 122)Chemical Engineering Guy · 5k+ views

3 minH2 academy: hydrodesulphurization (HDS) overviewJohnson Matthey · 1k+ views

U.S. Gulf Coast refinery demand for hydrogen increasingly met by merchant suppliers (US Energy Information Administration) · Global Hydrogen Review 2025 (IEA)

Haber-Bosch is the second-largest hydrogen sink in the world, consuming about 32 Mt a year to make roughly 180 Mt of ammonia, most of which becomes fertilizer. Nitrogen from an air separation unit and hydrogen are compressed to 150–300 bar, heated to 400–500 °C, and passed over a promoted iron catalyst; only 15–20% converts on each pass, so the unreacted gas is cooled, the ammonia condensed out, and the rest recycled, which is why the synthesis loop rather than the reactor is the defining piece of hardware. Entry 019 covers ammonia as a shipped energy carrier; this entry is about the plant as a customer for hydrogen. That customer is unusually attractive because the swap is clean: take out the reformer, put in an electrolyzer, and every compressor, catalyst bed, condenser, and storage tank downstream stays exactly as it is. Nothing else in this sheet offers a drop-in that simple at 32 Mt of scale.

The strength is that the demand already exists, at a single point, in industrial quantity, with no new downstream equipment and no new customer to convince. The weakness is that the loop needs steady state and renewables do not supply it. Iron catalyst and a 150–300 bar loop are happiest at constant rate; conventional plants turn down to perhaps 50–60% and ramp in percent per hour, and thermal cycling shortens catalyst life. That leaves three ways to bridge a variable electrolyzer to a steady loop, and all three cost money: buffer hydrogen storage, oversized electrolysis running at partial load, or a smaller loop designed for 10–20% minimum load. This is the specific reason green ammonia costs two to four times gray, and it is not a learning-curve problem that goes away with scale. The failure mode is a project that sizes electrolysis on average renewable output and discovers the loop needs a floor it cannot hold overnight.

If you are looking for the first large market a clean hydrogen plant can sell into, ammonia synthesis is the strongest candidate on technical grounds and the hardest on price. Pick it when you have firm cheap power or enough storage to hold a floor, and when the buyer faces a carbon cost on the product: CBAM on fertilizer imported into Europe, or a Japanese or Korean co-firing mandate, both of which pay for the carbon rather than the nitrogen. Do not build against the fertilizer market alone, because urea buyers price on nutrient content and will not pay a premium. If your power is intermittent and you have no storage, refinery hydroprocessing (entry 024) is the more forgiving offtaker, since a hydrotreater can take what it gets and top up from the existing hydrogen header. And if the plan is to ship the product rather than sell it locally, read entry 019 first, because the export chain changes the economics more than the synthesis loop does.

About 32 Mt of hydrogen a year making roughly 180 Mt of ammonia · 178 kg of hydrogen per tonne of ammonia · synthesis at 150–300 bar and 400–500 °C over promoted iron, 15–20% conversion per pass · best gas-based plants run 28–32 GJ per tonne against a thermodynamic floor near 21, world average nearer 46 · direct emissions about 450 Mt CO2 a year, roughly 1.3% of the global total · green ammonia costs two to four times gray · Yara's Herøya unit is 24 MW making 20,000 t/yr.

A modern gas-fed plant uses 28–32 GJ per tonne of ammonia, roughly 7.8–8.9 MWh, most of it as feedstock methane rather than fuel, and the world average is nearer 46 GJ because a large share of Chinese capacity runs on coal. Build the same tonne on electrolysis and it needs 178 kg of hydrogen at 50–55 kWh/kg, which is 8.9–9.8 MWh, plus roughly 0.5–1.0 MWh for the air separation unit and the loop compressors, so call it 9.5–10.8 MWh of electricity per tonne. The number that actually decides the project is not that total but the hours behind it: a loop running 4,000 hours a year instead of 8,000 doubles the capital cost per tonne, which is why buffer storage and oversized electrolysis appear in every green ammonia budget and why the flexibility problem shows up as a cost rather than an outage.

Yara's unit at Herøya Industrial Park in Norway is the cleanest small reference: 24 MW of electrolysis making about 10 t/day of hydrogen for 20,000 t/yr of renewable ammonia inside an existing plant. Fertiglobe's Egypt Green Hydrogen project at Ain Sokhna feeds electrolytic hydrogen into two operating ammonia plants, backed by 260 MW of solar and wind, with a 15 MW pilot running since 2022 and first export cargo in 2023. Envision's Chifeng base in Inner Mongolia is the largest built, shipping its first green ammonia cargo to LOTTE Fine Chemical in Korea in February 2026. The NEOM Green Hydrogen Project in Saudi Arabia, over 90% complete and targeting commercial production in 2027, is the largest under construction, with Air Products offtaking and Yara marketing part of the volume. Casale, Topsoe, and thyssenkrupp Uhde supply the flexible loop designs these projects depend on.

Gray ammonia's cost is roughly 60–80% natural gas, so the incumbent's price tracks the gas curve and a green project is really betting on the spread between power and gas in a particular place. Electrolysis dominates green capex, but the flexibility hardware is the line that surprises people: buffer storage, oversizing, and a loop rated for deep turndown together add a substantial fraction to a plant that would otherwise be an electrolyzer with a catalyst bed behind it. The demand pull is regulatory and regional. CBAM prices the carbon in fertilizer entering the EU, Japanese and Korean power co-firing mandates create an ammonia buyer with no interest in nitrogen at all, and India's SIGHT auctions subsidize domestic production directly. The useful diligence question is not the levelized cost of hydrogen but the delivered cost per tonne of ammonia against the importer's landed gray price plus whatever carbon charge applies, since that is the comparison the buyer actually runs.

11 minThe Haber-Bosch Process: Industrial Ammonia SynthesisProfessor Dave Explains · 10k+ views

3 minAmmonia synthesis How does it worktrkuwait · 100k+ views

1 minHow To Produce Green AmmoniaREVOLVE · 10k+ views

Ammonia Technology Roadmap: Towards More Sustainable Nitrogen Fertiliser Production (IEA) · Operating Envelope of Haber-Bosch Process Design for Power-to-Ammonia (RSC Advances)

Hydrogen direct reduction strips the oxygen out of iron ore with hydrogen instead of carbon, in a shaft furnace at 800–950 °C, and the product is solid sponge iron that goes to an electric arc furnace to become steel. The chemistry is the reason this is the strongest new industrial demand in the sheet: reducing hematite takes 54 kg of hydrogen per tonne of iron at stoichiometry, and 55–70 kg/t in a working furnace, so hydrogen is a reagent doing a job nothing else can do at that temperature rather than a fuel competing with cheaper fuels. The equipment is not new either. Midrex and Tenova's Energiron have licensed gas-based shaft furnaces for decades, and those furnaces already run on syngas that is more than half hydrogen, so the step to 100% is an engineering change rather than an invention. One difference matters. Reduction by carbon monoxide is mildly exothermic and reduction by hydrogen is endothermic at about 99 kJ per mole of Fe2O3, so a hydrogen furnace has to put heat in that a gas-fed one does not.

The strength is that this is real demand with real chemistry behind it, at a scale that would move the hydrogen market: steel is around 7% of global CO2, and a single 2.4 Mt/yr plant consumes more than 100,000 tonnes of hydrogen a year. The first weakness is power. That hydrogen needs 600–700 MW of electrolysis running essentially flat out, which means firm cheap electricity, which means a short list of places. The second weakness is ore, and it gets less attention: shaft furnaces need DR-grade pellets above about 67% iron with low gangue, and that grade is under 10% of world iron ore output, already fully consumed by the existing gas-based DRI fleet. The failure mode is a project that solves power and hydrogen and then cannot contract pellet at a price that works, or has to accept lower-grade feed and carry the gangue through an electric arc furnace that was not designed for it.

Build hydrogen DRI where you have firm renewable or nuclear power below roughly $30/MWh, a DR-grade pellet contract, and a customer who will pay a green premium under contract rather than in a press release. Northern Sweden and Canada have the first two; the Middle East has cheap power and pellet but weaker premium demand. If the power is not firm, look at natural-gas DRI first, which already makes well over 100 Mt a year, cuts emissions roughly 40–60% against a blast furnace on day one, and can be converted to hydrogen later when the fuel gets cheap enough. That intermediate step is what most of the industry is actually doing. Also weigh the electrolytic routes that skip hydrogen entirely, molten oxide electrolysis from Boston Metal and low-temperature electrowinning from Electra, which are less mature but avoid the reduction-gas loop altogether.

54 kg of hydrogen per tonne of iron at stoichiometry, 55–70 kg/t in a working shaft furnace · reduction endothermic at about 99 kJ per mole of Fe2O3, roughly 250 kWh of reaction heat per tonne of iron · shaft furnace at 800–950 °C · 2.8–3.9 MWh of electricity per tonne of iron for the hydrogen alone, plus 0.6–0.7 MWh for the electric arc furnace · DR-grade pellets above 67% iron are under 10% of world iron ore output · gas-based DRI already exceeds 100 Mt/yr · breakeven needs roughly $1.5–2.5/kg hydrogen plus a carbon price.

Because hydrogen reduction is endothermic at about 99 kJ per mole of Fe2O3, roughly 250 kWh per tonne of iron has to arrive as heat, normally by preheating the reducing gas to 800–950 °C, and that heat load is the main process difference from a gas-fed shaft furnace. The hydrogen itself dominates the energy bill: 55–70 kg per tonne of iron at 50–55 kWh/kg is 2.8–3.9 MWh of electricity, and the electric arc furnace behind it adds another 0.6–0.7 MWh per tonne of steel, so a hydrogen steel plant is roughly a 4 MWh/t electricity consumer. Excess hydrogen leaves the furnace mixed with steam, and the top gas is cooled, the water knocked out, and the hydrogen recycled, which is what holds consumption near stoichiometry instead of several times it.

Stegra's plant at Boden in northern Sweden is the flagship: 2.4 Mt/yr of steel on more than 100,000 t/yr of hydrogen, 37 alkaline electrolysis modules from thyssenkrupp Nucera all installed, over EUR 6 billion raised, and first deliveries expected in 2027. HYBRIT, the SSAB, LKAB, and Vattenfall venture, made the first hydrogen-reduced sponge iron at its Luleå pilot in 2021 and is building a demonstration plant in Gällivare. On the other side of the ledger, ArcelorMittal canceled or postponed its German DRI conversions in 2025 and thyssenkrupp Steel pulled back on its Duisburg hydrogen plans, both citing power and hydrogen cost. Midrex and Tenova are the shaft-furnace licensors every one of these projects depends on. Boston Metal and Electra are the electrolytic alternatives that skip hydrogen.

A DRI-EAF plant costs on the order of a billion euros per megatonne of capacity before the electrolyzers, and the electrolyzers can add as much again, so these are project-finance-scale bets that need a decade of contracted power and offtake to close. Hydrogen is the dominant operating cost once built, and the breakeven against a blast furnace lands near $1.5–2.5/kg with a carbon price behind it, which is why the European projects tracked EU ETS prices and CBAM design more closely than they tracked electrolyzer quotes. The demand-side question is whether a green steel premium exists as a contract. Stegra's multi-year supply agreement with thyssenkrupp Materials Services is the kind of evidence that counts; a memorandum of understanding with an automaker is not. If you are underwriting a supplier into this market, the sequence to watch is power contract, then pellet contract, then offtake, because the projects that died in 2025 died on the first of those.

4 minHow a direct reduction plant works – explained in 5 minutesthyssenkrupp Steel · 10k+ views

10 minGreen Steel From HydrogenEngineering with Rosie · 10k+ views

4 minThe MIDREX® ProcessMidrex Technologies · 10k+ views

15 Insights on the Global Steel Transformation (Agora Industry and Wuppertal Institute) · Green Ironmaking at Higher H2 Pressure: Reduction Kinetics and Microstructure Formation During Hydrogen-Based Direct Reduction of Hematite Pellets (Journal of Sustainable Metallurgy)

Burning hydrogen for industrial process heat covers the temperatures that electricity and biomass struggle with: cement clinker at about 1,450 °C, glass tanks at 1,500–1,600 °C, and steel reheat and forging furnaces at 1,100–1,300 °C. Reaching those temperatures is not the difficulty. Hydrogen's adiabatic flame temperature in air is around 2,100 °C against methane's 1,960 °C, so a hydrogen burner has margin to spare. The difficulty is that hydrogen burns differently in three ways that all matter to a furnace designer. Its laminar flame speed is roughly 2.9 m/s against methane's 0.38, which makes flashback into the burner a design constraint rather than an edge case. Its flame is nearly non-luminous because there is no soot, so radiant heat transfer to the load drops, and radiation is the dominant mode in glass and cement. And it makes about 1.7 times more water vapor per unit of energy while producing less total flue gas, which changes convective transfer, refractory chemistry, and NOx formation together.

The strength is genuine: hydrogen is the only clean fuel that can be dropped into a fired furnace and reach 1,500 °C with a combustion system that resembles the one already there. The weakness is that it competes on price with natural gas, and gas at $8–12/MMBtu is the energy equivalent of hydrogen at roughly $1.1–1.6/kg, which is below the plant-gate cost of every clean production route in this sheet, including coal gasification. That gap is not closing this decade. The second weakness is that a burner swap is not a fuel swap: flame speed, flame length, radiant profile, and NOx all change, so the furnace has to be re-engineered and in a glass tank or cement kiln the refractory lining may not survive the new heat distribution. The failure mode worth naming is a cement project sold as decarbonization: roughly 60% of a cement plant's CO2 comes from calcining limestone, so even a kiln fired entirely on hydrogen removes only about 40% of the emissions.

Use hydrogen for heat when you already have it inside the fence. A refinery, chlor-alkali plant, or ammonia complex with surplus byproduct hydrogen (entry 011) is burning it for process heat today at close to zero marginal cost, and that is a real and sensible application. Beyond that, the test is whether an electric route physically works on your product. If resistive, induction, plasma, or an electric melter can reach the temperature and touch the material without contaminating it, electricity wins on energy alone, because 50–55 kWh of electricity becomes 33.3 kWh of hydrogen heat while a resistance element delivers essentially all of it. Thermal storage systems from Rondo and Antora now deliver 1,000–1,500 °C from cheap off-peak power and take that argument further. If neither electricity nor hydrogen is available, biomethane is usually the cheaper drop-in. Reach for hydrogen when the process needs a flame, the site has no clean gas, and someone is paying a carbon price large enough to cover a dollar or more per kg of premium.

Cement clinker at about 1,450 °C, glass tanks 1,500–1,600 °C, reheat furnaces 1,100–1,300 °C · hydrogen's adiabatic flame temperature about 2,100 °C against methane's 1,960 · laminar flame speed about 2.9 m/s against methane's 0.38 · natural gas at $8–12/MMBtu is the energy equivalent of hydrogen at roughly $1.1–1.6/kg · about 1.7 times more water vapor per unit of energy than methane · roughly 60% of cement CO2 is process emissions no fuel switch touches · 50–55 kWh of electricity delivers 33.3 kWh of hydrogen heat.

The burner itself is efficient, but the chain feeding it is not: 50–55 kWh of electricity per kg of hydrogen comes back as 33.3 kWh of heat, about 60%, where a resistance element delivers essentially all of the electricity as heat and an industrial heat pump delivers more than it consumes at lower temperatures. Stack losses shift as well, since hydrogen produces roughly 1.7 times more water vapor per unit of energy than methane and a 1,400 °C furnace never condenses it, so more of the fuel's heating value leaves as latent heat. Against a fired natural gas furnace hydrogen is roughly a wash on delivered heat per unit of fuel energy, so the whole comparison is decided upstream at the point where electricity became hydrogen.

Hanson's Ribblesdale cement works in Lancashire, now part of Heidelberg Materials, ran the first cement kiln trial on a fuel mix including hydrogen in 2021, in a UK government-backed program. NSG Pilkington fired a float glass furnace at St Helens on hydrogen the same year and made architectural glass from it, and Encirc, working with Glass Futures and Diageo, made hydrogen-fired glass bottles. On the alternatives side, Rondo Energy and Antora Energy sell thermal storage that stores off-peak electricity as heat in brick or carbon blocks and delivers it above 1,000 °C, and Sublime Systems and Brimstone are attacking cement's process emissions directly rather than its fuel. Air Liquide, Linde, and Honeywell UOP supply the industrial hydrogen burners.

This is the end use with the weakest economics in the sheet, and the reason is straightforward: process heat is a commodity bought per unit of energy, and hydrogen is the most expensive way to deliver a joule that any of these plants has ever been offered. There is no cost structure story that fixes it, because the competition is a fuel that costs a tenth as much and an electric route that uses a third as much primary energy. What can change the answer is regulation with a specific shape. A carbon price above roughly $150/t starts to close the gap against gas in cement and glass; a mandate that a share of industrial fuel be non-fossil closes it by fiat. Anyone building a business on hydrogen burners should be selling to sites that already have the molecule, and should treat greenfield industrial heat demand as an option on carbon policy rather than a market.

1 minSimulating Flashback in a Hydrogen Swirl Burnerconvergecfd · 1k+ views

Industrial Decarbonization Roadmap (US Department of Energy) · Alternative Fuel Switching Technologies for the Glass Sector: Phase 3 (UK Department for Business, Energy and Industrial Strategy)

A fuel-cell truck carries hydrogen at 700 bar or as a cryogenic liquid, runs it through a PEM stack that converts 50–60% of its energy into electricity, and drives an electric motor through a small buffer battery that handles peaks and regenerative braking. The stack is the same technology as a PEM electrolyzer run backwards, which is why the two share a supply chain and a platinum-loading problem. A Class 8 truck uses roughly 7–9 kg of hydrogen per 100 km, so a 60–70 kg tank set gives 300–600 miles, and refueling takes 10–20 minutes against five for diesel. That combination is the pitch: battery-like driveline, diesel-like range and turnaround, no depot dwell time. The reason the pitch has worked for buses and not for trucks comes down to where the vehicle sleeps. A bus fleet returns to one depot every night, so a single dispenser serves fifty vehicles and runs at high utilization; a long-haul truck needs a station network that nobody has built.

On the vehicle itself the engineering works. Hydrogen storage is lighter than the equivalent battery, so a fuel-cell truck keeps more payload on long routes, and it refuels fast enough not to disturb a driver's hours of service. The weakness is the fuel price and the station network, and they reinforce each other: few stations means low throughput, low throughput means high prices, and high prices mean nobody buys the trucks. California retail hydrogen has been near $30/kg while the economics need roughly $4–6/kg, a gap of five times rather than a few percent. Durability is the third issue: the US Department of Energy's heavy-duty target is 25,000 stack hours, about a million miles at highway speed, and fielded stacks are still working toward it against diesel engines that reach that routinely. The failure mode is a company that raised capital on truck orders and discovered its customers could not fuel them. Nikola filed for bankruptcy in February 2025 and Hyzon's board voted to liquidate the same month.

Pick fuel-cell if the fleet returns to one or two depots nightly, runs a predictable daily distance above what a battery comfortably covers, and cannot afford the dwell time or the grid connection that megawatt charging needs. Transit buses fit that description exactly, and so do some regional distribution fleets and refuse operations. Do not pick it for long-haul over routes with no station, and do not model the fuel at a price your supplier has not contracted, because retail hydrogen and contracted hydrogen differ by a factor of three or more. Compare against battery-electric first in almost every case: energy per mile is roughly two and a half times cheaper, megawatt charging is arriving, and the depot grid upgrade is usually cheaper than a hydrogen station. If you want hydrogen's fuel logistics but not the fuel cell's capital cost, entry 030 covers hydrogen internal combustion, which trades 20–40% more fuel for a much cheaper powertrain.

PEM stack systems run 50–60% efficient · storage at 700 bar or as liquid hydrogen · roughly 7–9 kg of hydrogen per 100 km for a Class 8 truck · 300–600 miles of range and 10–20 minute refueling · about 30% of the input electricity reaches the wheels, against 70–80% for battery-electric · economics need roughly $4–6/kg against California retail near $30/kg · DOE heavy-duty stack durability target of 25,000 hours.

Trace the electricity and the case gets hard: 50–55 kWh makes a kg of hydrogen, 3–4 kWh more compresses it to 700 bar, the fuel cell converts 50–60% of the 33.3 kWh in that kg, and the drivetrain takes another 10%, so roughly 30% of the original electricity reaches the wheels. A battery-electric truck delivers 70–80% of what it charges. At $0.15/kWh that is about $0.30 per mile for the battery truck against $0.77 for hydrogen at $6/kg, or $3.90 at a $30/kg retail pump, with diesel at 6.5 mpg and $4/gal sitting near $0.62. Nothing about stack efficiency closes a factor-of-two-and-a-half gap in energy, so the fuel-cell case has to be made on refueling time, payload, and depot infrastructure cost.

Hyundai's XCIENT Fuel Cell is the most-driven heavy fleet in the world: 165 trucks across five European countries had covered 20 million km by January 2026, starting from the Swiss Hyundai Hydrogen Mobility program, with a new North American version launched in 2025. Toyota and Kenworth ran the Port of Los Angeles drayage demonstration and Toyota now supplies powertrains to Hyroad, which bought 117 trucks and the IP out of Nikola's bankruptcy auction in 2025. Buses are where the volume is: Skyworth delivered 249 fuel-cell buses to Guangzhou Public Transport, Bologna's TPER is deploying 127, and Solaris expanded hydrogen bus production for German, French, and Scandinavian orders, with Asia-Pacific taking roughly two-thirds of the global market. Ballard Power and Toyota supply most of the stacks that go into other people's buses.

The vehicle costs more than a diesel and much more than a battery-electric equivalent, but capital cost is not what decides this market, because at $30/kg the fuel alone outweighs the entire purchase price difference within a year of heavy use. The cost structure that matters is the station's: a hydrogen dispenser is expensive and its price per kg falls steeply with throughput, so the only configurations that work are ones where many vehicles share one dispenser. That is why buses succeed and long-haul trucking does not, and it is a structural feature rather than a stage on a cost curve. Anyone building here should sell into captive depot fleets with a delivered-hydrogen contract in hand and treat the retail station network as somebody else's problem. Subsidies decide the rest: California's HVIP and LCFS, the EU's Alternative Fuels Infrastructure Regulation, and Chinese municipal fleet programs are what every deployment in the examples above is standing on.

10 minHow to build a Hydrogen Fuel Cell Semi Truck? - Hydrotec Factory TourJerryRigEverything · 100k+ views

13 minFuel cell stack explainedPragma Industries · 100k+ views

8 minHow Do Hydrogen Fuel Cells Work?Reactions · 500k+ views

Fuel Cell Electric Bus Status Report 2025 (National Laboratory of the Rockies) · Hydrogen Class 8 Long Haul Truck Targets (US Department of Energy Hydrogen Program)

Material handling is the one hydrogen mobility market that has made money, and the reason has nothing to do with emissions. A high-throughput warehouse running three shifts has to keep every lift truck moving, and a lead-acid battery gets in the way twice: swapping it costs 15–20 minutes of paid operator time, and its voltage sags as it discharges, so lift and travel speeds fall through the shift. A fuel-cell power unit drops into the same battery compartment, refuels in two to three minutes at a dispenser on the warehouse floor, and holds constant power until the tank is empty. It also frees the battery room, which in a large distribution center is hundreds of square feet of conditioned space plus two or three spare packs for every truck. Plug Power built the business on that argument, deploying more than 60,000 fuel cell systems and over 180 fueling stations, with more than 17,000 units at over 80 Amazon fulfillment centers.

The strength is that the customer is buying operator minutes and floor space rather than energy, which is why these fleets tolerate $6–10/kg hydrogen that no truck fleet or power plant would touch. The weakness is that the comparison has moved. Lithium-ion lift trucks charge opportunistically during breaks, need no swap, no battery room, and no hydrogen infrastructure at all, and they have taken back much of the market the fuel cell won from lead-acid. That leaves the fuel-cell case intact only at the top end: very large sites, three shifts, hundreds of trucks, and enough throughput to justify on-site liquid storage and a dispenser network. The failure mode is a site sized on a hydrogen price the supplier cannot sustain. Plug Power's fuel business has run at negative margin for years while it built out production, so a customer signing a long fuel contract should ask what happens when that subsidy stops.

Choose fuel cells when the site runs two or three shifts, has more than roughly a hundred trucks, and is short of floor space, and when you can get delivered hydrogen under $10/kg. Under those conditions the payback comes from labor and space, and it is a real payback that operators have been renewing for fifteen years. Choose lithium instead at single-shift sites, at smaller fleets, and anywhere the electrical service is already adequate, because it avoids the whole hydrogen supply chain for a similar operational result. The same reasoning extends off-road: port terminal tractors and yard equipment, airport ground support, and mine haul trucks all run long duty cycles in a fixed area with a central fueling point, which is the configuration hydrogen suits. Do not extend it to equipment that works away from a depot, since that is the same station problem that has held back on-road trucking.

Refueling in two to three minutes against 15–20 minutes for a battery swap · roughly 1.5–2.5 kg of hydrogen per truck per shift · these fleets tolerate $6–10/kg · a shift's hydrogen costs 85–140 kWh of electricity to make and compress, against 25–35 kWh from the wall for a lithium truck · Plug Power has deployed more than 60,000 fuel cell systems and over 180 fueling stations · more than 17,000 units at over 80 Amazon fulfillment centers · Amazon's Aurora site runs a 1 MW electrolyzer fueling over 225 trucks.

Efficiency is the one thing that does not decide this market. Making and compressing a shift's 1.5–2.5 kg takes 85–140 kWh of electricity, the stack converts about half of the 33.3 kWh in each kg, and the truck ends up doing roughly the same work a lithium truck does on 25–35 kWh from the wall, so the hydrogen route uses about three times the electricity. Warehouses pay it because the alternative costs are measured in operator minutes and square feet rather than kilowatt-hours: two swaps a day at 15–20 minutes each is over half an hour of paid time per truck, and a battery room in a large distribution center occupies space that would otherwise hold inventory. On-site electrolysis, as at Amazon's Aurora facility, changes where the losses happen but not their size.

Plug Power's GenDrive units at Amazon, Walmart, Home Depot, and Kroger distribution centers are the fielded base, with GenFuel dispensers and on-site liquid hydrogen storage as the standard site build, and BMW has run them at Spartanburg since 2010. Amazon's Aurora, Colorado fulfillment center added a 1 MW electrolyzer to make its own hydrogen for over 225 trucks. Toyota Material Handling and Hyster-Yale, through its Nuvera subsidiary, are the OEMs building fuel-cell lift trucks directly. Off-road, Anglo American's nuGen program converted a 290 t haul truck at Mogalakwena, and Toyota's port tractor work at the Port of Los Angeles is the terminal-equipment reference. The competing product is any of the lithium lift trucks from Toyota, Crown, and Jungheinrich sold with opportunity charging.

This is the only entry in the sheet where the customer's willingness to pay is set by labor rates rather than by an energy price, which is why it supports $6–10/kg while every other transport application needs $4–6 or lower. The cost structure at a site is dominated by the fueling infrastructure, not the power units: liquid storage, a vaporizer, compression, and dispensers, which is why the economics only close above roughly a hundred trucks and why suppliers sell a site package rather than a product. The supplier side has been much harder than the customer side. Plug Power spent years selling hydrogen below its own cost to keep the fleets fueled, which is a warning about how thin the margin is once you own both the equipment and the molecule. If you are building here, the question to answer first is whether the fuel can be delivered profitably at the price the warehouse will pay, because the equipment economics have been proven and the fuel economics have not.

4 minPlug Power GenDriveplugpower09 · 10k+ views

2 minGenFuel hydrogen refueling for fuel cell forkliftsPlug Power · 10k+ views

3 minBMW: Using Plug's GenDrive Fuel Cells since 2010 - part 1Plug Power · 10k+ views

An Evaluation of the Total Cost of Ownership of Fuel Cell-Powered Material Handling Equipment (NREL) · Work Efficiency and Economic Efficiency of Actual Driving Test of Proton Exchange Membrane Fuel Cell Forklift (Molecules)

A hydrogen internal combustion engine is a diesel or gas engine block with a new fuel system, ignition, and control strategy, burning hydrogen with air the way it would burn methane. Brake thermal efficiency lands at 35–45%, against 50–60% for a fuel cell, but the engine costs a fraction as much, uses the same castings, bearings, and turbochargers as the diesel it came from, and can be serviced by any mechanic who has worked on one. Two injection choices define the design. Port injection puts hydrogen into the intake manifold at around 10 bar, which is simple and cheap but lets the gas displace roughly a third of the intake air, cutting power density; direct injection at high pressure recovers it and is what Cummins uses on the X15H. Combustion runs lean, with substantial excess air, which cuts flame temperature and therefore thermal NOx, though not to zero. The result is a powertrain that is worse than a fuel cell on fuel and much better on capital cost, supply chain, and fuel purity.

The strengths are practical rather than thermodynamic. The engine tolerates hydrogen that would poison a PEM stack, which matters when the supply is byproduct gas that would otherwise need expensive cleanup to reach ISO 14687 fuel-cell grade. It keeps the existing manufacturing base, dealer network, and service skills, and it costs far less to buy. The weaknesses are fuel consumption and combustion behavior. Burning 20–40% more hydrogen per mile than a fuel cell is a permanent operating penalty. Hydrogen's minimum ignition energy is about 0.02 mJ against gasoline's 0.28, and its flammability range is 4–75% in air, so a hot exhaust valve, a carbon deposit, or a film of lubricating oil can light the charge before the spark does. Pre-ignition and backfire into the intake are the classic failure modes, and crankcase blow-by needs active ventilation because hydrogen accumulates there. NOx aftertreatment is still required to meet Euro VII or EPA 2027.

The deciding variable is annual utilization, not the technology. A fuel cell's 20–40% fuel saving has to pay back its capital premium, so the higher the annual hours or miles, the more the fuel cell wins; below that crossover the engine is cheaper over the life of the machine. Construction and agricultural equipment sits well below it, which is why JCB went this way, and long-haul line-haul trucking sits above it. Also pick the engine when the hydrogen you can get is impure, when the duty cycle involves shock loading and dust that a fuel cell's air handling would struggle with, or when the customer will not accept a service network that does not exist yet. Compare against entry 028 on total cost per mile at your actual mileage, and against a diesel with renewable fuel, which usually beats both on cost and loses only on tailpipe carbon.

Brake thermal efficiency 35–45%, against 50–60% for a PEM fuel cell · burns 20–40% more hydrogen per mile than a fuel cell · roughly 22–27% of the input electricity reaches the wheels, against about 30% for a fuel cell and 70–80% for a battery · hydrogen's minimum ignition energy about 0.02 mJ against gasoline's 0.28, flammability range 4–75% in air · port injection displaces about a third of the intake air · Cummins X15H rated 400–530 hp and 2,600 Nm at about 44% thermal efficiency · JCB's production engine is 55 kW.

An engine converts 35–45% of a kilogram's 33.3 kWh into shaft work where a fuel cell converts 50–60%, so a hydrogen engine burns 20–40% more fuel for the same job, and across the whole chain roughly 22–27% of the electricity that made the hydrogen reaches the wheels. What the engine avoids is purification loss and cost. A PEM stack needs hydrogen at ISO 14687 grade, with carbon monoxide below 0.2 ppm and sulfur below 0.004 ppm, and getting a byproduct or reformer stream to that specification is a real energy and capital charge; an engine will run on gas that never sees a final polishing step. On lean-burn calibration the engine also gives up some power density to hold NOx down, so the efficiency figures quoted at rated power are usually not the ones you get at the calibration a fleet actually runs.

JCB is the most committed developer: over GBP 100 million invested, a 55 kW production engine approved for commercial sale in Europe in January 2025, hydrogen backhoes and telehandlers in customer hands, and a twin-engine 1,600 hp land speed record car built from the same block. Cummins built hydrogen into its fuel-agnostic X-series platform, with the 15-liter X15H rated at 400–530 hp and 2,600 Nm, over 1,000 km of range on 700 bar storage, 15-minute refueling, and full production expected in 2027; Werner has signed for 500. Deutz sells the six-cylinder TCG 7.8 H2 at 200 kW and has paired it with RheinEnergie for stationary power. Toyota has run hydrogen-engine Corollas in Japan's Super Taikyu series since 2021. On water, MAN Energy Solutions, Wärtsilä, and Rolls-Royce mtu all offer hydrogen or hydrogen-blend versions of existing engine families.

This is a cheaper powertrain built from an existing supply chain, which changes the shape of the business rather than the size of the market. Development cost is modest compared with a fuel cell program because the block, the crank, and most of the accessories already exist and are already in volume production, so an engine maker can bring a hydrogen variant to market for a fraction of what a stack platform costs. That also means there is little proprietary advantage to accumulate: every major diesel manufacturer can do this, and several have. The market is therefore set by hydrogen price and by regulation, not by engine technology. In Europe, where CO2 standards for heavy vehicles count tailpipe emissions, a hydrogen engine qualifies as zero-emission and a renewable-diesel engine does not, which is most of the commercial argument for it. If that regulatory treatment changed, most of these programs would lose their business case overnight.

4 minJCB's new Hydrogen powered engine explained by Engineering Director, Ryan BallardContent With Media · 10k+ views

6 minJCB shows off its hydrogen internal combustion engineLandPowerTV · 10k+ views

2 minCummins hydrogen ICE truck powered by B6.7HCummins · 10k+ views

Overview of Hydrogen Internal Combustion Engine (H2ICE) Technologies (US Department of Energy) · Research and Development of Hydrogen-Fueled Internal Combustion Engines in China (ACS Omega)

The power sector is the demand sink most often named in hydrogen strategies and the one with the weakest willingness to pay, and one number explains why. Hydrogen holds 33.3 kWh/kg on a lower heating value basis, and any device that converts it to electricity at around 60% gives back about 20 kWh, so a megawatt-hour takes 50 kg of hydrogen no matter which machine you pick. At $4–6/kg delivered that is $200–300/MWh in fuel before a dollar of capital, several times what a combined cycle costs all in. Two conversion routes exist, gas turbines and stationary fuel cells, and the arithmetic above barely distinguishes them, which is the useful thing to understand about this market. The `power-generation` sheet owns the comparison between those machines and everything else on the grid, entry 023 for hydrogen-fired turbines and entry 032 for stationary fuel cells, with capacity credit, ramp rates, and LCOE against a CCGT. This entry asks the other question: at what delivered price does a generator buy hydrogen at all, and how many tonnes would that be.

The strength is that power is the only sink that could absorb hydrogen in tens of megatonnes if the price ever worked, and it needs no new molecule, no purity spec beyond what a turbine tolerates, and no chemistry. It is also the sink that pairs naturally with underground storage (entry 017), because a plant running a few hundred hours a year needs months of inventory and a salt cavern is the only place to keep it. The weakness is that electricity is a commodity sold by the MWh into a visible merit order, so a generator will not pay a premium it cannot recover, and its alternative is gas at $25–65/MWh of fuel depending on the region. The failure mode is a project that budgets hydrogen at production cost and forgets the 50 kg/MWh multiplier, which turns a $2/kg difference in delivered price into $100/MWh on the bid. Round-trip efficiency is the second problem: power to hydrogen to power runs about 30–35%, so two thirds of the electricity is gone before the plant sells anything.

Sell hydrogen into power only when the buyer has a reason other than energy cost. Those reasons exist and they are specific: a capacity market or scarcity price that pays per kW-year for a few hundred hours of firm zero-carbon output, a mandate like Japan's and Korea's co-firing targets, a data center that needs power on a date and cannot wait for an interconnection, or a site with byproduct hydrogen already inside the fence. Outside those, the generator needs roughly $1–2/kg to match gas plus a carbon price, and almost nothing in this sheet is produced in that band. If you are the generator rather than the supplier, buy hydrogen capability rather than hydrogen: a hydrogen-ready combustor costs little on a machine you were buying anyway and preserves the option. And read the `power-generation` sheet before choosing between a turbine and a fuel cell, because that decision turns on dispatch, permits, and lead time rather than on the fuel.

33.3 kWh/kg LHV, about 20 kWh of electricity per kg at 60% conversion, so 50 kg per MWh · at $4–6/kg that is $200–300/MWh of fuel alone · the generator needs roughly $1–2/kg to match gas plus a carbon price · 50% hydrogen by volume is only about 24% by energy · fielded F- and H-class combustors are certified for 30–50% by volume · PEM fuel cells 45–55% electrical, solid oxide 55–65%, in 100 kW to 10 MW modules · power to hydrogen to power round trip about 30–35%.

Making a kilogram takes 50–55 kWh of electricity and another 3–4 kWh to compress it into storage, and the generator gets about 20 kWh back, so the power-to-power round trip lands near 33% and closer to 30% once cavern injection, withdrawal, and plant parasitics are counted. A lithium battery returns 85–90% of what it takes, which is why hydrogen only competes for the duration a battery cannot serve rather than for daily cycling. On the demand side the multiplier is what matters: 50 kg/MWh means a 500 MW plant running at a 10% capacity factor burns about 22,000 tonnes of hydrogen a year and the same plant at baseload burns 110,000 tonnes, as much as an entire hydrogen steel plant. If 1% of world electricity came from hydrogen it would take roughly 15 Mt a year, more than fifteen times all the low-carbon hydrogen made today.

The Intermountain Power Project at Delta, Utah is the reference plant: two Mitsubishi M501JAC turbines, 840 MW, in commercial operation from 2025 on a 30% hydrogen blend by volume, fed from salt caverns at the adjacent ACES Delta project. Wärtsilä has delivered the first large-scale 100% hydrogen reciprocating engine plants. JERA's Hekinan co-firing program is the largest volume of hydrogen the power sector is actually buying anywhere, and it buys it as ammonia under a Japanese mandate, which entry 019 covers. On the fuel-cell side, Bloom Energy's 2026 data-center agreements with Oracle, AEP, and Brookfield are the commercial story, and all of it is gas-fired. South Australia's canceled 200 MW Whyalla plant is the counter-example worth remembering.

Hydrogen power is an option contract, not an operating business, and pricing it that way clarifies everything. A hydrogen-ready combustor adds a small premium to a turbine order and buys thirty years of optionality, which is why nearly every new machine ordered in 2025 and 2026 is specified that way while almost none of them burn hydrogen. The operating case needs three things at once, and the projects that died were missing at least one: delivered hydrogen near $1–2/kg, bulk storage in geology rather than in tanks, and a revenue stream that pays per kW-year for firm capacity rather than per MWh for energy. For a hydrogen supplier the implication is blunt. Power is a very large potential market at a price nobody can currently supply, and a much smaller real market today made up of mandated co-firing, byproduct hydrogen burned on site, and data centers paying for speed. Sell into those, and treat bulk power generation as a market that opens only if production costs fall below what any route in this sheet has demonstrated.

3 minCan gas turbines run on hydrogen fuel? | GE VernovaGE Vernova · 10k+ views

3 minWorld’s first large-scale 100% hydrogen engine power plant | WärtsiläWärtsilä Corporation · 10k+ views

4 minPlug's Megawatt-Scale Fuel Cell Power Generation System - Technology ExplainedPlug Power · 5k+ views

Hydrogen for Power Generation (GE Vernova) · Large-Scale Electricity Storage (Royal Society)