Home » Green Ammonia Capex and Opex: The Benchmarks I Check First
Green Ammonia Capex and Opex: The Benchmarks I Check First
10 min read · Dec 2, 2025 · By Alex
A published cost figure for a new ammonia plant is usually a production cost per tonne rather than one capital number, and I treat it that way. I keep the green ammonia guide open beside me as the parent page for this file. Green ammonia capex and opex are quoted the same way: modelled production costs run from 407 to 1,239 USD per tonne in one October 2024 study, and the range moves with the price of electricity, the plant efficiency, the installed capacity and the year of the study. That is why I never compare two figures until their electricity assumption and their date are in front of me. I read the assumption before I read the answer.
I am going to order this in five moves: what the number measures, what drives the electricity side, what the plant itself takes to build, what the operating cost does, and the checks I run before I believe a published comparison. The vocabulary comes first, because much of the disagreement I see is a definition problem rather than a modelling problem.
In short
- What it measures: I read a cost figure as a production cost per tonne first, then ask what capital and what operating cost it already contains (2024).
- Power first: The electricity price drives the spread inside every modelled range I use (2024).
- Capital split: The electrolyser stack is 15 to 20 percent of installed investment, so most of the money sits elsewhere (2025).
- Operating case: Reported conventional production costs run from 100 to 1,000 USD per tonne by region, averaging 305 USD per tonne since 2013, as published in January 2024.
- My four checks: The date, the power price, the utilisation and plant life behind the figure, and the project stage.
In this article
What the number actually measures
Capex is the money spent to build the plant. Opex is the money spent to run it. A levelised cost spreads the whole lifetime cost across the tonnes produced. The electrolyser is the unit that splits water into hydrogen, and the balance of plant is everything around it: pipes, utilities, compression. The connection is the link to the power supply, and capacity factor, in plain terms, is how much of the year the plant actually runs. A benchmark is a published figure used for comparison. A retrofit adapts an existing plant, an offtaker is the party bound to buy, take-or-pay makes the buyer pay for the contracted volume, and a bankable number is one a lender will model. With that vocabulary in place, one fact does most of the work: published cost figures for the same molecule span roughly a factor of three in a single year (2024), which is why I read the assumption before the answer.
What this means for you: when a deck shows a single tonne cost with no label, I ask which of these words it belongs to, because a capital estimate, an operating cost and a lifetime levelised cost are not the same object.
The first driver is the price of electricity
Everything downstream of the meter starts here. In the Oxford study of October 2024, the cost of electricity is the crucial factor in an e-ammonia model, and modelled average production costs of green ammonia vary between 407 and 1,239 USD per tonne in 2024. The spread comes from electricity costs, plant efficiency, investment and operating cost estimates, and installed capacity, so I treat it as four separate assumptions rather than one uncertain answer. The same study projects 262 to 696 USD per tonne by 2030, a reduction of about 36 to 44 percent, and 202 to 482 USD per tonne by 2050, while making clear that significant reductions remain contingent on supportive government policy. I work the power side separately in an economics read on hydrogen plant costs. Any cost figure handed to me without its power price goes back to its author (2024).
What this means for you: before I accept a tonne cost, I ask for the electricity price sitting under it and the date of the assumption, because those two lines decide whether the number can travel into my own model.
What the plant itself takes
Buying the main unit is only a slice of the cheque. In the 2025 review’s five key questions, the electrolyser stack is only 15 to 20 percent of installed investment, so the majority of the capital goes to the process balance, the works and the connection (2025). Where that kit is made also moves the input price: China holds close to 60 percent of global electrolyser manufacturing capacity, and renewable hydrogen there is 40 to 45 percent cheaper than in Europe or the United States (2025). Unit scale has moved quickly as well: the largest unit built was 30 MW in 2021, a 500 MW unit went into service in China in 2025, and one project targets 2 GW by 2027, a 75 times increase in six years. The technical build-up sits on the plant page. A plant quote that ignores the connection and the works is not a quote (2025).
What this means for you: when I see a capital number, I split it into the stack and everything else first, then ask what the connection and the civil works were priced at, because that is where most of the money went.
What the operating cost does
A cheap build can still lose money every hour it runs. The paper from BloombergNEF and Climate Technology Coalition of January 2024 reports production costs running from 100 to 1,000 USD per tonne by region, against an average of 305 USD per tonne since 2013, and says the largest producer is normally the marginal producer that sets the world price. World capacity in that January 2024 paper stands at 243 million tonnes a year, with utilisation around 80 percent and plants lasting 20 to 50 years. On the input side it expects the cost of low-carbon hydrogen to fall 66 percent by 2030, to 1 to 4.1 USD per kilogramme. I check the operating case against a retrofit of an existing plant. An operating cost built on a fossil power curve is the fastest way to make a model look healthy and a plant look broke (2024).
What this means for you: I ask which utilisation and which plant life sit behind a tonne cost, because a figure quoted at a long life and full running hours will not survive contact with my own assumptions.
Why published benchmarks disagree
Two ranges can both be correct and still be useless next to each other. A modelled range for 2024 and a modelled range for 2030 answer two different questions, and the same study moves from 407 to 1,239 USD per tonne in 2024 to 262 to 696 USD per tonne in 2030 without anyone changing a plant (2024). One literature estimate carried inside that study puts green ammonia at 634 EUR per tonne in 2025, falling to about 380 EUR per tonne by 2040 (2024). Announcement adds a further layer: more than 300 Mtpa of low-carbon ammonia projects had been announced in May 2025 against a conventional market of 200 Mtpa, with 85 projects and 35 Mtpa at FEED and 48 projects and 15 Mtpa at final investment decision (2025). A benchmark with no date attached is a number, not a benchmark (2024).
What this means for you: when two published figures disagree, I line up their years and their project stages before I argue about methodology, because much of the gap disappears once the dates are visible.
What a lender will model
A cost figure that nobody has agreed to pay is a scenario, not a revenue line. In the Hydrogen Council and McKinsey roadmap of November 2025, one scenario uses imports closing around 600 USD per tonne as the low end (2025), and an exhibit carries 130 USD per tonne of carbon cost as a 2030 assumption. It cites contracts for difference as the instrument that reduces financing cost and improves bankability, and it describes pooled demand as the mechanism that lowers counterparty risk, creates bankable offtake volumes, and moves a producer from FEED to final investment decision. A cost figure only becomes relevant once an offtake covers it, and 45 of 185 tracked projects held such agreements in January 2024. The financing mechanics sit in the walkthrough on financing a hydrogen project.
What this means for you: I do not let a modelled cost turn into a revenue assumption until I have seen the offtake behind it, and until I know whether that offtake is an announced agreement or a signed contract.
What the equity cheque must clear
Cheap renewable potential does not by itself fund a plant. In the December 2025 Southeast Asia briefing, regional potential reaches nearly 11 Mtpa below 6 USD/kg by 2030, most of it in Indonesia, while 8 billion USD went into hydrogen production investment in 2025, 38 billion USD of public funding advanced since the previous evaluation, and 6.6 billion USD of development finance lending was committed in 2021 to 2025. High cost of capital is cited as the main barrier, ahead of slow renewable deployment and weak fuel switching incentives, and no certification scheme is yet in place in the region, with work under way in three countries (2025). I size the equity case in the investment view on low carbon fuels. Capital is the constraint named before technology in this region (2025).
What this means for you: before I price a project, I ask what the capital actually costs and whether a certification scheme exists to prove the product, because both sit in front of the engineering.
The four checks before I believe a figure
I run the same four questions on every published number, and none of them needs data that the page does not already carry.
- Check one, the date. A 2024 range and a 2050 range answer different questions (2024).
- Check two, the power price. The electrolyser is 15 to 20 percent of installed investment, and the rest is works and connection (2025).
- Check three, utilisation and plant life. Capacity around 80 percent and plants lasting 20 to 50 years decide what a tonne actually costs (2024).
- Check four, the stage. Only 6 percent of announced regional capacity has reached final investment decision (2025).
What this means for you: if a figure clears all four checks I will put it in the model, and if it fails any one of them I send it back with a question rather than a rejection.
Sources
- Oxford Institute for Energy Studies, Fueling the future: A techno-economic evaluation of e-ammonia production for marine applications, OIES Paper ET40 https://www.oxfordenergy.org/wpcms/wp-content/uploads/2024/10/ET40-Fuelling-the-future-final.pdf, October 2024.
- IEA, Global Hydrogen Review 2025, Five key questions about hydrogen https://www.iea.org/reports/global-hydrogen-review-2025/five-key-questions-about-hydrogen, 2025.
- BloombergNEF and Climate Technology Coalition, Scaling Up Hydrogen: The Case for Low-Carbon Ammonia https://assets.bbhub.io/professional/sites/24/CTC-whitepaper_Ammonia_Final.pdf, January 2024.
- Hydrogen Council with McKinsey & Company, Fertilizing the future: A roadmap to scale low-emission ammonia fertilizers https://hydrogencouncil.com/wp-content/uploads/2025/11/Fertilizing-the-future-FINAL.pdf, November 2025.
- IEA, Global Hydrogen Review 2025, Southeast Asia, 9 December 2025 https://iea.blob.core.windows.net/assets/745a259b-9cb7-4e1d-9ba1-a8b63d1eaae8/GHR2025_SEA_Webinar.pdf, 2025.
Frequently asked questions
How much does green ammonia cost per tonne today?
It depends on the assumptions, and it depends on the year. Modelled production costs run between 407 and 1,239 USD per tonne in 2024, moving to 262 to 696 USD per tonne by 2030 and 202 to 482 USD per tonne by 2050 in the same study (2024). None of those ranges is a market price and none is a figure available today, so the honest answer I give is a range plus its date rather than a single number.
What share of the capital does the electrolyser take?
A minority of it. The electrolyser stack is 15 to 20 percent of installed investment, with the rest going to process balance, works and connection (2025). That is why I ask for the full capital breakdown before I compare one plant quote with another, because a stack price on its own tells me very little about what the plant will cost.
Why do published cost figures disagree with each other?
Because the inputs differ. Electricity cost, plant efficiency, installed capacity and the year of the study all move the range, and reported conventional production costs already run from 100 to 1,000 USD per tonne by region against an average of 305 USD per tonne since 2013 (2024). Two studies can therefore print numbers that look contradictory while describing two different sets of assumptions.
What should an investor check before believing a benchmark?
Four things: the date of the figure, its power price, the utilisation and plant life behind it, and the project stage, since only 6 percent of announced regional capacity had reached final investment decision (2025). I also confirm that the figure is modelled rather than contracted, because a modelled cost is not a market price.
Not sure your project would survive a credit conversation?
A short call is usually enough to see where your project stands.
If you are at feasibility stage and want the numbers pressure tested before they enter your model, this is where project finance work comes in: I help you interrogate the cost basis, the power assumption and the project stage claims behind a published figure.
I do not provide investment advice or arrange investments. I act as a strategic advisor and facilitator.
Alex, Founder of WeConnect Asia
I help low-carbon fuel and carbon removal project developers in Southeast Asia secure the commercial and financial partners needed to reach FID.
Related articles

Pitching a Climate Tech Company to Strategic Investors: What the Corporate Has to Believe Before It Writes a Cheque
How to pitch a climate tech company to strategic investors: the metric cluster, the readiness evidence and the ask that corporates fund.

Biomethane: What It Is, How It Is Made and Who Buys It
Biomethane is biogas cleaned to pipeline quality. I explain how it is made, how it differs from biogas and who buys it in Southeast Asia.

CO2 Sourcing for e-SAF Projects: What the Rules Allow and What the Contract Must Say
Biogenic CO2, direct air capture or industrial flue gas: which sources an e-SAF plant may use, what they cost and which contract terms make the fuel certifiable.

Operating Performance Guarantees for SAF Plants: What Lenders Ask For
Minimum guarantees, performance damages, retesting and who stands behind them: the operating guarantees a SAF plant needs to keep debt and offtake intact.

EPC and Construction Risk in SAF Plants: What a Lender Actually Tests
FEED scope, convertible EPC, performance guarantees and damages caps: the construction risk questions a SAF project must answer before financial close.

Matching Feedstock and Offtake Risk Back-to-Back in a SAF Project: How Lenders Read the Chain
Lenders do not buy a feedstock plan and an offtake plan separately. They buy the link between the two, and the link is where most SAF files break.