The constraint on a low carbon hydrogen project is the offtake rather than the production cost, since a plant with defensible levelised cost and no binding buyer does not reach a final investment decision. The IEA's Global Hydrogen Review 2023 put the share of announced 2030 capacity that had cleared final investment decision at roughly four percent, which is a demand side result. Ammonia is the usual carrier because it ships and stores as a liquid, and the conversion and reconversion steps consume a material share of the energy in the molecule, so they belong in the delivered cost. Certification decides whether the molecule qualifies under a buyer's scheme at all, which puts the carbon intensity threshold and its accounting rules in the term sheet rather than in a later compliance exercise.
The investment paper is thorough. Levelised cost per kilogram, a sensitivity grid across power price and utilisation, an engineering estimate with contingency, a permitting schedule, a grid connection queue position. The technical work has been done properly and the numbers are defensible.
Then somebody asks who has committed to buy the output, and the answer is a list of memoranda of understanding, none of which is binding, none of which names a price, and most of which were signed by people who wanted a photograph. The paper goes back for another cycle, and the project joins a very long list of announced capacity that never gets built.
The IEA has made the same observation in successive editions of its Global Hydrogen Review: announced project capacity runs far ahead of capacity that has reached a final investment decision, with the 2023 edition putting the share of announced 2030 capacity that had cleared that hurdle at roughly four percent. The binding constraint is on the demand side, and it behaves like a supply chain problem rather than an energy one.
Why the offtake decides, and the production cost does not
Projects of this size get financed against contracted cash flow. A lender sizes debt against a debt service coverage ratio computed on revenue that can be demonstrated in advance, which in practice means a contract with a named counterparty, a term, a price mechanism and enough credit quality behind it that the revenue can be modelled as certain.
Merchant volume gets no such treatment, and hydrogen has almost no merchant market to speak of. There is no exchange, no settled index, no forward curve, and no hub price a bank can mark against. The existing industrial hydrogen market is mostly captive production behind refinery and ammonia plant fences, or pipeline supply under bilateral contracts in a handful of industrial clusters. A developer who reaches financial close with unsold volume is asking a lender to take price risk in a market where price discovery does not exist.
The practical effect is that the production cost analysis, which is where most of the engineering effort goes, determines whether a project is worth pursuing rather than whether it happens. Two projects with identical delivered costs will diverge entirely on the strength of their offtake, and the one with a fifteen year contract to an investment grade buyer will be built while the cheaper one with three memoranda will not.
That reordering has a consequence for how these projects should be planned. Offtake development is normally treated as a commercial workstream that runs alongside engineering. It belongs on the critical path, ahead of it, because the offtaker's requirements determine the plant's specification: what purity, what delivery profile, what certification regime, what carrier, what location.
The chicken and egg, and what actually breaks it
The deadlock is easy to state. Industrial buyers will not convert an asset or sign a long contract at a large premium until supply is reliably available. Producers will not build until buyers have signed. Both sides are waiting for the same signal, and both are behaving rationally.
Four things have shown they can break it, and none of them is a better cost curve.
A mandate that makes the premium unavoidable. The EU's revised renewable energy directive (2023) requires that a substantial share of hydrogen used in industry be renewable fuels of non-biological origin by 2030, with a further step by 2035. Where a buyer faces a compliance obligation, the relevant comparison becomes the cost of failing to comply, and the willingness to sign a long contract changes accordingly.
A subsidy that closes the gap directly. The European Commission's first Hydrogen Bank auction, settled in April 2024, awarded a fixed premium per kilogram produced over ten years, with winning bids between roughly 0.37 and 0.48 euros per kilogram. A production-side premium of known size lets a developer quote a price a buyer can actually pay.
An intermediary that takes the mismatch onto a public balance sheet. Germany's H2Global mechanism uses a state-backed intermediary that signs long purchase agreements on the supply side and resells on short contracts to whoever will pay the most, with public funds covering the difference. Its first auction round concluded in 2024. The structural insight is that producers need long contracts and buyers want short ones, and somebody has to hold that tenor mismatch.
An anchor buyer whose economics already work. Existing ammonia and refinery hydrogen demand is real, priced and continuous. Substituting into it requires no new asset, no new application and no consumer behaviour change, which is why the first commercial volumes in this sector are moving into applications that already exist.
Nothing on that list is a market signal in the ordinary sense. Planning a project as though a merchant market will emerge on the timescale of your construction schedule is the single most expensive assumption available in this sector.
Ammonia as the carrier, and what the conversion costs
For any distance beyond a pipeline, hydrogen moves as something else. Ammonia is the practical option today for reasons that have little to do with hydrogen and a lot to do with existing infrastructure. It is already produced at scale, with the IEA's Ammonia Technology Roadmap (2021) putting global production at around 185 million tonnes, roughly a tenth of which crosses borders. Ships, terminals, storage tanks and handling procedures exist. It liquefies at minus 33 degrees Celsius at atmospheric pressure, so it stores in conventional refrigerated tanks rather than requiring cryogenic engineering at minus 253.
The penalty is in the conversion, and it compounds along the chain. Synthesis consumes energy and capital. Shipping consumes fuel and incurs boil-off. Cracking back to hydrogen consumes a further substantial fraction, typically supplied by burning part of the cargo, and purification to the grade a fuel cell needs consumes more. Published assessments including the IEA's The Future of Hydrogen (2019) put the round trip penalty for conversion and reconversion in the region of a quarter to a third of the energy content, with the range depending on heat integration and the assumed cracker design.
That arithmetic drives a planning conclusion that is easy to miss. If the end use is ammonia, the chain stops before the expensive step. Fertiliser production, ammonia as a marine fuel under the tightening intensity limits of FuelEU Maritime from 2025, and power sector co-firing all consume the molecule as delivered. If the end use requires hydrogen back, the cracking penalty sits on top of a delivered cost that was already high, and the total has to compete against hydrogen produced locally by whatever route is available at the destination.
The consequence for project selection is that a producer should be choosing offtake markets by whether they consume the carrier directly. A project designed around ammonia buyers has a shorter chain, fewer counterparties and a lower delivered cost than one that assumes a cracking facility at the destination which somebody else is expected to finance.
Certification decides whether the molecule qualifies
A tonne of ammonia is a commodity. A tonne of ammonia that qualifies under a specific buyer's subsidy or compliance regime is a different product, and the difference in value is larger than any operational efficiency available on the plant.
The rules are detailed and they are not aligned across jurisdictions. The EU's delegated act on renewable fuels of non-biological origin, adopted in 2023, sets conditions on additionality of the power, on temporal correlation between generation and electrolysis, which moves from monthly to hourly matching at the end of 2029, and on geographic correlation between the generator and the electrolyser. The US Treasury's final rules for the section 45V production credit, issued in January 2025, set out their own version of the same three ideas, with hourly matching from 2030. ISO 19870 (2023) provides a methodology for quantifying the emissions associated with hydrogen production and delivery, which is the common technical language underneath the various schemes without being a substitute for any of them.
Three planning consequences follow.
The power procurement contract is a product specification. Whether the electrolyser is supplied by a dedicated new-build generator in the same bidding zone, on a contract with the right time granularity, determines whether the output qualifies. That decision gets made years before the first molecule and is difficult to reverse.
Location determines eligibility as much as cost. A site with excellent renewable resource in a jurisdiction with no mutual recognition arrangement produces a molecule that a European buyer cannot count, whatever its actual emissions intensity.
Certification failure is a commercial event with a defined size, and the contract should say who carries it. A cargo that loses its certification is worth the conventional price of the commodity, which is a fraction of the contracted price, and the gap is large enough that it belongs in the risk register rather than in a boilerplate clause.
A plant whose output cannot be stored economically
Electrolytic hydrogen production follows the power supply. Demand from an industrial buyer follows their production schedule. Storage is what normally reconciles two profiles that do not match, and hydrogen storage at meaningful scale is available only where the geology cooperates, principally in salt caverns. Everywhere else the practical buffer is hours in pressurised vessels, which is enough to smooth a shift pattern and nothing more.
That leaves a small set of options, each with a cost.
Run the plant flat and require a buyer who can take flat volume. This gives the best utilisation, which matters because utilisation is one of the two dominant terms in unit cost, alongside power price. It requires a buyer whose own operation is continuous and whose outages you have to plan around.
Follow the buyer's take profile. This protects the buyer and pushes the utilisation risk onto the producer, where it lands directly on unit cost. A plant running at forty percent utilisation carries its capital charge over forty percent of the volume.
Convert to ammonia and store the ammonia. This is the reason so many projects have an ammonia loop even where the buyer wants hydrogen. Ammonia storage is conventional and cheap by comparison, so the synthesis unit acts as the buffer that the hydrogen system cannot provide. The synthesis loop itself prefers steady operation, which pushes back toward flat running and toward batteries or oversized renewables to smooth the input.
Oversize the renewable supply and curtail. Hourly matching rules push in this direction, and the result is a plant with a lower utilisation than its nameplate suggests and a power cost that includes generation you paid for and did not use.
Each of these is a legitimate design. What they have in common is that the take profile in the offtake agreement, which commercial teams often treat as a detail to be settled later, sets the utilisation, and utilisation sets the cost. Negotiating volume flexibility without pricing it is how a project arrives at financial close with an economic case that no longer holds.
The limit
Delivered cost figures in this sector should be read as one point on a wide range. The two dominant inputs, power price and utilisation, are both assumptions rather than observations at the planning stage, and reasonable variation in either moves the answer substantially. Estimates from different sources also use different boundaries, some stopping at the plant gate and some including conditioning, shipping and reconversion, so two published numbers frequently describe different things.
The practical response is to stop quoting a single figure internally. Carry a range, state the power price and utilisation behind each end of it, and make clear where the boundary sits. A committee comparing a plant gate cost against a delivered cost without noticing is a common and expensive failure.
There is a second uncertainty worth naming. The subsidy and compliance regimes that make these projects viable are younger than the assets they support, and the detailed rules have moved more than once already. A project whose economics depend on one regime remaining exactly as drafted for fifteen years is carrying policy risk that no commercial term sheet fully covers. Blue routes with capture attached face a related question through a different mechanism, and that belongs in a separate discussion.
Take the offtake term sheet you have in front of you, extract the take profile the buyer is asking for, and re-run the plant economics at the utilisation that profile implies rather than at nameplate.
More on this in energy on the DemandLabs platform.
