The standard sentence about hydrogen demand is that buyers are slow to commit. It is a comfortable sentence, because slowness is temporary and implies nothing about the price. It also happens to be untestable, which is part of why it survives.
There is now a better question available, and it has an answer with a published boundary: what is the most a buyer could pay for hydrogen and still be no worse off than it is today? That is a number, it is computable, and the IEA has computed it.
The ceiling, and the sentence that defines it
Before the number, the boundary — because a threshold like this is worthless without knowing what it holds constant. The IEA states it in the same chapter:[2]
Analysis of the costs of hydrogen in different end-uses enables identification of the maximum acceptable costs for hydrogen users, i.e. the maximum amount that can be spent on the hydrogen feedstock within a low-emissions pathway while maintaining the same total levelised cost of production as the incumbent pathway to produce the same commodity.
Read what that holds constant. Not the buyer's willingness, not its climate commitments, not its access to capital. The buyer's own total levelised cost of production. The threshold is the price at which switching to hydrogen leaves the buyer exactly as competitive as it is today. Above it, switching makes the buyer's product more expensive than its competitor's.
And the finding is that in most sectors and regions, without policy support, that ceiling sits below USD 2 per kilogram.[2]
For steel, the number is negative
The same analysis puts steel outside the range entirely:[2]
For steel production, the hydrogen cost must be negative (i.e. would require incentives) to reach parity with the incumbent route.
A negative acceptable cost is not a rhetorical flourish and it is worth stating plainly. It means that at a hydrogen price of exactly zero — the gas delivered free — the low-carbon steel route still costs more per tonne than the blast furnace it replaces. Free hydrogen does not close the gap. Something else has to pay, and in every project structure that reaches financial close, something else does.
This is the single most important line for anyone modelling a steel-linked hydrogen project, because it reframes what the model is for. The model is not demonstrating that the project is competitive. It is quantifying how large a subsidy, carbon price or green premium has to be, and for how long, before it is.
The floor, and its own boundary
Now the other side, from a different publisher, in a different currency, on a different basis. Hydrogen Europe reports:[6]
The average levelised cost of hydrogen for renewable hydrogen production projects submitted to the second Hydrogen Bank auction in 2025 was 7.1 €/kg
Note what that figure is and is not. It is an average across submitted projects, not awarded ones. It is a cost the bidder declared while asking for a subsidy, which is a context that shapes the number in a knowable direction. And it is a levelised production cost, which under this blog's standing complaint about LCOH could have been computed a dozen defensible ways.[6]
The buyer has already said this out loud
The IEA's threshold is modelled. It is worth checking it against a buyer describing its own market, unprompted, in a regulatory disclosure. ArcelorMittal, announcing that it would not proceed with two German DRI-EAF investments:[5]
there is limited willingness among customers to pay premiums for low-carbon emissions steel
And, in the same statement:[5]
Green hydrogen is evolving very slowly towards being a viable fuel source
Two sentences, and between them the whole causal chain. The steelmaker will not sign for hydrogen because its own customers will not pay for what the hydrogen produces. The missing contract is not at the hydrogen plant's gate. It is one link further down, in a market for green steel that is being described by its largest potential supplier as thin.
This is why "buyers are slow to commit" misleads. It locates the hesitation in the buyer's decision-making. The disclosure locates it in the buyer's order book.
Where the demand actually is
Figure 1
All the hydrogen, and the hydrogen this industry is about
Global hydrogen demand in 2025, against the low-emissions share of it.
Grew almost 3% in 2025, concentrated in traditional uses in industry and refining.
Grew 20% in 2025 — from a base small enough that the growth rate carries almost no volume.
The second bar is not a rendering error. Low-emissions hydrogen grew 20% in 2025 and still amounts to close to 1 Mt against a market of more than 100 Mt. The IEA expects it to pass 1% of global production for the first time in 2026.
Source: IEA, Global Hydrogen Review 2026 — Demand · Data as of 2026-06-18
Show the data ▸Hide the data ▾
| Measure | 2025 | Growth in 2025 |
|---|---|---|
| All hydrogen demand | more than 100 Mt | almost 3% |
| Low-emissions hydrogen demand | close to 1 Mt | 20% |
Hydrogen is not a nascent commodity. Global demand grew almost 3% in 2025 to surpass 100 Mt, concentrated in traditional uses in industry and refining.[3] There is an enormous, established, price-sensitive market for hydrogen, and it is buying the cheapest molecule available.
Low-emissions hydrogen grew 20% in 2025 and reached close to 1 Mt.[3] A 20% growth rate on that base moves less volume in a year than a single large ammonia plant consumes in a month, which is why growth rates are the wrong instrument for reading this market.
The strategic consequence is uncomfortable but clean: the addressable demand is not waiting to be created. It exists, at scale, today. It is simply already being served, at a price the low-emissions route cannot match without help.
What the pipeline did while this was true
Figure 2
The same 2030, measured three years running
Potential low-emissions hydrogen production by 2030 based on industry announcements, as reported in three consecutive editions of the same review.
Cited retrospectively in the 2025 edition as the figure the 2024 edition carried.
The 2025 edition attributes the fall to cancellations and delays.
The 2026 edition reports the pipeline has shrunk to 27 Mt by 2030.
These are announcements, not capacity. Nothing was built or unbuilt between these three bars — what changed is how much of what was announced the IEA still counts as capable of reaching 2030. The 2024 and 2025 figures are stated as Mtpa; the 2026 figure is stated as Mt by 2030.
Source: IEA, Global Hydrogen Review 2025 and 2026 · Data as of 2026-06-18
Show the data ▸Hide the data ▾
| Edition | Potential production by 2030 |
|---|---|
| Global Hydrogen Review 2024 | 49 Mtpa |
| Global Hydrogen Review 2025 | 37 Mtpa |
| Global Hydrogen Review 2026 | 27 Mt |
Three consecutive editions of the same review, measuring the same target year, using the same method, produced 49, then 37, then 27.[4][1] The 2025 edition attributes the fall to cancellations and delays.[4]
Nothing physical happened between those three bars. No plant was demolished. What contracted was the set of announcements still credited with a path to 2030 — which is to say, the pipeline was never capacity. It was a list.
The peer-reviewed literature reaches the same place from independent data. Tracking 190 projects over three years, Odenweller and Ueckerdt found that only 7% of global capacity announcements finished on schedule in 2023, and that more than 97% of announced 2030 capacity is not yet backed by a final investment decision.[7]
Europe, specifically
Figure 3
Europe's electrolyser pipeline, and what is actually being built
Announced electrolytic project pipeline, capacity under construction, and capacity installed — all three from the same compilation, in the same unit.
Stated as 9.3 Mt / 84.2 GWel of announced pipeline capacity.
The same document also states this as 3% of the pipeline. The EU Hydrogen Strategy target of 6 GWel by 2024 was not met.
Across 229 identified plants. Only 517 MWel reached FID between September 2024 and July 2025.
Same document, same unit, same database — so these three bars are comparable to each other in a way that figures from three different publishers would not be. The installed figure is described as 'at least' 571 MWel across 229 identified plants, and the nominal output attached to it, about 94,000 tonnes a year, is nameplate rather than measured production.
Source: Hydrogen Europe, Clean Hydrogen Monitor 2025 · Data as of 2025-09
Show the data ▸Hide the data ▾
| Stage | Capacity | Note |
|---|---|---|
| Announced pipeline | 84.2 GWel | 9.3 Mt of hydrogen |
| Under construction | 2.84 GWel | 3% of the pipeline |
| Installed by June 2025 | at least 571 MWel | 229 plants |
| Reached FID, Sep 2024 – Jul 2025 | 517 MWel | in eleven months |
The European numbers are worth isolating because they come from one compilation in one unit, so the funnel is internally consistent. An announced pipeline of 84.2 GWel. Under construction: 2.84 GWel, which the same document also expresses as 3% of the pipeline. Installed by June 2025: at least 571 MWel across 229 identified plants. And in the eleven months from September 2024 to July 2025, 517 MWel reached final investment decision.[6]
The EU Hydrogen Strategy target was 6 GWel of electrolysers deployed by 2024. It was not met.[6]
What to do with this before your next model
None of this argues that green hydrogen projects cannot be financed. Projects are being financed. It argues that a specific and very common modelling posture — treat demand as given, solve for cost — has the causality backwards.
- Find your buyer's ceiling before you refine your own cost. The relevant question is not what your hydrogen costs. It is what your buyer's product costs today, and how much of that cost your molecule is allowed to become before the buyer's own competitiveness moves. That is a calculation about their industry, not yours.
- Name the instrument that closes the gap, and its expiry. Contract for difference, Hydrogen Bank premium, 45V, CBAM, a mandate, a green premium. One of them is carrying the project. Put it on the page with its legal basis and its end date, because that date is the project's real tenor.
- Treat a green premium as merchant revenue until a contract says otherwise. A buyer's stated willingness to pay more is a market view, not a receivable. ArcelorMittal has published its assessment of how many such buyers exist.[5]
- Distinguish displacing grey hydrogen from creating new demand. Displacement targets a buyer who already has the plant, the offtake and the logistics, and whose only question is price. New demand needs the buyer to build something first. The first is a price problem; the second is a price problem plus a capital cycle.
- Be precise about which number you are quoting. A cost of production, a buyer's acceptable cost, a subsidy strike price and an auction premium are four different quantities. Every one of them gets quoted as "the cost of hydrogen".
The uncomfortable reading of the IEA's threshold is not that hydrogen is too expensive. It is that for most uses, in most places, the honest business case does not currently run through the buyer's economics at all — it runs through a policy instrument, and the quality of that instrument is the quality of the project.
That is a knowable, model-able, underwritable position. It is just not the one on most slides.