Two beliefs do a lot of load-bearing work in hydrogen project models. The first is that the learning curve will bring costs down on the usual schedule. The second is that Chinese equipment will get there faster. Both are now old enough to be tested against published evidence rather than assumed, and the evidence is more specific than either belief.
The equipment gap is real, and most of it does not travel
Figure 1
The equipment gap, and what survives shipping it
Cost of making and installing an electrolyser in 2024, by where it was made and where it was installed.
The IEA's 2024 figure for electrolysers manufactured and installed outside China.
The headline gap that gets quoted, four to six times on equipment.
Once transport costs and tariffs are considered. Most of the gap does not travel.
All three bars are on the same boundary — made and installed — from the same source and the same year, so they are comparable to each other. The third bar is the one that matters for a European project: it is what a Chinese electrolyser costs once transport and tariffs are included, and it overlaps the bottom of the first bar rather than reaching the second.
Source: IEA, Global Hydrogen Review 2025 · Data as of 2025-09
Show the data ▸Hide the data ▾
| Where made | Where installed | USD/kW |
|---|---|---|
| Outside China | Outside China | 2,000–2,600 |
| China | China | 600–1,200 |
| China | Outside China | 1,500–2,400 |
The IEA's 2024 figures are unambiguous about the gap at source:[2]
The cost of making and installing an electrolyser outside of China in 2024 was USD 2 000 to USD 2 600 per kilowatt (kW) […] compared with USD 600 to 1 200 per kW for electrolysers manufactured and installed in China.
Four to six times, on the same boundary. That is the number the headlines carry, and it is correct.
The next sentence is the one that gets dropped:[2]
When transport costs and tariffs are also considered, the cost of installing a Chinese electrolyser outside China is to USD 1 500 to USD 2 400 per kW
A Chinese electrolyser installed in Europe lands at USD 1,500–2,400/kW, against USD 2,000–2,600/kW for one made and installed outside China. The ranges overlap. What survives the journey is a discount at the bottom of the range, not a change of regime.
Why the remaining gap does not reach the LCOH
Because the stack was never most of the cost. The IEA states it directly:[3]
the electrolyser stack represents only 15-20% in the investment cost of an installed electrolyser
This is the same finding as the €/kW problem, arriving from the other direction. There, three published costs for the same alkaline machine differed by a factor of eight because they were drawn on three different boundaries: stack, system, installed project. Here, the consequence of that structure is a decision rule.
The JRC does the arithmetic and states the result:[1]
A system built with a Chinese stack and BoP would only provide a LCoH reduction of 3%-13%, depending on the electricity sourcing design.
Not 4× cheaper hydrogen. Three to thirteen per cent, from a stack and balance of plant that are several times cheaper as equipment. Everything else — civils, grid connection, land, permitting, EPC, contingency, and then electricity for twenty years — is unmoved by where the stack was made.
The factories that make this equipment are nearly idle
Figure 2
What Europe can build in a year
EU operational electrolyser manufacturing capacity by technology, estimated at the end of 2025. Total: 8.9 GWel per year.
The largest single technology block in European manufacturing capacity.
The mature technology, and close behind PEM in European capacity.
The smallest block, and the one where a factory has since been hibernated.
These factories are running at about 10% utilisation, a figure the JRC attributes to the IEA and explains by a lack of orders that is in turn holding back economies of scale. For scale, and being careful about units: this is annual manufacturing capacity, while the EU27's entire installed electrolysis stock is expected to be 514 to 800 MWel at the end of 2025. One year of European factory output at full rate would be more than ten times everything Europe has installed.
Source: European Commission, JRC / Clean Energy Technology Observatory · Data as of 2026-01-14
Show the data ▸Hide the data ▾
| Technology | EU manufacturing capacity, end 2025 |
|---|---|
| PEM stacks | 4.2 GWel/year |
| Alkaline stacks | 3.8 GWel/year |
| Solid oxide (SOE) | 0.9 GWel/year |
| Total | approximately 8.9 GWel/year |
| EU27 installed electrolysis stock, end 2025 (different unit) | 514–800 MWel |
Europe can manufacture about 8.9 GWel of electrolysers a year: 4.2 GWel of PEM stacks, 3.8 GWel of alkaline, 0.9 GWel of solid oxide.[1]
What those factories are actually doing is the finding:[1]
The IEA reports an utilisation rate of today's factories of about 10%
and the JRC gives the cause as a lack of orders, which is in turn holding back the economies of scale that were supposed to lower the price.
Set that against the installed base, being careful that these are different units: the EU27's entire installed electrolysis capacity is expected to be between 514 and 800 MWel at the end of 2025.[1] A single year of European factory output at full rate would be more than ten times everything Europe has installed in total.
That is not a supply constraint. It is the opposite, and it is why the cost-down mechanism most models assume — volume driving unit cost down a curve — has nothing to run on.
The manufacturers have said this in their own filings
Figure 3
One order book, two consecutive years
Order intake in the green hydrogen segment of thyssenkrupp nucera, fiscal years ending 30 September.
The prior-year comparison stated in the same release.
Green hydrogen segment order intake for the year ended 30 September 2025.
Segment sales in the same year were €459 million against €524 million, and segment EBIT was −€56 million against −€76 million. The order intake is the forward-looking number of the three, and it is the one that fell by more than nine tenths. The company's chief executive attributes it to continued restraint in final investment decisions.
Source: thyssenkrupp nucera, annual results for fiscal year 2024/2025 · Data as of 2025-11-24
Show the data ▸Hide the data ▾
| Measure, green hydrogen segment | FY 2023/2024 | FY 2024/2025 |
|---|---|---|
| Order intake | €356M | €26M |
| Sales | €524M | €459M |
| EBIT | −€76M | −€56M |
thyssenkrupp nucera's green hydrogen segment took €26 million of orders in the fiscal year ended 30 September 2025, against €356 million the year before. Segment sales were €459 million against €524 million, and segment EBIT −€56 million against −€76 million. Its chief executive:[4]
The situation on the market for green hydrogen became even more challenging in the reporting year. Restraint in final investment decisions continues.
Nel halted production at its alkaline plant at Herøya in January 2025, affecting around 20% of full-time employees, with its chief executive stating that while the long-term outlook remains strong, the decision followed lower order intake in 2024 than expected.[5]
Topsoe announced in May 2026 that its Herning factory would be hibernated until demand is robust, and that it would not proceed with a second solid oxide factory in the United States, taking DKK 3,500–3,900 million of impairment and restructuring costs and affecting around 440 positions.[6]
Green Hydrogen Systems filed for bankruptcy in June 2025, stating that it had not been possible to reach a sustainable solution within the framework of its restructuring process, and that it did not expect any distributions to shareholders.[7]
Four companies, four disclosures, one message: the order book is the constraint, and it has been the constraint for two years.
The learning curve, stated precisely
Here the evidence refuses the simple version in both directions, which is why it is worth quoting exactly. The peer-reviewed finding is that learning has been substantial:[8]
Our findings indicate a significant learning effect for PEM and AEL electrolysers in the last 20 years, with learning rates of 32.1% and 22.9%, respectively.
Those are high learning rates. Anyone claiming that electrolysers have not come down the curve at all is wrong, and this is the paper that says so.
And then the sentence that matters for a project:[8]
Small-scale electrolysers exhibit a stronger learning effect (25%), whereas large-scale projects show no clear cost reductions due to their early stage of deployment.
The learning happened at small scale. At the scale a project-financed plant is built, no clear cost reduction is observed. So a model that applies a historical learning rate to a 100 MW plant is extrapolating from a segment where the effect was measured to a segment where the same study says it was not.
And the plant that was built is not running as designed
One more caution, on the other end of the same chain. Sinopec's Kuqa project, at 260 MW the largest operating green hydrogen plant in the world, was designed for 20,000 tonnes a year. Consultancy analysis of its 2023 output puts it at 2,010 tonnes, operating at around 20% of rated capacity, and notes that even hitting the 20,000 tonne target would correspond to the electrolyser running at 51% capacity.[10]
That figure is consultancy commentary and not an operator disclosure, and it is labelled as such here. But the design assumption is the part worth carrying: the flagship project's own target implies a 51% capacity factor. A model assuming 70–90% load on an electrolyser fed by variable renewables is outside the range that the largest built example was designed for, never mind what it achieved.
What to do with this before your next cost-down assumption
- Stop solving for equipment price. On the Commission's own assessment the entire Chinese-sourcing decision is worth 3–13% of LCOH. If the model needs more than that, the answer is not in procurement.
- Name the boundary of every learning rate you apply. A learning rate measured on stacks does not apply to installed project cost, because the stack is 15–20% of it.
- Do not apply a small-scale learning rate to a large-scale plant. The study that measures the high rates says explicitly that large-scale projects show no clear reduction. Using its headline number while ignoring its finding is citing it against itself.
- Check that your supplier will exist at delivery. One liquidation, one bankruptcy, one hibernated factory and one halted line, in eighteen months, across the European supply base. Supplier credit is now a real diligence item, not a formality.
- Take capacity factor from the built examples, not the datasheet. The largest operating project in the world was designed around 51%.
- Treat idle factories as a symptom, not an opportunity. Ten per cent utilisation is often read as spare capacity waiting to serve a boom. The manufacturers read it as absent demand, and they are the ones closing lines.
The version of this that is easy to say — Chinese electrolysers will make green hydrogen cheap — is not supported by the arithmetic of where the money goes in a built plant. The version that is harder to say, and better supported, is that the equipment stopped being the interesting variable some time ago, and that most of the cost a project has to attack has never been in the machine at all.