There is one number from the last seven weeks that a project analyst should carry, and it is not a price.
The number, and its boundary
Figure 1
Electrolyser capacity that came online in Europe, first half of the year
Megawatts of electrolysis commissioned in H1 2025 and H1 2026.
The comparison the report itself uses.
The report calls it a slowdown. Measured against the same half-year, it is roughly a quarter of it.
Read the boundary before the number. This is capacity COMMISSIONED, not installed; in EUROPE, not globally; in a HALF-YEAR, not a year; and in megawatts of electrolysis, not of renewable generation. Over the same window the average size of a commissioned project fell from 18 MWel to around 5 MWel, which is the part that is hard to read as a pause.
Source: Hydrogen Europe, Clean Hydrogen Monitor 2026 · Data as of 2026-09
Show the data ▸Hide the data ▾
| Measure | H1 2025 | H1 2026 |
|---|---|---|
| Capacity commissioned | 177 MWel | 46 MWel |
| Average size of a commissioned project | 18 MWel | around 5 MWel |
Hydrogen Europe's 2026 monitor puts it in a single clause:[1]
the first half of 2026 saw a slowdown with only 46 MWel coming online, compared to 177 MWel becoming operational in the same period in 2025
Before anything else, the boundary, because this figure will be quoted without one within a week. It is capacity commissioned, not installed. It is Europe, not the world. It is a half-year, not a year. And it is megawatts of electrolysis, not of the renewable generation feeding it. Each of those four would change the number by an order of magnitude if swapped.
Inside that boundary, Europe connected roughly a quarter of what it connected a year earlier.
The part that is harder to read as a pause
A slowdown in volume can be a timing artefact: three large projects that slipped from June to September would do it. The second figure is the one that resists that reading:[1]
the average size of projects commissioned between June 2025 and June 2026 fell to around 5 MWel, down from 18 MWel during the corresponding period a year earlier
Less capacity, in smaller units. That is not three delayed projects. That is a different population of projects reaching the end of construction — and a 5 MWel average is below the scale at which most of the sector's infrastructure assumptions make sense. A trunk pipeline, a salt cavern, a ten-year take-or-pay and a project finance package are all instruments sized for plants an order of magnitude larger.
What Europe actually has
Figure 2
What Europe has, is building, and has committed to
Installed electrolysis capacity, capacity under construction, and capacity reaching final investment decision — all from the same compilation, in the same unit.
78% concentrated in five countries.
677 MWel, up 19% from 571 MWel in July 2025. Around 1% of total hydrogen demand.
Six months of investment decisions across the whole of Europe.
78% of the 3.7 GWel under construction sits in just five countries, with Germany at about 1.3 GWel and Sweden at 0.8 GWel. Installed capacity grew 19% in a year — from 571 MWel in July 2025 — and amounts to around 1% of Europe's total hydrogen demand. The FID bar is the forward-looking one, and it is the smallest.
Source: Hydrogen Europe, Clean Hydrogen Monitor 2026 · Data as of 2026-09
Show the data ▸Hide the data ▾
| Stage | Capacity |
|---|---|
| Under construction | 3.7 GWel |
| Installed, July 2026 | 677 MWel |
| Installed, July 2025 | 571 MWel |
| Reached FID, H1 2026 | 0.5 GWel |
Three numbers from the same compilation, in the same unit, which makes them comparable to each other in a way that figures from three publishers would not be.[1]
Installed: 677 MWel by July 2026, up 19% from 571 MWel a year before — and amounting to around 1% of Europe's total hydrogen demand. Under construction: 3.7 GWel, of which 78% sits in just five countries, with Germany at about 1.3 GWel and Sweden at 0.8 GWel. And reaching final investment decision in the first half of 2026: 0.5 GWel, across the whole continent, in six months.
The FID bar is the one that describes the future, and it is the smallest of the three.
The diagnosis moved, and it moved somewhere harder
For two years the sector's self-description was a cost problem: green hydrogen is expensive, the compliance rules make it more expensive, bring the cost down and the market appears. The figures published this autumn point somewhere else.
The Hydrogen Council's 2026 compass reports $130 billion of committed clean hydrogen investment, 6.9 Mtpa of committed capacity across over 570 projects, of which 90% are under construction or operational, and operational capacity up 70% to approximately 1.7 Mtpa.[2] Those are not the numbers of a collapsing industry.
And then the sentence that explains the gap between those numbers and the 46 MWel:[2]
with 4.2 Mtpa of binding offtake contracted, 75% of which for existing hydrogen uses like refining and ammonia
Three quarters of all binding offtake in the world is for hydrogen that is already being consumed today, made from natural gas, in refineries and ammonia plants. The European construction pipeline says the same thing from the other end: refining accounts for more than half of planned end-use demand at 54%, followed by steel at 31%.[1]
This is the finding, and it is not the one the sector tells about itself. The contracted market is a substitution market. It is a known buyer, with an existing plant, existing logistics and a single question — price — which this blog has already examined from the buyer's side: without policy support, the maximum acceptable hydrogen cost is below USD 2/kg for most sectors and regions, and for steel it is negative.
New demand — the mobility, the power generation, the aviation fuel — is where the announced gigawatts live and where the binding contracts do not.
The cost figure that is about to circulate, and what it is not
Where the capital that still moves is going
Figure 3
Where the capital that is still moving goes
Order intake by technology at one listed manufacturer, third quarter of 2026.
Order intake up 473% year on year. Segment revenue NOK 89 million, up 65%.
Segment revenue NOK 23 million, down 91%. The report gives no year-on-year percentage for this order intake figure.
Group order intake rose 339% to NOK 253 million while group revenue fell 63% to NOK 112 million — a company taking more orders and recognising much less income, which is what the start of a thin recovery looks like from inside. The split is the finding: PEM revenue rose 65% in the quarter while alkaline revenue fell 91%. One manufacturer is not a market, but it is a clean read on which technology the scarce capital is choosing.
Source: Nel ASA, third quarter 2026 financial results · Data as of 2026-10-15
Show the data ▸Hide the data ▾
| Segment | Order intake, Q3 2026 | Revenue, Q3 2026 | Revenue change |
|---|---|---|---|
| PEM | NOK 245M | NOK 89M | up 65% |
| Alkaline | NOK 8M | NOK 23M | down 91% |
| Group | NOK 253M (up 339%) | NOK 112M | down 63% |
One manufacturer's quarter is not a market, but it is a clean instrument. Nel's third quarter, published 15 October 2026: group revenue down 63% to NOK 112 million, group order intake up 339% to NOK 253 million, order backlog up 34% to NOK 1,314 million.[3]
Taking many more orders while recognising much less income is what the beginning of a thin recovery looks like from inside a factory. The split is the interesting part. PEM revenue rose 65% in the quarter, with order intake up 473%. Alkaline revenue fell 91%.[3]
Inside the same company, with the same sales force and the same market, one technology is taking almost all of the orders. For a project selecting equipment, that is a supplier-continuity question before it is a technology question.
Two slower forms of failure
No large project cancellation with a company document behind it appeared in September or October 2026 — and that absence is itself informative, because the previous twelve months were full of them. What the period produced instead were two quieter failure modes.
The first is cost overrun that needs new capital. Stegra, Europe's flagship hydrogen-based steel project, published the result of a 100-day review under new owners on 30 September 2026, stating that additional capital is required to complete the project and that estimated costs are significantly higher than assumed in June, attributing it to ramp-up costs following a prolonged scaling back of work and to inflation.[4] The company gives no figure, and none is invented here. It also states that the timeline for green steel production remains as expected.[4]
The second is the indefinite postponement. A regional Spanish newspaper reported on 19 October 2026 that Naturgy has postponed its La Robla hydrogen plant, with the reason given as awaiting greater visibility from offtakers on their future demand for renewable hydrogen.[6] That is a single source reporting a statement to its newsroom, with no company release located — so it carries none of the argument on this page and is included only because the stated reason matches, word for word, what the figures above describe.
For a lender, the second is the more awkward of the two. A cancellation is priced the day it is announced. A project that is neither cancelled nor progressing has no date on which anyone has to mark anything.
The clock that is actually running
The IEA's June review put a date on the arithmetic: 22 Mt of potential production in announced projects may lose any chance of beginning operation by 2030, with two-thirds of it in Europe, North America and Latin America, unless investment decisions are taken by early 2027.[5]
Against which Europe took 0.5 GWel of final investment decisions in six months.[1]
What to do with this before your next model
- Carry the 46 MWel with its boundary or not at all. Commissioned, Europe, half-year, electrolysis. It will be quoted without those four words and it means something different each time one is dropped.
- Check whether your project is the size the rest of your plan assumes. A 5 MWel average is below the scale at which a pipeline connection, a cavern or a project finance package is coherent. If your model has infrastructure sized for 100 MW, the comparable population is no longer what is being built.
- Classify your demand as substitution or new. Substitution has a buyer with a plant and a price threshold. New demand needs the buyer to build something first. Three quarters of all binding offtake is in the first category.
- Do not let €3.9/kg into an LCOH comparison. Its own publisher says it is not a levelised cost.
- Add supplier continuity to equipment selection. One manufacturer's alkaline revenue fell 91% in a quarter while its PEM order intake rose 473%. Which production lines will exist at your delivery date is now a diligence item.
- Put a date on your own FID. The published arithmetic says the European 2030 cohort is decided by early 2027. If your decision is after that, you are in a different cohort and should model a different one.
The honest reading of this autumn is not that the sector is failing. $130 billion is committed, 90% of committed projects are under construction or operational, and operational capacity grew 70% in a year. The honest reading is narrower and more useful: what is being financed is the replacement of hydrogen that is already being burned, at the scale a refinery needs, by buyers who already exist — and almost everything else is still waiting for a contract.