A hydrogen policy headline usually arrives on a materials desk already converted into optimism. A government announces an auction volume in gigawatt-hours, and someone asks the powder supplier what it means for next year. There is no honest way to answer from a headline denominated in energy or money when the question is about kilograms of stabilized zirconia, in a named grade, at a particular plant and date. Between the two sit an auction rule, a construction deadline, a financing decision and a choice of electrolyte chemistry — each able to break the chain.
What a Buyer Cannot Read Out of a Hydrogen Headline
Korea is the cleanest case, because it runs the first dedicated hydrogen power bid market: a clean hydrogen portfolio standard in force since December 2022 and managed by the Korea Power Exchange since January 2023, with volumes published in gigawatt-hours and winners by plant. Policy now sits with Korea's climate and energy ministry, which took the energy portfolio on 1 October 2025.
That is more disclosure than most hydrogen programmes offer, and it still yields no order forecast: the results name the plant, not the technology inside it. The split of Korean bid-market awards between SOFC, PAFC, MCFC and PEM is not published for any round. What is public is which manufacturers appear — Doosan Fuel Cell principally with phosphoric-acid cells, Bloom SK Fuel Cell, Mico Power and HD Hydrogen with solid oxide — and Doosan's share of awarded volume, 73 per cent in 2024 and 60 per cent in 2025. For a ceramics buyer, the number you want is the one nobody prints.
Korea's Two Bid Markets Are Different Instruments
Treating "the hydrogen auction" as one thing is the common error. There are two, and they behave nothing alike.
The market that has actually bought fuel cells is the general one, open to grey hydrogen. It awarded 1,430 GWh in 2023 against a 1,300 GWh target, 1,314 GWh in 2024, and 1,355 GWh across 52 plants announced 17 August 2025.
The clean-hydrogen market, which is what people mean by "clean hydrogen policy", has awarded one contract in its existence: 750 GWh to Korea Southern Power for Samcheok unit 1 in December 2024, out of 6,500 GWh offered — a co-firing thermal plant, not a fuel cell. The 2025 round offered 3,000 GWh and was cancelled on its closing day.
Both markets now face 2026 uncertainty. The ministry set 930 GWh — about 125 MW on its own figures — for the general market and 500 GWh for the clean market: roughly 30 per cent below the 2025 general award, one-sixth of the clean volume announced for 2025; and, as reported at the time, it would re-examine whether to continue the general market after three years. Those auctions were to be held in the second half of 2026, and as of this writing no 2026 opening, closing or award had been announced for either market.
The Lag Is in the Rules, and an Award Is Not an Order
General-market winners get 24 months to build, clean-hydrogen winners 36, with generator penalties for overrun, and commercial operation for 2026 falls in 2028 and 2029. That timing comes from the tender terms, not a forecast.
The project level agrees. Mico Power won two sites in the 2025 general round, 9.9 MW at Namyangju and 2.85 MW at Yangju; design runs from 2025 and equipment supply completes in 2027. So does the revenue level: Doosan Fuel Cell states that main-equipment deliveries for its 2023-round wins began in earnest in Q4 2024, roughly twelve to fifteen months after award.
That is the whole lag claim the public record supports, and it is award-to-equipment. We found no published interval between a stack order and the electrolyte-powder purchase behind it. A powder lead time derived from an auction date is an assumption wearing a citation.
Nor does an award reliably become equipment. From a 29 June 2026 rated credit opinion on Doosan Fuel Cell: 2025 net new orders were negative — minus KRW 43.1 billion overall, main equipment minus KRW 140.1 billion against KRW 97.0 billion of long-term service agreements, with a KRW 346.9 billion contract terminated in April 2025. Some of the projects won in 2024 and 2025 were cancelled or cut, and the main-equipment backlog fell to about KRW 30 billion at 31 March 2026, from KRW 651 billion at the end of 2024. The same phenomenon appears elsewhere: the IEA's Global Hydrogen Review put announced 2030 low-emissions hydrogen production at 49 Mt in its 2024 edition, 37 Mt in 2025 and 27 Mt in 2026, against 4.3 Mt committed. Announcements are being withdrawn faster than they are converted.
Not Every Solid-Oxide Cell Contains Zirconia
"Solid oxide" names an architecture, not a material — the point the market discussion almost never reaches, and the one that matters most to a powder buyer.
The third row changes how a Korean auction result reads. Doosan's solid-oxide line at Gunsan runs on a Ceres Power licence from a November 2023 technology agreement; the plant completed in June 2025 and entered mass production that July at 50 MW a year. The peer-reviewed literature on Ceres metal-supported cells has described a gadolinia-doped ceria electrolyte with a ceria-nickel cermet anode. One qualification: Ceres's own current materials do not name the electrolyte. It describes steel-supported solid oxide cell technology and stops there, so the attribution rests on literature rather than a manufacturer statement, and we will not put it higher.
Even at that caution the consequence holds. A Korean award won by that line is not automatically a zirconia order; one won by Bloom SK Fuel Cell or Mico Power comes from a different stack family, whose electrolyte has to be checked separately. Which manufacturer won is a materials question before a commercial one, usually answerable because stack families are public even where auction technology splits are not. Choosing between the zirconia grades is then a separate decision, set out in the YSZ versus ScSZ comparison.
Electrolyte-Supported or Electrode-Supported: Roughly Ten Times the Electrolyte
Solid-oxide electrolysis is the same chemistry family and a different order book. 8YSZ is the SOEC electrolyte of choice, selected over scandia-stabilized zirconia and doped ceria on cost, with long-term stability demonstrated in fuel-electrode-supported cells at Jülich, Topsoe, EPFL, CEA, DTU, DLR and EIFER.
The difference that reaches a powder order is thickness. Electrolyte-supported cells of the Sunfire and Kerafol type use a nominal 90 µm 3YSZ substrate; published anode-supported cells carry 8–10 µm of YSZ on a 300–400 µm nickel-YSZ support. Electrolyte mass per unit of active area is linear in thickness, so 90 µm against 8–10 µm is roughly a nine- to elevenfold difference in mass for the same cell area, nothing else changed. That is the whole of the claim — not a ratio of powder demand per megawatt, because power density, yield and stack replacement differ too. Why megawatts cannot be turned into tonnes at all is set out in our piece on fuel cells for AI data centres.
It is a different qualification problem as well as a different mass: a 30-cell 3YSZ electrolyte-supported stack has been reported degrading at 10.5 mV per thousand hours (0.8 per cent per thousand hours) — the kind of degradation-rate figure a buyer has to ask for separately by architecture.
And 2025–26 is when the solid-oxide electrolysis build-out stopped rather than started. Topsoe opened Europe's first industrial-scale SOEC factory at Herning in October 2025, then said on 18 May 2026 that it would hibernate the plant "until demand is robust", and cancel a planned second factory in the United States, citing "slower than expected development in targeted clean hydrogen markets". On 11 August 2026 thyssenkrupp nucera abandoned its own SOEC mass-production plans and impaired its Arnstadt pilot line, about a year after that line opened. The counter-example is Elcogen, whose Tallinn plant opened on 17 September 2025 raising capacity from 10 MW to 360 MW of combined SOFC and SOEC capacity. Either way, the SOEC story is another instance of the watch-list's caution below: track survival to financing, not award count.
What the Deployed Fleet Actually Runs On
Bloom Energy's Form 10-K for the year ended 31 December 2025 says it plainly: "While most Energy Server deployments use natural gas as fuel, there is no combustion...". Nearly all product revenue has come from that power-generating server; its electrolyser is an early-stage market. Read that against a Korean general-hydrogen market open to by-product and reformed hydrogen. The demand a solid-oxide powder seller feels in 2026 is a power story, not a hydrogen-policy one.
Other jurisdictions reinforce that. The US 45V credit is a production credit for hydrogen and says nothing about fuel cells; the July 2025 reconciliation law moved its qualifying test to beginning construction before 1 January 2028 — a construction test, not a placed-in-service test. The DOE hydrogen hubs are a lesson in vocabulary: two were terminated in October 2025 and five listed as retained on 14 April 2026, yet by August 2026 those five remained stalled, with only 26 per cent of Phase I disbursed. Retained is not funded.
The European Hydrogen Bank's third auction, announced 7 May 2026, awarded €1.09 billion to nine projects in seven countries totalling about 1.1 GW. The published results do not state electrolyser technology — in any of the three — so they cannot be read as solid-oxide demand. What the EU does specify is origin: no more than 25 per cent of stack capacity — counting surface treatment, cell unit production and stack assembly — may be sourced from China, evidenced at financial close, at start of operations and at the end of the subsidy period.
Japan makes the same point differently. Of the four projects certified under its Hydrogen Society Promotion Act on 27 March 2026, two are ammonia import routes and two are membrane water electrolysis for process heat; none of the four buys a ceramic cell.
A Watch-List Instead of a Forecast
Five things are checkable, and they beat any tonnage projection.
- Read the statutory deadline in the award, not the press release. A 2026 general-hydrogen win carries a 2028 operation date and, under the terms in force when the round was set, a 24-month build deadline; a clean-hydrogen win, 2029 and 36 months. That sets the equipment window.
- Track survival to financing, not award count. Award volume is a ceiling, not a forecast.
- Identify the stack family, then the electrolyte. Metal-supported ceria, anode-supported thin-film YSZ and electrolyte-supported 3YSZ are three different purchase orders, and the licence behind a line tells you more than its megawatts.
- Ask which support architecture the cell uses — electrolyte-supported or electrode-supported — before sizing anything. The thickness difference alone is roughly tenfold per unit area, and it cuts across fuel cells and electrolysers alike.
- Write the specification so a substitution cannot arrive silently — stabilizer and mol% in the grade name, named dopants, measurement method on the certificate. The grades on our SOFC electrolyte powders page are stated that way, and the ScSZ supplier landscape covers who can quote against it.
Frequently Asked Questions
Does a Korean hydrogen auction win mean a zirconia order?
Not automatically. The results do not publish the technology behind each winning plant, so no round's solid-oxide share is public. And a confirmed solid-oxide win may still not be zirconia: Doosan's line runs on a Ceres Power licence, and the peer-reviewed literature on Ceres metal-supported cells describes a gadolinia-doped ceria electrolyte — an attribution from the literature, not from Ceres, whose own current materials do not name it.
How long after an award does the electrolyte powder get bought?
We could not find one. What is published is award to equipment: a 24-month construction deadline for general hydrogen and 36 for clean; a 2025 winner with equipment supply completing in 2027; and one manufacturer's 2023-round deliveries beginning in earnest in Q4 2024. The powder purchase sits inside that, and we will not invent the offset.
Is SOEC the same market as SOFC for a powder supplier?
Same material family, different order book. 8YSZ is the SOEC default, chosen over scandia-stabilized zirconia and doped ceria on cost. The roughly tenfold difference in electrolyte mass is a property of support architecture, not of application: an electrolyte-supported cell of the Sunfire and Kerafol type — used in both SOFC and SOEC mode — carries a nominal 90 µm 3YSZ substrate against 8 to 10 µm in a published fuel-electrode-supported cell, and electrolyte mass per unit area is linear in thickness.
Is hydrogen policy the demand driver for solid oxide fuel cells today?
Not on the evidence. The deployed fleet runs largely on natural gas by its own manufacturer's account, and the Korean auction that has actually bought fuel cells is the one open to reformed and by-product hydrogen. Hydrogen policy is a real channel with a rule-based lag — and in Korea, currently, a shrinking one under review.
References (Public Sources)
- Korea's clean hydrogen portfolio standard, the designation of the Korea Power Exchange as managing agency, the clean hydrogen certification notice in force 4 March 2024, bid market results for every round, and the climate and energy ministry's June 2026 announcement of 2026 volumes and operation years, and press reporting on the general-market review.
- KIS (한국신용평가) credit opinion on Doosan Fuel Cell, 29 June 2026 — construction deadlines, 2025 net new orders, the April 2025 contract termination, the backlog series, and the Gunsan solid-oxide plant under a November 2023 Ceres Power licence; plus Korean trade reporting on the Mico Power schedule.
- IEA Global Hydrogen Review, 2024, 2025 and 2026 editions; Bloom Energy Form 10-K for the year ended 31 December 2025; Topsoe investor release of 18 May 2026; thyssenkrupp nucera's 11 August 2026 announcement as reported; Elcogen release of 17 September 2025.
- European Commission, European Hydrogen Bank third auction results of 7 May 2026 and the auction terms on stack sourcing; US Section 45V final regulations and the July 2025 reconciliation law; DOE hub terminations, the 2026 retained-project list and disbursement reporting; Japan's four certified projects, 27 March 2026.
- Materials literature: a 2023 Journal of Materials Chemistry A review of solid oxide electrolysis cell materials; Journal of Power Sources modelling of metal-supported cells with gadolinia-doped ceria electrolytes; an International Journal of Hydrogen Energy lifespan study of 30-cell electrolyte-supported stacks; Elcogen cell development papers; Sunfire and Kerafol electrolyte-supported cell characterization; tape-casting studies of thin-film YSZ on nickel-YSZ supports.
Figures are as published at the time of writing; several moving parts, the 2026 Korean auction outcome above all, were not decided.
Nami Tech Solutions (NTS) sources solid-oxide electrolyte powders against a written specification rather than a market narrative: stabilizer and mol% in the grade, named dopants, measurement method on the certificate.
