Doosan Fuel Cell

Phase A — Understand the business

Company Overview

Doosan Fuel Cell makes stationary fuel-cell power-generation systems — refrigerator-to-shipping-container-scale units that sit on the ground and turn natural gas or hydrogen into electricity and heat, 24/7, without a flame. It was carved out of Doosan Corporation in October 2019 as a standalone listed entity, HQ in Iksan-si / Gunsan, Jeollabuk-do, South Korea, ~541 employees. The US arm operates as HyAxiom, Inc. (formerly Doosan Fuel Cell America), led by Jeff (Hyung Rak) Chung.

What it actually sells today (the P&L business):

• Phosphoric Acid Fuel Cells (PAFC) — the legacy cash engine. Model families "400 NG" (natural-gas-fed, ~440 kW class) and "M400" (hydrogen-fed). PAFC is a mature, ~40%-electrical-efficiency technology; Doosan is the dominant domestic supplier into Korea's policy-driven stationary market.
• Solid Oxide Fuel Cells (SOFC) — the new business, licensed from Ceres Power (UK), mass production begun July 2025 (see Lens 4). Higher efficiency (60%+), aimed at AI data centers, microgrids, and marine auxiliary power. Zero commercial orders booked as of Q3 2025.

How it makes money — the demand engine is Korean industrial policy, not a free market. Doosan's revenue is overwhelmingly a function of Korea's Clean Hydrogen Portfolio Standard (CHPS), which since 2024 obliges power companies to procure a rising share of generation from hydrogen/fuel cells via reverse auctions offering 15-year purchase contracts. Contracts are effectively take-or-pay, ultra-long-dated (deals run to 2046), which is a structurally attractive revenue profile — if the units get built and the policy holds.

Recent contract structure evidence:

• UH Power — 40 MW, ₩100B (~$73M), running through 2046.
• KEPCO + local utilities — 20-year PPA worth ₩96.4B, Nov 2025.
• A $2.65B fuel-cell order is referenced as a 2025 catalyst — magnitude unverified against a primary filing; treat as ``.

Customers/suppliers/competitors: Customers are Korean IPPs/utilities and hydrogen-power developers (KEPCO, Korea Hydro & Nuclear Power, UH Power, SK ecoplant, Hyosung Heavy). Key supplier/licensor = Ceres Power for SOFC stack IP. Competitors: Bloom Energy (US, SOFC) globally; FuelCell Energy, Plug Power (US); domestically Doosan is near-monopoly in PAFC but faces SK ecoplant/Bloom's Korean SOFC push.

``

Supply Chain

The chain is short, policy-gated at the demand end, and IP-gated at the tech end:

Upstream inputs Doosan Fuel Cell End customer / offtake
────────────── ──────────────── ─────────────────────
• Ceres Power (UK) ── SOFC stack • Cell/stack manufacturing • KEPCO + local utilities (20-yr PPAs)
 IP + design licence ─────────► (PAFC: own tech; • Korea Hydro & Nuclear Power
• Platinum-group catalysts SOFC: Ceres metal-supported) • UH Power (40MW→2046)
 (PAFC electrode) • Balance-of-plant integration • SK ecoplant, Hyosung Heavy (partners)
• Steel / BoP components • Gunsan/Saemangeum SOFC plant • CHPS auction market (15-yr contracts)
• Natural gas / hydrogen (fuel, 50MW/yr line (from Jul 2025) • US market via HyAxiom (stalled 2025)
 supplied by offtaker) • Samchully (SOFC city-gas JV) • Marine auxiliary (SOFC, type-approved)

Named chokepoints / single-source dependencies:

1. Ceres Power is a single-source IP dependency for the entire SOFC growth story. Doosan is Ceres's first licensee to reach mass production — but Ceres reported −26% H1-2025 revenue and launched a "business transformation" restructuring. A distressed licensor is a real supply-chain risk to the crown-jewel product. (Offsetting: Goldman upgraded Ceres to "Buy" on the data-center opportunity, and Ceres signed a multi-GW Centrica UK/Europe deal in 2026 — the ecosystem is not dead.)
2. Demand is single-sourced to Korean government policy (CHPS). No CHPS auction, no order book. This is not a diversified customer base; it is one regulatory mechanism.
3. Fuel is the customer's problem (offtaker supplies gas/H₂), which de-risks Doosan on input-cost volatility but ties unit economics to Korea's hydrogen-price/subsidy regime.

Names or it didn't happen: Ceres Power, KEPCO, Korea Hydro & Nuclear Power, Kumho, LS Electric, UH Power, SK ecoplant, Hyosung Heavy, Samchully, HyAxiom, Air Products are all confirmed chain participants.

Competitive Advantages (moats)

Real moats:

• Domestic policy incumbency / share. Doosan took 110.42 MW = ~63% of the general-hydrogen CHPS bid volume across 2024 H1+H2. In a market defined by a national auction, ~two-thirds share is a genuine incumbency moat — reference plants, local service network, and a track record utilities can underwrite for 15-year contracts.
• Manufacturing-first-mover on Ceres SOFC. Being the world's first Ceres metal-supported SOFC mass-production line is a process/learning-curve head start if — and only if — the units sell.
• Chaebol backing. Parent Doosan Enerbility (largest shareholder, from the ~16.78% legacy Doosan Corp stake moved in 2020 ) can bundle fuel cells with gas turbines + SMRs into an integrated "AI-data-center energy" pitch (CES 2026). Balance-sheet and channel support a startup couldn't buy.

Bargaining power — weak on both sides, which is the problem:

• Over customers: low. Buyers are sophisticated utilities bidding in a reverse auction explicitly designed to compress margin. The record-revenue/record-loss 2025 (Lens 5) is the tell — Doosan is winning volume by pricing at or below cost.
• Over the licensor: low. The SOFC IP is Ceres's, not Doosan's; royalties flow out, and the technology roadmap is set in the UK.
• Durability: the PAFC moat is durable-but-shrinking (mature tech, policy-dependent). The SOFC moat is potential, not proven — a licensed, not owned, technology, against a far-better-capitalized Bloom Energy that already has the Oracle/Brookfield hyperscaler relationships (Lens 7).

``

Segments

No our figures on the shelf and Doosan does not break out clean product/geography segments in English-language sources. What is sourceable:

By product/technology (qualitative, ``):

• PAFC (legacy) — still ~all of recognized revenue through 2025; the ₩455B FY2025 top line is essentially PAFC deliveries into CHPS/RPS contracts.
• SOFC (new) — ₩0 recognized revenue as of Q3 2025; first sales targeted by end-2025, unconfirmed. The 50 MW/yr Gunsan/Saemangeum line (construction from 2022, mass production from July 2025, Ceres tech) is the segment that the entire equity re-rating is built on.

By geography: overwhelmingly South Korea. The US (HyAxiom) push stalled — the $560M of cancellations in April 2025 were partly the US/Korea tri-gen and domestic contracts that fell over (Lens 5/13). International is aspiration (CES marketing), not yet revenue.

Trend and cause: revenue decelerated then re-accelerated — FY2021 ₩381B → FY2022 ₩312B → FY2023 ₩261B (trough) → FY2024 ₩412B → FY2025 ₩455B → TTM (Mar '26) ₩500B. The 2023 trough coincided with the pre-CHPS policy gap; the 2024–25 recovery is CHPS auction volume flowing into deliveries. Crucially, the recovery is volume-led, not margin-led — see Lens 5.

``

---

Phase B — Measure performance

Earnings Result

The single most important fact in this dossier: FY2025 was record revenue and a record loss.

``

What drove it:

• Gross margin went from +8.8% to −13.7% in one year — the company is now losing money on the cost of goods itself, before opex. This is not a demand problem (revenue +10.5% YoY); it is a cost/pricing problem. Two forces: (1) CHPS auction pricing compressed PAFC margins as Doosan bid aggressively for ~63% share; (2) SOFC ramp costs — a brand-new 50 MW Ceres line running from July 2025 with near-zero utilization and no offsetting SOFC revenue absorbs fixed cost straight into COGS ``.
• Operating loss −₩105B, net loss −₩133B — the gap between them (~₩28B) reflects interest on a net-debt ~₩375B balance sheet plus non-operating items.

Balance-sheet flags (Lens 10 detail): debt ₩438.2B, cash ₩63.0B, net debt ~₩375.2B; current liabilities ₩429.5B. A company burning >₩100B/yr operationally with ₩63B cash and ₩429B of near-term liabilities has a financing question, likely met with parent support and/or capital raise (a family stake transfer of ~24%/₩700B to Doosan Heavy was referenced — intra-group, not fresh external capital).

Market reaction: despite this, the stock is +~400% over the trailing year and market capMarket capitalisationThe share price multiplied by the number of shares. What the market says the equity is worth. re-rated +195%. The market is explicitly not trading the current P&L — it is trading the SOFC/AI-data-center option (Lens 8/12).

Earnings Calls (sentiment trend)

No transcripts on the shelf (Korean-market, no Fool/Insider-Monkey coverage). Sentiment reconstructed from management actions and press:

• 2024 tone (Partners Meeting, KED coverage): confident, volume-focused — "63% of CHPS bid," "performance to improve as 2023 contracts recognized in H2". Classic order-book optimism.
• 2025 tone (pivot narrative): the messaging shifted hard from PAFC volume to "SOFC / AI data center / Ceres first-to-mass-production." The company leaned into the July 2025 mass-production milestone and CES 2026. This is a narrative pivot ahead of financial proof — third-party analysis flagged "the widest gap of the year between PR and tangible commercial events".
• What they started saying: "SOFC," "data centers," "60% efficiency," "Ceres," "AI era."
• What they stopped emphasizing: the deteriorating gross margin and the US/HyAxiom pipeline (post the April 2025 cancellations).

Read: management is doing what a policy-supplier pivoting to a growth-tech does — selling the future to bridge the ugly present. Credible only if end-2025/2026 SOFC bookings arrive. ``

Comps

Peer table — Doosan vs global fuel-cell peers. Multiples are `` with source/date or n/a. No multiple is fabricated.

The comp that matters — Doosan vs Bloom. Bloom trades at P/S ~29× on $2.02B FY25 revenue (+37%) with real hyperscaler contracts (Oracle up to 2.8 GW; Brookfield; Federal Pacific) and stock +1,247% in 52 weeks. Doosan trades at a derived P/S ~7.4× `` on ₩500B (~$365M) revenue that is loss-making and SOFC-orderless. So the market is giving Doosan a fraction of Bloom's sales multiple — arguably correct (Doosan has no booked SOFC revenue, worse margins, policy-single-demand), yet Doosan has already moved +400% on the same AI-data-center-power narrative that drove Bloom. The read: Doosan is the cheap, high-beta, higher-risk Korean proxy for the Bloom trade — more torque if SOFC works, more downside if it doesn't.

Stock-Price Catalysts

What has actually moved the stock >5% (5-yr lens, ``):

• 2021–2022: the original "hydrogen economy" policy hype and hangover — Korea's RPS/hydrogen roadmap drove the 2021 highs, then a multi-year de-rate into the 2023 revenue trough.
• May 2024 — CHPS auction market launches → order-book optimism; Doosan wins 63% share.
• April 2025 — $560M of contract cancellations → negative, but partly reframed as re-bidding into richer CHPS terms.
• July 2025 — SOFC mass-production start (Ceres) → the pivotal positive catalyst; the AI-data-center-power narrative attaches here. This is when the stock's trajectory decouples from the P&L.
• Late 2025 → 2026 — AI-data-center-power melt-up. Bloom +1,247%, FCEL/PLUG +200%+; Doosan +~400% on the $2.65B order reference + SOFC ramp. Goldman upgrades Ceres to "Buy" on the DC opportunity — a sympathy catalyst. CES 2026 Doosan/HyAxiom AI-energy showcase.
• 52-wk range ₩21,150 → ₩108,900 — a 5.1× peak-to-trough band in one year. Currently ₩52,300, roughly mid-range.

What the pattern reveals: this name reacts to (1) Korean hydrogen policy events and (2) the global fuel-cell/AI-data-center thematic far more than to its own earnings. It is a policy-and-theme beta vehicle, not a fundamentals-graded compounder. That cuts both ways: the next leg is set by the SOFC order print and the durability of the Bloom-led theme, not by margin recovery alone.

---

Phase C — Judge people & books

Management

• CEO: Lee Doo-soon (Doosoon Lee) — appointed to the CEO title in the Sept 2024 reshuffle, succeeding Jeong/Jung Hyung-rak. Awarded the Order of Industrial Service Merit at the 4th Hydrogen Day, Nov 2025 — signals tight alignment with the Korean government hydrogen agenda (which is the demand base). US arm HyAxiom run by Jeff (Hyung Rak) Chung.
• Track record: the executable achievements are 63% CHPS share and first-Ceres-to-mass-production — genuine operational wins. The debit is the 2025 margin collapse and $560M cancellations on their watch.
• Skin in the game / ownership: this is a chaebol subsidiary, not founder-owned. Largest shareholder Doosan Enerbility (from the legacy ~16.78% Doosan Corp stake); a 24% family stake (₩700B) transfer to Doosan Heavy was referenced. Insider alignment is via the group, not personal founder equity — archetype: professional managers inside a family conglomerate. Implication: capital-allocation and strategy answer to Doosan Group's portfolio logic (bundling fuel cells with turbines/SMRs for AI data centers), which can be a strength (deep pockets, integrated pitch) or a governance risk (related-party transactions, cross-subsidy — see Lens 13). our figures not on shelf → ownership detail ``.
• Capital allocation: the defining decision — build a 50 MW SOFC line (from 2022) and ramp it into a loss ahead of orders. Aggressive, forward-leaning, and entirely dependent on the SOFC market materializing. ROE has been negative FY23–25. This is a management team spending the balance sheet to buy a call option on SOFC/AI-data-center power.
• Red flags: loss-making while re-rating; heavy narrative/PR cadence vs. thin commercial proof; chaebol related-party structure. Not fraud-flavored — more "policy-supplier making a big, unproven, capital-intensive bet."

Forensic Red Flags

Accounting/financial risks (`` — no filings on shelf to ground income-statement forensics; this is the key limitation of a web-only dive):

1. Margin/COGS quality. A swing to −13.7% gross margin warrants scrutiny of (a) how CHPS long-term contract revenue is recognized vs. cost (percentage-of-completion vs. delivery — a classic area to probe on a 15-year/2046-dated contract book), and (b) whether SOFC ramp costs are expensed to COGS vs. capitalized. Cannot verify without the K-IFRS filings — flagged, not concluded.
2. Order-book vs. revenue conversion. Large multi-year headline orders ($2.65B referenced; UH ₩100B→2046; KEPCO ₩96.4B/20yr) create a backlog-recognition gap — the risk that reported "orders" are contingent (permits, offtake) and slip or cancel. This risk already materialized once: the April 2025 $560M cancellation. Backlog quality is the number-one forensic question.
3. Leverage / going-concern-adjacent. Net debt ~₩375B against ₩63B cash and a >₩100B annual operating burn — solvency leans on parent support / capital raises. Not a going-concern warning, but a real dilution/financing overhang.
4. Related-party (chaebol) transactions. Sales/support flows within the Doosan group (Enerbility, HyAxiom, group offtake) require related-party scrutiny — standard for Korean conglomerate subsidiaries.

Regulatory findings (required sub-section):

• SEC (EDGAR EFTS — LR + AAER): none. Doosan Fuel Cell has no CIK and is not an SEC filer; no EDGAR enforcement search is possible.
• Non-SEC / web search ("Doosan Fuel Cell" (FTC OR DOJ OR FDA OR settlement OR fine OR penalty) enforcement): no material regulatory enforcement actions surfaced. The April 2025 contract cancellations are commercial (permit delays + CHPS re-bidding), not regulatory penalties.
• Item 3 / Legal Proceedings: n/a — no 10-KForm 10-KA company’s audited annual report to the US regulator. The most complete thing it publishes. on shelf (non-US filer). Korean DART filings (KRX equivalent) not ingested this run.
• Verdict: No material regulatory or legal enforcement findings — verified via SEC EDGAR EFTS (LR/AAER, zero) and web search as of 2026-07-06. The genuine risks here are financial (margin, leverage, backlog conversion), not regulatory.

---

Phase D — Project & stress-test

Forward Projection

Web-only, no our figures, no consensus model on shelf → this is `` with explicit arithmetic and wide error bars. Two variables swamp everything: (1) does SOFC book real orders, and (2) does gross margin normalize off the CHPS-auction/ramp trough. Currency KRW; FY = calendar (Dec YE); shares ~65.5M. No our model create logged (unattended watchlist rule).

Base anchors: FY2025 revenue ₩455B, gross margin −13.7%, operating margin −23.1%, net −₩133B.

Bear (SOFC stalls, margins stay compressed):

• FY2026 rev ₩480B (PAFC/CHPS deliveries only, minimal SOFC) ; gross margin recovers to ~−2% as new-plant absorption improves but no SOFC scale; operating margin ~−12%; **net EPS ~ −₩900** . Stock de-rates toward the analyst-low cluster (₩22–35K).

Base (SOFC lands first orders in 2026, PAFC margins normalize toward mid-single digits):

• FY2026 rev ~₩560B (first SOFC + steady CHPS) ; gross margin ~+6%; operating margin ~−3% (still investing); **net EPS ~ −₩250** — narrowing loss, "credible pivot" year.
• FY2027 rev ~₩700B (SOFC ramp toward ~50 MW line utilization + CHPS) ; gross margin ~+12%; operating margin ~+4%; **net EPS ~ +₩350** — return to profitability.
• FY2028 rev ~₩850B; operating margin ~+8%; net EPS ~ +₩900 ``.

Bull (SOFC becomes an AI-data-center product with export/hyperscaler traction, Bloom-style):

• FY2026 rev ~₩620B, FY2027 ~₩950B, FY2028 ~₩1.3T ; operating margin to ~12%+; **FY2028 net EPS ~ ₩1,800+** . This is the case the +400% move is pricing.

The number that actually matters (per the +clinical logic, applied to a ramp story): not EPS — it's cash runway to the first commercial SOFC sale and to operating breakeven. With ~₩63B cash and >₩100B annual burn, Doosan needs either the SOFC order flow to inflect within ~4–6 quarters or fresh capital (parent or market). Runway-to-catalyst is the binary.

Forecast to log later (if promoted to conviction): 336260.KS books its first commercial SOFC order by 2026-06-30 and 336260.KS FY2026 operating margin > −5%. Not logged this run.

Bull vs Bear

Bull case. Doosan is the only mass-production Ceres SOFC licensee on Earth, sitting on a 50 MW/yr line, aimed at the single hottest demand vector in energy — AI-data-center power — where Bloom just proved the TAM by signing Oracle for up to 2.8 GW and re-rating +1,247%. Domestically Doosan owns ~63% of a government-mandated, 15-year-contract auction market, giving it a policy-protected annuity to fund the SOFC bet. Parent Doosan Enerbility can bundle fuel cells with gas turbines + SMRs into an integrated AI-power pitch (CES 2026). If SOFC books even a fraction of Bloom's traction, the current ~7× sales multiple `` is a fraction of Bloom's 29× — enormous re-rating headroom. Earnings surprise vector: the first SOFC order announcement.

Bear case (permanent-impairment risks).

1. The SOFC market never comes to Doosan. Bloom, with US hyperscaler relationships, capital, and a decade head start, takes the data-center SOFC market; Doosan's Ceres-licensed units find no export buyers and Korea's DC market is small. The 50 MW line becomes a stranded, loss-absorbing asset. Zero SOFC orders as of Q3 2025 keeps this live.
2. Margin never recovers. CHPS is a reverse auction engineered to compress price; Doosan's 63% share was bought with the margin (−13.7% gross in 2025). If policy pricing stays punishing, the legacy business is a low-margin volume trap.
3. Licensor/financing double-squeeze. Ceres is restructuring (−26% H1-2025); Doosan is net-debt ₩375B burning >₩100B/yr. A distressed licensor + a stretched balance sheet + a dilutiveDilutionIssuing new shares, so each existing share owns a smaller slice of the same company. raise is a plausible value-destruction path.

Pre-mortem (18 months out, thesis broke): It's early 2028. Doosan announced a handful of small domestic SOFC pilots but no hyperscaler/export order; Bloom took the AI-DC SOFC market. CHPS auction prices stayed low; gross margin limped to +3% but never scaled. A dilutive rights issue in 2026–27 capped the equity. The stock round-tripped from ₩100K+ back toward ₩25–30K as the AI-fuel-cell theme cooled and the "Korean Bloom" premium evaporated. What killed it: the order print never came, and the balance sheet forced a raise before it did.

Are multiples too high? On current fundamentals (loss-making, orderless SOFC), yes — most analyst targets sit below spot ₩52,300. The stock prices an option, not the P&L. Justified only if you underwrite the SOFC pivot.

Contrarian view (what the market is refusing to see): The bulls are trading Doosan as "Korean Bloom" — but the market may be underpricing the domestic annuity and overpricing the SOFC export dream. The durable, sourceable value is the 63%-share, 15-year-contract CHPS book; the SOFC/AI-DC leg is a lottery ticket the tape has already mostly paid for. If SOFC disappoints, the CHPS annuity is a floor far below today's price; if it works, the torque is real. The asymmetry is wide and binary — this is an option, priced like one.

Devil's Advocate (short-seller)

Dismantling the bull case:

• Revenue concentration is total and policy-dependent. ~All revenue is Korea, ~all demand is CHPS/RPS. One policy change (auction volume cut, price-cap tightening, subsidy rollback) breaks the model. There is no diversified commercial base to cushion it.
• The moat is rented, not owned. The SOFC crown jewel is Ceres's IP — Doosan pays royalties and does not control the roadmap, against a licensor that is itself restructuring. Bulls call it "first-to-mass-production"; a short calls it "a contract manufacturer for someone else's stack, with no orders."
• Most dangerous competitor bulls underestimate: Bloom Energy. Bloom already has the hyperscalers (Oracle 2.8 GW), the capital ($71.7B cap), positive FCFFree cash flowCash left after paying to run and maintain the business. Unlike profit, it is hard to flatter with accounting choices., and a US home market that is the AI-data-center buildout. Doosan is trying to sell into that from Korea, licensed, loss-making. That is not a fair fight for the export dream.
• Worst capital allocation: building the 50 MW line into a −₩105B operating loss with no orders. Aggressive spend against an unproven market, on a stretched balance sheet — the April 2025 $560M cancellation proves the order book is soft/contingent.
• What must hold for today's price: SOFC must convert to real, scaled, ideally export/hyperscaler orders within ~2 years and margins must normalize and the theme must stay hot and no heavily dilutive raise. That's a stack of conditionals.
• Growth disappoints by 20–30%: if FY2026–27 revenue lands 20–30% under the base and SOFC stays orderless, the loss persists, a raise becomes likely, and the "Korean Bloom" premium collapses — a plausible path back to the ₩22–35K analyst-low zone, i.e. −40% to −55% from ₩52,300.
• Single scenario that permanently impairs: Bloom (and/or Chinese SOFC entrants) lock up the global data-center SOFC market while Korea's CHPS pricing stays margin-negative — Doosan is left as a sub-scale, loss-making, licensed also-ran with a stranded SOFC line. Plausibility: moderate — not the base case, but far from tail.

Management Questions (ordered by information value)

1. How many binding, priced commercial SOFC orders have you booked to date, and what is the confirmed delivery/revenue-recognition schedule for the 50 MW Gunsan line through 2027? (This one answer moves the thesis most — it's the entire pivot.)
2. What is your cash runway at the current operating burn, and will funding the SOFC ramp to breakeven require an equity raise? If so, size and timing?
3. Break down the FY2025 gross-margin collapse to −13.7%: how much is CHPS auction pricing on PAFC vs. under-absorbed SOFC ramp cost, and what is the normalized gross margin at line utilization?
4. What are the exact economics of the Ceres licence — royalty rate, term, exclusivity, and your exposure if Ceres's restructuring impairs its roadmap or support?
5. What is your realistic path to export / non-Korean data-center SOFC orders, and who are the target hyperscaler/IPP customers — given Bloom's incumbency with Oracle/Brookfield?
6. Of the multi-year headline orders (the ~$2.65B reference, UH, KEPCO), how much is firm/unconditional vs. contingent on permits/offtake, and what is the realistic cancellation risk after April 2025?
7. What CHPS auction volume and price assumptions underpin your FY2026–27 plan, and how sensitive is profitability to a policy or price-cap change?
8. What is the delivered-cost and levelized-cost-of-electricity gap between your SOFC and Bloom's, and how does it close over time?
9. How do related-party flows with Doosan Enerbility / HyAxiom / the group affect reported margins, and how are they arm's-length-priced?
10. What SOFC line-utilization and volume gets you to operating breakeven, and in which quarter do you expect it?
11. What is the US/HyAxiom strategy after the 2025 cancellations — re-enter, or concentrate on Korea + Asia?
12. How defensible is the ~63% CHPS share as SK ecoplant/Bloom push SOFC into Korea?
13. What is your capital-allocation priority order over the next three years — SOFC capacity, deleveraging, or the legacy PAFC book?
14. What are the marine/maritime SOFC (type-approved) revenue prospects, and on what timeline?
15. What single external event (policy, competitor, technology) would most change your strategy, and how are you hedged against it?

---