Mitsubishi Heavy Industries

Phase A — Understand the business

Lens 1 · Company Overview

MHI is a ~¥5tn-revenue diversified heavy-industry conglomerate — one of Japan's three great heavy-engineering houses (with IHI and Kawasaki Heavy) and the industrial anchor of the Mitsubishi keiretsu. It sells large, long-lead-time capital equipment and the multi-decade service contracts attached to it, across four reportable segments:

1. Energy Systems — the crown jewel and the reason this name sits in the "energy" beat. Gas-turbine combined-cycle (GTCC) power islands via Mitsubishi Power, steam power, nuclear (reactors, fuel cycle, steam-supply systems), aero-derivative/aero-engines, plus CO₂ capture (CCUS), hydrogen/ammonia-fired turbines. ~36% of revenue.
2. Plants & Infrastructure Systems — metals machinery (Primetals JV), chemical plants, environmental/flue-gas, commercial ships, engineering (MHIENG), material handling. ~17% of revenue.
3. Logistics, Thermal & Drive Systems — the high-volume "general machinery" bucket: forklift trucks (Logisnext, a global top-3 lift-truck maker), turbochargers, HVAC/air-conditioning, engines. ~26% of revenue.
4. Aircraft, Defense & Space — Japan's prime defence contractor (missiles, warships, tanks, fighter structures, the GCAP next-gen fighter with UK/Italy), space launch (H3 rocket), and commercial-aerospace work-share (Boeing 787 wing boxes). ~21% of revenue.

Contract structure. The economics are project-and-annuity, not consumer: multi-year EPC-style equipment contracts (often with customer advances — see Lens 5/10) plus decades-long parts-and-service agreements on the installed base of turbines and reactors. A recent example: Mitsubishi Power + LNGPH signed a long-term parts-and-services deal on 22 June 2026. Customers are utilities, independent power producers (IPPs), national governments (defence), and — increasingly, and indirectly — hyperscalers driving behind-the-meter and grid gas build-out for AI data centres.

Frame for this dossier (per the assignment): analyze MHI as Energy Systems (gas turbines + nuclear = the AI-power beneficiary) + Aircraft/Defense/Space (the Japan-rearmament kicker) + two steadier industrial legs. Defence is a segment, not the identity.

Lens 2 · Supply Chain

Map: specialty materials → MHI heavy-forging/casting + assembly → power-island integration → utility/IPP/government end-customer → multi-decade MHI aftermarket.

• Upstream inputs (the chokepoints). Large gas turbines and nuclear vessels are gated by single-crystal/directional superalloy castings, large forgings, thermal-barrier coatings, and precision machining — a genuinely scarce industrial base. Japan's Japan Steel Works (JSW / Muroran) and Japan Casting & Forging are the classic domestic sources of ultra-large nuclear/turbine forgings; MHI also runs deep in-house forging/casting at its heavy works. Turbine hot-section castings are a named global bottleneck across all three OEMs.
• The company (conversion). Takasago Machinery Works (48 km west of Kobe; ~1.06 million m² site) is the centre of gravity — R&D, design, manufacturing, and the grid-connected T-Point 2 full-scale validation plant all on one campus, the only place an OEM can validate a full-size advanced turbine before shipment. US assembly/hot-section work runs through Savannah Machinery Works, Georgia.
• Downstream / named customers. J-POWER USA and Capital Power (Genesee, Alberta — 2× M501JAC ordered Dec 2020) are named GTCC buyers. FY2025's order surge was 35 large GTCC units, up from 25, concentrated in North America and Asia, plus 6× M501JAC (3.6 GW) for Saudi Arabia's Rumah-1 / Al-Nairyah-1, Mar 2025. On defence, the customer is effectively a monopsony — Japan's Ministry of Defense — plus the landmark export: Australia's general-purpose frigate program (Mogami-class), contract concluded April 2026.
• Single-source dependency (the moat and the risk): only three firms worldwide — MHI, GE Vernova, Siemens Energy — can build the largest heavy-duty turbines, from a handful of factories in Japan/Germany/US, and they are structurally unable or unwilling to add capacity fast enough. This lens is not generic: the named chain is superalloy forgings → Takasago/Savannah → J-POWER/Capital Power/Saudi utilities/MoD → MHI 20-yr service.

Lens 3 · Competitive Advantages (moats)

• Oligopoly / scale (the primary moat). MHI + GE Vernova + Siemens Energy account for >70% of global large-gas-turbine production capacity and build ~two-thirds of turbines for gas plants under construction. In the top-5 (adding Ansaldo, Baker Hughes) the group held 64.5% share in 2025. This is a three-body problem with no credible fourth entrant on the largest frames — the capital, the validation campus (T-Point 2), and the 20-year reliability record are the barriers.
• Technology edge. The M501JAC (J-series air-cooled) runs a 1,650°C turbine inlet temperature, ~540 MW combined-cycle block, >63% efficiency, and a demonstrated 99.5% reliability over 11,000 hours. MHI has demonstrated 30% hydrogen co-firing at grid-connected T-Point 2 (Nov 2023), is targeting 50% then 100% H₂ firing after 2030, and claims world-leading share in high-efficiency CO₂ capture — the decarbonisation option value on top of the gas franchise.
• Switching costs / installed base. A turbine or reactor is a 30–40-year asset locked to its OEM for parts, upgrades, and service — a recurring, high-margin annuity that competitors cannot contest mid-life.
• Defence national-champion moat. MHI is the central pillar of Japan's defence industrial base — effectively un-disruptable domestically; the constraint is Japanese budget and policy, not competition.
• Bargaining power. Currently high and rising on the sell side: multi-year backlogs and a capacity-constrained market let all three OEMs push price and take customer advances (Lens 5). Weaker on the buy side for scarce forgings/superalloys, where MHI competes with its own rivals for the same input base.

Lens 4 · Segments

Segment CSV is empty, so all figures are ``. Best authoritative split = MHI's own FY2024 (year ended Mar 2025) segment page, which resolves a live conflict in the sources: a secondary earnings-call summary put "Energy revenue ¥2.63tn" — that is wrong (it is ~orders/backlog, not revenue); MHI's own page shows Energy revenue ¥1,815.7bn (~36%), and the four segments sum to ~¥5.0tn (matching the ¥5,027bn FY2024 total).

FY2024 revenue by segment:

FY2025 directional moves (year ended Mar 2026):

• Energy Systems — accelerating, the profit engine. Orders +~50% YoY on GTCC + nuclear; revenue up on GTCC/nuclear backlog execution; ~60%+ of group business profit sits here (business profit ~¥267bn cited by the call summary, directionally the largest — treat the absolute as approximate ``). Cause: AI-data-centre gas build-out + Japan nuclear restart/fuel-cycle work.
• Aircraft, Defense & Space — fastest grower. Revenue +~35–40% YoY, business profit +~50%. FY2025 Defense & Space revenue target was ¥1.35tn (+31%), ~10.4% margin, ~20% of group revenue (up from 15% in FY2023). Cause: Japan's Defense Buildup Program (Lens 8) converting into deliveries.
• Logistics, Thermal & Drive — decelerating / flat. Revenue down YoY (forklift + HVAC volume softness), profit up on mix (engine shipments to Asia). The low-drama cash leg.
• Plants & Infrastructure — steady. Business profit +~40% on metals machinery / engineering; the smallest, most cyclical-industrial leg.

Geography: MHI does not cleanly break revenue by region publicly quarter-to-quarter; the incremental growth is disproportionately North America (GTCC) and Japan (defence + nuclear) ``. Exact geographic revenue split: n/a.

---

Phase B — Measure performance

Lens 5 · Earnings Result (FY2025, year ended 31 Mar 2026; announced 12 May 2026)

The print is a margin-and-orders story on flat revenue — quality up, top line not.

*FY2024 "business profit" base is approximate — MHI's own FY2024 release stated ¥383.1bn, which does not reconcile to a clean +21.8% off ¥432.2bn; the definitional line ("business profit" vs "profit from business activities" vs equity-method inclusion) shifts by year. Flagged, not smoothed.

REVENUE CONFLICT (surfaced, not resolved). MHI's press summaries call FY2025 "record revenue, +14.1%" — but the reported income statement (stockanalysis.com) shows FY2024 revenue ¥5,027.2bn > FY2025 ¥4,974.1bn, i.e. roughly flat / −1%, and below the ¥5.4tn guidance MHI had set. The two are only reconcilable if MHI restated/reclassified the FY2024 base downward (consistent with the noted "swing to positive equity-method contributions"). The absolute FY2025 figure (¥4,974.1bn) is agreed by both sources; the YoY growth and "record revenue" claim are not. Either way the earnings record is unambiguous.

• What drove it: GTCC (35 large units) + nuclear + defence execution; margin expansion from exiting low-margin legacy programs and a swing to positive equity-method income. Operating margin ~9.2% (¥460bn/¥4,974bn), up from ~6.6% — the whole bull case in one number.
• Balance-sheet flags — mostly good, one caveat: interest-bearing debt −20%, equity ratio to 37.3% — real deleveraging. BUT the record ¥893.4bn FCF is flattered by customer advances on GTCC orders (cash in before revenue) — a working-capital tailwind of an order boom that reverses if the order cycle rolls over. High-quality this year; not annuity-durable.
• Guidance (FY2026, ending Mar 2027): revenue ¥5,400bn (+8.6%), business profit ¥540bn (+24.9%), targeting a 10% business-profit margin, net income ~¥390bn, ROE ~12–12.8%, dividend ¥29; orders guided down ~12% to ~¥6.8tn on the absence of one-off mega-projects. Management flagging the order peak itself.
• Market reaction / what was priced: the stock had already run +100% into a 2 Mar 2026 all-time high of ¥5,208 and, by 2 Jul 2026, sat at ¥3,788 — a ~27% pullback. The record print did not make a new high — a tell that FY2025 was largely priced by early 2026.

Lens 6 · Earnings Calls (sentiment trend)

No transcripts on the shelf; synthesized from the FY2025 call summary + management commentary.

• Consistent drumbeat (last several halves): "record orders / record backlog," GTCC demand "staying strong", capacity is the binding constraint, defence backlog executing, energy-transition option value (H₂/CCUS).
• The tone shift worth noting: from "can we win the orders" (2023–24) to "can we build them fast enough" (2025–26). Management now leads with doubling gas-turbine capacity by FY2030 (>¥100bn / ~$618m capex across Takasago + Savannah) and concentrating engineers at Takasago — a supply-side, execution-risk narrative, not a demand-generation one.
• What they started saying: "business-profit margin to 10%," "shareholder returns" (dividend +¥4 to ¥29). What they stopped saying: anything about SpaceJet/regional-jet ambition — the aviation narrative is now purely Boeing 787 work-share + defence/GCAP (see Lens 9). The 2026 guide's honest "orders −12%" is a maturity signal — they are managing expectations off the peak.

Lens 7 · Comps

All multiples ``; where a figure could not be cleanly sourced it is n/a (never fabricated). MHI market cap ¥12.73tn ≈ $80–85bn.

Read: among the power-equipment majors, MHI is the cheapest on forward P/E (~32x) despite arguably the broadest franchise (gas + nuclear + defence + aero) — the core "conglomerate SOTP discount" bull point. The counter (bear): MHI also carries the lowest margins/ROE of the group — it historically lagged GEV/Siemens on ROIC through the weak-thermal 2010s and poor capital allocation, and only now is closing the gap. "Cheaper" is partly deserved discount, not pure inefficiency. GEV's ~85x EV/EBITDA is early-margin-ramp optics; none of these names is cheap in absolute terms — this is a relative-value trade within a hot sector.

Lens 8 · Stock-Price Catalysts (moves >5%, last ~5y)

Pattern: MHI re-rates on policy and order step-changes, not on quarterly EPS.

• Dec 2022 — the ignition. Kishida's three security documents (NSS/NDS/Defense Buildup Program) + the FY2023 defence budget +26.3% to ¥6.82tn, committing ¥43tn / $275bn over FY2023–27 and 2% of GDP by FY2027. This is the structural re-rating trigger for the whole Japanese defence complex.
• 2023–24 — AI-power supercycle overlaid. The gas-turbine shortage narrative (data-centre electricity demand) turned Energy Systems from a decarbonisation-headwind story into a scarcity-pricing story.
• Result: +100.7% in the 12 months to March 2026, into the ¥5,208 ATH (2 Mar 2026).
• Mar–Jul 2026 — the give-back. ~27% pullback to ¥3,788 despite a record FY2025 print — driven by (a) "orders to decline slightly in 2026", (b) a ~32x forward multiple flagged as elevated, and (c) China export-control geopolitical noise. The market reacts to the second derivative of orders, and that just turned negative.
• Ongoing catalysts (2026): capacity-doubling capex, the Australia frigate contract (Apr 2026), the LNGPH long-term service deal (22 Jun 2026).

---

Phase C — Judge people & books

Lens 9 · Management

• CEO: Eisaku Ito — President & CEO since 1 April 2025; previously EVP/CTO and Head of Technology Strategy from 2020; 30+ years at MHI, a gas-turbine man. Symbolically important: the company just put a turbine technologist in the top seat at the exact moment the turbine franchise is the value driver. Short tenure — track record as CEO is not yet gradeable.
• Chairman: Seiji Izumisawa — President/CEO 2019–2025, now Chairman. This is the capital-allocation turnaround architect. On his watch MHI restructured the loss-making thermal-power business (into the MHI Power reorganisation), divested machine tools, wound down and killed SpaceJet (2023), and delivered record FY2023 orders/revenue/profit. The re-rating is substantially his portfolio pruning finally showing up in margins.
• Capital-allocation history — genuinely improved, off a bad base. The 2010s were poor (weak thermal bets, the SpaceJet catastrophe — Lens 10). The last ~5 years are a credible pivot: exit low-return programs, concentrate capital on gas turbines / defence / aero-engines, deleverage (debt −20%, equity ratio 37.3%), and start raising the dividend (¥24→¥25→¥29). ROE ~12% and rising toward the ~12.8% guide — respectable for a heavy-industrial, still below GEV/Siemens.
• Skin in the game / insider ownership: Japanese-conglomerate norm — low direct insider ownership, cross-held within the Mitsubishi group; no insider-transactions.csv on the shelf → n/a. Governance is professional-manager, keiretsu-networked, not founder-owner.
• Archetype: professional managers executing a portfolio-cleanup + up-cycle. The right archetype for harvesting an oligopoly; the risk is the classic conglomerate temptation to re-diversify with the windfall rather than return it (watch this — see Lens 14 Q1).

Lens 10 · Forensic Red Flags

Acting as a forensic analyst; no filings on the shelf, so income-statement/balance-sheet detail is `` and coarser than a 10-K teardown.

1. Revenue-recognition & the "business profit" black box. MHI reports a proprietary "business profit" line (¥432.2bn) that sits below IFRS-style operating income (¥460.1bn) and folds in equity-method results — and the FY2024 comparative base does not cleanly reconcile to the stated +21.8% (Lens 5). Long-cycle percentage-of-completion accounting on EPC/defence contracts gives management real estimate latitude. Flag: the headline profit metric is company-defined and the YoY bridge is opaque.
2. FCF quality — advances, not annuity. Record ¥893.4bn FCF is driven by customer advances on the GTCC order surge. That is cash today against revenue tomorrow; it inflates FCF at the top of the order cycle and reverses as the backlog converts / if orders slow. Do not capitalise this FCF as steady-state.
3. Backlog-to-revenue conversion. Backlog +¥3tn to ¥13.2tn while revenue was flat. Bulls read a loaded spring; the forensic read is conversion risk — a turbine/defence backlog realises over 3–6 years and is exposed to input-cost inflation and the fixed-price/percentage-of-completion margin squeeze that has burned Siemens Energy (Gamesa) and GE.
4. Warranty / execution tail on complex kit — see the San Onofre case below. Advanced turbines and reactors carry catastrophic-defect risk; MHI's protection is contractual liability caps, not flawless execution.
5. KEIRETSU DISAMBIGUATION (critical, and a trap in the sources). The auto-defect and engine-certification suits that surface under "Mitsubishi" (hood/airbag class actions; the US DOJ engine-certification $237.5m settlement) are Mitsubishi Motors / Mitsubishi Fuso — legally separate companies, NOT MHI (7011.T). They must not be attributed to this issuer.

Regulatory findings (required). Per regulatory/regulatory-findings.md: MHI has no CIK → zero SEC EDGAR (LR/AAER) findings possible. Non-SEC + web check:

• San Onofre (SONGS) — the material MHI-specific matter. MHI supplied 4 replacement steam generators to Southern California Edison; one leaked radioactive coolant in Jan 2012 after <1 year; both reactors were shut and permanently closed June 2013. SCE arbitrated, escalating its claim from $4bn to $7.6bn. In 2017 the ICC tribunal (2-1) found MHI supplied defective equipment but capped the award at the contract liability limit — ~$125m (contract cap $138m, consequential/replacement-power damages excluded). Takeaways: (i) real quality-execution risk in MHI's most complex products; (ii) MHI's contracts cap catastrophic liability — a structural balance-sheet protection that repeated here.
• Current auto/keiretsu suits: not MHI (see red flag #5).
• Verdict: No open SEC action (no CIK); no material MHI-entity enforcement/litigation surfaced for 2024–2026 beyond ordinary-course; the one historically material case (SONGS) resolved in 2017 within contract caps. Verified via the regulatory-findings file (no EDGAR possible) + web search as of 2026-07-10.

---

Phase D — Project & stress-test

Lens 11 · Forward Projection

Shares outstanding ~3.37bn. FY2025 net ¥332.1bn → EPS ~¥98.5. FY2026 guide net ¥390bn → EPS ~¥115.7. Three fiscal years forward (FY2026 → FY2028, ending Mar 2027–29); every input labelled, outputs ``.

Interpretation: at ¥3,788, the base case is ~29x FY27 / ~26x FY28 EPS — the multiple compresses to the high-20s only if the base converts. The bear path leaves EPS essentially flat for three years, which at ~32x today is the real risk: not a blow-up, a de-rating on a stalled order cycle. Per the assignment and --watchlist rules, no forecast.ts forecast was created (that is a committed-base-case step, reserved for a human-gated /thesis pass).

Lens 12 · Bull vs Bear

Bull case. MHI owns one seat in a three-firm global oligopoly on the single most supply-constrained piece of the AI build-out — large gas turbines — at the exact moment demand structurally steps up (data-centre power) and the incumbents choose not to flood capacity. Layer on Japan's once-in-a-generation defence doubling (¥43tn/2% GDP by FY2027), where MHI is the un-disruptable national champion, plus a nuclear-restart/new-build optionality (SRZ-1200, SMRs, Japan's 20–22% nuclear-by-2030 target). Management has proven it can expand margin by pruning (thermal, machine tools, SpaceJet), the balance sheet is delevering, and the stock is the cheapest of the power majors on forward earnings with a credible SOTP re-rate still available. Earnings surprise vector: service/aftermarket margin on a swelling installed base.

Bear case (2–3 permanent-impairment / de-rating risks).

1. The order cycle peaked in FY2025. Management itself guides orders −12% in FY2026; industry gas-turbine demand slips from ~100 GW (2025) to ~70 GW/yr through 2030. You are buying after the second derivative turned negative — the exact setup for multi-year multiple compression even if earnings hold.
2. Margins/ROE still structurally below peers. MHI has never earned GEV/Siemens-class returns; the "cheap" multiple may be a permanent quality discount, not an arbitrage. Fixed-price long-cycle backlog is where Siemens (Gamesa) and GE destroyed value — conversion is not free money.
3. Expectations already in the price. ~32x forward is not a value multiple for a ~+9% revenue grower whose FCF is advance-flattered. A stalled cycle (bear EPS ~flat 3y) at 32x is a 30–40% de-rate.

Pre-mortem (18 months out, thesis broke): Data-centre gas orders air-pocket as hyperscalers pivot capex to on-site nuclear/SMR + grid, 2026 orders come in worse than the −12% guide, a fixed-price GTCC or frigate contract books a percentage-of-completion charge, customer advances unwind (FCF halves), and the yen strengthens — the stock round-trips to ¥2,800–3,000 (~pre-2025 levels) as the "peak-cycle Japanese industrial" tag reasserts.

Are multiples too high? In absolute terms, stretched (~32x fwd, ~26–29x on base FY27/28). Relative to GEV/Siemens, defensibly the cheapest — so the trade is relative-value + time, not deep value.

Contrarian view (what the market is refusing to see): the market is trading MHI as an order-momentum name and just marked it down on the −12% order guide — but the durable value is the 20-year aftermarket annuity on a record installed base being laid down right now, which is under-modelled and shows up as margin for a decade regardless of new-order cadence. The pullback is the market selling the peak of the wrong (order) metric.

Lens 13 · Devil's Advocate (short-seller)

• Where's the revenue really concentrated? Increasingly on two hot cyclical bets — GTCC and Japanese defence — booked at a cyclical/policy peak. GTCC rides a data-centre capex wave that hyperscalers can throttle overnight; defence rides a single government's budget that assumes political continuity on 2% GDP through FY2027 and beyond. Both are top-of-cycle.
• Why the moat is weaker than bulls think: the "oligopoly" is also a shared-scarcity story — MHI is behind GE Vernova (GEV leads global share ~18.5%; in Asia GEV 38% vs Mitsubishi 17%). The most dangerous competitor bulls underrate is GE Vernova, which is winning more units and monetising the cycle at 2x MHI's multiple — if turbine demand normalises, GEV's scale and share leave MHI as the marginal, lower-margin oligopolist.
• Worst capital-allocation moves: SpaceJet — ~$9bn / ~¥1 trillion incinerated over 15 years (2008–2023), never carried a paying passenger, killed by US pilot scope-clauses. Plus the 2010s thermal-power bets and the San Onofre defective-generator debacle. This management lineage has a documented capacity to torch capital on flagship ambition — and now it has a windfall to redeploy.
• What must hold for today's price: 10% business-profit margin achieved and sustained, backlog converting without fixed-price charges, orders not falling more than the guided 12%, and the yen not strengthening materially. Miss any two and the ~32x multiple is indefensible.
• Growth disappoints 20–30%: bear EPS ~flat → at a normalised ~18–20x heavy-industrial multiple, downside to ~¥2,300–2,600, i.e. ~30–40% below spot.
• Single permanent-impairment scenario: a major GTCC or nuclear quality/warranty failure (SONGS redux) at the moment of doubled output/rushed capacity — plausibility low-to-moderate, severity high, but contractually liability-capped.

Lens 14 · Management Questions (ordered by information value)

1. Orders are guided −12% in FY2026 — what is your internal view of the through-cycle (FY2027–30) gas-turbine order run-rate, and at what order level does the ¥5.4tn→¥6tn revenue plan break?
2. With record FCF, what is the explicit capital-return vs. reinvestment vs. M&A split for the next three years — and what is the guardrail that stops another SpaceJet-scale bet?
3. How much of FY2025's ¥893bn FCF is customer advances, and what is FCF ex-advances — i.e. the steady-state number?
4. What share of the ¥13.2tn backlog is fixed-price, and how are you hedged against input-cost inflation and percentage-of-completion charges (the risk that broke Siemens Gamesa)?
5. Bridge "business profit" (¥432bn) to IFRS operating income (¥460bn) and reconcile the FY2024 comparative — what exactly moved the base?
6. Where is the 10% business-profit-margin target by segment, and what is the sustainable Energy Systems margin once the current pricing-power window normalises?
7. What is the aftermarket/service revenue and margin today, and its trajectory as the record turbine installed base matures?
8. On hydrogen/ammonia and CCUS — what firm order value exists today vs. pilots, and when does it become a P&L line rather than option value?
9. Doubling turbine capacity by FY2030 — what is the incremental ROIC on that capex, and what utilisation assumption underwrites it if demand slips to ~70 GW/yr?
10. China export controls / geopolitics — what input (rare-earths, castings) or market exposure is at risk, and what is the mitigation?
11. Nuclear: what is the realistic revenue timeline for SRZ-1200 and SMRs vs. restart/service work, and how much is contracted?
12. Defence beyond Japan — after the Australia frigate win, what is the credible export pipeline, and how do you price the political risk of Japanese-origin defence exports?
13. What is the GCAP (next-gen fighter, UK/Italy) investment and margin profile, and how does it affect group capital intensity this decade?
14. How do you close the ROE/ROIC gap to GE Vernova and Siemens Energy — is it structural (mix) or executional?
15. What is your yen assumption in the FY2026 plan, and the earnings sensitivity per ¥10 move vs. USD?

---