Harmonic Drive Systems

Phase A — Understand the business

Lens 1 · Company Overview

Harmonic Drive Systems Inc. makes one thing supremely well: the strain-wave gear — the compact, high-ratio, near-zero-backlash precision reducer that sits in the joint of a robot and turns a fast, weak motor into a slow, strong, exquisitely controllable one. The company commercialised the technology in Japan in 1962 and the proprietary manufacturing process for the flexspline (the thin, elliptically-deforming steel cup at the heart of the device) remains a trade secret to this day. In plain terms: this is not a robot company. It is the picks-and-shovels supplier whose part is inside almost every high-end robot arm, cobot, semiconductor wafer-handler, machine tool, medical robot, and — the reason the stock has quadrupled — the wrists, elbows and shoulders of the humanoids now being prototyped in the United States.

Two reporting segments:

• Reduction Gears — $309.78M, +9.5% YoY, 77.8% of revenue. The core: harmonic (strain-wave) gears sold as components to robot OEMs and equipment makers.
• Mechatronics Products — $88.42M, −0.9% YoY, 22.2% of revenue. Actuators, servo systems and integrated motion units — the company's push up the value chain from component to sub-assembly, marketed as "Total Motion Control" to reduce integration work for robot builders.

Total FY3/26 revenue was ¥59,557M (+7.0% YoY) / ≈$398.2M.

Customers / suppliers / competitors. End customers are industrial-robot OEMs (Fanuc, Yaskawa, ABB, KUKA, Universal Robots and their Asian peers), semiconductor-equipment makers, machine-tool builders, medical-device firms, and — the swing factor — North American humanoid developers. Suppliers are specialty-steel and precision-machining inputs (the flexspline metallurgy is the guarded step). Direct competitors: Nabtesco (Japan, dominant in the adjacent RV-reducer segment and a harmonic supplier), the US-headquartered Harmonic Drive LLC (majority-owned by HDS — see structure below, so not a true rival), and the rising Chinese cluster led by Leaderdrive, plus Beijing CTKM.

Corporate structure — the three "Harmonic Drives" are one family. HDS Japan (6324.T) owns 74.7% of Harmonic Drive AG (Germany; the remaining 25.3% held by the Japanese state fund INCJ) and a majority of Harmonic Drive LLC (Beverly/Peabody, Massachusetts). So the "US market leader" and the "German maker" often cited as separate players consolidate up into 6324.T. This matters for Lens 2/4: the group is a genuinely global manufacturer, not a Japan-only exporter.

Contract structure. Component sales are largely book-and-ship purchase orders, not take-or-pay or recurring subscription — so revenue is order-driven and cyclical, tightly levered to robot-OEM and semiconductor-equipment capex. There is no annuity stream to cushion a downturn; the flexspline gets designed-in, which creates stickiness (see Lens 3), but each unit is still a discrete sale. Bookings, not backlog contracts, are the tell.

Commercial-layer anchor: the KB robotics map already names Harmonic Drive as the lead node under "Actuators / motors → joint torque," flags "high-torque compact actuators" as a supply chokepoint controlled by "Harmonic Drive + emerging Chinese," and lists "actuator cost curve" as an active bottleneck gating the sub-$50K humanoid. This dossier is the company-level fill of that node.

Lens 2 · Supply Chain

Upstream → HDS → end customer, named:

• Upstream inputs: specialty steels and precision alloys for the flexspline and circular spline; grinding/hobbing and heat-treatment tooling; bearings; servo motors and encoders for the Mechatronics units. The binding, proprietary step is the flexspline — the thin-walled elliptical steel cup that flexes millions of cycles without fatigue. HDS's 80-year sliding/metallurgy know-how here is the single-source dependency the whole robot industry has on HDS, not the other way around.
• HDS (the node): machines the guarded flexspline centrally in Japan, then does localized final assembly at three hubs — Japan, Harmonic Drive AG (Germany), Harmonic Drive LLC (Massachusetts). This "centralised component / localised assembly" architecture is deliberate: it keeps the trade secret in Japan while putting capacity next to Western customers (the Beverly +13%/+33% expansions, Lens 5).
• Downstream customers (named):
• Industrial robot OEMs — Fanuc (6954.T), Yaskawa, ABB, KUKA, Universal Robots, Kawasaki, Estun.
• Semiconductor-equipment makers — wafer-handling and lithography-adjacent motion (called out explicitly as a FY3/26 demand driver).
• Humanoid developers — the North American clients management now names as the growth pipeline (Lens 5). By strong inference from the KB positioning map this class includes the Figure / Tesla Optimus / Apptronik-tier Western builders, all of whom need exactly HDS's compact high-torque reducers — but HDS does not publicly name which humanoid OEMs are buying.

Chokepoints & single-source dependencies:

The supply chain's key asymmetry: HDS is itself the chokepoint for premium actuation. Its vulnerability is not an input it can't get — it is that a good-enough substitute (Leaderdrive at 40–60% of price, 70–85% of torque-density life) is being qualified in at its customers, starting in China. (Carry-forward candidate for refresh; unchanged structural map since this cold build.)

Lens 3 · Competitive Advantages (moats)

The moat is real and unusually clean for a component maker — but it is a share-of-a-growing-pie moat, not a pricing-power fortress.

1. Process IP / trade secret (the core moat). 60+ years of flexspline metallurgy and manufacturing tolerances that competitors have not fully replicated. HDS + Nabtesco + Leaderdrive together hold >63% of global strain-wave-gear share; HDS specifically is "widely regarded as the global market leader and technology pioneer" and "the dominant player" in high-end strain-wave reducers. In the premium tier the effective share is far higher than the headline.
2. Design-in switching costs. A robot joint is engineered around a specific reducer's backlash, stiffness and torque curve; re-qualifying a second source forces re-tuning of the control policy. The KB positioning file names this directly: "Actuator+sensor stack changes force re-training of policies". That is a genuine, if not impregnable, switching cost.
3. Brand / reliability in mission-critical use. In semiconductor and medical robotics, field failure is catastrophic; HDS's reliability reputation commands a price premium and incumbency.
4. Scale + vertical footprint. Three-continent assembly, R&D at 6.63% of sales ($26.41M, 151 R&D staff), and the move to integrated "Total Motion Control" modules that raise the switching cost from part to sub-system.

Bargaining power — the honest read. Over Western robot OEMs HDS has held strong hand: it is often the only qualified source for a compact high-ratio joint. Over Chinese OEMs its power has already broken — domestic harmonic suppliers went from <5% of units in Chinese-assembled robots in 2018 to >35% by 2024, with Leaderdrive taking the majority of that. HDS's own China revenue fell −28.0% in FY3/26 (Lens 4) — the moat leaking in real time in one geography. So: durable moat in the premium/Western tier; eroding moat in the commodity/China tier. The bull case needs the humanoid pie to grow faster than the China share leaks.

Lens 4 · Segments

By product:

The core gear business is re-accelerating off the trough (+9.5%); Mechatronics is flat, a mild negative for the "move up the value chain" story — the integrated modules aren't yet pulling their weight.

By geography:

The story the geography tells is the whole thesis in four rows. Japan/Asia roared (+22.5%) on the domestic robot + semiconductor-equipment recovery. Europe and North America were roughly flat in trailing revenue — even though North America is where the humanoid orders (not yet revenue) are booking (Lens 5); the design-win-to-shipment lag means the humanoid ramp barely touches the FY3/26 P&L. And China fell 28% — the single loudest data point in the entire dossier, the Leaderdrive displacement rendered in the accounts. Trend: core gears accelerating, Western revenue lagging the order book, China structurally leaking. (Segment mix moved materially — re-run on next refresh.)

---

Phase B — Measure performance

Lens 5 · Earnings Result (FY3/26, reported 2026-05-11)

The print in one line: revenue recovered, profit is still climbing out of a hole, and the market paid up for the guidance, not the result.

• Revenue ¥59,557M, +7.0% YoY — a beat vs the company's own forecast; management said "sales in Japan, China and Europe, mainly to robot manufacturers, exceeded the assumptions".
• Operating profit ¥2,567M, operating margin 4.3%. Off a depressed base — this is a trough-margin business right now, not a broken one.
• Ordinary profit ¥2,539M; net income ¥1,608M, −53.7% YoY. Net fell despite revenue rising — the profit-collapse headline.
• Gross margin 30.4% — structurally below Fanuc's ~37% (Lens 7); HDS is the specialist component maker, not the fully-integrated automation platform.
• Cash flow / balance sheet: operating cash flow ¥6,425M — comfortably above net income, a quality-of-earnings positive (the gap is depreciation on the capacity build). Equity ratio 72.2%, interest-bearing debt ~$88.4M against ~$127.7M cash — net cash, fortress balance sheet.

Why did profit fall while sales rose? — the crux of Lens 5/10. The company is carrying unabsorbed fixed overhead and depreciation from a capacity build-out ahead of the humanoid ramp: the Beverly, MA plant (+13% capacity Dec 2025, +33% more by Dec 2026), Japanese capacity for humanoid mass-production, and total FY3/26 capex of ¥5.69B / ≈$38.05M. Building the factory before the volume arrives depresses today's margin by design. Net margin compressed to 2.7% from 6.2% the prior year.

• Conflicting data point — surfaced, not silently resolved. One source cites operating profit "declined 94.4% YoY" and references an impairment loss, which is inconsistent with the audited FY3/26 OP of ¥2,567M at +141.5% guided growth into FY3/27. Most likely that stat describes an earlier interim trough or a specific prior-period comparison, not the full audited FY3/26 year. I treat the ¥2,567M OP / 4.3% margin / ¥1,608M net (−53.7%) figures (JPX TDnet-sourced) as the authoritative FY3/26 result and flag the −94.4% figure as unreconciled.

Guidance — this is what moved the stock. FY3/27: revenue ¥68,000M (+14.2%), operating profit ¥6,200M (+141.5%), net income ¥4,500M (+179.7%), dividend raised to ¥42.1 from ¥20. The upgrade rests on operating leverage: as the pre-built capacity absorbs recovering gear volume + the first humanoid orders, incremental margins are very high. Mid-fiscal-year the company had already revised FY3/26 up (+¥1B OP/ordinary) on 24 Apr 2026 citing robot + semi demand.

Market reaction: shares closed +8.2% after the earnings briefing when management outlined accelerating North American humanoid interest; the stock later traded up +670 to ¥7,790 on a Goldman Sachs target hike, and sat at ¥8,840 on 2026-07-05 — near the top of a 52-week range of ¥2,316–¥8,870 (a ~3.8x move off the low). What that says: the market is trading the FY3/27 leverage + the FY3/30 humanoid option, not the trailing 2.7%-net-margin reality.

Lens 6 · Earnings Calls (sentiment trend)

No transcripts on the shelf (web-only). From reported briefing coverage across the last several quarters, the tonal arc is clear and one-directional:

• 12–18 months ago (FY3/25 trough): defensive — weak robot/semiconductor capex, China softness, margins compressed. The stock underperformed in 2024 and 2025.
• Q2 FY3/26: the pivot. Management first put a number on humanoid demand — ¥1.3B booked in Q2, ¥2.5B expected for FY3/26 — and "outlined accelerating interest from North American clients". Recurring new phrases: "physical AI," "North American clients," "prototype production orders," "humanoid mass-production system," "Total Motion Control."
• Q4/FY3/26 (May 2026): confident and forward-leaning — the ¥68B / +141.5% OP guide, dividend more than doubled, and the striking forward-order comment that humanoid order size "could double in FY3/27, or even triple" on improving visibility into U.S. prototype orders.

What they stopped saying: the apologetic China / cyclical-weakness framing that dominated the trough. What they added: an explicit multi-year plan (the FY2030 target — Lens 11) that reframes HDS from "cyclical Japanese component maker" to "physical-AI infrastructure." Sentiment shift: decisively positive and promotional — which itself is a mild yellow flag (Lens 13): a management leaning into a re-rating tends to accentuate the option and downplay the China leak.

Lens 7 · Comps

Peer set pulled from the research index (topic: robotics) plus the obvious motion/reducer names the index and KB name: Fanuc, Nabtesco, HIWIN, Estun, Cyberdyne.

Read: even against Fanuc — itself a premium, high-quality automation name at ~27x EV/EBITDA and ~46x P/E — HDS is in a different universe: ~10x P/S vs Fanuc's ~7x EV/Sales, but HDS earns a ~4% operating margin vs Fanuc's ~26% EBITDA margin and a 2.0% ROE. On any normalised earnings basis HDS is the most expensive robotics-adjacent name in the peer set by a wide margin. The multiple only makes sense if you underwrite the FY3/30 plan and a humanoid TAM beyond it. Against the broad Japan machinery sector P/S of ~0.8x, HDS trades at a ~13x sector-relative premium — the definition of a story stock.

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

Web-sourced; the pattern matters more than any single date.

• 2024–2025 de-rate / underperformance: robot + semiconductor capex downcycle, China share loss → the stock lagged; it printed a 52-week low of ¥2,316.
• Q2 FY3/26 humanoid-order disclosure: first hard humanoid number (¥1.3B) + "accelerating North American interest" → +8.2% on the day. This is the catalyst that flipped the narrative.
• Broker upgrades: Goldman Sachs target hike → +670 to ¥7,790 in a session.
• FY3/26 results + FY3/27 guide + dividend doubling (May 2026): the +141.5% OP guide and ¥42.1 dividend → sustained re-rating toward ¥8,840.
• Macro/thematic beta: HDS is a constituent of humanoid-robotics thematic baskets (e.g. the KOID ETF universe) and moves with Tesla Optimus / physical-AI headlines and Nikkei robot-sector sentiment.

What the market actually reacts to for this name: (1) humanoid order data points and Western-customer commentary — the single biggest lever, wildly disproportionate to their current revenue contribution; (2) broker targets (thin float, high retail/thematic ownership amplifies these); (3) the robot + semiconductor capex cycle (the real, current earnings driver); (4) Tesla/physical-AI thematic beta. It reacts far less to the China erosion, which the tape has been willing to look past. That asymmetry — bidding the option, discounting the leak — is where the risk lives.

---

Phase C — Judge people & books

Lens 9 · Management

• Chairman & Representative Director: Mitsumasa Ito. President & Representative Director: Akira Nagai. CEO: Akira Maruyama (appointed June 2024). The Nagai name at President ties to the founding lineage (the company's post-war strain-wave commercialisation is a Nagai-family story) — a founder-adjacent, long-horizon governance culture typical of a Japanese precision-engineering house.
• Track record. The credential is the technology: 60+ years holding global leadership in a product they invented, with the flexspline process still un-cloned. That is a rare durable-execution record for a component maker. The recent, more mixed chapter: they let margins and China share slip through the 2024–25 downcycle before the FY3/26 recovery.
• Skin in the game. Modest at the individual level — CEO Maruyama owns ~0.014% of shares. This is normal for a large Japanese corporate (professional-manager model with founder-family board presence), but it means alignment runs through governance/culture and the state fund (INCJ holds 25.3% of the German subsidiary), not through large insider equity. Full domestic shareholder register (Nagai-family stake, cross-holdings) not sourced — a genuine gap; flag for a Japanese-filing pull.
• Capital-allocation history. Conservative and now investing counter-cyclically: net-cash balance sheet (72.2% equity ratio); FY3/26 capex ¥5.69B building humanoid capacity ahead of demand; a modest buyback (264,800 shares, ~$5.41M) and a dividend more than doubled to ¥42.1 for FY3/27. The read: prudent, reinvestment-first, willing to eat near-term margin to be ready for the ramp. The knock — ROE is only ~2% and the balance sheet is arguably over-capitalised ("highly liquid but capital-inefficient" ); a sharper allocator would lever the fortress balance sheet or return more.
• Founder vs professional-manager archetype. Hybrid — founder-family chair/president lineage over a professional-manager CEO. Implication for this stage: stability and technical continuity (good for guarding the trade secret and a multi-decade ramp), but likely slower and less aggressive on price/China response than a founder-CEO or a Chinese competitor would be.

Lens 10 · Forensic Red Flags

Web-only forensic read (no filings to tie out; every figure labeled):

• Earnings vs cash — clean/positive. Operating cash flow ¥6,425M exceeds net income ¥1,608M. The gap is depreciation on the capacity build, not accrual games. This is the opposite of a red flag — reported profit is conservative relative to cash generation.
• Margin compression is real and disclosed, not hidden. Net margin 6.2%→2.7% is driven by unabsorbed fixed cost / depreciation from pre-emptive capex, and (per one source) an impairment loss. This is an honest cyclical/investment trough, not manufactured — but it does mean the entire bull case is a bet on operating leverage that has not yet shown up.
• Inventory/receivables vs revenue: not sourced at line-item granularity (no filing on shelf) — flag as an open verification item; nothing in summaries suggests channel-stuffing, but I cannot tie it out.
• Balance-sheet quality: net cash, low leverage (D/E ~0.16x), 72.2% equity — extremely conservative; goodwill/intangible risk low for an organically-grown component maker (the German/US subsidiaries were acquired but are operating businesses, not goodwill balloons per available data).
• SBC / non-GAAP flattering: not a factor — Japanese GAAP, no material stock-comp add-back culture; HDS is not dressing up a non-GAAP number.

Regulatory findings (required sub-section).

• SEC (EDGAR): No CIK — HDS is not an SEC registrant, so no EDGAR LR/AAER search is possible. The "Harmonic … $16.4M securities settlement" hits in search are Harmonic Inc. (HLIT, a US video-delivery company) — a different entity; do not attribute to 6324.T.
• Non-SEC (FTC/DOJ/antitrust/recall): web search "Harmonic Drive Systems" (FTC OR DOJ OR fine OR penalty OR consent decree OR settlement OR lawsuit OR antitrust OR recall) returned no material enforcement action against 6324.T.
• Item 3 / legal proceedings: no 10-K on shelf (foreign filer); Japanese-GAAP securities reports not ingested — not sourced.
• Conclusion: No material regulatory or legal findings identified — verified via SEC EDGAR EFTS (no CIK, 0 findings) and targeted web search as of 2026-07-06. Japanese statutory-filing legal-proceedings disclosure remains unread and is an open verification item.

---

Phase D — Project & stress-test

Lens 11 · Forward Projection (EPS, next 3 fiscal years)

Built bottom-up from the FY3/26 actual + company guidance. Share count ~94.6M. Current price ¥8,840 (2026-07-05). No forecast.ts create in this unattended watchlist run — logged here as an estimate only.

Anchors: FY3/26 actual net income ¥1,608M → EPS ≈ ¥17. FY3/27 company guidance net income ¥4,500M → EPS ≈ ¥48.

Implied P/E at ¥8,840: Base FY3/27 ~186x, FY3/28 ~143x, FY3/29 ~110x. Even on the bull path (FY3/29 ¥130) the stock is ~68x three years out. The math is unambiguous: the price already discounts years of flawless execution and a humanoid ramp that is not yet in the numbers. Reported forward P/E of ~163x and P/S ~10.4x corroborate the extremity.

Company's own multi-year frame (FY2030 plan): revenue ~$668.6M (≈¥100B), 15.0% operating margin, 10% ROE/ROIC. If hit, FY3/30 net income ≈ ¥7–8B → EPS ≈ ¥75–85 → a ~110x P/E on the 2030 target, discounted to today — still rich. The plan is the bull's foundation; even the bull's own base leaves little margin of safety at ¥8,840.

Brier-scoreable base call (not logged this run): "6324.T FY3/27 (year ending 2027-03-31) net income ≥ ¥4,500M (i.e. meets guidance), p ≈ 0.55." Rationale for only-modest confidence: guidance is credible on the recovery + operating leverage, but +180% net growth off a trough with a promotional management leaves real slip risk.

Lens 12 · Bull vs Bear

Bull case (narrative). Harmonic Drive is the toll-booth on physical AI. Every humanoid needs ~20–40 precision joints; the compact, high-torque, zero-backlash reducer is the hardest of those to make well, and HDS invented it and still guards the process. As the robot + semiconductor cycle recovers (Japan/Asia already +22.5%) the pre-built capacity flips from margin drag to operating-leverage engine — hence the +141.5% OP guide. Layer on humanoids: orders went from zero to a named ¥2.5B run-rate and management says they "could double or triple" next year on U.S. prototype demand. The FY2030 plan (¥100B rev, 15% margin) reframes a sleepy Japanese component maker as core physical-AI infrastructure with a fortress balance sheet and a dividend they just doubled. If humanoids become a mass market this decade, HDS is one of the very few listed, profitable, monopoly-adjacent ways to own the picks and shovels — and picks-and-shovels is where the durable money in a gold rush is made.

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

1. China is the canary, and it already died. Domestic harmonic share in Chinese robots went <5%→>35% in six years; HDS China revenue fell −28%. Leaderdrive is at price 40–60% of HDS and torque-density life already 70–85% of the benchmark — and closing. If that commoditisation curve walks up-market into Western humanoids (where BOM cost is the gating constraint on the sub-$50K robot, per the KB bottleneck map), HDS's premium moat compresses structurally, not cyclically. Humanoid builders are ruthless cost-down engineers; several (Tesla, Figure) are exploring in-house actuators.
2. The multiple is the risk. At ~186x FY3/27-guided / ~10x sales / 2% ROE, the stock has priced the 2030 plan and then some. Any humanoid-order slip, cycle wobble, or a single "in-house actuator" headline from a flagship builder de-rates a story stock hard — the ¥2,316→¥8,870 range shows how violently this name moves.
3. Humanoid revenue is still a rounding error. ¥2.5B of orders on ¥59.6B revenue is ~4%, and orders ≠ shipments. The gap between prototype orders and volume production is where robotics timelines habitually slip 2–4 years. The market is paying volume-ramp multiples for prototype-stage reality.

Pre-mortem (it's Jan 2028, thesis broke — what happened?). Humanoid volume production slipped (prototypes didn't convert), the robot/semi cycle rolled over in 2027, and a marquee U.S. builder announced an in-house or Chinese-sourced actuator. FY3/27 guidance was missed. The stock re-rated from ~186x toward a still-generous ~40x on lower forward numbers — a 60–70% drawdown — with the balance sheet intact but the narrative broken.

Are multiples too high? Yes, on any near-term-earnings basis — unambiguously. They are defensible only on a multi-year humanoid-TAM call, and even then thinly.

Contrarian view (what the market refuses to see). The bull consensus is "monopoly on humanoid joints." The thing the tape is discounting is that HDS's own China numbers are the leading indicator of what happens to that monopoly when a good-enough Chinese substitute meets a cost-obsessed customer. The market treats China as an idiosyncratic loss and humanoids as pristine upside; the contrarian reading is that they are the same story running one geography ahead.

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

Dismantling the bull case.

• What structurally breaks the model: the customer's incentive is to design HDS out. Reducers are a cost line in a BOM the customer is desperate to shrink. Every quarter of Chinese torque-density improvement + every Western humanoid maker's in-house actuator team is an existential nibble. This is not a moat that widens with scale — it is one that a determined, subsidised competitor narrows.
• Revenue concentration / what shifts: concentrated in the cyclical robot + semiconductor-equipment complex; the "diversification" into humanoids is concentration into a handful of unproven North American prototype programs. If two of those pause, the growth story stalls.
• Why the moat is weaker than bulls think: Leaderdrive at 70–85% of benchmark life and ~half the price is already good enough for most industrial duty cycles — proven by the China share shift. "Trade-secret flexspline" is a real edge at the extreme high end, but the volume humanoid market may not need the extreme high end.
• Most dangerous underestimated competitor: Leaderdrive (up-market migration) and in-house actuator programs at Tesla/Figure (vertical integration removes HDS from the BOM entirely). Bulls treat both as contained; the China data says up-market migration is the base rate.
• Worst capital-allocation / governance items: near-zero insider ownership (CEO 0.014%), an over-capitalised balance sheet earning 2% ROE, and a management that has turned promotional into a re-rating (the "double or triple" order language). A skeptic reads the dividend-doubling as feeding a story, not fixing the return profile.
• Assumptions that must hold for ¥8,840: FY3/27 guidance met (+180% net), humanoid orders keep compounding, China stops being extrapolated, the multiple stays >100x for years. Break any one and the stock is worth a third less.
• Valuation if growth disappoints 20–30%: on a 25% haircut to the base FY3/29 EPS (~¥60), a de-rating to a still-premium 40x → ~¥2,400 — i.e. back to the 52-week low. The downside is the entire humanoid premium.
• Single permanent-impairment scenario & plausibility: a flagship Western humanoid platform standardises on an in-house or Chinese actuator and publicly says so. Plausibility: medium over 2–3 years given both are actively pursued — and it would reprice HDS from "physical-AI toll-booth" to "cyclical Japanese gear maker with a China problem."

Lens 14 · Management Questions (ordered by information value)

1. Your China revenue fell 28% while Leaderdrive scaled — is that price, qualification loss, or demand, and what is your realistic defensible share of the volume humanoid reducer market (vs the premium tier) at scale?
2. Which humanoid programs are behind the ¥2.5B of orders, how many are U.S. prototype vs committed-volume, and what is your assumed order-to-shipment conversion timeline?
3. What torque-density and cost-per-joint roadmap gets you to a BOM a sub-$50K humanoid can absorb — and how do you defend price when Leaderdrive is at 70–85% of your life at half the cost?
4. Two flagship builders (Tesla, Figure) are developing in-house actuators. What is your concrete moat against vertical integration, and have you lost any design-in to an in-house program?
5. FY3/27 guides operating profit +141.5%. Bridge it for me: how much is cycle/volume absorbing fixed cost vs price vs humanoid mix — and what fixed-cost base are you now carrying?
6. At 2% ROE and a net-cash, 72% equity-ratio balance sheet, why is this the right capital structure? What would change your mind on a larger buyback or leverage?
7. The FY2030 plan is ¥100B revenue / 15% margin / 10% ROE. What are the two or three assumptions it is most sensitive to, and what humanoid unit volume does it embed?
8. Gross margin is ~30% vs Fanuc's ~37%. Is that a structural component-vs-platform gap, or is there a path to close it as Mechatronics/Total Motion Control scales?
9. Mechatronics fell 0.9% while gears grew 9.5%. Is the move up the value chain working, and what does "Total Motion Control" contribute to revenue and margin by FY2030?
10. How much incremental capacity (Beverly +33%, Japan humanoid lines) is committed vs optional, and what utilisation/volume do you need to make today's depreciation earn its keep?
11. What is the current split of humanoid orders by region and by customer concentration — and what is your single-customer exposure limit?
12. How replicable is the flexspline process really? What would it take a well-funded competitor 5+ years and how do you know they haven't closed the gap in duty cycles that matter for humanoids?
13. What is the domestic (Japan) shareholder and cross-holding structure, and what is aggregate insider/founder-family ownership today?
14. Yen strength would compress your translated overseas earnings — how hedged are you, and how much of the FY3/27 guide assumes the current rate?
15. If the robot + semiconductor capex cycle rolls over in 2027 before humanoid volume arrives, what is your margin floor and how do you protect the FY2030 trajectory?

---