Harmonic Drive Systems: A Zen Horizon Framework Deep Dive

Questions readers submitted on this report and that have been answered.

• QWhich company produces planetary roller screws?

  A

  First, the key distinction: planetary roller screws are not a product made by Harmonic Drive Systems. They are a class of precision linear transmission components that handle heavier loads and offer higher precision than standard ball screws. They belong to a different field from Harmonic Drive Systems, the subject of this report, which makes harmonic reducers for rotary joints, and they are produced by another group of specialized suppliers. The report makes this point repeatedly: among the roughly 28 actuators across Tesla Optimus, about 14 rotary joints use harmonic reducers, which is Harmonic Drive Systems' territory, while about 14 high-torque linear actuators in the lower limbs use planetary roller screws, with no harmonic reducers involved.

  High-end global supply is highly concentrated among established European manufacturers:

  • Switzerland's Rollvis and Switzerland's GSA: Rollvis offers a full RV/RVR/RVI/RVD lineup of ground, high-precision roller screws. It has been part of the GSA system since 2016, and the two brands together have more than 50% market share;
  • Sweden's Ewellix: formerly SKF's linear motion business, acquired by Germany's Schaeffler in 2022, with completion in early 2023;
  • Other suppliers include Sweden's SKF, U.S.-based Moog (which says it has more than 30 years of planetary roller screw design experience), and the U.K.'s Power Jacks (heavy-load and niche automotive applications). On industry research estimates, leading suppliers such as GSA, Rollvis, and Ewellix together account for more than 70% of the market.

  Chinese suppliers still have a small share, but are entering through the humanoid robotics wave. According to Qianzhan Industry Research Institute's 2023 figures, domestic Chinese manufacturers together account for about 19%, represented by Nanjing Technical Equipment Manufacture and Bote Precision (each with roughly 8% share). Targeting Tesla Optimus demand for linear actuators, the domestic supply chain names most often include Wuzhou Xinchun (inverted planetary roller screws, aimed at high-torque joints such as hips and knees, with a Thailand plant planned to start production in 2026), Hengli Hydraulic (migrating from hydraulic screw technology and planning capacity for planetary roller screw electric cylinders), and Zhenyu Technology. Traditional linear motion leaders such as Japan's THK and Taiwan's HIWIN (listed in the report's peer comparison table under "linear guides/screws") mainly focus on ball screws, but are also extending into planetary roller screws.

  This directly relates to the investment logic of this report. Readers asking this are probably trying to clarify how much Harmonic Drive Systems can capture from the humanoid robotics boom. The answer is one of the report's core arguments: harmonic reducers cover only Optimus's upper-body rotary joints, and content per unit is structurally capped (about 7% of total BOM). The more valuable lower-limb linear actuators use planetary roller screws, with profits going to the screw suppliers above and little direct relevance to Harmonic Drive Systems. So even if humanoid robots scale, Harmonic Drive Systems' slice has a ceiling. This is also one reason the report assigns a "Watch (bearish end)" view and stresses that the humanoid pie can grow without necessarily making Harmonic Drive Systems' own slice grow with it.

  (Note: the market share figures above are third-party industry research estimates and may differ by source and statistical year; the companies listed are for supply-chain mapping, not investment recommendations.)

  Jun 9, 2026

Hunting ten-year five-baggers among great growth stocks — pressing the upside question: "Can it get much bigger?"

• QHow high is its market ceiling? Is it expanding an existing pie, or creating an entirely new market?5/10

  A

  The ceiling is "directionally high, but the share is likely to leak away": Harmonic Drive Systems is expanding an existing pie, harmonic reducers, rather than creating a new market from scratch. The truly new incremental market, humanoid robots, is opening up, but within that new pie it will probably capture only a smaller corner with compressed gross margin.

  Start with where it stands today: Harmonic Drive Systems is the clear global leader in harmonic reducers, with about 58% share by value and more than 70% in the high-end market. It is almost synonymous with the category, "HarmonicDrive® = harmonic drive." But the business itself is a mature, highly cyclical one. FY3/2026 revenue was ¥59.56 billion (+7.0%), and historical revenue has swung sharply between ¥37 billion and ¥71.5 billion, almost a 2x range. In other words, the traditional base of industrial robots plus semiconductor equipment is a mature, cyclical, modest-growth stock business, not one with a very high ceiling.

  Only one thing can support the "high ceiling" narrative: humanoid robots. This is indeed a potentially new market. Goldman Sachs estimates the humanoid robot market at about $38B in 2035, and harmonic reducers are core components for rotary joints, making Harmonic Drive Systems a classic "picks-and-shovels" supplier. The company has already invested about ¥10 billion specifically in humanoid reducer capacity at the Ariake plant, targeting FY3/2027 humanoid reducer sales of ¥10-20 billion. Directionally, this follows a "create a new market" logic by expanding both usage per robot and the robot installed base.

  But the ceiling needs two discounts. This is exactly why the essence is "expanding an existing pie" rather than "owning a new market":

  • Content per unit is structurally capped. Of Tesla Optimus's 28 actuators, only 14 rotary actuators use harmonic reducers. The high-torque lower-limb joints use planetary roller screws because harmonic reducers lack sufficient life. Harmonic content is only about 7% of total BOM, so this is not an "all-body joints" opportunity. Its layer in the new pie has a physical upper bound.
  • The largest incremental piece is likely to go to Chinese suppliers. The main incremental harmonic reducer share for Optimus has already gone to China's Leaderdrive (688017) (Mexico plant, prices 30-45% below Japanese suppliers, target share about 60%). The pie is getting larger, but Harmonic Drive Systems' piece is getting smaller, and defending share also requires price cuts.

  Conclusion: the absolute ceiling is not low if the humanoid story plays out, but Harmonic Drive Systems gets "an expansion of an existing pie plus a capped and diverted small corner of the new pie," rather than exclusively creating and leading a new market. This is the root reason the rating stays at "Watch, bearish bias": the story is large, but the part this company can actually pocket is far smaller than what the current price implies, with roughly 70% of market value representing the humanoid option.

  Jun 10, 2026
• QCan its revenue at least double over the next five years? Will growth mainly be driven by volume, price, or new businesses?4/10

  A

  A five-year revenue doubling is "what the company is betting toward, but not a high-certainty case": moving from FY3/2026 revenue of ¥59.56 billion to ¥120 billion requires both engines, a traditional cycle recovery and humanoid robot volume, to ignite at the same time. The driver would be volume (robot and semiconductor equipment shipments), not price. Price is a headwind because Chinese competitors are forcing prices down.

  Start with the baseline and the company's own framing. Latest FY3/2026 revenue was ¥59.56 billion (+7.0%). The company's FY3/2027 guidance is revenue of ¥68 billion (+14.2%) and operating profit of ¥6.2 billion (+141.5%). Further out, its medium-term plan targets FY2030 revenue of at least ¥100 billion. Note that even the company's own five-year target is only "approaching a doubling" (from about ¥59.5 billion to ¥100 billion, about +68%). To truly double within five years to ¥120 billion, it would have to exceed its own medium-term plan by a clear margin.

  The three sources of growth break down as follows:

  • Volume (traditional-base recovery): higher certainty, but only a cyclical return. Current domestic plant utilization is only about 50%. Capacity is already in place, so recovery leverage is large. Externally, SEMI forecasts record global semiconductor equipment sales in 2026 and further growth in 2027, which is cyclical support for Harmonic Drive Systems' semiconductor line. But this is a recovery from trough to mid-cycle. It can contribute high-single-digit to low-teens growth, but cannot support a doubling by itself.
  • Volume (new humanoid business): the largest upside, but with uncertainty in both share and timing. The company targets FY3/2027 humanoid reducer sales of ¥10-20 billion (about 15-30% of total revenue), the key increment in the doubling narrative. Reality is more modest: FY3/2026 actual growth came almost entirely from traditional lines; current AI robot orders, including humanoids, are only about ¥2.5 billion on a combined basis (about 4% of revenue), and Optimus mass-production timing has been repeatedly delayed, with high-volume ramp pushed to summer 2027.
  • Price: not a tailwind, a headwind. Harmonic Drive Systems' selling price is about 2-2.5x Chinese products (Chinese suppliers are only 40-60% of its price), leaving humanoid mass production facing a dilemma between "cut prices and destroy margin" and "hold price and lose share." Gross margin has fallen from about 39% in FY3/2022 to 30.5% in FY3/2026. Price increases are not the growth engine; the price side is bleeding.

  Conclusion: a five-year doubling is possible, but it requires both engines to run at full load: a real recovery in the traditional cycle and humanoid reducer revenue rising from about 4% of revenue to a double-digit percentage. The ignition timing and ultimate share of the humanoid engine have not been proven by public data. Orders are a positive signal, with FY3/2026 orders at a high ¥61.6 billion, +16.2%, and book-to-bill above 1, but that speaks to the next year or two, not a five-year doubling. Treating a doubling as the base case and pricing the stock accordingly, with about 70% of market value as the humanoid option, is the biggest expectation mismatch today. The growth direction is right, but the "at least double" threshold is an optimistic case, not a neutral case.

  Jun 10, 2026
• QFive years from now, what will take over as the next growth engine? Does this "second curve" exist today?4/10

  A

  The second curve is "humanoid robot reducers today, with real money already committed, but still far from realized": Harmonic Drive Systems' first curve, harmonic reducers for industrial robots and semiconductor equipment, is a mature and highly cyclical base. The only clear successor is reducers dedicated to humanoid robots. It does exist today, with capacity, orders, and customers in motion, but its contribution remains tiny (about 4% of revenue), more like an option than an already formed second growth pillar.

  First, the state of the first curve: the traditional base swings sharply with the peaks and troughs of factory automation and semiconductor capex. FY3/2026 operating margin was only 4.3%, at the early stage of recovery from a cyclical trough. This curve will not be disrupted overnight, but growth is modest and it will continue to be eroded by Chinese suppliers. So the successor engine the market wants has to come from somewhere else.

  There is only one clear candidate, and it has already moved from paper to the shop floor:

  • Real capital has already been committed: the company has invested about ¥10 billion specifically to build mass-production capacity for humanoid reducers at the Ariake plant, targeting FY3/2027 humanoid reducer sales of ¥10-20 billion (about 15-30% of total revenue). This shows the second curve is not a PPT story; capex and production lines are in place.
  • The historical DNA fits: harmonic reducers were originally designed for aerospace and were used in Honda's ASIMO biped robot. Humanoids are a natural extension of that lineage, not a cross-border gamble. This lowers the technical risk of a failed transformation.
  • The TAM narrative is large: Goldman Sachs estimates a humanoid market of about $38B in 2035. If it scales, this curve could theoretically become larger than the entire company today.

  But "exists" does not mean "has taken over." Three hard constraints keep it in the option stage:

  • The share is still tiny: current AI robot orders, including humanoids, are only about ¥2.5 billion on a combined basis (about 4% of revenue), and FY3/2026 actual growth came almost entirely from traditional lines, with no visible humanoid volume yet.
  • Share may go elsewhere: the main incremental harmonic reducer volume from the largest humanoid buyer, Tesla Optimus, has already gone to Leaderdrive (Mexico plant, prices 30-45% lower, target share about 60%). Harmonic Drive Systems may remain as a high-end dual source, but the pattern is "the pie grows while its own slice shrinks."
  • Timing keeps slipping, and content is capped: the Optimus high-volume ramp has been pushed to summer 2027, and 2026 industry shipment forecasts differ by an order of magnitude, "50,000 vs 900,000 units." Harmonic reducers are only used in upper-body rotary joints (about 14 per unit, about 7% of BOM), so content per unit is structurally capped.

  Conclusion: the second curve does exist today, and the company has put about ¥10 billion of real money behind it. It also fits its technical lineage. That makes it stronger than a pure story stock. But it remains an option with uncertain timing, uncertain share, and only about 4% revenue contribution, far from taking over as the main engine. The current price already values this unrealized curve as if it had succeeded, with roughly 70% of market value as the humanoid option. That is the risk.

  Jun 10, 2026
• QWhat is its core competitive advantage? Will this moat widen or narrow over the next three to five years?5/10

  A

  The moat is "real, but erosive (3/5)," and over the next three to five years it is more likely to narrow slowly than widen: the core advantage is decades of accumulated materials and manufacturing know-how plus customer qualification that makes it nearly sole-sourced at the high end. That part is wide and durable. But the foundational patents expired long ago, there is no IP gate, and Chinese suppliers are eroding the mass-production end with half-price products and rapid technical convergence. Overall, the wide end can hold, while the mass-production battlefield is narrowing.

  First, what exactly is the advantage? Harmonic Drive Systems' moat is not one wall, but several layers:

  • Materials and manufacturing know-how (strong and hard to replicate): special steel and ductile iron supplied by Japan Steel Works, flexspline fatigue life (fatigue fracture is the main failure mode for harmonic drives), gear-grinding precision, batch consistency, and miniaturization (Harmonic Drive Systems can make size 3 units, while Leaderdrive's smallest is size 5). This reflects decades of process accumulation. Chinese suppliers have caught up on strength indicators, but still lag in precision, life, and miniaturization.
  • High-end customer qualification and brand (strong, with very high switching costs): aerospace, semiconductor lithography, medical, and high-end collaborative robots are nearly sole-sourced, failure-is-not-an-option fields with extremely high switching costs. HarmonicDrive® is almost a category synonym.
  • Pricing power (present, but being eroded by the price gap): Chinese products sell for only 40-60% of its price, yet the company still explicitly stated in FY3/2026 that it would "improve profitability through product price adjustments," showing real pricing power remains.

  Why judge it as "narrowing" rather than "widening"? The evidence is hard:

  • No IP gate: the foundational patent (Musser 1959, US 2,906,143) expired long ago. This is the fundamental reason Leaderdrive and other global competitors can legally manufacture these products. Basic IP is no longer a moat.
  • Share is being lost quickly: Chinese domestic harmonic reducer share rose from about 48.7% in 2020 to about 75.1% in 2024, while Harmonic Drive Systems' China share has fallen to about 35.6% (2024H1, still declining). It remains the leader globally by value, about 58%, and in high-end markets above 70%, but volume and China-market share are being eroded.
  • Pricing power is bleeding through gross margin: gross margin fell from about 39% in FY3/2022 to 30.5% in FY3/2026, while operating margin, once 22-25% at the peak, collapsed to nearly 0% at one point (FY3/2025). A truly wide moat should not be this cyclical or this pressured by price gaps.
  • Qualification barriers in humanoid mass production are being deliberately diluted: OEMs are actively fostering domestic second sources to cut costs, rapidly diluting qualification barriers in general industry and humanoid mass production.

  A fair point needs to be acknowledged: complementary rivals are not the threat. Nabtesco (about 60% global share in RV reducers) serves heavy-load bases and upper arms (>20kg), while harmonic reducers serve forearms, wrists, and <20kg joints. They cover different joints and are complementary rather than substitutes. The moat threat mainly comes from Chinese competitors in the same lane, not from RV reducers.

  Conclusion: this is an erosive moat, "wide at the high end, narrowing at the mass-production end" (the report scores it 3/5). The sustainable part is high-precision, high-reliability, miniaturized niches such as aerospace, semiconductors, and medical, which can hold over the next three to five years and may even widen slightly with high-end humanoid demand. But the main battlefield of general industry and humanoid mass production is structurally eroding. Net-net, the moat is more likely to narrow over the next three to five years. That is why it deserves close tracking, but not the "wide-moat perpetual leader" valuation implied by the current price.

  Jun 10, 2026
• QIf its core business is disrupted, does it have the DNA to reinvent itself? How does it handle mistakes and bad news?5/10

  A

  The reinvention DNA is "present to a degree, but conservative and steady, not an aggressive paradigm-reinventing company": Harmonic Drive Systems has gone through multiple structural adjustments, including technology introduction and localization, overseas acquisitions, and listing upgrades, showing it can endure cycles and make strategic moves. Its handling of bad news is also reasonably honest: in 2024 it decisively took a large impairment on the German subsidiary in one shot and cleaned up the balance sheet. But its DNA is that of a steady, incremental precision-manufacturing incumbent, not a disruptive self-reinventor; facing a paradigm-level variable like humanoids, it reacts cautiously.

  First, is there historical evidence of "reinvention"? Yes, but all of it is incremental rather than disruptive:

  • Technology introduction → localization → reverse export: in 1970, it introduced strain wave gear technology from U.S.-based USM and localized it in Japan, then over half a century built the global leader and made HarmonicDrive® a category synonym. This was a successful case of digesting, absorbing, and innovating on imported technology.
  • Cross-border integration capability: it began cooperation with a German company in 1996, increased ownership and consolidated it in 2017, and made it a wholly owned subsidiary in 2021 (100% ownership of both Germany SE and U.S. LLC), showing organizational ability in cross-border M&A and integration.
  • Capital-market progression: OTC registration (1998) → JASDAQ → TSE Standard (2022) → TSE Prime (uplisted in 2026-02), with continued governance and disclosure normalization.

  How it handles mistakes and bad news is the most positive part of this question:

  • The FY3/2024 loss of -¥24.8 billion was not an operating collapse, but an active admission of error: it came from a one-time impairment of ¥28.16 billion at the German subsidiary (goodwill ¥15.25 billion + customer relationships ¥10.19 billion + technology assets ¥2.72 billion), acknowledging that the 2017 European acquisition plan had fallen short of expectations. The key is that it cleared the item in one shot and substantially removed goodwill and intangibles from the balance sheet, lowering future impairment risk. This was an honest, decisive handling of an error. FY3/2026 financials have already recovered on a cleaner balance sheet (equity ratio 72.2%, net cash about ¥3.6 billion).
  • Maintaining dividends countercyclically: DPS has been held at ¥20 for years, including the large-loss year, supported by net cash and positive operating cash flow, showing a stable commitment to shareholders.

  But the "reinvention DNA" deserves real discounts:

  • It is a single-business company: precision reducers and applications, with almost 100% of revenue from reducers. Focus is a strength, but it also means little room to start a new stove if the core is disrupted, and no diversification safety net.
  • Management is cautious toward the new paradigm: current president Akira Maruyama is a professional manager who joined in 1985, and has publicly been cautious about humanoid robot investment. In a structure where the main incremental Optimus harmonic reducer volume has been taken by Chinese suppliers and Harmonic Drive Systems is only defending a high-end dual-source role, this caution is both discipline and a possible lag that misses the largest upside.
  • The real disruption threat comes from process substitution: if lower-limb roller screws, cycloidal or planetary solutions for rotary joints further replace harmonic reducers (Agility Digit already uses series elastic actuators plus cable drives to bypass harmonic reducers), its response is still to "make better harmonic reducers," not to step outside harmonics and redefine itself.

  Conclusion: Harmonic Drive Systems has the foundations of cycle resilience, honest error recognition, and cross-border integration. The 2024 impairment cleanup is a mature and candid example of handling bad news, making it far better than cash-burning or poor-quality businesses. But its genes are those of a steady, incremental precision-manufacturing incumbent, not an aggressive self-reinventor. If the core harmonic business is truly displaced by another mechanism or cleared out by Chinese competitors, there is not strong evidence that it can reinvent itself quickly. This is another reason for "Watch" rather than "Buy."

  Jun 10, 2026
• QDoes management, especially the founder, have a long-term view and deeply aligned interests with the company? Is it willing to sacrifice current profit for five to ten years out?6/10

  A

  Management is "family-controlled in substance for the long term, with moderately strong alignment, but led by a professional manager and cautious toward the biggest growth opportunity": the Ito family effectively controls about 35% through parent company KODEN, providing a stable anchor with a decades-long horizon. But the current president is a professional manager rather than a founder, personal ownership is not material, and he has publicly been cautious about humanoid robot investment. Long-termism and alignment exist, but the willingness to sacrifice current profit for a five-to-ten-year new engine is relatively restrained.

  First, the axis Baillie Gifford cares most about, "long-term view plus alignment," has layered evidence:

  • Long-term family control (strong): the controlling parent is the Ito family's KODEN Holdings, with a stake of about 34.8-35.3%, originating from the 1989 integration into KODEN. The second-largest individual shareholder, Norimitsu Ito, owns about 3.16%, bringing the family group to about 38%. Effective family control through the parent company (but <50%) is a typical Japanese ownership structure with a long-term, stable, cross-generational horizon. This is not short-term financial capital; the bet is on the company's longevity.
  • Relatively standardized governance (neutral to positive): there are no dual-class shares or super-voting rights (control is not locked through super-votes), 5 of 10 directors are outside independent directors, and foreign institutions such as Norway's sovereign wealth fund, UBS, Morgan Stanley, and Goldman Sachs hold meaningful stakes, though each is below 3%. This is much cleaner than companies controlled through Class B super-voting rights and related-party transactions.

  But the discounts to alignment and aggressiveness should be stated honestly:

  • The operator is a professional manager, not a founder: current president Akira Maruyama joined in 1985 and became president in 2024-06. His personal economic exposure is not as deeply tied to the company as a founder's would be. The true owners, the Ito family, control at the parent-company level but do not directly run the front line. In the Baillie Gifford model, the tension of "deep founder alignment plus long-term vision" is replaced here by "stable family ownership," not "founder-spirit-driven management."
  • Cautious toward the biggest growth opportunity: the president has publicly been cautious about humanoid robot investment. On one hand, that is rare capital discipline and avoids blindly burning money for a theme. On the other, in a structure where the main incremental Optimus harmonic reducer volume has been captured by Leaderdrive at half the price and Harmonic Drive Systems is only defending a high-end dual-source role, this caution may mean it is not betting hard enough on the biggest prize and is moving slowly.

  The core Baillie sub-question, "is it willing to sacrifice current profit for five to ten years out," has evidence on both sides:

  • The willing side: it has already invested about ¥10 billion specifically to build humanoid reducer capacity at the Ariake plant, targeting FY3/2027 humanoid reducer sales of ¥10-20 billion. It also countercyclically held DPS at ¥20 in trough years, supported by net cash and positive cash flow. Neither action is purely focused on current-period profit. The medium-term plan targets FY2030 revenue of at least ¥100 billion and operating margin of at least 15%, also a clear long-term target.
  • The restrained side: because FY3/2026 trough earnings were low, dividend per share of ¥20 implied a payout ratio as high as 117.7% (excess dividend), returning cash to shareholders rather than investing more aggressively to expand humanoid capacity and seize share. Compared with a stance of sacrificing today and investing heavily for ten years out to capture a new market, this is conservative.

  Conclusion: management is a steady combination of "family effective control for the long term (about 35%) + clean governance + long-term targets and real investment." Long-term perspective and alignment are broadly adequate, clearly stronger than short-term or poorly controlled companies. But the operator is a professional manager, not a founder with deeply tied economic exposure, and the team is cautious toward the biggest growth opportunity, humanoids. This is a management team that can defend steadily, but may not attack hard enough. Against Baillie Gifford's high bar of founder-level aggressiveness and willingness to sacrifice current profit for five to ten years, it is "qualified but not outstanding," leaving a gap versus the aggressive growth expectations embedded in the current price.

  Jun 10, 2026
• QIf it disappeared tomorrow, how much would customers miss it? Is its growth model sustainable and not dependent on harming society or regulation?5/10

  A

  Indispensability is "nearly irreplaceable at the high end, replaceable in general mass production," and sustainability is "healthy, not dependent on social harm or regulatory arbitrage": if Harmonic Drive Systems disappeared tomorrow, customers in aerospace, semiconductor lithography, medical, and high-end collaborative robots, all failure-is-not-an-option fields, would miss it badly, with almost no equivalent substitute in the short term. But customers in general industry and humanoid mass production already have half-price Chinese second sources, so the pain would be much lower. Its growth model is clean, based on real industrial demand, and does not cross social or regulatory red lines.

  First, how much would customers miss it? It varies sharply by scenario:

  • High end: they would miss it intensely, with almost no substitute. Aerospace (lightweight and high-precision space mechanisms), semiconductor lithography, medical, and high-end collaborative robots are nearly sole-sourced, with extremely high switching costs. These are fields where failure is unacceptable. Customers in these segments cannot find an equivalent substitute in the short term. Harmonic Drive Systems can make size 3 miniaturized units, and flexspline fatigue life and gear-grinding precision reflect decades of know-how. Chinese suppliers still lag in precision, life, and miniaturization. At this layer, it truly is indispensable.
  • General industry plus humanoid mass production: customers would miss it, but alternatives exist and the pain is limited. At this layer, OEMs are actively fostering domestic second sources to cut costs. Leaderdrive and others supply at 40-60% of its price, and strength indicators have caught up. If Harmonic Drive Systems disappeared, these customers could switch relatively quickly to Chinese suppliers, with far less pain than high-end customers. China's domestic localization rate reached about 75.1% in 2024, itself showing that a large amount of demand no longer depends on it.

  Another hard data point supports the view that indispensability is being diluted: its pricing power is bleeding. Gross margin fell from about 39% in FY3/2022 to 30.5% in FY3/2026. A truly indispensable component should not be pressured this heavily by half-price competitors. So indispensability is real at the high end and weak at the mass-production end, with the overall direction slowly diluting.

  On whether growth is sustainable and not harmful to society or regulation, this is clean and positive:

  • Demand is real industrial demand: it comes from factory automation, semiconductor equipment capex, and robotics industry upgrades, not regulatory arbitrage, subsidy dependence, or a model that harms users. Even if the largest growth dividend, humanoids, materializes, it serves physical manufacturing and the robotics industry, with positive social value.
  • Governance and compliance are clean: the company itself has no major litigation, no financial deficiencies, and no governance scandal. The only visible litigation is a personal lawsuit involving one audit and supervisory board member because of his former role as an outside director at JDI. It relates to JDI and not Harmonic Drive Systems' operations, and should not be exaggerated as a company risk.
  • Regulatory tail risk is mostly external: the main external variables are U.S. tariff policy and rare-earth controls, which affect downstream humanoid mass-production timing. These are industry-level macro risks, not sustainability issues in its own business model.

  Conclusion: Harmonic Drive Systems is an indispensable supplier at the high end, where customers would miss it intensely if it disappeared tomorrow, and that part of the moat is real. But general industry and humanoid mass production have half-price Chinese alternatives, so overall indispensability is being diluted, as confirmed by gross margin falling to 30.5%. Its growth model is healthy, sustainable, and not dependent on social or regulatory harm, so it passes this dimension cleanly. Overall, it is a "good company" socially, but "indispensable" only holds at the high end and is not enough to support the full-market perpetual premium implied by the current price.

  Jun 10, 2026
• QHow are the unit economics of this business (gross margin, incremental return)? Do they improve or worsen with scale? Where does the money it earns go?5/10

  A

  Unit economics are "historically structurally strong, currently pushed to a trough by the cycle and price war, and scale has not brought clear improvement": in good years gross margin reached 45-47% and operating margin 24.5%, making this a structurally high-gross-margin precision manufacturing business. But gross margin has been compressed by the Chinese price gap to 30.5%, operating margin is only 4.3%, and even at cyclical peaks ROE/ROIC were only about 7-11%/7%. Capital and cash are too heavy, so getting larger has not improved returns. The money earned mainly goes to capex, including about ¥10 billion of humanoid capacity, and countercyclical dividends.

  Start with the ceiling of unit economics. History proves this business is naturally high gross margin:

  • Peak gross margin of 45-47% (FY3/2017-2019), with the recent cyclical peak still at 36-39%, typical of high-gross-margin precision components.
  • Operating margin as high as 24.5% in good years (FY3/2019), showing very high profit leverage.

  But the truth today and after scaling is under dual pressure:

  • Gross margin has structurally shifted down because of price gaps: from about 39% in FY3/2022 to 30.5% in FY3/2026. Part of this comes from consolidating Germany's HD SE, which lowers group margin, and part comes from Chinese suppliers pressing at 40-60% of its price. This is not purely cyclical; it includes a structural component.
  • Operating margin has collapsed to the trough: only 4.3% in FY3/2026, and once close to 0% in FY3/2025. Profitability swings sharply between about 0% and about 25%, showing the moat is not strong enough to smooth cycles.
  • Incremental return (ROIC) is mediocre: this is the most damaging point for unit economics. Even at cyclical peaks, ROE was only about 7-11% and ROIC about 7%; current ROIC is only about 2.4%. High gross margin has not translated into high returns on capital because capital and cash are too heavy (tangible fixed assets mainly in Japan of about ¥30 billion, capital-intensive, and cash on hand of ¥21.56 billion). Scale is actually dragged down by heavy assets and heavy cash, and returns have not improved.

  A horizontal comparison makes the gap clearer: Nabtesco currently has operating margin of 7.5% and ROIC of 6.1%, both better than Harmonic Drive Systems' 4.3%/2.4%. Among Japanese transmission leaders, Harmonic Drive Systems' unit-economic quality is currently weaker.

  Where does the money it earns go? Two lines, disciplined enough but not highly compounding:

  • Mostly capex: it has invested about ¥10 billion specifically to build humanoid reducer capacity at the Ariake plant, betting on future growth. This is an offensive investment for the second curve.
  • Countercyclical dividends: DPS has stayed at ¥20 for years, including the large-loss year. The FY3/2026 payout ratio was as high as 117.7% (excess dividend), supported by net cash and positive cash flow. Maintaining dividends at the trough shows cash-cow attributes, but with ROIC only in the single digits, returning cash to shareholders also reflects limited high-return reinvestment opportunities.

  One positive anchor must be noted: cash flow quality is solid. Operating cash flow has been positive for the past 10 years (¥6.4-15.1 billion). Even in FY3/2024, when it reported a -¥24.8 billion loss, operating cash flow was actually the 10-year high at +¥12.7 billion because the loss was a non-cash impairment. This makes it a cash cow that can survive a cyclical winter, not a bleeding company. That is far better than pure cash-burning names.

  Conclusion: unit economics are "good gross-margin foundations (historically 45%+), hard cash flow, but mediocre incremental returns (ROIC about 7% at peak, about 2.4% currently), with scale dragged by heavy assets and heavy cash and not improving returns." Current gross margin of 30.5% and operating margin of 4.3% are at the trough, pressured by both the cycle and the price war. Capital allocation is disciplined (humanoid expansion plus stable dividends), but low ROIC makes it hard to call this efficient compounding. Overall, cash-cow quality is adequate, but it lacks the unit-economic signature of a high-quality growth stock where larger scale means better profitability. That is the core financial basis for not scoring it highly.

  Jun 10, 2026
• QWhat conditions must all be true for it to rise fivefold over ten years? Are those conditions realistic? What expectations does today's stock price imply?2/10

  A

  A ten-year fivefold return "requires four high-bar conditions to all hold, and the realistic probability is low"; today's stock price (about ¥7,090 / now down to about ¥6,200) implies "a successful humanoid supercycle plus Harmonic Drive Systems fending off half-price Chinese suppliers at the same time." It prices the most optimistic case as the base case, leaving a negative margin of safety.

  First, break down what has to happen for a fivefold return over ten years from a Baillie blue-sky perspective. A fivefold rise from the current price requires market value to move from about ¥671.2 billion to about ¥3.4 trillion, which requires the following conditions to hold simultaneously:

  1. Humanoid robots truly scale: industry shipment imagination (million-unit levels in the 2030s, with Goldman Sachs estimating a 2035 market of about $38B) must turn into real orders rather than remain an option.
  2. Harmonic Drive Systems keeps high share without crushing gross margin: in a structure where the main incremental Optimus harmonic reducer volume has already gone to Leaderdrive (Mexico plant, prices 30-45% lower, target share about 60%), it must retain a meaningful share as a high-end dual source without having margins destroyed by the price war.
  3. The traditional cycle recovers and persists: semiconductor equipment plus industrial automation must recover from trough conditions (current utilization about 50%) to mid-cycle strength and stay there, pulling operating margin back from 4.3% to double digits.
  4. Valuation does not deflate materially: as earnings materialize, the market must continue to assign a high multiple, even though the stock is already the most expensive name in the Japanese transmission chain on forward metrics.

  How realistic are these four conditions? The third, cyclical recovery, has relatively high certainty. The first, humanoid volume, is real directionally but timing keeps slipping (Optimus high-volume ramp pushed to summer 2027, while 2026 industry shipment forecasts differ by an order of magnitude, "50,000 vs 900,000 units"). The second is the biggest flaw: harmonic content per unit is structurally capped (about 14 per unit, about 7% of BOM), and the largest incremental share is already going to Chinese suppliers. The fourth requires a high valuation not to break, which conflicts with the natural multiple normalization that follows earnings delivery. All four holding at once is a low-probability combination.

  Now, what does today's price imply? This is the part to see clearly:

  • Every valuation metric is extreme: trailing PE 413-417x (distorted by trough earnings), forward PE about 116x (consensus EPS about ¥60.8) / 149x (company guidance EPS ¥47.54), PB 8.35x (about 94th percentile over 2010-2026, close to the 2020 historical top), and PS 11.27x (Japan machinery industry about 0.9x).
  • About 70% of market value is a pure humanoid option: anchoring on a reasonable bottom-cycle industrial valuation of about ¥160-200 billion, about 70% of the current ¥671.2 billion market value (about ¥470-510 billion) is humanoid robot option value.
  • To support the current price at reasonable multiples, normalized net profit must reach about ¥16.8 billion, equal to 2.5x the historical peak (¥6.6 billion), meaning the humanoid business would have to become several times larger than the entire current company.
  • The current price is above all sell-side averages: analyst consensus is Buy, but average targets range from about ¥4,419 (TipRanks view, overvalued at ¥6,100) to about ¥5,746 (stockanalysis view), with the report's core cluster around ¥5,300. The current price is above these averages. Jefferies downgraded it to Underperform in 2026-02 with a ¥2,400 target (implying about -66%), and SWS DCF fair value is only ¥2,358.

  Downside/upside is highly asymmetric: downside to the sell-side average is about -25%, and to DCF/Jefferies about -66% (note: the price has already fallen from ¥7,090 to about ¥6,200, but even so it remains close to or above most sell-side averages, leaving the conclusion unchanged); upside requires all four conditions above to hit.

  Conclusion: a ten-year fivefold return is not impossible, but it requires humanoid volume, Harmonic Drive Systems retaining share without destroying margin, sustained cyclical recovery, and no valuation deflation all at the same time. The "retain share without destroying margin" condition directly conflicts with the reality that the largest incremental share is going to Chinese suppliers and content per unit is capped, making it low probability. Today's stock price already treats "a successful humanoid supercycle" as the base case, with about 70% of market value as option value, forward 116-149x, PB near the historical top, and a price above all sell-side averages. The margin of safety is negative. This is the fundamental reason the rating stays at "Watch, bearish bias" and the ideal entry price is ≤¥3,200.

  Jun 10, 2026
• QWhy has the market not realized all of this yet? Is it because investors do not understand, underestimate it, or cannot see far enough? What will become the narrative inflection point?2/10

  A

  The market is "not failing to understand, but selectively failing to look far enough, with theme beta in control": Harmonic Drive Systems' leadership, high cyclicality, and Chinese erosion are all understandable, and sell-side research states them clearly. The issue is that this is a stock driven by humanoid robotics theme sentiment. The market is pricing it as if "the humanoid supercycle will definitely succeed," while deliberately looking past near-term concerns such as share leakage, capped content per unit, and valuation already above prior peaks. The narrative inflection point will be any disconfirming signal on "humanoid dividend realization pace" or "share structure becoming clear."

  First, what kind of "not realized" is this? Among the three choices, it is closest to "cannot see far enough + theme capture," not "does not understand" or "underestimates it":

  • Not a failure to understand: sell-side coverage is sufficient. Jefferies already downgraded it to Underperform in 2026-02 due to "Chinese competition + overvaluation," with a ¥2,400 target. SWS DCF fair value is ¥2,358. Even bullish sell-side average targets (about ¥4,419 to ¥5,746) are below the current price. In other words, professionals in the market see the overpricing clearly. The disagreement is not over the facts, but over whether to pay this price for the humanoid option.
  • Not underestimation: it rose from ¥1,763 at the end of 2024 to an intraday ¥8,400 in 2026-05 (about +376%), gained +123% over the past year, and was promoted to TSE Prime. The market did not look down on it; it chased it too hard.
  • It is "not looking far enough + theme beta": the entire Japanese transmission chain nearly doubled over a year (the two harmonic leaders +123%/+121%, THK +88%), reflecting theme beta rather than stock-specific alpha. The market has locked onto the long-lens story that humanoid TAM is large, Harmonic Drive Systems is the only global scaled pure harmonic leader, and it has the highest humanoid purity, while discounting three near-lens facts: ① the main incremental Optimus harmonic reducer volume from the largest buyer has already gone to Leaderdrive (half-price, target about 60%); ② harmonic content per unit is structurally capped (about 7% of BOM); ③ current AI robot orders are only about ¥2.5 billion (about 4% of revenue), and humanoid volume has not arrived.

  In other words, the market is not "wrong" so much as "betting." It is using an option that has not yet been realized and may lose share to justify the current valuation (about 70% of market value as the humanoid option, forward about 116x / 149x on guidance, PB around the 94th percentile). This valuation only works in the single scenario where humanoids succeed in a big way and Harmonic Drive Systems captures a large share.

  What will become the narrative inflection point (signals that make the market switch from long lens back to near lens):

  • Downward inflection (more likely to arrive first): ① another Optimus production delay or initial production far below expectations, disproving "humanoids are about to scale"; ② Leaderdrive's Suzhou/Mexico capacity coming online, triggering a price war and confirming it as the main Optimus supplier, validating "the dividend goes elsewhere and Harmonic Drive Systems is left with only a high-end corner"; ③ a second weakening in Chinese industrial/auto demand, making the traditional-base recovery fail as well. If any occurs, the market could move from 116-149x toward the industry's normalized 15-40x, the report's pre-mortem "Davis triple kill." The fact that the price has already fallen from ¥7,090 to about ¥6,200 (roughly -12% within several days) may be an early signal that this near-lens risk is starting to be repriced.
  • Upward inflection (high realization threshold): ① FY3/2027 guidance (operating profit +141%) is actually delivered and utilization keeps rising from about 50%; ② Harmonic Drive Systems is named by Tesla/Figure/Boston Dynamics or other Western humanoid customers as a mass-production design win (Western customers are the more plausible incremental source versus Chinese customers); ③ AI robot orders (about ¥2.5 billion) scale quarter by quarter. These would reinforce the bull narrative, but all require factual delivery, not just emotion.

  Conclusion: the market understands Harmonic Drive Systems, but under theme sentiment it chooses not to look far enough. Professional views (Jefferies/DCF/sell-side averages) have already pointed out overpricing, but the long-lens humanoid story plus theme beta has overwhelmed near-lens risk. The narrative inflection point is more likely to appear first on the downside: a humanoid production delay, Leaderdrive confirmed as main Optimus supplier, or a failed traditional-cycle recovery. Any one would reprice the stock from option valuation back to fundamentals. This is the core of "a great company at an uninvestable price; wait for a deep pullback of more than 40% before revisiting". The bet is not that the company is bad, but that the market will eventually reprice a dividend that is unrealized and may go elsewhere.

  Jun 10, 2026