Harmonic Drive Systems

Investment Thesis

Inventor of the strain-wave (Harmonic Drive®) gear — the zero-backlash precision reducer for robot joints. Global share ~35% in harmonic reducers (NOT the ~50% the X buzz claims): China's Leaderdrive/Green Harmonic now also holds ~35% globally + 60%+ of China and is sole-source to Tesla's Mexico Optimus line — HDS and Leaderdrive are effectively co-leaders, split premium/Western (HDS) vs cost-tier/China (Leaderdrive). PHYSICAL-AI INFLECTION (the reason for this 2026-06 evolve): each humanoid uses 14-40 harmonic drives vs ~6 for an industrial arm, so unit demand scales exponentially as humanoids commercialize. Four converging catalysts in 2026: (1) MLCC ROTATION — Murata's president explicitly named humanoid robots the 'second growth curve' for MLCCs beyond AI servers (a GB300 server = ~30,000 MLCCs); capital that rode the cloud-AI passive-component supercycle (Murata/TDK, +15-35% price hikes Apr-2026) is rotating into edge/physical-AI picks-and-shovels, where HDS is the reducer analog of Murata's MLCC chokepoint. (2) NVIDIA Isaac GR00T + Jetson Thor (2026 GTC cycle; GTC-Taipei Jun-2026 GR00T Reference Humanoid on Unitree H2 Plus) lowers the humanoid software barrier industry-wide. (3) OpenAI stood up an in-house Robotics division (11 SF roles incl. Actuator Design Engineer, $210-310k) — validates actuator/reducer demand. (4) Korea robotics complex +65% in 6 months (₩25.3T→₩44.5T); Samsung-backed Rainbow Robotics ₩13T, govt ₩2T physical-AI fund. HDS is the cleanest TSE-listed, liquid, platform-agnostic exposure to the precision-reducer chokepoint. BUT: spot ¥7,800 (near 52w high ¥8,400) is ABOVE the ~¥5,400 analyst consensus target — consensus implies DOWNSIDE; this leg is momentum/narrative-driven. P/E 459x TTM on trough earnings (FY-Mar2026 net margin ~2.7%), Fwd PE 164x on company guidance EPS ¥47.54 / 126x on consensus EPS ~¥62. Right thesis, expensive entry. TESLA VALUATION BENCHMARK (rev 2, 2026-06-01): the market already sustains ~400x trailing / ~206x forward P/E on Tesla (verified: TSLA P/E TTM 399.81, fwd 205.57, May 30 2026) for the SAME Optimus/physical-AI narrative. HDS — the upstream reducer supplier — screens CHEAPER than Tesla on the durable ratios (P/B 9.2x vs ~19.5x; PEG 2.79 vs ~6.0) and offers earlier, more-leveraged exposure to the joint chokepoint. Reframe: HDS's 459x trailing is on TROUGH earnings; on forward it's 164x (guidance) / 126x (consensus) — below Tesla's trailing and ~in line with Tesla forward. PRODUCTION RAMP (verified): Optimus Gen 3 starts summer-2026 at Fremont but 'agonizingly slow' — realistic 2026 output low-thousands; industry ~10-50k; real volume inflection 2027-2028 (Giga Texas 10M/yr capacity from 2027+). DEMAND MATH: 25-35 reducers/humanoid x 100k units/yr = 2.5-3.5M reducer units/yr — transformative vs HDS's ~¥60B revenue — BUT that mass tier is exactly where Leaderdrive (China, 30-40% cheaper) competes; HDS captures the premium slice. UPSIDE (3-5yr): base 2-3x, bull 5x+, 10x requires near-dominant share (capped by Leaderdrive co-leadership).

Key Dates