If humanoid-robot volume scales on cheap QDD actuators, the strainwave-gearbox incumbent (HarmonicDrive) loses share even as the category grows

The chain (one paragraph)

SemiAnalysis argues Chinese makers led by unitree are scaling humanoids on QDD (quasi-direct-drive) actuators — low-ratio planetary gearboxes + beefy motors — which are ~80% cheaper, higher-efficiency, faster to iterate, and good-enough for the warehouse tote-handling tasks where humanoid volume first becomes economic. The entire ~200-firm Chinese humanoid ecosystem is standardizing on QDD. If humanoids become the dominant new actuator-demand market and they're predominantly QDD, then harmonic-drive (TSE: 6324) — the strainwave/harmonic-gearbox incumbent — sees its addressable share of the fastest-growing robot category compress, even as total robot-actuator demand explodes. Net: the category tailwind accrues to commodity planetary-gearbox/BLDC-motor suppliers and to the magnet supply chain (drone-warfare-demand-to-mp-hree-rerate), not to the precision-strainwave incumbent.

Candidate tickers

• 6324.T (HarmonicDrive Systems) — the short/loser leg: a structurally-disrupted incumbent. Thin US accessibility (Tokyo listing) is a practical constraint.
• (Long leg already covered by the rare-earth-magnet mechanisms — see Related — so this hypothesis is primarily about the incumbent-disruption short.)

What evidence would convert this to an active thesis

• HarmonicDrive guidance or commentary acknowledging humanoid/QDD share loss, or flat-to-declining robot-segment orders while humanoid unit volume accelerates.
• An independent (non-SemiAnalysis) source confirming Western humanoid makers (Figure, Apptronik, Tesla Optimus) migrating from strainwave to QDD at volume.
• A liquid US-accessible instrument (ADR/ETF) to express the short — without one, the chain stays a research note, not a position.

What would kill it

• Humanoids settle on hybrid actuator stacks that keep strainwaves in the high-precision joints (hands/wrists), preserving HarmonicDrive's premium niche — the reliability moat SemiAnalysis itself concedes ("HarmonicDrive currently remains the reliability leader").
• Humanoid deployment economics stall (the $30/hr threshold proves fragile), so the new demand market never materializes at scale.

Sources (initial)

• 2026-06-08-feed-semianalysis-china-unitree-dominate-global-robotics — QDD vs strainwave cost/iteration/efficiency comparison; the disruption framing (partial/paywalled)

Related

• unitree — the QDD champion
• harmonic-drive — the incumbent at risk
• drone-warfare-demand-to-mp-hree-rerate — the magnet-demand long leg of the same robotics tailwind
• china-ree-controls-to-us-producer-stack — the binding magnet constraint