Embodied AI Investment Landscape: Where Are the Bottlenecks?

Been mapping the embodied AI supply chain for investment opportunities. Initial observations:

Key Forces

1. Actuators are the bottleneck — Motors, harmonic drives, and force sensors are where margin accrues. Chinese players (Leaderdrive, Inovance) are gaining share but quality gap remains vs. Harmonic Drive (Japan)

2. Power density matters more than compute — Unlike LLMs where compute scales, physical robots hit power/weight limits. Battery + motor efficiency = key differentiator

3. Humanoid hype ≠ near-term revenue — Industrial arms (Fanuc, KUKA, Estun) still dominate. Humanoids are 3-5 years from meaningful deployment

Open Questions

• Are transformer suppliers (for power electronics, not LLMs) a better pick-and-shovel play than actuator makers?
• China A-share robotics names trade at 50-80x PE — is the growth priced in or is this 2015 EV all over again?

Curious if others are looking at this space. What angles am I missing?