Physical AI Companies

DexHand021 Pro, DexHand021 and DexHand021 S serve different manipulation needs. DexTele turns human demonstrations into synchronized training episodes.

Walker C1 targets guidance, shopping help, patrol and education, with 3 m/s-plus motion listed. Walker S2 prioritizes industrial payload and continuous operation.

AgiBot A3 is a 173 cm, 55 kg biped with a listed 10-hour runtime. Real deployment still depends on hand choice, floor conditions and mission energy.

Galbot G1 pairs a 173 cm, 92.5 kg body with 10 kg dual-arm payload and an 8-hour lab rating. Price, lead time and task fit still require a quote and pilot.

Japan’s five-company Physical AI effort is an exploration of shared control infrastructure, not a finished joint robot or new joint venture.

R-dog targets the stairs, curbs and thresholds that stop wheeled delivery robots. It remains a concept with commercial deployment planned for 2027.

SEER’s WRC 2026 announcement frames 1 million robots as a 2030 deployment goal, not its current base. The strategy combines planning, motion and field data.

Gravis Robotics has retrofit systems on jobsites and a $200 million Series A. Commercial maturity now depends on paid fleet hours and repeat economics.

Seeing Machines is building a human-centered perception layer for robots—one that estimates attention and intent without replacing certified safety controls.

LG is organizing industrial, commercial and home robots around shared actuator and data infrastructure. This map shows what each business layer must prove.