The first wave of general-purpose humanoid robots is moving from demo videos into pilot deployments. The supply chain has five layers: the foundation models that give the machines purpose, the chips that run the inference, the sensors that let them see, the actuators that let them move, and the OEMs and end-customers actually putting them on the floor.
Models at the apex. Deployment on the warehouse floor. Every US-listed public company actually operating in between.
The general-purpose AI models trained to control physical bodies — NVIDIA's Project GR00T and Cosmos world model, Google DeepMind's Gemini Robotics. The brains the humanoid OEMs license or partner around.
The accelerators and SoCs that run the inference on the robot itself — Jetson-class edge compute, custom ASICs, and the broader GPU ecosystem the robotics labs train and deploy on.
Machine-vision cameras, depth sensors, and lidar — the perception stack a humanoid needs to understand a workspace. Most of these companies were built for autonomous vehicles and industrial inspection; humanoid robotics is the next adjacent market.
Industrial automation incumbents whose precision motors, servo drives, and motion-control systems are the closest existing analog to humanoid actuation. Includes Teradyne, owner of Universal Robots and the MiR mobile robotics platform.
The publicly listed companies actually building humanoids or putting next-generation robotics into production environments — Tesla's Optimus program, Amazon's fulfillment automation, Intuitive Surgical's surgical robotics, and Symbotic's warehouse systems.
Constituent lists are reviewed quarterly against segment disclosures, R&D commentary, and stated robotics roadmaps. Quotes are sourced from licensed market data; the Veridion Score is computed from six published factors. Inclusion is not a recommendation.