Overview
This National Day holiday, embodied AI was the most visible front-stage star. Scenic spots welcomed “robot new hires”: a traffic robot guiding at West Lake Hangzhou, a “climbing buddy” on Mount Tai, and robots performing 320 meters above ground at Bund Shanghai. The hardest proof point: AgiBot and Chimelong launched the world’s first large-scale embodied-AI theme park - over 300 robots across 100-plus interaction points for guiding, retail and education, stress-tested by real holiday crowds for days.
Meanwhile, AgiBot rolled out its 20,000th unit; Galaxy General revealed humanoid ET1 with three price tiers starting at 79,000 yuan, 26 degrees of freedom and “learn by watching once”; Unitree released the 6B full-body VLA model UnifoLM-WLA-1.0 covering 64 tasks.
Background & Interpretation
1. Technical Background and Evolution
Embodied AI is crossing from “showing tech” to “doing work”. Per CNNIC’s 2026 generative-AI report, it is the most active direction, with humanoid output likely above 100,000 units in 2026. Mass production and real-scenario validation are the two gates to commercialization.
2. Core Drivers and Mechanisms
Supply and demand both push: on supply, VLA models, edge compute and dexterous manipulation matured while unit cost dropped fast (ET1 at 79k); on demand, tourism, retail and manufacturing want “interactive labor”. Chimelong’s 300-robot cluster validated multi-robot coordination and large-scale O&M.
3. Market Structure and Industrial Chain Effects
From “single-unit demo” to “ten-thousand-unit delivery”, the chain (motors, reducers, sensors, OS like PartyOS) matures fast. RoboParty’s open-source RP1 and Galaxy’s ET1 pricing signal a forming price system and ecosystem. China, with manufacturing and scenario advantages, may lead in embodied scale-up.
Implications & Outlook
Near term, tourism, retail and industrial scenes scale first; medium term, embodied AI shifts from “eye-catching exhibit” to “measurable productive tool”, with unit economics dictating diffusion speed.
Takeaways for Industry Participants
- Makers: feed real-scenario data back to models; refine reliability and O&M.
- Scenario owners: pilot with “human-robot collaboration”, not “robot replacement”.
- Investors: watch production yield, unit cost and scenario ROI.
This column compiles industry information and shares technical perspectives; it does not constitute investment advice.