Company and supply-chain layer
Who is building the embodied AI stack?
This map summarizes a company dataset into a reader-facing industry view. It is meant to help you orient quickly: region, supply-chain layer, category, and representative companies.
Dataset snapshot: 100 company records, last source update 2026-02-13, concentrated in Guangdong, Beijing, Shanghai, Zhejiang, and Jiangsu. Use this as a starting map, not a procurement database.
What This Map Helps You Decide
| Decision | Use this page to… | Do not use it to… |
|---|---|---|
| Learning direction | connect technical topics to real company categories | decide which company is “best” |
| Market scan | see where companies cluster by region and stack layer | replace a primary-source market report |
| Supplier research | identify whether a company is upstream, midstream, downstream, or algorithm-layer | make procurement decisions without source checks |
| Content planning | choose which robotics subfields need deeper explainers | treat the 100 records as globally complete |
How To Read This Map
Most robotics companies sound similar from the outside. The useful first split is not “AI company” versus “robot company”; it is where the company sits in the physical AI stack.
| Layer | What it usually means | Evaluation question |
|---|---|---|
| Upstream | sensors, actuators, chips, reducers, core controllers | Does it solve a hard bottleneck in cost, precision, latency, reliability, or supply? |
| Midstream | robot bodies, complete systems, robot intelligence stacks | Can it integrate hardware, perception, control, and deployment into one credible platform? |
| Downstream | logistics, service, cleaning, delivery, industrial applications | Does the robot solve a repeated paid workflow better than existing automation? |
| Algorithm | AI chips, models, perception, planning, foundation-model layers | Is there a real robot feedback loop, or only generic model capability? |
Regional Concentration
| Region | Company count in current snapshot | What the cluster suggests |
|---|---|---|
| Guangdong | 29 | Shenzhen/Dongguan hardware depth, service robots, sensors, supply-chain execution |
| Beijing | 27 | universities, research labs, AI models, humanoid startups, platform companies |
| Shanghai | 17 | industrial robotics, embodied AI startups, service robots, manufacturing pilots |
| Zhejiang | 12 | Hangzhou AI/software ecosystem plus robotics hardware companies |
| Jiangsu | 8 | Suzhou/Nanjing manufacturing, actuators, industrial automation |
The concentration is useful for reader orientation, but it should not be mistaken for global completeness. It reflects the current editorial source set.
Category View
| Category | Count | Reader interpretation |
|---|---|---|
| Industrial robotics | 27 | Factory automation and application-specific systems remain the deepest commercial base |
| Robot bodies | 20 | Humanoid, quadruped, biped, and general-purpose robot platforms |
| Service robots | 19 | Cleaning, delivery, hospitality, logistics, and consumer-adjacent robots |
| Actuators | 14 | motors, reducers, joints, controllers, and precision motion components |
| Algorithms | 13 | model, chip, perception, control, and scheduling layers |
| Sensors | 7 | vision, lidar, tactile, and machine-vision components |
Representative Companies By Layer
Robot Bodies
| Company | Region | Notes |
|---|---|---|
| Unitree Robotics | Hangzhou | Quadruped and humanoid platforms; widely visible through Go and H series robots |
| Fourier Intelligence | Shanghai | Rehabilitation robotics background, GR humanoid line, actuator integration |
| AgiBot | Shanghai | Humanoid and embodied AI platform company |
| GalaxyBot | Beijing | Multimodal and model-driven robot platform positioning |
| UBTECH | Shenzhen | Education, service, and industrial humanoid deployments |
| LimX Dynamics | Shenzhen | Biped and wheel-legged locomotion systems |
Upstream Actuators And Motion
| Company | Region | Notes |
|---|---|---|
| Inovance | Suzhou | Servo systems and industrial automation base |
| Leader Harmonic | Suzhou | Harmonic reducers and precision transmission |
| Sanhua | Hangzhou | Thermal and actuator-system manufacturing background |
| Moons' | Shanghai | motors, stepper systems, precision drive |
| Kinco | Shenzhen | low-voltage servo and joint-module related components |
| Shuanghuan | Hangzhou | precision gears and RV reducer capabilities |
Sensors And Perception Hardware
| Company | Region | Notes |
|---|---|---|
| Orbbec | Shenzhen | 3D vision sensors and machine-vision hardware |
| Hesai | Shanghai | lidar systems used across autonomy and robotics |
| PaXini Perception | Shenzhen | tactile sensing and electronic-skin direction |
| RoboSense | Shenzhen | lidar and perception systems |
| Livox | Shenzhen | lidar products in DJI ecosystem context |
| Huaray | Hangzhou | industrial cameras and machine-vision hardware |
Industrial And Service Applications
| Company | Region | Notes |
|---|---|---|
| Estun | Nanjing | industrial robot and motion-control stack |
| Rokae | Beijing | collaborative and flexible industrial robots |
| Dobot | Shenzhen | collaborative and desktop robotic arms |
| Hai Robotics | Shenzhen | warehouse ACR systems |
| Pudu Robotics | Shenzhen | delivery and commercial service robots |
| Gaussian Robotics | Shanghai | commercial cleaning robots |
How This Connects To The Learning Path
Use this map after you understand the technical stack:
- If you are studying perception, inspect sensor and machine-vision companies.
- If you are studying control, inspect actuator, reducer, controller, and robot-body companies.
- If you are studying robot learning, inspect companies with data, teleoperation, simulation, and deployment loops.
- If you are studying commercialization, compare downstream application companies against the workflow they replace.
How This Connects To Data Collection
The company map also explains why robot data is not a single market. A data-collection product can be:
- a teleoperation rig sold to labs,
- a tactile sensor used inside a hand,
- a robot-body platform that generates demonstrations,
- a data factory with trained operators,
- or a downstream fleet that collects operational data.
For the practical layer, continue with Robot Data Collection Methods Compared.
Source Note
This page is derived from an editorial company dataset. Company names, locations, tags, and highlights are consolidated for orientation. Treat counts and labels as an editorial map, and verify individual claims against primary company sources before citation, purchasing, investment, or hiring decisions.