In the field
Core
Inputs
Sources
Robots
38
Components
18
Data services19
Data producers
7
Capture hardware
10
Data tools
17
Simulation
16
Models
21
Testing & safety
3
Runtime & ops
16
Deployments
27
Also inData tools4
Also inSimulation2
Also inData producers1
Also inModels1
Also inCapture hardware1
Also inDeployments1
Build AI
Build AI

Collects first-person video of factory work and publishes the Egocentric-10K open dataset.

San Francisco, USA · 2025
Claru
Claru

Sells training data for physical AI, including egocentric and demonstration datasets.

2025
Human Archive
Human Archive

Wearable rigs (headsets, tactile gloves, mocap suits) record people doing everyday tasks, mostly via gig workers in India.

San Francisco, USA · 2026
Luel
Luel

Rights-cleared data marketplace and custom contributor collection, including egocentric video subsets for robot training.

San Francisco, USA · 2025
Mecka AI
Mecka AI

Captures human motion with body sensors and phones; sells the EgoVerse dataset, plus evaluation and deployment support.

New York, USA · 2024
micro1
micro1

Expert-demonstrated robotics data for frontier robotics labs, alongside its broader AI training data services.

Sieve
Sieve

Multimodal data lab that sources, filters and annotates video and other real-world data into training-ready datasets for frontier AI labs, including robotics and physical AI.

San Francisco, USA · 2022
Troveo
Troveo

Licensing marketplace for rights-cleared real-world video and other data sourced from content owners, including manipulation footage.

Austin, USA
XDOF
XDOF

Teleoperation and wearable-sensor robot data collection; released the ABC-130K bimanual manipulation dataset with UC Berkeley.

Berkeley, USA · 2024
Appen
Appen

AI data services firm with a Physical AI product line sourced from its contributor network.

Sydney, Australia · 1996
Objectways
Objectways

Annotation vendor that also records head-mounted task footage in India for robotics clients.

Scottsdale, USA
Scale AI
Scale AI

Physical AI data engine: teleoperated and egocentric collection, data factories, and multimodal annotation for robotics labs.

San Francisco, USA · 2016
Toloka
Toloka

Managed crowd data provider expanding into distributed physical AI data collection.

Amsterdam, Netherlands · 2014
Lightwheel
Lightwheel

SimReady asset library, EgoSuite egocentric data and RoboFinals simulation-based evaluation of VLA models.

2023
Midcentury
Midcentury

Closed egocentric video dataset plus Matrix, a cloud simulation platform; emerged from stealth September 2026.

New York, USA
Bones Studio
Bones Studio

Optical motion-capture studio selling human motion datasets for humanoids, including the open BONES-SEED set.

Warsaw, Poland · 2013
Config
Config

Bimanual robot data operations, the Loop demonstration-to-policy platform, and the CFG-1 manipulation model.

San Jose, USA · 2025
DreamVu
DreamVu

Sells the Alia360 360-degree depth camera and runs custom robot data capture programs.

Philadelphia, USA · 2017
microagi
microagi

Shift records industrial work as training data; Atlas post-trains models and deploys robots on factory floors.

Munich, Germany · 2025
20%