Skip to content
All research
Landscape2026-07

AI & Robotics

Mapping the humanoid robotics landscape and the technological stack: actuation, compute, teleoperation, and the policy problems nobody puts in the keynote.

USA

Tesla Optimus

Vision, end-to-end learning, in-house actuators

Factory tasks in limited form; public demos ahead of unsupervised hours

USA

Figure

Humanoid + BMW-class manufacturing pilots, foundation-model bets

High-visibility pilots; teleop still in the data loop

USA

Boston Dynamics Atlas

Electric Atlas, world-class mobility, Hyundai industrial path

Research and industrial transition

USA

Agility Digit

Biped for logistics, GXO-style warehouse work

Early commercial hours; form factor is the thesis

Norway / USA

1X Neo

Soft-robotics bias, home thesis, tendon-driven hardware

Consumer narrative, safety case still the gate

USA

Apptronik Apollo

Industrial humanoid, Mercedes and Google-class partners

Partner-led factory path

China

Unitree

Cost, volume, quadruped heritage into humanoids

Hardware accessible; stack catching up in public

Germany

NEURA Robotics

Cognitive robotics, European industrial customers

EU-native alternative in the conversation

A humanoid is a bet that the world is already built for humans, so the robot should be too. The stack underneath that bet is not one product: actuators and gearboxes, battery and thermal, hands, onboard compute, foundation models, teleoperation for data, and a safety case that industrial regulators will actually accept.

The interesting split in 2026 is not 'who has a demo'. It is who is collecting real hours in a real workplace, who still depends on a puppeteer, and who has a path from supervised factory tasks to unsupervised ones without rewriting the safety case every quarter.