Segment
Humanoids and embodied AI: the bet on general-purpose machines
Industrial robots work because the world around them is engineered to be predictable. The humanoid bet is that a learned policy can handle an unmodified world instead — removing fixturing and programming rather than performing faster. That is a claim about software, and the hardware is the easier half.
In one sentence
Embodied AI is the application of learned policies to physical control, most visibly in humanoid robots intended to work in spaces built for people rather than in cells engineered around machines.
The case for a human form is compatibility: stairs, door handles, tools and workspaces were all built for a human body, so a machine with that shape can in principle work in them without changing the building. The case against is that the form is a poor engineer's choice for almost any single task — a wheeled base is more efficient, a fixed arm is more precise, and legs cost energy and complexity.
The real question is not the shape. It is whether a policy can be trained to do useful physical work reliably enough for production, and whether it can be specified by demonstration rather than by programming. If that works, the economics of a very large amount of currently manual work change; if it does not, the humanoid is an expensive way to do what a purpose-built machine does better.
How this breaks down
Split by hardware, by the learning that drives it, and by where the training data comes from.
- Humanoids and embodied AIthis page
- Humanoid hardwareDegrees of freedom, torque density, hands, and why battery life is the specification nobody quotes.Breaks down into: Dexterous hands · Power and thermal2 pages belowChokepoint
- Embodied learningImitation learning, simulation-to-real transfer, and why physical tasks are harder than text.Definition pageChokepoint
- Teleoperation and dataHumans driving robots to produce the demonstrations, and the operation that turns that into a dataset.Definition pageChokepoint
What this depends on
1 of these is marked as a chokepoint: a handful of qualified suppliers, a multi-year lead time, or a single geography.
- Supply chainChokepoint
Actuators, reducers and magnets
A humanoid needs dozens of high-torque actuators, drawing on the same constrained reducer and magnet supply as everything else.
Actuators and reducers - Technology
Real-time motion control
A policy issues targets a few times a second; the servo layer underneath still has to hold balance and track them, and that is the same control problem an industrial arm has.
Control and software - Standard
Machinery safety standards
A machine of this mass moving beside people is assessed under the same standards as any other, and whether the form can be certified for that at all is still unresolved.
Safety and collaboration - Technology
Navigation in an unmodified building
A humanoid that has to walk to the work needs the same mapping, localisation and dynamic obstacle handling as a wheeled robot; legs change the balance controller, not the navigation problem.
Autonomous navigation
Companies across Humanoids and embodied AI
Every company named on a step below this page, ordered by how many of those steps it appears at. Compiled from the pages themselves rather than written separately, so the two cannot disagree. Not a ranking and not a recommendation.
- Figure AIPrivate4 steps
Humanoid hardware · Dexterous hands · Embodied learning · Teleoperation and data
- Physical IntelligencePrivate2 steps
- Unitree RoboticsShanghai2 steps
- 1X TechnologiesPrivate1 step
- Agility RoboticsPrivate1 step
16 more companies appear at a single step each; they are named on the pages for those steps.
How these pages are written
Each page explains one technology in plain language, states what it depends on, and names companies by what they supply at that step. Company roles are described qualitatively and deliberately carry no market shares, revenue figures or rankings — those change faster than an explainer can, and a stale number is worse than none. Ticker links point at company pages on this site and are provided for reference only.
Nothing here is investment advice, a recommendation, or a forecast. A company named on a page about a technology is not thereby a good investment, and the chokepoints described are structural facts about supply chains rather than predictions about prices. Technology moves; where a page describes something as unresolved or in development, that was true when it was written.
Plutux is not an investment adviser. Market data and AI-generated analysis are for information and education only, not investment advice. Disclaimer