Humanoid Robot Companies and Robot AI Startups in 2026

Humanoid robots, machines shaped roughly like people with arms, hands and legs, have moved from science fiction and research labs into factories and warehouses. A wave of investment since 2023 has pushed several robotics startups to multibillion-dollar valuations, driven by the belief that the same AI advances that produced ChatGPT can give robots the general intelligence to work in human environments. This guide covers the private humanoid robot companies and robot AI startups in our database, the major players outside it, how the technology works and the obstacles that remain.

Each company in the table links to a full profile with sourced valuation, funding and investors. Some important players, such as Tesla and Unitree, are public or listed and appear only in the discussion.

Private robotics leaders in our database

Two approaches: building bodies versus building brains

Robotics startups fall broadly into two groups. Some build complete humanoid robots, designing their own hardware, motors, hands and batteries along with the AI that controls them. Others focus on the "brain": foundation models that can control many different robots, which they license to hardware makers. Both approaches have attracted billions of dollars.

Full-stack humanoid makers

Figure AI, based in California, is the most valuable private humanoid company in our database at US$39 billion. It builds its own humanoid robots and AI system, called Helix, and has worked with BMW on manufacturing tasks. Outside our database, Agility Robotics makes Digit, a humanoid used in logistics warehouses; Apptronik makes Apollo and has partnered with Mercedes-Benz and Google DeepMind; 1X, a Norwegian-American company backed by OpenAI's startup fund, makes the Neo robot aimed at homes; and Boston Dynamics, owned by Hyundai, makes the electric Atlas. Canada's Sanctuary AI focuses on highly dexterous hands and general-purpose robots for work tasks.

Robot foundation model companies

Skild AI, founded by Carnegie Mellon professors, builds Skild Brain, a model designed to control many kinds of robots, and raised at a valuation above US$14 billion in 2026. Physical Intelligence, founded by leading robotics researchers, develops general-purpose robot policies, its π models, that have shown robots folding laundry, cleaning tables and handling unfamiliar objects. Large technology companies are also active: Google DeepMind's Gemini Robotics models and Nvidia's GR00T models for humanoids are offered to robot makers.

Public and listed players

• Tesla: develops the Optimus humanoid and has said it aims to use thousands in its own factories before selling them more widely.
• Unitree: a Chinese company known for low-cost quadruped and humanoid robots, which have become popular with researchers and in demonstrations.
• UBTech and other Chinese makers: China has dozens of humanoid companies supported by government policy, many focused on manufacturing and cost reduction.
• Hyundai (Boston Dynamics): the long-time leader in dynamic robots, now commercialising the electric Atlas.

Why investors are excited

Three trends converged. First, AI models became much better at understanding images, language and actions, making it plausible for robots to follow instructions and adapt to new situations rather than repeat pre-programmed motions. Second, hardware costs fell as components such as motors, batteries and sensors became cheaper, partly thanks to electric vehicle supply chains. Third, labour shortages in manufacturing, logistics and care work in many countries created demand for automation that can work in spaces designed for people. A humanoid shape means a robot can, in principle, use the same tools, stairs and workstations as humans, avoiding expensive redesigns of factories and warehouses.

The potential market is enormous: if robots could perform even a small share of physical labour, the value would dwarf most software markets. That possibility, rather than current revenue, explains valuations far above what these companies earn today. Skild AI, for example, said it reached about US$30 million in revenue in 2025 while being valued above US$14 billion.

How robot AI works

Modern robot AI often uses vision-language-action models: neural networks that take camera images and an instruction, such as "put the dishes in the rack", and output the motor commands needed to do it. Training them requires large amounts of data showing robots, or humans, performing tasks. Companies gather data in several ways:

• Teleoperation: humans control robots remotely to perform tasks, recording the movements for training.
• Simulation: robots practise millions of times in virtual environments, which is fast and safe but imperfectly matches reality.
• Human video: models learn from videos of people doing tasks, though translating human movements to robot bodies is difficult.
• Fleet learning: robots deployed in real settings collect data that improves the model for all robots.

Data is the biggest bottleneck. There's no equivalent of the internet's vast text archive for physical actions, so companies that can collect diverse, high-quality robot data at scale may gain lasting advantages.

The obstacles

• Reliability: factories need robots that work correctly far more than 99% of the time; demos often show best cases.
• Safety: humanoids are heavy and powerful, and working alongside people requires strict safety certification.
• Cost: humanoids must become cheap enough to compete with human wages and simpler specialised machines.
• Dexterity: human hands are extraordinarily capable; replicating fine manipulation remains one of the hardest problems.
• Battery life: many humanoids can work only a few hours before recharging.
• Business models: companies are experimenting with selling robots, leasing them or charging per hour of work ("robots as a service").

Where humanoids are being used first

Early deployments focus on structured industrial settings: moving totes in warehouses, loading and unloading, carrying parts in car factories and simple assembly support. These environments are more predictable than homes and have clear economic value. Home robots that cook, clean and care for people are a longer-term goal; companies such as 1X and Figure have shown home demonstrations, but widespread home use will require far higher reliability and lower cost.

Competition between the US and China

China has made humanoid robots a national priority, with government funding, industrial clusters and a strong electronics and EV supply chain that lowers hardware costs. Chinese companies have released humanoids at prices far below Western competitors. US companies lead in AI models and have raised more private capital. The competition resembles the electric vehicle race, where Chinese manufacturers achieved scale and cost advantages quickly. Export controls on advanced chips may affect Chinese robot AI, while Western companies face higher hardware costs.

How to evaluate a humanoid robot company

• Are robots working in paid commercial deployments, and for how many hours?
• Is the AI general, or tuned for specific demonstrations?
• What is the cost to build each robot, and how will it fall with volume?
• How much data does the company collect, and how does it improve the model?
• Who are the customers, and are contracts firm or pilots?

What to watch

Watch for large-scale factory deployments, falling robot prices, new robot foundation models from startups and big tech, funding rounds and possible IPOs, and progress in China. Our guides to the most valuable AI startups and AI infrastructure companies cover the wider AI landscape.

The supply chain behind humanoids

A humanoid robot needs dozens of actuators, the motors and gearboxes that move its joints, along with sensors, cameras, batteries, computing chips and hands with many degrees of freedom. Actuators are among the most expensive components, and companies are racing to design cheaper, more reliable versions. Much of the global supply chain for motors, magnets and batteries is concentrated in China, which gives Chinese robot makers cost advantages and raises supply security questions for Western companies. Scaling production from hundreds to tens of thousands of robots a year will require factories and suppliers that barely exist today.

Funding sources

Humanoid companies have raised from venture firms, technology giants and sovereign funds. Nvidia, Microsoft, Amazon, OpenAI's startup fund, SoftBank and Jeff Bezos's investment firm have all backed robotics startups, partly because robots would use enormous amounts of AI computing and partly to secure strategic positions. Large rounds give companies years of runway to develop hardware, but also raise expectations for commercial progress.

New entrants to watch: 1X, Agility and Galbot

Three companies show how varied humanoid strategies have become. 1X Technologies sells NEO, a soft-bodied home robot priced at US$20,000 or US$499 a month, and was reported in 2025 to be seeking funding at US$10 billion or more. Agility Robotics, maker of the warehouse humanoid Digit, agreed in June 2026 to go public through a merger with Churchill Capital Corp XI at a US$2.5 billion pre-money value. And Beijing's Galbot, whose wheeled humanoids pick goods in pharmacies, raised RMB 2.5 billion in March 2026 after a round at a reported US$3 billion and is preparing a Hong Kong IPO.