Humanoid Robots

Humanoid robots are robots whose physical form resembles the human body, generally comprising a torso, a head, two arms, and two legs, and are designed to work in environments and tools built for people. Long a staple of research laboratories and science fiction, humanoid robots entered early commercial deployment in the mid-2020s, with fleets working on automotive assembly lines, in warehouses, and in consumer pilots.^[1][2][3] By 2026 more than 140 companies worldwide were manufacturing humanoid robots, and industry analysts began forecasting a multi-billion-dollar market.^[1][5]

History and Background

The first full-scale humanoid robots were built in Japan in the 1970s, beginning with WABOT-1 at Waseda University in 1973. Subsequent research milestones included Honda's ASIMO, introduced in 2000, and Boston Dynamics' hydraulically actuated Atlas, unveiled in 2013, which demonstrated remarkable locomotion and acrobatic capability. For most of this period, humanoid robots were demonstrations rather than productive tools: a robot on a factory floor meant a demo reel, not a shift schedule.^[2][3]

The field changed rapidly in the 2020s with the convergence of advances in embodied AI, neural-network-based control, and cheaper actuators and sensors. Companies such as Tesla, Figure AI, Agility Robotics, Unitree, 1X Technologies, and Apptronik attracted billions of dollars in investment and began deploying robots in real production environments. Tesla entered the field in 2021 with its Optimus programme, Figure AI reached a reported US$39 billion valuation in 2025, and Chinese firm Unitree shipped roughly 5,500 humanoid units in 2025.^[1][3][4]

Technology

Modern humanoid robots combine computer vision, natural language interfaces, and reinforcement learning-based control with high-torque electric actuators, tactile sensors, and onboard compute. Neural networks trained in simulation and on teleoperated demonstrations allow the robots to perceive their surroundings, grasp and manipulate objects, and navigate dynamic environments. Key technical challenges include software reliability, recovery from unexpected states, maintenance of complex hardware, and the large volume of data required to teach each task.^[3][6]

The market spans a wide price range. Unitree's G1, the lowest-cost production humanoid, starts near US$16,000, while 1X's home robot NEO is priced around US$20,000; full-size industrial platforms such as Agility's Digit and Boston Dynamics' electric Atlas cost substantially more, and Figure's commercial service is reportedly billed at about US$25 per robot-operating-hour.^[1][4]

Applications and Commercial Deployment

The first scalable market for humanoid robots is manufacturing, particularly automotive assembly, where structured environments, repetitive tasks, and labour shortages create a clear business case. Verified 2026 deployments include Figure robots logging more than 1,250 operational hours at BMW's Spartanburg plant, Tesla operating Optimus units inside its own factories, Apptronik's Apollo moving assembly kits at Mercedes-Benz facilities, and Agility's Digit having completed tens of thousands of operating hours across warehouse customers.^[2][3][4] In logistics, humanoid and mobile robots are used for material handling, parts transfer, bin picking, and kitting.^[2]

Analysts caution that production deployments remain limited to a handful of industrial partners, and that high-speed, high-precision tasks such as welding remain the domain of traditional fixed-arm industrial robots.^[2][6] Market forecasts differ widely: IDTechEx projects the humanoid robot market at roughly US$29.5 billion by 2036, while other analysts have estimated US$38 billion by 2035 and US$152 billion by 2040.^[5][7]

>See Also

• Physical AI
• AI in Malaysian Manufacturing
• AI in Malaysian Logistics
• NVIDIA
• Reinforcement Learning

References