Physical AI refers to artificial intelligence systems that perceive, reason, and act in the physical world through sensors, actuators, and embodied control. Unlike purely digital AI, Physical AI connects software intelligence with real-world interaction, enabling robots, autonomous machines, and smart devices to move, manipulate objects, and respond to dynamic environments. This topic covers perception, motion planning, control systems, and learning-based behaviors used in robotics, autonomous vehicles, and industrial automation. It also explores simulation, digital twins, and real-world training methods that bridge the gap between virtual models and physical execution. As AI increasingly leaves the screen and enters the real world, Physical AI represents a critical foundation for scalable, intelligent machines operating outside controlled environments.
BMW Group has deployed Figure AI’s Figure 03 humanoid robot at its Spartanburg plant for a logistics sequencing application, following the successful 11-month Figure 02 deployment that supported production of 30,000 BMW X3 vehicles, as the company expands its humanoid program from body shop to assembly logistics.
UBTECH’s U1 full-size companion humanoid robot has accumulated over 5,000 pre-orders on JD.com and Tmall within three weeks of launch, nearly five times the company’s total 2025 full-size humanoid robot sales of 1,079 units, as UBTECH targets 10,000 U1 units in 2026 alongside 10,000 industrial humanoids.
Realbotix has deployed its M-Series humanoid robot and Optio AI teaching assistant at Salamanca City Central School District in New York, with plans to expand the pilot to approximately 500 high school students in fall 2026 across AI and robotics courses.
Goldman Sachs Research estimates Korean companies will account for 30% of global humanoid robot production by 2035 through direct manufacturing and component supply, with Korean supply chains supporting approximately 74,000 units by 2030, driven by automotive parts expertise translating directly into actuator and dexterous hand capability.
Naver Labs Europe has unveiled DIVINE, a universal encoder that consolidates multiple specialized AI encoders into a single model, reducing robot memory consumption by approximately 90% and increasing encoding throughput by up to 12 times for compact autonomous robotic platforms.
MWC Shanghai 2026 opened at the Shanghai New International Expo Center with Honor’s humanoid robots Flash and Vita Boy appearing live on stage, as GSMA Director General Vivek Badrinath named humanoid robotics, the low-altitude economy, and autonomous vehicles as the three most significant frontiers emerging from Asia’s mobile ecosystem.
Morgan Stanley has raised its China humanoid robot shipment forecast to 50,000 units for 2026, nearly doubling its previous projection of 28,000, as commercial verification in factories, logistics, and retail accelerates faster than the bank anticipated at the start of the year.
Researchers at the University of Basel have developed a modular nanorobot comprising a magnetic propulsion module and a swappable payload capsule connected by a DNA-based self-assembly system, demonstrating targeted anticancer drug delivery that reduced HeLa cell viability to 16% within 72 hours.
Agility Robotics has announced a merger with Churchill Capital Corp XI to go public at a $2.5 billion valuation, providing more than $620 million in proceeds to fund commercial expansion of its Digit humanoid robot, backed by over $300 million in existing orders for a new Digit version.
Bear Robotics has signed a definitive agreement to acquire Kinisi Robotics, adding the KR1 humanoid robot and its vision-language-action manipulation models to Bear’s existing fleet of more than 16,000 deployed service robots, completing what the company describes as an end-to-end physical AI platform.