BMW has placed a wheeled humanoid robot on a German factory floor, marking the first automotive deployment of Hexagon Robotics’ AEON and a significant European test of physical AI.
The Leipzig pilot builds on BMW’s 2025 experience with Figure 02 in South Carolina. During ten months at Spartanburg, the robot supported production of more than 30,000 BMW X3 vehicles, worked ten-hour shifts and moved over 90,000 components.
AEON now brings those lessons into a plant that combines battery production, moulding, pressing, body work and final assembly.
An industrial body favours wheels over spectacle
AEON stands 1.65 metres, weighs 60 kilograms and travels at up to 2.5 metres per second. Instead of walking, it uses wheels because Hexagon found them faster and more energy-efficient on flat factory floors.
That decision reveals the project’s priorities. A robot does not need to imitate every aspect of human movement. It needs to perform dependable work in the environment where it is deployed.
The robot can exchange its own battery in 23 seconds, supporting continuous operation. Twenty-two sensors—including peripheral, infrared, time-of-flight and SLAM cameras—provide 360-degree spatial awareness. A human-like torso accepts different grippers, hands and scanning tools.
This combination supports material handling and quality inspection. Mobility lets the same unit move between work areas, while interchangeable tools allow several defined tasks rather than one fixed motion at a stationary cell.
Leipzig tests a deliberate rollout
AEON first entered Leipzig in December 2025. Another test is planned for April 2026, followed by a summer pilot with two units working on high-voltage battery assembly and exterior-part manufacturing.
BMW selected Leipzig because its range of processes makes it a demanding validation environment. Success there would provide evidence across much of the production spectrum rather than one isolated task.
The company has also created a Centre of Competence for Physical AI in Production. It consolidates expertise and gives technology partners a path from laboratory evaluation to factory pilot. That institutional structure matters because robots touch safety, facilities, data and workforce design simultaneously.
BMW is not dropping AEON onto an unprepared plant. It has worked to replace production data silos with a consistent platform. Autonomous systems can only learn and coordinate when information about equipment, tasks and quality is accessible in a standard form.
Simulation and data make the machine scalable
AEON uses Nvidia Jetson Orin computers onboard and was trained substantially in Nvidia’s Isaac simulation environment. Hexagon says simulation reduced core locomotion development from months to weeks. Microsoft Azure supports scalable model work, while Maxon supplies actuators.
Simulation allows engineers to expose a robot to many conditions before risking equipment or people. The factory then provides the real-world exceptions that models cannot perfectly anticipate. Data flows between both environments as performance improves.
The source cites a Deloitte survey in which 58% of companies reported some physical AI use, with 80% expected within two years. Such forecasts do not guarantee useful deployment, but BMW’s measured programme offers tangible evidence beyond a stage demonstration.
The key questions are operational: Can two robots work safely around people? Do battery exchanges and maintenance support expected availability? Can one platform perform enough tasks to justify integration cost?
BMW’s choice of a wheeled humanoid suggests physical AI will not converge on one ideal body. Form follows the workplace. Human-like reach is useful because factories are built around people; legs are unnecessary when wheels are better.
Leipzig is therefore less a race toward a general-purpose robot than a disciplined test of flexible industrial labour. Europe’s physical AI story begins not with a machine dancing, but with one proving it can show up for a shift and do useful work.
