.jpg?w=1024)
Europe’s humanoid robot race now has a Paris contender.
UMA, a physical AI startup founded in 2025, unveiled the design of its first humanoid robot at the Machina Summit and introduced Real-Time Learning, an AI architecture that teaches robots new tasks through demonstration rather than manual programming.
The launch gives France and the wider EU another entry in a robotics race still dominated by US and Chinese companies, with UMA targeting factories, warehouses, logistics hubs, and other industrial sites before homes.
UMA brings physical AI to Paris
UMA said its robot is designed for factories, warehouses, logistics centers, and industrial facilities, where it could take on repetitive, physically demanding, or hazardous work. The company described its humanoid architecture as a way to fit into environments already built for people, including existing tools and infrastructure.
“Demographic, industrial, and environmental challenges all point to the same reality: societies need greater productive capacity,” said Rémi Cadène, CEO and co-founder of UMA, according to Business Wire.
UMA’s design uses a neutral visor instead of facial features, a soft technical outer shell, and visible mechanical joints. The company called the approach a “dressed machine,” meant to make the robot look calm and capable without blurring the line between person and machine.
Europe gets a robotics trust test
The French Tech Journal reported that UMA showed journalists a prototype assembled in Paris, along with videos of its AI controlling robotic arms for industrial tasks. Cadène said the work was not scripted and described the system as a neural network connected directly to cameras and motors.
The company’s Real-Time Learning approach is the central pitch. Instead of requiring engineers to reprogram robots for each new application, UMA wants robots to learn from demonstrations, practice tasks, adapt to unfamiliar situations, and improve under supervision.
For European manufacturers, that matters because many automation projects fail when systems cannot handle edge cases. A robot that learns on the job could help warehouses, factories, and logistics operators automate more flexible work without having to rebuild entire facilities around machines.
Warehouses come before homes
UMA is not promising mass deployment overnight.
AI Insider reported that Cadène said humanoid robots will take years to reach large-scale deployment, comparing the timeline to the internet and smartphones.
“We’re still at the beginning of this journey,” Cadène said, according to AI Insider.
Cadène’s caution is a reminder that humanoid robots remain expensive, technically difficult, and hard to deploy safely around people. UMA’s first commercial path appears more practical than futuristic: logistics and warehouses first, then manufacturing, with consumer settings much later.
For France and the EU, UMA’s announcement fits a larger industrial question.
Europe has strong research talent and deep manufacturing experience, but the humanoid robotics race is still led largely by US and Chinese companies. UMA bets that Europe’s labor shortages, aging population, and reshoring efforts could make the region an early market for physical AI.
Related reading: Learn more about Portugal’s open-source Amália AI model, built specifically for European Portuguese.