Agile Robots is planning to build up to 2,000 compact humanoid robots in 2027 as part of a partnership with Nvidia aimed at universities and corporate research laboratories.
The Munich-based company expects production of its Agile ONE S and ONE SN models to start before the end of 2026. Chief executive Zhaopeng Chen told Reuters that anticipated output would range from 1,000 to 2,000 units next year, with Stanford University and ETH Zurich among potential customers.
The ambition is to give researchers a more accessible platform for developing and testing the capabilities that could eventually put humanoids to work in factories and warehouses.
Smaller robots, faster development
Standing 1.4 metres tall and weighing around 34kg, the new models are compact alternatives to Agile’s full-sized humanoids. Their smaller dimensions are intended to make them easier and safer to handle in research environments.
The ONE SN features a configuration developed jointly with Nvidia, combining dexterous hands and cameras with Nvidia’s Jetson Thor onboard computer. The partnership also brings Nvidia’s GR00T AI model together with Agile’s hardware to support data collection and robot skill development.
That integration could help tackle a practical obstacle for robotics teams: the time spent getting separately sourced hardware and software to work together.
Spencer Huang, Nvidia’s product director for robotics, told Reuters that a standardised configuration could save researchers two to three months of integration work.
From laboratory experiments to useful work
For researchers, the potential benefit is more time spent testing what robots can do and less time assembling the platform needed to run those experiments.
For Agile, the move opens a route into research institutions developing the next generation of robotic capabilities. It also places the company among a growing field of European humanoid developers, including Germany’s NEURA and Britain’s Humanoid.
The wider significance will depend on what researchers achieve with these machines. Easier access to an integrated humanoid platform could accelerate experimentation, but translating those results into reliable industrial work remains the commercial test.
