NVIDIA and Agile Robots Launch Humanoid Research Platform

NVIDIA and its German partner Agile Robots have unveiled a humanoid robot platform in a bid to spare researchers the months it can take to put together the hardware, software and AI systems for robotic functions. 

The Agile ONE SN was introduced at the Berlin leg of NVIDIA GTC on October 21. It is a combination of the humanoid hardware from Agile Robots and two of NVIDIA’s products: the Jetson Thor computing system and the Isaac GR00T AI platform. Universities and corporate labs are the target audience for this robot, those who are engaged in work on humanoid intelligence and how it is applied in the real world. 

NVIDIA says that by putting simulation, training and deployment tools under one roof in the development environment, The Agile ONE SN can cut weeks of setup down to a few days. The 1.4-metre tall robot has 31 degrees of freedom for multi-joint movement and weighs 40 kg; a Jetson Thor computer on board handles the AI reasoning and control in real time. There is also motion-tracking, lidar and cameras to let the unit make sense of what is around it and move accordingly. 

Rather than having to build and connect different systems prior to any experiments, developers will have a research platform right at their disposal. Spencer Huang, product director for robotics at NVIDIA, told Reuters that a standardised configuration can be a time saver for researchers, cutting roughly two or three months of integration work. "If we can bring researchers something that's already 80% of the way there, it accelerates their development," he said. 

Agile Robots is in the process of scaling up production of humanoids. Zhaopeng Chen, the company’s chief executive, has told Reuters they plan to be making 1,000 to 2,000 of the Agile ONE S and ONE SN models in 2027, with manufacturing to get underway by the end of 2026. While Stanford and ETH Zurich have shown some interest, no details on pricing or firm orders have been disclosed. 

As for its Isaac ecosystem, NVIDIA has some updates to offer, not least Isaac Lab 3.0, a software runtime and new resources for the training of robots.  

With the SimReady library now offering more than 2,000 simulation-ready assets, a developer can create virtual testing environments before transferring learned behaviors to physical robots. 

Availability of the Agile ONE SN is scheduled for early 2027, at which point NVIDIA’s Halos safety technology will be integrated. 

However, the platform's impact will depend on how easily researchers can use its tools and whether capabilities developed in simulation perform reliably on physical machines. 

It is a chance for Agile Robots to put standardised humanoid hardware in front of research institutions. For NVIDIA, it is further entrenchment in the provision of the software and computing infrastructure for tomorrow’s robots.