Agile Robots plans 2,000 research humanoids with Nvidia
Agile Robots says its Nvidia-backed research humanoids could enter production by the end of 2026 and become available in early 2027.
Agile Robots says it plans to produce between 1,000 and 2,000 compact humanoid robots next year through its partnership with Nvidia, with the first Agile ONE S and ONE SN units scheduled to begin production by the end of 2026. The company expects the models to become available in early 2027, with universities and corporate research laboratories among the target customers.1
The figures were reported by TechNode from comments that Agile Robots CEO Zhaopeng Chen gave to Reuters. They describe a research-focused expansion rather than a general consumer launch: the robots are intended to give developers a ready-made platform for building, training and validating humanoid skills before deploying them in real-world environments.2
Agile ONE SN is based on the company's compact Agile ONE S humanoid body and is built around Nvidia's Isaac GR00T development platform. The official specification lists a height of 1.4 metres, a weight of 43 kilograms and 31 degrees of freedom in the body. Its hands add another 20 degrees of freedom each, bringing the full system to 71 degrees of freedom.1
The robot also includes Nvidia Jetson Thor onboard computing, RGB-D cameras, wrist cameras and 3D LiDAR. Agile Robots says the system can deliver up to 140Nm of joint torque and carry up to 3kg per arm. Wi-Fi 7, Bluetooth and an emergency-stop button are among its other listed features.1
The important part of the partnership is the software workflow. Agile ONE SN ships with preconfigured Isaac GR00T software, including the GR00T 1.7 open vision-language-action model, whole-body control and a reference workflow covering demonstration capture, data generation, simulation, training, evaluation, deployment and real-world testing.1
That approach is designed to reduce the integration work that can slow robotics research. Instead of assembling a body, sensors, compute platform and training pipeline separately, a lab can begin with one open and modular system. Agile Robots says researchers can use the complete workflow or integrate individual components into an existing stack, while retaining control of their data.1
Agile Robots launched ONE SN with Nvidia at Nvidia GTC Berlin on October 21. The company says the platform is aimed at research and development, and that interest includes institutions such as Stanford University and ETH Zurich. Skills developed on the compact ONE SN can later be transferred to the full-size Agile ONE, which is positioned for industrial applications.1
Our opinion
A production target of up to 2,000 units would make Agile ONE SN more consequential than another humanoid prototype announcement, but the number is still a plan rather than a shipment total. The more interesting part is the attempt to package hardware, sensing, simulation and training into a research platform that labs can actually use.
That strategy also makes the Nvidia partnership practical rather than cosmetic. Isaac GR00T and Jetson Thor give Agile ONE SN a recognisable software and compute foundation, while Agile Robots supplies the physical system and the route into industrial deployment. The early-2027 availability target will be the real test: researchers will need reliable hardware, useful tools and transparent pricing, not just a compelling demonstration at a trade show.