Nvidia used its GTC Berlin keynote on October 21 to expand its Isaac robotics platform, unveiling the Agile ONE SN, a reference humanoid built with Agile Robots, and making Isaac Lab 3.0 generally available. The Agile ONE SN comes preconfigured with Nvidia's Isaac GR00T software, runs on the Jetson Thor computer and is scheduled for availability in early 2027.
The robot is built on Agile Robots' Agile ONE S body, which stands 1.40 meters (4.59 feet) tall, weighs 40 kilograms and has 31 degrees of freedom with a 3-kilogram payload per arm. It carries two WUJI Hand 2 five-finger hands with 20 degrees of freedom each, 1 kHz force control and tactile sensing. Jetson Thor supplies 2,070 FP4 teraflops and 128GB of unified memory, fed by head and wrist cameras, lidar and inertial measurement units.
Nvidia is aiming the Agile ONE SN at frontier humanoid research, and says Stanford University and ETH Zurich have expressed interest. The software stack includes the GR00T foundation model, a data-to-deployment workflow and whole-body control, and Nvidia says the preconfigured package cuts setup time from weeks to days. On-robot safety protocols from Nvidia Halos are planned.
“The breakthrough for humanoids is usefulness at scale: one adaptable machine that learns many skills”— Zhaopeng Chen, Founder and CEO of Agile Robots
Key facts
- 01Nvidia and Agile Robots unveiled the Agile ONE SN reference humanoid at GTC Berlin on October 21, due in early 2027.
- 02The robot stands 1.40 m, weighs 40 kg, has 31 degrees of freedom and runs on Jetson Thor with 128GB of unified memory.
- 03Isaac Lab 3.0, Nvidia's open source robot learning framework, is now generally available.
- 04Nvidia's SimReady Central Library puts more than 2,000 simulation-ready robotics and factory assets on Hugging Face.
- 05Nvidia also introduced the OpenShell Reference Architecture for monitoring and halting AI agents.
Isaac Lab 3.0, now generally available, is Nvidia's open source, agent-ready framework for training and refining robot policies from a desktop up to a data center, with multi-GPU workflows. It supports both the Newton and Nvidia PhysX physics engines. Nvidia also introduced Isaac OS, an open source runtime built on Nvidia Holoscan that links AI models, sensors, controls and robot hardware for deterministic, real-time performance from simulation to physical robots.
Training data got its own set of releases. The SimReady Central Library puts more than 2,000 simulation-ready robotics and factory assets on Hugging Face, with contributions from Comau, FANUC, Franka Robotics, LG Electronics, Universal Robots and Vertiv, among others. The Human 1M Dataset covers diverse human activities with synchronized three-view video and 3D hand tracking; it is on Hugging Face, with the full dataset available to Cosmos Coalition partners.
Nvidia added three open source reference workflows for its DGX Station: refining GR00T 1.7 models with reinforcement learning, training dexterous manipulation skills in Isaac Lab, and post-training Cosmos world models for synthetic data generation. Companies supporting the workflows include 1X, Agile Robots, Humanoid, Skild AI and Sereact.
Halos, Nvidia's safety platform built on IGX Thor for autonomous vehicles and robots, picked up robotics partners including Agile Robots, Agility, Dexory, FORT Robotics and KION. SICK is evaluating Halos for dynamic industrial safety, while Accenture, TÜV SÜD and Reynolds & Moore are working to speed certification.
Nvidia paired the robotics news with the OpenShell Reference Architecture, a blueprint for creating, restricting and monitoring AI agents. Models run on Nvidia's Vera Rubin system, agent software runs on Vera CPUs, and BlueField-4 data processing units act as guards that shut down agents attempting unwanted actions. Nvidia says the design can quarantine and halt rogue agents within milliseconds, with CoreWeave, Crusoe, Lambda, Nebius and Oracle Cloud Infrastructure among the rollout partners.
“AI capability and safety must advance together.”— Jensen Huang, Founder and CEO of Nvidia
Industrial partners are already building on the stack. ABB Robotics is using Omniverse libraries with RobotStudio HyperReality to virtually commission robotic systems, and Toyota Material Handling Europe and SoftServe are using Omniverse and Isaac Sim for autonomous material handling, cutting deployment time by up to 50%. Bellboy uses Jetson and GR00T for Billie, a robot already deployed at hotels run by Marriott, Hilton, IHG and OYO.
Nvidia frames the push around a $50 trillion manufacturing and logistics market that it says physical AI could transform. The company noted that many of the announced products are at various stages of development and that availability and timing may change.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
