NVIDIA unveiled a broad update to its Isaac robotics platform at GTC Berlin October 21, 2026, introducing a reference humanoid, new datasets, a fresh runtime and a newer version of its robot learning framework. The company says these additions link up the journey from gathering training data to putting robots to work in the real world.
“General-purpose robotics is advancing at extraordinary speed, the ChatGPT moment for robotics is fast approaching,” NVIDIA founder and CEO Jensen Huang said.
A Ready-Made Humanoid
The biggest announcement is Agile ONE SN, a humanoid robot made by Agile Robots on NVIDIA's technology. Building robots usually means stitching together different tools for simulation, training and real-world testing. NVIDIA says Agile ONE SN puts all of these on one platform, reducing setup from several weeks to a matter of days.
Also Read | Layoffs Are Rising In Tech, But These Skills Are Helping People Unlock Better Pay
ALSO READ | Super Micro Contractor Pleads Guilty In $2.5 Billion AI Chip Diversion Case Linked To China
At 1.40 metres and 40 kilograms, the robot has 31 degrees of freedom and can lift 3 kilograms with each arm. Its two WUJI Hand 2 hands have five fingers each, along with 20 degrees of freedom and a sense of touch. Computing is handled on board by NVIDIA's Jetson Thor, and the robot ships with NVIDIA's Isaac GR00T platform already set up. Cameras on its head and wrists, lidar and inertial sensors allow it to sense what's around it, and NVIDIA plans to add its Halos safety system.
Agile ONE SN is due in early 2027. Stanford University and ETH Zurich are among the institutions that have expressed interest.
New Tools for Developers
NVIDIA also made Isaac Lab 3.0 generally available. The open source framework lets developers train and fine-tune how robots behave, scaling from a single desktop up to a data centre. Alongside it comes Isaac OS, a new open source runtime that ties together a robot's AI, sensors, controls and hardware so it can respond in real time.
ALSO READ | Nvidia, Broadcom Shielded From US Power Crunch; Memory Makers Exposed: Morgan Stanley
More than 2,000 ready-to-simulate robot and factory models are now on Hugging Face, as well as part of the Human 1M Dataset, which records people carrying out many different activities. NVIDIA is also releasing three open source workflows for its DGX Station machines to speed up generating data and training robots.
ABB Robotics and KUKA are adding NVIDIA Omniverse libraries to their factory simulation software, and NEURA Robotics is teaching robots new skills in simulation using Isaac Sim and Isaac Lab. According to NVIDIA, Bellboy's hotel robot, Billie, is already working at hotels operated by Marriott, Hilton, IHG and OYO.
Essential Business Intelligence, Sharp Market Insights, Practical Personal Finance Advice, Daily Fuel, Gold and Silver Prices and Latest Stories — On NDTV Profit.
