NVIDIA boosts robotics development  

NVIDIA has unveiled a series of new open-source tools and AI models aimed at accelerating robotics research and development. The announcement includes the Newton Physics Engine in NVIDIA Isaac™ Lab, the open Isaac GR00T N1.6 reasoning vision-language-action model, and updated AI infrastructure designed to streamline robot learning and deployment. 

The Newton Physics Engine, developed with Google DeepMind, Disney Research, and NVIDIA, is a GPU-accelerated, open-source physics platform managed by the Linux Foundation. Built on NVIDIA Warp and OpenUSD frameworks, Newton enables developers to simulate complex robot actions, from navigating uneven surfaces to handling delicate objects, ensuring safer and more reliable real-world deployment. Leading research labs and robotics companies, including ETH Zurich, Technical University of Munich, Peking University, and Lightwheel, have already adopted Newton.

The open Isaac GR00T N1.6 model incorporates NVIDIA Cosmos™ Reason, a reasoning vision-language model that converts instructions into actionable plans and generalises robot behavior across diverse tasks. Developers can post-train these models using the NVIDIA Physical AI Dataset, which includes thousands of real and synthetic trajectories. Companies such as AeiROBOT, Franka Robotics, LG Electronics, and Techman Robot are exploring Isaac GR00T N models for general-purpose humanoid applications.

NVIDIA also updated its Cosmos World Foundation Models, enabling large-scale synthetic data generation for physical AI training. New workflows in Isaac Lab 2.3 help robots master dexterous grasping and multi-fingered hand tasks, with applications demonstrated on platforms such as Boston Dynamics’ Atlas.

To support these innovations, NVIDIA introduced new AI infrastructure, including the GB200 NVL72 rack-scale system, RTX PRO™ Servers, and Jetson Thor™, all designed to accelerate training, simulation, and on-robot inference. These technologies are already being adopted by major robotics research labs and companies globally, advancing humanoid robotics and physical AI capabilities.