NVIDIA Unveils Open-Source Newton Physics Engine to Accelerate Humanoid Robot Development

Key Takeaways

NVIDIA announced on September 29, 2025, at the Conference on Robot Learning in Seoul, South Korea, that its open-source Newton Physics Engine is now available in beta through NVIDIA Isaac Lab.

The announcement represents a significant industry collaboration aimed at solving one of robotics' most persistent challenges: bridging the gap between simulated training and real-world performance.

The physics engine was co-developed by Google DeepMind, Disney Research, and NVIDIA, and is managed by the Linux Foundation.

Built on NVIDIA Warp and OpenUSD frameworks, Newton enables developers to simulate complex robot actions, including walking through challenging terrain like snow or gravel and handling delicate objects such as cups and fruits.

Rev Lebaredian, vice president of Omniverse and simulation technology at NVIDIA, stated: "Humanoids are the next frontier of physical AI, requiring the ability to reason, adapt, and act safely in an unpredictable world.

With these latest updates, developers now have the three computers to bring robots from research into everyday life — with Isaac GR00T serving as the robot's brains, Newton simulating their body, and NVIDIA Omniverse as their training ground."

Enhanced AI models bring human-like reasoning to robots

Alongside Newton, NVIDIA released Isaac GR00T N1.6, an open robot foundation model that will be available on Hugging Face and integrates NVIDIA Cosmos Reason, a customizable reasoning vision language model designed for physical AI.

The system converts vague instructions into step-by-step plans using prior knowledge, common sense, and physics to handle unfamiliar situations.

Leading robot manufacturers, including AeiROBOT, Franka Robotics, LG Electronics, Lightwheel, Mentee Robotics, Neura Robotics, Solomon, Techman Robot, and UCR, are evaluating Isaac GR00T N models for building general-purpose robots.

The NVIDIA Physical AI Dataset on Hugging Face, which now includes thousands of synthetic and real-world trajectories, has been downloaded over 4.8 million times.

Widespread academic and industry adoption

NVIDIA technologies, including GPUs, simulation frameworks, and CUDA-accelerated libraries, were referenced in nearly half of the accepted papers at the Conference on Robot Learning.

Early adopters of Newton include research institutions such as ETH Zurich Robotic Systems Lab, Technical University of Munich, and Peking University, along with robotics company Lightwheel and simulation engine company Style3D.

Major industry players, including Agility Robotics, Boston Dynamics, Disney Research, Figure AI, Franka Robotics, Hexagon, Skild AI, Solomon, and Techman Robot, are adopting NVIDIA Isaac and Omniverse technologies.

Additionally, NVIDIA's Jetson Thor platform, powered by a Blackwell GPU, has been adopted by partners including Figure AI, Galbot, Google DeepMind, Mentee Robotics, Meta, Skild AI, and Unitree for real-time on-robot AI inference.

Addressing critical industry needs

More than a quarter-million robotics developers worldwide require accurate physics simulation to ensure that skills taught to robots in virtual environments can be executed safely and reliably in real-world applications.

Humanoid robots, with their complex joints, balance requirements, and movements, push current physics engines to their operational limits.

NVIDIA also announced updates to its Cosmos world foundation models, which have been downloaded over 3 million times and allow developers to generate diverse training data using text, image, and video prompts.

Read more: