Nvidia Unveils New Physical AI Models, Accelerating the Rise of Intelligent Robots

Nvidia Unveils New Physical AI Models, Accelerating the Rise of Intelligent Robots

San Jose, California — Nvidia has announced a major step forward in robotics with the release of a new suite of physical artificial intelligence (AI) models, designed to help robots better perceive, reason and act in the real world. The launch highlights Nvidia’s growing ambition to position itself at the center of next-generation robotics, from industrial machines to humanoid assistants.

At the heart of the announcement are two key technologies: Cosmos and Isaac GR00T. These physical AI models are built to bridge the gap between digital intelligence and physical action, allowing robots to move beyond rigid, pre-programmed behaviors and instead adapt dynamically to changing environments.

According to Nvidia, the new models enable robots to learn complex skills such as object recognition, spatial awareness, task planning and coordinated movement. By combining large AI models with physics-based simulation, the company aims to make robots more autonomous, flexible and capable of general-purpose work.

Cosmos focuses on teaching robots how to understand and navigate the physical world by learning from large volumes of simulated and real-world data. It helps robots predict outcomes, understand cause-and-effect relationships, and respond more naturally to their surroundings. Meanwhile, Isaac GR00T is designed as a foundation model for humanoid robots, enabling them to generalize tasks — such as picking up objects, using tools or collaborating with humans — without needing extensive reprogramming for every new scenario.

During the announcement, several major industry partners showcased robots powered by Nvidia’s physical AI technology. Boston Dynamics demonstrated advanced mobility and manipulation capabilities, while Caterpillar highlighted applications in autonomous heavy machinery and industrial robotics. Franka Robotics presented precision robotic arms for research and manufacturing, and LG revealed service-oriented robots aimed at commercial and consumer environments.

Nvidia CEO Jensen Huang described physical AI as the next frontier of artificial intelligence, noting that “the future of AI is not just thinking machines, but machines that can act.” By combining high-performance computing, AI foundation models and simulation tools, Nvidia believes it can significantly shorten the time it takes to design, train and deploy intelligent robots.

Industry analysts say the move could accelerate the adoption of robots across sectors such as manufacturing, logistics, healthcare and construction. Rather than building narrowly specialised machines, companies may soon deploy generalist robots capable of learning new tasks on the job.

With Cosmos and GR00T, Nvidia is signalling that robotics is no longer a distant promise but a rapidly approaching reality — one where intelligent machines can safely and effectively operate alongside humans in the physical world.