Nvidia aims to become the Android of generalist robotics

On Monday, Nvidia revealed details about its full-stack ecosystem for physical AI, which refers to artificial intelligence systems operating in the physical world. The company released a new stack of robot foundation models, simulation tools, and edge hardware. This launch signals Nvidia’s ambition to become the default platform for generalist robotics—robots capable of performing diverse tasks rather than narrow, single-use functions. The software release includes several models: Cosmos Transfer 2.5 and Cosmos Predict 2.5, which are world models used for synthetic data generation and robot policy evaluation; Cosmos Reason 2, a reasoning vision language model (VLM) that processes visual and language data; and Isaac GR00T N1.6, a vision language action (VLA) model purpose-built for humanoid robots.

To support this software ecosystem, Nvidia introduced Isaac Lab-Arena, an open-source simulation framework designed for virtual testing of robotic capabilities. The platform also utilizes OSMO, an open-source command center that serves as the connective infrastructure to integrate workflows from data generation through training. These tools aim to address critical industry challenges by allowing developers to safely test and validate robot behavior in simulated environments before physical deployment.

To power these software models on physical machines, Nvidia introduced the Blackwell-powered Jetson T4000 graphics card. The hardware is designed to provide on-device compute for robots. Its technical specifications include:

• AI Compute: 1200 teraflops of AI compute
• Memory: 64 gigabytes of memory
• Power Efficiency: Operating at 40 to 70 watts

Nvidia is also deepening its integration with Hugging Face, an AI platform and community, to make robot training more accessible. The collaboration connects Nvidia’s 2 million robotics developers with Hugging Face’s 13 million AI builders. This integration comes as robotics has become the fastest growing category on Hugging Face, allowing developers to experiment with robot training without requiring expensive hardware.

By standardizing both the software models and the underlying hardware, Nvidia wants to establish itself as the default platform for generalist robotics, mimicking how Android became the standard operating system for smartphones. This ecosystem strategy is already gaining traction among industry players. Companies currently using Nvidia’s technology to develop and deploy physical AI systems include Boston Dynamics, Caterpillar, Franka Robots, and NEURA Robotics.

Why it matters

Nvidia is positioning itself as the essential infrastructure layer for the robotics industry, aiming to standardize development much like Android did for smartphones.