NVIDIA wants to become the "Android" of "physical AI."

```

Nvidia is making a full effort to build the default platform in the robotics field, aiming to replicate Android's dominant position in smartphone operating systems.

On January 5, at CES 2026, Nvidia released multiple open-source foundation models, including several open-source base models that enable robots to reason, plan, and adapt across various tasks and environments, with all models made available on the Hugging Face platform.

At the same time, Nvidia launched the next-generation Jetson T4000 graphics card based on the Blackwell architecture, as well as an open-source command center called OSMO, to support the entire robotics development workflow. The company also deepened its partnership with Hugging Face, aiming to lower hardware barriers and technical thresholds for robot training.

This strategy reflects the industry trend of artificial intelligence migrating from the cloud to the physical world. As sensor costs decrease, simulation technology improves, and AI models gain more generalization capability, robots are evolving from performing single tasks to general-purpose roles. Companies like Boston Dynamics and Caterpillar have already begun using Nvidia technology, and the robotics category has become the fastest-growing area on the Hugging Face platform.

Building a Complete Model Matrix

The foundation models released by Nvidia this time constitute the core capability layer of physical AI.

Cosmos Transfer 2.5 and Cosmos Predict 2.5 are world models responsible for synthetic data generation and robot policy evaluation, allowing robotic behaviors to be validated in simulated environments.

Cosmos Reason 2, as an inference visual-language model, gives AI systems the ability to observe, understand, and act in the physical world.

Isaac GR00T N1.6 is a visual-language action model developed specifically for humanoid robots, using Cosmos Reason as the inference core, achieving full-body control functions and allowing humanoid robots to perform locomotion and object manipulation simultaneously.

The Isaac Lab-Arena launched by Nvidia at CES is an open-source simulation framework hosted on GitHub, aimed at addressing industry pain points in validating robotic capabilities.

As robots learn to handle precise object manipulation, cable installation, and other complex tasks, validating these abilities in physical environments tends to be costly, time-consuming, and risky.

This platform integrates resources, task scenarios, training tools, and existing benchmarks such as Libero, RoboCasa, and RoboTwin, providing a universal framework for an industry that previously lacked unified standards. The supporting open-source platform OSMO serves as a command center, integrating the entire workflow from data generation to training and supporting both desktop and cloud environments.

Lowering Hardware Barriers

The new member of the Thor series, the Jetson T4000 graphics card, is equipped with the Blackwell architecture. As a cost-effective device-side computing upgrade solution, it provides 12 trillion floating-point AI operations and 64GB of memory, with power consumption controlled between 40 and 70 watts.

Nvidia has also deepened its partnership with Hugging Face, integrating Isaac and GR00T technologies into Hugging Face's LeRobot framework, connecting Nvidia’s 2 million robotics developers with Hugging Face’s 13 million AI builders.

The open-source humanoid robot Reachy 2 now directly supports Nvidia Jetson Thor chips, enabling developers to test different AI models without being locked into proprietary systems.

Early indications show that Nvidia’s strategy is having an effect. Robotics has become the fastest-growing category on the Hugging Face platform, and Nvidia’s models are leading in download volume. Companies such as Boston Dynamics, Caterpillar, Franka Robots, and NEURA Robotics are already using Nvidia technology.

This strategy reflects the company’s intention to make robotics development more accessible, while positioning itself as a foundational hardware and software provider—similar to the role Android plays for smartphone manufacturers.

As AI moves from the cloud to machines with the ability to learn in the physical world, cheaper sensors, advanced simulation technologies, and AI models capable of generalizing across tasks are driving an overall industry transformation.

Risk Warning and DisclaimerThe market has risks; investment requires caution. This article does not constitute individual investment advice, nor does it take into account the special investment objectives, financial situation, or needs of any particular user. Users should consider whether any opinions, views, or conclusions in this article are suitable to their specific situation. Investments made based on this article are at your own risk. ```