Nvidia wants to be the Android of generalist robotics 

Nvidia launched a brand new stack of robotic basis fashions, simulation instruments, and edge {hardware} at CES 2026, strikes that sign the corporate’s ambition to turn into the default platform for generalist robotics, a lot as Android grew to become the working system for smartphones. 

Nvidia’s transfer into robotics displays a broader trade shift as AI strikes off the cloud and into machines that may discover ways to suppose within the bodily world, enabled by cheaper sensors, superior simulation, and AI fashions that more and more can generalize throughout duties. 

Nvidia revealed particulars on Monday about its full-stack ecosystem for bodily AI, together with new open basis fashions that permit robots to cause, plan, and adapt throughout many duties and numerous environments, transferring past slender task-specific bots, all of which can be found on Hugging Face. 

These fashions embrace: Cosmos Switch 2.5 and Cosmos Predict 2.5, two world fashions for artificial information technology and robotic coverage analysis in simulation; Cosmos Motive 2, a reasoning imaginative and prescient language mannequin (VLM) that permits AI programs to see, perceive, and act within the bodily world; and Isaac GR00T N1.6, its next-gen imaginative and prescient language motion (VLA) mannequin purpose-built for humanoid robots. GR00T depends on Cosmos Motive as its mind, and it unlocks whole-body management for humanoids to allow them to transfer and deal with objects concurrently. 

Nvidia additionally launched Isaac Lab-Area at CES, an open supply simulation framework hosted on GitHub that serves as one other part of the corporate’s bodily AI platform, enabling protected digital testing of robotic capabilities.

The platform guarantees to handle a vital trade problem: As robots be taught more and more complicated duties, from exact object dealing with to cable set up, validating these skills in bodily environments will be pricey, gradual, and dangerous. Isaac Lab-Area tackles this by consolidating sources, process eventualities, coaching instruments, and established benchmarks like Libero, RoboCasa, and RoboTwin, making a unified normal the place the trade beforehand lacked one.

Supporting the ecosystem is Nvidia OSMO, an open supply command heart that serves as connective infrastructure that integrates your complete workflow from information technology by way of coaching throughout each desktop and cloud environments. 

And to assist energy all of it, there’s the brand new Blackwell-powered Jetson T4000 graphics card, the most recent member of the Thor household. Nvidia is pitching it as an economical on-device compute improve that delivers 1200 teraflops of AI compute and 64 gigabytes of reminiscence whereas working effectively at 40 to 70 watts. 

Nvidia can be deepening its partnership with Hugging Face to let extra individuals experiment with robotic coaching with no need costly {hardware} or specialised information. The collaboration integrates Nvidia’s Isaac and GR00T applied sciences into Hugging Face’s LeRobot framework, connecting Nvidia’s 2 million robotics builders with Hugging Face’s 13 million AI builders. The developer platform’s open supply Reachy 2 humanoid now works immediately with Nvidia’s Jetson Thor chip, letting builders experiment with totally different AI fashions with out being locked into proprietary programs.  

The larger image right here is that Nvidia is attempting to make robotics improvement extra accessible, and it desires to be the underlying {hardware} and software program vendor powering it, very similar to Android is the default for smartphone makers.

There are early indicators that Nvidia’s technique is working. Robotics is the quickest rising class on Hugging Face, with Nvidia’s fashions main downloads. In the meantime robotics corporations, from Boston Dynamics and Caterpillar to Franka Robots and NEURA Robotics, are already utilizing Nvidia’s tech.

Comply with together with all of Trendster’s protection of the annual CES convention right here.