Nvidia is not building its own robot for every warehouse or home. It wants to supply the chips and software that help other manufacturers’ machines see, learn and move. Much as Android provides a foundation for phones from many brands.
An expansion of its Isaac platform brings those components closer together. Developers get tools to let robots practise digitally before testing their skills in a real machine.
Practise on a computer first
Teaching a robot to pick up a box takes plenty of repetition. Isaac Sim and Isaac Lab let developers practise in a simulated environment. GR00T supplies AI models for robot skills. A compact Jetson Thor computer then runs that AI inside the robot.
The Agile ONE SN from Agile Robots brings several components together. It has cameras, a laser sensor and hands that sense touch. It is intended for research and is due to become available in early 2027.
Earning money across robot brands
That is where the Android comparison comes in. Manufacturers build their own machines, while Nvidia supplies a shared technical foundation. It does not need to build the bestselling robot itself to benefit from growth in robotics.
For cars, Nvidia offers DRIVE and Hyperion: computers, software and a reference sensor setup. It is the same approach, but a different package. A manufacturer buying Nvidia chips is not automatically using all its software.
A robot is more complicated than a phone
Robots have different hands, joints and tasks. What works on one machine may go wrong on another. Nvidia provides a starting point; manufacturers still have to make their own robots work reliably.
The same issue arises with Chinese robotaxis, more powerful humanoid robots and autonomous delivery systems. The machine catches the eye, but software determines its capabilities. That is where Nvidia wants to become indispensable. Whether it becomes the standard is still undecided.
Unitree, Sharpa and Nvidia in one research robot
Another example is the open reference design announced in May 2026. Unitree supplies the body, Sharpa the hands, and Nvidia the Jetson Thor computer and software. Researchers can build on a combination whose components have already been matched to one another. The presentation below shows that design.
