Nvidia, a company known for its graphics cards and AI technologies, is apparently planning to launch a new chip called Jetson Thor next year, as reported by the Financial Times. The chip is specifically designed for humanoid robots and is expected to provide the necessary computing power for complex tasks. With this move, Nvidia could tackle potential revenue losses in the rapidly growing but more competitive AI market. Both Google and Amazon are reportedly working on their own custom AI chips, which would reduce their dependence on Nvidia.
Nvidia's Vice President of Robotics, Deepu Talla, has highlighted two key technologies that are utilized for Jetson Thor — generative AI models and the ability to train robots in a simulated environment. The latter is made possible by the Genesis project, which we previously reported on. This simulator allows robots to be trained in virtually, which generates years of training data in a short amount of time.
Nvidia itself does not plan to enter the robot production business. Instead, the company plans to act as a supplier for other manufacturers and establish Jetson Thor as a platform for the next generation of humanoid robots. Nvidia has already used the "Jetson" name for a series of powerful single-board computers (SBCs), which are designed for AI applications.
These SBCs can control actuators with image recognition data. The latest device in this series is the Jetson Orin Nano Super. It costs less than $250 and can be used for AI applications.