On August 25th, local time, NVIDIA announced the release of its Jetson Thor robotic computer developer kit and production-grade modules, with the first kits shipping next month. This “robot brain” can run multiple AI workflows and achieve real-time reasoning, enabling robots to interact intelligently with humans and the physical world in real time.
Robots’ body control, object manipulation, and sensory data processing all require powerful computing power to operate in real time. NVIDIA stated that Jetson Thor, based on an NVIDIA Blackwell GPU and equipped with 128GB of memory, delivers 2070 FP4 TFLOPS of AI computing power, a 7.5x increase in AI computing power, and 3.5x improvement in energy efficiency compared to its predecessor, Jetson Orin, released in 2023. It is designed to provide computing power for robots in a wide range of industries, including manufacturing, logistics, healthcare, and retail, as well as the agricultural industry. This performance leap enables robots to process high-speed sensor data and perform visual reasoning at the edge, workflows that were previously too slow to operate in dynamic, real-world environments.
Nvidia founder and CEO Jensen Huang stated that robotics represents Nvidia’s largest growth opportunity outside of artificial intelligence. Jetson Thor enables the construction of robotic systems that interact with and even transform the physical world, running multiple generative AI models simultaneously on edge devices. As a supercomputer, Jetson Thor will usher in the era of physical AI and general-purpose
Ensen Huang has stated that the next wave of AI will be embodied intelligence—intelligent systems that can understand, reason about, and interact with the physical world.
To seize the latest opportunities in robotics, Nvidia has made extensive investments in the field, not only investing in humanoid robot companies but also launching several new robotics chips since 2014. Last March, Nvidia released Project GR00T, a universal base model for humanoid robots. Robots powered by Project GR00T can understand natural language and imitate human behavior by observing it, rapidly learning coordination, agility, and other skills. To help robots better perceive their environment, the company also released the Isaac Perceptor software development kit, which features multi-camera visual odometry, 3D reconstruction and occupancy mapping, and depth perception.
Currently, robotics accounts for only approximately 1% of Nvidia’s overall revenue, but it’s growing rapidly. Nvidia recently integrated its business units, bringing its automotive and robotics divisions into the same segment. In May, the segment reported quarterly sales of $567 million, a 72% year-over-year increase. “We don’t make robots or cars, but we support the entire industry with infrastructure, computers, and related software,” said Deepu Talla, Nvidia’s vice president of robotics and edge AI.
Companies such as United Imaging Healthcare, Wanjie Technology, UBTECH, Galaxy General, Yushu Technology, ZhongQing Robotics, Zhiyuan Robotics, Agility Robotics, Amazon, Meta, and Boston Dynamics are already using Jetson Thor. Yushu Technology founder and CEO Wang Xingxing stated that Jetson Thor represents a significant leap in computing power, enabling robots with greater agility, faster decision-making, and higher levels of autonomy, which are critical for navigating and interacting in the real world.
Nvidia stated that its Jetson Thor can also be used in self-driving cars, particularly for Chinese automakers.
