Last night, Jensen Huang gave a gift to Chinese robotics companies.

Last night, Jensen Huang gave a gift to all robots.

It turned out that NVIDIA released a new robot computing platform called Jetson Thor, which is known in the industry as the "robot brain". Its biggest highlight is that it packs computing power close to that of a server directly into the robot's body: based on the latest Blackwell GPU architecture, the AI computing power soars to over 2000 TFLOPS, a full 7.5 times higher than the previous generation. It is equipped with 128GB of large memory, enabling the performance of edge computing devices to reach an unprecedented level.

This means that robots are no longer just executing simple instructions, but can perceive the environment, analyze information, and make quick decisions like intelligent agents with a "brain". Whether in factories, warehousing and logistics, or service and medical scenarios, Jetson Thor enables robots to perceive, judge, and act autonomously. NVIDIA emphasizes that it can run multiple AI models simultaneously, providing developers with a highly free and flexible development environment.

The US price of the Jetson AGX Thor developer kit is $3499 (equivalent to about 25,000 RMB). Many robot companies around the world have already adopted it. Among them, Chinese enterprises account for a larger proportion, such as Ubtech, Galaxy Universal, Zhongqing, and Zhipu Robotics.

Most current robots lack a "brain"

Jetson Thor is equipped with the latest Blackwell GPU architecture and has up to 128GB of memory. It can process multi-channel data from cameras, lidars, tactile sensors, etc. simultaneously, achieving high-speed, multi-modal perception and real-time decision-making.

Put simply, this means that robots can perceive the environment, judge the situation, and take actions like humans, rather than just mechanically executing instructions. NVIDIA CEO Jensen Huang pointed out: "Jetson Thor helps millions of developers around the world build robot systems that can interact with and even change the physical world. With unparalleled performance and energy efficiency, it can also run multiple AI models simultaneously on edge devices. As an excellent supercomputer, Jetson Thor is driving the real arrival of the era of physical AI and general-purpose robots."

Jetson Thor not only has powerful hardware but also a very complete software ecosystem. By combining with the Isaac GR00T humanoid robot model and the Isaac Sim simulation platform, developers can first let the robot practice skills in a virtual environment and then deploy the optimized model to real devices. Just like humans practice skills in simulated scenarios first, robots can make mistakes and learn in the "virtual world", which is not only safe but also significantly shortens the R & D cycle.

The Omniverse platform enables real-time linkage between the virtual environment and the real scenario, allowing robots to adapt to complex scenarios and changing environments in advance. In logistics and warehousing, it can handle autonomous handling and sorting; in the medical field, it can perform inspections, deliver drugs, and even assist in rehabilitation training; in industrial production, it can complete precise operations and collaborations. Domestic and foreign enterprises such as Unitree Robotics, Boston Dynamics, and Agility Robotics have begun to integrate Jetson Thor into their robot products to verify its performance in real environments.

Wang Xingxing, the founder of Unitree Robotics, said: "Jetson Thor brings a huge leap in computing power, enabling robots to achieve a higher level of autonomy and faster decision-making in the real world, and significantly improving their navigation and interaction capabilities." Wang He, the CTO of Galaxy Universal, also revealed that after adopting Thor in their G1 Premium robot, both the movement speed and the smoothness of actions have been significantly improved.

Robots are getting closer to us

In fact, due to the lack of investment in the "brain", most current robot products can only perform pre - set actions, and the success rate is not 100%. Wang Xingxing, the CEO of Unitree Robotics, also said in a recent speech that the reason why robots are not widely used at present is not because the hardware is not good enough. "The current robot hardware, whether it's the whole machine or the dexterous hand, is completely sufficient," he said. He also pointed out that the biggest problem now is that the development of embodied intelligence lags far behind, which is the key factor limiting the large - scale application of current robots, especially humanoid robots. In other words, the robot field lacks an intelligent brain that can control the body's actions.

Looking globally, there are not many enterprises that can afford to invest in building the "brain", and NVIDIA is one of them.

The release of Jetson Thor is therefore highly anticipated because technological breakthroughs by leading enterprises often determine whether the robot industry can enter a new stage of development. With the continuous improvement of AI computing power and the improvement of the software ecosystem, robots will not only be an extension of industrial machinery but real intelligent agents with autonomous learning and decision - making abilities.

In the future, robots can replace humans in factories to perform repetitive labor and achieve all - day production; in the logistics field, they can handle autonomous handling, sorting, and even manage the warehousing system; at home, they can do cleaning, take care of the elderly or children; in the medical field, they can assist in complex surgeries and rehabilitation training, and even provide psychological companionship and educational tutoring. As the hardware cost decreases and the technical threshold drops, these scenarios are getting closer to us.

Of course, Jensen Huang is never a philanthropist. Nowadays, global robot enterprises are closely tied to NVIDIA's chips. "If you want to do well in the robot field, you have to cooperate with NVIDIA," an embodied intelligence investor told Phoenix Tech.

Through Jetson Thor, NVIDIA is not just deploying a product but an ecological system for the era of intelligent robots. From hardware computing power to software platforms, and then to real - world scenario implementation, if robots truly move from science fiction to reality and gradually become part of our daily lives, the demand for computing power will support more valuation myths for NVIDIA.

This article is from the WeChat official account "Phoenix Tech", author: Jiang Fan. Republished by 36Kr with permission.