At CES 2025, Jensen Huang declared it is now the "ChatGPT moment for robotics," as Nvidia intensifies its development of robotics technology to build a comprehensive robotics ecosystem. Nvidia unveiled its generative world model platform, Cosmos, designed to address industry issues related to data shortages.
2025 is expected to be the year of mass production for humanoid robots. However, shortly after Huang showcased 14 robots at CES 2025, an intriguing prediction from SemiAnalysis analyst Dylan Patel sparked renewed discussion. Patel said, "Nvidia will be a robotics company at the end of the day, not just semiconductor."
The path to robotics transformation
Patel pointed out that only a few people fully understand Nvidia's underlying strategies, both in manufacturing and software. This raises important questions: What insights led this veteran analyst to form such an opinion, what is Nvidia's strategy in robotics, and why doesn't Nvidia directly enter the robot manufacturing business?
Recently, Huang visited multiple cities in China, including Beijing, Shanghai, and Shenzhen, attending year-end meetings at various branch offices. Though his public schedule did not show meetings with officials or clients, Chinese media outlet Chinastarmarket revealed that Nvidia's Vice President and General Manager of Robotics and Edge Computing, Deepu Talla, has led the sales rep team in China.
According to the official WeChat account of the National and Local Co-built Embodied Artificial Intelligence Robotics Innovation Center, after Deepu Talla and his team's recent visit, they stated that both parties would deepen cooperation in 2025.
During a two-hour discussion, the Innovation Center and Nvidia explored topics such as world models, data generation, high-precision advanced simulators, whole-body control, and planning-related collaborations.
The National and Local Co-built Embodied Artificial Intelligence Robotics Innovation Center, established in November 2023, is China's first innovation center focused on core technologies, product development, and the application ecosystem for embodied intelligent robots. It was jointly established by leading entities in the robotics industry, such as Ubtech, Beijing Jingcheng Machinery Electric Company, and Xiaomi.
Currently, the Innovation Center is advancing the development of embodied "brains" driven by AI large models for task planning, and data-driven brains for end-to-end skill execution, aiming to create universal embodied intelligent bodies that can handle massive task demands across multiple scenarios.
Strategic developments and partnerships
Before the visit to China, Nvidia had already set up teams for media seminars in its Beijing office to discuss its strategy and plans for robotics.
Talla stated that over the past year, Nvidia has developed Omniverse technology, which has sped up the robot development process. With these advancements, significant differences in robotics technology are expected over the next 5–10 years.
This is why so many companies are now focused on advancing this field. Nvidia has discussed with many companies that are researching robotics-related issues, including the development of a general-purpose robotic "brain" model.
The three pillars of robotic development
When people talk about robots, most think of physical robots like humanoids or industrial robots. Although this physical form is the ultimate goal, the path to achieving it is quite complex.
Nvidia believes that to truly address the challenges in robotics, three types of computers are needed:
Training System Computer
This system is used to train AI models, typically done on the cloud, data centers, or powerful systems like Nvidia DGX. This is a key step in building the robot's "brain."
Simulation System Computer
Once training is complete, testing is required. Nvidia promotes a better solution by introducing a "simulation layer" to conduct tests in virtual environments. Simulations allow thousands of scene tests to be run on a large scale, quickly, and safely, without being constrained by real-world time or cost limitations.
Deployment System Computer
The third system is installed inside the robot, acting as the robot's operational "brain." For Nvidia, this can be achieved with systems like Jetson or AGX.
By integrating these three computer systems, development time can be significantly reduced. With simulations running thousands of tests before real-world deployment, the entire process can be accelerated. This cycle (training, simulation, testing) allows robotics technology to advance within five years, not fifty years.
Article translated by Elaine Chen
