NVIDIA's upcoming GB300 NVL72 "Blackwell Ultra" rack-scale systems are said to be incorporating a humanoid robot assembly, as per sources close to Reuters. In the realm of silicon manufacturing, PCB manufacturing, and server manufacturing, most processes are automated, requiring minimal human involvement. However, up until now, humans were needed for the final assembly of rack-scale systems. Foxconn and NVIDIA are reportedly planning to establish the first AI-powered humanoid robot assembly plant in Houston, Texas. The main idea is that once the plant is finished in the upcoming months, humanoid robots will completely take over the final assembly process, eliminating the need for human workers in the manufacturing process.
This transition is actually beneficial. Since server assembly often involves lifting heavy server racks all day, the humanoid robot system will assist humans by handling the physically demanding tasks, thus reducing the strain on workers. Initially, humans will supervise these robots during their operations, with fully autonomous factories expected in the future. The human role will mainly focus on inspecting the work. NVIDIA has been preparing for humanoid robots for some time, having developed NVIDIA Isaac, a comprehensive CUDA-accelerated platform specifically for humanoid robots. Brands like Agility Robotics, Boston Dynamics, Fourier, Foxlink, Galbot, Mentee Robotics, NEURA Robotics, General Robotics, Skild AI, and XPENG require models that are spatially aware, prompting NVIDIA to create Isaac GR00T N1, the world's first open humanoid robot foundation model, which is accessible for customization and use by anyone.
These models work alongside various sensors and visual inputs that are processed on NVIDIA's Jetson-embedded modules, providing instructions to humanoid robots. The entire system also utilizes NVIDIA's GR00T blueprint for generating synthetic data to fine-tune humanoid robots, as well as the Newton AI physics engine to simulate robot behavior before deployment in manufacturing facilities like this one. Essentially, NVIDIA is constructing its own infrastructure for manufacturing its most advanced AI accelerators.
