Nvidia expanded its Halos safety architecture from autonomous vehicles to robotics with the June 2026 launch of Halos for Robotics, targeting safe deployment of autonomous mobile robots in warehouses, humanoid robots in factories, and surgical robots. The full-stack system integrates hardware such as the IGX Thor computing module, which dedicates an independent processor to safety-critical workloads, enabling functional and safety systems to operate on the same silicon. CEO Jensen Huang stated in March 2026 that physical AI already generates nearly $10 billion in annual revenue, identifying it as Nvidia's second-most important growth category since 2025. The company has invested directly in humanoid robotics startups Agility Robotics and Figure AI, as well as firms developing robotics foundation models, and partnered with Unitree to provide an open humanoid robot reference design for researchers. Amit Goel, head of robotics ecosystem and edge computing, emphasized that safety represents the next bottleneck as AI models and robot hardware mature, positioning Halos for Robotics to unlock these systems' capabilities.
Read original
arstechnica/ai