According to aws.amazon.com, Amazon Web Services has introduced support for training robot policies using NVIDIA Isaac Lab on Amazon SageMaker AI, addressing the computational demands of physical AI as it moves from research to production. The announcement, published on June 9, 2026, demonstrates training capabilities for the Unitree H1 humanoid robot.
According to the AWS blog post, robotics teams increasingly train robots in GPU-accelerated simulation rather than real-world environments because physical training is “slow, expensive, and often unsafe,” while simulation can “compress months of learning into hours.” The post notes that reinforcement learning for complex behaviors like humanoid locomotion is compute-intensive, with single-node training runs stretching from hours to days.
The solution provides two compute options through Amazon SageMaker AI: SageMaker HyperPod and SageMaker Training Jobs. According to aws.amazon.com, the service “provisions instances, configures drivers and networking, monitors node health, and tears down resources when jobs finish,” allowing engineering teams to focus on developing robot policies rather than managing infrastructure.
Full implementation code is available in the accompanying GitHub repository, according to the announcement. The blog post indicates this approach supports both short iterative experiments for tuning reward functions and longer production runs that train configurations to convergence.