AWS is betting that reference architectures—not just raw compute—are what robotics teams need next.
Amazon Web Services has introduced an open-source toolchain that combines AWS infrastructure with NVIDIA robotics software to support the development, training, simulation and deployment of physical AI systems.
The Physical AI Toolchain on AWS provides reference architectures, infrastructure-as-code resources and deployment automation for the development lifecycle, from collecting robot data and generating synthetic training scenarios to validating models in simulation and deploying them on physical hardware.
“We built the Physical AI Toolchain on AWS because customers told us that too much of their engineering effort was going to infrastructure instead of innovation,” said Uwem Ukpong, vice president of AWS Industries. “We want to flip that.”
The toolchain is designed to accommodate different robots and tasks. Developers can supply their own robot descriptions, teleoperation data and task definitions, and use individual components or combine them into a workflow spanning the physical AI development lifecycle.
From training to deployment
The toolchain covers synthetic data generation, model training, simulation and validation, edge deployment, and continuous improvement using operational data from deployed machines.
Its software components include NVIDIA Isaac Sim for simulation, Isaac Lab for reinforcement learning, Isaac GR00T for humanoid robot training, and NVIDIA Cosmos for synthetic data generation. NVIDIA OSMO supports workflow orchestration, while AWS services provide infrastructure for training, simulation, storage and edge deployment.
The AWS technical architecture identifies Amazon SageMaker for model training, Amazon EC2 GPU instances for simulation, Amazon S3 for data storage and AWS IoT Greengrass for edge deployment. Amazon’s corporate announcement also identifies Amazon Bedrock AgentCore among the orchestration capabilities.
The toolchain also incorporates development technologies and formats including PyTorch, Hugging Face, Gymnasium, the Robot Operating System 2 (ROS 2), the LeRobot data format, the Unified Robot Description Format (URDF) and the Open Neural Network Exchange (ONNX) format.
AWS provides implementation examples alongside the architecture. One example uses NVIDIA Isaac GR00T and 27 episodes of UR3 robot pick-and-place teleoperation data to demonstrate elements of the development workflow.
Industrial applications
AWS says industrial automation, warehousing and logistics, energy, healthcare, mining, agriculture, aerospace and defense are the industrial sectors where this would have the most impact.
The company describes possible uses including collaborative robot arms that adapt to new assembly tasks, autonomous robots trained to handle different parts, and factory systems that monitor production.
Amazon’s announcement also identifies NEURA Robotics, RLWRLD and Config as companies working on physical AI. NEURA is developing cognitive humanoid robots; RLWRLD is working on foundation models for dexterous manipulation; and Config has developed a data pipeline that captures robot-action data and uses generative AI to produce additional training scenarios.
“In Physical AI, speed is everything: how fast you can fine-tune models, deploy them into the real world and scale from individual systems to large fleets,” said David Reger, founder and CEO of NEURA Robotics. “The Physical AI Toolchain on AWS helps us accelerate exactly that cycle.”
Amit Goel, head of NVIDIA’s Robotics Developer Ecosystem and Edge AI Product, said the toolchain brings together training, simulation and deployment capabilities.
“Building physical AI requires a seamless integration of three computing platforms—training, simulation, and deployment,” Goel said. He added that the open-source toolchain combines AWS services with NVIDIA’s physical AI models, tools and libraries to support an end-to-end development workflow.
AWS has published technical documentation and implementation materials for the toolchain. The technical blog describes prerequisites including an AWS account with approved GPU capacity, an NVIDIA NGC API key for container images and a Hugging Face token for downloading model weights.
Amazon says the toolchain is intended to help manufacturers launch physical AI capabilities in weeks rather than the years it might take to build the infrastructure from scratch. This is the company’s stated expectation, not an independently verified performance result.
