NVIDIA Unveils Open Physical AI Data Factory Blueprint

New framework aims to accelerate robotics, vision AI, and autonomous vehicle development

Mar. 16, 2026 at 9:03pm

Got story updates? Submit your updates here. ›

NVIDIA has announced the NVIDIA Physical AI Data Factory Blueprint, an open reference architecture that streamlines the generation, augmentation, and evaluation of training data for physical AI systems. The blueprint enables developers to transform limited training data into large, diverse datasets, including rare edge cases and long-tail scenarios. NVIDIA is collaborating with cloud providers Microsoft Azure and Nebius to integrate the blueprint with their infrastructure and services, empowering developers to scale up their data production. Leading physical AI companies are already using the blueprint to advance their robotics, vision AI, and autonomous vehicle projects.

Why it matters

The success of physical AI systems, such as robotics, vision AI agents, and autonomous vehicles, depends on the availability of massive, high-quality training data. The NVIDIA Physical AI Data Factory Blueprint aims to address this challenge by providing a unified, automated framework for data curation, augmentation, and evaluation, reducing the time, cost, and complexity of building these systems at scale.

The details

The blueprint includes several key components: Cosmos Curator for processing and annotating real-world and synthetic datasets, Cosmos Transfer for exponentially expanding and diversifying the data, and Cosmos Evaluator for automatically scoring, verifying, and filtering the generated data. NVIDIA is also providing the open-source OSMO orchestration framework to unify and manage these workflows across compute environments, reducing manual tasks for developers. Cloud partners Microsoft Azure and Nebius are integrating the blueprint into their respective platforms, enabling developers to leverage accelerated infrastructure and services for their physical AI projects.

  • The NVIDIA Physical AI Data Factory Blueprint is expected to be available on GitHub in April 2026.

The players

NVIDIA

A world leader in AI and accelerated computing, providing the core technologies and framework for the Physical AI Data Factory Blueprint.

Microsoft Azure

A cloud service provider that is integrating the Physical AI Data Factory Blueprint into an open physical AI toolchain, available on GitHub.

Nebius

A cloud service provider that has integrated the OSMO orchestration framework into its AI Cloud, enabling developers to deploy production-ready data pipelines for their physical AI projects.

FieldAI, Hexagon Robotics, Linker Vision, Milestone Systems, RoboForce, Skild AI, Teradyne Robotics, Uber

Leading physical AI developers that are using the NVIDIA Physical AI Data Factory Blueprint to accelerate their robotics, vision AI, and autonomous vehicle development.

Got photos? Submit your photos here. ›

What they’re saying

“Physical AI is the next frontier of the AI revolution, where success depends on the ability to generate massive amounts of data. Together with cloud leaders, we're providing a new kind of agentic engine that transforms compute into the high-quality data required to bring the next generation of autonomous systems and robots to life. In this new era, compute is data.”

— Rev Lebaredian, Vice President of Omniverse and Simulation Technologies at NVIDIA (Globe Newswire)

What’s next

The NVIDIA Physical AI Data Factory Blueprint is expected to be available on GitHub in April 2026, enabling developers to access and integrate the framework into their physical AI projects.

The takeaway

The NVIDIA Physical AI Data Factory Blueprint represents a significant step forward in addressing the data challenges that have hindered the development of advanced physical AI systems, such as robotics, vision AI agents, and autonomous vehicles. By providing a unified, automated framework for data curation, augmentation, and evaluation, NVIDIA and its cloud partners are empowering developers to scale up their physical AI efforts and bring these transformative technologies to market more quickly.