NVIDIA Unveils Physical AI Data Factory Blueprint at GTC 2026

Developing robots and autonomous vehicles requires vast amounts of training data spanning different environments, actions and risk scenarios, while collecting such data in the real world is costly and slow. NVIDIA has therefore integrated generative AI, simulation platforms and computing infrastructure into a data factory designed to lower the data barrier for Physical AI developers and shorten model training and validation cycles.

NVIDIA unveiled its Physical AI Data Factory Blueprint at GTC in March 2026. It uses Cosmos technology to automatically generate high-quality synthetic data and incorporates the open-source OSMO orchestration framework, connecting with cloud services including Microsoft Azure. The blueprint targets large-scale robot and autonomous-vehicle development. NVIDIA has not disclosed partnership amounts or formal commercial pricing.