Nvidia has unveiled new Omniverse SDKs and Cosmos world foundation models, marking a significant leap forward for robotics developers in AI technology.

The Cosmos Reason model enhances robots' reasoning and planning capabilities by leveraging memory and physics understanding, which is particularly beneficial for tasks like data curation and video analytics.

Supporting these advancements, the RTX PRO Blackwell Servers and DGX Cloud infrastructure enable developers to execute complex robotics workloads remotely, facilitating demanding simulations.

In collaboration with Lightwheel, Nvidia is enhancing the policy evaluation process through an open-source framework that allows for large-scale evaluations.

Accessibility improvements have been made with NVIDIA Brev, enabling developers to quickly deploy NVIDIA RTX-enabled GPU instances for seamless access to Isaac Sim.

A highlight of the release is Isaac Sim 5.0.0, a scalable robotics simulation application that supports Neural Rendering and enhanced actuator models, along with synthetic data generation capabilities.

The integration of physical AI combines neural graphics, synthetic data generation, and reinforcement learning, which are crucial for modern robotics, self-driving cars, and smart spaces.

Nvidia's CEO emphasized the transformative impact of AI on computing, noting that it is reinventing data center architecture for the first time in 60 years.

Nvidia's initiatives aim to bolster the development of smart cities and enhance productivity and safety through advanced physical AI solutions in partnership with companies like Accenture and Belden.

The newly introduced Omniverse SDKs facilitate data interoperability across different simulation platforms, allowing over 250,000 developers to effectively simulate robots.

Nvidia has also launched advanced reasoning models, Nemotron Nano 2 and Llama Nemotron Super 1.5, which significantly enhance cognitive capabilities for AI agents in tasks requiring scientific reasoning and coding.

These updates encompass new libraries, models, and development tools designed to create physically accurate virtual environments and AI agents, further pushing the boundaries of robotics technology.