CoreWeave launches Physical AI templates for robotics and AV development

The templates cover use cases such as robot-arm manipulation, simulated driving scenarios, and reinforcement learning. CoreWeave also integrated skills for coding agents like Claude Code and Codex, allowing them to launch, monitor, and retrieve results from these workflows. The company emphasizes that while the templates handle the execution layer, teams remain responsible for designing meaningful evaluations and determining if results are valid for physical systems.

This move extends CoreWeave’s platform beyond raw GPU capacity into software workflow management. The company named partners including Cadence, Convergent Science, and Voxel51, though it clarified that not all partner products are immediately available as launch templates. CoreWeave did not provide specific benchmarks for setup-time reduction, positioning the launch as a tool for reproducibility and operational efficiency rather than a guaranteed performance gain.