EXCLUSIVE: NVIDIA's NVentures backs Saturn Dynamics' $8M bid to run world models at the edge
The founders say Atlas can simulate physical environments in real time while using a fraction of the compute required by larger video models.
By RuntimeWire Staff · Published
Scoop: RuntimeWire original reporting.
Why it matters
World models can become a costly bottleneck as robot makers generate more training and validation data. Saturn Dynamics is betting that edge-speed simulation, rather than ever-larger models, will determine how broadly autonomy can be deployed.
Saturn Dynamics, founded by Elisa Seghetti, Simone Totaro and Francesc "Cesc" Cunillera, has raised an $8 million seed round to build computationally efficient world models for robots and autonomous machines, according to people familiar with the financing. NVentures, NVIDIA's venture capital arm, participated in the round.
The investment puts NVIDIA behind a startup whose central pitch is that physical AI will need much cheaper simulation. Saturn Dynamics says its flagship Atlas World Model can run in real time on edge hardware, allowing robots and vehicles to predict how a scene will change without repeatedly sending data to a large cloud model. Its website claims Saturn's models are "1000x more computationally efficient," a figure that remains a company assertion. (nvidia.com)