Decart

Real-time AI is the next frontier and the infrastructure to run it at scale simply doesn’t exist yet. Every company building for video, robotics, or physical AI hits the same wall: models that are powerful in the lab but too slow, too expensive, and too fragile to deploy in production.

Decart is the missing infrastructure layer. The company’s vertically integrated stack handles everything from sub-30ms world model inference and full-HD video generation at 100fps, to physical AI simulation for robotics and autonomous systems, all running across Nvidia, Google, and Amazon silicon. Where most solutions drop a model into an existing cloud pipeline and call it done, Decart is rebuilding the real-time layer of the AI stack from the ground up.