As AI moves from screens into cars, robots, drones, factories and defence systems, the data problem is changing. The next generation of physical AI may depend less on collecting more data, and more on generating precisely the data that does not yet exist. At DiffuseDrive, Daniel Bin Schmid, Domonkos Haffner and Gabor Vecsei are working on different parts of that problem, from controllable generative models to production-grade, air-gapped AI infrastructure.