Inbolt's CEO says factory AI's real bottleneck is the loop between a robot's eyes and its arms

Rudy Cohen runs Paris-based robotics startup Inbolt, which builds camera-based control software that turns a digital twin of a factory cell into live robot motion. Inbolt's control layer is now running in more than 100 factories, the company says, with 40 million-plus production cycles logged on lines at Stellantis, Toyota, and Ford. Each project, Cohen argues, is held up less by model quality than by fixtures, wiring, and the six-figure bill for getting a vision-guided robot cell to hit six-sigma tolerances inside an existing line. That integration cost, he says, is what decides whether a 2026 physical-AI capex bid pencils out, and it is the thesis he is taking to RoboBusiness in Santa Clara on Oct 20.

Inbolt has raised €20 million, or about $23.2 million, and operates across Europe, the U.S., and Japan. The wider question is whether other vendors, integrators, and OEM line managers see the same bottleneck, or whether the data-gating consensus still holds on the factory floor. Cohen is putting that debate on the RoboBusiness stage.