Inbolt CEO: Physical AI's real bottleneck is deployment

Inbolt's argument challenges a prevailing assumption in physical AI that more data is the key to progress. Cohen counters that in factory settings, the constraint is not data but the integration and latency between perception and motion. His company's experience across over 100 factories gives weight to the claim that deployment, not data, is the true bottleneck.

The talk at RoboBusiness will use production data from major automakers like Stellantis, Toyota, and Ford to argue that physical AI is already commercially viable in manufacturing, even if it doesn't get the same attention as larger models. By focusing on servo-frequency control and reducing the need for extensive fixtures and wiring, Inbolt suggests a path to making deployments less costly and time-consuming.

The stakes for the industry hinge on whether the deployment-focused approach can scale beyond early adopters. If Cohen's case resonates, it may shift where companies invest in physical AI—toward integration and real-time control rather than just data collection and simulation. The company's presence in Japan, as part of its global operations, indicates the relevance of this debate for Japanese manufacturers exploring AI-driven robotics.