Physical AI Edges Closer to Real-World Deployments

Capgemini's framing positions enterprises as the active agents here, but the more consequential question is whether the operational infrastructure to support physical AI at scale actually exists yet, or whether 'live operations' is doing a lot of work to describe deployments that are still heavily supervised.

This lands at an interesting moment relative to several threads we've been tracking. Physical Intelligence's π0.7 announcement (covered April 16) showed a robot brain capable of novel task generalization, which is exactly the kind of capability gap that has historically kept physical AI in pilots. Google DeepMind's Gemini Robotics-ER 1.6 release (April 13) addressed spatial reasoning limitations that constrained real-world manipulation. Both suggest the model-side blockers are loosening. Meanwhile, the MIT Technology Review piece on treating enterprise AI as an operating layer (April 16) argued that competitive advantage now sits in deployment infrastructure, not model capability. If that framing holds for physical AI, Capgemini's consulting position gives it a structural interest in accelerating exactly this transition narrative.

Watch whether major industrial automation buyers (automotive, logistics) begin announcing multi-site physical AI contracts with named vendors in the next two quarters. Pilot-to-production claims that don't produce reference customers within six months are typically premature.