FANUC Brings Physical AI to Factory Motion

At Automate 2026, FANUC is highlighting cobots that adapt welding, fastening and handling to less predictable factory conditions.

FANUC America has outlined its Automate 2026 program with a set of demonstrations focused on industrial robotics driven by physical AI. The clearest example is the CRX-3iA performing vertical-up welding: the lightweight collaborative robot is presented as portable by an operator, then used on a structural steel beam with a motion profile designed to reduce weld puddle sag and reproduce part of the movement and precision of an experienced welder.

The useful point is not that another robotics booth has added the phrase "AI" to its signage. FANUC is showing how large robot makers are trying to push automation into tasks that are less regular and closer to the messy conditions of real factories. The company also lists a CRX-20iA/L cobot tightening bolts on a moving engine block using Inbolt technology and NVIDIA-powered processing. The goal is easy to describe, but hard to industrialize: track a part in motion, adjust the robot’s action in real time and avoid stopping a production line just to make the operation programmable.

Those demos give a practical meaning to physical AI. This is not a chatbot attached to a machine. It is the coupling of perception, simulation, trajectory planning and control. FANUC also points to Isaac Sim for digital-twin visualization of a palletizing cell, plus 3D vision and human-aware tracking for box handling. In each case, the promise is less theatrical than operational: shorten setup time, adapt motion to changing conditions, or make a workstation automatable without building a rigid cell around one fixed part.

For manufacturers, the signal is worth watching. Collaborative robots were first sold mainly as machines that were easier to program and safer near people. The next step is to make them more context-aware. If these systems hold up outside a trade-show demo, they could expand automation into jobs that conventional robotics often avoided, including out-of-position welding, fastening on moving conveyors and variable inspection. The limit is still obvious: a working booth demonstration is not the same thing as shop-floor qualification. But FANUC’s announcement shows that industrial robotics is starting to absorb AI through motion and control, not just through dashboards and planning software.