Mitsubishi Electric develops a CNC machining system with humanoid robots

Mitsubishi Electric said on Oct 21 that it developed what it calls the industry's first machining system in which humanoid robots and CNC machine tools collaborate through physical AI. The system lets humanoids take on work normally done by machine-tool operators — setup, cleaning, removal and transport of finished products, inspection and troubleshooting — in a bid to cut labour at machining sites. A demonstration is scheduled at the Japan International Machine Tool Fair, JIMTOF 2026, at Tokyo Big Sight from Oct 26.

The problem the company chose to solve is integration. Connecting a humanoid to a CNC tool has meant writing dedicated communication functions and control sequences for each robot, which makes every deployment a custom project. Mitsubishi Electric's version has three parts: an AI machining task planner that takes an operator's instruction in natural language, generates tasks and divides the work between the machine tool and the robot; a CNC machine tool carrying physical AI that exchanges operational information in natural language; and a general-purpose humanoid. The planner and the machine-side AI are built on the company's Maisart AI technology.

The robot is the commodity part

Everything specific to the system sits in the machine tool and the planner, which is why Mitsubishi Electric describes the humanoid as general-purpose. It has not named a robot supplier, and the short list of humanoid makers with a machine-tending product in 2026 suggests it will buy rather than build. The competition for this kind of work is the labour it replaces: setting and cleaning inside a machine enclosure is dirty, repetitive work that factories already automate with gantry loaders and bar feeders, at a cost that depends on volume and part variety.

The retrofit argument

Mitsubishi Electric's stated case is that conventional industrial robots need substantial capital investment to modify production lines, while a humanoid that fits existing spaces does not. That argument holds best on low-volume, high-variety machines where a fixed loader is hard to justify, and least on the long-run parts that are already automated.

Machining sites are one of several fronts in Japan's physical-AI push: Kawasaki Heavy has said it wants a fully autonomous humanoid by as soon as 2030, and Mitsubishi Electric's own framing is a step toward unmanned machining. The JIMTOF demonstration will show whether the natural-language split between operator, machine and robot works on a real machining task rather than a scripted one.