Humanoids won’t scale on factory floors until costs drop

Following a successful pilot with Figure 02 at its plant in Spartanburg, S.C., BMW said it will deploy the new Figure 03 humanoid. | Source: BMW Group

While the tech industry thinks AI is the main barrier for robots joining the manufacturing workforce, the real issue has less to do with intelligence and more to do with robotic unit economics and financial viability.

AI has come a long way, vision systems have matured, foundation models keep getting better and the hardware has significantly improved. But for most manufacturers, the cost of automating using humanoids at today’s prices just isn’t feasible.

Physical AI has evolved quickly and robots now handle manipulation tasks that seemed out of reach only a few years ago. Several vendors have shown machines running real warehouse operations. The hardware has advanced with stronger actuators, better mobility and effectors that keep improving. Even the hands, which have been a sticking point for years, are close enough that plenty of use cases work without them being perfect.

There are some organizations that have had real robotics success. Generalist ’s robots can perform precise manipulation tasks. Agility Robotics ’ humanoids have achieved proven results in the warehouse. The robot hands built by Alt-Bionics have improved both humanoid robots’ dexterity and broken new ground in prosthetics. All of these wins are paving the way for further innovation and accessibility in the space.

Many people in robotics will tell you the core technology problems have either been solved or are nearly there. The demos are becoming more impressive and the conversation has moved from “Can they do the work?” to “Can we justify the cost?”

Despite technological progress, human oversight will be key to filling in what gaps remain, especially when a robot encounters something it didn’t expect.

Some developers lean hard on teleoperation, where a person controls the robot’s movements directly. That requires almost all of the operator’s attention, which caps how far it can scale because every robot needs a human tied to it.

A better model is to treat people as high-level supervisors who step in only when a robot faces something unusual or needs direction on a task. This way, one person can supervise 50 or more robots at once, and every intervention serves as training data to improve the AI over time. Over time, that involvement may drop, but it never hits zero.

The promise of humanoids is their flexibility. In theory, one machine can do many different jobs without much reconfiguration. But that flexibility comes with a price. A factory automation system built for one workflow is optimized for it from end to end. In high-volume production, that specialization beats general-purpose humanoids.