Humanoids Tested in Factories Are Still Slower Than Humans

At Spartanburg, South Carolina, Figure 02 moves between containers of components. The humanoid robot from Figure AI picks them up and then moves them to production stations. BMW says the machine would have helped manufacture more than 30,000 BMW X3 vehicles over ten months. But Ulrich Wieland, the site’s head of logistics, acknowledges that it remains slower than humans.

The appeal of this form lies in its compatibility with the factory as it already exists. Human-shaped arms, hands and legs could make it possible to use the same tools and move through the same spaces, without completely rebuilding production lines. It is a promise of flexibility, not yet proof of profitability.

Manufacturers are multiplying their trials. Mercedes is testing Apptronik’s Apollo in Berlin to transport components, perform logistics tasks and carry out quality checks. Hyundai plans to deploy Atlas, designed by its subsidiary Boston Dynamics, at its US plant in Georgia. Tesla is continuing to develop Optimus for dangerous, repetitive or strenuous work.

The industrial path remains strewn with obstacles, however: handling objects, adapting to unexpected situations and repeating precise movements at a high pace. Ben Armstrong, director of the MIT Industrial Performance Center, told the New York Times that the humanoid could be a “million-dollar solution to a hundred-dollar problem” when a conventional robotic arm would suffice. Unions also fear job losses, despite the promise of reducing strenuous tasks and creating skilled jobs.

In practical terms, humanoids are currently factory testing tools, useful for exploring difficult or repetitive tasks, but not proven replacements for workers. For BMW, Mercedes, Hyundai and Tesla, the next step is less about showing that a robot can walk than establishing that it can handle parts quickly, repeat tasks for long periods and justify its cost.