The thing missing from every humanoid-robot story for the last four years has been the boring part. Highlight reels of robots folding shirts and unloading dishwashers, sure. Real operating data from an actual customer running them on an actual production line, not so much. That changed quietly this spring when Figure and BMW published the eleven-month numbers from the Spartanburg pilot, and the numbers do not require any squinting.
The deployment: Figure 02 units running ten-hour shifts, Monday through Friday, on BMW’s X3 sheet-metal line at Plant Spartanburg, which is BMW’s single largest factory in the world by volume. Eleven months. 1,250 hours of cumulative runtime. More than 90,000 individual parts loaded. Contribution credit on more than 30,000 finished X3 vehicles. Accuracy reported at above 99 percent on the specific tasks the robots were assigned. The pilot officially “retired” the Figure 02 units in late 2025 with what one trade press writeup memorably described as the robots looking “bruised” from the shift count, which is its own kind of milestone. Tools wear out. That is what tools do.
BMW is now expanding the program two ways. First, transitioning the Spartanburg deployment to Figure 03, which is the next-gen unit: 16 degrees of freedom in the hands instead of 12, 25 percent more payload, longer battery runtime, faster on-board inference. Second, opening a second deployment at Plant Leipzig in Germany. The unit count is increasing and the scope of tasks the robots are being asked to do is widening. None of this is a moonshot announcement. It is a customer placing a bigger order because the previous order paid out.
The reason this matters more than the demo videos is that humanoid robotics has spent two years in a credibility valley where the press releases are spectacular and the field deployments are mostly invisible. Tesla’s Optimus, the most-covered humanoid in the world by a factor of ten, had Elon Musk himself conceding in April that the units “are not in usage in our factories in a material way.” The contrast with Figure has now become a number contrast and not a vibes contrast. One company shipped a robot to a paying customer who ran it long enough to wear it out and then ordered the next version. Everyone else is still cutting promotional video.
What this does not prove yet: that the unit economics work outside an automotive line with a single repetitive task, that the same robots can do the dexterous work the marketing decks promise, or that the BMW deal isn’t being subsidized by Figure for strategic reasons. But “humanoid robots have started doing real work somewhere” is a phase change from where the category was a year ago, and the burden of proof on everyone else just got a lot heavier.