Figure AI has formally retired its second-generation humanoid robot, the F.02, closing out one of the most closely watched real-world deployments in the young humanoid industry. The California company pulled the machines from BMW Manufacturing's Spartanburg, South Carolina plant after roughly 11 months on an active assembly line — and used the send-off to counter long-running skepticism about how much real work the robots were actually doing.
A Deployment Measured in Cars, Not Demos
The headline numbers are substantial for a technology many still consider experimental. According to Figure, the F.02 units helped produce more than 30,000 BMW X3 vehicles and loaded over 90,000 sheet-metal parts during their time on the line. Each robot logged more than 1,250 hours of work and covered roughly 200 miles walking the plant floor.
Their assignment was narrow but unforgiving: lift sheet-metal parts from bins and place them onto welding fixtures within a 5-millimeter tolerance, after which conventional robotic arms performed the welds. Figure says the humanoids handled the loading with a cycle time of about 84 seconds and maintained placement accuracy above 99 percent — the kind of consistency an automotive line demands, shift after shift.
The rollout was staged. Within six months of bringing up the F.02, Figure had delivered robots to the plant and begun testing; by month ten they were in full deployment, running a standard schedule of roughly 10 hours a day, five days a week.
The "Battle-Scarred" Reveal
Rather than quietly decommission the fleet, CEO Brett Adcock turned the retirement into a statement. He shared images and video of the robots covered in scratches, scuffs, and grime accumulated over months on an industrial line — visible wear that Figure presented as proof of genuine, sustained operation.
The framing was deliberate. Earlier in the year, reports had characterized Figure's presence at BMW as a small-scale feasibility study rather than a functional integration, and Fortune had previously questioned whether the company overstated the robots' role. By foregrounding the physical "scars" of assembly-line duty, Figure is arguing that the deployment was exactly what it claimed: real-world work, not a staged pilot.
Why Retire Them Now
The move is tied to a generational upgrade rather than a retreat. With Figure 03 now released, the company is beginning a fleet-wide retirement of the F.02, bringing the units back to headquarters. The data gathered at Spartanburg — on reliability, wear, cycle timing, and the countless small failures that only surface in a live factory — is being folded directly into Figure 03's operational readiness, which is engineered for larger-scale production.
That transition mirrors a broader inflection in the humanoid sector. Recent industry tallies point to real, if early, scaling:
- Figure 03 has now passed 1,000 units produced.
- Boston Dynamics' electric Atlas is shipping initial units to partners including Hyundai and Google DeepMind.
- Chinese entrants such as AgiBot and, most recently, BYD are pushing into embodied AI, intensifying competition with Tesla's Optimus program.
Why It Matters
For years, humanoid robotics has been rich in flashy demonstrations and thin on evidence that the machines can hold down a productive job. Figure's Spartanburg run is one of the first deployments to offer concrete, factory-grade metrics — hours worked, parts loaded, cars built, accuracy sustained — rather than a highlight reel.
That distinction matters for the entire field. Automakers evaluating humanoids need to know whether a robot can survive the grind of a real line: the dust, the repetition, the incidents, the 10-hour shifts. A retirement announcement built around visible wear and hard production figures is, in effect, a durability report — and a more persuasive sales pitch than any polished demo.
It also underscores why automotive plants have become the proving ground for embodied AI. The task overlap between building cars and building robots — in perception, actuation, and computing — is high, and manufacturers already run the structured, repetitive environments where today's humanoids perform best. If Figure 03 can carry the lessons of the F.02 into higher-volume deployment, the industry's long-promised leap from pilot to payroll may finally be underway.
