Humanoid robots in factories: Figure 02 helped produce ~30,000 BMW X3s — what that really means

Key Points

Quick summary

Humanoid robots in factories just logged a meaningful pilot run.

Figure (Figure) reports that its second‑generation humanoid, Figure 02, completed a six‑month pilot at BMW Group (Baoma Jituan 宝马集团) — Spartanburg plant.

The pilot participated in the production of roughly 30,000 BMW X3 vehicles.

During the run, the robots handled more than 90,000 parts, operated for over 1,250 hours, and are estimated to have walked more than 1.2 million steps (~200 miles / ~322 kilometers).

What Figure reported — operational totals

Brett Adcock (Bu Lei Te · A De Ke Ke 布雷特·阿德科克), founder and CEO of Figure (Figure), reshared the company’s update on November 19, 2025.

The company disclosed these totals for the Spartanburg pilot:

• Vehicles involved: ~30,000 BMW X3s
• Parts loaded by Figure 02 robots: >90,000 parts
• Total operating time: >1,250 hours
• Estimated distance walked: >1,200,000 steps (~200 miles / ~322 kilometers)

How Figure 02 was used on the production line

Public images and descriptions show Figure 02 performing pick‑and‑place tasks.

The humanoid removes stamped metal panels from racks or bins.

The robot places panels onto welding fixtures.

Conventional six‑axis industrial robots then perform the welds and transfer assembled parts into the main production line.

Figure 02’s role was handling, positioning and loading upstream of standard robotic welding and downstream flow.

KPIs, targets, and how performance was measured

Figure defined specific KPIs for the pick‑and‑place work to quantify progress.

Target metrics disclosed included:

• Cycle time: total time to complete the sequence (including opening the welding fixture door, loading, etc.) — target < 84 seconds per cycle.
• Positioning accuracy: proportion of shifts in which all three stamped panels were loaded correctly — target > 99% success per shift.
• Intervention count: number of times human operators had to pause or reset the robot — target = 0 interventions per shift.

Known issues: the forearm remains the weak link

Figure acknowledged that Figure 02’s front arm (the forearm) had the highest failure rate among subsystems.

The vulnerability stems from three factors:

• Compact packaging required in that segment.
• Multiple axes of flexibility needed (three degrees of freedom).
• Thermal limits associated with that design area.

In short, packaging density, required mobility, and heat management make the forearm a particularly challenging subassembly.

The article notes that dexterous hands and forearms are among the toughest engineering problems in humanoid robotics.

Tesla (Tesla Tesla 特斯拉) reported similar issues with Optimus prototypes — cooling for motors and shortened component lifetimes were reported problems.

Elon Musk has repeatedly said that building hands and forearms for humanoid robots is an especially difficult engineering challenge because many actuators and control complexities that emulate human hand function live in the forearm.

Retirement of Figure 02 and the shift to Figure 03

Figure publicly unveiled Figure 03 and has begun retiring Figure 02 units.

The company recalled deployed Figure 02 machines from BMW’s headquarters.

Figure says the operational experience from Figure 02 will inform manufacturing processes, component architecture and mechanical redesigns for Figure 03.

Specifically, Figure says Figure 03 re‑architects electronics in the forearm and wrist.

The company removed a distribution board and dynamic cabling routes in that area to reduce complexity, improve reliability, and simplify thermal management.

Industry context — fragmentation and convergence

The pilot highlights that the humanoid robotics industry has not converged on standard designs or scene definitions.

Orient Securities (Dongfang Zhengquan 东方证券) commented the sector still lacks consensus on fundamentals like locomotion (biped vs. wheeled), mechanical architecture of dexterous hands, necessity of linear joints, component choices, and robot sizing.

That fragmentation creates many SKUs (stock keeping units).

Multiple SKUs can dilute industry capital and engineering focus and slow the emergence of embodied intelligence at scale.

The brokerage also noted sentiment for robot mass production has cooled and the robotics sector has pulled back recently.

Orient Securities expects clearer signals of small‑volume production and initial V3 deployments in early 2026.

The first quarter of 2026 could be a key signal if V3 units begin limited production.

What this means for investors, founders, and operators

If you build, invest in, or deploy industrial robotics, this pilot provides a handful of practical implications:

• Field data matters. Real production runs expose subsystem weaknesses — especially in compact, thermally constrained components like forearms and wrists.
• Iterate fast, but expect recalls. Retiring Figure 02 in favor of Figure 03 is a normal hardware cycle: deploy → learn → redesign → redeploy.
• Standardization unlocks scale. Lack of scene and design consensus can slow capital efficiency; convergence will likely be a precursor to broader mass production.
• Robots complement, not replace, existing automation. In this pilot, humanoids handled upstream pick‑and‑place while conventional six‑axis robots handled welding — a hybrid flow is realistic today.
• KPIs are critical for deployment decisions. Cycle time, accuracy and interventions are practical metrics for evaluating ROI and operational readiness.

Takeaway

The BMW pilot proves humanoid robots can contribute to complex automotive lines for repetitive handling tasks.

Hardware durability — particularly in forearms and hands — plus thermal and reliability engineering remain central obstacles.

Figure’s retirement of Figure 02 and redesign toward Figure 03 follows a classic hardware product‑iteration pattern: field data → targeted redesign → redeployment.

Industry consensus and standardization will be important before broad, high‑volume adoption can scale.

Humanoid robots in factories are making practical progress, but the path to mass adoption still demands engineering, standardization, and realistic deployment KPIs.

References

• 机器人“进厂”先头部队绩效如何？Figure披露：已参与3万辆宝马汽车生产 – 东方财富网
• 财联社 – Cailian Press
• Figure.ai – Figure