Humanoid robots in factoriesBMW humanoid robot

BMW Humanoid Robots Verified Pilots, Tasks and Results

BMW humanoid robot projects at Spartanburg, including Figure 02 production work and the Figure 03 sequencing trial, with published results and missing operating data.

By TechniaHQRobot8 min read

Introduction

BMW and Figure have published two distinct humanoid programs at Plant Spartanburg. BMW reports that Figure 02 ran for about 1,250 operating hours over a ten-month period, moved more than 90,000 sheet-metal parts and assisted production of more than 30,000 BMW X3 vehicles.

Figure 03 arrived at Spartanburg in June 2026 for a different logistics workflow: sequencing parts from variable positions while stepping, repositioning and pulling a cart. BMW also published the Figure 03 project through its PressClub. The Figure 02 operating totals should not be attributed to Figure 03, and neither company has published a full audited cost model or complete intervention history.

Key findings

  • Figure 02 handled sheet-metal loading at BMW Group Plant Spartanburg and Figure reports more than 1,250 runtime hours and 90,000 parts loaded.
  • Figure 03 is documented at the same plant in a sequencing workflow that combines picking, placement, stepping and cart manipulation.
  • The published Figure 02 totals are company-reported and do not include a complete intervention or cost breakdown.
  • Results from one structured automotive workflow should not be generalized to other parts, lines or plants.
  • Useful follow-up metrics include task cycle time, intervention minutes, failed grasps, recovery rate, maintenance hours and productive runtime.

BMW Humanoid Robots Verified Pilots, Tasks and Results - evidence comparison

The table records what each source establishes and keeps missing data visible.

System or methodWhat the evidence establishesEvidence classMain unresolved point
Figure 02 at SpartanburgFigure and BMW documented a production trial handling sheet-metal parts in BMW's South Carolina plant.Officially documented industrial pilotBMW and Figure do not publish a complete audited cost model.
Figure 03 sequencing at SpartanburgFigure and BMW document Figure 03 performing a sequencing workflow at Plant Spartanburg in June 2026. Figure describes whole-body manipulation of parts and a wheeled cart.Officially documented factory demonstration and projectFleet size, trial count, intervention rate and long-duration reliability are not publicly disclosed.
Other BMW testsBMW has evaluated multiple robotics approaches, but each supplier and site must be verified separately.Project-specific evidence requiredResults from one plant and task should not be generalized to all automotive work.

Evidence classification

  • Officially documented specifications, standards or project status stated by the responsible organization.
  • Real-system evidence demonstrations or deployments performed on physical hardware under described conditions.
  • Company claim a numerical or operational statement reported by the company and not independently audited.
  • Simulation or research evidence useful for mechanisms, but not proof of field deployment.
  • Insufficient public evidence control mode, trial count, version or operating conditions are missing.

Figure 02 production work

BMW describes a ten-month Figure 02 deployment at Plant Spartanburg. The robot retrieved and positioned sheet-metal parts for welding, operating five days a week for ten hours per shift. BMW reports more than 90,000 components moved, about 1,250 operating hours and roughly 1.2 million steps.

Those figures document repeated physical work, but they do not disclose failed grasps, intervention minutes, maintenance hours or a full cost per handled part. The numbers remain manufacturer-reported.

Figure 03 sequencing trial

In June 2026, Figure and BMW published a new Figure 03 project in Spartanburg. The task is sequencing: selecting thin-walled parts from variable orientations, placing them into the required order and moving a wheeled cart.

Figure attributes the whole-body behavior to Helix 02, with visual feedback coordinating the hands, arms, torso and feet. Public material shows the workflow, but it does not publish fleet size, trial count, intervention rate or full-shift reliability.

How to compare the two programs

Figure 02 has published operating totals from a production deployment. Figure 03 has a newer sequencing demonstration and project description. Combining those records would overstate the evidence for Figure 03.

For either generation, compare task cycle time, productive runtime, human intervention, recovery after failed grasps, charging or battery handling, maintenance time and the number of parts accepted by the production process.

Factory conditions that can change the result

Part orientation, bin position, fixture tolerances and human traffic can change perception and grasping difficulty. Walking or cart movement can also add time that is absent from a stationary manipulation test.

A workflow validated on one part family does not establish performance on another line. BMW and Figure have not published a plant-wide humanoid reliability benchmark.

Data still missing for a cost comparison

Public sources do not provide a complete audited cost model for the BMW programs. Robot purchase or service cost, integration labor, maintenance, spare parts, operator supervision and downtime are needed before calculating cost per accepted part.

A future deployment report would be easier to compare if it published fleet size, observation period, intervention minutes per operating hour and task-level failure categories.

Limitations and missing information

  • BMW and Figure do not publish a complete audited cost model.
  • Exact fleet size and daily staffing are not fully disclosed.
  • Results from one plant and task should not be generalized to all automotive work.
  • Specifications, prices, repositories and deployment status can change after publication.
  • Benchmarks from different robots or environments are not directly comparable.

Conclusion

BMW's published humanoid record now includes two different Figure generations at Spartanburg. Figure 02 has company-reported operating totals from sheet-metal loading. Figure 03 has a documented sequencing workflow with variable part positions and whole-body movement. Cost, intervention and long-duration reliability data remain incomplete, so the two projects should be compared on their published task boundaries rather than merged into one performance claim.

Frequently asked questions

What does BMW humanoid robot mean?

A BMW humanoid-robot project is a test, pilot or production activity conducted at a BMW Group site with a named robot supplier. A signing ceremony, lab video or future plan is not a deployment. Evidence must identify the site, task, date and current status.

How should BMW's humanoid projects be evaluated?

Record the robot generation, task, site, observation period, cycle time, intervention time, failed grasps, recovery behavior, productive runtime and maintenance. Keep Figure 02 operating totals separate from Figure 03 sequencing evidence.

What real-world evidence is available?

Figure reports more than 1,250 hours and 90,000 parts loaded by Figure 02 at Spartanburg. BMW and Figure also document Figure 03 performing a sequencing workflow at the same plant in June 2026. Public sources do not provide a complete intervention history for either program.

Sources and methodology

Sources were rechecked on September 25, 2026 using Figure's production-at-BMW report, Figure's June 2026 Figure 03 project page and BMW Group material. Company operating totals remain attributed to the publisher, and values absent from those records remain undisclosed.

Source review notes

Review date recorded July 11, 2026

Supported by the cited sources

  • Figure and BMW documented a production trial handling sheet-metal parts in BMW's South Carolina plant.
  • Figure reports larger-scale production use and published hours, parts and vehicle contribution figures; these remain company-reported.

Not confirmed or incomplete

  • BMW and Figure do not publish a complete audited cost model.
  • Exact fleet size and daily staffing are not fully disclosed.
  • Results from one plant and task should not be generalized to all automotive work.

Likely to change quickly

  • Commercial availability, prices, model versions and software access.
  • Deployment counts, company partnerships and repository maintenance status.

Share this article

Share the current TechniaHQRobot article page.

Continue reading

Open the latest robotics reporting, Physical AI analysis and hardware notes.

Browse robotics news
Article by @techniahqrobot