Humanoid Robotics
BMW Moves Figure 03 Into Spartanburg Logistics After 30,000-Vehicle Pilot
BMW is taking Figure 03 into a harder Plant Spartanburg logistics task: sorting unsorted components into sequence trolleys after Figure 02 supported 30,000-plus X3 vehicles.
BMW Group says Figure 03 will begin work at Plant Spartanburg on logistics sequencing, following a Figure 02 pilot that supported production of more than 30,000 BMW X3 vehicles over roughly 10 months.
The new use case is more revealing than another robot-walking video. Figure 03 is being moved from repeatable body-shop insertion work into the messy middle of automotive logistics, where parts arrive unsorted, must be staged in sequence, and then move to the assembly line through tugger trains or BMW's Smart Transport Robot system.
AI-generated image
The real story is logistics structure: bins, sequencing trolleys, sensors, and repeatable material flows. Source: Biped.News AI editorial image.
Key Stats
30K+
X3 Vehicles Supported
10 mo.
Figure 02 Pilot
03
New Figure Model
Hall 52
Spartanburg Focus
What BMW Announced
BMW is extending its humanoid robotics work with Figure AI at Plant Spartanburg in South Carolina. The previous Figure 02 project ran in the body shop, where the robot inserted sheet-metal parts for welding. BMW says that pilot supported production of more than 30,000 X3 vehicles and showed that a humanoid could complete precise, repeatable work under real production conditions.
The next step is Figure 03 in logistics sequencing. In BMW's description, components arrive in larger containers and are not yet sorted for final assembly. The robot will pick up parts, sort them into a sequencing trolley, and send that trolley toward a collection point. From there, an automated tugger train or Smart Transport Robot can carry the parts to the installation location so assembly workers receive them just in sequence.
That may sound narrower than the marketing phrase "general-purpose humanoid," but narrow is exactly why it matters. Production logistics has recurring work, known containers, defined destinations, and physical strain. It also has enough variety to expose whether a humanoid can move beyond a single carefully tuned body-shop task.
Key Insight
BMW is not asking Figure 03 to run the whole factory. It is testing whether a humanoid can handle a frequent logistics pattern: identify unsorted components, place them into the right sequence, and feed an existing automated material-handling system.
Why Spartanburg Matters
Plant Spartanburg is BMW's large U.S. manufacturing base and a natural proving ground because X-model assembly creates high logistics intensity, frequent part movement, and ergonomic pressure points.
Why Sequencing Matters
Sequencing is not just picking. It ties perception, grasping, order logic, line timing, trolley placement, and recovery from wrong or missing parts into one factory workflow.
From Body Shop to Logistics
The Figure 02 pilot was already a useful signal because it happened inside a real BMW production environment. The robot inserted sheet-metal parts into fixtures for welding, a job that demands accuracy and repeatability. For early humanoids, that type of task is a strong fit: the station can be instrumented, the fixtures are known, and success can be measured by cycle completion and quality checks.
Logistics sequencing adds a different kind of difficulty. Parts can arrive mixed in large containers. The robot has to select the correct item, orient it, and place it into the right position on a trolley. The downstream promise is that the assembly worker receives the right part at the right time, in the right order, without extra ergonomic burden upstream.
The shift also changes the business case. A body-shop task can prove dexterity and fixture interaction. A logistics task tests whether humanoid robots can become flexible automation inside the factory's material flow. If that works, the same pattern can be applied to many parts families and many stations. If it fails, the failure will show up quickly in missed sequence, rework, stopped lines, or humans spending too much time supervising the robot.
AI-generated image
Sequencing work depends on reliable perception, part identification, trolley logic, and material-flow integration. Source: Biped.News AI editorial image.
| Metric | Figure 02 BMW Pilot | Figure 03 BMW Project | What To Watch |
|---|---|---|---|
| Primary Area | Body shop | Production logistics | Whether the task transfers across areas |
| Task | Insert sheet-metal parts for welding | Sort unsorted components into sequence trolleys | Intervention rate and sequence accuracy |
| Reported Output | Supported 30,000+ BMW X3 vehicles | Not yet reported | Parts per hour and usable uptime |
| Environment | Fixture-heavy station | Containers, trolleys, transport links | Robustness to variation |
| Commercial Signal | Completed production pilot | New deployment step | Repeat deployment beyond one line |
What Figure 03 Adds
Figure says the newer robot brings features aimed at factory availability and safer human proximity. The announced list includes soft components, wireless charging, speech-to-speech audio functions, improved hands, tactile sensors, and palm cameras. Each of those features maps to a practical factory problem rather than only a spec-sheet upgrade.
Soft components matter when a robot shares space with workers, trolleys, racks, and vehicles. Wireless charging matters because manual plug-in routines become a hidden labor cost when fleets grow. Audio functions could help workers understand robot state or intervene faster, though voice interfaces on noisy factory floors still need proof. Tactile sensors and palm cameras matter because sequencing is a manipulation problem. The robot has to know not only that an object is visible, but that it has actually been grasped, oriented, and placed correctly.
The risk is that every added sensor and feature also adds integration burden. Automotive plants care about uptime, maintainability, safety validation, and predictable quality. A humanoid that completes a task in a demo can still lose the business case if it needs too much attention, pauses too often, or cannot recover from common mistakes without pulling an associate away from higher-value work.
Technical Context
• Tactile sensing: Useful for confirming grip quality when parts vary in shape, finish, or weight.
• Palm cameras: Useful for close-range placement checks where overhead vision may be blocked.
• Wireless charging: Useful if the robot needs more availability without adding manual battery-handling work.
• Speech functions: Useful only if they survive factory noise, safety rules, and real operator habits.
BMW's Physical AI Stack Is Bigger Than One Robot
BMW framed the project as part of a broader Physical AI strategy. That framing is important because the robot is not the only system being tested. Plant Spartanburg is also tied into BMW's iFACTORY approach, virtual factory simulation, ergonomic planning, and AI-assisted quality systems.
The sequencing project sits inside that larger production stack. Before parts reach the line, BMW can model process flows in virtual 3D simulations. The company says the BMW Virtual Factory helps make complexity manageable and can simulate human movement sequences to improve ergonomics. In quality assurance, BMW uses Artificial Intelligence Quality Next, or AIQX, for visual and acoustic inspection with camera systems and sensors that provide line workers with feedback.
For humanoid robotics, this is the part that is easy to miss. The robot is not succeeding alone. It succeeds if the plant gives it a well-modeled process, readable work objects, reliable transport interfaces, safety procedures, and enough data to improve over time. The commercial question is whether that surrounding infrastructure makes the humanoid more valuable than another piece of fixed automation.
AI-generated image
Physical AI deployment depends on the plant's digital layer as much as on the robot body. Source: Biped.News AI editorial image.
Deployment Reality Check
BMW and Figure have named the customer, site, prior production context, task family, and the next robot model. That makes this more substantial than a generic humanoid demo. The previous Figure 02 work also gives the announcement a useful baseline because BMW linked it to more than 30,000 X3 vehicles over roughly 10 months.
The missing numbers are just as important. BMW has not disclosed the number of Figure 03 units in the new logistics project. It has not published target parts per hour, intervention rate, mean time between faults, injury reduction, or cost per useful hour. Those are the numbers that decide whether a humanoid robot is a manufacturing tool or an impressive technology trial.
The honest read is that BMW is scaling the complexity of the test, not declaring scaled humanoid production. Figure 03 is being asked to work in a common logistics workflow with clear potential for reuse. If BMW publishes hard operating metrics later this year, this could become one of the cleaner benchmarks for factory humanoids.
Editorial Read
This is a serious pilot with a named industrial customer and a concrete task, but the robot count and productivity targets remain undisclosed. Treat it as a strong signal of BMW's intent, not proof that humanoids have already reached scaled automotive deployment.
What Comes Next
The next useful milestones are practical. BMW or Figure should disclose how many Figure 03 units are working in the logistics application, how many hours they run without intervention, how often they miss or misplace a part, and whether the same workflow expands beyond the initial area. A second plant or a second parts family would matter more than another polished demo video.
For Figure, Spartanburg is valuable because it links the company to one of the world's most demanding manufacturing customers. For BMW, the project is a way to test flexible automation without rebuilding every workstation around a fixed robot cell. For the humanoid robotics market, the announcement narrows the debate. The question is no longer whether a humanoid can be filmed doing factory work. The question is whether it can create dependable capacity in a production system that already measures everything.
Frequently Asked Questions
What is BMW using Figure 03 for?
BMW says Figure 03 will work on logistics sequencing at Plant Spartanburg. The robot will pick unsorted components from larger containers and place them into sequence trolleys for onward movement to assembly.
How is this different from the earlier Figure 02 pilot?
The earlier pilot involved inserting sheet-metal parts for welding in the body shop. The new work moves into production logistics, where the robot must handle sorting, sequencing, and integration with material transport.
How many robots are being deployed?
BMW and Figure have not disclosed the Figure 03 robot count for the new sequencing project. That keeps the announcement in pilot territory until fleet size and operating metrics are published.
Why does the 30,000-vehicle number matter?
It gives the previous Figure 02 work a production context. BMW says the pilot supported production of more than 30,000 X3 vehicles over roughly 10 months, which is more informative than a standalone lab demonstration.
Bottom Line
BMW's Figure 03 project is one of the more useful humanoid robotics signals of 2026 because it points at an actual factory bottleneck. Sequencing work is repetitive, physically demanding, and frequent across automotive production. It is also measurable. If Figure 03 can do it with low intervention and acceptable uptime, humanoids move closer to being a real factory automation layer.
The open question is scale. BMW has shown enough to justify watching closely, but not enough to declare the economics solved. The next proof will be boring on purpose: more hours, more parts, fewer interventions, and a second deployment that looks less like an experiment and more like standard equipment.