Industry News
BMW Is Running Two Humanoid Robot Programs at Once
BMW has become the first automaker to run two separate humanoid robot programs in parallel. At Spartanburg, Figure 02 moved over 90,000 components across 1,250 operating hours while building X3s. At Leipzig, Hexagon's wheeled AEON humanoid is preparing for battery assembly work starting summer 2026.
BMW Runs Two Humanoid Robot Programs at Once: Figure 02 Hits 90,000 Parts, Hexagon AEON Starts Battery Work
The German automaker is the first to operate legged and wheeled humanoids side by side in production, turning its factories into the most active testbed for humanoid robotics in the automotive industry.
By The Numbers
Overview
BMW has quietly become the most aggressive automaker in humanoid robot deployment, and the gap between it and every other car manufacturer is widening. While competitors are still running isolated pilots or signing letters of intent, BMW is now operating two distinct humanoid programs at two different factories on two different continents, using two completely different robot architectures.
At its plant in Spartanburg, South Carolina, BMW wrapped up an extended deployment of the Figure 02 robot, a fully legged humanoid that spent ten months working ten-hour weekday shifts on the X3 assembly line. Over that stretch, Figure 02 logged roughly 1,250 operating hours, moved more than 90,000 sheet metal components, and took over 1.2 million steps on the factory floor. The robots were removing and positioning parts for welding, a task that requires consistent force, repeatable accuracy, and the ability to handle variable part geometries.
At its Leipzig factory in Germany, BMW has started test deployments of a completely different machine: the Hexagon AEON, a wheeled humanoid that rolls on flat surfaces but can step over obstacles when needed. The AEON is being evaluated for high-voltage battery assembly and component manufacturing, with a full pilot scheduled for summer 2026.
Running two programs simultaneously is not a coincidence. It reflects a deliberate strategy to evaluate different robot form factors against different factory tasks, gather comparative data, and avoid locking into a single vendor or architecture before the technology matures.
Technical Deep Dive
The two robots BMW is currently running could not be more different in their mechanical philosophy. Figure 02 is a conventional biped: two legs, two feet, fully ambulatory. The AEON takes a hybrid approach, using wheeled legs that give it speed and stability on flat surfaces while retaining the ability to navigate uneven terrain or step over obstacles.
| Specification | Figure 02 | Hexagon AEON | Boston Dynamics Atlas |
|---|---|---|---|
| Locomotion | Fully legged biped | Wheeled legs (rolls / steps) | Fully legged biped |
| End-effectors | Fixed dexterous hands | Interchangeable grippers / tools | Dexterous hands |
| Primary use case at BMW | Sheet metal handling for welding | High-voltage battery assembly | Not deployed at BMW |
| Deployment status (BMW) | Completed 10-month run; Figure 03 evaluation underway | Test phase (Dec 2025, Apr 2026); full pilot summer 2026 | Not deployed; Hyundai is primary customer |
| Factory location | Spartanburg, SC (USA) | Leipzig (Germany) | N/A |
| Shift structure | 10-hour shifts, Mon-Fri | Test deployments; pilot TBD | N/A |
| Operating hours logged | ~1,250 hours | Early-stage test deployments | N/A |
Figure 02: Lessons from 10 Months on the Line
The Spartanburg deployment gave BMW and Figure real production data at a scale few humanoid programs have reached. Sheet metal handling is a physically demanding, repetitive task: the robot had to reliably pick up formed steel panels, orient them correctly, and position them for welding, cycle after cycle. Doing that for 1,250 hours across a fleet operation that supported 30,000-plus vehicle builds is not a lab result. It is production data.
The 1.2 million steps metric is more than a headline number. It speaks to the mechanical durability of Figure 02's legs and drive system under sustained factory conditions. Industrial floors are not perfect; they have vibration, slight gradients, foot traffic, and other robots moving through the same space. The fact that the deployment ran for ten months without a reported stoppage due to hardware failure (at least none disclosed publicly) is a meaningful data point for the broader industry.
BMW and Figure are now evaluating the Figure 03 for additional use cases. Figure 03 is the next-generation platform and is expected to offer improved manipulation capabilities and onboard AI processing. The transition from 02 to 03 at a working customer site is a sign that the partnership is deepening rather than winding down.
Hexagon AEON: The Wheeled Humanoid Bet
The AEON's design philosophy is worth examining. Most humanoid robots in factory deployment today are fully legged, which gives them maximum flexibility to navigate any environment but at the cost of speed, stability, and energy efficiency on flat surfaces. Hexagon took a different view: most factory floors are flat. Wheels are more efficient on flat surfaces. So why not put wheels where the feet would be?
The result is a robot that rolls quickly and stably across open floor space but can switch to stepping when it needs to navigate stairs, ramps, or threshold changes. The upper body retains full humanoid proportions, which matters for working in spaces designed around humans, reaching into machines, and using standard tooling. The AEON also supports interchangeable end-effectors, meaning BMW can swap in different grippers, scanners, or tools depending on the task.
For battery assembly, this flexibility is useful. High-voltage battery packs require precise torque application, connector seating, and quality scanning steps. Being able to swap a torque tool for a scanner without replacing the entire robot arm is a practical manufacturing advantage.
Key Players
BMW
The Munich-based automaker has transformed itself into the most active automotive testing ground for humanoid robots globally. With active programs at both Spartanburg and Leipzig, BMW is gathering cross-vendor, cross-architecture data that no other automaker currently has. Milan Nedeljković, BMW's Board of Management member responsible for production, has framed the strategy around AI and engineering convergence: "The symbiosis of engineering expertise and artificial intelligence opens up entirely new possibilities in production."
Figure AI
The San Jose-based startup completed a landmark 10-month deployment at Spartanburg with Figure 02 and is now positioning Figure 03 for expanded work at BMW. Figure has raised over $675 million and counts Microsoft, OpenAI, and Nvidia among its backers. The Spartanburg deployment is by far the most publicly documented example of a humanoid running in sustained automotive production.
Hexagon
The Swedish industrial software and metrology company branched into humanoid robotics with the AEON platform, which is now beginning its factory trial at BMW Leipzig. Hexagon's manufacturing intelligence background shapes the AEON's design priorities: tool interchangeability, measurement integration, and precision assembly tasks. The Leipzig deployment puts the AEON in a real EV battery production environment far earlier than most new entrants reach.
Boston Dynamics
Not a BMW partner (yet), but relevant context. Boston Dynamics and its Atlas humanoid are at the center of the industry's biggest pending contract: Hyundai plans to purchase thousands of Atlas units for its manufacturing operations. Hyundai owns Boston Dynamics, which gives the Korean automaker a structural advantage in supply. The question is whether owning the robot company is smarter than running a competitive multi-vendor program like BMW's.
Industry Impact
The automotive sector is emerging as the proving ground for industrial humanoid robots, and BMW's multi-vendor approach is setting a template that other manufacturers are watching closely.
Agility Robotics and its Digit robot are the other major data point. Digit is deployed at distribution centers operated by GXO Logistics for Spanx, at Schaeffler's bearing manufacturing facilities, and at Toyota Motor Manufacturing Canada. Those deployments are real but spread across logistics and light manufacturing, not the high-intensity, high-volume environment of an automotive final assembly line.
The collapse of K-Scale Labs in late 2025 added a sobering note to the industry's trajectory. The YC-backed startup, which had positioned itself as an open-source humanoid platform, shut down operations and released its intellectual property publicly. The lesson was not subtle: hardware is harder than software. Building a reliable, durable humanoid robot that can handle ten-hour daily shifts without mechanical failure is a different engineering challenge than building capable AI software. K-Scale ran out of runway before it could solve the hardware problem.
That context makes BMW's results with Figure 02 more meaningful. Sustained production deployment is the test that startup claims cannot substitute for. The 1,250 hours and 1.2 million steps are data points that K-Scale never reached, and that most current humanoid programs have not either.
For the broader automotive supply chain, BMW's two-program approach signals something important: automakers are not waiting for a single dominant platform to emerge. They are running parallel evaluations, treating humanoid robots the way they treated early collaborative robot deployments in the 2010s, running multiple vendors, learning what works, and scaling the winners.
What's Next
Figure 03 Evaluation at Spartanburg
BMW and Figure are actively evaluating the next-generation Figure 03 for additional use cases at the South Carolina plant. Figure 03 is expected to offer improved dexterity and processing compared to Figure 02. No timeline has been announced for a production deployment.
Hexagon AEON Full Pilot at Leipzig (Summer 2026)
After a first test deployment in December 2025 and a second in April 2026, the AEON is scheduled to move into a full production pilot at Leipzig during summer 2026. The focus will be high-voltage battery assembly, a safety-critical task where robot reliability is non-negotiable.
Hyundai and Atlas Scale-Up
Hyundai's plan to purchase thousands of Boston Dynamics Atlas robots will create the largest single humanoid fleet in manufacturing if it proceeds. That deployment will generate a massive volume of real-world data and could accelerate the entire industry's understanding of what humanoid robots can and cannot do at scale.
Agility Digit Expansion
Agility Robotics is pushing Digit further into automotive supply chains, with Toyota Motor Manufacturing Canada as its most prominent automotive customer. Schaeffler's deployment in component manufacturing provides another data point for how Digit performs in precision industrial tasks outside logistics.
Frequently Asked Questions
Why is BMW running two different humanoid programs at the same time?
BMW is deliberately testing multiple robot architectures against different factory tasks to build vendor-independent knowledge. Figure 02 is suited for ambulatory tasks in mixed human-robot environments; the AEON's wheeled design makes it faster and more energy-efficient on flat factory floors. Different tasks call for different designs, and BMW wants data on both before making larger commitments.
What did Figure 02 actually do at Spartanburg?
Figure 02 handled sheet metal parts for the BMW X3 assembly process. Specifically, it removed formed steel panels from one position and placed them correctly for welding operations. The robot ran 10-hour weekday shifts for 10 months, moved over 90,000 individual components, and logged roughly 1,250 operating hours total.
What is the Hexagon AEON and how does its wheeled design work?
The AEON is a humanoid robot from Hexagon that uses wheels in place of traditional feet. On flat surfaces it rolls, which is faster and more energy-efficient than walking. When it encounters stairs, ramps, or obstacles, it switches to stepping. The upper body is humanoid in proportion, allowing it to work in spaces designed for humans. It supports interchangeable grippers and scanning tools, making it adaptable to different assembly tasks.
How does BMW's approach compare to what Hyundai is doing?
Hyundai owns Boston Dynamics and plans to buy thousands of Atlas humanoids for its own factories. That is a concentrated bet on a single platform from a captive supplier. BMW's approach is the opposite: running competitive evaluations across multiple vendors and architectures without a single dominant supply agreement. Both strategies have merit. Hyundai gets scale advantages and deep integration; BMW gets comparative data and vendor flexibility.
What happened to K-Scale Labs and what does it mean for humanoid robotics?
K-Scale Labs was a Y Combinator-backed humanoid startup that shut down in late 2025 and open-sourced its intellectual property. The closure underscored that building reliable humanoid hardware is extraordinarily difficult and capital-intensive. Software can iterate quickly; hardware cannot. The failure added credibility to established players like Figure and Boston Dynamics that have already demonstrated sustained production deployments.
The Bottom Line
BMW has moved past the pilot-project phase and into sustained, parallel humanoid deployment. Two robots, two factories, two vendors, two completely different mechanical architectures, all running simultaneously. Figure 02 delivered 10 months of production data at Spartanburg and is being succeeded by Figure 03 evaluation. The Hexagon AEON is beginning its battery assembly trial at Leipzig with a full pilot targeted for summer 2026. No other automaker is operating at this breadth. The gap between BMW and the rest of the automotive industry in humanoid deployment experience is growing with each shift that runs.