A robot picks up a component, identifies where it belongs, places it in the correct sequence and moves on to the next task.
It does not look like an industrial revolution.
But inside BMW's Spartanburg factory in South Carolina, this apparently ordinary operation represents a significant shift in the relationship between artificial intelligence, manufacturing and human labour.
The robot is Figure 03, developed by the American robotics company Figure AI. Its task is part of BMW's move towards using humanoid robots in production logistics, where components arrive unsorted and must be organised for assembly.
The system uses Figure's Helix 02 artificial intelligence technology to coordinate its hands, arms, torso and feet while manipulating parts and repositioning its body. BMW says the new application builds on an earlier deployment of Figure 02, which supported the production of more than 30,000 BMW X3 vehicles. <Cite refs={["turn417346view0","turn304460search4"]}/>
The significance is not that a robot can pick up an object. Industrial robots have been doing that for decades.
The significance is the attempt to build machines that can operate in environments designed for humans, adapt to changing physical situations and perform a wider variety of tasks without requiring a separate, rigidly programmed machine for every operation.
If that becomes commercially viable at scale, it could change the economics of manufacturing.
And Europe, Africa and every country hoping to industrialise should be paying attention.
1. The old factory robot had a weakness. AI is trying to overcome it.
Industrial automation has been remarkably successful when manufacturers can control the environment.
A conventional industrial robot can weld, paint, cut, lift and position components with exceptional speed and precision. But these systems generally work best when their tasks, tools, movements and surroundings are carefully engineered.
Change the position of a component, introduce an unfamiliar object or alter the task, and the system may require new programming, sensors, fixtures or engineering work.
This is not a universal limitation of industrial robotics. Modern industrial systems already use machine vision, adaptive controls and advanced sensing. But conventional automation can become expensive when a task varies substantially.
Humans are naturally better at dealing with many such variations. We can look at a container of parts, recognise their shapes, understand what is needed and adjust our movements accordingly.
Figure AI is attempting to combine that flexibility with robotic consistency.
Its Helix system uses visual information to help translate perception into physical action. The objective is to make a robot capable of performing more tasks in changing environments rather than relying exclusively on preprogrammed movement sequences.
That distinction matters because the economics of automation depend not only on what a machine can do, but also on how expensive it is to adapt the machine when production changes.
If a robot can be redeployed across several tasks through software and training rather than extensive physical reconfiguration, its potential economic value increases.
The real breakthrough would not be a robot that can perform one task.
It would be a robot that can learn to perform many tasks reliably enough for a factory to depend on it.
2. BMW has already tested the concept in real production
There is a difference between a technology demonstration and an industrial deployment.
A carefully edited video can show a machine completing a task successfully. A factory must establish whether it can do that task repeatedly, at the required speed, with acceptable safety and reliability.
BMW's earlier Figure 02 project provides useful evidence of progress.
According to Figure AI's November 2025 report, its robot deployment at BMW's Spartanburg plant involved:
- More than 90,000 parts loaded.
- Over 1,250 hours of operation.
- Approximately 1.2 million steps.
- Work performed across ten-hour weekday shifts.
- Contributions to the production of more than 30,000 BMW X3 vehicles.
The robot's task was to retrieve sheet-metal parts and position them on welding fixtures. Conventional industrial robots then carried out the welding operations. <Cite refs={["turn417346view2","turn304460search2"]}/>
This last detail is important.
The humanoid robot did not independently manufacture 30,000 cars. It performed a defined operation within a much larger production system.
That is still significant.
The deployment demonstrates that humanoid robotics can move beyond laboratory experiments into demanding industrial environments. It also generated operational lessons about reliability, hardware design and integrating robots into existing factory systems.
Figure has said that experience at BMW informed changes to the next-generation robot, including a redesign of wrist electronics after the forearm emerged as its most significant hardware failure point during the earlier deployment. <Cite refs={["turn417346view2"]}/>
That is what industrial progress often looks like: not instant perfection, but repeated testing, identification of weaknesses and engineering improvements.
Figure 03 now represents the next step.
Its sequencing task involves retrieving unsorted components, organising them on a trolley and delivering them for onward transportation to the assembly area. <Cite refs={["turn417346view0"]}/>
The larger ambition is to make this flexibility repeatable across more production tasks.
3. The biggest economic opportunity may be redesigning the factory
The most obvious interpretation of humanoid robotics is that robots will replace workers.
That is part of the potential economic impact, but it is not the whole story.
Manufacturers already automate tasks where automation is economically attractive. The more consequential opportunity is to automate activities that have remained difficult or expensive because they require human-like movement, adaptation or manipulation.
Consider the movement of components through a factory.
Parts arrive at a warehouse or production area. Workers or machines sort them, transport them, position them and deliver them to the correct station at the correct time.
A delay in one part of this process can affect the entire production line.
A manufacturer may already have sophisticated welding robots but still rely on human workers to handle variable material flows between machines.
A more adaptable robotic system could potentially bridge some of these gaps.
If it can work with existing equipment, tools and workspaces, a manufacturer may not need to rebuild an entire facility around a new automation system.
That could lower the cost of automating certain tasks.
It could also change how factories are designed. Engineers might increasingly organise production around a combination of specialised industrial robots, autonomous transport systems, humanoid machines and human workers.
The future factory may not be completely robotised.
It may instead be a more flexible combination of different kinds of automation.
The strategic prize is not simply replacing a worker with a machine. It is reducing the cost and complexity of coordinating an entire production process.
4. Nvidia's involvement reveals where the next industrial competition is heading
Physical AI requires more than a capable robot.
It requires training data, computing infrastructure, machine-learning models, simulation systems, sensors, actuators, batteries and reliable mechanical engineering.
Figure's September 3, 2026 announcement with Nscale provides a striking illustration of this infrastructure requirement.
The companies announced a partnership targeting access to up to 100,000 Nvidia GPUs using the forthcoming Vera Rubin platform. The initial deployment is targeted for the second half of 2027 in Barstow, Texas.
Figure described an initial computing commitment of $3.5 billion, with an intention to scale beyond $6 billion. The agreement also includes a strategic investment by Nscale in Figure. <Cite refs={["turn417346view1","turn304460search5"]}/>
These figures require careful interpretation.
The 100,000 GPUs are a planned maximum, not evidence that all those processors are already installed and operational. The stated computing commitment is also not the same as a completed cash payment or a demonstrated return on investment.
Nevertheless, the scale of the proposed arrangement shows how capital-intensive the development of general-purpose robotic intelligence may become.
Training a system to interact with the physical world requires information about more than language or images.
It needs to learn how objects move, how forces affect them, how hands interact with surfaces, how bodies maintain balance and how actions produce consequences.
The system must then translate its learned capabilities into reliable physical behaviour.
This creates a growing relationship between robotics and the AI infrastructure industry.
Nvidia supplies computing technology. Nscale supplies planned computing capacity. Figure develops the robots and AI systems. BMW provides a real manufacturing environment in which industrial applications can be tested.
Each part of this ecosystem contributes something different.
The result is a new competitive landscape in which manufacturing leadership may increasingly depend on access to both physical engineering and advanced computing.
5. The new economics: when does a robot actually pay for itself?
This is where the excitement should meet financial discipline.
A robot performing a task successfully does not automatically mean that deploying it is profitable.
Manufacturers must compare the total cost of the robotic system with the cost and performance of the alternatives.
The relevant expenses include:
- The robot's purchase price or leasing cost.
- Integration into the existing production system.
- Training, software and computing.
- Electricity and charging infrastructure.
- Repairs, replacement parts and maintenance.
- Safety systems and operating procedures.
- Human supervision and intervention.
- Downtime when the robot cannot complete a task.
- The cost of redesigning processes around the new technology.
There are also potential benefits beyond wages.
A robot may reduce exposure to physically demanding work, improve consistency, operate during additional shifts or reduce errors in material handling.
But each benefit must be measured in the specific application.
A robot that works reliably for many hours with minimal supervision could be valuable. A robot that frequently stops, requires technicians to intervene or cannot meet production-cycle requirements may be much less attractive.
The correct comparison is therefore not simply:
Robot cost versus worker salary.
It is:
Total cost of delivering the required productive output with the robot versus total cost of delivering the same output through the available alternatives.
And the alternatives may include a conventional industrial robot, a specialised automated machine, a human worker or a hybrid arrangement.
For some operations, a traditional fixed robot will remain cheaper and faster. For others, a humanoid machine could offer greater flexibility.
The commercial outcome will depend on the task.
6. Europe should be concerned about industrial competitiveness, not just the appearance of humanoid robots
The original argument about Europe raises an important question: what happens when the technology that determines manufacturing productivity is developed elsewhere?
Europe has extensive industrial capabilities, established automotive manufacturers, advanced engineering expertise and a substantial research base.
It is not starting from zero.
Indeed, BMW's own work illustrates that European manufacturers can help develop and deploy these technologies. BMW's September 2026 announcement also described a humanoid-robot pilot at its Leipzig plant in Germany, using a different system called AEON to support production logistics. <Cite refs={["turn417346view3"]}/>
That is evidence of European industrial participation, not proof that Europe has already lost the race.
But the competitive challenge is real.
If American or Asian companies establish advantages in robotic foundation models, semiconductor supply, deployment data and manufacturing-scale integration, European factories may have to purchase or license increasingly important components of their automation systems.
That could affect costs, bargaining power and the distribution of industrial profits.
It could also create dependencies in areas such as software updates, technical support, spare parts and access to advanced computing.
Yet buying foreign technology is not necessarily a strategic failure. Manufacturing has always relied on international supply chains.
The question is whether European firms retain the ability to understand, modify, integrate and improve the technology—or become dependent on a small number of external providers for capabilities essential to production.
Europe's response need not be to manufacture every chip or robot domestically.
It could focus on strengths in industrial systems, precision engineering, sensors, safety, robotics integration, software and specialised manufacturing applications, while ensuring reliable access to critical components.
The goal should be technological capability and negotiating power, not autarky.
7. What does this mean for workers?
The impact on employment will not be uniform.
Robots may reduce demand for some repetitive material-handling tasks. They may also create demand for technicians, robotics engineers, maintenance specialists, safety professionals and people who can integrate automated systems into manufacturing operations.
Some workers may be reassigned to higher-value activities. Others may face displacement, particularly where retraining is difficult or where employers use automation to reduce staffing.
The balance will depend on how quickly the technology improves, how widely it is deployed, whether new demand expands production and how employers organise the transition.
It is therefore premature to conclude that humanoid robots will eliminate factory employment.
It is equally premature to assume that they will have no effect on jobs because the first deployments are limited.
The most useful questions concern measurable changes:
How many hours of human labour does the robot displace or augment?
How many workers are needed to supervise a fleet?
How frequently does the system require human intervention?
Does automation expand total production enough to create other jobs?
Who captures the productivity gains?
If a company reduces costs, do the benefits translate into higher wages, lower prices, increased investment or primarily higher profits?
These are economic and policy questions, not simply engineering questions.
Governments, manufacturers and workers need to discuss them before large-scale deployment makes the consequences harder to manage.
8. Africa should not watch this as a spectator
For African countries, the story is not primarily about whether humanoid robots will soon appear in every factory.
It is about the changing requirements for industrialisation.
A country seeking to build a manufacturing sector must compete on productivity, reliability, logistics, skills, energy and access to capital.
AI-enabled robotics may change the economics of some activities, particularly where flexible automation can make smaller or more variable production runs viable.
But the technology will not remove the fundamental requirements of industrial development.
A factory still needs reliable electricity, transport infrastructure, spare parts, skilled technicians, quality control, financing and customers.
A humanoid robot cannot compensate for a broken supply chain or unreliable power.
Nor does buying a robot automatically create domestic technological capability.
If an African factory imports the machine, imports the critical components, depends on foreign computing infrastructure and cannot repair the equipment locally, it may gain productivity while remaining dependent on external suppliers.
That may still be commercially worthwhile. But it is not the same as developing an indigenous robotics industry.
African governments and businesses should therefore consider several complementary priorities.
First, invest in practical technical skills. Industrial maintenance, mechatronics, machine vision, control systems and robotics integration will become increasingly important as factories adopt advanced automation.
Second, build local integration capabilities. African engineers should be able to install, maintain and adapt robotic systems to local manufacturing conditions rather than relying entirely on foreign contractors.
Third, use automation where the economics justify it. A specialised machine or a conventional robot may be more suitable than a humanoid for many applications. Technology choices should follow the task, not the hype.
Fourth, support local manufacturing and component supply. Opportunities may exist in tooling, grippers, sensors, industrial software, power systems, maintenance and other parts of the robotics ecosystem.
Fifth, develop industrial data capabilities. Factories that collect reliable operational data can improve production planning, quality control and automation, even before deploying advanced humanoid robots.
Sixth, make power reliability a strategic priority. AI infrastructure and industrial automation both depend on dependable energy. Without it, the productivity advantages may not justify the investment.
For Nigeria in particular, the objective should not be to import the most expensive robot available and announce that the country has entered the future.
It should be to identify industries where automation can improve domestic productivity, develop the workforce and suppliers needed to support it, and build businesses that can retain more of the value created.
9. The decisive question is not whether the robot looks human
Humanoid robots attract attention because their appearance makes their progress easy to understand.
But a human-shaped machine is not automatically the most efficient way to automate a task.
A fixed industrial arm may be better for welding. An autonomous mobile robot may be better for moving pallets. A conveyor may be better for transporting large volumes of standardised products.
Humanoid robots become interesting when their form offers practical advantages: using equipment designed for people, navigating human workspaces, handling varied objects or switching between tasks without extensive physical changes.
That is the economic proposition Figure is testing at BMW.
The technology still faces substantial challenges, including reliability, cycle times, safety, maintenance, deployment cost and the amount of human oversight required.
The factory video cannot answer all those questions.
But the fact that BMW is progressing from a defined material-handling task to more complex sequencing work shows that industrial applications are being explored through real operating environments rather than demonstrations alone. <Cite refs={["turn417346view0","turn417346view2"]}/>
That is worth watching closely.
Conclusion: The next industrial advantage may belong to those who combine machines, models and manufacturing
The robot at BMW is a glimpse of an industrial possibility, not proof that the factory of the future will be entirely autonomous.
The earlier Figure 02 deployment established a useful operational foundation. Figure 03 is now being tested on a more complex logistics task. The planned Nscale computing partnership demonstrates the scale of infrastructure Figure believes it needs to advance its AI systems.
Together, these developments illustrate a wider shift: manufacturing is becoming a contest not only over machines and labour, but also over data, computing, software and the ability to translate intelligence into reliable physical work.
Europe has substantial industrial assets and opportunities to participate in that transition. Its challenge is to preserve and expand its technological capabilities while avoiding excessive dependence on external suppliers.
Africa faces a different but related challenge: turning access to automation into domestic productivity, technical expertise and industrial capacity.
Neither continent can assume that progress elsewhere will automatically translate into prosperity at home.
The real measure of success will be the productive output a system delivers, the cost and reliability with which it delivers it, and the share of the resulting economic value that remains with the companies, workers and societies that deploy it.
The robot sorting parts at BMW may look like a small development.
But if machines become more adaptable, the implications could extend far beyond one factory or one automaker.
Industrial revolutions often begin when a task that once required human judgement becomes reliable enough to automate. The countries that recognise the economic consequences early will be better positioned to shape what comes next.
Comments
Post a Comment