Automotive Companies Developing Humanoid Robots for Production and Customer Interaction

Text | Feng Wei Feng Da Bai

On October 21, local time in the United States, Tesla announced in its third-quarter earnings report that Optimus has begun production at the Fremont factory. Musk provided the same external delivery target as before during the subsequent conference call, aiming for around the end of 2027: the robot's hands, durability, and AI still need further improvements. To obtain training data, Tesla also plans to manufacture between 10,000 and 20,000 robots, allowing them to practice various tasks at the 'Optimus Academy.'

Automakers are looking for more efficient car manufacturing and also want to sell robots to new customers. It may first appear in stores: providing reception and explanations for cars, while observing how humans interact with robots.

From production lines to customers' hands, robots must go through task training, fault diagnosis, design modifications, and mass production. Tesla's plan precisely outlines this distance: the production line has already started, and the company still needs to manufacture tens of thousands of units, allowing them to keep practicing.

Why does the automotive industry have the conditions to take this path? Existing businesses can bear the initial investment, factories provide real-world tasks, and engineering and supply chains can incorporate improvements into the next generation of products.

The automotive industry has been continuously rewriting its production methods

The automotive market is large enough, and continuous improvements in production methods can translate into substantial profits. According to data published by the International Automobile Manufacturers Association (OICA), global car production in 2025 is expected to be about 96,400,000 units. It covers passenger and commercial vehicles, with a massive supply chain, factory operations, and after-sales demand behind it. For an industry of this scale, reducing the number of scrapped parts, minimizing inventory, and enabling new models to enter production faster are all worth companies investing in continuously.

Existing businesses can provide funding for new projects, but how long they can sustain the investment is still constrained by the performance of the core business. When Xiaopeng released its first-quarter performance results, He Xiaopeng proposed using the profitability and cash flow from the automotive business to support physical AI development, and then cultivate new revenue streams including robots. Afterwards, the sources of funding expanded further: according to Xiaopeng's disclosure, its robot business completed its first round of equity financing on August 24, raising over 900,000,000 dollars. This allows development risks to be shared by more investors. Automakers can first provide technology, manufacturing, and application conditions, and then introduce external capital for new businesses, thereby reducing the pressure of long-term investment that the core business alone would have to bear. Funding supports continued development, but whether it can form stable revenue still depends on the sales and usage of the product.

This type of investment has a clear historical lineage in the automotive industry. In 1913, Ford's moving assembly line allowed vehicles to be assembled station by station, reorganizing mass production. In 1961, Unimate entered General Motors' factory, taking over dangerous and repetitive tasks such as unloading from casting machines. By the stage of smart manufacturing, trial and error was moved earlier into the digital environment: BMW used virtual factories to check for collision issues between new car models and production lines, shortening the previously nearly four-week physical testing to a three-day virtual simulation.

Although the technologies involved in each stage are different, the requirements faced by enterprises are similar: the next batch of cars must be made better, new models must be delivered on time, and production costs must remain competitive. Toyota's production system emphasizes quality anomaly handling and just-in-time production, which is also about making process improvements serve overall delivery. In order to sell the next generation of products, automakers must continuously transform their production systems, and suppliers must also adjust parts, processes, and delivery arrangements. This long-term pressure is the real driving force behind the automotive industry's pursuit of productivity changes.

After capital investment, automakers still need to organize scattered technologies and manufacturing resources into products. A price and delivery target needs to be broken down into part performance, manufacturing costs, quality requirements, and procurement plans, so that suppliers can decide what to do and how much to invest. In the context of robot business, the role of automakers should also be judged along this line of relationship: what tasks they propose, how high the cost can be, and how they plan to procure, will provide suppliers with the basis for developing components. The vast supply chain is mobilized through clear demands and order expectations, transforming industry scale into real engineering power.

Automobiles are also a highly matured complex consumer product. Factories care about rhythm, yield, and downtime, while stores are the first to hear consumer questions about price, design, and functionality. The after-sales network continuously encounters real-world usage issues. These feedbacks influence how the next car is defined, which parts need to be modified, and which features are worth continuing to invest in. Once entering the robot business, the automotive network can provide early access opportunities, and robots still need to collect their own usage feedback. Whether the movement in the factory is stable, whether the communication in the store is effective, and whether home users are willing to use it daily correspond to different product requirements.

Of course, the electronics industry also has a massive market and precise manufacturing capabilities. According to IFR's 2024 industrial robot installation data, the electronics industry is still slightly higher than the automotive industry. The automotive industry is worth being discussed separately, because the demand for mass production, the development of complex electromechanical products, and the concentration of consumer brands and sales service networks are all within the same industry, and it continuously experiences product updates. It has the conditions to simultaneously become a user of robots, a product operator, and an early market organizer. Young automakers can also participate in this system; the key is how to use these conditions, rather than how many years the company has been around.

Machines begin to learn new tasks

Automotive factories are not lacking in robots. Fixed mechanical arms are adept at welding and painting tasks, which have already formed mature solutions. Today, the study of humanoid robots is still being pursued because companies hope robots can handle more varied tasks: the position of items may change, the production model may change, and the original tasks of picking up, moving, and operating, which were done by humans, are difficult to fully cover with a single set of fixed actions.

Visual, language, and action models are attempting to connect the environment, instructions, and actions. The end-side robot model published by Google DeepMind in 2025 has already demonstrated research progress on adapting to new tasks with minimal demonstrations. For automakers, the direction worth investing in is: after car model changes and part repositioning, can robots continue working through training, reducing the burden of redeveloping dedicated solutions each time.

The appeal of a humanoid structure is also related to this. A large number of workstations, tools, and passageways were originally designed for human use. Being close to human operational ranges allows robots to potentially utilize some existing environments. Companies can then attempt to bring machines to the tasks rather than having to rebuild an entire work setup around the equipment each time. Bipedalism is not the optimal choice for all tasks; stable wheeled devices can also accomplish many tasks. Automakers will ultimately purchase solutions that are suitable for the tasks. The spatial adaptability brought by humanoid robots must be demonstrated together with their action capabilities.

The collaboration between BMW and Figure has already demonstrated how real tasks can change robots. Figure 02 is responsible for loading sheet metal parts at the Spartanburg factory, requiring it to pick up items from the rack and place them into welding fixtures. According to Figure's project review, the forearm was the main hardware failure point during the cumulative 1,250 hours of operation. The third-generation product therefore redesigned the wrist electronics, removing the communication distribution board and dynamic cables from that location. Problems encountered in the workshop were thus incorporated into the design of the next generation.

This collaboration allows BMW to validate new equipment on its existing production lines, while Figure gains feedback on failures and maintenance that laboratory environments cannot continuously provide. The returns each party seeks are different: BMW needs better production tools, while Figure aims to incorporate improvements into its own products to serve more customers.

Automakers' recent investments in intelligence have also given them the opportunity to further participate in development. Xiaopeng uses its self-developed Turing chip in IRON, while simultaneously building a dedicated production line for robots. Chips, computing facilities, and development tools can serve different products, but robots still need to establish their own motion control and task capabilities. For enterprises, the significance of such investments is not just about building fewer systems: the technical resources originally used for cars are now beginning to have the opportunity to validate other product directions.

Cars can recognize roads, but robots still need to learn to grasp, touch, and operate objects, and the constraints they face are not the same. However, if a company believes that intelligent technology can continuously expand the range of physical tasks it can perform, it has a reason to maintain a research direction beyond cars. Waiting until the robot market is fully mature to enter means leaving the early product definition, usage feedback, and partnerships to the pioneers. The early investment purchases both technological progress and the time to choose the next business venture.

How does the investment return to business?

When an enterprise simultaneously advances multiple robot applications, it must allocate limited R&D human resources and budgets. Stores hope for robots with friendly appearances and smooth communication, while factories are more concerned with payload, motion reliability, and continuous operation. Products optimized for hospitality may not be suitable for entering production lines; capabilities developed for factories may not attract consumers. Management needs to first determine who the priority service recipients are and which department will bear the development and trial costs, so that they can accept the requirements accordingly. Otherwise, if the project gains attention through demonstrations, but the production department cannot use it, the project will struggle to sustain itself with just a few sets of conflicting performance reports.

Modern placed the training of Atlas in manufacturing tasks, providing an observable path. RMAC first conducted validation around part sorting and material handling, and the formal deployment in the factory was arranged in subsequent stages. The investment during training is to prepare for future production applications. If robots can gradually take on repetitive handling or heavy tasks, factories can reassign some positions, using people for tasks that require judgment, handling anomalies, and maintaining equipment. Changing the human resource structure includes adjustments to positions and skills, and cannot be directly equated to laying off a certain number of workers.

The production department verifies the changes in the entire project: how much it costs for training, maintenance, human intervention, and downtime, and whether the qualified output is more stable. If the same equipment can also be trained to adapt to new parts and new workstations, there is an opportunity to reduce dedicated modifications during changeovers. When costs are still high and output is unstable, factories have even more reason to narrow the task scope, switch to other equipment, or delay procurement. The operational records Figure accumulated on-site at BMW answer how the product should be modified; expanding procurement still requires proving that robots are more worth investing in than existing solutions for this task.

The group can arrange its own factory for trial use, but external customers will compare robot solutions with human labor, fixed mechanical arms, and other options in terms of price, deployment, and maintenance costs. An internal procurement by the group cannot prove that the group as a whole has gained new revenue. The second curve needs to be established, and the product must still be something that external customers are willing to pay for continuously.

The family market is another challenge. Consumers will look at whether the robot can complete tasks such as moving and picking up items, whether it is easy to use and repair, and whether it is worth the price. The experience of car companies in operating durable goods can help them consider delivery and long-term service, but the purchase rationale still needs to be provided by the robot itself.

This market expansion will also change product design. Companies cannot build a new machine for every single customer, or else the more they sell, the heavier the burden of customization and service will be. Robots need to gradually form hardware that can be mass-produced and task capabilities that can be expanded, so that subsequent customers can benefit from the previous development. The value of manufacturing experience here is to help new business organize product updates and continuous delivery. Only after orders and designs stabilize can scale have the opportunity to further reduce costs.

Before this new business matures, robots may also enter the car sales process. MoJia has publicly introduced its application in overseas car marketing service scenarios and disclosed its use of Qiche's distribution channels, logistics, and service networks. Car dealerships provide robots with places to interact with consumers, and robots add an interactive method to the dealership.

Many investments in intelligent technology are invisible to consumers. A robot that can naturally communicate and complete reception tasks will make some of these technological capabilities tangible on-site, also giving salespeople more opportunities to introduce cars.

Factory trials, dealership receptions, and external sales do not have a unified return order. Modern first verified around manufacturing tasks, and Xiaopeng and MoJia considered service scenarios for people earlier. Store applications, in particular, need to distinguish between one-time display and long-term operation: activity budgets can support short-term appearances, but whether dealers are willing to continuously bear maintenance and service costs is closer to the real demand of this channel.

Figure 1: Support and return validation for robot iteration in the automotive business. Solid lines indicate the investments, usage, or design feedback disclosed by the companies in the text, dashed lines indicate the yet-to-be-validated returns and their relationship with subsequent investments; each relationship does not occur at the same company.

Commercial goals are beginning to diverge along different paths

Which type of return a company hopes to obtain determines how much robot business the automaker will take on. Developing products in-house, holding specialized companies, or purchasing solutions from external companies can all bring robots into the automotive industry, but the direction of funds and the responsibilities of operations differ.

Hyundai holds Boston Dynamics and lets Mobis participate in the development and supply of Atlas actuators, enabling it to coordinate product development, component manufacturing, and internal applications within the group. It must invest in robot R&D and manufacturing, and also face the external market for this business. BMW's collaboration with Figure allows it to propose requirements around car production and participate in validation, with Figure responsible for robot product development. BMW needs to judge whether the solution is worth adopting, and Figure must determine how to sell the improvements to the next customer.

Xiaopeng chose to develop IRON in-house and build dedicated production lines. The responsibilities of product definition, R&D, manufacturing, and delivery fall more heavily on its own business system. It can decide on its own how the robot will collaborate with existing intelligent technologies and also needs to organize sales and after-sales services for this product line. According to its announced plan, the robot will first start from stores and campuses. These scenarios can provide interaction and service tasks. After internal trials, it will also need to establish delivery and maintenance capabilities for external customers.

MoJia uses Qiche's channels to enter overseas car marketing scenarios, enabling it to leverage existing outlets to reach customers and arrange delivery using its logistics and service networks. However, having an automotive sales service network does not mean that it already has robot service capabilities. Robot companies still need to provide support for local languages, tasks, and repair needs, enabling outlets to actually operate such products.

Figure 2: Three business relationships and their public progress. BMW is a user and collaborator, Figure is the robot product developer. Future deployment, mass production, and delivery times are all planned by the companies; Figure 03 logistics tasks are for demonstration purposes.

This division also leaves room for specialized robot companies and component manufacturers. Automotive customers propose requirements and participate in validation, while product and component suppliers are responsible for turning experience into deliverable results that can be resold. Original equipment manufacturers can benefit from better production solutions, while suppliers can continue to operate for other customers. The role of the automotive industry in promoting robot development will also occur through such collaborations, and cannot be measured solely by how many robots the automaker has sold.

Modern has Atlas practice part sorting first, Xiaopeng has IRON start from stores and campuses, and BMW has Figure find problems on production lines. They have already handed the next round of R&D to different sites: factories continuously expose issues such as data collection, malfunctions, and downtime, while stores will expose issues such as interaction, reception, and maintenance. The automotive industry can mobilize capital, manufacturing, and sales networks, and whichever site repeatedly solves problems first, will first develop that type of robot. Return to the previous page for more information