Humanoid Robotics: From Pilot to Demand to Scale

Across leading robotics markets, humanoid robots are beginning to move beyond staged demonstrations into factories, warehouses, and other industrial environments. During its 2025 pilot at BMW Group Plant Spartanburg, Figure 02 handled more than 90,000 components and accumulated approximately 1,250 operating hours within an active automotive production line. Such deployments suggest that demand is beginning to emerge most clearly in structured applications where robots can perform useful work within facilities and workflows originally designed for people.

However, successful pilots should not be mistaken for commercial maturity. To progress towards repeatable deployment at scale, humanoid robots must become safer, more dexterous, capable of sustaining productive uptime, and economically competitive with human labour and existing automation alternatives. Achieving this will depend not only on advances in artificial intelligence, but also on the maturity of the supply chain for actuators, sensors, batteries, processors, precision components, and supporting infrastructure.

This article examines where genuine demand for humanoid robots is emerging, which commercialisation gaps still separate pilots from scale, and how China and the United States are pursuing different pathways to overcome them. It also considers where value may accrue across the broader ecosystem as humanoid robotics moves towards reliable and economically viable industrial adoption.

Robot Figure 02 during its 2025 pilot at BMW Group Plant Spartanburg, where it supported production of more than 30,000 BMW X3 vehicles, handled over 90,000 components, and logged approximately 1,250 operating hours — Source: BMW Group’s official website. 

1. Why Humanoids, and Why Now? 

Conventional industrial robots have historically created value through specialisation Many current humanoid platforms are being developed as general-purpose or multipurpose robots capable of performing different tasks within human-designed environments. 

McKinsey groups general-purpose robots into three principal forms: wheeled robots, quadrupedal robots, and humanoid robots. Wheeled robots move efficiently across flat, structured environments such as warehouses; quadrupeds are better suited to stairs, uneven terrain, and inspection in hazardous locations. Meanwhile, humanoids are general-purpose, bipedal robots modelled on the human form, combining human-like arms and hands with the mobility needed to operate in environments designed for people and, where safety requirements are met, alongside human workers. 

General-purpose robotic architectures: wheeled, quadrupedal, and humanoid systems - Source: McKinsey & Company. 

Why, then, build a robot in human form? The answer is not aesthetics, but compatibility. Many factories, warehouses, and commercial facilities are already designed around human reach, movement, tools, and workstations. A humanoid could therefore extend automation within existing brownfield environments while reducing, rather than eliminating, the need for major facility redesign.  

The near-term value proposition appears strongest in repetitive, physically demanding, or hazardous tasks that are difficult to automate economically with fixed, single-purpose systems. In material handling, machine tending, inspection, and component transfer, humanoids could reduce worker exposure to heavy lifting, heat, chemicals, or repetitive strain, while moving between workstations and adapting to changing layouts. This flexibility may also improve the economics of automation: instead of installing a dedicated machine for each task, one platform could potentially support several workflows as demand changes. 

2. Beyond the Hype: Is Commercial Demand Emerging? 

The clearest evidence of humanoid demand is beginning to come from operating environments. At BMW Group Plant Spartanburg, Figure 02 moved from laboratory training into an active automotive production line. During its 2025 deployment, the robot retrieved and positioned sheet-metal components for welding, supported the production of more than 30,000 BMW X3 vehicles, moved over 90,000 parts, and accumulated approximately 1,250 operating hours and 1.2 million steps (BMW Group, 2026).  

Commercial activity is also beginning to extend beyond individual pilots. Following an initial proof of concept, GXO and Agility Robotics entered a multi-year Robots-as-a-Service agreement to deploy Digit in live logistics operations (Agility Robotics, 2024). At GXO’s SPANX facility, Digit works alongside existing autonomous mobile robots, moving totes from other robotic systems onto conveyors. Mercedes-Benz has separately entered a commercial agreement with Apptronik to test Apollo for parts delivery, component inspection, and the movement of kitted totes within manufacturing facilities (Apptronik, 2024). These programmes remain limited in volume, but they represent a stronger demand signal than prototypes alone: customers are committing facilities, workflows, engineering resources, and operating time to determine whether humanoids can create measurable value. 

Taken together, the strongest demand signals are emerging in factories and logistics facilities. The task categories most frequently tested in live industrial environments include material movement, component transfer, package handling, machine tending, and routine inspection. In these settings, humanoids are not replacing all existing automation; rather, they are being tested as a flexible layer between fixed industrial robots, wheeled mobile systems, and human workers, particularly where a task requires both mobility and manipulation within infrastructure designed for people. 

Market signals point to rising confidence in humanoid robotics, although realised demand remains at an early stage. According to McKinsey & Company (2025), investor interest in general-purpose robotics, including humanoids, has accelerated sharply. Annual funding increased fivefold between 2022 and 2024 to exceed US$1 billion, with most capital concentrated in China and the United States. Meanwhile, Mordor Intelligence (2026) projects the global humanoid market to grow from US$3.93 billion in 2026 to US$17.80 billion by 2031, representing a 35.26% CAGR.  

Humanoids Market - Source: Mordor Intelligence 

At the same time, Morgan Stanley (2025) estimates that the wider humanoid economy could exceed US$5 trillion by 2050, with more than one billion units in operation and approximately 90% deployed in industrial and commercial settings. These figures indicate substantial market expectations, but commercial agreements, repeat deployments, and operating performance remain the more reliable evidence of actual demand. 

Taken together, the market is moving beyond experimentation, but its commercial shape remains highly selective. Demand is real, yet it is emerging primarily in narrow, structured industrial workflows, especially within factories and logistics facilities, rather than in open-ended, general-purpose applications. More complex use cases, including household assistance and operation in highly variable environments, remain largely prospective. Near-term adoption will therefore be driven by platforms that can perform a limited set of valuable factory tasks reliably, not by robots claiming universal capability. 

3. From Pilot to Scale: Four Commercialisation Gaps. 

The emergence of industrial pilots shows that humanoids can perform useful work under defined conditions. Commercial scale, however, requires a higher standard: robots must operate safely alongside people, remain productive throughout a working shift, perform tasks with sufficient mobility and precision, and deliver economics that justify deployment beyond a single site. McKinsey & Company (2025) frames these requirements as four bridges between pilot validation and repeatable commercial adoption.  

Safety readiness is the first condition for wider deployment. Humanoids will need to operate in shared workspaces without depending on constant supervision or extensive physical separation. This requires vision, proximity detection, tactile sensing, force-limited actuation, compliant joints, and fall recovery to function as a coordinated safety system. Technical safeguards must also be supported by consistent testing and recognised certification pathways; until then, many deployments are likely to remain partially segregated.

Sustained uptime determines whether a robot can become a productive asset rather than an intermittent demonstration. Current humanoids typically operate for only 02 to 04 hours per charge, compared with the 08 to 12 hours expected in many industrial shifts. Battery swapping and fast charging offer practical near-term responses, while lighter structures, more efficient transmissions, stronger thermal management, and faster fault recovery can extend productive time further.  

Dexterity and mobility define the range of work a humanoid can perform. Current systems can already support transport, basic handling, and inspection, and low-variability conditions but remain materially behind humans in fine manipulation and adaptability. Human hands possess approximately 20 to 27 degrees of freedom, while robotic hands generally offerfewer independently controlled movements and less effective tactile feedback. Reliable manipulation also requires the continuous integration of vision, touch, force, balance, and real-time learning, capabilities that remain strongest in structured settings.  

Cost competitiveness ultimately determines whether technical capability can translate into broad adoption. A humanoid may perform a task successfully, but customers will not expand deployment unless its cost compares favourably with human labour, fixed automation, or other robotic alternatives. Affordability depends not only on lower component prices, but also on utilisation, maintenance, serviceability, and the amount of supporting infrastructure required. 

Although the four bridges address different dimensions of commercial viability, each is influenced by the maturity of the underlying component supply chain, with the most direct impact on cost reduction.

Safety depends on reliable sensors, force-controlled actuators, redundant control systems, and components that can be validated consistently. Uptime is shaped by battery performance, power electronics, thermal management, spare-part availability, and ease of maintenance. Dexterity and mobility rely on precision actuators, gear systems, tactile sensing, and lightweight structures that can deliver repeatable motion at scale. Cost competitiveness, meanwhile, depends on whether these components can be standardised, sourced from multiple qualified suppliers, and manufactured at sufficient volume. 

The humanoid bill of materials shows where these dependencies are concentrated. Five hardware domains account for approximately 85% to 90% of total unit cost. Actuation represents an estimated 40% to 60%, followed by sensing and perception at 10% to 20%, compute and control at 10% to 15%, structural components at 5% to 10%, and battery modules at 5% to 10%.  

Component cost by level of differentiation in humanoid robots Source: McKinsey & Company. 

This also provides part of the context why the leading ecosystems are pursuing different routes to commercialisation: China is drawing on manufacturing depth and rapid field deployment, while the United States is building from strengths in AI, simulation, and software-led autonomy. 

4. Two Paths to Scale: China and the United States.

As humanoid robotics moves from pilot deployment toward commercial scale, competition is increasingly taking place at the ecosystem level.  

China and the United States approach this challenge from different starting points, although the distinction is one of relative emphasis rather than an absolute division. China benefits from manufacturing depth, component availability, and rapid physical deployment, while the United States draws more heavily on strengths in artificial intelligence, computing infrastructure, simulation, and software-led autonomy. Both are seeking the same commercial flywheel: improved hardware and models enable wider deployment; deployment generates operating data; and higher production volumes support further performance improvements and cost reduction. 

China’s model is supported by the depth of its industrial base. According to the International Federation of Robotics (IFR, 2025), China remained the world’s largest industrial robotics market in 2024, installing a record 295,000 units and accounting for 54% of global deployments. This was nearly six times the 50,100 units installed across the entire Americas, where installations exceeded 50,000 for the fourth consecutive year but declined 10% from 2023. China’s operational stock exceeded two million robots, while domestic manufacturers supplied 57% of its home market, up from approximately 28% a decade earlier.  

China’s humanoid sector can draw on established electric-vehicle, industrial-robot and electromechanical supply chains for motors, harmonic drives, batteries, power electronics, sensors and precision components. The country processes around 90% of the world’s permanent magnets, while dense manufacturing clusters in Shenzhen, Suzhou, Hangzhou and Ningbo allow OEMs to source alternatives and revise designs through relatively short production cycles. McKinsey & Company has also cited recent data indicating that approximately 7,700 humanoid-related patents over the past five years, reinforcing a model in which higher production volumes, faster hardware iteration and real-world data collection support one another.  

The United States approaches the same problem from the intelligence layer. Its ecosystem includes Google, Nvidia and Tesla, alongside specialised companies such as Physical Intelligence, Figure AI, Apptronik and Agility Robotics. Nvidia’s open-source GR00T foundation model and investments across the robotics sector strengthen a software infrastructure focused on autonomy and capability transfer across tasks and platforms. In 2024, US private AI investment reached approximately US$109 billion, nearly twelve times China’s US$9.3 billion, illustrating the difference in capital available for frontier-model and compute-intensive development.  

US companies therefore rely more heavily on simulation, teleoperation, human demonstrations and purchased training data. This approach can advance model development before large fleets are deployed, but its commercial value depends on whether skills learned in controlled or virtual environments transfer reliably into variable workplaces. According to Bloomberg (2026), the constraint is particularly important in embodied AI: capable models may ultimately require tens of millions of hours of physical-interaction data, while leading companies are currently estimated to have accumulated only around 500,000 hours. 

The two pathways also carry different risk: China’s manufacturing scale and cost advantages coexist with restrictions on access to certain advanced computing technologies, while overseas expansion may be constrained by cybersecurity, data-governance and certification requirements. The United States has greater strength in models and computing but remainsmore dependent on international hardware supply chains and must still demonstrate that advanced intelligence can deliver dependable customer-site performance. 

The distinction between the two models is unlikely to remain absolute. Commercially viable humanoids will require both intelligence capable of adapting across tasks and hardware that is safe, reliable, and affordable enough to deploy at scale. The strongest position may ultimately belong not to the purest hardware-first or intelligence-first model, but to the ecosystem that integrates both capabilities most effectively. 

5.  Where Value May Accrue?

 

Humanoid robotics is entering a more commercially credible, but still selective, phase. Demand is forming around structured industrial workflows, while scale still depends on closing persistent gaps in safety, uptime, dexterity, and cost. The investment landscape is therefore broader than the race to build the robot itself.  

Full-stack OEMs may capture value through control of the platform, customer relationship, and embodied-data loop. Yet equally important opportunities may emerge in the enabling layers: actuators and precision components, tactile and perception systems, batteries and charging, fleet software, safety and certification, and brownfield integration. Actuation is particularly significant, accounting for roughly 40%–60% of the bill of materials.  

For investors, the strongest signals will be operational rather than promotional: pilot-to-paid conversion, customer-site uptime, intervention frequency, cost per productive hour, BOM reduction, and manufacturing readiness at scale. The companies best positioned to capture value may not be those producing the most striking demonstrations, but those whose deployments customers choose to repeat, expand, and integrate into everyday operations. 

References: 

Agility Robotics (2024), GXO signs industry-first multi-year agreement with Agility Robotics, https://www.agilityrobotics.com/content/gxo-signs-industry-first-multi-year-agreement-with-agility-robotics  

Apptronik (2024), Apptronik and Mercedes-Benz enter commercial agreement, https://apptronik.com/news-collection/apptronik-and-mercedes-benz-enter-commercial-agreement  

Bloomberg (2026), China sends robots out into the world to learn how to be human, https://www.bloomberg.com/news/articles/2026-07-15/china-sends-robots-out-into-the-world-to-learn-how-to-be-human  

BMW Group (2026), BMW Group: First humanoid robot introduced in Plant Leipzig, https://www.bmwgroup.com/en/news/general/2026/humanoid-robot-in-leipzig.html  

International Federation of Robotics (2025), Global robot demand in factories doubles over 10 years, https://ifr.org/ifr-press-releases/news/global-robot-demand-in-factories-doubles-over-10-years 

McKinsey & Company (2025), Will embodied AI create robotic coworkers?, https://www.mckinsey.com/industries/industrials/our-insights/will-embodied-ai-create-robotic-coworkers  

McKinsey & Company (2026), Turning humanoid supply chain constraints into billion-dollar wins, https://www.mckinsey.com/industries/industrials/our-insights/turning-humanoid-supply-chain-constraints-into-billion-dollar-wins  

Mordor Intelligence (2026), Humanoids market size and share analysis—growth trends and forecast (2026–2031), https://www.mordorintelligence.com/industry-reports/humanoids-market  

Morgan Stanley (2025), Humanoids: A US$5 trillion market, https://www.morganstanley.com/insights/articles/humanoid-robot-market-5-trillion-by-2050