25 September 2026 | Interaction | By Editor Robotics Business NEWS <editor@rbnpress.com>
Humanoid robotics is moving from an era dominated by prototypes and demonstrations toward early commercial deployment, with manufacturing and logistics emerging as important starting points. In this Robotics Business News interview, Julia Wahlen, Market Analyst at Berg Insight, examines the technologies and market forces shaping this transition. She discusses the growing role of embodied AI, the importance of real-world training data, challenges around reliability and production costs, and the expanding ecosystem of AI, semiconductor, manufacturing and industrial technology partnerships. Wahlen also explores connectivity, fleet management and the long-term growth outlook for humanoid robots as the industry moves toward larger-scale deployment.
How do you see the humanoid robotics industry evolving as it moves from prototypes and pilot projects toward large-scale commercial deployment?
The humanoid robotics industry is now starting to move beyond the prototype stage. During 2025 and 2026, several developers have moved into pilot production, customer trials and early commercial deliveries. However, there is still a considerable gap between demonstrating a capable robot and deploying thousands of units that can perform useful work reliably every day.
We expect the first larger deployments to take place in manufacturing and logistics. These environments are relatively structured, which makes it easier to introduce humanoid robots gradually and focus on specific tasks. Material handling, machine tending, inspection and certain assembly tasks are examples where we are already seeing interest.
These early deployments will also be important for the development of the technology itself. Humanoid robots need large amounts of real-world interaction data to improve their AI models, so putting robots into actual working environments gives developers both commercial experience and valuable training data.
Cost will naturally be important as well. Humanoid robots are still expensive to produce, partly because many components are manufactured in relatively small volumes. Higher production volumes, more specialised suppliers and greater standardisation should gradually bring costs down.
Over time, we expect humanoid robots to move into less structured environments. Service and retail applications are likely to become more relevant, followed by broader domestic use. Homes are particularly difficult environments for robots because they vary considerably and require robots to handle many different objects and unexpected situations.
What role does embodied AI play in making humanoid robots more capable, autonomous, and useful in real-world industrial environments?
Embodied AI is a key part of making humanoid robots useful beyond highly predefined tasks. Conventional industrial robots are very effective when performing the same operation repeatedly in a controlled environment. Humanoid robots are intended to be more flexible.
They need to perceive their surroundings, understand a task and decide how to physically perform it. Recent developments in foundation models and Vision-Language-Action models are helping connect perception, reasoning and physical action. Instead of programming every movem individually, the aim is for robots to understand higher-level instructions and adapt their actions to the situation.
One of the main bottlenecks is training data. Large language models have been trained on enormous amounts of text and images available online. Equivalent datasets do not exist for physical interaction. Data showing how a robot should pick up an unfamiliar object, use a tool or respond when something changes in its surroundings is much more difficult and expensive to collect.
This is why simulation and synthetic data have become so important. Robots can practise tasks many times in simulated environments before those capabilities are transferred to physical systems. Real-world deployments can then provide additional data that is used to refine the models.
There is still a long way to go before humanoid robots can autonomously handle the full range of situations encountered by people. At the moment, the development of embodied AI is largely about increasing the number of tasks robots can perform reliably and reducing how much task-specific programming is required.
What are the biggest challenges your company faces in scaling humanoid robot production while maintaining reliability, safety, and cost efficiency?
As Berg Insight is a market research company rather than a humanoid robot manufacturer, we see this question from an industry perspective. From our research, reliability, cost and manufacturing scale are some of the biggest issues developers need to solve.
A robot performing well in a demonstration is not the same as a robot operating for many hours every day in a factory. Commercial customers need consistent performance and high uptime. Current humanoid robots still have limitations in areas such as manipulation, locomotion, battery runtime and the ability to respond reliably to unexpected situations.
Cost is closely connected to this. Hardware currently accounts for a very large share of the cost of a humanoid robot. Actuators are particularly expensive and can account for a substantial part of the bill of materials. As the supplier ecosystem matures and production volumes increase, we expect these costs to come down.
There are also difficult engineering trade-offs. Increasing battery capacity can improve runtime but adds weight. More computing power can improve perception and AI capabilities but also increases power consumption and cooling requirements. Improvements therefore have to be considered at the level of the complete robot rather than one component at a time.
Safety becomes increasingly important as robots move out of controlled testing environments. Humanoids are ultimately designed to operate around people, often without physical barriers. This puts high demands on perception, motion control and the reliability of the overall system.
Scaling production therefore involves more than simply increasing manufacturing capacity. The robots also have to become more reliable and affordable at the same time.
Which industries and applications do you believe will drive the earliest large-scale adoption of humanoid robots, and why?
We expect manufacturing to be the main application area in the early market, together with logistics and warehousing.
Factories provide relatively structured environments and often contain repetitive or physically demanding tasks that are suitable for automation. Material handling, machine tending, inspection and selected assembly tasks are some of the more obvious examples.
There is also a clearer business case in these environments. A manufacturer can evaluate how many hours a robot operates, what tasks it performs and what costs it replaces or reduces. This makes it easier to assess whether the investment makes economic sense.
One reason humanoid robots are interesting for manufacturing is that factories are already designed around people. A humanoid robot could potentially use existing tools, workstations and infrastructure without requiring the same degree of redesign that some traditional automation systems require.
We expect service and retail applications to grow as the technology becomes more capable. Domestic robots are likely to take longer to reach large volumes because homes are much less predictable. A household robot has to deal with different layouts, objects, people and situations, while also meeting very high safety and reliability requirements.
How important are strategic partnerships with AI, semiconductor, manufacturing, and industrial technology companies to accelerating humanoid robotics commercialization?
Partnerships are already playing a major role in the industry. A humanoid robot combines a very broad range of technologies, from AI and semiconductors to actuators, sensors, batteries and manufacturing systems. Developing all of these technologies internally would require enormous resources.
We are seeing different approaches. Some humanoid developers are highly vertically integrated and design many important components themselves, while others rely more heavily on specialised suppliers. Even the more vertically integrated companies remain dependent on a wider technology and manufacturing ecosystem.
AI and semiconductor partnerships can provide access to computing platforms and models. Manufacturing partners can help companies move from prototypes to larger production volumes. Component suppliers become increasingly important when thousands rather than tens of robots need to be produced.
Industrial customers are just as important. Deploying robots in factories and warehouses allows developers to test their systems under real operating conditions. It also provides feedback and data that can be used to improve both the hardware and AI.
As the market matures, we expect the surrounding ecosystem to become much larger. System integrators, software companies, component suppliers and maintenance providers will all have roles to play. Humanoid robotics is unlikely to develop as an isolated industry.
What role will connectivity, including cellular networks, cloud platforms, remote monitoring, and fleet management, play in the future operation of humanoid robot fleets?
Connectivity will become more important as companies move from operating individual robots to managing larger fleets.
The robot itself still needs to handle time-critical functions locally. Balance, obstacle avoidance, perception and motion control cannot depend on a continuous connection to the cloud. A humanoid robot has to be able to respond immediately to what happens around it.
Cloud platforms are more useful for functions that are less sensitive to latency. This could include model training, software updates, analysis of operational data and coordination across a fleet. Data collected by deployed robots can also be used to improve AI models and distribute updated capabilities to other robots.
Fleet management will become particularly important once companies operate hundreds or thousands of units. Operators will need to monitor battery levels, system status, maintenance requirements, task completion and software versions. Remote assistance may also be needed when a robot encounters a situation it cannot resolve autonomously.
Cellular connectivity and private 5G networks could be useful in large factories, warehouses and other industrial sites where robots need reliable connectivity while moving across the facility.
We therefore expect a combination of onboard computing and cloud infrastructure rather than either approach replacing the other. Immediate decisions will largely remain on the robot, while the cloud supports areas such as fleet management, data analysis and model development.
With forecasts pointing to a potentially US$554 billion humanoid robotics market by 2040, what is your vision for the industry over the next decade, and what will determine which companies emerge as market leaders?
The next decade will be an important period for the humanoid robotics industry. Berg Insight estimates that annual shipments will grow from approximately 16,000 units in 2025 to around
1.2 million units in 2030. We expect growth to accelerate further during the 2030s, reaching around 26 million annual shipments and an installed base of almost 83 million units by 2040.
There are still many uncertainties behind a forecast over this time horizon. The speed of adoption will depend on how quickly developers can improve reliability, autonomy and battery performance while bringing production costs down.
It is also too early to say which of today's companies will eventually become the largest players. The market is developing very quickly, with strong activity in both China and the US as well as a growing number of European companies.
The companies that are best positioned will need more than a capable robot. AI and access to real-world training data will be important, but so will hardware reliability, manufacturing capabilities, supply chains and access to capital. Companies also need customers willing to deploy robots in real environments.
Manufacturing scale could become an increasingly important differentiator over the next few years. Several companies are now preparing for significantly larger production volumes, and the ability to manufacture robots consistently and at a competitive cost will become more important as the market moves beyond pilot projects.
The industry is still at an early stage, and many of the robots being developed today will change substantially over the coming decade. What we are seeing now is the beginning of the shift from humanoid robots as research platforms towards products that need to prove their value in everyday commercial operations.
