Humanoid robots: The next frontier in physical AI

The evolution of humanoid robots

Morgan Stanley Research estimates that the market for humanoid robots is likely to reach about $5 trillion by 2050, representing more than one billion humanoid robots in service. Although adoption is expected to be modest into the 2030s, the adoption rate is projected to accelerate dramatically through the 2030s and 2040s.

While a booming technology today, humanoid robots can trace their early conceptual evolution back centuries. An abridged timescale for humanoid robot evolution includes the following:

Core capabilities of modern humanoid robots

Humanoid robots exhibit an array of capabilities that focus on advanced AI for independent learning and decision-making, along with physical design characteristics that enable the robot to interact seamlessly in human environments. A broader scope of core capabilities includes the following:

Perception

Robots use perception to understand and operate safely within their environments. Perception involves an array of advanced sensors, including high-resolution color cameras; distance-sensing and mapping devices such as LiDAR or radar; human-sensing devices such as thermal imaging sensors; microphones; and tactile sensors that enable robots to feel contact. These sensors provide data to the AI system, enabling the robot to respond to changing environments or situations.

Interaction

Modern humanoid robots are increasingly capable of working alongside humans. These robots focus on natural language processing, speech recognition and expressiveness. This is intended to make humans more comfortable and support intuitive human-robot interaction (HRI).

Mobility and manipulation

Humanoid robots achieve human-like movement by using complex actuators and controllers. Today's robots can mimic the fluidity of human movement across a range of everyday tasks, enabling robots to walk, run, jump, navigate stairs and uneven terrain, manipulate doors, grasp containers and perform precise tasks such as delicate assembly and surgical procedures.

AI autonomy

Humanoid robot capabilities are driven by machine learning (ML) and AI. Agentic AI enables robots to perceive, analyze, reason, act and learn with a high level of autonomy. Humanoid robots can adapt and learn by observing human actions as well as reinforcement learning techniques.

Versatility

Humanoid robots are devices capable of performing a wide range of tasks in human environments, such as homes, offices and factories. However, they might lack the speed or sheer strength found in industrial robots.

5 real-world use cases of humanoid robots in 2026

Humanoid robots are here. Leading robotics developers are beginning to field practical robots for a variety of real-world tasks:

1. Logistics

Humanoid robots can handle physical tasks such as sorting and palletizing, picking, packing, loading and unloading. They can also support other inventory tasks such as material transport, labeling and manifesting.

Humanoid robots can operate in unstructured warehouse environments and interact with infrastructure such as aisles, shelving, boxes and barcode scanners, while avoiding objects, other robots and human counterparts. Humanoid robots can also use vision capabilities to perform basic quality control, identifying objects for correctness before accepting them for inventory.

Examples of humanoid robots for logistics tasks include Agility Robotics' Digit, Figure AI's Figure 03 and RobotEra's L7.

2. Manufacturing

Humanoid robots can assemble products. This process involves gathering and sorting components, using tools to assemble them and then packaging the build. At each step, humanoid robots can apply visual observation and measuring to ensure the quality of each part or the functionality of the completed build.

Other examples of manufacturing tasks include polishing or applying paints and sealants with precision to eliminate overspray and drips; performing precise drilling and fastening, such as in automotive body assembly; handling sensitive composite materials and packaging pharmaceuticals in sterile environments. The scope of possible manufacturing tasks is almost limitless, depending on business types and needs.

Examples of humanoid robots for manufacturing tasks include the Apptronik Apollo, Figure AI's Figure 03 and RobotEra's L7.

3. Other industrial tasks

Industrial tasks can be tedious or hazardous for humans. In addition to logistics and manufacturing, there are numerous industrial examples of maintenance and remediation where a robot's human form can be an advantage for navigating and working in human-designed environments.

Humanoid robots can perform inspections and basic maintenance in hazardous areas, such as power plants or locations where hazardous materials pose a danger to people. They can also perform basic custodial tasks such as cleaning and sanitizing medical facilities or collecting trash and moving waste.

Examples of humanoid robots for industrial tasks include Boston Dynamics' Atlas, Tesla's Optimus Gen2, NEURA Robotics' 4NE-1 and DEEP Robotics' Dr02.

4. Commercial and service industries

Humanoid robots also hold enormous potential in commercial and service businesses, including office, hotel, retail and healthcare environments, where the emphasis is on human interaction, contextual understanding and support.

Retail stores can use humanoid robots to stock shelves and assist shoppers in locating items or answering questions. Hotels and other hospitality establishments can use robots to offer concierge services. In offices, humanoid robots can provide reception duties, make deliveries, handle basic cleaning and maintenance, restock consumable items and tackle scheduling tasks.

Humanoid robots in healthcare can perform tasks such as patient monitoring and companionship, medication reminders, bathroom assistance, physical therapy support, surgical and other procedural assistance, handling of infectious samples and routine cleaning.

Examples of humanoid robots for commercial services include Unitree Robotics' G1, Figure AI's Figure 03, Sanctuary AI's Phoenix, and Engineered Arts' Ameca.

5. Households

There is broad potential for humanoid robots in homes. Smaller robots are well-suited for the dynamic and confined environments that many everyday households present. Advanced AI and complex sensors enable robots to navigate the environment and the people and pets that move within it.

Everyday tasks humanoid robots can perform include vacuuming, mopping, sorting and folding laundry, loading and unloading dishwashers, gathering cooking ingredients, serving food, preparing drinks and organizing items.

Beyond chores, humanoid robots in the household can assist as information interfaces. For instance, they can deliver information on weather and driving conditions before a spouse leaves for work, remind family members of scheduled activities or make reservations at a favorite restaurant.

Humanoid robots can also provide increasing levels of assistance to young, elderly, injured or ill humans, offering timely medications, bathroom assistance and other support for a variety of human needs and situations.

Examples of humanoid robots for household service include 1X's NEO, Tesla's Optimus Gen2 and NEURA Robotics' 4NE-1.

Challenges in humanoid robotics

Despite the enormous promise of humanoid robots, users must consider numerous challenges before adopting them. Common challenges of humanoid robots include the following:

AI mistakes and hallucinations

AI lacks fundamental common sense and is prone to making mistakes. The real challenge with humanoid robots is that AI tends to guess if it doesn't know something, and the consequences can be deadly. For example, an uncertain or hallucinating robot could bring the wrong medication to an elderly patient, or the wrong food to someone with a specific food allergy.

The AI systems that humanoid robots use should be able to identify uncertainty and then take cognizant steps to find correct and reliable answers. AI must also have a means of checking for and correcting hallucinations, thereby establishing common sense in humanoid robots and their decision-making processes.

Cybersecurity

Cybersecurity is perhaps the most alarming challenge in humanoid robots. Robots will work alongside humans, which demands a high level of trust. However, robots often interconnect and depend on advanced networking. Security vulnerabilities in robots and their interconnecting networks can enable hackers to steal data, take control of robots and drive virus-like botnets. This poses unprecedented dangers to human safety.

Energy efficiency and power costs

The motors and actuators for bipedal locomotion and other human-like articulations -- such as arms, hands and fingers -- often demand significant amounts of energy. This means that humanoid robots must include advanced battery technology and require efficient, seamless charging mechanisms. However, even then, operating time will be finite, and the electricity robots need will incur operating expenses.

Training

Users must train humanoid robots to perform their assigned tasks. Fundamental training might be standard, but robots require extensive training to perform complex job-specific tasks. These robots will need human guidance to teach them how to perform, refine and update their behaviors. This requires human expertise; therefore, the methodologies and processes used to administer robot training must be carefully developed and documented.

Ethical and societal concerns

The presence and use of humanoid robots will lead to significant ethical and societal implications. There are concerns about the potential for human job displacement. There are also legal or legislative issues about robot ownership and the liabilities resulting from robot actions. These issues have been barely discussed at the legislative level, and few frameworks exist to address the unintended consequences of robot behaviors or mistakes.

Availability of advanced materials

Humanoid robot components depend on the availability of advanced materials. Rare earth metals like neodymium and dysprosium are necessary for magnets in motors; lithium, cobalt and nickel are needed in batteries; and advanced materials such as titanium alloys, carbon composites and specialized polymers are essential for structural parts. Some of these materials are difficult to manufacture or are so rare that their supply might be disrupted by geopolitical contention. This can lead to limited and expensive robotic components.

Humanoid robots and safety

Although humanoid robots can operate in human-built environments -- often alongside humans -- the need for safety is crucial for the adoption and acceptance of these robots.

One inescapable reality is that robots are machines, and machines fail. Designers must understand how their robots can fail and then implement a multi-layered approach that ensures safety during both normal operations and unexpected failures. Consider the following:

• Decision-making. The first layer of safety is decision-making. Humanoid robots must use sensory data from AI algorithms to make safe decisions. For example, a humanoid robot must perceive and avoid humans, other robots and objects within the environment. This demands dynamic, acute collision prediction and avoidance capabilities.
• Fail-operational designs. Techniques such as redundancy and fail-operational designs are essential to avoid simple emergency stops, which can result in falls, spills, crushing and impacts. If a humanoid robot is carrying a load and a problem is detected, the safest behavior might be to place the load safely, then place the robot into a safe state and address the root of the problem.
• Emerging safety standards. AI must integrate existing and emerging standards for machine safety, such as ISO 13849-1:2023, which regards the safety of machinery and safety-related parts of control systems. These standards must evolve to recognize human interaction in a wide assortment of dynamic environments.

The future of humanoid robots

Humanoid robots involve a mix of integrated technologies, including AI, sensors and actuators. These technologies are advancing rapidly and converging to make humanoid robots more attainable in various settings, including factories, homes and the healthcare industry.

The demand is noteworthy. Goldman Sachs Research found that the global market for humanoid robots could reach $38 billion by 2035. The global population decline is creating numerous opportunities for tasks that humans prefer to avoid, ranging from factory work to elder care. Morgan Stanley Research predicts that as many as one billion humanoid robots will be in service by 2050, and major players, such as Tesla and Figure, are leading the development of affordable, resilient and self-learning robots. Analysts expect the cost per humanoid robot to decrease to $13,000 to $17,000 by the early 2030s, dramatically lowering the purchasing cost per robot.

However, serious concerns remain. There are ethical and societal impacts of humanoid robots to contend with, as well as the more dangerous cybersecurity threat. Robots depend on networks, and flaws in network security or communication protocols can be catastrophic. Considering the sheer number of potential humanoid robots in homes, offices and factories in the years ahead, the possible dangers of robotic security vulnerabilities are serious.

Stephen J. Bigelow, senior technology editor at TechTarget, has more than 30 years of technical writing experience in the PC and technology industry.