By Allan Lagasca, Robotics Segment Strategic Program, Smart Industrials Worldwide Leader at STMicroelectronics
Robots used to live in fiction as obedient machines.
Then they became extensions of human intelligence.
Now they learn, move and adapt in the real world.
The robotics sector is entering a transformational phase, largely driven by physical AI and advancements in large language models (LLM) in computation – the processing required to enable machines to understand instructions, interpret data and generate real-time responses.
In short, we are entering an age of robots that can sense and adapt to real-world environments in real-time, much as a human would. And like your average human, they combine perception, intelligence and motion, allowing them to understand their surroundings, plan actions and respond in real time.
This rise of humanoid robots is rapidly shifting from an aspirational concept to a new frontier with growing investment from leading tech companies and manufacturers.
What underpins these advancements is an improvement in sensors, actuators, edge AI processing, and efficient power systems driven by the semiconductor industry. And across numerous industries and applications, these ‘humanoid’ robots are beginning to work safely alongside people in dynamic environments.
We’re entering an age of machines that are just like us.
Why Human-like Robots?
The industrial interest in humanoid robots is also being driven by several key factors.
The first is an increasing workforce gap in vital manufacturing, logistics, healthcare, and service sectors. These industries require physically demanding, repetitive, or labor-intensive tasks – so it’s unsurprising that they often experience labor shortages. Precision work is ideal for robots that can optimize and extend operations. But humanoid robots still retain a high level of dexterity and flexibility that allows them to more quickly adapt to changing conditions, especially compared to traditional robots that are often bolted in place.
Second is compatibility with existing workplace environments. Built to imitate humans, humanoid robots are similarly suited to operate in spaces built for humans. This means they can navigate existing spaces built for humans, such as stairs, doorways, shelving layouts, workstations, etc. without requiring costly facility redesigns. This dramatically lowers barriers for business adoption, especially in industrial settings.
Third, is that humanoid robots serve as promising testbeds for advancing biomechanics, robotics engineering, and embodied intelligence – all of which is part of Physical AI. As humanoid robots are approaching cost parity with human labor, and hardware and sensors advances are improving their abilities to respond and react to evolving environments, they are poised to join the workforce in the coming years.
Humanoid robots are already moving out of labs. The first meaningful applications are concentrating in logistics and manufacturing for material handling and simple assembly steps – but there are also early pilots in healthcare support, retail, and basic inspection roles.
The Foundations of Robot Intelligence
Physical AI refers to the integration of AI with physical systems, where AI is embedded directly into physical entities that can sense, act, and adapt in real-time. Unlike LLMs that operate in data center environments through data simulations, Physical AI functions in tandem with a physical body interacting with its environment. The core elements allowing technology to do this in real time are sensors for perception, computation for interpretation, and actuators for movement.
What makes this possible is not any single breakthrough but a tightly integrated semiconductor platform that combines sensing, compute, motion control, power management, connectivity and functional safety into deterministic, real-time systems.
True 3D perception requires depth sensing, environmental mapping, and object detection. For robots, this is achieved through MEMS and image sensors, global shutter CMOS sensors, and Time-of-Flight (ToF) modules. The first two in particular are vital in addressing challenges with diverse, occluded objects and maintaining accurate spatial representations despite sensor noise and environmental complexity. These allow robots to better track motions and changes in their environment to better react and precent accidents.
Embedded computation is where the most drastic advancements have been in the last decade. Edge AI processing & ML capabilities have notably accelerated to bring more intelligence to today’s robots. New innovations enable distributed intelligence for perception, motion planning, SLAM, and motor coordination at the edge. By reducing bandwidth and offloading low-priority tasks from the main CPU, real-time responsiveness is enhanced. And modern deep learning and reinforcement algorithms are leading to continual improvements and optimizations on the software side.
Precise motion control is more straightforward. Humanoid robots utilize motor drivers and controllers for various robot joints – such as the shoulders, neck, elbows, wrists, fingers, etc. – to replicate physical movements similar to humans.
It is the true combination and complete integration of these components that enables robots to become intelligent, autonomous agents rather than just programmed machines. A complex and comprehensive portfolio of semiconductor solutions allows a perfect integration of the mechanical and electronic systems. This allows humanoid robots to perform complex tasks with precision, safety, and intelligence in shifting environments.
In humanoid development, this system-level integration is the main technical challenge and differentiator between experimental prototypes and scalable platforms.
Remaining Challenges & The Road Ahead
There are major issues to be overcome before wide scale adoption of humanoid robots. These span not only the practical concerns related to deployment, but span software improvements, data security, cost scaling, standards and regulations. Yet these concerns are being dissipated rapidly – in fact, Goldman Sachs expects humanoid robots to be economically viable for consumer applications by 2028-2031.
Innovations across AI, advanced sensors, high-performance semiconductors, and next-generation mechanical systems are unlocking technology improvements at a rapid rate. As these technologies mature, humanoid robots will shift from isolated pilots to broad deployment across factories, hospitals, warehouses, and more. Robots will eventually become collaborators and coworkers, supporting workforces in environments that demand physicality, adaptability, and efficiency.
Humanoid robots are emerging as the next major intelligent platform following smartphones and cloud computing, driven by labor shortages, aging populations and advances in physical AI.
As physical AI matures with robots that can think, move, perceive, and collaborate with humans, the line between digital intelligence and physical capability will blur.
The convergence of advanced semiconductors, edge AI, vision systems, and actuation technology positions humanoid robotics as the key transformative technology of the next decade.