Contributed by Luca Verre, group vice president, head of Strategy, MDRF Products Group
There is a growing misconception about robotics: that the next generation of machines will be defined by a bigger brain. It is an appealing idea, but it is incomplete. In the real world, robots do not succeed because they have one powerful processor. They succeed because they can sense, decide, and act at the right place and at the right time
That is the real shift now underway in artificial intelligence. AI is leaving the cloud, moving beyond the edge, and taking physical form inside machines that move through the world. This is physical AI: intelligence that must operate in real time, under power constraints, in dynamic environments, and often in close proximity to people.
Once AI enters the physical world, the rules change. Latency becomes a safety issue. Power becomes a systems issue. Reliability becomes non-negotiable. A robot arm cannot wait for a remote server to interpret a signal. A vehicle cannot depend on a round trip to the cloud to maintain stability. A humanoid robot cannot afford uncertainty in sensing, balance, or motion control. The future of robotics will not be built on centralized intelligence alone. It will be built on distributed intelligence across the entire machine.
This is where semiconductors matter most.
Physical AI depends on a chain of functions that must work together seamlessly: sensing, embedded processing, power delivery, actuation, and connectivity. If any one of those layers is too slow, too power-hungry, or too brittle, the system fails. That is why the industry is moving away from passive sensing architectures, where raw data is simply streamed to a central processor, and toward more intelligent systems in which information is interpreted locally and only what matters is passed upstream.
This is a fundamental rethinking of how machines perceive the world.
At STMicroelectronics, we see this shift as a defining opportunity. ST is building the foundation of intelligent sensing for AI. As the #1 European IDM in optical sensing and a global leader in MCU and MEMS, we are well positioned to support the next generation of robots, vehicles, and industrial machines with technologies that sit at the point where the physical and digital worlds meet.
What gives ST an edge is not a single product. We combine one of the industry’s broadest sensing portfolios with embedded processing, edge AI tools, analog and power technologies, and the manufacturing control of our IDM model. That matters because robotics is a systems problem. The interaction between sensor, processor, power stage, and actuator is where performance is won or lost.
In practical terms, that means moving from a sensor that merely reports motion to a sensing system that helps classify it; from an MCU that runs code to an embedded platform that can infer locally; from a power device that simply drives a load to a power architecture that supports efficient, responsive motion.
This kind of integration is especially valuable in Physical AI because the system itself is the product. A robot is not built around one chip. It is built around a network of signals and decisions, all of which must remain stable while the machine moves through a messy, uncertain, human environment.
That is why humanoid robots are attracting so much attention. They are not just a futuristic concept. They are a stress test for everything Physical AI must solve. A humanoid must perceive distance, texture, movement, force, and human presence. It must balance continuously. It must react without hesitation. It must do all this while keeping power consumption, cost, and safety within bounds.
The market is still early, and it would be a mistake to pretend otherwise. But the direction is clear. Humanoids are emerging as a serious growth opportunity, and the semiconductor content per system is substantial. At ST, we estimate that a typical humanoid platform can represent around $600 of addressable semiconductor content, depending on the architecture. That includes sensing, embedded intelligence, power management, motor control, and connectivity. The content opportunity is distributed across the machine.
That distribution is exactly why the semiconductor industry matters so much in this next phase of AI. Physical AI needs companies that can help machines understand the world, respond in real time, and do so reliably at scale.
ST already operates at the junction of sensing, embedded processing, power, and control. We have the technology breadth, the customer engagement model, and the manufacturing control to support Physical AI as it moves from prototypes to deployment. That combination gives us a strong position in consumer sensing today which is expanding to more demanding industrial, automotive, and robotic systems tomorrow.
The next wave of AI will not be defined only by larger models or faster data centers. It will also be defined by how intelligence is distributed into the machines.
That is why robotics is moving beyond the brain, and why the companies that can make the entire system smarter will be at the forefront of the next wave of innovation.
For more information, visit: Humanoid robots | Application – STMicroelectronics
