Robots are becoming the physical arm of AI

In robotics, AI-optimised edge processors and SoC architectures handle the local evaluation of sensor, image and motion data directly in the system. This makes robots faster, more autonomous and more responsive, because latency falls and cloud connections become less critical.

The most important trends include specialised AI accelerators in the chip, stronger integration of CPU, GPU, NPU and image processing in a single SoC, and energy-efficient designs for mobile and collaborative robots. Added to this are more compact form factors, higher memory bandwidth and optimisations for real-time inference, for example for navigation, object recognition and quality inspection. In parallel, the term “physical AI” is gaining weight, because AI works directly in machines, vehicles and robots as part of the control logic.

In semiconductor technology, three developments are particularly relevant: advanced packaging processes that enable more computing power in a small area, a stronger focus on energy efficiency, and closer coupling of computing and sensor systems. For robotics applications, what is decisive is therefore less the maximum peak performance than the ability to run AI robustly, economically and in real time at the edge of the network.

Wide-bandgap power semiconductors, in particular gallium nitride (GaN) and silicon carbide (SiC), enable compact, highly dynamic and energy-efficient drives. Thanks to their higher switching frequencies and lower losses compared with classic silicon, motors and gearboxes can be built smaller, which is particularly decisive for mobile, collaborative or humanoid robots.

The most important trends include the integration of GaN-based drivers directly into joint modules, the increase in power density through higher voltages (up to 800 V and more), and the combination with advanced cooling concepts and Ethernet-based real-time communication architectures. In parallel, SiC solutions are gaining importance where high continuous loads and thermal robustness are required, for example in heavy industrial robots or automated guided vehicles.

Particularly relevant at present is the further development of GaN HEMTs and SiC MOSFETs with lower switching losses, improved EMC properties and higher reliability under dynamic load changes. In addition, new packaging technologies and system-in-package integration are driving miniaturisation, while at the same time the cost per watt is falling, a decisive factor for series readiness in robotics.

Highly integrated, safe sensor technology , including 3D imaging, forms the technological backbone of modern robotics. It enables machines to capture their environment precisely, detect obstacles and interact safely with people. Without such sensor systems, neither the sensitive control of robot arms nor the autonomous navigation of humanoid platforms in dynamic environments would be conceivable.

Current trends include the fusion of multimodal sensor data from lidar, time-of-flight cameras, radar and tactile sensors in order to create real-time 3D models of the environment. Safety-certified 3D ultrasonic sensors are also increasingly used, reliably detecting people even in poor lighting conditions and meeting standards such as IEC 61496 or ISO 13849. In parallel, optical force-torque sensors and infrared-based depth-sensing systems are gaining importance in order to ensure intuitive human-robot interaction and collision-free movements.

In semiconductor technology, development is driven above all by miniaturised, energy-efficient image sensors with integrated on-chip signal processing. New CMOS technologies enable higher resolutions with lower latency, which is decisive for safety-critical applications. Specialised ASICs and edge-AI chips that pre-process sensor data directly and thus reduce the computing load on central controls are also gaining relevance. These advances make robots not only safer, but also more compact and more economical to use.

Connected real-time data architectures based on Ethernet/TSN (Time-Sensitive Networking) and power line communication (PLC) form the neural backbone of modern robot systems: they enable deterministic, synchronised data streams between sensors, actuators and controls, and thus precise movement sequences, collaborative scenarios and seamless integration into higher-level IT and cloud systems. The most important technology trends are the convergence of wired TSN with mobile 5G networks for hybrid, real-time capable architectures, the use of standardised Ethernet hardware instead of proprietary solutions, and the integration of OPC UA over TSN for vendor-neutral interoperability. In parallel, PLC is gaining importance where additional cabling is uneconomical, for example in retrofittable or rotating robot components.

In semiconductor technology, three developments are currently driving adoption: first, integrated TSN-capable Ethernet controllers in SoCs and microcontrollers that implement real-time functions directly in the chip. Second, energy-efficient, highly integrated PLC modems with a small footprint for compact robot joints. Third, hardware-accelerated time-stamping and synchronisation logic (IEEE 802.1AS), which ensures sub-microsecond accuracy even under high network load, a basic prerequisite for multi-robot coordination and digital twins.

Quantum-resistant hardware security and an anchored root of trust are becoming the security foundation in robotics: they protect the identity, firmware and update paths of connected robots against manipulation, supply chain risks and future attacks using quantum computers. In practice, the focus is shifting from pure software security towards hardware-supported trust, because autonomous systems are increasingly coupled with the cloud, the edge and industrial networks.

The most important trends include post-quantum cryptography for signatures and key exchange, secure boot chains, hardware-based attestation, and the integration of TPMs, secure elements and HSM-like functions directly into controls and sensors. In parallel, zero-trust architectures, identity management for machines and the protection of over-the-air updates are gaining importance, because robots are operated over long lifecycles and extended regularly.

For semiconductor technology, three developments are particularly relevant: specialised security controllers, more cryptographic functions on chip, and application-oriented quantum security at silicon level. Manufacturers are working to integrate root-of-trust mechanisms directly into microcontrollers, SoCs and secure elements in order to reduce latency and shrink attack surfaces.