Increasing automation in industry and manufacturing, along with rising demands on connectivity solutions such as 5G, requires ever greater computing power and real-time capability. Electric and hybrid vehicles likewise rely on advanced embedded systems for motor control and battery management. All of this is driving the embedded processing market. Today’s solutions are already powerful enough to enable AI applications and machine learning for edge computing. Sensor fusion is also becoming an increasingly important driver of embedded processing. Particularly in Europe and the Asia-Pacific region, automation and smart systems are fueling growth, while power consumption continues to fall and development tools become more streamlined.
Convergence of MCUs and MPUs continues
Edge AI, IoT and Industry 4.0 require real-time data processing with AI capabilities, while keeping power consumption low, system costs under control and time-to-market short through higher functional integration. As a result, traditional microcontrollers (simple control, low power consumption) are increasingly converging with the computing power of microprocessors (MPUs). This leads to hybrid system-on-chips that combine peripherals, memory and high-performance cores on a single chip.
Trends in MCU–MPU convergence
The convergence of MCUs and MPUs enables highly integrated SoCs that combine control peripherals with high computing performance. This saves board space and reduces both cost and power consumption. Selecting the right embedded processing solution requires the correct balance between real-time capability (a traditional MCU strength) and AI-capable performance (an MPU strength). Thermal management must also be considered. Hybrid architectures benefit from dedicated software ecosystems, including open-source toolchains. Key design aspects include selecting NPUs for edge AI, ensuring RISC-V and Arm compatibility, and validating determinism in IoT and automotive applications to optimise latency and energy efficiency.
“Due to ever-increasing energy efficiency, neural network accelerators can now be integrated into both MPUs and MCUs.”
- Ulrich Schmidt, Director Embedded Processing Technology, EBV
AI moves into the device
Edge AI and the integration of artificial intelligence into embedded systems are gaining enormous importance, as they enable real-time processing, reduce latency and improve both cybersecurity and energy efficiency – particularly in IoT, autonomous systems and industrial applications. Current embedded processing trends are creating the foundation for this development: specialised AI accelerators such as NPUs, GPUs, FPGAs and integrated microcontrollers deliver high TOPS performance at low power consumption, supported by model compression, quantisation and hybrid edge-cloud architectures.
Trends in AI Integration
Dedicated NPUs integrated into SoCs deliver high TOPS-per-watt performance for efficient edge inference
Techniques such as 4-bit quantisation and TinyML reduce computational and memory requirements, enabling AI on resource-constrained MCUs
Scaling to 7 nm, 5 nm and 3D stacking increases transistor density and bandwidth for compact, high-performance edge hardware
Chiplet-based combinations of CPUs, GPUs, NPUs and FPGAs enable optimised parallel processing of AI workloads
TOPS – Tera Operations Per Second TOPS measures how many basic AI-related operations a processor can perform per second.
Integrating AI into devices also increases thermal load due to parallel workloads, raises power demand despite optimisations and introduces additional security requirements, such as protection against side-channel attacks. PCB layout must therefore prioritise cooling and ensure mixed-signal integrity. Model quantisation and pruning techniques are essential, as are robust firmware updates, hybrid edge-cloud interfaces and validation tools for TinyML to ensure real-time capability and reliability under constrained hardware resources. At chip level, heterogeneous architectures such as chiplets and 3D stacking are key, complemented by neuromorphic approaches for maximum efficiency.
New technologies reduce the energy demand of embedded systems
Demand for energy-efficient embedded systems continues to grow, driven by the IoT boom, increasing need for low-power solutions in Industry 4.0 and edge computing, and the general push towards energy-efficient devices and systems. The use of low-power processors, optimised software and advanced power management techniques can significantly reduce energy consumption. Low-power microcontrollers alone can reduce energy usage to around 30 percent compared to standard MCUs through technologies such as subthreshold operation. New models combine this with AI-optimised power management, achieving efficiency gains of up to 50 percent in edge computing applications.
Current technology trends in energy-efficient embedded systems:
In general, energy consumption in embedded systems can be reduced through low-power MCUs with sub-1 µA sleep modes, dynamic voltage and frequency scaling, efficient software optimisation such as event triggering instead of polling, and energy harvesting. When integrating low-power MCUs, designers must consider peripheral compatibility, precise power management algorithms, thermal analysis and real-world load testing to ensure stability and efficiency.
Embedded Processing at EBV Elektronik: Comprehensive Support, Tailored for Engineers & Leaders
EBV Elektronik empowers manufacturers, developers, and integrators with a complete embedded processing ecosystem — from low-power MCUs to edge AI accelerators, supported by custom software, industrial-grade hardware, and global production services. For engineers, this means faster development, improved reliability and scalability; for decision-makers, it’s reduced risk, lower costs, and a strong innovation edge.
“An important trend in embedded processing is the convergence of MCUs and MPUs. More energy-efficient MPU cores are now addressing markets that were previously covered by powerful MCUs.”
- Ulrich Schmidt, Director Embedded Processing Technology, EBV
