The MCU is hailed as the "nerve center" of modern electronic devices and is a core component that controls various functions in embedded electronic systems. At present, new business forms such as edge AI, embedded intelligence, new energy vehicles, and digital transformation of manufacturing are opening up more incremental markets for MCUS and driving technological upgrades in MCUS. MCU manufacturers have promoted the functional expansion and performance leap of MCUS through a series of innovative measures such as architectural innovation, process iteration, and toolchain, meeting the demands of downstream applications for customized design and personalized solutions. Under the collision of market demand and technological innovation, five major trends in MCUS have emerged.

    Break the limitations of eFlash and move towards processes of 22nm and below
    The MCU manufacturing process has broken through 40nm and is now moving towards 22nm, 18nm and even 16nm. In March this year, NXP released the S32K5 series of automotive MCUS, which is the industry's first 16nm FinFET MCU with embedded MRAM (Magnetic Random Access Memory). It is reported that this series of MCUS adopts TSMC's 16nmFinFET embedded MRAM technology, featuring a durability of 1 million update cycles, supporting reflow soldering, and data retention for 20 years at 150℃. Stmicroelectronics has introduced a process technology based on the 18nm fully depleted silicon on insulator (FD-SOI) process and integrated with ePCM (phase change memory). The first STM32 MCU based on this technology is scheduled to enter mass production in the second half of 2025. Compared with the currently adopted 40nm embedded eNVM (Non-volatile Memory) technology, the combination of 18nm FD-SOI and ePCM will bring more than 50% improvement in energy efficiency ratio and 2.5 times increase in non-volatile memory density. It is reported that the technology was jointly developed by stmicroelectronics and Samsung Foundry.
    It is not difficult to see that while MCU manufacturers adopt more advanced manufacturing processes, they also simultaneously embed new types of storage. Previously, the commonly used code and data memory for MCUS was eFlash. However, the traditional eFlash solution faced cost and reliability challenges at process nodes below 28nm, which also limited the process miniaturization of MCUS. Under this trend, new types of memory such as MRAM and RRAM are regarded as the main solutions for embedded storage in 28nm and below process nodes. NXP said that it takes about one minute to update 20MB of code with Flash memory, while it only takes about three seconds with MRAM, reducing the downtime caused by software updates. In addition, MRAM offers one million update cycles and has a durability far exceeding that of flash memory. Renesas Electronics' 22nm RA8P1 MCU, launched in July this year, also integrates embedded MRAM. Renesas stated that compared with flash memory, MRAM has a faster write speed, higher durability and stronger data retention capability.
   Embodied intelligence and humanoid robots have become the new blue ocean
   Embodied intelligence is regarded as the next wave of AI, and humanoid robots are the best carriers of embodied intelligence. The former puts forward new design requirements for MCUS, while the latter broadens the market capacity of MCUS. In terms of quantity, for tasks like controlling all body joints alone, more than 30 MCUS may be required. Li Yi, the product marketing director of the MCU Division of GigaDevice, said in his speech at NEPCON China that a robot has about 20 degrees of freedom and joints, and each joint has specific load requirements. When performing an action, precise coordination is required among all the joints. The GD32G553 series of MCUS from GigaDevice, which adopt the Arm Cortex-M33 core, are suitable for controlling large joints. The GD32H7 series of MCUS, which adopt the Arm Cortex-M7 core, are more suitable for fine joints that require precise motion control and have higher real-time requirements for algorithms. Some customers will use over 20 M7-core MCUS and more than ten M33-core MCUS for a single robot to perform different joint tasks. These joints communicate with each other through the built-in bus to ensure the synchronization of movement and data between each joint. Li Yi said.
    The technological paradigm of embodied intelligence, which integrates artificial intelligence into physical entities such as robots, endows them with the ability to perceive, learn and interact dynamically with the environment like humans, and also poses multiple demands on MCUS. The first is higher integration, integrating computing power, storage, ESC (EtherCAT slave controller), PHY (external signal interface chip), etc. into a more compact package. Second, it features higher real-time performance and communication data bandwidth, and can achieve low-latency communication with units such as sensors, AI inference, servos, and actuators. For instance, it can communicate with more sensors at high speed and low power consumption through the I3C serial communication protocol. The third is to build a data security mechanism from configuring trusted startup, execution to encrypted storage.
    The trend of MCU+AI is deepening
    The trend of AI moving towards the edge and edge has made edge AI a must-win territory for MCU manufacturers. To enhance the AI computing capabilities of MCUS, an increasing number of manufacturers are integrating AI accelerators such as Npus into MCUS to improve the execution speed of AI inference and training tasks. Stmicroelectronics believes that introducing NPU into MCUS will trigger the "aha moment" of new application scenarios for edge AI. Its STM32N6 is equipped with stmicroelectronics' self-developed NPU, with a computing throughput of 600 GOPS (600 billion operations per second), which is 600 times higher than that of the STM32H7 without an NPU. Compared with running neural network models such as image classification, object detection, and speech recognition on the Cortex-M55 core of STM32N6, when these neural network models are run on the NPU of STM32N6, the inference performance has increased by 26 to 134 times.
    As the complexity of AI use cases increases, multimodal scenarios are becoming more widespread. The MCU needs to continuously enhance its system integration capabilities and configure rich hardware interfaces to meet the multi-modal processing requirements in high real-time scenarios. For instance, XMOS has launched an edge multi-core controller integrating an AI accelerator, high-performance DSP, control MCU and flexible I/O, supporting audio, image, visual and other various sensor signals, enabling real-time and continuous operation of AI applications. This controller can also serve as an interface for large models, clouds and networks, providing sensor information preprocessing for face detection, feature extraction, identity verification, image classification, offline local autonomous operation and intelligent sensor interfaces. Since edge and end-side devices are often powered by power supplies, it is also necessary for the MCU to optimize the energy efficiency ratio, ensuring the battery life of the devices while completing the AI load. In addition, when the MCU+AI operates functions such as face detection and voice interaction, it involves users' personal data, and the security standards of the MCU need to be upgraded from "functional safety" to "AI trusted computing".
Embrace RISC-V actively
    To better meet the customized demands of downstream applications, MCUS are actively embracing the RISC-V open-source architecture and making breakthroughs in multiple fields such as automobiles, smart terminals, and various industries. In the automotive field, automotive-grade MCUS designed based on the RISC-V architecture will accelerate their application in vehicles in the next two years. In April this year, the R&D Institute of Dongfeng Motor Corporation announced that its fully domestically produced high-performance automotive-grade MCU chip DF30 has completed its first tape-out verification and is scheduled to go into mass production and be launched on the market next year. It is reported that the DF30 chip is based on the RISC-V multi-core architecture and adopts the domestic 40nm automotive-grade process, with a functional safety level reaching ASIL-D. Recently, Nanjing Zijing Semiconductor announced that the mass production version of its high-performance automotive-grade MCU M100 has been successfully taped back. It is expected to enter the mass production stage in the third quarter of 2025 and will be first applied to models such as Great Wall Motor's Blue Mountain, High Mountain, Tank 300, Tank 400, and Tank 500. It is reported that the Bauhinia M100 is built on the open-source RISC-V core, featuring a modular design and reconfigurable cores. The 4-stage pipeline enables it to have a faster processing speed and less time consumption, meeting the ASIL-B grade requirements.
    In the terminal field, hisilicon launched the Hi3066M in March this year to meet the intelligent demands of white goods. This MCU uses hisilicon's own RISC-V core and is equipped with an eAI engine, supporting a main frequency of 200MHz. It can be applied to innovative application scenarios such as edge AI in air conditioners, refrigerators, and washing machines, for instance, it supports AI energy saving for air conditioners, AI weighing and eccentricity detection for washing machines, as well as AI noise reduction and energy saving for refrigerators. In addition, Shanghai hisilicon is promoting the in-depth adaptation of RISC-V with OpenHarmony and has contributed millions of lines of code to the open-source community in total. Zhang Tao, the chief expert of RISC-V at hisilicon     Technology Co., LTD., stated in his speech in May this year that RISC-V builds the core foundation to support technological innovation, while OpenHarmony provides a unified foundation to support the industrial ecosystem. The deep integration of the two provides developers with a unified, rich and efficient software platform, which can effectively reduce development costs and improve development efficiency, thereby promoting the prosperous development of the industrial ecosystem.
    Focus on providing personalized solutions for specific applications
    Nowadays, the intelligent transformation in various fields such as consumption, automobiles, and industry is constantly expanding the application boundaries of MCUS. The application requirements in different fields vary greatly, which prompts MCU manufacturers to deeply understand the needs of specific application scenarios and provide tailor-made product solutions. In the automotive field, the "three electric" systems (battery, motor, and electronic control) of new energy vehicles impose strict requirements on the computing power, reliability, and safety of the MCU. Take the battery management system (BMS) as an example. It requires an MCU to precisely monitor parameters such as battery voltage, current, and temperature, and to precisely control the charging and discharging process of the battery to ensure its performance, lifespan, and safety. Infineon's AURIX TC4x series MCUS are specifically designed for automotive powertrains, chassis and safety applications. They support functional safety ASIL D systems and can meet    the strict standards of reliability and safety in the automotive industry.
    In the industrial field, application scenarios such as factory automation, process control, and robotics require MCUS to have high real-time performance, high reliability, and rich communication interfaces. For instance, in industrial automated production lines, MCUS need to collect sensor data in real time to control the operation of motors and achieve precise motion control. Renesas Electronics' RZ/T2M microcontroller is specifically designed for industrial Ethernet applications, integrating a dual-core Arm Cortex-R5F core. It features outstanding real-time performance and communication capabilities, meeting the demands of industrial automation for high-speed and reliable data transmission. In the field of consumer electronics, with the popularization of smart homes and wearable devices, MCUS need to perform well in terms of low power consumption, miniaturization and cost control. For instance, in a smartwatch, the MCU not only needs to drive the display screen and process sensor data, but also achieve functions such as Bluetooth communication, while ensuring the device's battery life. NXP's i.MX RT1010 crossover MCU features a low-power design and integrates a rich array of peripherals, capable of meeting the high performance and low power consumption requirements of wearable devices.
    Overall, the MCU is standing at a crossroads of transformation. With the advancement of technology and the diversification of market demands, these five major trends will continue to reshape the industrial landscape of MCUS. From breakthroughs in manufacturing processes to the expansion into emerging application fields, from functional upgrades brought about by AI integration to innovations in the RISC-V architecture, as well as in-depth customization for specific applications, MCU manufacturers can only gain an edge in the fierce market competition and promote the key role of MCUS in more fields by closely following the trends and constantly innovating. Facilitate the intelligent upgrade of industries.