Recently, Embedded World 2026 was held in Nuremberg, Germany. As the annual flagship event for the embedded systems industry, each edition of the show brings together the most cutting-edge technologies and products. As everyone knows, AI is currently the hottest field, and both physical AI and OpenClaw have further fueled its popularity. Today, EEWorld will highlight some of the new offerings from select vendors, helping engineers stay on top of the latest trends and developments in the embedded space.
Infineon has once again topped the global MCU market this year. According to Omdia’s latest research, the company’s total MCU market share is projected to rise to 23.2% in 2025 (compared with 21.4% in 2024), an increase of 1.8 percentage points year on year—the largest gain among its competitors. Notably, this market-share expansion comes despite a modest overall decline of 0.3% in the MCU market.
Infineon’s key exhibits this year include robotic solutions and prototypes, such as a humanoid robot head equipped with XENSIV sensors, PSoC MCUs, EZ-USB FX10, and NVIDIA Jetson Thor, showcasing edge AI capabilities. In the AI domain, Infineon’s Live Lab offers remote, real-time semiconductor solution testing with no lab-access restrictions, reimagining hardware and software development workflows. In the automotive sector, the TRAVEO SDV regional demonstration is on display; in IoT, a combined Wi-Fi 7, Bluetooth, and IEEE 802.15.4 solution is showcased; and in quantum computing, Infineon highlights TPMs that support post-quantum cryptography, secure MCUs, and OPTIGA security devices.
Notably, Infineon also unveiled three DRIVECORE software suites at the exhibition. At the heart of this product expansion is the new DRIVECORE suite tailored for Infineon’s RISC-V virtual prototyping platform, representing a critical step in building an automotive ecosystem around Infineon’s upcoming AURIX series of RISC-V-based microcontrollers (MCUs).

During the exhibition, TI unveiled two new MCU families—the MSPM0G5187 and AM13Ex—both of which integrate TI’s TinyEngine neural processing unit (NPU). TI believes that while much of the world is focused on AI acceleration and NPUs in large-scale SoCs, it has been shown that some of the most interesting and far-reaching AI applications can actually be realized on small chips like MCUs.
The MSPM0G5187 features an 80 MHz Cortex-M0+ core, 128 kB flash memory, 32 kB SRAM, USB 2.0 FS, I2S, an ADC, and a TinyEngine NPU with 2.56 GOPS; the AM13E23019 is equipped with a 200 MHz Cortex-M33 core, making it a real-time motor-control MCU with edge AI capabilities, a TMU, 512 kB flash memory, and enhanced security.

NXP has launched the i.MX 93W integrated wireless application processor for physical AI, combining computing and connectivity through the integration of the i.MX 93 application processor, the IW610G tri-radio module, and a complete IW610G radio bill of materials (BOM) that includes passive components, crystals, and the RF front-end. The device is equipped with Cortex-A55 cores running at 1.7 GHz and Cortex-M33 cores at 250 MHz, along with EdgeLock secure enclave, a dedicated Neural Processing Unit (NPU) called Ethos U-65, dual-band Wi-Fi 6, BLE, and 802.15.4 support. By replacing up to 60 discrete components with a single pre-validated solution, it reduces system scale and supply-chain complexity. The i.MX 93W processor is scheduled for release in early 2027.

During the exhibition, ST unveiled its ST64UWB series, which supports the latest UWB standard IEEE 802.15.4ab and multi-millisecond ranging (MMS), including narrowband auxiliary radio (NBA).
The ST64UWB-A100 is a CCC-certified digital key for automotive applications, supporting UWB+NB distance ranging. It is also designed for automotive use, offering UWB+NB ranging and radar capabilities, along with in-cabin protection, radar sensing, and access control. The ST64UWB-C100, on the other hand, targets consumer and industrial applications, providing UWB+NB distance ranging that remains accurate even under non-line-of-sight (NLOS) conditions, while optimizing access control through Aliro.

As in previous years, Microchip has a robust exhibit lineup. This year, we’ve highlighted the following key offerings: First, the PolarFire Ethernet sensor bridge to NVIDIA Holoscan, designed for low-latency sensor data acquisition and pre-processing; second, the 10BASE-T1S endpoint solution, which integrates control, lighting, and audio functions—including a 60x60 RGB lighting system, multiple sensor inputs (proximity, light, temperature, humidity, motor, and joystick), quad-microphone audio capture, and Power over Data Line capability; third, an agile AIoT system adoption platform that showcases the appeal of technologies such as 10BASE-T1S, Time-Sensitive Networking (TSN), Zephyr, Linux, secure MCUs/MPUs, advanced graphics, and touch-screen and camera systems; and fourth, automotive SerDes solutions for the Qualcomm Ride platform, demonstrating four ADAS cameras connected via ASA-ML to the Qualcomm Snapdragon Ride platform, which leverages the VS775S serializer and VS776Q deserializer devices.

Renesas has announced the official global launch of “Renesas 365,” an intelligent modeling platform powered by technical support from Altium. Leveraging modeling-based evaluation and optimization technologies, engineers can now use Renesas 365 as an intelligent design environment that proactively helps select the most suitable MCU based on comprehensive system requirements. Rather than manually sifting through datasheets one by one, engineers can now receive guided recommendations based on pin usage, peripheral configuration, timing constraints, power consumption, and the degree of compatibility between components and system building blocks.

During the trade show, Silicon Labs announced a groundbreaking collaboration with BANF, a South Korean provider of smart tire systems, in the field of tire-monitoring technology. By integrating Silicon Labs’ ultra-low-power BG22 Bluetooth system-on-chip (SoC) into its in-tire sensor platform, BANF has successfully developed a real-time, high-resolution tire-data processing system specifically designed for autonomous vehicles and connected-fleet applications.
For decades, tires have been the industry’s “black box.” Conventional tire-pressure monitoring systems (TPMS) only trigger an alert when tire pressure drops significantly, which limits their ability to fundamentally prevent fuel inefficiency or mitigate safety risks. BANF’s solution transforms tires into connected, intelligent nodes that deliver actionable data in real time.

Installation Diagram for the iSensor Sensor from BANF
Nordic Semiconductor officially unveiled Nordic Fuel Gauge v2.0 at the trade show. This major upgrade represents a significant enhancement to its high-precision software-based fuel-gauge solution, designed specifically for the award-winning nPM1300 and nPM1304 power-management ICs. The new version introduces advanced features such as precise state-of-health (SoH) estimation, adaptive battery modeling, and long-term fleet analytics, extending cutting-edge battery-management technology to a wide range of low-power IoT products. In addition, the updated solution integrates seamlessly with nRF Cloud—Nordic’s cloud lifecycle service powered by Memfault—enabling efficient condition monitoring without the need to build custom cloud infrastructure.
The new version will help manufacturers develop products that are more reliable, more sustainable, and have a longer service life, while also complying with emerging battery-swapping regulations. These include the EU’s Battery Regulation (EU) 2023/1542, which stipulates that portable batteries must be “easily removable and replaceable by end users at any point throughout the product’s entire life cycle.”

At this year’s exhibition, GigaDevice announced the launch of the GD32M531 series of 32-bit MCUs, specifically tailored for motor-control applications. Powered by the Cortex-M33 core and featuring a dedicated hardware accelerator for motor control, the family also boasts highly integrated peripheral resources. With outstanding computational performance, precise control capabilities, and industrial-grade reliability, the GD32M531 series enables accurate dual-motor plus PFC control, delivering efficient and cost-effective solutions for a wide range of home-appliance applications, including outdoor air-conditioning units, air-source heat pumps, washing machines/dryers, dishwashers, and multi-burner induction cooktops. Samples and development boards for the GD32M531 series are now available, with mass production and full-scale supply scheduled to commence in April.
The GD32M531 series is built on a hardware architecture specifically optimized for motor control, delivering significant improvements in computational performance, control accuracy, and protection mechanisms: it features a built-in hardware accelerator for trigonometric functions and space-vector pulse-width modulation (SVPWM), tailored for FOC algorithms, which reduces CPU workload, enhances the efficiency of sensorless FOC algorithms, improves the precision of motor speed control, lowers operating noise, and boosts energy efficiency; it is equipped with two channels of enhanced AD-Timers that support independent operation of two sets of FOC. Hardware-based phase-shift ADC triggering enables synchronous and precise current and voltage sampling, eliminating the latency common in traditional software-triggered systems; the integrated POC>OC (Port Output Controller combined with General Timer Output Controller) provides filtered overcurrent protection without CPU intervention, achieving microsecond-level response times.

At the edge AI exhibition area, Rockchip highlighted the technical capabilities of its flagship SoC, the RK3588, and its next-generation 3D-architecture edge inference coprocessor, the RK1828. A dynamic demonstration showcased a video analytics solution powered by the Qwen3-VL-2B model, which delivers real-time multi-channel video summarization with response times as low as 0.5 seconds. By leveraging the RK3588 to invoke and deploy an offline large model on the RK1828, the system can concurrently analyze up to four video streams, automatically compress and condense video content for specified time periods, and precisely describe key events. In addition, the GUI Agent solution, built on the open-source Step-GUI-Edge framework from Jieyue Xingchen, demonstrated fully edge-side GUI automation running natively on Android—without any reliance on cloud-based computing power. The exhibit also featured an AI digital human integrated with an edge-side large model, supporting deployment of either the Qwen3-VL-4B or Qwen2.5-7B large language model to enable completely offline large-model inference.
In addition to edge-side AI, Rockchip showcased a range of industrial applications at the exhibition, including the lightweight RK3506J HMI solution, the RK3588J web-based human-machine interface, an RK3506-based PLC, an RK3568J motion-control card, the RK3572 Linux real-time system solution, and an industrial defect-inspection system built on the RK3588 + RK1828 platform. In the robotics zone, the RK3588 was on display, while the automotive zone featured the latest AI BOX solution based on the RK3588 + RK1828 architecture.

Under the theme “Intelligent Integration, Robotic Rebirth,” the Sweet Potato Robot showcases its latest innovations, with key demonstrations including the deep integration of RDK S100 with the Hugging Face LeRobot ecosystem, as well as a visual computing demo featuring RDK X5 paired with the newly released Ultralytics YOLO 26.
At this exhibition, Diguo Robotics officially launched the RDK S100—the nation’s first single-SoC integrated computing-and-control robot development kit—into the European market. Equipped with over 80 TOPS of high-performance computing power and featuring a human-like cerebellum–cerebrum architectural design, the kit enables efficient collaboration between large and small models for embodied intelligence, seamlessly closing the end-to-end “perception–decision–control” loop. This drives the co-evolution of the embodied robot’s cerebellum and cerebrum, providing global robotics developers with robust, stable, and highly efficient core computing support.
A major highlight of this exhibition was Digu Robot’s focused showcase of a comprehensive portfolio of innovative products developed by the members of its “Gravity Project (DGP)” initiative, including the Vbot Super-Powered Robotic Dog, the Aura Family Companion Robot, the BirdieSense Golf Embodied Intelligence Coaching System, and the Aurora Arc Portable AI Vision Assistant—demonstrating the program’s robust incubation outcomes and dynamic spirit of innovation. According to reports, Digu Robot’s Gravity Project (DGP) is a one-stop empowerment platform designed specifically for robotics innovators and teams, offering end-to-end services such as technical support, resource matchmaking, and real-world application deployment to accelerate the transition of innovative projects from concept to market. Looking ahead, Digu Robot will continue to intensify its efforts to connect robotics entrepreneurs and innovation teams worldwide, building an international innovation network that injects new momentum into the high-quality development of the global robotics industry.

Espressif Systems unveiled three new RISC-V products slated for release in 2026: the tri-band Wi-Fi 6E high-performance coprocessor ESP32-E22, the next-generation low-power Bluetooth SoC ESP32-H4, and the ultra-low-power Bluetooth SoC ESP32-H21. In addition, the company showcased its AI solutions, including the LLM solution, the ESP Private Agents Platform, ESP-WHO, ESP-SR, and ESP-DL.

At Embedded World, MediaTek unveiled a new family of Genio platforms designed to power a wide range of IoT products and applications: the MediaTek Genio Pro, Genio 420, and Genio 360. The Genio Pro series is MediaTek’s high-end offering for performance-intensive IoT and embedded solutions, while the Genio 420 and Genio 360 deliver efficient, system-level edge AI capabilities for smart home, retail, industrial, and commercial IoT devices.
The Genio Pro is built on TSMC’s 3nm process and features a full big-core Arm v9.2 configuration, including one Arm Cortex-X925 core, three Cortex-X4 cores, and four Cortex-A720 cores. This CPU architecture delivers advanced application computing performance of up to 260K DMIPS. The Arm Immortalis-G925 GPU provides 3.1 TFLOPS of compute and graphics performance. Equipped with MediaTek’s eighth-generation NPU, the platform supports system-level generative AI acceleration exceeding 50 TOPS, while the NeuroPilot framework enables support for PyTorch, ONNX Runtime, and TFLite (LiteRT), delivering unparalleled acceleration for traditional AI workloads such as machine vision and object classification. For generative AI applications, the platform supports a generation rate of up to 23 tokens per second, making it suitable for large language models (LLMs) with up to 7 billion parameters.

Overall, this year’s Embedded World in Nuremberg features a high degree of AI integration, with both edge AI and physical AI serving as key buzzwords for exhibitors. In addition, enhanced sensing and connectivity are also top priorities, as the biggest challenge for edge AI over the next three to five years will be seamlessly integrating display, perception, computing, and other functions into a cohesive system.
Wei-Min Dai, Chairman and President of Verisilicon, has stated: “The world does not need that many large models—such models would consume far too much power, a burden the global energy infrastructure simply cannot sustain. Under these circumstances, deploying domain-specific large models, as well as medium- and small-sized models, at the edge becomes all the more critical.”
Currently, AI is accelerating its migration from the cloud to the edge, and it is foreseeable that agentic AI and physical AI will be the primary focus of future embedded-system development.
As AI increasingly moves to the edge, it is bound to disrupt the established paradigms of the embedded systems industry. From an engineering perspective, the rise of edge AI poses significant challenges for embedded development. However, as evidenced by this year’s exhibition, vendors are stepping up their investments in software tools and algorithms.

Article source: EEWORLD Electronic Engineering World. Thank you for learning!