embedded world 2026: The Evolution of Edge AI, Wireless Connectivity, and Low Power Innovation

The convergence of edge Artificial Intelligence (AI) and wireless connectivity solutions was the key theme of embedded world 2026, most notably demonstrated by the growing integration of Neural Processing Units (NPUs) alongside various connectivity Systems-on-Chip (SoCs) and Microcontroller Units (MCUs). Meanwhile, various new product announcements, demonstrations, and partnerships highlighted continued innovation in the broader connectivity ecosystem, ranging from low-power, low-latency Ultra-Wideband (UWB) connectivity, new IEEE 802.15.4ab secure access solutions, Bluetooth® Channel Sounding, Wi-Fi 7, energy harvesting, and multi-protocol solutions.   

In terms of headline announcements, there were many.

On the UWB front, STMicroelectronics introduced its long-awaited ST64UWB family of IEEE 802.15.4ab solutions, consisting of three SoCs that take advantage of Narrowband Assist (NBA) to significantly improve range, reliability, and overall user experience across multiple UWB use cases. The transition to IEEE 802.15.4ab will be pivotal in enabling longer-range, lower-power, and more robust deployments across consumer, enterprise, and industrial markets. The ST64UWB-A100 focuses on digital key and vehicle localization; the ST64UWB-A500 integrates AI acceleration for edge AI radar applications such as gesture sensing, child presence detection, and other externally facing vehicle use cases, while the ST64UWB-C100 targets commercial and consumer applications with support for the Aliro standard. For more information on next-generation UWB, please see ABI Research’s IEEE 802.15.4ab: Unlocking UWB’s true potential whitepaper in partnership with STMicroelectronics.

CEVA announced its next-generation Ceva-Waves™ UWB Internet Protocol (IP), the industry’s first IEEE 802.15.4ab-compliant UWB IP, capable of supporting up to 30X range and 4X faster data rates. Meanwhile, Infineon was also demonstrating its upcoming UWB solutions at the show. Meanwhile, CEVA’s NeuPro-Nano NPU IP also won the Artificial Intelligence category of the 2026 embedded award, and its combination of edge AI and wireless connectivity IP will be a fundamental enabler of wider adoption of these combined solutions from vendors in the years to come.

While UWB has gained the most attention within a secure range of applications to date, SPARK Microsystems continued to demonstrate its Low Energy UWB (LE-UWB) technology at the show alongside its growing ecosystem of devices and partners. These solutions are capable of providing ultra-low latency, ultra-low power consumption, and high data throughput, while maintaining robust and reliable connectivity. Key demonstrations included lossless audio with sub-3 ms latency, 8K Human Interface Device (HID) peripherals, low-power proximity and presence detection solutions capable of supporting a battery life of over 5 years. ABI Research believes this technology has the potential to find a sweet spot between Bluetooth® and Wi-Fi and enable high-throughput for Extended Reality (XR), audio, personal AI devices, wearables, and gaming applications, essentially acting as a short-range cable replacement technology.  This year, we expect to see a significant increase in real-world products across consumer and Internet of Things (IoT_ applications, highlighting UWB technology’s unique role.

Synaptics introduced its SYN765x, an innovative AI-native wireless solution that combines the latest Wi-Fi 7 multi-protocol connectivity, optimized for IoT applications, with advanced edge AI to enable unique inference capabilities on a diverse range of IoT devices. Alongside Wi-Fi, the SYN765x incorporates concurrent support for additional technologies central to the IoT market, including Bluetooth® Low Energy (LE) (with support for Bluetooth® Channel Sounding) and 802.15.4 (with support for Zigbee and Thread). Channel Sounding was also widespread at embedded world 2026, with the technology capable of providing a low-cost, low-complexity alternative to the integration of an additional UWB radio.

NXP unveiled its i.MX 93W application processor, combining an AI NPU capable of up to 1.8 Equivalent Trillions of Operations Per Second (eTOPS) alongside an integrated IW610 tri-radio with support for Wi-Fi 6, Bluetooth® LE, and 802.15.4. The solution is a highly integrated SoC that is capable of significantly reducing the size, cost, and complexity of edge AI rollouts. 

Texas Instruments (TI) expanded its microcontroller portfolio with its MSPM0G5187 and AM13Ex MCUs. These integrate TI's TinyEngine NPU to optimize deep learning inference operations to reduce latency and improve energy efficiency. This announcement followed the news in early February that TI will be acquiring Silicon Labs, a market leader in low-power wireless connectivity, and the combined company will be well placed to take advantage of the scalable rollout of edge AI wireless technologies in the coming years.

The IoT market continues to expand at a rapid pace, enabling the creation of valuable services and digital transformation across consumer, enterprise, and industrial settings alike. The connected IoT installed base is projected to surpass 50 billion devices by the end of this year. Much of this growth and maturity has been enabled by the ongoing innovation in the wireless connectivity market, including growing demand for wireless solutions, the continued transition to new wireless standards, the availability of new spectrum, and the proliferation of multi-protocol solutions.

This wireless connectivity innovation comes alongside the enormously transformative shift in compute capabilities, driven by the rapid expansion of Artificial Intelligence (AI) and Machine Learning (ML) algorithms and models, whether in the cloud, at the edge, or within hybrid AI architectures. Most importantly, AI is increasingly migrating away from data centers and the cloud and shifting toward pre-trained models that operate on the device itself. These models continue to shrink in size and increase in efficiency, enabling complex tasks to run on a whole host of IoT devices without the need for cloud computing. This shift toward AI inference at the edge is an enormous area of disruption and is a key driving force of future IoT connectivity innovation. 

As the IoT is a hugely diverse market, consisting of an ever-growing number of device types, often with wildly divergent priorities and needs. If edge AI is to reach its true potential, it needs to be effectively combined with high-performance wireless connectivity designed from the ground up with these varied applications and requirements in mind, offering a flexible platform from which developers can fine-tune to specific applications. These solutions will be critical in enabling evolving concepts such as the AI home that will rely on edge AI to deliver greater automation, personalization, and enhanced security and privacy.

However, these trends also bring new challenges and threats to the ecosystem.

First, the huge shift in focus to edge AI capabilities, in which connectivity will have a central role, will result in the need for new product portfolios, partnerships, and platforms that can help accelerate AI seamlessly across multiple consumer and IoT markets. Vendors are increasingly embedding NPU support into their AI-native solutions, while the shift toward connected/wireless MCUs is also accelerating to optimize power efficiency, reduce Bill of Materials (BOM) cost and size, and simplify development. These are also now beginning to embed NPU capabilities, scaling edge AI even further, while also supporting multi-protocol connectivity that combines Wi-Fi, Bluetooth®, and 802.15.4. However, vendors need to offer the right combination of edge AI capabilities and connectivity innovation, both in terms of their portfolio and go-to-market strategy. There also needs to be an effective balance of compute capabilities with memory, power, cost, robustness, and security requirements. Meanwhile, new entrants are emerging all the time, and there is huge innovation coming from different parts of the ecosystem. Therefore, a vendor that was once strong on connectivity cannot simply ignore the integration of compute capabilities. This requires a rethink in strategy and developing the right partnerships.

Vendors also need to enable compelling inference models and use cases that edge AI can solve for specific applications, rather than just giving a blank page for customers to build on. These solutions need to be able to solve specific problems and bring Return on Investment (ROI) to end users.  Finally, there is the issue of what is going to be done with all this data when it is generated, and how to standardize this across different applications to deliver real-world outcomes. The expansion of edge AI will bring intelligence to the edge, but it will remain ineffective if there is not a strong framework in place to utilize this effectively across different ecosystems.

Above all, the wider memory shortage impact has not been felt yet and looms over everything—this could cause some significant challenges to end-market opportunities in the next 2 to 3 years across multiple consumer, enterprise, and IoT verticals.