Embedded World 2026: A Field Report on the Latest in Embedded Electronics

Embedded World, one of the biggest events in the calendar for the embedded electronics industry, has come to a close for another year — having hosted a record 36,000 visitors to nearly 1,300 exhibitors showcasing everything from the latest in Bluetooth technologies to self-driving vehicles and robotics.

"This year once again impressively demonstrated the importance of Embedded World in the global embedded community," boasts Benedikt Weyerer, executive director of the annual show, of Embedded World 2026. "The energy in the halls — packed aisles, lots of good discussions and an impressive variety of ideas — really inspired us. This strong feedback motivates us enormously and encourages us to continue on the successful path of the event.

For companies in the embedded space, Embedded World is a chance to showcase everything they've been working on over the past year. For visitors, an opportunity to see not only what's new but what's around the corner — with exhibitors showing off technologies that are shipping now alongside those that are still in the pipeline, including fully-autonomous vehicles and humanoid robotics.

"An exceptionally broad spectrum of innovative technical presentations characterized this year's Embedded World conference," says event chair Axel Sikora, celebrating a 13 percent growth in visitors compared with last year. "The depth of the technical discussions and the quality of the scientific input show how actively and future-oriented our developer community is working. Rarely has it been so clear how intensively research and practice are working together on new solutions."

For those who couldn't be among the 36,000 lucky attendees, we've picked our highlights from the show floor below.

Nordic Semiconductor

While companies like AMD were showing off almost everything a vehicle needs bar wheels and an engine, Nordic Semiconductor's stands focused primarily on lighter-weight embedded applications, including in industrial and smart home spaces. The technology the company was keenest to show off? Artificial intelligence at the edge — not the controversial large language model (LLM) chatbot technology currently making headlines, but more traditional machine learning models running on low-power embedded devices for everything from voice control to gesture recognition.

"Here I have a little nRF54L15 board; this has no dedicated hardware for running neural networks, no neural processing unit or anything, but this runs [a] custom Newton model that does gesture recognition based on the data from the accelerometer on the board," Nordic's Robin Saltnes told us as he walked us through a trio of edge-AI demos at one of the company's two booths. "All of [the] gestures are recognized directly on this little device, and then only the result is sent over Bluetooth to the PC."

Despite being little larger than your thumbnail, the circular nRF54L15 board at the heart of the demo performs all recognition on-device — using only its microcontroller core. As you'd expect, though, Nordic also has something for those who need to run larger and more complex models: its energy-efficient Axon neural processing unit (NPU), as found in a pair of nRF54LM20B boards at the same demo station.

"We have one model for wake words that is recognizing one phrase, and this is tuned to reject false positives. So, if I try to say all the keywords before activating it with a wake word, it won't recognize them. But if I say 'okay, Nordic,' it's recognized," Saltnes explained.

"This keyword spotting algorithm [running] on the normal Arm Cortex CPU takes almost 74 milliseconds to complete on inference, and this translates to 187 micro-Coulombs of charge being used by one event; while on the Axon NPU it takes less than seven milliseconds, six and a half about, and less than 20 micro-Coulombs to complete one inference — so in this case it's 10 times faster and uses almost 10 times less energy."

Elsewhere on the stand Nordic was showing off its smart-home chops with a Matter-based smart lock system, built around the Connectivity Standards Alliance's new Aliro vendor-neutral specification, how the new Shorter Connection Intervals (SCI) feature in the latest Bluetooth standard can drop latency for sensors and input devices, its Internet of Things (IoT) connectivity options including cellular and non-terrestrial network (NTN) satellite options, and its nRF70-series low-power Wi-Fi communications controllers being used with both Linux-based operating systems and the Zephyr real-time operating system (RTOS) — about which more later.

More information on Nordic's line-up is available on the company's official website.

NXP

At the NXP booth, we were treated to a hands-on demonstration of another new feature in the latest Bluetooth specification: Bluetooth Channel Sounding, which delivers up-to-centimeter-level positioning accuracy between two compatible Bluetooth devices — without the need for any other hardware, nor reliance on non-Bluetooth technologies like Ultra-Wideband (UWB) transceivers.

"It's actually an improvement from the older technology [used] for locationing: often it was using RSSI [Received Signal Strength Indication] signal measurements," NXP's Kathleen Jachimiak explained as we took a look at the demo, one of a number to be found at NXP's FRDM Lab at Embedded World, "and those can be affected by people walking by, walls, reflections, all of that affects accuracy.

"What [this] is doing is having [one] board send tones to the receiving board, and the receiving board interprets those tones, takes information like time-of-flight and other measurements of that tone, and can accurately predict the measurement between the two locations."

Moving one board away from the other and seeing the estimated distance update on a compact display isn't a groundbreaking demo on the surface, until you think about how it's achieved — with no ultrasonics, no lasers, no magnets, no stereo cameras, none of the additional sensors you'd expect to find, just two unmodified NXP development boards.

"It's pure Bluetooth, taking advantage of Channel Sounding," Jachimiak explained. "We're seeing Channel Sounding in the real world in things like asset tracking, smart home automation, access control, and even retail and logistics."

Elsewhere at the FRDM Lab and NXP's second booth at the event the company was showcasing its S32N7 "Super-Integration Processor" for intelligence automotive applications, how embedded systems can — and should — migrate to post-quantum cryptography that can't be broken by as-yet theoretical but under-development quantum computing systems, and a variety of use-cases for NXP technologies in the smart home and industrial automation. The company's partner wall, meanwhile, showcased a range of real-world implementations from a variety of companies — under the watchful eyes of a pair of quadcopter drones designed for last-mile deliveries.

More information is available on the NXP website, along with access to the source code for the demos on display at the event.

Zephyr Project

Zephyr's presence at this year's Embedded World comes at a milestone for the organization: the 10-year anniversary of the free and open real-time operating system (RTOS) project, which is held under the auspice of the Linux Foundation.

"Back when Zephyr started, the joke in the industry was 'Security is the S in IoT,'" Zephyr Project director Kate Stewart told us in an interview at the organization's booth, in amongst demos from Zephyr users and blue kites being offered as giveaways in celebration of the project's first decade. "And so we got developers together, did a bit of a survey, and we did some focus groups to figure out what's really needed, and security was highlighted, and safety was highlighted. And so we found some people that were very interested in collaborating on these topics and improving the communication stacks and sharing effort. And that's what started Zephyr."

To say Zephyr has had a successful decade is no exaggeration: since its launch in 2016 the project has grown from an initial pool of around 100 contributors to more than 3,000 spread across the globe — while the RTOS now supports over 1,000 boards across multiple different architectures, delivering on the project's original promise of a vendor-neutral operating system that makes it easy to migrate projects across hardware platforms.

"We are at the point where the contributors, they come and go, but this is fine, because we have a critical mass of people who maybe contribute new functionality to Zephyr because they need it in their products," developer advocate Benjamin Cabé notes. "They ship the products. Maybe they sort of lose interest because, I mean, it's there, it's in their products, and someone else is going to step up to maintain the functionality, because someone else is going to need it and ship it in other products and make the feature compatible with the hardware. That's where and how we get to this ecosystem of 1,000s of boards."

Cabé also had advice for those interested in experimenting with the OS themselves. "What I usually tell people [is], you likely have dev kits hanging around on your desk, be it [Espressif] ESP32, Arduinos, [STMicroelectronics] STM32, whatever, really, thousands of boards. Pick one, pick any of our 100-plus code samples, and go and play with Zephyr," he told us. "Whether you're interested in building some kind of Bluetooth-enabled devices, anything sensor related, we have, likely, the drivers already for them. It's really, really easy to get started, and if people get stuck, just like [Hackster], we have those forums, Discord, people can [use]. We're pretty friendly, and we are usually pretty helpful when people get stuck. It's a really nice community."

Those looking to get stuck in, whether simply as a user or as a contributor to the project itself, can find out how on the Zephyr Project website, while anyone looking for more on the project's origins, growth, and plans for the future should turn to the report published by the Linux Foundation celebrating the project's first decade.

Honorable Mentions

With almost 1,300 exhibitors spread across seven halls, it's simply impossible to speak to everyone — even with three days to play with before the doors close on Embedded World for another year. Amidst the crowds of attendees thronging the halls and the blur of booths small, large, and absolutely gargantuan, we caught sight of everything from weightlifting robot arms and autonomous vehicles to connectors and cases aplenty — and our highlights of these follow.

Perhaps the biggest news for the hobbyist and education markets was Arduino's unveiling of the VENTUNO Q, its second Qualcomm-powered single-board computer and a bigger and more powerful stablemate for last year's Arduino UNO Q. Launched to celebrate its 21st anniversary, the Arduino VENTUNO Q features a Qualcomm Dragonwing IQ-8275 eight-core 64-bit application-class processor with a 40 tera-operations per second (TOPS) accelerator for on-device machine learning and artificial intelligence (ML and AI) workloads, plus an STMicroelectronics STM32H5F5 coprocessor for real-time workloads. What the company wasn't sharing, though: pricing and availability, with attendees able to whet their appetites but not place an order for a board of their own.

IoT connectivity specialist Blues, meanwhile, was showing off its latest Notecard communications module — built in partnership with satellite communications specialist Skylo to pack Wi-Fi, terrestrial cellular, and non-terrestrial network (NTN) satellite connectivity in a single module. Designed for use with the company's Notecarrier range, the Blues Notecard for Skylo is a drop-in replacement which includes 500MB of cellular data and 10kB of satellite data without the need to sign up for a subscription or pay recurring fees.

Aside from its role in powering the microcontroller side of the new Arduino VENTUNO Q, STMicroelectronics was also showcasing its own hardware — including the ST64UWB, described the company as "an industry-first system-on-chip supporting the latest ultra-wideband specification, IEEE 802.15.4ab." This, the company says, will target everything from vehicle keyfob systems to industrial applications — and will launch with support for the Aliro vendor-neutral smart lock specification.

Texas Instruments' booth leaned heavily, as did many others, on interest in the AI boom, with the company showcasing its TinyEngine neural processing unit (NPU) for ultra-low-power on-device machine learning at the edge. "TI invented the digital signal processor almost 50 years ago, laying the groundwork for today's edge AI processing," claimed TI's Amichai Ron of the launch. "Now TI is leading the next phase of innovation by integrating the TinyEngine NPU across our entire microcontroller portfolio, including general-purpose and high-performance, real-time MCUs."

Arm, likewise, was pushing its low-power compute for intelligence at the edge, showing off what it describes as a "unified foundation" built atop its Armv9 microprocessors, Ethos-U low-power neural coprocessors, and software stack. Live demos included an adaptive "swarm" of collaborative robots, smart cameras, and industrial controllers, an always-on wake-word-driven speech recognition system, and a smart camera system using transformer-based vision and language models running on-device to present a plausible description of items in its field of view.

These were far from the only companies showing off their creations at the event, and if we rushed by your booth on the way elsewhere we can only apologize — and hope to see your latest and greatest creations next year at Embedded World 2027.