Hand selecting AI on a digital circuit board, representing artificial intelligence and edge computing innovations

AI was the Magic Word in Embedded World North America This Year

Posted on November 18, 2026

AI has been dominating the tech landscape for a while now, so it comes as no surprise to see it center stage at the recently concluded Embedded World North America. But how exactly is the technology being integrated into embedded computing, and how does it seek to change things? Is it just hype chasing, or are there actual shifts in processes? What even can AI do for embedded applications? Here is what we saw. 

It’s All About Physical AI

When you usually hear about AI models and their successes, it is usually about cloud AI. Models like ChatGPT or Gemini run on third-party servers and are offered as online services. But when it comes to embedded computing, the focus is on physical AI, that is, AI applications that can be run on-site, directly processing the data collected and acting on it. 

This is basically Edge computing with a twist. At Embedded World North America this year, NVIDIA and Qualcomm talked about the emerging field of physical AI, and their partnerships with manufacturers to produce smaller, more integrated sensor suites to make such systems practical. 

Robotics, of course, is the most talked-about use case for physical AI, but there are more relevant applications too. Integration with cameras for smarter AI security systems, or AI-powered automation in factories, where robotic arms can accurately perform tasks they normally struggle with. And things like warehouse management, where AI vision can enable better object detection and pathing, letting machines move stock and keep track of inventory without constant human intervention. 

FPGAs Are on the Rise

We talked about how essential sensor suites are to the deployment of AI at the Edge, but that’s not the only piece of the puzzle. The variety of sensor data collected cannot directly be piped into a regular embedded system, or be interfaced by it. For this, a more flexible form of low-level computation is required. And FPGAs, or Field Programmable Gate Arrays, fit the bill perfectly. 

Since FPGAs are programmable on a hardware level, they are the best way to integrate multiple channels of real-world data measurements and process them in parallel. Since FPGAs also feature extremely low latency, their introduction into the computing pipeline is seamless, bringing the processed data to the actual AI computer without the delay associated with cloud AI approaches. 

As a result, we are seeing manufacturers across the board roll out FPGA products. Some are integrated directly with GPUs as hybrid AI boards, while others simply slot into the ecosystem as highly efficient signal processing solutions, essential for applications like AI vision or robotics. 

RF Integration Is Becoming Key

While wireless networks are essential for home computers, they are often skipped in embedded systems, relying instead on wired connectivity standards. But with the Internet of Things (IoT), embedded systems started relying on wireless communication a lot more, and with the push toward AI-powered edge systems, this need is becoming even more pronounced. 

AI applications require extensive data, and the best way to obtain that data is through a network of devices and simple computers that are equipped with their own sensors and can transmit the information collected to the main Edge system for actual processing. And for communicating with such a high number of devices, wired standards like Ethernet prove too expensive and impractical. 

RF, simply short for radio frequency, is an umbrella term for communication using wireless wavebands, which includes Bluetooth and WiFi as well, but also other bands for a more diverse communication network. 

Zephyr Adoption Is Becoming Universal

Small-scale embedded systems working directly with field interfacing hardware generally run on Real-Time Operating Systems (RTOS). There are many variants of these, all usually built on the Linux kernel, and for many years there hasn’t been a clear winner. But the open-source project Zephyr RTOS is quickly emerging as the default choice, with more and more products shifting to the Zephyr ecosystem.

We saw many teams demonstrating projects using Zephyr in Embedded World North America, apart from well-attended technical sessions by Zephyr itself, going into topics like AI-assisted Board Support Package (BSP) development. This lets teams quickly put together a Zephyr BSP for any hardware configuration, and is already being used for fast prototyping of new-age embedded systems by engineers around the world.

The leading reason for the industry to gravitate towards Zephyr, however, is not the fancy AI development feature, but the more robust ecosystem. It is striking vendor partnerships and providing a consistent software platform to access the variety of hardware that can be leveraged by embedded computers, making it much easier to take a project from start to finish without reinventing the wheel. 

Embedded World North America: The Upshot 

Embedded World is always an opportunity to showcase the emerging trends of the industry, including new technologies and products being made to leverage them. For the past few years, AI and Edge computing have been buzzwords for pretty much all such events, but as of this year, we are seeing the conversation move to actual practical use cases. 

Physical AI is now a reality, not a hypothetical concept, and thanks to better sensors, efficient FPGAs, and RF integration, it can actually be rolled out in real-world applications to start adding value. Vendors and OEMs are making it even easier by creating hybrid products that integrate the whole foundation of an AI embedded PC on one board, including signal processing and a GPU as well. 

In other words, Edge computing is here, and it’s going to stay. Embedded systems are investing in AI capabilities and adding intelligent features that were not possible before. If you have not considered it yet, it is time to do so now, before you are left behind the curve of innovation.

 

Contact Our Experts

 

Share: