We’re now reaching the point where the term edge AI is becoming ingrained in the industry. This was especially evident at the recent Embedded World 2025 event; edge AI is no longer a mythical creature or two separate words in a sentence, but a set of technologies and practices that is proven to deliver real value and is being adopted by largest players in all kinds of verticals. But how does this latest and greatest category advance IoT — a recent technology that has found mass adoption across all sectors?
The convergence of IoT, AI, and edge computing is now reshaping industries. It enables smarter, faster, and more efficient systems. In my recent talk, I highlighted how these technologies are evolving and why their integration is critical for the future of innovation.
Defining IoT and the Shift to Intelligence
IoT can be viewed as the combination of sensing, connectivity, and intelligence.
We have been very well covered on the sensing front for many years with sensors available virtually for every possible signal and application.
Connectivity has effectively commoditised in the last years, with technologies like LoRaWAN and LTE-M fully maturing. Now the market is comfortable taking those off the shelf to enable connected devices.
But what about intelligence?
This aspect has been playing catch up with the ones above, with many devices having to rely on the rule that "low power = low performance." The last two years alone have proven that this is not true anymore. Edge AI is now enabled by hardware, more than ever.
The development of energy-efficient GPUs and dedicated neural processing units (NPUs) has significantly enhanced the capabilities of end nodes, or edge devices. These advancements enable complex AI tasks to be performed locally, further reducing reliance on the cloud.
With hardware vendors like Alif Semiconductor, NXP Semiconductors and Himax Technologies, Inc. adapting Ethos NPU design from Arm, and vendors like STMicroelectronics and Renesas Electronics coming up with their bespoke ASICs, the intelligence budget of an IoT device, including battery-powered, has increased exponentially over what was available even just a few years ago.
Qualcomm’s recent introduction of Dragonwing, its leading edge AI, high-performance, low-power line of hardware and solutions, shows that the biggest leaders in cross-industrial tech are recognising this shift as well, and are starting to build full end-to-end solutions with edge AI at the core.
The shift is crucial as it enables devices to not only collect data but also analyze and act on it in real time.
The Gartner Hype Cycle for AI reveals that while computer vision is a mature technology, edge AI is rapidly evolving alongside it. Moreover, computer vision itself is actively moving to the edge, being accessible on more and more restricted hardware. Edge AI is expected to reach the plateau of productivity in less than two years, making it a game-changer for industries requiring low-latency, real-time decision-making.
Benefits of Edge AI
Running machine learning models on edge devices offers significant advantages, including reduced energy costs and bandwidth usage compared to cloud-based processing. This is particularly beneficial for applications like smart manufacturing, where real-time data analysis can prevent costly downtime.
Model Cascading
The integration of several machine learning models, often called cascading, offers different ways to make the application more intelligent and move beyond constraints of one model or one device, and offer edge-to-cloud and edge-to-edge model combinations, some examples of which are discussed in the video.
The most novel and powerful edge chips, like Qualcomm’s IQ9 series, are powerful enough to run a full VLM (Visual Language Model) locally that does not require any training. Edge Impulse has shown this in action at the Embedded World 2025 in a joint demo with Qualcomm.
Edge Impulse and IQ9 bring object detection directly to hardware, enabling real-time #AI deployment on edge devices. Watch: pic.twitter.com/9Rn6KA4OYa
— Qualcomm (@Qualcomm) March 13, 2025
The Importance of Accessible Tools and Developer Community
Platforms like Edge Impulse are democratizing edge AI by providing tools for building and deploying machine learning models on edge devices. This allows developers to quickly iterate and deploy solutions across various hardware, making edge AI more accessible and scalable. After all, developers are the most important aspect of any emerging and existing technology — without developers nothing is going to be built!
There is a clear need for free and accessible tools to empower developers to adopt these emerging technologies, and that is the role Edge Impulse has created and taken. Join our community of 170,000+ developers for free today to harness the full potential of IoT, AI, and edge computing.