If there was one theme impossible to ignore at embedded world 2026, it was artificial intelligence. Walking the exhibition halls in Nuremberg, it quickly became clear that AI is no longer a future ambition for embedded systems – it is becoming a core capability across processors, microcontrollers, FPGAs, and development platforms.
From demonstrations of Edge vision systems to discussions around “physical AI” and AI-enabled IoT devices, vendors consistently highlighted how intelligence is moving closer to the device itself. Rather than relying solely on Cloud computing, companies across the embedded ecosystem are working to bring machine learning and AI inference directly onto low-power hardware.
Conversations with exhibitors including Lattice Semiconductor, Texas Instruments, MediaTek, Arduino, and Tria Technologies revealed a common focus: enabling developers to deploy AI at the Edge more efficiently, more affordably, and with fewer barriers.
Enabling physical AI
Lattice Semiconductor discussed how its low-power FPGA platforms are helping to enable what it described as “physical AI”.
Unlike traditional AI applications that analyse data in the Cloud, physical AI refers to systems that can interact directly with the real world through sensors and actuators. Robotics, machine vision, and industrial automation are prime examples.
Lattice’s FPGA technology allows developers to implement AI processing in compact, energy-efficient devices, making it possible to run inference at the Edge without requiring large processors or high power budgets. By placing AI closer to where data is generated, these systems can respond in real time – a critical requirement for applications such as autonomous machines and smart factories.
The company’s message reflected a broader industry shift: embedded hardware is evolving to support intelligent decision-making directly within devices.
Expanding AI across microcontrollers
Microcontrollers remain the backbone of embedded systems, and AI capabilities are increasingly being integrated into this space.
Texas Instruments highlighted its strategy of expanding Edge AI capabilities across its microcontroller portfolio, particularly within its MSP family. Rather than restricting machine learning functionality to specialised processors, the company is working to make AI features accessible on widely used embedded hardware.
This approach allows developers to integrate machine learning models into systems that already rely on microcontrollers for sensing, control, and connectivity. For many IoT and industrial applications, the ability to run lightweight AI models directly on an MCU could unlock new functionality without requiring major architectural changes.
The move also reflects growing demand from developers who want AI capabilities without significantly increasing system cost or power consumption.
Unlocking AI for IoT devices
Connectivity specialist MediaTek also emphasised the growing role of AI in IoT deployments. The company highlighted how its platforms combine connectivity, compute, and AI capabilities to enable intelligent Edge devices.
In many IoT environments, transmitting large volumes of sensor data to the Cloud is inefficient or impractical. Edge AI allows devices to analyse data locally, sending only relevant insights or events to Cloud systems.
This not only reduces bandwidth usage but also enables faster responses in time-critical applications. As a result, AI-enabled IoT devices are becoming increasingly viable across industries ranging from smart homes to industrial automation.
Making AI accessible to developers
While advanced processors and accelerators are important, another major focus at embedded world was lowering the barrier to AI development.
Arduino, long known for its role in prototyping and maker communities, discussed how AI capabilities are becoming integrated into more accessible development platforms. Arduino had unveiled its latest platform, the VENTUNO Q – a powerful, AI-capable development board built on a Qualcomm processor following Arduino’s recent acquisition by Qualcomm.
VENTUNO Q is built on Qualcomm’s IQ8-8275 processor — an eight-core CPU combined with a GPU for advanced computer vision and a Neural Processing Unit (NPU) capable of up to 40 TOPS. This means the board can run large language models (LLMs), visual language models (VLMs), and YOLO-style object detection entirely on the Edge, with no Cloud connectivity required.
Similarly, Tria Technologies highlighted how demand for AI-enabled embedded computing platforms is shaping its hardware roadmap. Customers across multiple industries are seeking systems capable of handling machine learning workloads at the Edge, particularly as AI-driven features become expected in modern devices.
A defining theme for the industry
Taken together, these conversations paint a clear picture of where the embedded industry is heading.
Artificial intelligence is no longer confined to data centres or high-end GPUs. Instead, it is being integrated across the entire spectrum of embedded hardware – from microcontrollers and FPGAs to IoT modules and development boards.
The emphasis is increasingly on Edge AI, where devices can process data locally, make intelligent decisions, and interact with the physical world in real time.
As developers continue to push for smarter, more autonomous systems, this trend is only expected to accelerate.
Check out more AI related news from embedded world on the Electronic Specifier website.
