Electronica 2026: Forlinx Showcases Qualcomm Dragonwing AI Solutions

Forlinx showcases new modules, robotics controllers and on-device AI solutions built on technologies from Qualcomm Technologies, NXP, NVIDIA, Rockchip and DEEPX.

MUNICH, Germany, November 12, 2026 — Forlinx Embedded is showcasing new embedded computing platforms and live on-device AI solutions at electronica 2026, taking place November 10–13 at Messe München. At Hall C5, Booth 331, we combine processor technologies from across our semiconductor ecosystem with Forlinx hardware and application software to support industrial and robotics product development.

A key addition to this year’s showcase is our Qualcomm Dragonwing-based portfolio, featuring two system-on-modules and live demonstrations of local image search and multi-stream video analytics. We are also introducing platforms based on NXP, Rockchip and NVIDIA technologies, alongside DEEPX AI accelerator cards.

Our demonstrations address practical development requirements: processing multiple camera feeds, combining sensor data, running AI locally and integrating these capabilities into an embedded product. Visitors can explore the systems in operation and discuss their project requirements directly with our team.

New Modules and AI Acceleration Platforms

Qualcomm Dragonwing-Based Modules and Ecosystem Engagement

Forlinx is showcasing two Qualcomm Dragonwing-based system-on-modules: the
FET9075AA-C, based on the Qualcomm Dragonwing IQ-9075 processor, and the
FET8275AA-C, based on the Qualcomm Dragonwing IQ-8275 processor.

The FET9075AA-C configuration on display is rated at
100 dense TOPS, while the FET8275AA-C configuration is rated at
40 dense TOPS. These modules expand our portfolio for customers developing industrial vision and on-device AI applications.

During electronica 2026, our team met with Qualcomm representatives and participated in an on-site interview with Qualcomm, alongside our showcase of Dragonwing-based modules and live edge AI demonstrations.

We combine Qualcomm Dragonwing processor technology with Forlinx embedded hardware and application software to support the development of industrial vision and on-device AI systems.

Our team can help you evaluate camera integration, AI workloads and system requirements using these modules and the applications demonstrated at the booth.

NXP i.MX952 and Rockchip RK3588 Series Modules

Our
FET-MX952x-C system-on-module, based on NXP’s i.MX952 processor, combines application processing, real-time control, on-device AI acceleration and hardware video processing. It supports industrial and robotics designs that need to integrate control functions with camera or display workloads.

The new
FET3588S2-C expands our Rockchip RK3588 series portfolio for applications requiring multimedia processing, display connectivity and multiple data interfaces.

Our team can help you evaluate the module, carrier board and software together against your application requirements, including camera inputs, display configuration and peripheral connectivity.

DEEPX AI Acceleration

We are also showcasing the
DEEPX DX-M1 and DX-M1M accelerator cards. Using an M.2 interface, the cards deliver
25 TOPS of INT8 AI performance, with typical power consumption of
3–5 W.

The cards give customers an additional option for integrating AI acceleration into embedded systems for local vision inference and other AI workloads.

NVIDIA-Based Controllers for Robotics and Embodied AI

Forlinx is introducing two controllers for robotics and embodied AI: the
FCU3021, based on NVIDIA Jetson Orin NX, and the
FCU3031, based on NVIDIA Jetson AGX Thor.

These controllers expand our robotics portfolio to address different computing requirements for applications combining AI inference, environmental perception and motion-control integration.

Our live FCU3021 demonstration combines a depth camera with LiDAR, bringing camera imagery and range measurements together for near-field environmental perception. The demonstration presents the camera view, LiDAR point cloud and combined output for comparison.

For your robotics project, our team can discuss sensor connectivity, data synchronization and how perception processing fits alongside the robot’s control requirements.

On-Device AI Solutions in Action

Local Image Search and Video Analytics with Qualcomm Dragonwing

Our
IQ-9075-based image search demonstration retrieves previously captured images using a text description. The system detects and tracks people, encodes images of those targets using SigLIP2 and stores their features in a local index. It then compares a text query with the indexed features to return images ranked by similarity.

Capture, indexing and search all run locally. You can try different descriptions at the booth and see how the system retrieves matching images from the indexed targets.

Our
IQ-8275-based video analytics demonstration processes
four live USB camera feeds and 32 local video streams, with detection results overlaid on a tiled display. Individual channels show how industrial safety analytics are applied to the video.

The 36-input workload combines live USB camera feeds with local video streams. It does not represent 36 concurrent physical cameras.

Two-Stage Vision Processing with NXP i.MX9596 and ARA240

Our
i.MX9596 +
ARA240 vision solution divides processing into two stages. The i.MX9596 first detects a person in the camera feed. Selected frames then pass to the ARA240 accelerator for further analysis, such as describing clothing or recognizing an action.

Both stages run locally. This architecture separates initial detection from more detailed image analysis, allowing the two tasks to run on different computing resources. Our team can discuss how this approach could fit your camera configuration, target model and response-time requirements.

Panoramic Vision and Multi-Channel Analytics

Our
RV1126B-based panoramic vision solution corrects fisheye lens distortion, stitches multiple camera views into a panoramic image and performs person detection on the combined view. The solution addresses wide-area monitoring and robot surround-view requirements.

Our
FCU3502 edge AI computing box, based on an RK3588 + RK1828 architecture, runs analytics across
36 network video streams at 720p and 20 fps using H.265 encoding in the booth demonstration. The multi-channel display shows the video feeds and their corresponding analytics results.

Synchronized Power Monitoring and Local AI Interaction

Our
RK3576 AMP + FPGA 64-channel power data acquisition solution combines synchronized acquisition, data processing and visualization. The FPGA coordinates synchronized sampling, while the processor handles subsequent processing and waveform display for industrial power-monitoring applications.

Our
offline digital human Q&A solution integrates speech recognition, retrieval-augmented generation and an animated avatar. In the demonstration, the system retrieves information from product documentation, generates a response and delivers it through the avatar, with the workflow running locally.

This approach supports interactive applications where cloud connectivity is limited or product information needs to remain on site.

From Platform Evaluation to Product Development

We support embedded product development with system-on-modules, development platforms, hardware integration and software support. Across our semiconductor ecosystem, we help customers evaluate platform capabilities against the requirements of their products.

At the booth, our team is available to discuss processor selection, camera and display interfaces, AI model deployment, sensor integration and custom carrier-board development. Bring your input types, channel count, target models and interface requirements so we can discuss what to evaluate and how to plan the next stage of development.

Our demonstrations continue through
November 13 at Hall C5, Booth 331, Messe München. Visit us to explore the platforms, see the solutions in action and discuss your next design with the Forlinx team.