Arduino Opens VENTUNO Q Preorders for Physical AI

Arduino's VENTUNO Q moves from announcement to preorder

Arduino opened preorders for VENTUNO Q on August 25, 2026, moving its high-end physical-AI development board from an earlier product announcement into commercial preorder availability.

The board's central design is a dual-brain architecture: a Qualcomm Dragonwing IQ-8275 processor handles Linux and AI workloads, while a dedicated STM32H5 microcontroller provides deterministic real-time control for motors, CAN-FD, PWM and GPIO.

Arduino lists an introductory preorder price of $299, with its U.S. store saying delivery is expected in about four weeks. That means the product is available to preorder now, but broad customer delivery is still upcoming.

On-device AI and real-time control on one board

Qualcomm says the Dragonwing IQ-8275 portion of VENTUNO Q delivers up to 40 dense TOPS of NPU performance. The platform includes 16 GB of LPDDR5 memory, 64 GB of eMMC storage and an M.2 NVMe expansion path.

The Linux side ships with Ubuntu, while the real-time controller runs Arduino Core on Zephyr RTOS. The two processors communicate through an RPC bridge, giving developers a way to combine relatively heavy AI inference with time-sensitive control loops without splitting the application across multiple development boards.

The board also includes features aimed directly at robotics and machine vision, including ROS 2 compatibility, CAN-FD, PWM, triple MIPI-CSI camera inputs and 2.5 Gb Ethernet.

Local LLMs, VLMs and robotics workflows

Arduino and Qualcomm position VENTUNO Q for local inference rather than cloud-only AI. Through Arduino App Lab, developers can run NPU-optimized models for use cases including language, vision-language, speech recognition, object detection and gesture detection.

Qualcomm says developers can also bring GGUF-format models from Hugging Face or train and quantize custom models through Edge Impulse Studio. The platform supports Qualcomm's GenieX runtime for on-device generative AI.

That combination makes VENTUNO Q relevant to autonomous mobile robots, industrial inspection, predictive maintenance, smart-city systems, drones, robotic arms and other applications where sensing, inference and physical action need to happen locally.

A planned path from prototype to production

A second part of the announcement is the new Works with Arduino pathway. Qualcomm says the same Dragonwing IQ8 architecture is expected to become available in production-grade system-on-module formats from partners including SECO and Toradex.

This production pathway is upcoming, not yet the same thing as the VENTUNO Q preorder. The idea is that developers could prototype on Arduino hardware and later move the same software stack and model logic into industrial SOM-based systems.

Why this matters for edge AI

Many edge-AI boards can run neural-network inference, while many microcontrollers can provide deterministic control. VENTUNO Q is notable because it deliberately places both roles in the same developer platform.

For physical-AI builders, that reduces the gap between perception and action: a vision or language model can run on the Linux/NPU side while the MCU executes hard real-time control.

The board is not a robot and does not by itself provide autonomous behavior. Its significance is as infrastructure for developers building robots, industrial machines and local AI systems that need a combination of multimodal inference and real-time actuation.

The correct status as of the launch is: VENTUNO Q is open for preorder at $299, customer deliveries are still upcoming, and the broader Dragonwing IQ8 production-SOM pathway is planned for later.