At Computex 2026, Qualcomm introduced the Dragonwing IQ10 Robotics Reference Design (RRD), the company's most comprehensive commitment to the physical AI market to date. The platform consolidates compute, sensor interfaces, deterministic I/O, networking, and a layered software stack into a single enclosed reference design for autonomous mobile robots (AMRs), industrial robotics, and humanoid platforms.
Early access begins in June 2026, with global commercial availability scheduled for September 2026.
A Full-Stack Platform for Production Robotics
The Dragonwing IQ10 RRD is built on the Dragonwing IQ10 processor, delivering up to 700 TOPS of AI performance through 18 Qualcomm Oryon CPU cores, multi-core NPUs, and a GPU architecture designed for on-device perception, planning, and reasoning — without requiring external accelerators.
Key hardware capabilities include:
- Native multimodal sensor ingestion: Up to 12 GMSL2 high-speed camera inputs alongside LiDAR, Time-of-Flight (ToF), IMU, and other sensors — all without separate bridging hardware, reducing latency between sensing and processing.
- Deterministic real-time I/O: PCIe Gen5, Time-Sensitive Networking (TSN), 10GbE, EtherCAT, CAN-FD, and USB interfaces for precision motion control and timing-critical actuation loops.
- Integrated connectivity: Wi-Fi 7, 5G, and 10GbE for both local robot control and cloud-connected fleet operations.
- Environmental hardening: Fully enclosed unit operating across -40°C to +70°C with integrated forced-air cooling, supporting 12V/24V nominal power inputs with over-voltage protection at 26V.
- Functional safety and security: An integrated safety island and platform-level OS security services for production industrial deployments.
Software Stack
The platform ships with an end-to-end robotics software stack:
- On-device AI runtimes for low-latency perception and decision-making
- ROS2 middleware for hardware abstraction and ecosystem compatibility
- Platform services for sensing, planning, and actuation
- Cloud-connected lifecycle management via Qualcomm AI Hub for deployment, monitoring, and OTA updates
- MLOps and DevOps tooling for AI model development, validation, and lifecycle management
- Ubuntu Linux OS for broad robotics software ecosystem compatibility
Out of the box, the platform supports core robotics building blocks: perception, navigation and localization, planning and control, manipulation, task orchestration, and natural language interaction for human-robot interfaces.
The Automotive Playbook Applied to Robotics
Qualcomm's entry into robotics mirrors the strategy that transformed its automotive business from a connectivity component supplier into a $4+ billion annual-revenue platform business. The Snapdragon Digital Chassis followed a consistent arc: enter with connectivity and cockpit compute, expand into safety-critical subsystems, build a software stack and OTA infrastructure, accumulate design wins, and generate long-duration revenue through software content.
Qualcomm's Nakul Duggal, EVP of Automotive, Industrial and Embedded IoT, explicitly described the robotics platform as building on the company's foundational low-latency, safety-grade technologies developed for automotive applications.
Early Access Partners
Early access partners include NEURA Robotics, Advantech, APLUX, Booster, Innodisk, MeiG, NEXCOM, Radxa, Thundercomm, and VinMotion, who are already exploring the platform's full capabilities.
On-Stage Incident Underscores Physical Challenges
The launch event was marked by an unexpected moment that quickly went viral. During Qualcomm's live presentation, a NEURA Robotics 4NE-1 humanoid — one of the platform's early access partners — suddenly collapsed on stage and had to be covered with a blanket by staff. The incident, captured on video and widely shared on social media, served as a vivid reminder of the "sim-to-real" gap that remains one of the embodied intelligence industry's most persistent challenges.
The irony was not lost on observers: the very platform designed to bridge simulation and physical deployment witnessed a real-world hardware failure in real time. NEURA Robotics, which had just announced a record $1.4 billion Series C funding round led by Tether, acknowledged the physical engineering hurdles that even well-funded companies must overcome.
The incident underscores a fundamental tension in the industry: while AI software, silicon, and capital ecosystems are rapidly maturing, the physical mechanics of bipedal robots — balance, reliability, and robustness under real-world conditions — remain works in progress.
