
Radxa's 2026 Qualcomm Lineup: Dragon Q8B, Q5E SBCs and AI-Ready NAS
Radxa unveiled an ambitious 2026 roadmap of 22 Qualcomm-powered products, led by the Dragon Q8B and Q5E Arm SBCs and AI-capable DragonStation and DragonBay NAS.
Radxa came to its joint developer day with Qualcomm carrying one of the most ambitious roadmaps the Arm single-board computer world has seen this year: 22 Qualcomm-powered products for 2026, headlined by two new SBCs and a pair of AI-capable network storage systems. For homelab builders and Arm enthusiasts, there's a lot to like here.
Dragon Q8B and Q5E Lead the SBC Charge
The flagship board is the Dragon Q8B, built around a Snapdragon 8cx Gen 3 octa-core SoC with up to 32GB of LPDDR4x, 29+ TOPS of AI throughput, dual 2.5GbE networking, and a PCIe 3.0 slot on a compact 100 x 75 mm footprint. It's a genuine workstation-class Arm board.
Alongside it, the smaller Dragon Q5E trims things down for efficiency, pairing a 4nm Dragonwing QCS6690 octa-core Kryo SoC with 6 TOPS and the same dual 2.5GbE networking on a tidy 65 x 56 mm board. Together they give developers a high-end and a mainstream option from the same family.
AI-Capable NAS: DragonStation and DragonBay
The storage side is where things get especially interesting. The DragonStation is a 6-bay all-NVMe NAS with 10GbE and an onboard AI accelerator card that Radxa says can run local models with up to 120 billion parameters and agent frameworks — a self-hosted AI box and a fast file server in one chassis. The more approachable DragonBay is a 4-bay NAS aimed at media streaming, photo archiving, and multi-user collaboration.
A Roadmap Built for Builders
Beyond the headline products, the full lineup spans Raspberry Pi CM-compatible compute modules, a Dragonwing IQ-9075 cluster, and even a RoboX Q1000 robotics platform. That breadth matters: it gives the maker and homelab community a coherent set of Arm building blocks, from tiny modules to clustered compute, all sharing a software ecosystem.
Why Radxa's Qualcomm Push Is Good News
More competition in the Arm SBC and self-hosted AI space means more choice, better tooling, and lower prices for everyone building at the edge. With capable NPUs across the range and NAS systems designed to run local large language models, Radxa's 2026 roadmap is a strong vote of confidence in open, self-hosted computing. Pricing and availability will roll out across the year.
Sources: CNX Software (June 1, 2026); Liliputing (June 2026).
