ASRock Launches AI BOX-A395: Compact, Powerful AI Workstation with Ryzen AI Max+ 395 | Edge AI & Deep Learning

TAIPEI, March 23, 2026 – ASRock Industrial today unveiled the AI BOX-A395, a compact system designed to deliver high-performance artificial intelligence capabilities in a small form factor. The system, powered by AMD Ryzen™ AI Max+ 395 processors, aims to bring advanced AI processing closer to the data source, reducing reliance on cloud infrastructure.

The AI BOX-A395 integrates a CPU, GPU, and NPU, offering up to 50 TOPS of AI acceleration, according to ASRock Industrial. It supports up to 128GB of LPDDR5x-8000 unified memory, enabling the processing of large AI models and data-intensive workloads directly on the device. The company positions the system for use by enterprises, developers, and system integrators seeking to deploy AI applications locally.

The AMD Ryzen™ AI Max+ 395 processors feature Zen 5 CPUs with up to 16 cores and 32 threads, coupled with an AMD Radeon™ 8060S Series GPU and an AMD XDNA™ 2 NPU. The AI BOX-A395 is compatible with both Windows and Linux operating systems.

Connectivity options include two USB4 ports, one USB 3.2 Gen2 Type-C port, two USB 3.2 Gen2 ports, and two USB 2.0 ports. The system also features dual Ethernet ports, offering 10GbE and 2.5GbE LAN speeds. Quad-display output is supported via two HDMI 2.1 and three DisplayPort 2.1 interfaces, enabling resolutions up to 8K.

Storage options include two M.2 Key M slots supporting PCIe Gen4 x4 SSDs with RAID 0/1 configuration, and one M.2 Key E slot for Wi-Fi 7 and Bluetooth 5.4. The system employs a thermal architecture with six heat pipes and a copper base to manage heat dissipation during intensive AI workloads.

Housed in a 200 x 100 x 232mm aluminum chassis, the AI BOX-A395 includes onboard TPM 2.0 and redundant BIOS support. ASRock Industrial, established as an independent entity in 2018, specializes in industrial motherboards, edge computers, and related solutions. The company’s stated vision is to “cocreate an intelligent world,” focusing on industries including commerce, automation, robotics, entertainment, and security.

According to a recent post on Reddit, some users have expressed reservations about the system’s use of RDNA 3 architecture, suggesting that waiting for RDNA 4, with its FP8 support and anticipated performance improvements, might be a more advantageous strategy.