AMD’s showcase concept for a high-performance local AI desktop, built around its Strix Halo platform, has now become an orderable product. A listing on MicroCenter’s online store confirms that the Ryzen AI Halo mini PC is slated for release on July 10, priced at $3,999, with the retailer identified as the exclusive sales partner for the system.
Hardware built for demanding local inference
The compact machine is driven by the Ryzen AI Max+ 395 APU, housing an integrated GPU with 40 compute units and access to as much as 128 GB of shared memory. That resource pool is designed to handle sizable large language models on-device, removing the need for cloud offload. AMD has highlighted compatibility with extensive model families, listing support for GPT OSS at 120 billion parameters, Qwen 3.5 at 120 billion parameters, and smaller architectures such as GLM 4.7 Flash and Qwen 3.6 35B.
Familiar silicon with a sizable price position
The same Strix Halo silicon already appears in other premium mini PCs targeting gaming and local AI workloads. One recent example, the GMKtec Evo-X2, currently sells for $3,399. In testing, that device delivered CPU throughput on par with a Ryzen 9 7945HX Dragon Range processor, while its Radeon 8060S integrated graphics traded blows with a discrete RTX 4070 Laptop GPU and surpassed the RTX 4070 Laptop variant inside the AtomMan G7 Ti mini PC.
When AMD positions the new system against Nvidia’s DGX Spark at a similar price level, it claims a performance lead ranging from 4 to 14 percent. The mini PC also adds flexibility by supporting both Windows and Linux, whereas the Spark is confined to Linux. The networking story, however, tilts in Nvidia’s favor: where the DGX Spark includes a 200 GbE ultra-fast interconnect, AMD’s box is equipped with a considerably more modest 10 GbE port, a trade-off that could shape deployment decisions for multi-node setups.
A controlled retail launch
By anchoring distribution to a single retailer at launch, AMD appears to be taking a measured approach to the halo-class mini PC segment. The listing underscores a growing category of desktop hardware that repurposes potent integrated graphics and unified memory architectures not just for gaming, but as practical platforms for running frontier AI models on the edge.
Sources: www.microcenter.com, videocardz.com