Local AI enthusiasts tend to look for compact, high-performance computers, such as the Mac mini M4 Pro, Mac Studio or DGX Spark from Nvidia (like the Dell Pro Max with GB10). AMD obviously thought that this market had a future, and has just announced that the Ryzen AI Halo development platform, presented in January at CES, was going to be released. Pre-orders should arrive quickly.
The mini PC is produced directly by AMD, which is rather unusual for the brand, which tends to rely on partners for graphics cards. The first generation will be available in June from Micro Center, an American reseller. It is powered by a Ryzen AI Max+ 395 processor (16 cores, 32 threads) which integrates a Radeon 8060S graphics card. The machine integrates 128 GB of RAM on a 256-bit bus (like the M4 Pro chips) and AMD has chosen a 2 TB SSD for storage. This is often the norm in this type of machine, we have fast Ethernet (10 Gb/s) and Wi-Fi 7. The mini PC is powered by USB-C, with consumption announced at 120 W.
AMD highlights compatibility with Windows, which neither Dell nor Apple’s machine can offer, but also GNU/Linux. In addition to the computing capacity of the GPU and its 40 units, the system on chip also integrates an NPU (50 TOPS). The machine is expected to be priced at $4,000 (excl. VAT), which is a little below the price of the Nvidia models… which nevertheless retains some advantages, such as CUDA compatibility (a widely used in-house API) or a more efficient GPU.
The most surprising thing, in reality, is that AMD is already planning what comes next. Models equipped with Ryzen AI Max+ Pro 495 chips are expected from major manufacturers (HP and Lenovo at least) in the third quarter of 2026, that is to say in less than six months. If the system on chip is not fundamentally faster and keeps the same structure – 16 cores, 40 units for the GPU, a 50 TOPS NPU – the RAM will increase to 192 GB, which probably implies a significantly higher price in these times of RAMpocalypse. Variants with Slower systems-on-chips are also planned.
In any case, AMD is trying to find its place among local AI enthusiasts, especially for those who want to integrate agentic systems like OpenClaw without having to use expensive cloud servers.
