I took a look at the Minisforum AI X1 Pro because I keep seeing “AI everything” plastered all over product pages, and I wanted to know if the NPU (AMD’s XDNA engine) is actually useful for home lab stuff. On paper, the idea is awesome: efficient object detection for cameras, facial recognition for a photo library, maybe even lightweight LLMs—without hammering your CPU or GPU. In reality though, the short answer is no… at least not yet. The biggest theme in this video is software support: getting the NPU to do anything useful took a lot of tinkering, and Linux support is still lagging behind. Outside of the NPU, the X1 Pro is legitimately a really impressive mini PC. It’s quiet, efficient, and surprisingly powerful for its size, with a built-in power supply, dual 2.5GbE, USB4, and even OCuLink for PCIe expansion. I tore it down, talked through the specs (including the 3x NVMe slots—one of which is only a single PCIe lane), ran a few benchmarks, and messed with Proxmox where I got idle power down to about 7W. For the NPU testing, I tried Windows features like Live Captions (which finally worked after updates), played with NPU image generation in Amuse, and followed AMD’s tutorials for image recognition where the NPU was way faster and more power-efficient than the CPU in my little “benchmark.” I also got AMD’s hybrid LLM setup running (GPU + NPU), but you’re pretty limited on models and it’s not exactly plug-and-play yet.