Hardware
What I actually recommend for running AI at home - picked for value, not for the biggest spec sheet.
Complete Systems
Ryzen 7 Mini PC (32GB RAM)
A buying profile for a 32GB Ryzen 7-class mini PC: enough RAM and CPU headroom for 7-8B models, no GPU required.
$500-600Prebuilt RTX 4090 Workstation PC (24GB VRAM)
For running 27B-30B class models comfortably without building a PC yourself - the tier above the mini PC and above a mid-range GPU upgrade.
$3,000-4,500+Mac Studio (M5 Max / M5 Ultra)
For running local AI as part of how you make money, not just how you learn - unified memory up to 512GB, quiet, low power draw, no PC to maintain.
$5,499-10,000+Upgrades & Components
NVIDIA GPU (12GB+ VRAM)
For when a mini PC's CPU-only ceiling isn't enough - larger models, faster responses, or local image generation.
$500-750+DDR5 RAM Upgrade Kit (32GB)
The single biggest lever on a CPU-only setup, and still worth getting right on a GPU build - what to check before buying a kit.
$375-475+ (see note on pricing below)PSU / Power Supply (850W+, 80+ Gold)
The unglamorous part of a GPU build that is genuinely worth not cheaping out on - real headroom above the GPU's rated draw.
$90-180External NVMe SSD (1TB+)
Portable storage for when your model collection outgrows internal storage - fast enough that load times barely suffer.
$160-260Accessories
As an Amazon Associate I earn from qualifying purchases. Some links on this page are Amazon affiliate links; they may earn a small commission at no extra cost to you - one way this site stays free.