The honest first-homelab build, 2026

Everyone starts with “I just want Plex.” Six months later you’re running 14 Docker containers, Home Assistant, and half a *arr stack on a box you bought for $189. That’s fine. That’s the point.

This guide skips the mythology and gives you the real numbers.

Why a mini-PC beats a NAS for starters

A purpose-built NAS is great hardware. But for a first build it has two problems:

1. Locked software. DSM and TrueNAS are excellent but constrain your stack. You end up wanting Docker, VMs, or a custom app — and the NAS fights you.
2. Wasted spend. A 4-bay NAS enclosure + drives + RAM frequently costs 2× a mini-PC that does everything and runs on 6W.

The exception: if you need >40TB from day one, start with a proper NAS. Otherwise, a mini-PC is the right call.

The $189 build: Intel N100

The Beelink EQ12 has become the default recommendation for a reason:

• Intel N100 processor — 4 cores, Quick Sync hardware transcoding
• 16GB DDR5, 500GB NVMe included
• Dual 2.5GbE — enough for a separate LAN and IoT VLAN
• ~6W idle. At €0.30/kWh that’s about €1.30/month to run 24/7

Quick Sync matters. Software transcoding a 4K HEVC stream burns 3–5× more CPU than hardware. With Quick Sync, Plex or Jellyfin handles 4K → 1080p in real-time at under 30% CPU.

What it runs well

• Jellyfin / Plex (1–2 simultaneous 4K transcode streams)
• Home Assistant
• Full *arr stack (Sonarr, Radarr, Prowlarr, Bazarr)
• ~15 Docker containers concurrently
• Nextcloud (files + photos)

Where it hits the ceiling

• 3+ simultaneous software transcodes
• Heavy databases (PostgreSQL, InfluxDB at scale)
• Proxmox with multiple VMs running simultaneously

Upgrading the storage

The N100’s included 500GB NVMe is your OS + app drive. Separate your media.

Option A — USB NAS expansion: A USB 3.2 4-bay enclosure (~$80) + two 14TB WD Reds gives you ~24TB usable in RAID1 pairs. Simple, works fine for media.

Option B — Dedicated NAS node: A second mini-PC with a HBA card and 4× drives. More overhead, more flexibility.

Option C — Grow into it: Start with the 500GB NVMe and an external 8TB drive. When you need more, you’ll know what you actually need.

Most people do Option C and are glad they didn’t over-engineer it.

RAM: 16GB is usually enough

If you’re running a lot of VMs, 32GB is worth it. For pure container workloads, Linux’s memory management and container limits mean 16GB handles 20+ containers without swapping.

Monitor actual usage with free -h after a week. Only upgrade if you’re consistently above 85%.

The real costs

Compare to cloud: a basic VPS with 4 cores, 16GB, 500GB runs $60–80/month. Your homelab pays for itself in under 9 months.

Software stack recommendation

OS:          Proxmox VE (bare metal, free)
Containers:  Docker via Portainer or Dockge
Media:       Jellyfin (open source) or Plex
*arr:        Sonarr + Radarr + Prowlarr + Bazarr
Automation:  Home Assistant OS (in a VM)
Files:       Nextcloud AIO (Docker)
Reverse proxy: Caddy (dead simple HTTPS)

Proxmox on bare metal gives you one Linux VM for Docker containers and one isolated VM for Home Assistant — clean separation with near-zero overhead.

FAQ

Do I need ECC RAM? No, not for media and containers. Yes if you’re running ZFS on spinning rust for important data. The N100 doesn’t support ECC anyway.

Should I get the 8GB or 16GB version? Always 16GB. The $20–30 difference is not worth the headache of running out.

Is the N100 fast enough for Stable Diffusion? For CPU inference on small models, barely. For any real AI workload, you want a GPU. See our AI mini-PC guide.

What about a Raspberry Pi? Fine for Home Assistant alone. Too slow and too storage-constrained for a full homelab stack.

When to step up

If you find yourself wanting any of these, it’s time to look at the N305 or MS-01:

• 3+ simultaneous streams
• Proxmox with 4+ VMs
• PCIe GPU for local AI
• 10GbE networking
• 8+ drive bays

The Homelab Planner walks through your specific workload and gives you a concrete build recommendation.