n8n is light to start and hungry under load. A single user testing workflows is happy on 2 GB of RAM; a production instance running dozens of active workflows, queue mode, and a Postgres database wants 4 vCPU and 8 GB. Pick the VPS for where you’re going, not where you start.
The numbers below are estimates from n8n’s docs and community reports, not measured benchmarks. Use the cost calculator to size your exact setup.
What n8n needs
| Setup | vCPU | RAM | Storage |
|---|---|---|---|
| Testing / single user | 2 | 2 GB | 20 GB |
| Production (many workflows, queue mode) | 4 | 8 GB | 80 GB |
| n8n + local model (Ollama) | 4 | 12 GB+ | 80 GB |
n8n itself is Node.js and modest; the RAM goes to concurrent executions, the bundled Postgres, and anything you self-host alongside it.
The four hosts, for a production n8n box (4 vCPU / 8 GB)
| Host | Plan | Specs | $/mo | Notes |
|---|---|---|---|---|
| Hetzner | CPX22 | 4 vCPU / 8 GB / 80 GB NVMe | ~$9 | Best price-to-performance. Editorial pick — no affiliate. |
| Contabo | VPS S/M | 4–6 vCPU / 8–16 GB / 150–400 GB | ~$6–11 | Most RAM and storage per dollar. |
| DigitalOcean | Basic 8 GB | 4 vCPU / 8 GB / 160 GB SSD | ~$48 | Best docs and dashboard; priciest. |
| Vultr | Regular 8 GB | 4 vCPU / 8 GB / 160 GB NVMe | ~$36 | Many regions, hourly billing. |
Our pick
- Cheapest that’s genuinely good: Hetzner CPX22. If you’re comfortable in a terminal, nothing beats it on value. We don’t earn anything if you pick it — we still recommend it.
- Best value with the most headroom: Contabo VPS M (16 GB RAM) for a few dollars more — great if you’ll add Ollama or a vector DB later.
- If you want polish and support: DigitalOcean, at a premium.
Size your specific workload — including a local model or Supabase alongside n8n — with the VPS cost calculator.