On-prem vs. cloud: where's your break-even?

Move the sliders to your situation. The math is the same sheet we open in every discovery call — including the honest answer when the cloud wins. Nothing you enter leaves your browser.

Requests per day 400 Avg. tokens per request (prompt + answer + RAG context) 2500 Cloud price, € per 1M tokens (blended) 8.00 On-prem hardware budget, € 4000 Electricity, € per kWh 0.30 Maintenance & ops, € per month 80

Cloud / month

€0

On-prem / month

€0

Assumptions: ~350 W average GPU-server draw at 50% duty cycle; one server handles this volume up to ~30M tokens/month on a 7B-class model. Orientation only — your real numbers belong in a worksheet. See the full cost-math guide.

Want this calculated with your real workload?

We run the full version — multiple use cases, hardware quotes, electricity tariffs, your actual token logs — in the first call. Free, 30 minutes.

Run my real numbers