ROI calculator: cloud vs local AI
Compare the total cost of inference on your own hardware with a cloud API, using published API prices. Move the controls and the result updates instantly.
10050.000
1004.000
660 (5 years)
Default 0.25 €/kWh. Use the one on your bill.
It adds nothing to the figures: support is quoted by scope, in writing, after the call. Ticking it adds it to your request.
Optional: value of the time AI saves
These hours are saved whether the AI runs in the cloud or on your hardware, so they do not change the comparison. They are shown separately.
040 h
€10€80
If the volume changes
| Queries/mo | Cloud | Local | Difference | Payback |
|---|
Your current setup, varying only queries per month.
Sources and assumptions
- API prices: the official OpenAI and Anthropic pricing pages, checked 2026-09-09. Models marked indicative have not been re-checked since April 2026. Dollar prices are converted at €0.92 per dollar.
- Token split: 60% input and 40% output, typical of chat and document search.
- Local cost: purchase price plus electricity with the device on 24 hours a day at its maximum draw. Installation, support and maintenance are not included.
- A local model is not identical to the paid model you compare it with. Measure it on your tasks before you decide.
Compare with per-person subscriptions
ChatGPT Plus €18/mo –
Claude Pro €18/mo –
Gemini Advanced €22/mo –
Copilot Pro €18/mo –
Price per person. The total is for one person over the chosen months.