Member of Inception Program

On-premise or cloud: run the numbers

Set how many people work with AI. The calculator estimates GPU demand and compares the cost of your own servers with rented cloud GPUs over three years.

20
Chat, search, and documents in daily work
5
Heavy use: coding assistants, agents, batch jobs

On-premise, 3 years
Cloud, 3 years rented GPUs, no residual value
Your savings
Break-even

Cumulative cost over 3 years, in euros

Six-month overview, in euros

MonthOn-premiseHardware book valueCloudDifferenceDifference incl. hardware sale

Assumptions

  • One developer occupies 1/8 GPU, one daily user 1/16 GPU.
  • On-premise: entry with Server S2 (2 GPUs) at 40,000 € one-time, then 10,000 € per year per 2 GPUs for software and operations. Capacity grows in 2-GPU steps. These amounts are not final: GPU and hardware prices are currently volatile and are confirmed in the quote.
  • The hardware is yours and loses 20% of its value per year. The book value stays on your balance sheet as an asset; the cloud leaves no residual value.
  • Cloud: 2.15 € per GPU hour in continuous operation (8,760 hours per year), plus 15% DevOps overhead.
  • Indicative figures excluding power and facility costs. For a binding quote, talk to us.