What does AI really cost per head?
The average lies. Calculate with median consumption, activation and full cost — and see why consumption is the smallest item and adoption the biggest.
Before you calculate, one line of framing: “cost” here means the actual compute consumption of usage, not the license price — and the average across all users lies, because a small circle of power users drives most of it. The calculator therefore shows the evidenced median consumption, the activation and the full cost separately. How the numbers come about and how to read them is right below the calculator.
Your inputs
- Adoption & enablement~50%
- Integration & operations~20%
- Governance & security~20%
- Usage / consumption~10%
The visible license price only covers the smallest block. The biggest lever is adoption.
Directional: the usage consumption (smallest item) and adoption as the largest block are evidenced. Total TCO is a range, not a measured benchmark.
Value (3 factors) − full cost (TCO incl. adoption)
Value assumed at average usage. A range, not a point forecast — full cost is a directional estimate.
And what does AI bring? Open the value calculatorYour numbers as a personal report
Your cost, the maturity assessment and the next steps for your stage — as a printable document, tailored to your inputs.
Turn cost into a business case
The playbook shows you how to steer adoption where use cases create real value — and make full cost controllable.
Open the playbookA benchmark from real usage, not from a brochure
The consumption figures in the calculator come from anonymised meinGPT usage data: over 10,000 users, twelve months, analysed as a distribution rather than a marketing promise. “Cost” here means the actual compute consumption of usage — what requests and generated text cost in compute — not the license price. A fixed price per head only shifts the consumption risk, it does not remove it.
The median is €3.55 per active user and month (12-month median). Since the premium models launched, the running median sits at around €6.50 — that stays compute consumption, not license. These are exactly the values the usage level in the calculator uses, so you compute with evidenced numbers instead of a gut feeling.
The average lies
Why the average lies — and which three numbers really count
AI usage follows a power distribution, not a bell curve: a small circle of power users drives most of the consumption. In the benchmark, the top 10 % of users cause around 78 % of the cost. That is why median and mean diverge sharply — and an average that presses both groups into one number carries no decision at all.
For planning you therefore need three numbers instead of one: the median as the baseline for the mass, the peak values (P90/P99) as where the budget really sits, and the activation rate — the share that actively uses AI at all. These three levers are exactly what you find in the calculator above: usage level, consumption and activation.
Why the average lies — read the full analysisFour cost blocks — and the smallest is the one on the quote
The total cost of ownership of an AI rollout has four blocks: adoption & enablement, integration & operations, governance & security, and running usage. The compute consumption — the item that holds the €3.55 to €6.50 — is the smallest. Around half of TCO goes to adoption: training, use-case development, change management. Exactly the block that is on no quote.
Where the money really goes
That also explains why, according to Bitkom, 33 % of companies say AI is more expensive than expected: the surprise is not in the consumption, but in the other blocks. And it makes adoption the biggest lever — cutting training lowers the bill and the usage at the same time. The visible license price is only the wrapping around the smallest block.
The four cost blocks in detailThe full analysis: AI cost per employee
Median, peaks and activation, the four TCO blocks and what Gartner (~3 %) and Forrester (116–353 % over three years) say about ROI — the detailed pillar behind this calculator.
Read the full analysisGet the full benchmark as a whitepaper
The median consumption, the 78 % concentration and the steerability playbook in three levers — from 10,000+ users over 12 months. Instantly as a PDF, by email.
Get the whitepaperWhat this is about
The AI Cost Calculator shows your real AI cost beyond the license. What's evidenced is the usage consumption per active user (median benchmark from 10,000+ users) — the smallest cost block. Around it sits the total cost of ownership: adoption, integration and governance, with adoption typically the largest item. Total TCO is a directional range, not a measured benchmark. For anyone planning an AI budget they can defend.
Good to know
- What does “cost” mean — the license or the consumption?
- The actual compute consumption of usage, not the license price. A fixed per-head price only shifts the consumption risk, it doesn't remove it.
- Why is total TCO a range?
- The usage consumption (smallest item) and adoption as the largest block are evidenced. Integration and governance are directional — total TCO is therefore an estimate expressed as a range, not a measured benchmark.
- Does the calculator cost anything?
- No, it's free.