Graue Zellen
Measure it, don't spin it.
Companies now use AI everywhere and have no idea what it costs in energy. Graue Zellen pulls usage from the admin interfaces of the AI tools in play, converts it into energy, CO₂ and water, and turns that into a traceable monthly statement.
What it's about
Graue Zellen makes a company's AI consumption visible. Usage comes from the admin interfaces of the tools in use: token-level for Claude, usage counts for Microsoft 365 Copilot and GitHub Copilot. From that we model energy, CO₂ and water. Every figure is labelled as measured or estimated, uncertainty range included.
Today that footprint is small in most companies. But it grows fast, and it grows unmanaged: a single agentic sprint multiplies consumption. Start measuring now and in two years you have a history when clients, tenders or the sustainability report ask. Start then and all you have is an estimate.
It is built at Degeler Consulting together with our partner Phillipp Risch, who brings the energy side. In parallel, the calculation method is being verified scientifically with dedicated measurements before any number goes into a report. As of August 2026: we are looking for pilot companies.
What Graue Zellen deliberately is not: an offsetting shop. Measure first, then reduce. Compensation is the last step at best, and it replaces neither of the two before it.
Who it's built for
Companies using AI at scale that don't want a blank space in their first sustainability report.
Sustainability leads who need defensible numbers instead of a gut feeling derived from the licence agreement.
Managing directors who want to know which AI usage in the house is expensive, in euros as well as in kilowatt-hours.
Organisations with a green profile that don't want their own AI use to stay a blind spot.
Why it works
Connect, don't fill in forms
The company grants access to its AI tools once, after that usage data flows in automatically. Nobody fills in questionnaires or retypes invoices.
Measured and estimated kept apart
Usage is measured, energy is modelled. No vendor publishes kilowatt-hours per request. That is exactly why every figure carries its confidence level instead of a decimal place that fakes precision.
A statement every month
Instead of one number in the annual report there is a monthly statement per AI service, with the trend against the previous month. That lets you evidence a reduction rather than claim one.
Reduce, don't buy your way out
The data produces the uncomfortable recommendation: which model is oversized for the task, which automation runs idle. Compensation comes after that, clearly labelled as such.
Who builds this with us
Every project has a different constellation. These are the people behind this one.
Graue Zellen
How much energy, CO₂ and water a company's AI use actually costs.
Companies now use AI everywhere and have no idea what it costs in energy. Graue Zellen pulls usage from the admin interfaces of the AI tools in play, converts it into energy, CO₂ and water, and turns that into a traceable monthly statement.
Let's justtalk.
You'll leave with an honest read you can actually use internally – whether we end up working together or respectfully part ways. No pitches. No pressure.