AI Footprint Estimator

Build your monthly AI footprint profile from real workloads, compare vendors and regions — honest ranges, not fake-precise numbers.

Your workloads

per month
Cloud region
Datacenter PUE1.20
Electricity / month
0.72–1.44kWh
Emissions / month
0.27–0.53kgCO₂e

That is roughly…

  • 🚗 1.57–3.13 km in an average car
  • ☕ 5.33–10.7 espressos
  • 🔋 60.0–120 smartphone charges

Cleaner region

Same profile in Europe (Stockholm) (25 gCO₂/kWh): 0.018–0.036 kgCO₂e — −93%

Vendor comparison — same workloads, kgCO₂e / month

VendorTypical hostingCleanest option
OpenAI
Serves from Microsoft Azure, predominantly US — typically US East (Virginia).
0.22–0.53
US East (N. Virginia)
0.015–0.036
Europe (Stockholm)
Anthropic
Serves from AWS and Google Cloud, US-heavy.
0.17–0.41
US West (Oregon)
0.17–0.41
US West (Oregon)
Google Gemini
TPU serving on GCP; the only provider with a published per-prompt figure (0.24 Wh median).
0.16–0.45
US Central (Iowa)
0.13–0.36
Europe (Amsterdam)
Self-hosted (open weights)
Own GPUs: lower batching and utilization than hyperscale serving, usually higher PUE.
0.44–1.78
US East (N. Virginia)
0.030–0.12
Europe (Stockholm)

Vendor factors include typical serving efficiency and PUE — the PUE slider applies only to the main estimate above.

Grid right now — US East (N. Virginia)

Carbon intensity windows for your region are a Pro feature.

Save this profile & track it monthly

Sign in to save workload profiles, export CSV and generate branded PDF reports.

Sign in →

Ranges reflect real uncertainty (±3–5× on per-token energy). Location-based grid intensities. Training is not amortized into queries. Full methodology · v2026.06

AI Footprint Estimator — CO2Radar