🌍 GPU Carbon Footprint Calculator
Calculate the environmental cost of your AI training workloads in real-time
326
kg CO2 / MWh (US avg)
700W
H100 TDP
30%
Renewable energy (avg)
1.4x
Cooling overhead
Training Configuration
Environmental Impact
↵ Configure your training
Enter parameters to calculate environmental impact
🌱 How to Reduce Your Carbon Footprint
- Choose green regions: Iceland (100% geothermal) or Nordic countries (95% renewable)
- Optimize batch sizes: Better GPU utilization = less wasted energy
- Use mixed precision: FP16 training uses 50% less energy than FP32
- Schedule wisely: Train during off-peak hours when grid is cleaner
- Consider TPUs: 2-3x more energy efficient for certain workloads
📊 Carbon Intensity by Provider
🔬 Methodology
Our calculations follow the ML CO2 Impact framework:
- Power consumption: GPU TDP × Utilization × PUE
- CO2 emissions: Power (kWh) × Grid Carbon Intensity (kg CO2/kWh)
- PUE factor: Includes cooling, power conversion, and infrastructure overhead
- Grid data: Updated monthly from EPA eGRID and EU EEA databases