AI Model Training Cost Calculator
Estimate the energy, electricity cost, carbon footprint and infrastructure required to train an AI model based on GPU count, GPU type and training hours.
Important: These tools provide educational estimates only. Actual AI electricity use varies by model, data centre, cooling system, hardware, location and workload. Do not use these results as official energy, planning, investment or engineering advice.
Training run
Infrastructure requirements
This run draws roughly 0.3 MW of continuous power and consumes about 180,633.6 kWh — equivalent to 66.9× a UK household's annual electricity. Sustained loads at this scale typically require dedicated grid connections, high-density cooling and backup power.
How This Calculator Works
IT energy = training hours × GPUs × per-GPU power. We multiply by a PUE of 1.4 for cooling and overhead. Cost = total kWh × your electricity rate; carbon = kWh × 0.207 kg CO₂/kWh.
What The Result Means
The energy and cost figures size a single training run; peak power indicates the connection and cooling capacity the hardware needs while running.
Limitations
Real runs vary with utilisation, networking, idle time and hardware generation. Use as a planning estimate.
Related Tools
Continue exploring the UK AI Energy Intelligence Centre.
Related Downloads
Free UK AI energy resources from AcrossAI.