Electricity consumption is the electrical energy AI systems use during training and inference. Measuring it is harder than it sounds: providers report inconsistently, studies use mixed units, and some estimates rely on indirect assumptions. section 6.9
Table 6.1 Average electricity consumption across different models compared with typical household consumption
| kWh / 1,000 responses |
% of daily household kWh (US) |
% of daily household kWh (UK) |
|
| Text generation |
0.047 |
0.15% |
0.5% |
| Image generation |
2.907 |
9.83% |
30% |
Compare text and image generation by 1,000 uses: image generation consumes far more electricity than text generation, and intensive use can accumulate quickly. section 6.9
Why usage volume matters
A single AI request may seem minor, but a coder can generate hundreds of autocompletions per hour. Search engines and online stores may also trigger AI queries, adding consumption without users directly opening an AI tool. section 6.9
The sensible takeaway is not to treat one reported figure as universal. Consumption varies by model and task, while comparisons may express the impact as electricity, carbon emissions, or equivalent household use. section 6.9
