AI Energy Use
Pemakaian Energi AI
The electricity and hardware behind the models, and what it means for who gets to use them.
Training and running large models consumes real energy, and the numbers are large enough to matter: a frontier training run is measured in gigawatt-hours, and serving inference at scale is a continuous draw on power and water for cooling.
Quantization, distillation, and caching all reduce the cost of the same output, and smaller models displace larger ones for most routine work. Those are engineering decisions with an environmental consequence, and they mostly point the same way.
The wider question is about access rather than carbon. Training frontier models needs capital and power that only a handful of organisations have, which makes the tooling, the models, and the pricing decisions of that small group part of everyone's working conditions.
Related terms
Quantization
Storing weights in fewer bits so a model runs with less memory and cost.
Distillation
Training a small model to imitate a large one, keeping most of the capability at a fraction of the cost.
Model
The trained file of learned parameters that turns input into output.
Local LLM
Running the model on your own machine or server, with no provider in the loop.
Cost per Token
What you pay for each unit of text the model reads or writes.
Want this working inside your team?
It starts with one conversation about your workflow.