A new kind of load
Evernex observes that data center energy use is rising with the development of generative AI, and has published a separate article on how AI affects data center sustainability. Training large models and answering millions of prompts requires dense clusters of specialized chips running at high power.
DoublingThe International Energy Agency has projected that global data center electricity demand could roughly double by 2030, with AI a major driver. Check the IEA for its latest figures.
Three footprints in one
- Electricity. Training a model is a one-off burst of demand. Running it for every user, every day, is a constant one. Where the grid relies on fossil fuels, that means carbon emissions.
- Water and cooling. Dense AI racks produce a lot of heat. Evernex notes that data centers consume significant amounts of water as well as energy.
- Hardware. AI accelerators are manufactured from mined materials and are often replaced quickly. That creates embodied carbon and future e-waste.
How to balance progress with responsibility
- Use the right-sized model. A small model is often enough, and it needs far less power.
- Optimize software. Efficient code and inference reduce compute per answer.
- Run work when the grid is cleaner. Carbon-aware scheduling suits training and other flexible jobs.
- Choose efficient facilities. Better cooling, Energy Star equipment and renewable electricity cut impact.
- Extend hardware life. Maintenance, refurbishment and certified recycling slow the flow of e-waste.
- Measure and report. What is not measured is rarely reduced.
The honest picture
AI can also help, for example by optimizing energy use and software, as Evernex notes. Whether it is a net gain depends on how much it is used, how efficiently, and on what power.





