Original title: Artificial Intelligence for Energy Optimization in Data Centers
Mohammed Basharath Ullah, Summaiya Unnisa Begum, Mohammed Nadeem Ullah · September 2026
Data centers run on huge amounts of energy. Artificial intelligence now helps manage this power use, but running AI also adds extra computing work. Right now, researchers treat these two sides as totally separate issues. The authors reviewed existing studies to see how well AI actually saves energy in data centers. They found serious problems in the research. Most control studies test their ideas only in computer simulations instead of real physical facilities. Not a single study counted water withdrawal or embodied carbonThe total greenhouse gas emissions from manufacturing and disposing of data center equipment., which is the pollution caused by manufacturing the equipment. In addition, the reported energy savings across different technique families overlap so much that nobody can tell which method works best. The researchers conclude that the field cannot currently rank its own tools.
Imagine shopping for a washing machine where all ads claim 20% less electricity, but none mention how much water is used or the pollution from factory production. You cannot choose the most efficient model because critical factors are missing. In this paper, researchers found that studies on AI for data centers have the same issue: they ignore water and carbon costs, and the energy savings are too similar to rank methods. They propose a system to report all these details, like a full comparison chart for appliances.
Data centers consume significant electricity and water, so identifying effective AI optimizations saves costs and reduces environmental impact. Ignoring factors like carbon and water leads to incomplete decisions that affect sustainability and resources.
The work is an architectural and methodological proposal, not a trained AI system. It relies on existing literature, so it may not capture the latest methods, and it does not provide new experimental data or validation on real hardware.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
Artificial Intelligence for Energy Optimization in Data Centers — Mohammed Basharath Ullah, Summaiya Unnisa Begum, Mohammed Nadeem Ullah
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.