Artificial Intelligence for Energy Optimization in Data Centers

Data centers are simultaneously optimized and burdened by artificial intelligence. Current literature separates control studies, which view workload as an exogenous process, from sustainability studies that treat infrastructure as a fixed multiplier. A review of 194 papers, with 63 coded, reveals that 18 of 28 control-oriented studies rely solely on simulation. Only five reached production facilities or physical hardware. Notably, none of these studies account for embodied carbon or water withdrawal during the energy optimization process.