Most operators focus on PUE, a measure of how much of a data center’s power consumption goes on computing rather than on ancillary functions such as cooling or lighting. The closer it is to 1, the more efficient the data center.
“While improving PUE helps, more is needed,” according to Kukreja. “Demand for AI tech, which sucks up power with those processors, is growing faster than what local power grids can provide worldwide, not just in Singapore.”
Fixing this requires a comprehensive approach, Kukreja said: “One key strategy is grouping lightly-used servers to save on energy, computer licenses, and machine costs. Additionally, storing data more efficiently with deduplication software significantly cuts down stored data, saving energy.”
