Researchers Found an Innovative Way to Cut Data Center Energy Use

Researchers at the University of Illinois Urbana-Champaign developed copper cold plates using electrochemical additive manufacturing (ECAM) to improve data center cooling efficiency by up to 32%, reducing pressure drop by 68%. The optimized fin design, created through topology optimization, could drastically cut energy consumption in data centers, with a 1-gigawatt facility potentially lowering cooling power use from 500 megawatts to just 11 megawatts.
Researchers have discovered a way to significantly reduce energy consumption in data centers by enhancing direct-to-chip cooling technology. A team led by Nenad Miljkovic, professor at the University of Illinois Urbana-Champaign, created copper cold plates using a proprietary electrochemical additive manufacturing (ECAM) process developed by Fabric8Labs. These plates achieve up to 32% better cooling efficiency than conventional designs while reducing pressure drop by 68%, improving coolant flow. The innovation lies in the cold plates’ fin design, optimized through topology optimization—a mathematical algorithm that refines shapes for maximum efficiency. After 1,000 iterations, the team produced a tree-like structure ideal for heat transfer. Miljkovic noted that this method enables intricate, three-dimensional structures impossible with traditional manufacturing. Testing showed that data centers with 1 gigawatt of computing power currently use 500 megawatts for air-cooling, totaling 1.55 gigawatts of energy consumption. With the optimized cold plates, cooling energy use could drop to just 11 megawatts, drastically improving efficiency. The next phase involves testing the cold plates on real servers to validate performance in practical applications. Miljkovic expressed optimism about scaling the technology, which could alleviate strain on power grids as data centers expand. The research was published in *Cell Reports Physical Science*, highlighting a potential breakthrough for sustainable data center operations.
This content was automatically generated and/or translated by AI. It may contain inaccuracies. Please refer to the original sources for verification.