MIT Professor Targets Data-Center Energy Waste with AI-Driven Efficiency Tools

Christina Delimitrou, an associate professor in MIT’s EECS department and member of CSAIL, is investigating how artificial intelligence can curb the rising electricity demand of data centers. Her research indicates that a substantial portion of existing compute capacity remains idle, with systems often operating at roughly 15 % of their potential. By deploying machine-learning algorithms for automated resource management and redesigning server architectures, her team seeks to improve performance while slashing energy use.

One result, the Seer platform, predicts and prevents online-application problems, reducing latency and downtime for services such as video streaming and video-conferencing. Delimitrou, who studied at the National Technical University of Athens, Stanford and Cornell, also stresses the importance of teaching independent thinking to MIT students. She argues that addressing AI’s carbon footprint requires smarter use of current infrastructure rather than building more facilities.