The rapid expansion of artificial intelligence has created significant strain on global energy infrastructure, with data center energy consumption projected to more than double by 2030 compared to 2024 levels. While proponents argue that AI will eventually improve efficiency across industries, current energy demands are driving increased construction of natural gas power plants and extending fossil fuel reliance in the global energy mix, complicating efforts to meet climate goals.
A major challenge in addressing AI’s energy impact is the lack of transparency regarding resource consumption and insufficient measurement systems. Researchers currently lack precise data on real-time AI energy usage, making it difficult to identify efficiency opportunities or implement targeted improvements.
To address this gap, researchers from the University of Michigan Engineering are collaborating with colleagues at the University of Illinois Urbana-Champaign and BAE Systems to develop E-CALC: Energy Calculus for ML Lifecycle Optimization. The proposed system would enable developers to estimate energy usage at every stage of AI workflows, from initial data processing through model training to final deployment, eliminating the need for extensive trial-and-error testing across numerous hardware and software configurations.
The project has received particular attention for its applications in military and defense contexts, where remote computing systems must operate complex AI algorithms under strict power constraints. The Defense Advanced Research Projects Agency has committed $1.83 million in funding over two years to support the initiative. According to project leadership, reducing AI’s energy requirements could lower costs, reduce infrastructure strain, and facilitate broader deployment of advanced AI systems across civilian, business, scientific, and national security applications.
Article Attribution | Read More at Article Source
Article summary produced by Claude AI