DARPA Funds AI Energy Tracking Project

Researchers are developing a system to help developers measure and reduce electricity waste during AI model training.

Updated on Oct. 31, 2026 in Artificial Intelligence

Isometric editorial illustration featuring a modular industrial power transformer component in a clean, abstract setting. — DARPA has awarded $1.83 million to a research team to develop E-CALC, a system designed to measure and reduce electricity consumption during AI training. AI Illustration. Upload story photo >

Live Poll

Do you believe the rapid energy consumption of artificial intelligence is worth its technological benefits?

DARPA has awarded $1.83 million to a research team to create an AI energy tracking system called E-CALC. The project aims to improve software design and reduce electricity waste caused by hardware-software misalignments.

Why it matters

The project addresses critical challenges in the AI sector, where extreme opacity in energy usage strains power infrastructure and increases operational costs. Managing this electricity consumption is essential as data center energy use is expected to surge by 2030.

The E-CALC project targets electricity waste in AI model training, which currently accounts for up to 30% of energy usage due to hardware-software misalignments. The team intends to build tracking capabilities for data processing, model training, and deployment.

The players

DARPA

The Defense Advanced Research Projects Agency is a research and development agency of the United States Department of Defense responsible for developing emerging technologies.

University of Michigan

A public research university in Ann Arbor that is leading the multi-institution effort to develop the E-CALC system.

BAE Systems

A British multinational aerospace, defense, and security company that is participating as a partner in the research project.

University of Illinois Urbana-Champaign

A public land-grant research university in Illinois that is serving as a research partner for the E-CALC initiative.

The details

The research team, led by the University of Michigan with partners from the University of Illinois Urbana-Champaign and BAE Systems, will work to provide developers with granular insights into power consumption. By optimizing software design, the project seeks to curb the significant energy demand currently straining infrastructure.

The project has a two-year funding duration.
Data center energy consumption is projected to double by 2030 compared to 2024 levels.

The Tech Race

This effort serves as a direct intervention into the accelerating AI power crisis, attempting to reverse the trend of extreme opacity in resource usage. It follows the establishment of the DARPA-funded E-CALC energy tracking project to modernize energy accounting in high-compute sectors.

For users and businesses, this technology could eventually lead to more stable power pricing and reduced environmental impacts associated with large-scale AI deployment. Improved energy efficiency in software development helps ensure that AI services remain sustainable and accessible in the long term.

The takeaway

Transparency in energy consumption is becoming a mandatory requirement for the future scalability of the AI industry. Developers should prepare for increased scrutiny regarding the energy efficiency of their models as infrastructure demands continue to climb.