DARPA backs project to measure AI's energy use at every workflow stage

· Oil & Gas · United States

DARPA is backing a project to measure the energy consumed by artificial intelligence at every stage of the workflow. The effort addresses a problem the AI boom has created: machine learning is being woven into everything from energy grids to consumer devices, driving a global race to build data center capacity and pushing energy demand projections higher. Current measurement practices, as described in the item, are not granular enough to attribute consumption accurately across the AI pipeline. Better measurement would give grid operators, data center developers and energy suppliers a clearer basis for planning capacity and for judging how much new generation AI workloads actually require.

Why it matters：Data centers have become a major source of new electricity demand, so a credible way to attribute AI's energy use could reshape how operators size generation and grid investments.

Key points

  • The project covers every stage of the AI workflow, not just total consumption.
  • The AI buildout has driven projections for energy demand sharply higher.
  • The item names no specific budget or participating organizations.

Sources (original reporting)

More stories from this day