AI’s Energy Demand Surges: Efficiency Solutions and Challenges Ahead

AI's energy demands strain global power grids. Researchers aim to boost AI efficiency, cutting costs and energy use.
DARPA Backs Project to Measure AI Energy Use at Every Stage of the Workflow

The rapid expansion of artificial intelligence is sparking a significant increase in global energy demand. From integrating AI into power grids to even electric toothbrushes, the push to expand data center capacity is leading to soaring projections for energy use. Predictions indicate that energy consumption by data centers will more than double by 2030, from the 2024 level of approximately 420 terawatt-hours (TWh).

Advocates of AI suggest that it will eventually offset its energy consumption by enhancing efficiency across various economic sectors. The Financial Times previously highlighted AI’s potential in solving complex problems such as discovering new materials and processes, a task likened to finding a “needle in a haystack.”

Nonetheless, current trends show a significant rise in water and energy usage within the AI sector, which poses challenges in a world striving for sustainable energy solutions. Despite efforts to boost clean energy production, the demand driven by AI is fueling the construction of natural gas plants and reinforcing the role of fossil fuels in the energy mix. This surge in energy demand, coupled with delayed climate action, makes it increasingly likely that global emissions goals will be exceeded.

Instead of waiting for AI to naturally enhance efficiency, researchers are focusing on making AI inherently more efficient. A team from the University of Michigan Engineering is developing a system to optimize AI software energy usage. According to Michigan Engineering, current misalignments between software and hardware can waste up to 30% of electricity during AI model training.

Researchers from the University of Michigan Engineering, in collaboration with the University of Illinois Urbana-Champaign (UIUC) and BAE Systems, are working on a system to improve energy tracking and efficiency in AI. The project addresses the lack of transparency in the sector’s energy usage. Understanding the exact energy consumption of AI remains a challenge, yet it is acknowledged to be substantial.

This new initiative aims to enhance energy tracking capabilities, enabling developers to estimate energy usage throughout the AI lifecycle. Michigan Engineering notes that this approach will allow energy estimation across various stages, from data processing to deployment, avoiding costly trial-and-error methods.

Such advancements are particularly vital in military contexts, where equipment needs to run complex AI algorithms within strict power limits. The E-CALC project, supported by the Defense Advanced Research Projects Agency (DARPA) with a budget of $1.83 million over two years, focuses on optimizing energy use in AI.

Mosharaf Chowdhury, an associate professor at the University of Michigan and the principal investigator for E-CALC, explains, “AI is becoming part of everyday life, business, science and national security, but it requires enormous and rapidly growing amounts of electricity. If that energy use continues to rise unchecked, it could increase costs, strain power infrastructure, and make advanced AI harder to deploy.”

By Haley Zaremba for Oilprice.com

More Top Reads From Oilprice.com

Original Story at oilprice.com