Upgrading AI by Making It More Efficient

AI currently consumes as much power as a small nation, with a significant portion dedicated to powering data centers that train and operate AI models. As AI expands, improving energy efficiency presents a critical challenge that SK companies, including SK Hynix, are committed to addressing.

SK Hynix is advancing more efficient AI memory solutions, including High-Bandwidth Memory (HBM) technologies as part of this commitment. These are essential components of modern AI systems that manage substantial data flows at exceptional speeds. The company anticipates doubling the energy efficiency of its HBMs by 2026, alongside a parallel initiative to nearly double the efficiency of its SSD data storage platforms.

Absolics, a U.S.-based SK company, also contributes to making AI more efficient through its innovative glass substrate technology. When implemented at scale across data centers, this technology could reduce energy consumption by approximately 50 percent.

Industry analysts forecast that power demand from data centers may more than double within the next decade. Further compounding this challenge, AI computational power is estimated to double every 100 days, which could raise global energy demand attributable to AI by up to 36% annually in the coming years.

These developments underscore the necessity of sustainable approaches to AI infrastructure as computational demands continue to accelerate.

Watch the video to learn what SK companies are doing to develop a more energy-efficient AI.