Researchers at the U of M aim to ease data center energy load

Researchers at the University of Minnesota, led by Professor Jean Ping Wang, are working on a new technology to mitigate the high energy consumption of AI data centers. Their innovation, called computational random access memory (CRAAM), aims to eliminate the constant back-and-forth movement of data between memory and processors. This new approach would allow computing to happen directly within memory, drastically reducing the energy load. Professor Wang's research, which began 24 years ago, involves building chips using silicone wafers to transform memory into processors. His startup, Besimax, is dedicated to advancing this technology with the goal of bringing it to major data center players like Amazon and Microsoft within 5 to 8 years, with sufficient funding. The International Energy Agency notes that a single ChatGPT search uses ten times more electricity than a Google search, highlighting the critical need for such advancements. Besimax estimates that CRAAM could reduce energy consumption by nearly 99%. The team is building layers of materials onto wafers to create these new chips, which represent a significant breakthrough in addressing the resource demands of the rapidly growing AI industry and data centers.