Imec reports advances in energy-efficient AI memory research
Leuven-based research centre imec has announced two breakthroughs in the development of next-generation memory technology that could help make artificial intelligence systems more energy-efficient and compact. The findings are being presented this week at the IEEE/JSAP Symposium on VLSI Technology and Circuits in Honolulu.
As demand for AI continues to grow, existing memory technologies are increasingly reaching their technical and economic limits. Imec is therefore researching ferroelectric memory, which can store electrical charge without requiring a constant power supply.
The first breakthrough involves ferroelectric capacitors, a key component of memory chips. Researchers succeeded in developing capacitors that operate at lower voltages while maintaining performance and durability. According to imec, the technology could eventually replace conventional DRAM memory and reduce energy consumption.
The second advance concerns memory density. Imec successfully stacked five layers of memory cells vertically — a first for this type of technology — and improved the process used to erase stored data. The approach could pave the way for smaller memory chips with significantly greater storage capacity.
According to imec, both developments could ultimately contribute to more compact and affordable memory systems for AI and other data-intensive applications. However, the organisation stressed that the work remains at the research stage and will require further development before commercial deployment becomes possible.
#FlandersNewsService | © BELGA PHOTO DIRK WAEM
Related news
