Texas Engineers Collaborate with TSMC on Energy-Efficient Memory Tech for AI
Engineers from Texas are exploring SOT-MRAM technology to mitigate the rising energy demands of AI systems. This collaboration with TSMC aims to enhance memory efficiency and reduce reliance on data centers.
Theia Market Signal Identification - AI Assisted
Researchers from Texas have collaborated with Taiwan Semiconductor Manufacturing Company (TSMC) to develop SOT-MRAM, a memory technology designed to meet the increasing energy needs of artificial intelligence. SOT-MRAM boasts a write speed of 2 nanoseconds and consumes only 2 picojoules per write operation, significantly outperforming existing memory technologies that can take several milliseconds and consume hundreds of picojoules.
As Texas anticipates a 5x increase in energy usage from data centers, improving memory capabilities like SOT-MRAM could address both efficiency and reliance on data centers. Continued refinement of this technology aims to enhance speed and reduce variability, potentially allowing SOT-MRAM to replace CPU-based accelerators in various AI applications.
Comments