One of the causes of global boiling, which has been inducing abnormal weather around the world, is the enormous power consumption of data centers, especially in arithmetic logic unit (ALU) operations and memory data transfers. We aim to achieve ultralow-power consumption by replacing Si-based cache memories with analog in-memory computing (AiMC) and randomaccess memories using crystal indium oxide, which enables field-effect transistors (FETs) with excellent on-state characteristics, including a mobility of up to 142.7 cm2 /V s, and an off current lower than that of a Si FET by 11 orders of magnitude. We are also studying ultralow-power artificial intelligence (AI) ALUs using AiMC FETs in the subthreshold regime. This report will introduce the trend of these technologies. Further development of devices based on these technologies will achieve high integration and power saving, which is expected to be important countermeasures against global boiling.
Loading....