Fujitsu Laboratories Ltd. had recently made an announcement regarding the development of a circuit technology. It would improve the energy efficiency of deep learning hardware by reducing the data bit width used in the process. Also, it won’t require changing network structures or training algorithms. In deep learning, massive calculations are involved based on training data. However, the upper limit to processing performance is governed by the electricity volume used by servers and other hardware that carry out the processing. Hence, increasing performance per watt has become a crucial issue. So, the latest circuit technology from Fujitsu aims to address this problem.