This paper proposes a research on real-time analysis and visualization for big data of energy consumption. In this research, we access real-time energy consumption data from cloud storage by a Transmission Control Protocol/Internet Protocol (TCP/IP). In order to optimize K-Means clustering algorithm, we implement CUDA C programming to finish data-intensive calculation in the Graphic Processing Unit (GPU), which enhances the efficiency of analysis for big data of energy consumption. Meanwhile, to realize data visualization, we draw the data mining results in a multidimensional plane utilizing DirectX, which is a standard graphics API. We also render the original energy consumption data directly in the form of four-dimensional geometry with the plane together, so as to obtain more useful information intuitively.