In this study, we propose a cloud computing service for fast audio source signal separation. To implement this service, we have developed a GPU-based ICA (Independent Component Analysis) program by using CUDA software development toolkit (SDK). This program can rapidly separate n independent components from m sources which are composed of these components mixed by a random way. On the other hand, we also have implemented a web server to provide users with the service of fast audio signal separation. Users can upload their audio files onto this web server through the browser interface. After users click the execution command button in the web page, the web server will invoke the GPU-based ICA program to fast separate the independent audio signals from the uploaded files. Finally, users can download and play the wave files of the separated audio signals from the web server by clicking the hyperlink in the website.