Data mining has become a popular buzz word but, in fact, promises to revolutionize commercial and scientific exploration. Databases range from millions to trillions of bytes of data. There are interesting implications because conventional data analysis and visualization tools for databases of these scales fail. Moreover, the present and near-term computer networks are not capable of transferring such massive data sets in reasonable times. Thus there are requirements for a whole new style of analysis and visualization tools as well as new methodologies for distributing data. This talk discusses what is meant by data mining, explores the issues of computational complexity for data mining algorithms, especially the limitations of conventional methods, and proposes some novel approaches both algorithmically and architecturally.