The Virtual Observatory (VO) is driven partly by new science goals, partly by external technological changes, and partly by the flood of data coming our way. More astronomy is done on line through organised ‘science ready’ archives, and with processed data (source catalogues etc) from large survey projects. Users assume on-line availability of anything useful. The normal methods to date are to download files and analyse them at home, but very likely in the future data will be analysed in situ and analysis software will be as standardised as data reduction software is now. Many science goals require combining data from multiple archives (e.g. crossmatching different wavelengths), and often demand the processing of huge amounts of data, e.g. rare object searches, computing correlation functions, etc. All users want to be ‘power users’.