Present information technologies use semiconductor devices and magnetic/optical discs; however, they are all foreseen to face fundamental limitations within a decade. Therefore, superseding devices are required for the next paradigm of high-performance information technologies. "Single molecule devices" have been expected to be the most probable candidate; however, they have not been made practical since they were first proposed more than a quarter of a century ago. The major obstacles are the extreme difficulty in accessing a single molecule, and the very complicated electron states of a molecule connected to electrodes. With the advancements in scanning tunneling microscope (STM) and simulation technologies, the design and demonstration of single molecule devices are close to reality. This paper first reviews the architectures suitable for single molecule information processing, in which it is claimed that the performances of information processing is higher, if speed and element number product is larger in almost all architectures. Then, prospects for single molecule devices, including switching devices, wires, diodes, nanotubes, optical devices, storage devices and sensing devices for future information technologies and other advanced applications are described. Four milestones for realizing the peta/exa-floating operations per second (FLOPS) personal molecular supercomputer are proposed. Current status and necessary technologies of the first milestone are described, and necessary technologies for the next three milestones are also discussed.