A hardware and software environment with performance above 1 Teraflops (teracomputing) is presently required to face the leading computational challenges not only in fundamental sciences, but also in an increasing number of fields related to applied sciences and engineering. A short survey of the top level computing platforms is made, with a special attention to United States and Japan. The technology options adopted for teracomputing are evaluated. In this perspective, a European line of supercomputers is described, which was originated by the Italian Institute for Nuclear Physics and is now in use also in Germany and France: this line of massively parallel platforms is based on the use of CMOS VLSI custom chips, carefully designed to match the computing needs, so that only the silicon strictly necessary to the operations to be performed is used.