Measures of the performance of digital electronics have increased steadily for four decades. The essential ingredient of progress has been miniaturization. The road to advances beyond a decade into the future has always been obscure and has stimulated much speculation as to where miniaturization must end. Thus far, new ideas have regularly met the challenges posed by new problems and have allowed the trends to continue. The persistence of silicon as the basis of information technology is the one constant. The problems and solutions arising from changing physics caused by miniaturization are examined in the perspective of the environment in which millions of closely packed devices must function and the external world that information processing technology serves. The complex structure of devices offers considerable room for ingenuity and novel approaches. The large number of devices collected on a single chip of silicon must communicate with one another via a complex array of wires. Miniaturization of the wires poses a different set of questions. The simple structure of a wire offers less scope for invention than is possible with devices and it is more difficult to see solutions to problems. Nevertheless, as has happened in the past, it is difficult to find obviously insuperable barriers to progress in fundamental physics.

Uncertainties inherent in physical phenomena and structures obscure the physical representation of information. Time-dependent uncertainties in physical parameters are conventionally regarded as noise. Time-independent uncertainties, fixed in the physical structure of the devices used to represent information, although determinable in principle, are in fact unknown to the system designer and play a role in information theory similar to that of conventional noise. Several examples of the way in which power is used in logic and memory devices to overcome the effects of physical uncertainty are presented.