Informatique : l'après-Moore ? Le progrès des semi-conducteurs en question : le monopole Intel face à des stratégies alternatives

The phenomenon labelled “Moore’s law” is well known: over the last 30 years the density of microprocessors per integrated circuit, i.e. the number of transistors integrated on the same silicon chip, has doubled every 18 months.
Readers of Futuribles will perhaps recall, however, the article we published in September 2002 (n° 278), in which Jean-Paul Colin already argued that this pace of progress could not be maintained indefinitely. He returns to the topic here, stressing that this amazing miniaturization of transistors gave rise to many desirable but also undesirable effects.
According to Colin, other factors influence the overall performance of an integrated circuit and of the system which uses it, such as its architecture and the power lost by a microprocessor in the form of heat. First, this power loss is doubling every 36 months: such an increase will eventually become a major problem. Secondly, improvements in the architecture of microprocessors -a factor whose importance is rapidly increasing as systems become more complex- is not keeping pace with this growing complexity.
Colin insists that these two factors are likely to be decisive in future in determining whether or not Moore’s law continues to operate. It is virtually certain, moreover, that Intel, having failed either to make the investment or to acquire the necessary expertise in this regard, will see its position challenged by new competitors (including IBM, American start-ups and Asian firms) which will perhaps play a much more important role hereinafter.