Any one of several similar folk theorems that fit computing capacity or cost to a 2t exponential curve, with doubling time close to a year. The most common fits component density to such a curve (previous versions of this entry gave that form). Another variant asserts that the dollar cost of constant computing power decreases on the same curve. The original Moore's Law, first uttered in 1965 by semiconductor engineer Gordon Moore (who co-founded Intel four years later), spoke of the number of components on the lowest-cost silicon integrated circuits — but Moore's own formulation varied somewhat over the years, and reconstructing the meaning of the terminology he used in the original turns out to be fraught with difficulties. Further variants were spawned by Intel's PR department and various journalists.
It has been shown that none of the variants of Moore's Law actually fit the data very well (the price curves within DRAM generations perhaps come closest). Nevertheless, Moore's Law is constantly invoked to set up expectations about the next generation of computing technology. See also Parkinson's Law of Data and Gates's Law.