The Integrated Circuit was invented in 1963. The principle had been conceived earlier but the technology was not available. The planar junction transistor was not heavily invested in because IBM had opted to use surface barrier transistors in its mainframes. The decision to adopt the diffusion type motivated the development of the planar junction, which in turn enabled transistors to be placed next to each other on the same semiconductor wafer. A planar transistor enables a digital circuit to be laid out on a plane and therefore projected directly onto a flat surface so that photolithography can occur and enabling integrated circuits to be mass-produced. This was first done in 1963.
The earliest integrated circuits carried single digit numbers of transistors. This doubled every two years or so for several years but hit a limit imposed by Vannevar's Law after an unspecified period. This is generally true of digital technology with the exception of the Compact Disc. The general pattern of progress is that improvements occur exponentially until a particular point in time before becoming linear.
Let x=(date when progress ceases to be exponential)-(date of invention). Then at an interval of 2x after invention, a device of the same size, price and so forth will be twice as good as it was at an interval of x afterwards, but at 3x it will only be three times as good, whereas given exponential progress it would then be four times as good. The CD is not so much an exception as a radically new technology in digital terms.
Examples of such change are seen in RAM chips, which are static. SRAM chips use flip-flops which use more than six transistors to store a bit. These are placed in multiples of three, i.e. half a byte. By 1975, IBM was able to use forty-eight RAM chips to store 8 Kw of data. In 1979, Sinclair was able to double the capacity using the same number of chips, and by 1985 it had doubled again. After that time, however, the increase in capacity became linear. Similarly, in 1974, IBM needed an entire board of integrated circuits to implement its six-bit CPU, in 1984 it was able to do the same with half the number and in 1990, could manage the same processing power on a single chip, but was not able to improve beyond this point in the same way. After that time, it took a bit-slice approach and was able to use two chips to double word-length but otherwise improvements in processing power and speed were only linear in nature from that point onwards.