Although Fairchild gave the first ISSCC paper on the feasibility of CMOS in 1963, and though RCA made the first working CMOS devices in 1964, it was over a decade later before anyone thought of the technology as a go-er.
"Nobody really recognized CMOS as a viable future product until the late 70s, early 80s", said the late Bernie Vonderschmitt, founder and former CEO of Xilinx who took over as general manager of RCA's solid state division in 1972, "and that's an astounding statement to make given the fact that Silicon Valley is supposed to be where the action is."
RCA launched its CMOS line in 1970 but, two years later, when it pulled out of the computer business with an unprecedented half a billion dollar write-off, that was the end of its involvement.
"Consumer electronics saw no need for digital electronics at all", said Vonderschmitt, "and therefore really had no interest in CMOS."
Consumer electronics is very conservative - and also, CMOS in the early 80s just wasn't very good. The 4000 series was mostly seen as rather slow, sloppy and static-sensitive TTL; sure, it'd run on batteries while TTL couldn't, but why on earth would you? CMOS turned up in some of those pocket computers from Casio et al, but it was a mildly esoteric technology with no real market. There's not that much you can do with a handful of gates that's very exciting for consumers.
At Sinclair, we had quite a long running project to replace the Ferranti ULA (which was TTL, I think) with a CMOS version from - oh, I forget. Saga? Sage? Not a company that succeeded, anyway. That was necessary for Pandora, the portable version, but I don't think we ever got one that worked properly. Certainly, even in the late 1980s I can remember problems with BiCMOS ASICs, mostly to do with substrate impedance and noise.
And I know you don't much like Intel, but when you see them talking about pure CMOS 60GHZ RF systems it is a very impressive experience - especially if you remember the days of trying to get an 80C85 (which I'm not sure Intel ever made) running at more than 2.5 MHz.
That's really interesting, thanks Rupert.