If only there had been open access, maybe it wouldn't be called Moore's law:
I didn't go to Midland after all, but went instead to the Applied Physics Laboratory at Johns Hopkins University, which has roughly the same relationship to Johns Hopkins that JPL has to Caltech, and where I could continue to do basic research in areas related to what I had done before. But I found myself calculating the cost per word in the articles we published and wondering if the taxpayers were really getting their money's worth at $5 per word. Just as I was starting to worry about the taxpayers, the group I was working in was, for various reasons, breaking apart. So I decided to look for something that had a bit more of a practical bent, and at the same time see if I could get myself back to California.
From Gordon Moore's The Accidental Entrepreneur.
Interesting, but I don't think his comparison of APL to JPL is entirely accurate (which is odd because he had first-hand experience). As I understand it, NASA contracts out the operation of JPL to Caltech, i.e. JPL is a NASA lab. APL, on the other hand, is an actual division of Johns Hopkins on par with the Colleges and Schools.
You did read the original "Moore's Law" paper, of course, right? It has not that much to do with the limits of "what is possible" in given technology, but a lot to do with basic economics, so I am not at all surprised that he would be calculating "cost per word"!*
The famous straight line in log space chart was, actually, Fig. 2. Fig. 1 had a number of skewed "parabolas" showing cost vs. the number of components on a single packaged die in any given year, actual Moore's law is plotting the minimum of those parabolas vs. time. You pack in too little, and packaging cost dominates your chip price; you pack in too much, and limited yield kicks in and wafer cost eats away your margin...
* My good buddy, to protect the innocent I'd call him Dr. Q. :), once estimated that it costs US taxpayer between $1.25 and $2.5 megabucks to educate/train a superconductor electronics designer, and I do agree with this estimate...