Nothing gets misused these days more than Moore’s Law.  I worked with Gordon Moore at Intel for a few years, and we used to chat about his “law.”  If you know Gordon, you know he graphs everything… ask him about anything, and Gordon will dig around in his desk and pull out a chart he’s been keeping on the subject, points hand-plotted and trend lines drawn in.  He told me how he had been worried about when Intel would have to build a new, more advanced wafer fab and had been plotting the increase in chip complexity year by year.  That plot led to the famous conclusion that the transistor count on chips doubled every two years (actually he modified this a couple of times over the many years he kept that chart).  The answer that Gordon had been searching for came out clearly — Intel should not simply double chip complexity with each generation, it should quadruple it because otherwise it would not be able to build new fabs fast enough.  (It turns out he kept a second chart on how long it took to build a new fab as well, and how that was changing over time).

The big difference between Intel and other companies in the chip business is that, if someone like Gordon charts trends like this, the company acts on it and builds it into its annual planning process.  So at Intel, people translated Gordon’s conclusion into the implications for photolithography, fab equipment, wafer sizes, new chip designs, and so on.  The entire Intel planning system was built on the idea that chip complexity would double every two years.  And the continual drive to meet that trend on Gordon’s chart carried over to other companies as well.  Moore’s Law became a self-fulfilling prophecy.  It’s the basic reason that memory chips go from 4 to 16 to 64 etc… instead of 4 to 8 to 16 to 32…

Since those early chats with Gordon, I’ve reflected on just why his chart was correct (besides the fact that since everyone assumed it was correct, they all drove to keep up).  My conclusion was that what he was getting at was actually the change in “feature size” in the basic chip geometries, and that it was the feature size that was being cut in half every two years and THAT fact was driving the quadrupling of complexity.  Because feature size reductions help you in two directions  — x and y — you’re seeing an “area” or two dimensional phenomena based on a one-dimensional change.  Cut the feature size in half and you can pack four times as many transistors into the same area.

Another technology that follows this geometric trend is hard-drive capacity.  If you cut the size of a magnetic bit on a drive in half, you can pack twice as many of those bits on the track that circles the disk.  But you can ALSO CUT THE WIDTH OF THE TRACK IN HALF, and therefore double the number of tracks.  Twice as many bits on a track and twice as many tracks quadruples the capacity.  Bingo… another technology following the geometric progression laid down in Moore’s Law.

Let’s think of some others.  Um, those are the only two I know of.  Yet in book after book, presentation after presentation, people call down Moore’s Law to show why their particular business is going to undergo staggering growth.  The one I find really funny is the software business.  Not too long ago, Bill Joy, who certainly should know better, expressed his concern in a widely quoted letter to Wired magazine, that as software capability increased according to Moore’s Law, computers would take over the earth and displace us.  Joy, one of the founders of Sun Computer, should know that software capability inches along at a snail’s pace and if anything, has failed miserably to keep up with even the most conservative predictions of new capabilities.

It turns out that Moore’s Law has become the universal arrow in the quiver of every visionary and futurist.  But the next time you run into it, and believe me, it won’t be long, ask yourself, “what is there about this technology that makes it the EXCEPTION that will actually follow Moore’s Law?”