Posting in Design
Moore's Second Law. Just as chip density rises exponentially, so does the initial cost of making the chip. It's this law, and its implications, that Intel now has to face in dealing with the 1 billion Euro fine the EU has levied for what it calls willful violations of its competition rules.
(That's Gordon Moore, the former Intel CEO for whom the law is named, at left, from Wikipedia. He turned 80 in January. Belated best wishes.)
Chip circuit lines were growing so close together it seemed impossible to see anything like a constant doubling of circuits and, thus, a doubling of chip speed every 2 years.
Engineers found ways past it. A hundred million miracles are happening every day.
But there is a corollary to Moore's Law, something I have long called Moore's Second Law. Just as chip density rises exponentially, so does the initial cost of making the chip.
Throughout their history Intel and its competitors have thrown sharp elbows and competed fiercely. When there are several market competitors this is called good business. When there are just two, and the big guy has the little guy on the ground, it's called an antitrust violation.
I am generally in favor of antitrust law. I find monopoly profits to be a tax the economy should not have to pay.
But in this case we are talking about an inevitable result of Moore's Second Law. Had things worked out differently perhaps AMD would be facing this crisis now. Or Texas Instruments. Or Motorola. Or IBM.
As costs rise it becomes impossible, at some point, to support two market competitors. Throughout this decade we have seen the rise of "fabless" chip companies, like nVidia, and the creation of general purpose chip foundries -- that's the model AMD has chosen for itself.
Intel does not have to do this for financial reasons but it's time to consider doing this. Because as certain as Moore's Second Law has broken its competitors, it's coming after Intel next. By carefully planning the move now Intel guarantees it won't be caught flat-footed when it becomes necessary.
By splitting manufacturing from design Intel's current problems go away. We can have multiple design companies and, if necessary, move toward sharing foundry costs throughout the industry.
Moore's Second Law will still be lurking, of course. Eventually you run into innovations that cost too much to make. But that assumes innovation continues along its current path. The company that moved five years ago to dual core and low power designs knows there are many others.
May 12, 2009