The cost of building chip fabrication plants will continue to increase (and the return on investment to decrease) until it becomes fiscally untenable to build new plants.
i.e. while it may be technologically possible to continually double the density of chips every 18 months, the cost of achieving this goal will eventually surpass the profit.
-- MichaelLeach
It is said that Intel and AMD financial statements confirm a trend toward increasing cost and deminishing returns. ( detail? )
I thought the law was something more like this:
"The cost of building a modern chip fabrication plant will double every 18 months until it becomes infeasible to build new plants." Sort of. But this is basically just a restatement of Moore's actual quote above, but in laymans terms.
"Return on investment" has nothing to do with it. The problem is that the entire concept of money breaks down in large enough quantities.
Moore's second law predicts that, eventually, a modern chip fabrication plant will cost more money than the net yearly salary of every human on earth. Then a few years later, a modern chip fabrication plant will cost more money than the total gross lifetime income of every human on earth. And since money represents labor this means it would take more than the gross lifetime work done by every human on earth to build.
Exactly. The idea of "return on investment" doesn't work with very large investments. It reminds me of the way "F=ma" doesn't work at high speeds, or the idea of "brightness" doesn't work with very dim lights, or the idea of free-markets doesn't seem to work with monopolies (the reason for anti-trust laws, right?). So it makes me think that there's something very interesting to investigate here.
I think big players are betting on UtilityComputing as the next gen alternative for when the CPU fabrication investment vs. profit return curve maxes out.
Perhaps the reason why the Linux SCO lawsuit was/is so vicuous is that a significant part of the IP allegedly lifted by the Linux community from Unix is related to multi-processor management.
So, if you believe in MooresSecondLaw, then MP architecture becomes an important issue. --MichaelLeach