I was watching a TV show recently about the incredible increase in computing power and the concerns that someday computers could be smarter than us and destroy us. What i found interesting was one part that was referring to Moore's law, the statement that computing power doubles approximately every 2 years. But what everyone seems to miss is that by definition Moore's Law must come to an end. Why? Because at some point you cannot make things any smaller.
Being in the chip design business i have seen first hand the progression of Moore's law. But with each generation of process technology the problems get more difficult to overcome and the costs continue to sky rocket. The original law has changed from doubling every 18 months to the current 2 years. There is discussion the law will need to change to a doubling every 5 years. This pushing out of the "schedule" highlights the challenges involved and brings into question when and if it ends. But it needs to be noted the death of Moore's Law is both a technical issue and the financial issue.
On the technical issue. The dimensions of the devices are getting so small that the components being made on the chip must be measured in widths of atoms vs. nanometers. At this size quantum effects are the order of the day vs. conventional EM theory. Does a wire only 50 atoms wide act as a wire? The answer is maybe?
In addition can light be used to image the devices at these dimensions? Again the answer is maybe? Although in the lab we can move individual atoms with very expensive equipment, we cannot and most likely will not be able to manufacture the millions transistors at this scale.
The solution appears to be carbon nanotubes. They have the potential to solve some of the technical issues, but they are not yet mature. And if they do come into use, they do not scale. The size of the carbon nanotubes will not shrink significantly over time. Another problem with carbon nanotubes is revealed by recent studies of ill health effects (
http://www.sciencedaily.com/releases/2007/03/070330185008.htm). This brings into doubt how quickly this technology can safely be deployed. The simple way to understand this is you cannot shrink an atom. So even if you magically got to the point you could make a transistor out of one atom, you could not go further.
Here is a link to the latest available road map. The challenges described to continue Moore's Law are very daunting in deed.
http://www.itrs.net/Links/2007ITRS/ExecSum2007.pdfThe cost issue is the other limitation. Only Intel and a few other companies and groups can afford the latest 45nm fabs. At a 3-5 billion dollar cost they are the only ones that can make it pay. This will also limit the speed of integration as it will take longer with each generation to reach break even on your investment. Companies will be forced to stay at a larger generation longer even if they posses the technical skill to move on. In addition the mask costs continue to sky rocket. They currently are over a million dollars for the latest devices.
The crux of this is that i think there is a natural limit to the scale we can reach. Let's call this Mike's Law. My educated guess is that Moore's Law will fail no later than the year 2018 at the 18nm process node for technical reasons. However it may fail sooner for financial reasons. So what effect would this have on the semiconductor industry and all the other industries that rely on ever increasing advances in computing power.
At the expiration of Moore's Law the semiconductor industry will be mature. Therefore all industries that rely on it will also be mature. You will see the stagnation of the software, consumer electronics, and all Internet related industries. Fundamentally this ever cheaper and more powerful chip technology has driven growth in all the tech and consumer sectors. If the chip companies cannot provide more for less anymore then there can be no new computers, software, phones, and consumer products. Look why buy a new phone or computer if the new ones don't offer anything more than the current ones. These products cannot offer anything more since the underlying chip technology cannot advance after the end of Moore's Law.
I know this is a stretch for most people as we have been living with Moore's Law for so long, but it is just common sense to know it must change. As far as the replacement, there is none. Carbon nanotubes, quantum computers, can only offer a one time step down in size, one last burst of increased integration before there is no where else to go.
Going back to the doomsday scenario of machines out smarting us? Hopefully the Mike's Law limit will be reached before machines become self aware. I can only wonder if this limiting factor is God's way of saving us from ourselves.
--