How bright is the future of Moore’s Law?

Every now and then you hear a pessimist predicting demise of Moore’s Law. It is too difficult, if not impossible to continue they say, give several compelling scientific and economic arguments, is the end for real? Hardly…!

When you say electronics, the first thing that comes to mind is Moore’s Law, not a law of nature, but a discipline that the industry followed for the last four decades, delivering faster and affordable electronics that we enjoy today and take for granted.

Moore’s Law doubles the number of transistors every two years, improves performance, and lowers energy consumption, (all for the same dollar cost)—too good to be true, but it is really true, short of a miracle, and explained by the laws of Physics. A main-frame computer in the 60’s and 70’s occupied a floor, consumed kilo-watts of power, cost millions of dollars, but now you have that compute power in your laptop running on batteries, such is the power of Moore’s Law. And now you can see why I am so obsessed by it.

It is true that following Moore’s law is getting difficult, but when has it not? When I was a graduate student, my professor predicted that Moore’s Law would end in the 90’s when it reaches sub-micron dimensions. Did it? In the late 90’s they predicted demise of Moore’s Law based on power and leakages, yet the law survived, and now they are making general claims that it will end sometime soon. Gordon Moore gave a keynote at the International Solid State Circuits Conference (ISSCC) in 2003 where he stated that no exponential is forever, but we can delay “forever”, and that is the strategy I urge you to follow. Let’s take a look at why it seems so difficult.

First, the laws of physics which helped in the past are now making it difficult to continue. Today’s transistors are so small that the geometries are approaching atomic limits. For example, the wavelength of light being used to pattern these geometries is much longer than the pattern itself, sort of like trying to shoot a duck with huge cannon! Also at such atomic dimensions, quantum mechanical effects start surfacing and making things difficult overall.

Second, economics of Moore’s law is getting harder. The cost of a fabrication facility (Fab) is now in billions of dollars. Compare this to the 70’s when you could afford to build a fab in your garage for a few million dollars, and almost everyone did, it was a fashion then! A mask set (for photolithography) alone now costs more than a million. If you make a mistake in your design, then correcting the mistake would cost you millions of dollars in a new mask set, not to mention months of delay in design and fabrication. Design technology too has advanced over time, but not at the same rate as Moore’s Law, demanding increasing design resources.

Third, establishing the value proposition to follow Moore’s Law becomes increasingly difficult. If you double the number of transistors, then what will you do with it, and how does it add value? Difficult to justify following Moore’s Law in the face of growing economic concerns. When the 386 was introduced in the 80’s, the pessimists claimed that it delivered plenty of compute power, and would not know what to do with a 486. Of course, there was no internet, no world-wide-web, or even windows then, and look what followed! You cannot underestimate creativity, and that is why, I believe that you need both, steady evolution of hardware (compute power), and evolution of value proposition for it, with new users and usages, both evolving hand-in-hand. Expecting to identify a killer-application upfront to establish the value proposition is futile.

Therefore, those who know how to follow Moore’s Law by bending the laws of Physics, adding value, and making it economical, will succeed. And those who don’t will start falling off pointing fingers at those who do. I am a fighter and am not playing dead. In my subsequent posts, I will highlight how do deal with the three issues one by one, and prove it to you that the future of Moore’s Law is indeed bright!

Comments are closed.