The end of Moore’s Law – What will drive progress now?
It is considered a fundamental law and landmark of the digital revolution: Moore’s Law. In 1965, a few years after the invention of the integrated electronic circuit, the semiconductor pioneer and later co-founder of Intel, Gordon Moore, claimed that the number of components on an integrated circuit per area unit will double each year. He later modified the doubling time period to two years, and finally his Intel colleague David House formulated the law in its present popular form: the processing power of computer chips doubles every 18 months. It has accurately described the dynamics of the evolution in computer chip design over the last 50 years and has thus become the backbone of the digital progress in the past 50 years.
But would the world look much different had the doubling occurred in slower steps, for example, only every five or ten years? Whoever asks this question, does not understand the mathematical dynamics of exponential growth. In the case that the computing power of the chip had doubled only every five years since 1965, today’s chips would only be the 15 millionth part as fast of what they are today, comparable to computer chips in 1981. The smart phone (introduced in 2007) would have made its appearance in the year 2104. Even more dramatic would be the consequences had the doubling time been 10 years: computationally we would be in the year 1974, and the smart phone would not have appeared before the year 2243! Moore himself was well aware of the scope of ‘his’ law what made him predict such wondrous things as home computers, digital watches, self-steering vehicles and “personal portable communications devices” (i.e. mobile and smart phones).
Now,of course Moore’s law is not a natural law, if only because eternal exponential growth breaks any of such. Rather, it can be understood as an economic law. Once formulated and popularized it constituted the framework of the developmental plans in the semiconductor industry, thanks to whom the growth of computing power has actually proceeded exponentially so far. Since the 1990s the semiconductor industry has explicitly laid out a biennial road maps to coordinate the work of hundreds of manufacturers and suppliers in order for Moore’s Law to remain valid – making it more of a „self-fulfilling prophecy“ than a „law“. Here the manufacturers benefited from the fact that until today computer chips have been so versatile that it took only a few different types of them – mainly processing and memory devices which could be sold in large quantities. This enabled the companies to deploy the necessary amount of capital for the development of a new chip generation (the market leader Intel alone invests around ten billion US dollars per year in research and development, a significant proportion of which isused to meet Moore’s Law).
But first warnings that the components on the chips were to become too small go back to 1989. The first problem thereby is the heat inevitablygenerated when more and more circuits are integrated into smaller and smaller spaces. For electronsmoving ever faster through ever smaller circuits the chips heat up more and more. More fundamentally, however, is the second problem: The electronic structures currently entail sizes in the order of magnitude of about 10 nanometers, and in about five years, these should be down to only two or three nanometers. This corresponds to the size of about ten atoms. At that point, the laws of quantum mechanics reign fully on what is happening, and the uncertainty principle ruling in the quantum world will make the behavior of electrons and thus the entire transistors hopelessly unreliable (which is primarily due to the “quantum mechanical tunneling”occurring inevitably on that scale). Thus Moore himself in 2007predicted the end of his law giving it at that time another 10 to 15 years, until a fundamental physical limit would be reached. Meanwhile we are almost there. Thus in March 2016 the semiconductor industry presented a roadmap for further chip development for the coming years which for the first time is no longer based on Moore’s Law. And in its last annual report Intel wrote that going forward it could no longer comply with Moore’s Law.
So one of the fundamental drivers of technological progress in the last 50 years willlikelysoon be exhausted, and one of the mostground breaking developments of our age will come to an end. Is this the beginning of the end of electronic and digital progress per se? Not at all, the chip makerssay. What will come to an end is the uniform and joint effort of an entire industry to comply with Moore’s Law. Going forward the chipmakers will foster a much more differentiateddevelopment. It will no longer be about making all chips better and better and then use the muniformly for all of the many diverse applications. The producers will rather start from the applications themselves, which mean while have become as multi-varied as smartphones, video game graphics, supercomputers, and cloud data centers, and then determine which chips are most appropriate for which application. With the development of mobile computers,which come with very diverse needs, such distinctions in chip design have already become an economic necessity. The new mobile devices require many different processing units, each of which however sells in rather low quantities. And this limits the available capital for their development and production. Daniel Reed, Vice President for Research of the University of Iowa, says: „My bet is that we run out of money before we run out of physics.“ He compares the future development in chip production with what happened to the aircraft industry. A Boeing 787 is not faster than a 707 from the 1950s – but they are completely different planes. The 787 comes with innovations such as fully electronic controls and carbon fibers fuselage. This is what will happen with computers:„Innovation will absolutely continue — but it will be more nuanced and complicated.“
But there is also the prospect for fundamentally new approaches, many of which researchers already pursue today, and which will continue driving technological progress-for computers and may even enable Moore’s Law to further apply. These include alternative materials for electronic circuits such as graphene or carbon nanotubes, spintronic elements (which instead of conventional electronic circuits using the flow of the electrons employ the orientation of the electrons’ spin to process and memorize information), neuromorphic systems (whose elements are based on the neural structure of the brain), the integration of memory and processing functionality into the same unit (so called ‘mem-computers’), a three-dimensional chip architecture (in which the elements are stacked on top of each other in multiple thin silicon layers rather being put on a single two dimensional surface on a wafer, which, however, only works for memory chips, where there is no heat problem, as those consume energy only upon memory being accessed), and most recently the possible development of a quantum computer, which could revolutionize the entire digital world.
Although these alternatives have not yet managed to take the step out of the laboratory, the chip and computer industry will hardly run out of ideas to further develop digital capacities. If we thus interpreted Moore’s Law such that simply the benefits of electronic building blocks for end users double every 18 months, we are safe to assume that the law is far from losing its validity, as human creativity barely ever runs out of steam. The next digital revolution could already be on its way soon after the first one has been completed with Moore’s Law in its current form coming to an end. In any case, we will enter a new and certainly exciting age of technological progress.