More than Moore: what next, after silicon?

Computer chips processing power

Since the 1960s, the microprocessor industry has kept fairly on track with “Moore’s Law,” which states that the transistor count on a microprocessor doubles every 2 years. Gordon Moore’s model has been quite useful for anticipating computing power over time.

Very few people realize that, within the broad microprocessor industry, Gemalto personalizes and ships one of the largest numbers of computer chips produced by any business each year. We have three billion secure, individually customized devices in various form-factors, or embedded in other systems. The computing power of these “secure elements” is dedicated to digital security, and is often not very well known by the public. And of course, it also grows at the speed of Moore’s Law.

100% of these computers’ brains are made with Silicon, a material that can be a conductor, or an insulator (a semi-conductor) within a mature technology called CMOS (“Complimentary Metal-Oxide Semi-conductor”). Grids’ “channel” sizes are as small as 14 nano-meters today (around 80,000 times smaller than the thickness of a sheet of paper, just 140 silicon atoms).  In the years to come, CMOS production processes will continue to be refined, shrinking to 9nm at first, and probably all the way down to 5nm by 2020. Each time CMOS production processes shrinks its gate size, CMOS transistors require less current to switch and they get physically smaller, so you can put more of them in the same small amount of silicon. In short, you get more speed, less power consumption, less heat to dissipate…exactly what mobile and IoT technologies will always need.

However, physics being physics, silicon will reach its limits very soon.  For a transistor to be a transistor, you need to have some form of a gate that, when switched to a logical state of “1” , connects as source to a drain. When CMOS technology reaches the 5nm production process, i.e. 50 Angstroms/50 silicon atoms wide, the channel side barely contains a handful of free electrons to open or close the channel. The bottom line is that the industry will need a new material or a new technique to continue to deliver on the Moore’s law, and in order to deliver microprocessor chips containing tens of billions of transistors.

A transistor is like a water tap:  It’s on or off depending on whether you twist the tap open.   Silicon CMOS has performed in this way for 5 decades, but it required several tenths of a volt to activate the tap, and its size won’t be further shrinkable very soon.  The next transistor is already being nurtured at IBM labs.  It’s organic, needs only millivolts to switch and, unlike semiconductors, it does not have the constraints of a depleted zone below the grid to act as a transistor.  By departing from silicon and finding new materials for integrated circuits, there is a good chance the new “Moore’s Law” will deliver far more than twice the number of transistors every two years.

For the security industry, this means the computing power available for off-line decisions, within devices in the hand of consumers, will grow effectively infinitely. We’ll then enter an era in which an AI could live in your pocket, so you can image how powerful, secure and limitless the IoT could be.

What are your expectations for the future of Moore’s Law? Let us know by tweeting at @Gemalto or by posting a comment below.

Leave a Reply

Your email address will not be published. Required fields are marked *