Most of my articles are dissertation-length manifestos and many have commented that this fact, in combination with their extremely technical nature, boils down to make them a tough read. My latest effort is an attempt to condense some of the good takeaways into bite-sized nuggets with a sauce that is tasty but not overly technical.
In this article, I will discuss a very valuable silicon and transistor replacement technology that Micron (NASDAQ:MU) owns.
Trees Don't Grow to the Sky
In the mainstream media, there's very little discussion about the end of Moore's Law, which is the doubling of transistors per dollar every couple of years for the last five decades. Cost per transistor stopped decreasing below 28 nanometers and Intel (NASDAQ:INTC) recently announced that their 10 nanometer process would take substantially longer than the previous two-year cadence.
Even if they can find appropriate materials to build transistors at sub-10nm levels, power consumption is expected to be very high. Additionally, chip speeds hit the 4Ghz speed limit well over a decade ago. The smart people in the industry expect that we'll see a formal end in scaling in the next few years. While FinFET allowed the fabs to crank out a couple of extra nodes, it isn't a technology - it is a band-aid on a gunshot wound.
So there's definitely a need for a technology to replace the transistor with something that can continue scaling. There's bonus points if that technology can provide a leap in speed and power efficiency. I believe that Micron possesses such a technology.
The Shining Briefcase
US Patent 6969867
Field effect chalcogenide devices
This device looks a lot like a transistor because its operation isn't much different. At rest, terminals 110 and 120 are conductively isolated from each other (electricity cannot flow between them because material 100 is non-conductive). However, if you apply an electrical pulse to control terminal 140 (which is insulated from the other two terminals by insulating layer 130), material 100 becomes conductive and electricity can now flow between terminals 110 and 120.
This is identical to a transistor's function with one difference: as long as a holding voltage (with a sufficient current) is maintained on one of the terminals 110 or 120, then material 100 will remain in its conductive state without any voltage on the control terminal 140. This allows the cell to function as a memory latch in addition to its transistor-like switching capability.
For example, if terminal 110 was permanently connected to a holding voltage (say 0.2 volts) while terminal 120 was permanently connected to a resistor, then you could write a "1" into the cell with a short pulse on the control terminal 140. That "1" would persist until terminal 120 was shorted to ground (or voltage removed from 110), causing material 100 to "snap back" into its non-conductive state.
In this embodiment, such a "one switch, one resistor" latch is a low-cost, high-density, high-speed replacement for DRAM and SRAM in addition to its transistor-replacement qualities. And no silicon of any variety is required - this stuff can be conjured out of thin air (or oxide, as the case may be). This is the most important characteristic of chalcogenide switches: they are constructed of amorphous material.
This is essentially the opposite of traditional silicon transistors, which are constructed of a single crystal of silicon, which are only formed through a complex, specialized process. As such, you can print amorphous devices with ink jet technology or roll them on with printing press technology. I fully expect to see hobbyists printing their own chalcogenide-based memory and processors at home, for example.
In my opinion, it doesn't matter what memory technology lies at the core of 3D XPoint - the fact that they are using these transistor-less chalcogenide devices is even bigger news because we now know that this stuff is real. While we don't yet know the performance of the devices themselves, there is potential for them to provide an immediate leap in speed and power efficiency (think 10 or 20Ghz with no need to charge your device every day).
In my opinion, this patent and its supporting patents are priceless. However, the priority date is 2003, which means that this will become public domain in 2023. The other patents roll off through 2026 so there is about another decade of exclusive rights that are established.
For those who haven't followed my aforementioned more complicated articles, I am a Micron investor by way of Energy Conversion Devices ("ECD"), which initially developed this technology. ECD was put into bankruptcy about four years ago and, while that bankruptcy is still open, Micron purchased this technology from ECD.
However, last summer, Micron transferred this technology (and many others) to a company called Carlow Innovations LLC. I can only assume that there is a good reason for this because this patent is the primary reason that I'm a Micron investor. It might be worth $100 billion, for example, if the currently undisclosed performance, durability and power efficiency are all good.
If Moore's Law had broken down in 2000, we wouldn't have smartphones and tablets or all of these Internet of Things. More importantly, we wouldn't know that we were missing these devices. With silicon costs now increasing for the first time in history, this sort of innovation is hindered. AMD's (NASDAQ:AMD) processors are still built with 32 nanometer technology, for example. If something can re-enable this innovation, that's worth something.
A lot of something.
Disclosure: I am/we are long MU, INTC, ENERQ.
I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.