More About The Intel Investor Meeting

| About: Intel Corporation (INTC)
This article is now exclusive for PRO subscribers.


Air gaps are a significant technology breakthrough.

Intel's 3D NAND process is fundamentally different from other proposed 3D NAND approaches.

Intel could secure $35 billion in SSD business in as short as three years.

Time has a way of making some things more clearly understood. I believe this is the case with regard to the recent Intel (NASDAQ:INTC) investor meeting.

The meeting took place on November 20th and was unusually upbeat.

We are only now getting very well thought out articles on Seeking Alpha, such as this one by Ajay Palekar that parallels my own thoughts that Windows 8 was at least as responsible as tablets for the slowdown in PC sales. Ajay feels that Windows 10 and the fifth generation Core CPUs will restore the PC business to some level of sustainable growth.

I want to take the rest of this piece to focus on a couple of items that apparently are not understood as well as they might have been.

The first of these is the first ever use of "air gaps" between signal wires in a logic process to maintain signal integrity in the realm of 14nm fab processes.

This is a remarkable materials and process breakthrough in the category of strained silicon, HKMG and TriGate transistors.

Air gaps are infinitesimally small air pockets built into the circuitry to reduce signal coupling between these equally small wires connecting over a billion transistors.

Even Daniel Nenni of SemiWiki, no cheerleader for Intel, is impressed by this technological move. Nenni admits that "one of the foundries" (probably TSMC (NYSE:TSM)) has a "research" effort underway on air gaps

Perhaps Intel didn't absolutely need to use air gaps to make 14nm work. Perhaps they didn't need strained silicon to make 90nm a reality or perhaps 45nm could have been done without HKMG, and if high leakage isn't a problem, 20nm could have been done without TriGate transistors.

The idea here is that the sustainability of Moore's Law depends on materials breakthroughs as much as lithography magic. It is critical for Intel to master these new materials technologies and build years of experience and expertise with them to move Moore's Law forward, while maintaining a high level of signal integrity on the integrated circuits in question.

So, score another process breakthrough for Intel that will cause a lot of head scratching at competing foundries.

Next, of course, is the issue of a "cost disruptive" 3D NAND process.

As I have been claiming for most of the past year, Intel is up to their ears in memory technology, as shown by the above chart.

We can see that the Intel/Micron (NASDAQ:MU) joint venture has been responsible for a number of "firsts" in NAND technology. This includes multiple patterning, and word line air gaps in 2010, they claim to be first at 3x and 2x nm processes. Intel/Micron was first to a 128Gb chip using HKMG.

By now there should be a picture forming that Intel is using the JV to prove and implement new processes that ultimately end up in the Intel logic fabs.

The last two elements of this chart are the most important for Intel investors. The chart acknowledges the "industry" (Samsung) first of 3D NAND in 2014. Right after this, Intel claims the high ground of being first with "3D NAND with disruptive cost" in 2015.

The last two elements of this chart are the most important for Intel investors. The chart acknowledges the "industry" (Samsung) first of 3D NAND in 2014. Right after this, Intel claims the high ground of being first with "3D NAND with disruptive cost" in 2015.

The repetition of the above paragraph is no mistake. This is important stuff.

Intel is saying that the Intel/Micron implementation of 3D NAND is fundamentally different. Different cost, of course, and probably a different approach to the design and fabrication of 3D NAND. The first product from the JV is a 256Gb chip, twice the density of the best available today.

I am shocked that no analyst has even expressed an opinion on this remarkable development.

in my previous article, I speculated on the manufacturing cost of this first chip from the joint venture at sub-$2.

That sub-$2 cost involves some informed guessing and I believe it to be reasonable and consistent with the industry term of "disruptive cost".

For discussion sake, let's assume that the disruptive cost is about the cost of a 128Gb chip today. That would put the cost at about $3.50.

A 128GB SSD could be built from four of these chips at a total cost of ~ $18. A selling price, with 65% gross margin would be ~$51.

A 256GB SSD could be built with eight of these parts at a total cost of ~$32. A selling price, with 65% gross margin would be ~$91.

A 1TB data center drive could be built with 32 of these chips and could cost ~$330.

To trigger a total replacement of HDDs in client PCs, those in a position to know, use a selling price of about $60 per 256GB SSD as the tipping point price. From the above, we can see that with the first product of the Intel/Micron JV, the 256GB SSD is only 50% above the tipping point price. The very first shrink of this process will drive the selling price below the tipping point price of $60.

The big question is, "How fast will the PC and data center industry forego HDDs in favor of SSDs?" Another question is, "How much capacity will it take to make the conversion as fast as the market wants it to happen.

By the middle of 2015, the JV will have Lehi, Utah running at 80,000 wafer starts per month or 960,000 wafers per year. The output of Lehi will be split between Intel and Micron. By the end of 2015, the Micron Singapore plant should be at full capacity of 100,000 WSPM. By the end of 2016, Micron should have the twin to the Singapore plant in operation for another 100,000 WSPM. There is some kind of foggy "deal" within the JV where Intel gets some percentage of the output of the Micron Singapore plants.

The NAND market is, roughly, made up of four segments - PC, enterprise and data center, mobile and consumer (thumb drives, toys, etc.)

Intel is shipping about 400 million CPU chips per year (100 million last quarter).

With the understanding that Intel will target the high-end, high-function data center SSD market, the size of which I frankly don't know very accurately, let's see what it would take to make 400 million 256GB SSD for the CPU chips that Intel will ship each year.

It will take 3.2 billion 256 Gb chips to make the 400 million SSDs.

It will take 5.5 million wafers to build the 3.2 billion NAND chips, or about 460,000 wafers per month. Right now Intel only has committed capacity of 40,000 wafers per month through the JV, less than 10% of what would be needed to provide SSDs for all the CPU chips.

Just the PC client segment will produce recurring revenue of $24 billion per year at the tipping point price of $60 per 256GB SSD.

From the above we can see that Intel and Micron will only have about 3.4 million wafers available by the end of 2016. That is only about 50% of what will be needed to support this one segment of the NAND business - the PC client SSD segment.

Mobile, smartphones and tablets, will need about another 2.8 billion chips or another 4.8 million wafers per year.

There is something like 10 million server chips sold per year. Let's assume each of those chips will have an associated 2TB SSD. That's another 640 million chips or 1 million wafers.

The consumer market is probably another .5 million wafers.

So, the industry will need about 12 million 3D NAND wafers to make the inevitable conversion from HDDs to SSDs.

I could argue that that much capacity will have to be operating and stable by the end of three years.

SanDisk (SNDK) and Toshiba (OTCPK:TOSYY) won't even start on 3D NAND until 2016.

With the emphasis on thinness, most PCs and PC-like devices will be too thin to even consider using HDDs.

The next thing to think about is that if non-availability of NAND capacity and SSDs began to impinge on Intel's ability to ship CPU chips, what do they do?

My very strong opinion is that if the above scenario begins to play out, it will be a zero latency decision for Intel to fire up their own NAND capacity to ensure that there are no choke points in the PC supply chain.

I can't believe that Intel management isn't considering this watershed conversion every single day.

I believe that, with the Intel position on CPU and server chips, the company can reasonably lay claim to as much as 50% of the rapidly developing NAND based SSD business. In three years the company could add $35 billion of SSD business to what is likely to be $65 billion of legacy business. I believe that a $100 billion Intel can earn $7 across four billion shares and trade at 20 times earnings, or $140/share.

Since fancy lithography is not the limiting factor in the manufacture of 3D NAND, it could be entirely possible that Intel is poised with 3D NAND fab capacity that could be brought online in a matter of months.

Disclosure: The author is long INTC.

The author wrote this article themselves, and it expresses their own opinions. The author is not receiving compensation for it (other than from Seeking Alpha). The author has no business relationship with any company whose stock is mentioned in this article.

Editor's Note: This article discusses one or more securities that do not trade on a major U.S. exchange. Please be aware of the risks associated with these stocks.