The Dynamics of Semi Pricing and How it Drives Flash     3:53 PM -- 5/2/07

Q) I recently read that a report from iSuppli predicted that by the fourth quarter of 2009, over half notebooks sold will use flash memory for storage, rather than hard drives.

Matthew Wilkins, principal analyst at iSuppli, said "the shift to flash is being aided and abetted by a dramatic drop in prices for NAND memory."

This seems like an opportunity to buy into SSD manufactures such as SanDisk (SNDK) and others for the long term. I also read that SNDK will be the manufacturer used for Dell (DELL) laptops using SSDs. Can you over some other companies poised to take advantage of this change in technology, especially companies that focus on SSDs?

I'd like to identify stocks worth holding on to for a few years within this market.

A) I can't remember exactly when I wrote it, but it seems like a couple years ago when I predicted that we would see flash memory take a significant chunk of the laptop market in 2009. It's good to see the large research companies are catching up to this thinking.

Before we get to the meat of this situation, I think its best to understand why this change is rapidly approaching and, to do that, we really have to start with the basics. Besides, I think this is a great case study that will help readers understand some of the important dynamics that drive the semiconductor market.

In 1965, Gordon Moore wrote a white paper exclaiming that the density of components capable of being economically fit on a given size piece of silicon would double every year through at least 1975. This white paper covered technical and economic issues facing integrated circuit designers and was published in April 1965 in Electronics Magazine. Following this publication, renowned Cal Tech professor, Carver Meed, dubbed Dr. Moore's prediction as "Moore's Law."

Moore's Law has survived butchering, misquotes, and a slew of naysayers during the last four decades, but has proven accurate through the test of time.

In 1965, the economic limit for integrated circuits (IC) was about 50 "components." It wouldn't be totally accurate to describe these components as transistors, but for our purposes, that will do just fine. Looking towards the future, Dr. Moore prophesized that by 1970, the economic limit would be about 1,000 "components" per IC and wrote in his landmark paper that the implication was that the cost per component would be one order of magnitude less by that time. The following is what Dr. Moore wrote we could expect beyond 1970:

"The complexity for minimum component costs has increased at a rate of roughly a factor of two per year (see graph on next page). Certainly over the short term this rate can be expected to continue, if not to increase. Over the longer term, the rate of increase is a bit more uncertain, although there is no reason to believe it will not remain nearly constant for at least 10 years. That means by 1975, the number of components per integrated circuit for minimum cost will be 65,000."

By and large, Dr. Moore's projections have held true for a bit longer than four decades so far and will likely make it through at least one more. To give you an idea as to the aggregate result of what forty years of doubling does, there were enough transistors sold last year to put over 200 on the back of every ant on planet earth - tens of millions for every human - and all this from only about one square mile of silicon. In the western world, the average human owns over a billion transistors - since you're reading this on a computer screen, it's probably safe to say that you own more transistors than you do anything else - even carpet fibers.

Now, let's take a step back from the overwhelming realities of Moore's Law and look specifically at NAND Flash to get a better understanding of what is going on in that technology and why it is poised to take market share from the hard disk drive space.

In 2005, I wrote that due to the heavy capacity commitments being made to NAND Flash, the technology was destined to move through at least a few years of Moore's Law at hyper-speed. More specifically, my prediction was that NAND Flash prices would fall in half about every nine to twelve months rather than the twelve to eighteen months that has been more typical of Moore's Law during the last decade. If we look back at the aggregate results, that is pretty much what we've seen in the Flash industry and it is exactly why profits have all but vanished for Flash producers. The reason why NAND Flash moved so fast was that there has been and still is an overabundance of supply and this supply imbalance has driven prices down a very steep curve. When I looked at this steep curve a couple years ago and ran some calculations, I wrote that we would see NAND Flash take niche laptop applications in 2007 and take meaningful market share in mainstream laptops in 2009.

If you've not had a chance to read it yet, I think you'll get tremendous value from the report we published in 2005 covering semiconductor cycles. This report covers not only the dynamics that drove past cycles, it also looks to the future and predicts what I thought at the time was heading our way. So far, what's transpired since the paper was published has been very much in line with these predictions and I think what I predicted that has yet to come will prove to be fairly accurate as well. This report will also help you understand why no one can provide an accurate answer to your question as to how to invest in NAND Flash.

While it's very easy to predict that the number of NAND Flash memory bits sold will radically increase over time, it's impossible to predict if the companies selling these bits will be profitable and, if so, to what degree. To give you some scope on this, the number of DRAM bits sold increased by well over two orders of magnitude from 1995 through to 2006. However, even with this amazing increase in sales (a 100x+ increase), DRAM revenue during any of these subsequent years has never come close to matching the peak revenue set in 1995 and even with Vista driving the average PC DRAM content to 2GB per machine, it won't happen this year either.

Obviously, the reason for this seemingly dichotomous situation is not a lack of demand. The growth in DRAM demand has been amazing and in reality, my "two orders of magnitude" estimate could be as much as a whole order of magnitude towards the conservative side.

The problem with DRAM technology as a business model has been strictly driven by the fact there has been more than enough supply to fulfill this demand and, ever since the revenue peak in 1995, the overabundance of supply has driven prices down much faster than costs. As a matter of fact, I'm of the opinion that if you were able to aggregate all the profits and losses of the DRAM industry over the last decade, you would find that the entire industry has operated at or very near to a net loss for the last decade.

Heretofore, memory has lived by the "law of closets;" the more closet space you have the more stuff you will buy. In the DRAM space, Microsoft (MSFT) has been more than happy to add to the complexity (I'll let you decide if this complexity is matched by utility) of its operating systems which, in turn, has filled all the DRAM the average PC buyer wants to afford. In other words, we've not reached a point yet where we've been willing to say, "that's good enough, I have as much memory as I need."

In the NAND storage space, it started with relatively low-resolution digital pictures and music, but has now evolved to high-resolution pictures and video. With video now growing into a mainstream driver, this industry could also be quite a ways from "good enough." However, a considerable percentage of the laptop market is not driven by the desire to store a library of movies; we'll be content with storing these memory-eating videos on home servers and mass storage systems and we'll download only what we want to carry to our laptop. In this significant percentage of the laptop market, size, weight, power consumption and boot time are often much more important concerns than storing the entire Star Wars library. Due to this, I think we're going to see an intersection of "good enough" (enough memory to satisfy the application of a meaningful market share) and cost hit by NAND Flash in about two years and the added volume from the emerging niche laptop applications will be one of the cost curve drivers.

What happens in these situations is the spirals of demand, which are actually driven by lower costs, are quite visible and there is more-than-ample competition striving to hit the cost objectives that will spur the demand. And, as each point on the curve is reached, we can see the next knee in the demand curve and that continues to feed the spiral.

As I wrote in my SNDK preview as well as in the last State of Tech report, SNDK does have some advantages in the current early adoption stage of NAND Flash in both the PC and industrial markets. SNDK acquired two vital technologies from MSystems that I think it will be able to leverage effectively during the next year. These were enhanced multi-level-cell (MLC - where multiple bits of information can be stored in a single memory cell) and the technology that allows system designers to boot directly from NAND Flash.

We've already seen Hynix sign a MLC license with SNDK and I suspect we'll see Samsung do the same later this year. We've also already seen SNDK introduce NAND Flash drives for notebooks and industrial applications that incorporate the MSystems direct boot technology. While I continue to believe that SNDK is trading in "no man's land" (too high to buy and too low to short, which indicates at least some of this thinking is already built into the price), I do believe we'll see some positive results later this year from the aforementioned efforts.

However, the reality is the cost curve that will be necessary to fulfill the prophesy of NAND Flash laptops will be driven by not only SNDK, but the partnership of Intel (INTC) and Micron (MU), Samsung, Hynix, and Toshiba.

Investor Considerations:

Rising demand does not necessarily lead to rising profits. And, since investment returns are governed by profits, you cannot make investment decisions based solely on demand.

What both iSuppli and I have predicted is a cost curve that will stimulate demand. If demand does materialize as we have predicted, it will be due to (not just "aided and abetted by, but absolutely due to) declining costs - think from the supply side, not the demand side (this applies to all economic thinking).

What I predicted two years ago is now abundantly obvious and you can't make money by investing in the obvious. Sure, NAND Flash stocks like SNDK will continue to be volatile, and if you are lucky enough to time the troughs and peaks you can make a killing. Of course, if you're lucky enough to guess the next national lottery number you can make a fortune too. However, I think that when you start depending on luck you are speculating rather than investing.

To win as an investor in this foray, you'll have to be able to predict when demand will outstrip supply ahead of not only Wall Street, which will quickly factor its thinking into the price of the associated stocks, but also ahead of the six major players in the NAND Flash space. Ironically, while Wall Street is factoring its thinking into the price of these stocks, these companies will try to sense a shortfall in capacity and, if they do, they will quickly bring on more capacity and end up either subverting the capacity shortfall and, in many cases, end up in another oversupply situation. If you think you can beat all these highly motivated parties with billions on the line, including six huge corporations with the ultimate inside view of the industry, in predicting the next NAND Flash shortage, I wish you the best of luck.

Bottom Line: While the headlines and research companies focus on what's really abundantly obvious, I'll continue to look for what the market is missing; what will be driven by the obvious that few have realized yet. Of course, this implies that we'll occasionally be wrong and in other cases be too far ahead of the market and have to suffer through long periods of the market telling us we're wrong before it says we're right. However, since races are seldom won by the fearful, we'll not let fear stand in our way of trying to win at what might be described as the most competitive sport in the world - beating the tens of thousands of skilled investors we call simply "The Street."

Disclosure: At the time of this publication, out of the stocks discussed herein, Paul McWilliams had a long position in INTC.