Most investors in the technology/semiconductor/memory area understand that, due to crushing competitive pressures, investing in memory manufacturers has been a losing proposition for individuals, companies, and even countries.
The reason for that is that there is simply no differentiation in the memory business; a bit is a bit is a bit. Given an acceptable quality level, the next parameter has always been price. Japan and Taiwan, as countries investing in the memory business, have been brought to their knees by the next country/company willing to throw money away into the memory business. The list of memory suppliers over the years has declined from 40-50 suppliers to the present four suppliers, with only three of them in the DRAM business.
A 1024 bit DRAM of the early 1970s cost $8 in high volume. Today you can buy 10,000,000 times that amount of memory for the same $8. It would come in the form of 3 1/3 4Gb DRAM chips. If that doesn't put a funny look on your face, you have no emotion.
What are the factors that will keep this new oligopoly from continuing the cycle of murder and suicide by price?
Let's take a walk down memory lane for some of the answers. Anyone remember the early calculators? Virtually every semiconductor company was pursuing the calculator chip business in the early 1970s. As a matter of fact the pursuit of the calculator chip business (the first killer app) led Intel (INTC) to invent, out of necessity, the first microprocessor. OK, OK, enough history. The fact is that the competition for these calculator chips became so intense that many semiconductor companies thought they should control their destiny by making the entire calculator themselves, so they did. They did the same thing with electronic watches.
I used to make business visits to National Semiconductor in those days and load up on $5-8 watches and calculators for friends and family.
The problem with that was that, in a very short period of time, the enabling integrated circuit became so cheap that it only represented about 5% of the cost of those calculators and watches. The semiconductor companies had no advantage when 95% of the value of these products was in bands and crystals, displays, and plastic parts. Consequently the semiconductor companies lost their britches and finally quit and left the consumer business to those who understood it.
Something similar, but subtly different is going on in the memory business today. The memory manufacturers are producing high level assemblies and even consumer products themselves.
Haven't we seen this movie before?
Not really, and I'll tell you why. Let's take a little stroll around the memory business as it exists today.
Samsung is consuming as much as half of its memory output to support other Samsung consumer product divisions. They supply DRAM for their PCs, mobile DRAM and NAND for their exploding smartphone business. Samsung is also producing Solid State Drives (SSD) for their own consumer products as well as other PC manufacturers.
Micron has made DRAM memory modules for years where the Micron content is 90% plus of the end product. The Micron Crucial Division supports this business today and sells everything from DRAM memory modules to Hard Disc Drives bought from third parties. Crucial is one of the early SSD manufacturers and, as such, is a large captive consumer of the NAND output from Micron. The LEXAR division has a great deal of the look of a SanDisk (SNDK) in that they supply "Thumb" drives, compact flash cards for photography, and micro SD cards of all descriptions. Of course, Micron is supplying component level products to the entire industry with large mobile DRAM business going to Apple from the newly acquired Elpida division.
SanDisk, of course, set the model for selling high functionality NAND based product, such as Thumb Drives and micro SD cards to the consumer market. The company has recently entered the SSD business through a couple of acquisitions.
Missing DRAM will cost SanDisk in the high value embedded business where both DRAM and NAND are inside the same package.
Hynix brings up the rear with a recent entry into the SSD business, but has a robust business in the component level mobile DRAM and NAND business.
So, is this just a calculator and watch business all over again?
No way! The common thread in the new consumer model for memory companies is that the products are 90-95% semiconductor (memory) content.
A micro SD card has a few passive components and a NAND memory chip.
SSDs are a controller chip, a printed circuit board and… umm, a whole bunch of NAND memory.
When 8Gb PC DRAM chips start shipping, a 4Gb PC module will be four 8Gb chips, or just four individual chips on the mother board for thinness. Not much opportunity for independent memory module manufacturers, those wonderful people who have historically skinned the memory manufacturers for a nano-cent on price, thereby contributing to the profitless DRAM cycles.
No, this time it's really different for the memory industry. For every one of those 128 or 256Gb SSD shipped by the four NAND makers, you can decrement one unit from Seagate (STX) or Western Digital (WDC). That's $30 billion worth of HDD business that, over time, will convert to $60 billion is SSDs. The HDD business is a slow dying business, much like the core memory business of the 1970s. I will have long since run out of birthdays by the time the last HDD is sold, but that day will come.
At 90-95% NAND dollar content, independent SSD companies will have precisely no chance of long-term survival. They will either die or be bought for pennies by one of the NAND manufacturers.
The best play in this new world of memory is Micron because investors and analysts have demonstrated that they really don't know what the company is worth since the closing of the Elpida acquisition. Micron has very quickly jumped from an also-ran fourth place supplier, leap-frogged Toshiba/SanDisk and SK Hynix, to rival Samsung as the world's largest memory supplier. Considering the Samsung internal consumption of memory, Micron has really become the world's largest merchant supplier of a broad line of memory technology.
You hear Micron talk about "scale" in their investor presentations. The first thought that comes to mind when scale is mentioned is cost… as in lower cost. While that is true, larger scale also confers on Micron the ability to "throttle" production in order to keep prices from collapsing as in 2012. If the old Micron shut a plant for a while, they were out of that business. As the swing player, Micron can convert DRAM capacity to long-term growth of NAND and back again, if necessary.
Micron, with Elpida, now has a total annual wafer capacity of 7.2 million wafers. At $2500 revenue per wafer, Micron would be an $18 billion enterprise from a level of $8.2 billion and losing $1 billion in 2012. You don't need to be real sharp on arithmetic to see that this is a different company from a profitability standpoint.
The Memory business IS different this time and Micron IS a different company this time.