On Thursday, October 10, we will get the first clear picture of the Micron/Elpida (NASDAQ:MU) combined companies' operations since a pro forma filed back on February 5, 2013 that covered combined operations through September of 2012. Any good information on the new Micron is over a year old, and that was a very eventful year with PC DRAM contract prices up over 130% in that time and already profitable NAND up a more moderate 30%. All this time, costs have continued to decline due to migration to newer, smaller production nodes.
By way of review, Micron Technology is in the memory business. All types of memory: All variations of DRAM, all complexities of NAND and NOR flash memory, high value future technology such as the HMC (Hybrid Memory Cube).
On July 31, Micron officially absorbed the biggest and best Japanese memory manufacturer, Elpida, for what many are calling the "steal" of the century. Elpida was made up of the memory remnants of NEC, Mitsubishi, and Hitachi. From personal experience, I can say that these three companies had, and still have perhaps the world's best memory technology. Elpida's demise was driven by an ever more expensive yen and a persistent over-supply of memory product by state sponsored and funded competitors.
So what will we see on October 10th?
Very bright analysts find it frustrating to analyze Micron because of the many moving parts within the company and the incredibly violent moving parts of the memory market itself.
I'm going to take a simple approach by looking at 300mm wafers as "widgets;" widgets that have an average selling price and average cost.
In the Micron press release of July 31, 2013, it was said that Elpida could produce "more than 185,000 wafers per month" and that this would amount to 45% of the total output of the existing Micron. So, 185 divided by .45 should give us the original Micron wafer capacity of 400,000 wafers per month. The combined capacity is about 585,000 wafer starts per month, give or take. Annualized, this gives us about seven million wafers per year of raw capacity.
Elpida is 100% DRAM with most of that being the very profitable mobile DRAM. For simplicity, think of the old Micron as 44% DRAM, 44% NAND, and 12% NOR and "other."
DRAM for Elpida and the old Micron combined would be about 4.33 million widgets (wafers) per year.
NAND for the new Micron is the same as NAND for the old Micron at 2.11 million widgets per year.
NOR and other is about 576,000 widgets per year. The total adds up to just a tad over seven million widgets per year.
DRAMeXchange lists the current PC DRAM contract price for a 4Gb DRAM as $3.44. Since PC DRAM is typically the least expensive DRAM device, this price would be a conservative price to use for guesstimating the average selling price of the DRAM widgets. I'm using the 4Gb here since Micron said months ago that 4Gb bit production passed though the 50% point for total DRAM production, so the 4Gb, not the 2Gb DRAM is the new industry price benchmark. That 4Gb DRAM is about a 60 sq. mm chip, which would produce 950 chips per 300mm "widget" times $3.44 would give about $3250 of revenue per widget. Times 4.33 million DRAM widgets is just over $14 billion in DRAM revenue.
I'm going to use 128Gb NAND chips as the benchmark for NAND products since this is, or will become, the most efficient part to use in solid state drives, which is, or will be the major driver of NAND growth.
The 128Gb NAND chip is about 190 sq. mm (soon to be about 125 sq. mm on Micron's 16nm process), which gives us about 300 chips per wafer. This site gives us an average contract price for an MLC 128Gb NAND of $9.00. That produces a price for the NAND widgets of $2700. Times 2.11 million is about $5.6 billion in NAND revenue.
If I average the DRAM and NAND widget prices, I come up with about $3000 per widget for the 576,000 "other" category, or about $1.7 billion in "other" revenue.
All added together, a reasonable annual revenue number for the new Micron is $21.3 billion either now or ramping to that number in the next year.
OK, now what do these widgets cost to produce?
Over the past year and the most recent quarter, the cost of goods sold by Micron divided by three times the monthly wafer produced appears to give a cost of less than $1500 per wafer.
In the February 5th pro-forma, the same calculations for Elpida produce a cost of an individual widget of about $1050, much less than Micron because most of the Elpida assets were written down during the bankruptcy proceedings.
The same pro-forma list total operating expense for the combined companies as $531 million for the quarter ending September-November 2012. In the same pro-forma, the total year operating expense come to about $2.350 billion. Since probably everyone got raises, I think we should use $4 billion for op ex going forward.
Per the same pro-forma, the total interest expense for the new Micron will be about $241 million.
Here we go:
|Revenue ($ mil)||21,300|
|Op Ex (Total)||4,000|
|Income tax (20%)||1,452|
|Dilute shares (BIL)||1.2|
There we are folks, something to discuss for the next four days. Whatever comes out, the bottom line is that the new Micron is a very different company from the Micron that the Street has come to know and hate.
On the other side, Goldman Sachs (NYSE:GS) has put Micron on their ten most overpriced (risky) stocks. But then GS also put the 4% yielding Intel (NASDAQ:INTC) on the list. How can Intel at 4% yield be risky? Maybe Intel and Micron, being connected at the hip, have cut off Goldman Sachs from financial services business?
Thursday will be interesting. I'm ignoring Goldman and holding all-in through earnings. Good luck.
Russ Fischer - in Lake Havasu City, AZ, at the personal watercraft world final races, hoping for my son to win, or at least not get hurt badly.
Disclosure: I am long MU, INTC. 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.