Summary
- Micron has had a very busy 3-1/2 years.
- Remember the Micron/Intel joint venture?
- Expect to hear about some game changing new memory technology in the near future.
Some recent Micron history:
At the beginning of 2012 Micron (NASDAQ:MU) was in deep trouble. DRAM sales had a negative 50% gross margin. Micron had no position in mobile DRAM. NAND did not have the necessary scale to be a successful operation. Micron was the industry joke, with one foot in the grave and the other on a banana peel. The stock was selling for under $7 per share.
In February 2012, Intel (NASDAQ:INTC) and Micron renegotiate their joint venture agreement which bolsters Micron's balance sheet in preparation for a serious offer on Elpida, the bankrupt Japanese manufacturer of DRAM … Primarily mobile DRAM. Micron stock is under $8 per share.
On May 16 2012, Apple (NASDAQ:AAPL) gives large mobile DRAM orders to Elpida. The orders represent 50% of Apple consumption of mDRAM. Micron stock closed May 2012 at $5.45.
July 2, 2012, the Micron offer for Elpida is accepted for $2.5 billion.
Dec 2012, the spot price of 2Gb DRAM chips begins to increase and doesn't stop until it is three times the recent low.
August 1, 2013 Micron closes deal to buy Elpida for $2.5 billion. Stock closed August 13 at $13.57
August 2014, the spot price of 4Gb DRAM chips begins to decline in price and doesn't post a green day yet in over 11 months.
Dec 1, 2014, Micron stock price peaks at $36.49. June 26, 2015 stock price closes at $18.73, 49% below the high.
This is how you go from zero to-hero-to-zero in 3 ½ years.
So, from the beginning of 2012 to the closing of the Elpida deal on August 1, 2013, Micron was understandably intensely focused on the Elpida acquisition to the exclusion of nearly everything else. There were no DRAM shrinks. There were no new fabs.
It took about a year for the attributes of the combined company to be understood after the deal closed. It was an incredible deal. We learned that the $2.5 billion price was in yen, and the yen had spent the previous 12 months increasing in value, so the deal was really $2 billion, in dollars … minus the brilliant yen hedges the former CFO engineered. We learned that during the 12 months prior to the closing, Elpida had managed to put together an unexpected billion dollars in cash. Of course, the Elpida assets had previously been written down to near zero, so there would be little depreciation for a couple of years.
All in all this deal will go down in acquisition history as a VERY sweet deal.
About the middle of 2014, I recognized that everything was not as wonderful as I thought, and the price had rapidly caught up with the "new valuation" that was only then becoming clear. The company was capable of producing $16 billion in revenue and $3-4/share in earnings, however, the distraction of the acquisition caused DRAM manufacturing technology to lag by one node.
I thought NAND, at 16nm, had managed to keep up with, and maybe even lead, the competition a bit. Then we learned that Micron had not really moved to 16nm technology and they hadn't moved to TLC technology (my speculative opinion is that they knew what was coming in 3D NAND and 2D NAND would not be worth the engineering investment). So, Micron was, in fact, behind on NAND technology, which I considered critical to the company's future. I sold my call options about a year ago.
Here's where Micron is now:
In DRAM, Micron is pulling off a 3-fer. They are ramping 20nm lithography wherever possible, converting to the new DDR4 technology, and finally expanding basic chip density a generation to 8Gb monolithic chips. These are high risk maneuvers. And not everything is going perfect. Yield is apparently less than wonderful at Inotera. Low yield causes cost to increase temporarily, but the increased cost is a sign that the new processes are being ramped, and they have to get ramped to lower cost ... Semiconductor Business 101. Unfortunately prices are also down because of further weakness in the PC market.
While PCs might be soft, mobile DRAM consumption is expected to surge with the next round of smart phone offerings later this year. A PC recovery predicated on Windows 10 and the Intel Skylake processor's near simultaneous release would be icing on the DRAM cake. By the end of the year Micron could be producing 50-60% more DRAM bits at 30% lower cost per bit than today.
In NAND, the company seems to have recovered nicely with 16nm in MLC and TLC ramping about a quarter earlier than they mentioned this time last year. The real story in NAND is, of course, 3D. 3D NAND will not be a big revenue needle mover this year, but there will be enough parts in the wild soon (60 days?) for detailed evaluations to be made. I believe the dramatically improved endurance and data retention of these parts will make this technology the must-have NAND for the next 3-5 years. I think all competitive 2D and 3D NAND will look like junk next to the Micron/Intel product. The SSD (Solid State Drive) business is expected to total $65 billion, all based on high performance NAND memory. This should peg the needle on the Micron growth-o-meter for the next five years.
As discussed in Stephen Breezy's excellent article, there are emerging memory technologies that are near enough in the future to be seriously considered in investment decisions.
Micron is showing technology roadmaps with "New Memory Technology A Gen 1" appearing this year and Gen 2 appearing next year. The roadmap also includes "New Memory B" scheduled for 2017.
Micron has been discussing "Storage Class Memory", without exactly letting us in on all the secrets. We can take a stab at what it will be: It will be Random Access Memory (RAM), just like DRAM. It will be non-volatile, like NAND (data remains with power off). It will insert itself between system RAM and NAND storage, kind of like a transition from DRAM to NAND. It will be fast, but not as fast as DRAM; it will be cheap, but not as cheap as NAND (not right away). It will tend to reduce the amount of DRAM necessary in a system. Let me repeat that, SCM will reduce the amount of DRAM required in a system. The emerging memory technology will reduce system cost, and increase system performance; it will surely reduce the power of future mobile systems.
Micron will do well to transfer as much DRAM production to 3D NAND as possible because DRAM consumption will be shrinking in the future.
In 2014, Micron moved 40,000 wafer starts per month of DRAM capacity in Singapore to NAND production. There were a lot of questions and second guessing about that move, but in retrospect it seems like the right thing to do.
As mentioned in Breezy's article, the on-chip cache memory in CPU chips now takes up over half the silicon area of a CPU chip. This cache memory has stopped scaling so that Intel has talked about taking it off-chip. To this end, Intel has dabbled with in-package DRAM used as a huge 128GB level four cache that is shared by the CPU and the GPU. Of course, this chip is built by Intel for their own use. If that trend continues, we will see Intel increasing the amount of in-package DRAM and the new SCM taking over the job presently done by system DRAM. Since the SCM is non-volatile, this change will also take place in the higher end mobile SoCs as a power reduction strategy. What I am saying here is that perhaps the advent of SCM will cause the DRAM market to begin shrinking.
This is a good place to remind the reader that Micron and Intel have a joint venture for the purpose of developing and producing NAND and "emerging non-volatile." It is entirely possible that 2015 will see the very beginning of some new, disruptive direction in memory technology. Micron even refers to the 2015 new memory technology as "disruptive" in their bullet points: - "Enable disruptive new memory technology and prepare for ramp in 2015"
Intel and Micron seem to be connected at the hip, so a stock swap combination would not surprise me. A combination would make a $75 billion dollar business with the potential to become a $100 billion company in 3-5 years.
The analysts were wrong/late in the beginning and they are wrong now.
With all this in mind, I am calling Micron a strong buy again. This will be new territory, so we can expect many more speed bumps than with the Elpida phase of the Micron story. That means that this might be the time for stock instead of options, or some options bought with funds from writing covered call options on the stock. Micron will be bought or go on to new highs, at which point covered calls could provide serious income for a long time.
No comments:
Post a Comment