After decades of failures and consolidation in the memory industry, few participants remain, with the playing field shrinking from 40 to 50 a couple of decades ago to just 3 to 4 significant players now. Despite this, the revenue and range of products from DRAM, NAND, and NOR is greater than ever, and the few surviving combatants in this battle royal stand to finally make sustained and significant profits which should result in multiple expansion, similar to what has been observed in hard-drives with Seagate (NASDAQ:STX) and Western Digital (NASDAQ:WDC). Micron (NASDAQ:MU) in particular stands to benefit, as it has been looting the battlefield as its foes have fallen.
DRAM is volatile memory, the thinking memory of your phone, tablet, computer, gaming system, car, or other electronic device. It's produced by Samsung, SK Hynix, Micron, and Elpida, with some minor players taking a couple of percent of the market. NAND flash (and the less common NOR, which is quickly being replaced with NAND) is non-volatile memory, the long-term storage of your electronic device. It's produced by Samsung, SK Hynix, Micron, and the partnership of Toshiba/SanDisk (NASDAQ:SNDK). Samsung, SK Hynix, and Micron periodically shift production back and forth between DRAM and NAND to match market demands.
The most recent and remarkable casualty in the memory market was Elpida, which filed for bankruptcy in 2012. Micron and Hynix both bid for the company, and Micron won at 200 billion yen (about $2.5 billion at the time, a steal considering the $6 - $8 billion replacement cost of the facilities.) The yen has since depreciated dramatically vs. the dollar which has had the dual effect of making Elpida's production cheaper and the purchase price less expensive at just $2 billion today.
Unfortunately, Micron will not enjoy the benefit of about half of the improvement in the purchase price from the depreciating yen due to currency hedges. At the same time as the purchase price and cost of production were falling, the price of memory has been rising. The spot price of DRAM has doubled since the start of 2013, and NAND has risen by about 30%.
Elpida and its subsidiary Rexchip add about 200,000 wafers/month of DRAM to Micron's current 400,000 wafers/month split between DRAM and NAND. More importantly, Elpida adds a strong presence in mobile DRAM which is in short supply and sells for about double the price of PC DRAM at only 25% higher production cost. This is a market that Micron has struggled to enter, with just 2.7% market share in the first quarter, compared to Elpida's 18.5%.
Micron's 400,000 wafers/month resulted in $2.32 billion in revenue in the March/April/May 2013 quarter, with 47% from DRAM, 40% from NAND, 8% from NOR, and 5% from "other." While this led to a $.04 EPS gain, it's important to note that excluding one-time items, it was a $.15 gain. Based on pricing trends, it's likely that the third month of the quarter was most profitable, and Micron should easily beat the current quarter earnings estimates even without the inclusion of Elpida.
In a pro forma filing to the SEC, Elpida was reported to have $794 million in sales in the quarter ending September 30, 2012. This was before the dramatic rise in DRAM prices due to chronic shortage, and also before the signing of an important contract in which Apple (NASDAQ:AAPL) now buys 80% of Elpida's mobile DRAM. In the fourth quarter of 2012, Elpida had $968 million in revenue. Most importantly, according to Japanese press reports, March/April/May 2013 (Micron's most recent quarter) Elpida had revenue of $283 million/$400 million/$442 million ($1.125 billion total) and profit of $45 million/$100 million/$148 million ($293 million total) or about 28 cents per Micron share. Both revenue and profits have rapidly accelerated in recent months.
From here we can easily estimate the current quarter's revenue and profits. During the most recent earnings release, we heard the following statement from Micron's CFO, Ronald C. Foster:
So guidance for Q4 DRAM is as follows; using quarter-to-date selling prices and projected product mix for the quarter, ASPs would be up mid-to-high single-digits compared to the Q3 average. Projected bit costs are expected to be flat relative to Q3, projected production volume is up high single-digits.
I believe this to be a conservative projection due to the phrase "using quarter-to-date selling prices" -- 80% of Micron's sales are under contract rather than spot, and contract prices are still about 20% below the level of spot, but they are converging. Quarter-to-date prices should be lower than the ASP for the quarter.
Likewise with NAND, which had contract prices increase by 2-5% in the 2-week period following this conference call:
Bit costs are expected to be down high single-digits, while bit production is expected to be up high single-digits.
NOR is expected to drop by a third, replaced by NAND.
Inotera, a joint venture between Micron (35.5% ownership) and smaller Taiwanese ally Nanya, was a loss of $13 million for Micron in the previous quarter, and will be a gain of $41 million in the current quarter. Inotera posted an impressive 48% gain in revenue quarter-over-quarter which is indicative of the kind of recovery that the memory industry is undergoing.
Using the conservative Micron estimates for DRAM, NAND, and NOR, plus Inotera's results, I get revenue of approximately $2.55 billion and EPS excluding one-time items of approximately $.42. Elpida should close on or around July 31, and if they elect to include it in the current quarter's results, it should add something on the order of an additional $1.3 billion in revenue and $.40 EPS, which would give a total of $3.85 billion in revenue and $.82 EPS. There will also be some one-time gains including a gain of about $.50 per share on the yen depreciation after the purchase of Elpida, and a gain on increased investment in Inotera by Nanya.
As impressive as this is, it's just the beginning. Industry predictions are for increases in demand to outstrip supply through 2015, and yet an unprecedented thing is happening -- investment in new equipment and facilities is decreasing compared to last year. There are a whole host of reasons including impending changes in technology, less market share to be gained by participants, and the increasing cost of building a new fab, resulting in lower returns on capital. Samsung is the only one currently building a new fab. It will be in Xi'an, China, it will cost $7 billion, and it isn't expected to have a significant effect on supply through 2014. Clearly with a cost of $7 billion for even a single fab, not to mention significant technological requirements, there's a high barrier to entry by new participants.
Meanwhile technology improves and costs just keep falling -- typically cost per bit falls 30% per year. Micron will be manufacturing 128 Gb NAND at 16 nm by the fourth quarter of 2013, the most advanced process in the business. According to memory expert Jim Handy, NAND currently sells at $.50/GB and costs $.32/GB to manufacture, but should cost about $.23/GB to manufacture with the transition to the smaller process. Micron is also already looking at ways to reduce back-end costs such as testing and packaging. ChipMOS (NASDAQ:IMOS), Micron's current packaging provider, is widely expected to announce a new contract within the next few months for the chips Elpida produces. At less than 3x EV/EBITDA, I believe ChipMOS shares are extremely undervalued, particularly since the company stands to be one of the prime beneficiaries of sustained stable pricing in the memory sector. This cost reduction will only serve to multiply Micron's profits with time.
The merger with Elpida should happen July 31, and a Micron analyst day is scheduled for August 9. We'll see the first post-Elpida earnings shortly after in September. I think that the recent pullback represents an excellent entry point. While there may be short-term pricing noise, the long-term outlook for the memory sector looks bright. The market is missing what I expect will be a new era of rationality, leading to significant sustained profitability, and a multiple re-rating. I believe that the market will react quite favorably to these upcoming events, and that the time to invest is now.