- Node shrinks are reducing wafer output in the industry.
- Demand for DRAM is strong and getting stronger.
- Oligopoly rules will prevail as MIPO transitions to ReRAM.
- Profits in DRAM are good now and will get better as margins expand through 2016.
"My other piece of advice, Copperfield,' said Mr. Micawber, 'you know. Annual income twenty pounds, annual expenditure nineteen nineteen six, result happiness. Annual income twenty pounds, annual expenditure twenty pounds nought and six, result misery."
David Copperfield (1850) - Charles Dickens
"On two occasions I have been asked, 'Pray, Mr. Babbage, if you put into the machine wrong figures, will the right answers come out?' I am not able rightly to apprehend the kind of confusion of ideas that could provoke such a question." - Charles Babbage
Frankly, dear readers, I had not anticipated writing this article, but the recent news regarding DRAM capacity increases in the industry drove me to drop what I was doing and return to this oh so well-tilled ground. As investors, I'm sure many of you are like me in that we try to diligently stay abreast of current information regarding our focus companies and industries. Sometimes the information we get is encouraging, sometimes it is discouraging, but either way it can be very valuable for the conscientious investor. For me, a "long" and "deep value" investor, information related to the markets that our companies compete in is always compelling and potentially actionable. One of the publications I follow is DIGITIMES, a Taiwanese newspaper and online publication that closely follows all aspects of the semiconductor market, including the DRAM business.
The question I would pose for you today is the following: when does an article reach a standard that we can call "information" and when is it merely reporting that contains a conglomeration of "facts" that tie to ill-founded speculation that is then conflated into misleading conclusions? Let's consider a recent piece dating from August 28 distilling the DRAMeXchange report on the state of the DRAM business:
Herein is the payoff sentence.
"An increasing number of DRAM makers are planning to expand their capacity in the fourth quarter of 2014 as DRAM market supply tightens, while bit supply could grow nearly 30% in 2015 following recent decreases in die size and capacity adjustments. As a result, the industry's oligopoly structure could face considerable challenges in 2015, among them oversupply that would eat into producers' profit margins, said Avril Wu, assistant vice president of DRAMeXchange." [Bold emphasis added]
Now, rather than regurgitate this rather unfortunate example of business journalism in this document, I encourage you to read the full piece yourself here. What possible conclusions can one draw from the combination of data and speculation that was contained in this article? It certainly sounds ominous, doesn't it? For Micron (NASDAQ:MU) investors this is potentially very dark news. After all, fully two-thirds of Micron's bits are DRAM bits. DRAM last quarter was almost 80% of Micron's gross margin. The company is on the verge of getting better in NAND, but for the remainder of 2014 and the balance of 2015 the company is depending on DRAM to carry the business. Maybe the sky is falling! Before we run right out and liquidate our MU position, let's think about the news and the implications of this "news" in a different way - so here goes:
As yours truly and other contributors have previously discussed, the DRAM business is an oligopoly, and because of that industry structure there are certain rules that apply. We have discussed those rules ad nauseum so I'm not going to repeat them here - for those interested in a quick review, follow the link. The Dickens quote above captures the essence of these rules, in that oligopolies exist for one reason only, and that is to impose artificial limitations on the supply of valuable goods and services - to put a self-imposed budget on it so to speak. Come in under budget so that supply is less than demand - profits will increase as the buyers bid up prices to procure precious inventory. Come in over budget - not good! When supply exceeds demand, prices will plummet and profits will go down as buyers play off the suppliers against each other. It seems pretty straightforward, doesn't it? Happiness versus misery - if you're a stakeholder in any of the memory companies that is. It seems so simple. Are we missing something?
Well, maybe we are - maybe things are a bit more complex than I have painted them. For one thing, it matters a lot what type of product the oligopoly is producing - pure commodities such as barrels of oil, bushels of wheat, or PC DIMMS are much more likely to be exposed to the full set of roller-coaster like pricing effects that occur because of supply/demand imbalances. Product like mDRAM, which is now specialized and which has multiple highly specific variants which are not interchangeable and which furthermore is increasingly being tailored to a specific customer's requirements will see much more muted reaction to the market's current supply demand reality.
All that being said, the rule remains: MIPO (the Memory Industry Producers Organization) has NO interest in generating excess DRAM supply. MIPO has only three members - Samsung (OTC:SSNLF), SK Hynix (OTC:HXSCF), and Micron, who control 90% of the DRAM bits that come to market. Currently, these three offer slightly more than 1 million wafers a month of capacity. Offering more bits to the market than the market can consume, particularly in PC DIMM configurations, is simply - dare we use this word in a professional context? - STUPID.
Careful readers of the DIGITIMES piece may recall this particular passage:
"Samsung has announced that it will reserve major portions of its S3 Plant's capacity for DRAM production, and is currently in the process of deploying the necessary equipment and materials. The South Korea-based electronics giant will likely begin manufacturing DRAM wafers by the end of 2014, and is expected to focus on PC DRAM. In the second half of 2015, Samsung's maximum S3 DRAM capacity is forecast to reach 60,000 per a month. Based on the worldwide DRAM industry's existing 1,050/month [ED. In thousands - equal to 1.05 million wafers per month] production rate, this would represent an increase of approximately 5% for the entire industry." [emphasis added]
If this report is to be believed, not only is Samsung increasing production of DRAM, they are increasing production in the most commodity like memory component. What's going on here? Does Samsung see something that DIGITIMES and DRAMeXCHANGE don't see? Let's take a deeper look.
Maybe there are good reasons the industry would be adding capacity. Let's hear from Micron's Kip Bedard in the recent Pacific Crest Conference when asked about DRAM supply dynamics.
"[…] one thing that hasn't been taken into account […] is that from here on out each node […will cost us] up to 10-15% of our wafer output. If you keep that in mind - we're running a little over a million wafers a year in DRAM [in Hiroshima, which is the one Micron fab that makes PC DRAM]- so conceivably [we would have to add] a little over 100 to 150k new wafer starts just to make up for what we are going to lose […] That's […] still misunderstood. […For example,] Inotera is running [about] 120k wafer starts a month and they could be as low as 95k by the time they finish [the 20nm] transition […] next year.
Seems pretty clear, right? Because of node shrinks the industry digs a hole for itself every year in terms of wafer output. If DRAM demand is increasing then at some point wafer output has to increase.
"[The Hynix and Samsung capacity increases] sounds reasonable to me because they need to replace wafers. Next year [the industry] must grow wafers to get to 20-25% capacity increase - [this year] the market has absorbed that [level of capacity increase] and prices have gone up and inventories have been stretched even thinner." Next year […] we're looking at low twenties [of supply growth] if we can make up some of this reduced [wafer] output. If we don't bit growth will be as low as - at any given company - as 12-15%. That's all you'd be able to accomplish […because] you're losing the wafer count and just using the node shift to make up for it." […] I think we'll see 20-25% [growth] because I do expect we'll make up that reduced wafer count."
Let's hear some more commentary from Pacific Crest, this time related to DRAM pricing trends:
"DRAM continues to be quite strong […and is] getting stronger as the PC space strengthens. […] DRAM ASP's [are] continuing to go up and will stay that way through next year. […On balance] ASP's […] have a pretty solid runway going into next year." […We'll] see improving margins next year."
In other parts of the DRAM commentary Bedard provides some pertinent detail regarding the key sub-markets.
"Mobile [is doing] the heavy lifting for demand and [is] staying strong [at] 45-50% a year demand growth."
"[We will see a] very big 40-45% YOY [growth] the next couple of years."
He then goes to say that Micron sees network DRAM being pulled along by the rapid growth of LTE networks worldwide and that trend, among others, will drive 30% growth in that sub-market.
How good is this market? Listen carefully:
"DDR4 is going [to be] an interesting transition. For the first time DDR [deployment] won't be led by the server space. [This is] a good problem to have."
Folks, in the DRAM business the above rates as what the journalism people call a "man bites dog" headline. Servers have been practically the sole consumer of DDR memory since it was invented. Low power DDR4 is going into phones and other mobile devices and Micron will be supplying a lot of that product at excellent margins. Oh yeah, the spot market (the subject of the DIGITIMES article above):
"Less and less product is going to [the] spot [market though there will be] some DRAM still going in because of bad yields. Contracts are stretching out."
Finally, let's hear some commentary on margins and Micron strategy going forward - relating to both NAND and DRAM:
"[We won't be commenting on our market share goals but] the business unit guys all have targets on where they think their optimized business mix might be. It might not be market share but more so margin profile. [...] that's what's going to drive where we put parts rather than pure market share." […Regarding capacity additions -] "Micron's strategy will be to get in a position where we can add small increments at a time - 5 to 10k [ DRAM and NAND wafers…] at a time because the risk will be that we don' have enough bits […] We want to keep moving up the gross margin profile […] without throttling [demand]. […This is more important in NAND] than DRAM because it is so elastic."
So let's wrap up with what we know of the current DRAM demand/supply environment.
- PC DRAM (counter to all expectations earlier in 2014) demand is stabilizing because the PC market is better than expected.
- PC DRAM is now less than a third of total industry DRAM production and will fall to only a fourth of the production next year.
- Samsung has decided to build a plant which, when it finally gets into full production at the end of next year, will be making PC DRAM ("primarily") and such an expansion will add roughly 5% to the MIPO's output, but in 2016, not in 2015 as the DIGITIMES article states, because of the long ramp in 2015 to achieve full output of the expansion.
- "[…] SK Hynix is not looking to make [announcements of] any official capacity expansions at the moment […]" - DIGITIMES Aug. 28
- Micron, which has only 25% of its current production in PC DRAM, is focusing on the non-commodity, high-growth, high-margin mobile, server, and networking markets - not on PCs.
- Margins are good and have strong upward momentum in the target markets and (for now, at any rate) in the PC space as well.
- The industry as a whole must add about 10% more new wafer output in order for supply to grow in the 20-25% YOY range.
To be clear, as I have said in previous articles on this topic, I firmly believe that long-term bets on human rationality are ALWAYS going to be losing bets. Just because it is "STUPID" (i.e. sub-optimal, irrational, etc. etc.) to do something doesn't mean that individuals and companies don't do it. Human folly and black swans will forever haunt us. In this specific case, however, in my humble opinion, there is no reason to believe that Samsung is intending to act in any way other than one that is fully congruent with their interests as the largest member of an oligopoly.
Samsung is on record as saying that they are interested in stable and steady markets. Personally - count me crazy - I doubt that Samsung's leadership got to their current positions because they were "stupid". Ergo, I do not believe Samsung is going to over-produce and thereby ruin their good (and expanding!) margins by over producing PC DRAM or any of the semi-conductor memories that they make. Especially when one considers that such a move only helps their ravenous consumer-product competition with lower prices on an absolutely essential component of their product.
One last comment regarding the DIGITIMES article. As Mr. Babbage stated so well, to assume that one can load bad information into a machine and expect a right answer is rather extraordinary. In this case the bad input was (at least) three very large and totally bogus assumptions about the memory industry and its participants.
First, they apparently assumed that the industry has not changed - ergo since it used to over-produce on a regular cycle, it will therefore always be on an over/under production cycle. To be fair, they are not alone in this feeling. No less than the eminent Jim Handy of Objective Analysis has expressed the same opinion. Paradigmatic shifts in industry and market conditions almost always challenge practitioners and consultants to keep up with current on the ground reality. Mark Twain's cat again comes to mind.
"We should be careful to get out of an experience only the wisdom that is in it and stop there lest we be like the cat that sits down on a hot stove lid. She will never sit down on a hot stove lid again and that is well but also she will never sit down on a cold one anymore."
Second, that (and this even though the article uses the term "oligopoly") members of an oligopoly would consciously act contrary to their express best interests. Does it strike you as odd that an article which refers to an industry as "oligopoly" would then go on blithely to conclude that the industry participants are going to act in a way that would weaken it without providing some semblance of explanation for this seemingly inexplicable behavior?
Third, that PC DRAM still dominates the memory market. As we have discussed PC DRAM is becoming less and less of MIPO's output as the years go on. And there could potentially be many more bad assumptions floating around that I haven't mentioned.
So, what's the right answer for investors in Micron? How about this. Stop paying attention to articles, headlines, analyst recommendations, etc. when they speculate that industry pricing and margins will be dropping due to over-expansion of capacity, except when they provide a convincing rationale for Micron, Samsung, or SK Hynix acting against their best interests.
What would such a rationale look like, you ask? Here's one example. Let's just add this as a final paragraph to the DIGITIMES piece to see how it works.
Inside sources at Samsung have reported that, in a complete shift of Samsung corporate governance policies, Dong-Soo Jun, President of the Memory Business at Samsung, has been given total independence to craft and execute a Semiconductor Segment production strategy. Mr. Jun is previously on record as saying that he considers market share gains the most important metric with which he would measure the success of the business.
If this last paragraph is added to the DIGITIMES piece we now have some information/analysis to take seriously. We still, as investors, need to do our due diligence to check out the information (which in this case is totally wrong), but at least the article is not wasting our time. When (and if!) we start seeing serious analysis that addresses this fundamental question of motivation to over-produce, we can tune back in. Until then, let's avoid this "confusion of the mind".
Long Micron, with the next article to address Micron short-and-intermediate-term profits and cash flow. (Hint, they are good and are going to get a lot better.)
Editor's Note: This article discusses one or more securities that do not trade on a major U.S. exchange. Please be aware of the risks associated with these stocks.