General "Buck" Turgidson: Mr. President, we are rapidly approaching a moment of truth both for ourselves as human beings and for the life of our nation. Now, truth is not always a pleasant thing. But it is necessary now to make a choice, to choose between two admittedly regrettable, but nevertheless *distinguishable*, postwar environments: one where you got twenty million people killed, and the other where you got a hundred and fifty million people killed.
President Merkin Muffley: You're talking about mass murder, General, not war!
General "Buck" Turgidson: Mr. President, I'm not saying we wouldn't get our hair mussed. But I do say no more than ten to twenty million killed, tops. Uh, depending on the breaks.
"Dr. Strangelove Or: How I Learned to Stop Worrying and Love the Bomb"
For those of us who are old enough to remember Stanley Kubrick's classic send-up of the cold war, the quote above will probably bring a wry smile and a shake of the head. Hard to believe, but the doctrine Kubrick was skewering in the movie, Mutually Assured Destruction ("MAD"), was the glue that held the world together in those tense years between the end of WWII and the fall of the Wall in 1989. The idea, you may remember, was that in a world without the Bomb, war was inevitable. With the Bomb, however, war was unthinkable. All the players had too much to lose. The logic of MAD was ironclad. What could possibly go wrong?
So what has MAD got to do with the Memory Industry, you may ask? Stay with me here and we're going to examine that connection in depth, because the answer will help us better navigate the investment opportunities that the industry will provide over the balance of this decade.
As you will recall from previous articles in this series, our theory of the case is
- The industry is basically a four-member oligopoly I have called the Memory Industry Producer's Organization (MIPO) that has the power and the motivation to control the supply of NAND and DRAM memory.
- In a business environment that features strong secular demand for memory of all types but in particular NAND flash, limiting supply will result in relatively higher prices that offer MIPO members the opportunity to harvest much higher margins.
- MIPO members are on record that no new fab construction is contemplated in the foreseeable future; only incremental new wafer capacity will be added in the short term. Longer term, as the varying 3D NAND approaches that the suppliers have taken prove out, 50% more wafer capacity will have to be added simply to accommodate 3D tech node tool expansion.
- Memory supply will thus increase only to the extent that tech node transitions are productive.
So here we have it, our brave and wonderful new world, where the long-suffering solid-state memory suppliers stride confidently toward the sunlight highlands of ever increasing revenues and profits. What could possibly go wrong?
Well, if you're a memory industry analyst, a lot. At least you would think so, as with every new analyst conference event, the opening question is always some version of: "What's happening with supply and pricing out there? Is it REALLY different this time?" So they are nervous, and so are investors, who know, deep down, that nothing this good can last. I mean, if there is one thing we battle scarred investors know, it's this:
IF IT SEEMS TOO GOOD TO BE TRUE, IT IS! (too good to be true)
And it's not as if the stakes could be any higher, right? If you are an OEM, an OED, whatever, you have a deep interest in the outcome of this epoch discontinuity that the entire IT industry seems to be facing. We're talking about memory here, folks. Metaphorically, if logic is the internal combustion engine, then memory is the oil of our new, smart, connected lives. There is really no substitute. Right? Right!
IoT without boatloads of NAND? Forget it. In every IT device category, product designers and planners are used to having this spigot of ever flowing, ever cheaper IT oil available for the next great thing. What's the price going to be next year? Two years? Five years? No problem, just track Moore's law from today through the period of product launch. Easy. That is the way the world has always worked in product design for IT. And now there is this possibility that the product that is being designed may either, A., not be economic at all because of the cost of the memory it requires, or B., may be constrained in supply because I can't get enough memory to make all the shiny new things I want to make?
So, yeah, this issue is a little bit bigger than just us MIPO company investors and the Industry. That's why the title "Memory OPEC" is apropos, because it captures the emotional and business import of the issue. The whole IT world is watching, and 2014 is just like 1973 when the first embargoes and the long lines at gas stations developed. The outcome of this issue affects a broad set of powerful entities and interests, and almost all of these interests would like to see this turn out very badly for the MIPO companies. They want cheap memory - as much as they want, whenever they want it - and they really don't care if memory suppliers make a dime, just so long as enough of the suppliers stay in business to keep supplying the oil that makes the world run.
Unlike the OPEC situation, the memory industry is not a cartel with the ability to formally plan and coordinate production. MIPO is an oligopoly and, as we have discussed, while there are abundant communication mechanisms for these inter-dependent companies to utilize to signal their production intentions, they can't be transparent about their plans without running afoul of ferocious governmental entities. This is, ultimately, a game of trust - trust in the fundamental rationality of companies and their leaders. This seems so simple. Why would Samsung (OTC:SSNLF), Micron (NASDAQ:MU), SanDisk (SNDK), and Hynix (OTC:HXSCF) over produce when restraining supply will improve their bottom line? As my grandmother used to say, why on God's green earth would you do that? Why? Seems unthinkable, doesn't it? So let's think about it.
Here's why. Just like they did in Buck Turgidson's day, we will turn to game theory to explain it. Game theory, as you probably know, was invented by John von Neumann and more fully developed by a mathematician named John Nash in an attempt to better explain decision options in complex environments. (Some of you may recall the movie made about Nash's life - "A Beautiful Mind".) Vastly over simplifying it, the goal of all oligopolies/cartels is to maintain stability. As long as the situation doesn't radically change, all members in an oligopoly benefit by nothing (much) changing - they seek equilibrium. In such a multi-player game, Nash defined a state of equilibrium as follows: no player, with full knowledge of the other players' strategy, would benefit by changing theirs.
The problem, though, is that this so-called "Nash Equilibrium" is very hard to achieve. Turns out there is always a General Jack Ripper lurking in the background with the means and the motivation to launch that rogue bomber wing attack that will trigger massive retaliation. By definition, MIPO's success means that some amount of marginal demand is unsatisfied at the (higher) prices that MIPO oligopoly pricing wants to offer. That in turn means that there is always a marginal benefit for a member to cheat by offering more product to satisfy that demand. There is, is other words, built in bias for oligopolies/cartels to fail. On the way to failing, however, some cartels take their own sweet time going about it. Check out OPEC, for example. Let's consider some scenarios that will face MIPO and see what we can learn.
Let's start out by clarifying what success and failure looks like. We're going to use some game theory terminology to describe a scenario and company actions within that scenario. Let's start with success. What will the next few years look like from a company performance perspective if MIPO solidarity holds and capacity is restrained from exceeding supply? Let's use Micron and SanDisk and offer a "success" pro forma for them as a way of illustrating the overall gains. We will use a four square grid. Outcomes will be one axis - either stable, or unstable. Intentions are the other axis - they can be malign or benign.
Scenario #1 - Benign Stability. Currently the industry exists in what is called in game theory a weak equilibrium. Samsung is building out the Xian NAND plant, Toshiba (OTCPK:TOSYY)/SanDisk is building out the FAB 5 NAND plant, Micron has transitioned 100 Kwpm in Singapore from DRAM to NAND, and SK Hynix is building a new DRAM plant in Korea. No additional wafer capacity beyond this has been announced at this point, and all MIPO members except Samsung have publicly stated that they have no intention of building anything other than incremental additional new wafer capacity in the foreseeable future. Demand in both the DRAM and NAND technologies is exceeding supply. This is success.
Scenario # 2 - Malign Stability - "Crimea" - In this scenario, one of the MIPO companies announces a new Fab that may change the relative market share positions within the industry but does not promise to overwhelm demand with excessive capacity. The rational game theory response by the other MIPO members? Allow the preemptive effort to result in small share gains for that member but overall profitability of the business is not affected. The key to this decision is a signal from the offending company that this is a "one-off" that won't be repeated. Failing that, countermeasures along the lines of scenario 3 may be required.
Scenario #3 - Malign Instability - "Munich" - Like Scenario #2, one MIPO company (Company 1) preemptively announces a big new FAB, but unlike #2, this action promises to throw the industry into oversupply. The other MIPO companies react to that announcement by announcing their intention to bring on smaller, incremental capacity, which in cumulative effect will result in significant industry overcapacity. What could be seen as "enforcement action" by the other companies responding to the actions of the first will result in a massive amount of "unqualified" memory coming to market from Company 1. The new product will be particularly hard hit from a price perspective and will degrade the investment decision rationale made by Company 1 that underlay its decision to build. Their actions, however, will ensure that Company 1 is much less likely to pursue such a unilateral approach in the future, thereby eventually returning the industry to profitability. In any case, this is the only rational response to this threat. Any other response will encourage the offending company to continue unilateral overbuilding.
Scenario #4 - Benign Instability 1 - Demand Drought - This scenario is "benign" because no overt action of a MIPO enterprise to build additional wafer capacity is the stressor to MIPO solidarity. Here, global economic conditions or some other external factor lead to a dramatic decrease in demand for memory, thereby directly causing an oversupply situation. Rational strategy in this scenario is for MIPO members to shutter fab operations or build inventory that will be withheld from the market. This scenario is particularly challenging from a legal perspective because the companies must be very careful not to be seen as colluding for purposes of market manipulation.
Scenario #5 - Benign Instability 2 - "Technology Breakout" - Here, an unanticipated oversupply environment is triggered by a process breakthrough by one MIPO supplier that proves much more productive than anticipated by other MIPO members, thereby leading to dramatically lower costs. In this case, it may well be in the best interest of the MIPO member producing the breakthrough to deliver the "excess" product to market because it is uniquely capable of thriving in a lower price environment. If this happens, the rational response from the other MIPO members is to take no action in response to the breakout as long as the breakout supplier refrains from taking full advantage of the situation. "Restraint" is the watchword in this situation.
Scenario #6 - Benign Instability 3 - Demand Explosion - Once again, no overt action by a MIPO supplier takes place, but some external factor drives a strong and unexpectedly large increase in memory demand (e.g., some killer app drives an IoT use case that blows up rapidly). In this case, memory shortages drive such rapidly rising memory prices that outside entities are tempted to enter the business. MIPO strategy in this scenario must be to react rapidly and decisively to build new capacity. The new capacity added by each member must be congruent with that member's market share for this scenario to remain benign.
As these scenarios illustrate, there is a lot that can go wrong with our wonderful oligopoly and the imagined treasures that investors are anticipating. What are the odds on any one of these happening? Your guess is as good as mine. Personally, I don't see scenario 4, demand drought, happening any time in the foreseeable future. The global economic picture is muddled, but generally good enough to drive continued growth in the major economies of the world. The "malign" scenarios, 2 and 3, are also unlikely in my view, though one always has to wonder about Samsung launching a "Crimea" situation as in 2.
My personal favorite in scenario 6. NAND is a textbook case of a technology driving use cases that radically change the environment and the economics of an industry. In other words, we have a "discontinuity" developing before our eyes and the only question is when does it REALLY head north on the hockey stick? My gut is telling me that there are two areas, one consumer and one enterprise, that will drive an explosion of demand and do it in the 2015 timeframe. If this is correct then MIPO will be stressed to the max to bring on sufficient capacity. Whatever the case, initially (2015, 2016) gross margins and profits will climb to heretofore-unseen levels and investors will be delighted. Customers will have a different view however, and therein lay the seeds of MIPO destruction as any number of ugly responses to the NAND shortages start to play out. Think, for example, here about Apple (NASDAQ:AAPL) deciding to get into the NAND business. $5B is lunch money to Tim Cook.
So when, not if, scenario 6 comes about, what this mean for our two favorite companies, Micron and SanDisk? It's better news for SanDisk than Micron, but 2015 and 2016 are going to be great years by any standard. SanDisk's Target Financial Model (TFM) metrics? Forget about them. Instead of revenue growth at 15%, try 25-30%. Instead of 50% gross margins, let's think about 65%. Micron, given the strong growth in the DRAM side, will also do very well, just not as good as SanDisk in the NAND side of the business.
My old boss Ken Olsen used to say that of all the challenges an entrepreneur/CEO faces, none is so daunting as success. Micron and SanDisk, indeed all of MIPO, will have a chance to test his theory - it's just a matter of time. Over the course of the next two to three years, investors will be glad they get to take the test. For you worriers out there? Relax. Mutual Assured Destruction is still an option for MIPO CEOs and General Turgidson is always ready and happy to offer them his sage advice when the mushroom clouds start blooming.
Long SanDisk, long Micron.
Disclosure: I am long MU. 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.
Additional disclosure: Also long SanDisk.