Saturday 3 March 2018

Samsung, and how to make money without a crystal ball

It is popular in markets to make definitive predictions about the future, and then position one's portfolio to benefit should those predictions come to fruition. Superficially, this approach seems to make sense - after all, future outcomes will be dictated by future events, and so surely you should try to predict what is going to happen, because what happens will drive future outcomes.

The problem is that the world is infinitely complex, and any one person's judgement about what will happen in the future, derived from an array of mental models applied to the current situation, will inevitably prove to be an excessively simplistic view of reality that is prone to error.

A much better approach, in my view, is to acknowledge this inherently irreducible uncertainty, and instead think probabilistically and try to find situations where prospective payoffs are asymmetric in nature - heads I win; and tails I don't lose very much. If you can find bets like that, you do not need to predict whether a heads or a tails will be flipped. You just need to know that the probabilities - given the payoffs on offer - are favourable.

Core to this process is considering in advance what is likely to happen if you are wrong. This is too seldom done by hubristic market participants who - while smart and insightful - frequently have excessive degrees of confidence in their ability to predict the future. They confidently predict that the coin will toss heads, and disregard what will happen if it actually ends up tossing tails.

An example of the wrong way to do it is what Kyle Bass has done in recent years. In early 2016, he confidently declared that China had experienced credit and real estate bubbles (something I somewhat agree with), and was accordingly imminently about to experience a banking and currency crisis (something I view as a possible outcome, but not that likely, for reasons I may discuss in a future blog article, nor a reason to sell all 'risk assets').

"This is going to happen in the next 12-18 months", he declared with absolute assurance. As a result, he sold all his risk assets and positioned his fund for that outcome, and that outcome alone. If heads is seemingly 100% assured, why worry about what will happen if you flip tails? Go all in. The problem was not so much his reasoning (although there were issues with that as well), but the fact that he did not consider what the consequences would be if he was wrong. As a result, he has sat out a huge 24 month rally in global markets, and missed some of the best buying opportunities offered this cycle (there were extraordinary bargains available in early 2016 on account of China fears).

An example of a better approach is to consider the shares of Samsung at present, and in particular, the preferred class, which trade at an approximate 15% discount to the ordinary shares (despite having slightly superior economic rights). At the preferred price, Samsung is trading at approximately 6x current earnings, and about 4.5x earnings excluding its tremendous 70tr won net cash hoard.

The reason it is trading so cheap - aside from the fact that South Korean stocks are relatively cheap overall - is that two-thirds of its earnings currently come from semiconductor components/memory (DRAM and NAND), which have experienced a surge in pricing over the past 18-24 months, and have driven an approximate doubling in Samsung's earnings. Forget smartphones - memory profits are Samsung's key profit driver at present. Historically, semiconductor prices have been extremely cyclical, and the market is sceptical earnings will be sustained at current levels. It is believed capacity increases will quickly follow, as has historically been the case, which are likely to drive an equally significant decline in memory prices over the next few years, likely halving Samsung's earnings.

There are two sides of the debate to be had here. On the bearish side, one can point to the historical pattern of repeated boom-bust cycles in memory prices, and the fact that prices are currently at elevated levels (with price increases having already sharply slowed), and industry heavyweights are increasing capex to boost supply. An argument was also made back in 2015 that industry consolidation was likely to result in improved industry profitability and reduced cyclicality, but that proved a false dawn, as memory prices collapsed into early 2016, taking down the share prices of producers alongside these dashed hopes. Markets have remembered, and have not been willing to extrapolate current industry super-profits ('fool me once, shame on you, fool me twice...).

One the more bullish side, however, consider the views of Sun Microsystem co-founder Bill Joy, who joined Water Street Capital a few years ago, and convinced them to not only close their Micron short (Micron is another major memory producer, that currently trades at 4-5x PE), which was premised on the above bear thesis, but also to go heavily long. His argument is something to the effect that (1) Moore's 'Law' is starting to break down, as the ability to increase transistor density has begun to reach its physical limits; this will make increasing supply more difficult/costly/time-consuming in the future; (2) demand is accelerating, due to a rapid pace of cloud data center buildouts and smartphone/device adoption, as well as various IoT trends; and (3) the global memory market has consolidated down to a few players, which may instill greater supply-side discipline.

He believes that the combination of #1-2, in particular, may mean that memory prices remain 'stronger for longer' as producers struggle to keep pace with growing demand for many years, potentially resulting in a sustained period of robust profitability for industry incumbents, whereas #3 may also mark a more permanent shift away from the temporary booms and rapid cyclical busts of yesteryear when supply eventually does catch up with demand. If he is right, high levels of profitability (and profit growth) may be sustained by the industry for many years from here. It would be a high-growth, high-profitability industry trading at 4x earnings.

Who is right? I have no idea. I'm not going to try and have better insights into the decay of Moore's Law and the possible structural implications on industry demand and supply dynamics than one of the co-founders of Sun Microsystems. However, it is also hard to completely dismiss the industry's long history of cyclicality, given the commoditised nature of memory components, coupled with the fact that demand and is at a relatively high ebb at present (given rapid cloud server buildouts). I'm agnostic, but am inclined to believe the reality is probably somewhere in between the two extremes (i.e. that memory prices stay higher than the bears currently believe for longer, but do eventually mean revert several years out, and experience another sharp cyclical downturn at some point).

The beauty of the situation with Samsung, however, is that you do not need to know the answer, because at it's current price, the market is already pricing in the mean-reversionary scenario with Kyle Bass-esque certitude. If memory prices collapse in the next few years, Samsung's earnings will likely approximately halve, pricing the preferred stock on about 9x earnings ex-cash - a valuation that is still far from expensive - particularly given the company's improving capital allocation (the company is now returning approximately 50% of its annual FCF in the form of dividends and buybacks), and in the meantime, they will continue to make boatloads of money. However, if Joy is right, Samsung will prove much much too cheap, and will be able to buy back boatloads of stock at extremely high earnings yields. Samsung is an easy double (at least) over the next couple of years if memory prices prove resilient and markets warm to Joy's view.

Heads you win; tails you don't lose very much (although it is still likely Samsung's share price will fall in the short term if news emerges that memory prices are rolling over). I like these odds. I currently have a 1.6% position in Samsung Preferred GDRs (SMSD LI).*


*I also own a small 10bp position in Micron, acquired recently at $41 (about a 4x PE of its current earnings run-rate), but I much prefer Samsung, as it was still making loads of money at the recent trough of the memory cycle, whereas Micron was unprofitable. Micron also has some debt, whereas Samsung is in a significant net cash position.

Samsung trailing 12-month EPS - still highly profitable at cyclical memory price lows, due to its smartphone and other businesses

Micron 10yr trailing 12-month EPS - loss making at historical cyclical memory price lows