Micron: Cycles and Risk

I've been reading The Most Important Thing Illuminated by Howard Marks, a legendary value investor. He spends a lot of time talking about cycles and risk, two very timely subjects for the Micron Technology (MU) investor.


I think it's essential to remember that just about everything is cyclical. There's little I'm certain of, but these things are true: Cycles always prevail eventually. Nothing does in one direction forever. Trees don't grow to the sky. Few things go to zero. And there's little that's as dangerous for investor health as an insistence on extrapolating today's events into the future - Howard Marks, The Realist's Creed

The long term argument for MU is strong. The memory industry is experiencing long term, secular growth (via a variety of new use cases like AI, 5G, IoT, automotive, datacenter, and surveillance). More data will need more memory, both for processing (DRAM) and for storage (NAND). The market is controlled by three large players (Samsung, SK Hynix and Micron), who presumably have strong incentives to informally coordinate on preventing oversupply.

Management hasn't explicitly said that the cycle has been abolished, but not-so-subtly hints at these trends by branding themselves "The New Micron"; in other words, "this time it's different." Or is it?

These secular growth drivers are no doubt important, but it is unclear how they differ in magnitude from growth drivers in the past (which proved to be cyclical). The same arguments for "this time it's different" were likely being made with: the personal computing boom, the internet boom, notebook computing and the rise of the smartphone. Are current trends really as game changing as widespread adoption of the internet? Or the smartphone? Probably not.

I think it's safe to say that "cycles will prevail eventually" - and that we are already (and have always been) in one. The question is: where in the cycle are we and how much risk does that entail?


Marks defines risk as "first and foremost - the likelihood of losing money" - not volatility. However, he also concludes that "looked at prospectively, much of risk is subjective, hidden and unquantifiable" and that they key to understanding risk is that "it's largely a matter of opinion." Risk is even hard to quantify retrospectively:

It's hard to be definitive about risk, even after the fact. You can see that one investor lost less than another in bad times and conclude that that investor took less risk. Or you can note that one investment declined more than another in a given environment and thus say it was riskier. Are these statements necessarily accurate?...The performance of your portfolio under the one scenario that unfolds says nothing about how it would have fared under the many "alternative histories" that were possible - Howard Marks, The Most Important Thing Illuminated

These answers are insightful, but also frustrating in their ambiguity (as most truths often are). Pretty much the only concrete guidance he gives on risk is:

At bottom, the riskiest thing is overpaying for an asset (regardless of its quality), and the best way to reduce risk is by paying a price that’s irrationally low (ditto). A low price provides a “margin of safety,” and that’s what risk-controlled investing is all about - Howard Marks, Risk Revisted Again

So where does that leave Micron?

The Pendulum

Marks makes an analogy between cyclical investment markets and a pendulum swinging back and forth. This visualization helps tie together cyclicality and the risk of overpaying.

The pendulum swings:

  • between euphoria and depression,
  • between celebrating positive developments and obsessing over negatives, and thus
  • between overpriced and underpriced

I can't predict when the memory cycle will end or where it will eventually bottom (or even if it has already bottomed), but a few things seem clear:

  • On the spectrum between "euphoria and depression", investors are clearly depressed. MU is down nearly 50% from its recent highs. DRAM prices are expected to fall 20% in 1H2019. MU has guided revenue down 18% YoY; Samsung and SK Hynix have also warned about similar declines. Fears about the trade war with China, slowing global economic growth, slowing hyperscale capex spending and rising interest rates are rampant.
  • Investors are obsessing over the negatives and ignoring many positive developments, such as: the consolidation of the industry into a handful of players, who have cut capex spending more rapidly than in any other cyclical downturn; MU's significant structural cost savings and stronger balance sheet; and, perhaps most importantly, increased IP enforcement that should slow China's rise in semiconductor manufacturing, which in my opinion is the biggest long term threat to Micron's growth prospects.

Based on these factors alone, it seems reasonable to conclude that the pendulum has already swung towards underpriced.


Unfortunately, knowing that Micron is likely to be underpriced based on cyclical, primarily psychological factors tells us nothing about how far the pendulum has yet to swing. It could continue to decline. Who knows what the bottom will be?

If you accept that Micron's long term business prospects are strong, what is an "irrationally low" price that provides a "margin of safety"?

Ben Graham's Cyclical Valuation

Ben Graham, father of value investing, argued that for cyclical stocks you should take the average of the last ten years of earnings, which is the average length of a business cycle. By this metric, MU's "cyclically adjusted" EPS is $2.228, and at current prices ($35.76) it trades at a 16x TTM multiple. Not expensive, but also not "irrationally low."

Price to Sales Cyclical Valuation

I tried to take the analysis one step further by looking at the price-to-sales (PS) ratio from Jan 1, 2013 to Jan 16, 2019. The PS is less "noisy" than the PE ratio (in that it doesn't turn negative during downcycles). This time period was chosen because it is modern enough to be relevant, yet also excludes the excesses of the dotcom bubble. It also includes five memory downcycles, where a downcycle is defined as 2+ consecutive quarters of negative QoQ growth.

One might argue that the PS ratio says nothing about margins, and that earnings are ultimately what matter. This may be true, but I think it is reasonable to assume that margins will remain similar. If anything, this might prove to be too conservative, as the industry has become more concentrated into an oligopoly; even if the constituents do not overtly collude, they all clearly telegraph their expectations in quarterly earnings calls, and thus are better able to control margins by coordinating supply and demand.

This trend seems to be supported in the data (which also probably reflects Micron's improving cost structure):

In particular, it is interesting to note that the "trough" operating margin has increased with each cycle.

The median price-to-sales (PS) ratio for MU for this time period is 1.597. By this metric, MU's "cyclically adjusted" price per share is $42.84, and at current prices MU shares are 20% undervalued.

But What If It Gets Worse?

I also looked at each memory downcycle individually, to see how depressed the PS ratio has been during similar periods:

Time Period No. of Quarters Cumulative QoQ Decline Median PS Ratio
12/1/2006 - 5/31/2007 2 -16.05% 1.6355
6/30/2008 - 2/28/2009 3 -35.68% 0.5356
3/1/2011 - 2/29/2012 4 -11.4% 0.8644
5/31/2012 - 11/30/2012 2 -16.19% 0.7726
12/1/2014 - 5/31/2016 6 -43.57% 1.2521

The average PS ratio during all of these memory downturns is 1.01. Applying this multiple to different standard deviations of analyst revenue estimates for 2019 (and assuming the share buyback is successful), we arrive at the following share prices:

CY 2019 Analyst Revenue Estimate PS Ratio Shares Out. Share Price
$31.148B (+2 std) 1.01 936MM $33.61
$28.754B (+1 std) 1.01 936MM $31.027
$26.36B (mean) 1.01 936MM $28.44
$23.966B (-1 std) 1.01 936MM $25.86
$21.572B (-2 std) 1.01 936MM $23.277

If you exclude the Great Recession from the calculation of mean PS ratios during downturns, which seems reasonable,  you get a PS ratio of 1.13, resulting in the following scenarios:

CY 2019 Analyst Revenue Estimate PS Ratio Shares Out. Share Price
$31.148B (+2 std) 1.13 936MM $37.603
$28.754B (+1 std) 1.13 936MM $34.71
$26.36B (mean) 1.13 936MM $31.82
$23.966B (-1 std) 1.13 936MM $28.932
$21.572B (-2 std) 1.13 936MM $26.04

Again, it is important to note that these are estimates for trough valuations. On a cyclically adjusted basis, and after the pendulum starts swinging in the other direction, the valuation premium increases.


I'd be fairly comfortable accumulating additional shares of MU around $30 (which it was trading at just a few weeks ago). The valuation metrics above seem conservative, and the psychological factors outlined above also seem to indicate that the cyclical pendulum has swung in the underpriced direction. The risk of overpaying at this level, especially if you are a long term investor, seems low.

With that said, I'll leave you with an important caveat, which may be painfully relevant to anyone who has bought MU recently:

...don't expect immediate success. In fact, you'll often find that you've bought in the midst of a decline that continues. Pretty soon you'll be looking at losses...
This makes it very difficult to hold, and to buy more at lower prices (which investors call "averaging down"), especially if the decline proves to be extensive. If you liked it at 60, you should like it more at 50...and much more at 40 and 30.
But it's not that easy. No one's comfortable with losses, and eventually any human will wonder, "Maybe it's the not me who's right. Maybe it's the market." The danger is maximized when they start to think, "It's down so much, I'd better get out before it goes to zero." That's the kind of thinking that makes bottoms...and causes people to sell there - Howard Marks, The Most Important Thing Illuminated

Disclosure: I am long MU.

Show Comments