March 15, 2020 - 12:01am EST by
2020 2021
Price: 43.99 EPS 0 0
Shares Out. (in M): 1,129 P/E 0 0
Market Cap (in $M): 48,536 P/FCF 0 0
Net Debt (in $M): -1,427 EBIT 0 0
TEV ($): 47,109 TEV/EBIT 0 0

Sign up for free guest access to view investment idea with a 45 days delay.

  • Semiconductor
  • Semis
  • Technology
  • Cyclical


I own shares of Micron (MU US).  Micron is a member of the DRAM semis oligopoly (the other 2 primary competitors being Samsung and SK Hynix). It also has a smaller, much less profitable NAND business. Micron has been written-up both long and short for VIC before. Currently, I think the investment case for Micron boils down to some short-term pain ahead in operating results as covid-19 pushes out the cyclical bottom of the memory cycle further into the future. Perhaps anticipating that pain, Micron shares have declined 28% in the past 3 weeks. I am of the view that the memory industry is a better industry than many believe, that Micron has the capacity to weather the storm, and that Micron is undervalued relative to its mid-cycle earnings power.


Micron is a long-term attractive business to own on a secular basis.

The total amount of memory consumed globally will continue to grow due to the data-intensive nature of modern technology.  End-markets can shift in their relative importance (e.g. smartphones are not as important as they were in the past; data-centres are becoming more important, etc) but the overall volume of memory will grow.  See some charts below. DRAM bit demand will grow 14%+ per year on average for the foreseeable future. It will continue to be lumpy, but the average growth rate over time should be significant. 



The other attractive feature of the market is its oligopolistic nature with extremely high capital barriers to entry.  Over time (and especially after the industry shrunk from 4 major key competitors to 3), this has allowed the remaining participants to enjoy relatively high margins and high RoE. 




Micron historically badly lagged both Samsung and SK Hynix in its DRAM profitability, but recently Micron managed to get ahead of SK Hynix and is gradually closing the gap vs Samsung. 



Similarly, on a % ROE basis, Micron historically lagged both SK Hynix and Samsung but recently got ahead of them.  Impressively, the ROE improvement was achieved at a time where Micron moved from a net debt position of $5B in 2016 and 2017 to a net cash position of $2B currently.  While the industry still has boom-and-bust cyclical properties, this data demonstrates that the industry quality has been getting better, most likely due to the increasing degree of oligopoly and more diverse drivers of end demand.  



In my base case, I assume a Micron ROE of 19% at exit, even though average industry ROE over the past 3 years is closer to 25% and Micron’s average ROE since the closing of the Elpida transaction in 2013 is 24%. I believe that this gives my base case assumptions some margin of conservatism.  


Good position to survive a recession and thrive as memory rebounds. 

If you take a look at Micron’s quarterly financial results for the past 6 quarters, it would be perfectly understandable if your first reaction was surprise as to why anyone would be interested in owning a company like this, with a current streak of 6 quarters of declining consolidated EBIT in a row.  If you then look at the stock chart and see how Micron’s share price ran from $32 a year ago to $60 just prior to the emergence of covid-19, you’d probably be even more surprised - why is the share price rising if EBIT is still declining, you might ask. Unfortunately, I have been learning that semis share prices never follow fundamentals in lock-step and apparently tend to anticipate both first and second order derivatives.  Prior to covid-19, the memory market looked like it was on the cusp of a very profitable period in calendar year 2020. The memory market peaked in the summer of 2018, with significant declines in the spot and contract prices for DRAM and NAND memory. NAND prices started to rebound in the fall of 2019 and DRAM spot prices started to go up a bit later but still earlier than anyone had predicted. This current memory downturn was set to go down in recent history as perhaps the most benign one we have seen yet, with trough margins (especially in DRAM) much higher than before.  For instance, on an EBIT basis, DRAM makers lost money at the trough of the prior downturn (2016) but this time, DRAM EBIT margins were expected to trough at 15-25%. Micron was on track to come out of this cycle without a quarter of negative FCF. Prior to covid-19, Micron and its peers had essentially called the coming calendar Q1 2020 quarter as the trough of the cycle. I think covid-19 is likely to derail this memory recovery. Covid-19 will undoubtedly damage employment, consumption and corporate investment across multiple sectors of the global economy, affecting demand across all memory end-markets (smartphone; PC/notebook; server; industrial).  Memory stocks are very high-beta and thrive on consumer and business confidence. The memory business model is capital intensive, and has extremely high fixed costs. If bit demand declines and suddenly becomes less price-elastic, margins can drop very quickly. The current consensus EPS numbers for the memory makers are probably too high and can be safely ignored. Analysts cite some reasons to be bullish for memory from a demand perspective due to reports that server demand for memory is currently surging as there is a rush to install more capacity as social distancing (due to covid-19) means more WFH and puts a strain on internet infrastructure. This might be true but the overall impact of that and how long that lasts is uncertain, and I would expect the negative impact of covid-19 to dominate.  Thus, my base case assumptions is that covid-19 extends the memory downcycle for longer and pushes DRAM EBIT margins into the red (see the DRAM EBIT margin chart earlier in this memo). I assume the bottom of the memory cycle is pushed forward by a year, from Feb 2020 to Feb 2021. While profits drop, I also expect inventory to build again. Micron’s annualized DOI is currently at 119 and peaked at 151 in May 2019; this time, I expect it to peak closer to 190 in May/August 2020, before gradually recovering. Micron also has an $8B annualized capex programme. As a result, I think Micron will burn $4B of cash over the next 6 quarters, cumulatively. Luckily, Micron currently has $7.6B of cash and equivalents, $6.2B of debt, and total liquidity of $10B+. The company also has a $7B remaining share repurchase authorization, though given the upcoming cash burn, I do not expect them to tap it entirely in the short-term and put the company into a net debt position of greater than $2B.   The important point here is that Micron clearly has the financial strength to weather the extended downturn, and will probably avoid issuing equity, too.


Shares undervalued on a mid-cycle basis. 

Coming out of the down-cycle, I expect Micron to reach mid-cycle in fiscal year ending August 2023. At that point, I expect Micron to generate an 19% ROE, $6.10 of EPS and $4B of FCFE. The company typically converts c.60% of net income to FCFE due to the working capital drag and capex that runs significantly above D&A.  Since acquiring Elpida in 2013, Micron has grown its revenue by a 7% CAGR and its EBIT by a 19% CAGR. Going forward, I think a more sustainable growth-rate of mid-cycle EBIT is in the low-teens. I think the combined quality and growth prospects of this business are no lower than the average quality in the Russell 2000, and accordingly, Micron should deserve a healthy multiple of earnings. The current share price of $42.99 (and fully-diluted market cap of $48.5B) does not entirely reflect that.  I believe fair value for Micron shares should be above $60. 


As an aside, it is interesting to me that semiconductors as an asset class have this interesting tendency of being very cyclical yet they still somehow manage to allocate capital well and generate respectable ROEs and growth metrics. As a result, they manage to outperform some other indices like the S&P500 and R2K over long time periods.  Some segments of the semiconductor market also have the tendency of being frequently undervalued due to that cyclicality. As long as short-term indicators like spot chip prices are moving in the wrong direction and the speed of the weekly chip price declines is not moderating, sometimes it seems companies like Micron just have no multiple support.  The stylized example below illustrates how I think the market frequently perceives and values Micron relative to a company with a more stable, more predictable, lower-volatility stream of cash flows.



Reasonable downside protection.

Besides Micron’s balance sheet giving it the ability to weather an extended industry downturn, Micron also has a more valuation-focused source of downside.  As an asset-heavy business, investors and analysts frequently think of Micron’s valuation in terms of P/TBV. Micron’s current P/TBV multiple is 1.4x (with a tangible book value per diluted share of $30.60). Since Micron’s merger with Elpida, its average P/TBV is around 1.8x, with a trough of 0.9x, during the 2015/16 semi downturn and 1.1x during the earlier half of the current downturn in late 2018 / early 2019.   Using 1x tangible book gives c.29% of short-term downside from the current share price.    


Newsflow around China will continue to be a source of volatility, and China may be a long-term risk.

- Micron shares are often a punch-bowl proxy for newsflow relating to US/China trade and Huawei.  Huawei accounts for less than 15% of Micron’s revenue, and Micron continues to ship some products to Huawei.  An escalation of the trade war or toughening of the measures against Huawei will damage both sentiment and Micron’s actual fundamentals / EPS. 

- Longer-term, China’s sovereign semiconductor self-sufficiency plan (‘Made in China 2030’) can be a threat to the existing memory oligopoly of Micron, Samsung and SK Hynix.  China is already key to the industry, in many respects. From a production point of view, 10-15% of global semis production capacity is physically located in China; the proportion of memory capacity that is in China is thought to be lower.  E.g. of Micron’s 16 fabs, only 1 is located in China (most are in Taiwan, Japan and Singapore). Then, in one way or another, China ‘consumes’ 40-50% of the global semis produces, e.g. a large portion of semis produced globally end up being handled or put together into final products (smartphones; serves; notebooks) by OSATs or IDMs with operations in China. 



Micron’s shares have 30% potential downside as operating results worsen due to covid-19 elongating the current memory downcycle, but I see significant medium-term upside (70%-130%), with mid-cycle EPS of $6+ that is attractive relative to the current share price of $42.99.  Memory industry's highly volatile results and high capex-intensity tend to discourage investors, but I think the industry is under-appreciated as an oligopoly that delivers reasonable FCF, ROE and EBIT growth through the cycle and has substantial secular tailwinds to demand.  


I do not hold a position with the issuer such as employment, directorship, or consultancy.
I and/or others I advise hold a material investment in the issuer's securities.


Quarterly results of Micron and competitors (Samsung, SK Hynix)

Spot / contract prices for DRAM and NAND

    show   sort by    
      Back to top