MU - Micron Technology Inc.

[samwise]

I am building a personal position in MU via out of the money LEAPs since it is currently an opportunity rich environment.  I would plan to exercise and hold the stock longer-term if thesis is correct.  I think the fair value on this company is more than $100 BN enterprise value today and growing.

 

Elements of my thesis below... some of which already has been mentioned in the thread.

 

- very healthy long-term returns on tangible capital

- long-term consolidation helps to enhance investment and supply dynamics in memory to protect margins

- geopolitically, ability to manufacture critical technology components only growing in importance; Micron is last major co. standing in memory with a North American nexus (even if they produce much outside the U.S.)

- management is for real. very impressed with the recent strategy presentations and the business trajectory over last five year

- huge scale requirements in a cyclical industry leading to protection against new entrants

- good capital allocation: they do acquisitions in the down cycle to further consolidate the industry, have been investing sufficiently internally in R&D to grow future value, and have been smartly retaining earnings in the up-cycles to de-lever

- ability to add value on top of what was previously an extremely tough commodity business

- overlooked by the market because it is hardware, not software. software is all the rage.  Venture capitalists don't want to fund a hardware competitor because it's boring.

- long-term, absolutely fundamental to several very long-term secular trends that look to accelerate due to COVID-19 (gaming, cloud, etc.)

- likelihood of being mispriced: screens poorly on simple metrics due to recent cyclical lows in memory pricing (P/E, etc.)

- no net debt, lower risk during period of economic uncertainty

 

+ one more reason: They claim to be on the leading process node 1z now. In a commodity market, they have now caught up with the low cost producer. (This also answers Spek’s question on their tech above).

 

Some questions on your thesis:

What strike do you buy? By a rough estimate they produced an average of $6 earnings in the last 3 years. So for a cyclical company do you apply a 10x multiple or 16x?

 

And while there will be bit growth, do you expect much revenue growth through cycles?

 

Any views on what happens with the rumoured end of Moore’s law?

[HJ]

And while there will be bit growth, do you expect much revenue growth through cycles?

 

Any views on what happens with the rumoured end of Moore’s law?

 

Those seems to be the most pertinent question and are related.  I've seen numbers that show DRAM industry having virtually no growth in revenue terms from the mid 90's until 2013-2014.  Basically all the technological advances benefitted the users rather than the producers. 

 

Until recently it's been a miserable industry to be in, and now there is the threat of potential government backed uneconomic entrants.  How do we know the world is wiling to pay more for memories consistently year in and year out?

[Spekulatius]

RAM is a commodity. That’s all one needs to know.

[ValuePadawan]

RAM is a commodity. That’s all one needs to know.

 

I'm not sure it is a commodity in the same way cotton, copper and cattle are. I think it will deliver greater returns than a lumber company or soybean farmer for the simple reason that only 3 companies can make DRAM chips. It takes large amounts of capital for the fab in addition to lots of patents, IP and talented individuals to manufacture the chips. If you think my logic is flawed I'd be thrilled if you could explain where I've made a mistake. The moat in my eyes is the IP and technical skill to make more and more advanced chips with more storage capability and use less and less energy per bit. As long as demand keeps outpacing supply every few years these companies will hit the lottery every 4 or 5 years by being able to increase ASP's dramatically. Not only that but in the nadir of this current cycle the DRAM companies have managed to stay FCF positive this last cycle showing the improvement in profitability in the business. All disagreements are welcome.

[Guest eatliftinvestgolf]

Sorry for the delayed response.  I don’t think my personal LEAPS strike is so important because it needs to be understood within the context of position sizing and overall portfolio exposure.  I went far out of the money for furthest possible duration at purchase.   

 

On Moore’s law:  I believe the CEO at face value when he goes on the record to say that the capital intensity to remain in this industry has increased.  I don’t have a hard view on the scientific constraints. 

 

I don’t think memory will be so commodity like in the future. I do think that application specific memory applications will create better pricing power.  I don’t think this is a short term benefit, but more of a long term trajectory. In some ways, this is more a market structure bet than a bet on pricing power. Money is a commodity but banks with large market share have Meaningful pricing power !

 

On revenue growth, I think something has fundamentally changed in the industry that will lead to overall revenue growth at Micron.  Let’s check back in five years!

[shamelesscloner]

Is there a case to be made for the big cloud companies (Amazon, Google, Microsoft) bringing memory manufacturing in-house since it's such a large capital expense for data centers? How would that impact the long-term investment case for Micron?

[Jurgis]

I think more likely scenario is that China gets into memory manufacturing at some point. Now whether rest of the world will buy chips/memory/electronics made in China... ask Apple I guess.  8)

[mwtorock]

Is there a case to be made for the big cloud companies (Amazon, Google, Microsoft) bringing memory manufacturing in-house since it's such a large capital expense for data centers? How would that impact the long-term investment case for Micron?

 

If it make economic sense, I am sure the Apples, HPs, DELLs would have done that long time ago for all computer parts.

 

CapEx for manufacturing is not among the best things to show investors these days. No one is investing to build memory factories other than Chinese govt for non economic reasons.

[HJ]

A technical question for those of you technically oriented, Apple makes a big deal out of its M1 chip which is a SOC and has a  "Unified Memory".  How does it affect the prospect of independent memory makers?  Do these SOC obviate the need for third party memory makers for these chips?

[nwoodman]

A technical question for those of you technically oriented, Apple makes a big deal out of its M1 chip which is a SOC and has a  "Unified Memory".  How does it affect the prospect of independent memory makers?  Do these SOC obviate the need for third party memory makers for these chips?

 

Good question and one that has been on my mind too.  The iFixit teardown offered the following:

 

https://www.ifixit.com/News/46884/m1-macbook-teardowns-something-old-something-new

 

Next to the shiny silver M1 chip on each board you’ll notice two small black rectangles. Those are the new “integrated” memory chips: 8 GB (2x 4 GB) of SK hynix LPDDR4X memory. Apple calls this UMA, or Unified Memory Architecture. If it looks familiar, it might be because you’ve seen one of our recent iPad teardowns. It’s no surprise that Apple copied some of its own homework here. By baking RAM into the M1 package, each part of M1 (CPU, GPU, Neural Engine, etc) can access the same memory pool without having to copy or cache the data in more than one place.

 

So while not Micron it is at least one of the big three - SK Hynix

 

 

Cheers

 

nwoodman

 

 

[HJ]

A technical question for those of you technically oriented, Apple makes a big deal out of its M1 chip which is a SOC and has a  "Unified Memory".  How does it affect the prospect of independent memory makers?  Do these SOC obviate the need for third party memory makers for these chips?

 

Good question and one that has been on my mind too.  The iFixit teardown offered the following:

 

https://www.ifixit.com/News/46884/m1-macbook-teardowns-something-old-something-new

 

Next to the shiny silver M1 chip on each board you’ll notice two small black rectangles. Those are the new “integrated” memory chips: 8 GB (2x 4 GB) of SK hynix LPDDR4X memory. Apple calls this UMA, or Unified Memory Architecture. If it looks familiar, it might be because you’ve seen one of our recent iPad teardowns. It’s no surprise that Apple copied some of its own homework here. By baking RAM into the M1 package, each part of M1 (CPU, GPU, Neural Engine, etc) can access the same memory pool without having to copy or cache the data in more than one place.

 

So while not Micron it is at least one of the big three - SK Hynix

 

 

Cheers

 

nwoodman

 

Thank you.  That's what I was suspecting, just haven't found any corroborating evidence.  There's obviously sophisticated circuitry in the M1 to "unify" the memory, which is getting rave reviews.  With Intel falling behind TSM in process technology, it does look like semiconductor design is entering a new "golden era". 

[valuebet]

If I get it right Apple use MU chips for their A14 Bionic SoC in new iPhones:

 

[HJ]

For what it's worth, Tsinghua Unigroup, which is one of the major entities in China involved in the semiconductor space defaulted on a couple of small domestic bonds, USD$~200MM.  YMTC, considered by some to be an existential threat to the NAND suppliers, is a subsidiary of the Unigroup.  The default might end up being just a blip to the operations of YMTC, but is note worthy.  As with most things in China, they are opaque until they are thrust right in front of your eyes, and you go "how could I have missed it?".

 

[valuepickr]

Next Con Call will be interesting we might get some better insight

 

https://newsfilter.io/articles/update-1-chinese-phone-maker-honor-partners-with-key-chip-suppliers-after-huawei-split-6b301bbe2ae855482bb946381acea875

[samwise]

By a rough estimate they produced an average of $6 earnings in the last 3 years. So for a cyclical company do you apply a 10x multiple or 16x?

 

And while there will be bit growth, do you expect much revenue growth through cycles?

 

These questions have become more urgent as price is almost at 90, i.e. 15x P/E assuming EPS= $6 through the cycle. I've made a 3x on this and psychologically I have an itch to sell. But I'll make an attempt at valuation here and am looking for feedback and other people's thoughts.

 

Q: Is MU a long term compounder with cyclicality , or will its earnings say at the $6 level and just cycle around his over time? I am trying to answer this below; come up with a valuation with underlying assumptions.

 

Assumptions:

1. The oligopoly will succeed in managing supply to match demand over the long term and get a good ROE.

2. They cannot match demand over the short term. Demand moves in a month, but supply takes years to build. So mismatches will occur and cause cycles.

3. The oligopoly will not be broken by new entrants.

 

Assume that the company gets 20% ROE, and book value has been growing almost 20%. This leads to $6 over an average equity of $30  over the past 3 years, but implies that current earning potential is $36*0.2 = $7. Then the P/E multiple is reduced to 13x. If this EPS is growing with book value (at a rate much faster than GDP), then this definitely deserves a higher than 13x multiple.

 

Of course we have to assume that the increase in book is not creating oversupply which destroys the returns. So one question is, at what rate will book value increase to match long term demand growth? demand growth is 10-15% bit growth, but not sure how that translates to Capex .

 

Thoughts? growth views? price targets ? disagreement with the assumptions?

[RadMan24]

By a rough estimate they produced an average of $6 earnings in the last 3 years. So for a cyclical company do you apply a 10x multiple or 16x?

 

And while there will be bit growth, do you expect much revenue growth through cycles?

 

These questions have become more urgent as price is almost at 90, i.e. 15x P/E assuming EPS= $6 through the cycle. I've made a 3x on this and psychologically I have an itch to sell. But I'll make an attempt at valuation here and am looking for feedback and other people's thoughts.

 

Q: Is MU a long term compounder with cyclicality , or will its earnings say at the $6 level and just cycle around his over time? I am trying to answer this below; come up with a valuation with underlying assumptions.

 

Assumptions:

1. The oligopoly will succeed in managing supply to match demand over the long term and get a good ROE.

2. They cannot match demand over the short term. Demand moves in a month, but supply takes years to build. So mismatches will occur and cause cycles.

3. The oligopoly will not be broken by new entrants.

 

Assume that the company gets 20% ROE, and book value has been growing almost 20%. This leads to $6 over an average equity of $30  over the past 3 years, but implies that current earning potential is $36*0.2 = $7. Then the P/E multiple is reduced to 13x. If this EPS is growing with book value (at a rate much faster than GDP), then this definitely deserves a higher than 13x multiple.

 

Of course we have to assume that the increase in book is not creating oversupply which destroys the returns. So one question is, at what rate will book value increase to match long term demand growth? demand growth is 10-15% bit growth, but not sure how that translates to Capex .

 

Thoughts? growth views? price targets ? disagreement with the assumptions?

 

I've also held Micron but its been under a year so have no desire to sell. SK has embarked on some serious equipment expenditures and will try and iron out those new tech production kinks before Micron bites the bullet. A lot of this is new found territory. With everything needing more and more memory, particularly from the automotive sector, and with modest valuations on a 'mid-cycle'  EPS (which doesn't assume growth in midcycle EPS going forward), it's a tough case to make a sell (confirmation bias notwithstanding). At the end of the day, Micron is in a fantastic position with a lot of new industry moving in. This has the possibility of extending the cycle longer than many expect. I am not though expecting AMD or Nvidia type returns.

[samwise]

RadMan,

 

I decided to sell my MU position a while back, as I think the compounding scenario is going to be hard.

 

MU needs to thread a needle o achieve any compounding

1. Invest enough Capex to stay ahead of China

2. Invest at the same rate as competitors like Samsung

3. Invest slowly enough to get good ROIC.

 

This can also be thought in terms of price decreases passed on to customers.

MU needs to

1. Pass on cost decreases to customers , enough to stay ahead of China as they decrease their costs. Also enough to make customers switch to the lower cost bits.

2. Keep enough of the cost decreases to ensure that MU earns enough incremental EBITDA to cover the new CAPEX.

 

All this while managing a cyclical end market with Capex controls, inventory buildup etc as they did in the last cycle.

 

The more I thought about this, the harder this seemed to achieve. Now China is still many years behind, and maybe the tech is hard enough that they never catch up. But it seems the incumbents have to run as hard as possible to stay ahead, and also run at a measured pace where they earn great ROIC on incremental capital. Coke and cornflakes is a much easier business.

[ValuePadawan]

It is gonna be awful tough for the Chinese DRAM national champion Changxin Memory Technologies (CXMT) to catch up to where Micron and SK Hynix are going without being able to get their hands on an EUV tool. Now hopefully Micron doesn't pull an intel and mess up EUV adoption and end up behind the Korean memory makers.

 

On cost improvements the transition to the 1 alpha node alone will cause a 40% decrease in cost due to producing more chips per wafer as chip size decreases. Imagine China was 1 node behind 1 alpha (which they are not) that means Micron could sell a 1 alpha chip 20% cheaper than a CXMT chip's breakeven cost and still make money. It's like a race where every step you fall behind the frontrunner you gain ten pounds. With DRAM, continued scaling is only going to get more and more difficult and without earning any profit Changxin could very well become a money pit.

 

As scaling and shrinkage gets more difficult and demand continues to increase the only way to keep supply up with demand is by building more fabs which means increasing supply is now a boardroom decision instead of a survival necessity as it has in the past.

 

If I've made any mistakes please show me where my blindspots are y'all.

[LearningMachine]

It is gonna be awful tough for the Chinese DRAM national champion Changxin Memory Technologies (CXMT) to catch up to where Micron and SK Hynix are going without being able to get their hands on an EUV tool.

 

Looks like you're agreeing that probability of them getting their hands on an EUV tool is not zero.  What do you think the probability range for that is?

[samwise]

Micron could sell a 1 alpha chip 20% cheaper than a CXMT chip's breakeven cost and still make money.

 

Yes I don't dispute that. But will MU earn more or less than what it currently earns? IF earnings stay constant at $6 mid-cycle, then I am happy to sell at 15x P/E. Only reason to hold on (for me) would be growth in earnings, which I have no confidence in. of course P/E could go to 25 and you could do very well in the stock, or underlying EPS might be higher.

 

I don't have answers, but happy to share my confusion below about incremental returns on capital.

 

For growing EPS you need one of the following

1. Book value is constant and ROE keeps rising. unlikely.

2. Book grows at 15%, so depreciation grows at 15%, for constant ROE, EBITDA should grow at 15%. Now I don't know what 15% book means in terms of bit growth, but if I assume 15% bit growth, then you get 15% EBITDA growth by keeping ASP constant a constant margin (unlikely that ASPs don't drop). Or you could get 15% EBITDA growth by constantly increasing EBITDA Margin (also unlikely).

 

Maybe I just don't understand the industry in enough detail. But to believe this is a compounder with bits growing at 15%, you need to know how much Capex is required above depreciation and how much of cost decreases need to be passed on to customers.

 

I don't think this is purely a boardroom decision. Yes the industry is better than before, but I am not sure it is that much better, since the oligopoly only lasts if the players maintain their lead on nodes. A lot of MU's increased earnings have come from catching up with Samsung, so it can be done.

[ValuePadawan]

It is gonna be awful tough for the Chinese DRAM national champion Changxin Memory Technologies (CXMT) to catch up to where Micron and SK Hynix are going without being able to get their hands on an EUV tool.

 

Looks like you're agreeing that probability of them getting their hands on an EUV tool is not zero.  What do you think the probability range for that is?

 

ASML is currently making somewhere between 30-40 EUV tools per year and each one has a price tag of around $120m USD it is not like the Chinese can just buy one with a shell company and smuggle it into China. Still lets assume they did fool ASML and got their hands on an EUV tool. The tool is shipped from Veldhoven, Netherlands in pieces then rebuilt onsite by ASML experts and this isn't exactly Lego. It takes 6 cargo planes to ship the whole thing. The building that houses the EUV tool has to be specially built with a special foundation because the machine is so heavy.

 

Then when it arrives ASML experts have to put together and calibrate the EUV tool. How good does their calibration have to be? Well the room that the tool is in has to have constant humidity and stay at 21C so that the tolerances in some pieces of the machine don't become to tight or too loose due to thermal expansion or contraction of the metal.

 

Imagine the Chinese are able to do all that on their own. Then they have to correctly operate the EUV tool and achieve high yields on their wafers. Intel which has been working with ASML for decades to create the EUV technology had a lot of issues trying to produce their 7nm chips. If Intel with the capital, talent and pedigree had issues with making the EUV process work I can't imagine the Chinese could anytime soon and that is with ASML's collaboration.

 

Fine let's assume the Chinese get it all working and then the EUV needs a replacement part or the laser has an issue. With the restrictions on ASML set by the US it's not like they can ship that part and help the Chinese troubleshoot.

 

As long as the US keeps sanctions on Chinese chip companies for cutting edge tech and processes I give the chance CXMT gets and can use an EUV tool maybe 1%. Very low. I would say effectively 0 but I'm not creative enough to think of what can happen in the future. If the US allows ASML to sell EUV's to the middle kingdom all bets are off though.

[ValuePadawan]

Micron could sell a 1 alpha chip 20% cheaper than a CXMT chip's breakeven cost and still make money.

 

Yes I don't dispute that. But will MU earn more or less than what it currently earns? IF earnings stay constant at $6 mid-cycle, then I am happy to sell at 15x P/E. Only reason to hold on (for me) would be growth in earnings, which I have no confidence in. of course P/E could go to 25 and you could do very well in the stock, or underlying EPS might be higher.

 

I don't have answers, but happy to share my confusion below about incremental returns on capital.

 

For growing EPS you need one of the following

1. Book value is constant and ROE keeps rising. unlikely.

2. Book grows at 15%, so depreciation grows at 15%, for constant ROE, EBITDA should grow at 15%. Now I don't know what 15% book means in terms of bit growth, but if I assume 15% bit growth, then you get 15% EBITDA growth by keeping ASP constant a constant margin (unlikely that ASPs don't drop). Or you could get 15% EBITDA growth by constantly increasing EBITDA Margin (also unlikely).

 

Maybe I just don't understand the industry in enough detail. But to believe this is a compounder with bits growing at 15%, you need to know how much Capex is required above depreciation and how much of cost decreases need to be passed on to customers.

 

I don't think this is purely a boardroom decision. Yes the industry is better than before, but I am not sure it is that much better, since the oligopoly only lasts if the players maintain their lead on nodes. A lot of MU's increased earnings have come from catching up with Samsung, so it can be done.

 

As long as the only companies making DRAM are Samsung, SK Hynix and Micron I have no doubt they will in time behave rationally. This has become a Cournot Oligopoly (https://en.wikipedia.org/wiki/Cournot_competition) They will act rationally and ROIC will continue to be high. If CXMT is able to produce cutting edge DRAM all bets are off but if the ecosystem of companies stays at it's current equilibrium they are the toll gates on the information superhighway and they can use that market power to increase returns through-cycle. I fully expect them to be able to increase ASPs over the next decade as long as they keep up on node development with SK and Samsung. Node development and CXMT progress are what I'm staying laser focused on as I see those as the most important building blocks holding up my thesis.

 

Data is becoming ever more important in the global economy and with the DRAM company positions stable, their leverage will increase and EPS will follow that. Software is eating the world and chips are it's teeth.

[samwise]

I fully expect them to be able to increase ASPs over the next decade as long as they keep up on node development with SK and Samsung. Node development and CXMT progress are what I'm staying laser focused on as I see those as the most important building blocks holding up my thesis.

 

 

If that happens, all three companies will do really well and margins will be very high. Theoretically, I guess they could all increase price until just below CXMT costs, even as node development produces cheaper bits. So they could get huge margins with increasing revenue and reducing costs. Supposedly DRAM demand is inelastic so price increases should be possible.

 

Les see if it happens. Thanks for the interesting discussion.

[ak1979]

CXL is shaping out to be the standard for Memory interface with CPU,GPU and other accelerators.  it's not surprising to see Micron drop out of developing 3D Xpoint memory and plan to sell Utah Fab by end of the year.

https://blocksandfiles.com/2021/03/25/cxl-and-the-developing-memory-hierarchy/

[backtothebeach]

There is a guy posting his MU and memory chip cycle analysis on the Fool regularly. I started reading with this post: https://boards.fool.com/micron-valuation-33136322.aspx in 2018, where he pretty much nailed the downside prediction. (I only sold a few puts in the the $30-40 range figuring I could roll them forever until they expire worthless at very little risk. Should have put a limit buy below $30 instead of using options...).

Looking forward to his analysis of the next earnings report. IIRC his prediction for this up-cycle is MU topping out at $100-120+.