Seeking Alpha
Deep value, growth, contrarian, long only
Profile| Send Message|
( followers)  

Summary

  • Micron’s "indicators” for Q3 were cautionary.
  • DRAM performance is the key and should be very strong.
  • A model is proposed and scenarios offered.

Micron (NASDAQ:MU) eschews offering guidance to its investors; so coming up with an earnings estimate has some interesting challenges. What it does offer instead of guidance is an estimation of how well it will do in its two major product lines, those being DRAM and NAND memory. Estimates are offered in each of three areas, which are bit production growth, bit pricing levels, and bit costs. The estimates offered are relative to the last quarter, and therein, in the combination of these six factors, lay the clues to financial output in the quarter to come. Summing up, (for those of you with an analog bent), we have six levers which in combination will give Micron investors and analysts the power to lift the curtain up and reveal the future quarter's financial potential.

These six factors, three per technology, all directly affect one of two (and sometimes both) critical elements of corporate performance, those being revenue and gross margin. Since Micron and all of its peers in the semi-conductor memory industry are called upon as a routine matter to invest billions of dollars in plant and assets in these two technologies, it seems only right that, failing to provide real guidance, forecasting these elements - ASP, bit growth, and cost per bit - is exactly the right fall back approach. After all, if investors can feel comfortable that the company can manage top line revenue growth and manage appropriate levels of profitability with that growth, expense levels can be managed to levels that ensure investment in future technologies is possible.

Unfortunately for investors, Micron offers us the lever but disdains to offer the fulcrum. Micron does not in fact provide us with gross margin reporting for DRAM and NAND. We have the rod to move the world but we do not have the rock on which to place it. This is disappointing but there it is. How do we go about finding our own fulcrum? This article will discuss that process and, as an endpoint, offers earnings scenarios for Q3 that utilizes our purpose-built fulcrum substitute.

Now, I don't know about you, but I don't work day-to-day in the semi-conductor memory business and I certainly don't work for Micron. Because I don't, I don't have access to any inside information or even rumors about Micron's performance other than what can be gleaned in the trade press. So where does that leave us? Is there no hope for anything approximating a scientific approach to getting an answer?

Well, before we throw our hands up in frustration and just let loose with our very best swag*, let's gather up what we do know and see what we can make of it. So what do we know? First, we know what Micron has estimated. For Q3 2014, here are their estimates:

DRAM:

Bit Growth - Up "low single digits"

ASP - (As Sold Price) - "Down low single digits"

Bit Cost - "Down low single digits"

What is Micron telling us with these estimates? My take is that Micron saw a marginal improvement in the DRAM business with price reductions and cost reductions canceling each other out, but the business is still growing because of increased production. This makes sense to me because they are in the midst of a big transition to 20nm process with all the day-to-day uncertainties that such a project entails. Another issue is the built-in pricing drag that their contract laden DRAM business brings with it. "Spot" pricing may be rising, but if Micron just locked in a 3-6 month contract with say, Apple, their pricing may not be changing (much) in that period.

Where are we with NAND?

NAND

Bit Growth - "Down high single digits"

ASP - "Down low single digits"

Bit Cost - "Flat"

Here, Micron seems to be very pessimistic. Production is down significantly; prices are projected to be lower, and costs are level with the previous quarter. Depending on how we define "high single", "low single" and "flat", Micron's NAND business could be down 10% or more, top and bottom line. Not pretty. This too makes sense to me because Micron has made it abundantly clear that they are not going to "dump" NAND on the market just to clear inventory. Also, very recently in the Sanford Berstein conference, Mark Durcan, CEO, reiterated that approach. They are building the as yet unqualified memory that they manufactured into SSDs and that will involve delays in bringing the product to market. One other nugget: 16nm process ramp is going slowly (perhaps deliberately?), resulting in only 10% of the NAND bits coming from Micron being of that variety.

What else do we know? We know a little bit about Micron's Q2 results. Not all that we would like to know by any means. We do not actually know, for instance, what Micron makes in DRAM and NAND. Instead of providing investors with that information, Micron reports by four business units, and two of those combine both DRAM and NAND, so the overall results for two memory technologies are a muddle. Here is what Micron gives us:

Div. Rev.

Division

Op. Income

Op. Margin

$1,888

DSG

$520

27.5%

$910

WSG

$185

20.3%

$902

NSG

$77

8.5%

$365

ESG

$59

16.2%

$42

Other

$28

66.7%

$4,107

$869

The DRAM Solutions Group (DSG) has a revenue stream focused entirely on selling DRAM technology, so that's a pure number. The NAND Solutions Group (NSG) is entirely NAND product, so there's another pure number. WSG (Wireless Solutions Group) and ESG (Embedded Solutions Group), on the other hand, sell products that combine the technologies. Micron, in its Q2 10-Q does say that of the two technologies "DRAM products account[ed] for a significant majority of the sales" in WSG, so that's a clue we can use for consolidating sales by technology. For ESG, Micron goes on to tell us that NAND flash exceeds DRAM in revenue and NOR brings up the rear. Ron Foster, on the Q2 earnings call, estimated Q3 NOR revenue to be between $100-110M, so the NOR distribution needs to approximate that number.

Let's normalize the data as best we can to come to estimate of revenues and profit for DRAM and NAND. Here's a stab at it

The table above is from Micron's Q2 10-Q and shows their revenue by each of their four divisions plus a category they call other. If we divide the "Other" category into thirds and prorate those amounts into the DRAM and NAND product lines based on the relative bit production levels for each (two-thirds DRAM and one-third NAND), and then go on to make some assumptions about technology content in WSG and ESG, here's an approach that made sense to me.

Division

DRAM %

NAND %

NOR %

WSG

75%

15%

10%

ESG

46%

50%

4%

Which gives us the imputed proration of DRAM and NAND revenue in the divisional reporting. (shown below)

DRAM

NAND + NOR

Total

$850

$425

$1,916

$916

$2,766

$1,341

Here is the breakout that derives gross margin by product line.

Margin Distribution

WSG

ESG

NSG/DSG Op. Income

Total

NAND

$46

$32

$77

$155

DRAM

$139

$27

$520

$686

And here is the outcome of the revenue and operating expense proration that derives gross margin: (notes: the top line is NAND - operating expenses are added back into operating income to provide a gross margin amount).

Total Imputed Op. Income

Total Imputed Revenue

Total Imputed Op. Inc. %

Op. Ex ProRata (1/3, 2/3)

Gross Margin

GM %

$155

$1,341

11.6%

$176

$331

24.8%

$686

$2,766

24.8%

$358

$1,044

37.4%

Now, admittedly this is only an approximation. If you don't like the estimates I made of DRAM and NAND/NOR product that I made for ESG and WSG please feel free to make your own.

The surprise I got from this exercise is that NAND is making as much money as this shows. At 25% gross margin, the business is obviously under performing, but it's not horrible, even under conditions in the last quarter that included Micron selling some unqualified product on the market. On the other hand, the numbers feel accurate because the overall NAND operating income should go up as it did from 8.2% to 11.3% when the prorated profits from value-added sales in WSG and ESG products are added in.

In any case, we now have a base line for Q2 that will allow us to apply the bit growth, ASP, and cost estimates that Micron gave for Q3. Now let's make an attempt at providing some rigor to Micron's estimates. First off, just what exactly does "low single digits" mean? Obviously, it could mean something different to each of us, but here is my attempt to pin these terms down:

Description as Percentage

flat

0

low single

2

mid single

5

high single

8

mid to high

7

low teens

12.5

mid teens

15

high teens

17.5

So now we have a tool to apply a certain percentage to the Q2 actuals based on the Micron estimates, right? Before we do that, though, let's analyze Micron's predictions to see how good they are. If we can have some confidence about Micron's ability as a forecaster then this whole exercise is going to go a lot smoother. So here's the data from the last four quarters, when the numbers assigned above are totaled based on the difference between the quarterly forecast and the outcome. Note that positive variance means that Micron's estimate was low and the outcome was better. If the variance is negative, Micron was too optimistic and the actual result was worse than they forecast. So here goes, starting in Q3, 2013.

Q3 Estimate

Category

Q3 Actual

Variance

Q4 Estimate

Category

Q4 Actual

Variance

Bit Growth

Bit Growth

up mid to high single

Trade NAND

0%

-7.0%

up high single

Trade NAND

13%

5

couple

DRAM

6%

4.0%

up high single

DRAM

42%

N/A

ASP

ASP

down mid single

Trade NAND

8%

13.0%

down mid single

Trade NAND

-9%

-4

up mid single

DRAM

16%

8.0%

up mid to high single

DRAM

5%

-2

Cost / Bit

Cost / Bit

down mid single

Trade NAND

1%

-4.0%

down high single

Trade NAND

-10%

2

down mid to high single

DRAM

-5%

-2.0%

flat

DRAM

4%

-4

Note the "N/A" for the variance for DRAM bit growth in Q4. This was the period when Micron was bringing on Elpida and I felt that circumstance was a one-off that would excuse some poor forecasting or, more likely, could be explained as a forecast for their pre-Elpida business. Here are the numbers for Q1 and Q2 of 2014.

Q1 Estimate

Q1 Actual

Variance

Q2 Estimate

Q2 Actual

Variance

Bit Growth

Bit Growth

up mid teens

Trade NAND

17%

2

up high teens

35%

17.5

up mid-40's

DRAM

69%

24

flat

0

0

ASP

ASP

down high single

Trade NAND

-6%

2

down high teens

-18%

0

up mid single

DRAM

Flat

-5

flat

-1%

-1

Cost / Bit

Cost / Bit

down high single

Trade NAND

-7%

1

down mid teens

-12%

-3

down low single

DRAM

-11%

9

down high single

-8%

0

And here is the variance table abstracted from the above:

Category

DRAM Variance Map

Q3 2013

Q4 2013

Q1 2014

Q2 2014

AVG.

Bit Growth

4

N/A

N/A

0

2.0%

ASP

11

-2

-5

-1

-2.7%

Cost/Bit

-2

-4

9

0

-2.0%

4.5

-3.0

2.0

-0.5

0.0%

So what can we discern from the last year of forecasting? This is admittedly a very small sample, but overall forecasting wasn't too bad from an "average miss" perspective, particularly if you throw out quarterly results where the deviation from the norm was greater than 5 points. (Those of you with a statistical bent are welcome to go crazy on this and give us some better insights in the comments section.) Granted, this is horseshoes and hand grenades in terms of the data points, but with that caveat Micron turns out to be, within forecasting categories, pretty good and generally conservative (i.e. their actual turned out better than the forecast). Bit growth seemed to be the area that poses the biggest challenge for them, and that may be simply because they are ramping tech nodes (the 20nm for DRAM and the 16nm for NAND). The good news is that their wrong forecasts and big misses are generally being too conservative - the actual was better than the forecast.

Prices, except for Q3 2013 in both technologies, were generally very consistent, although this is an area (for DRAM at any rate) where their forecasts proved optimistic. Cost is another area that is pretty consistent though they did have a couple of big misses and there again that may have to do with a tech node ramp outcome being one that they didn't foresee.

I'm going to apply this data as a modifier to Micron's Q3 estimates as a way of incorporating a "confidence factor" into the "Base Case". Before I apply the variance data, I'm going to throw out one of the data points in a category where the overall variation is greater than or equal to 5 points higher than the next closest value and then average the 3 data points that remain. The "AVG" column above reflects this methodology. Now that we've got a gross margin number, we can now play out some earnings scenarios. The first scenario - "Micron Base Case" will convert Micron's forecast, modified by their historical forecast variance, into a Q3 revenue and gross margin result. The second scenario - let's call this one "DRAM Upside" - will modify scenario 1's assumptions about DRAM. There is good reason to believe that Micron was way too conservative about DRAM prices. Inotera, for example, reported very strong Q1 results at the end of March and there is no reason to believe the DRAM pricing momentum changed through Q3's end in May. Accordingly, scenario 2 will increase DRAM prices by 3% and leave intact Micron's forecast for DRAM bit growth. Because Micron's effective cost per bit goes up because of the Inotera contract price increases which track DRAM pricing levels, Micron's cost is bumped up .25%. NAND will be left unchanged from Scenario 1.

The third scenario, which we might term "High End", will bump up both DRAM and NAND performance. DRAM prices will be increased by 5%, costs will be unchanged from scenario 2 - up .25% - and DRAM capacity remains at "Base Case" levels. NAND performance will be increased by assuming a price increase for NAND at 1.75%, a bit growth of 5% and a bit cost reduction of 1.00%

Let's run the numbers and see how they play out for revenue and margins. Note that the "variance weighted cell" (highlighted in blue for your convenience) is the sum of "ASP Guidance", Variance, and the Modifier for that estimate factor for the Base Case but is equal to the "Modifier" value for Scenarios 2 and 3.

Micron Q2 Forecast - Base Case

Q2

Q3

% Chg

Micron Forecast

Variance Weighted Factor

Variance

Modifier

DRAM Revenue

$2,766

$2,610

-5.64%

Bits*

3000

2985

-0.5%

Rev/Bit

$0.92

$0.87

-5.2%

ASP guidance

-2.5%

down low single

-5.2%

-2.7%

0.00%

Bit Growth

-2.5%

down low single

-0.5%

2.0%

0.00%

DRAM COGS %

62.3%

63.0%

1.2%

DRAM COGS $

$1,723

$1,645

-4.5%

Cost guidance

-2.5%

down low single

-4.5%

-2.0%

0.00%

Cost per Bit

$0.57

$0.55

-4.5%

DRAM GM$

$1,044

$965

-7.52%

DRAM GM%

37.7%

37.0%

-2.00%

NAND Revenue

$1,341

$1,231

-8.19%

Bits*

1000

948.1

-5.2%

Rev/Bit

$1.3409

$1.2985

-3.2%

ASP guidance

-2.5%

down low single

-3.2%

-0.7%

0.00%

Bit Growth

-7.5%

down high single

-5.2%

2.3%

0.00%

NAND COGS %

75.3%

77.4%

2.7%

NAND COGS $

$1,010

$953

-5.7%

Cost guidance

0.0%

flat

-0.5%

-0.5%

0.00%

Cost per Bit

$1.01

$1.00

-0.5%

NAND GM$

$331

$278

-15.88%

NAND GM%

24.7%

22.6%

-8.37%

*Bit numbers for DRAM and NAND are arbitrary and reflect relative production levels.

In this case, both the top line and margins take a big hit in both technologies. Revenue drops more than 5% in DRAM and more than 8% in NAND. Not surprisingly, gross margins also drop significantly, with DRAM declining by 2% and NAND by 8%. Not good!

DRAM Upside Scenario

Q2

Q3

% Chg

Micron Forecast

Weighted Factor

Variance

Modifier

DRAM Revenue

$2,766

$2,835

2.49%

Bits

3000

2985

Rev/Bit

$0.92

$0.95

ASP guidance

-2.50%

down low single digits

3.00%

0.00%

3.00%

Bit Growth

-2.50%

down low single digits

-0.50%

2.00%

0.00%

DRAM COGS %

62.3%

58.2%

DRAM COGS $

$1,723

$1,650

Cost guidance

-2.5%

down low single digits

-4.25%

-2.00%

0.25%

Cost per Bit

$0.57

$0.55

DRAM GM$

$1,044

$1,186

13.60%

DRAM GM%

37.7%

41.8%

10.85%

NAND Revenue

$1,341

$1,231

-8.19%

Bits

1000

948.1

Rev/Bit

$1.34

$1.30

ASP guidance

-2.50%

down low single digits

-3.17%

-0.67%

0.00%

Bit Growth

-7.50%

down high single digits

-5.19%

2.31%

0.00%

NAND COGS %

75.3%

77.4%

NAND COGS $

$1,010

$953

Cost guidance

0.0%

flat

-0.51%

-0.51%

0.00%

Cost per Bit

$1.01

$1.00

NAND GM$

$331

$278

-15.88%

NAND GM%

24.7%

22.6%

-8.37%

The DRAM Upside Scenario features a more than 8% swing in DRAM ASP from the Scenario 1 Base Case model. Predictably DRAM swings to 2.5% revenue increase and a more than 10% gross margin increase. This is much better news for Micron investors!

Let's take a look an even stronger scenario that features even more improvement in DRAM effective ASP gains and throws in a less dire NAND forecast.

DRAM & NAND Upside

Q2

Q3

% Chg

Micron Forecast

Weighted Factor

Variance

Modifier

DRAM Revenue

$2,766

$2,934

6.05%

Bits

3000

3030

Rev/Bit

$0.92

$0.97

ASP guidance

-2.50%

down low single digits

5.00%

-2.67%

5.00%

Bit Growth

-2.50%

down low single digits

1.00%

2.00%

0.00%

DRAM COGS %

62.3%

56.2%

DRAM COGS $

$1,723

$1,650

Cost guidance

-2.5%

down low single digits

-4.25%

-2.00%

0.25%

Cost per Bit

$0.57

$0.55

DRAM GM$

$1,044

$1,284

23.05%

DRAM GM%

37.7%

43.8%

16.03%

NAND Revenue

$1,341

$1,433

6.84%

Bits

1000

1050

Rev/Bit

$1.3409

$1.3644

ASP guidance

-2.50%

down low single digits

1.75%

-0.67%

1.75%

Bit Growth

-7.50%

down high single digits

5.00%

2.31%

5.00%

NAND COGS %

75.3%

72.9%

NAND COGS $

$1,010

$1,044

Cost guidance

0.0%

flat

-1.51%

-0.51%

-1.00%

Cost per Bit

$1.01

$0.99

NAND GM$

$331

$388

17.28%

NAND GM%

24.7%

27.1%

9.78%

So now we're really looking good with the top line growing by 6% + in both technologies and gross margins increasing by 16% in DRAM and almost 10% in NAND over Q2. NAND gross margins are still terrible but at 27% NAND is at least contributing positively to the overall business. Imagine how good the earnings would be if NAND could match DRAM at 43.5%!

Which scenario is more likely to match the Q3 earnings call number? For my money it's Scenario 2 - DRAM upside. Let me count the ways:

  1. Inotera's strong results have demonstrated that at least Inotera's share (30%) of Micron's business probably benefited from strong demand. As I understand it, Inotera serves primarily the server memory and pc market and thus it seems logical that market reception and this pricing should be very strong.
  2. DRAM spot prices were up during Q2. While Micron probably doesn't sell that much of its output in the spot market the fact that prices were up indicates that upward pressure was lifting the prices in Micron's contracts.
  3. The counter to point 1 is that Micron buys Inotera memory at an agreed upon discount to "market" pricing. This means that in rising markets, Micron's costs will actually rise for the roughly one-third of its production that it gets from Inotera. That cost increase was modeled in the scenario, but I don't know how accurate that .25% cost increase will be.
  4. While pricing was up and rumors of memory allocation were rampant, one damping factor that shouldn't be ignored is that much if not all of Micron's DRAM business is with customers who do business with Micron through short-to-intermediate supply contracts. Just because spot pricing trends were up does not mean that Micron benefited from higher prices with customers.
  5. The NAND forecast seems to be a likely outcome because Micron has been adamant that they won't unduly dump product when there is a reasonable alternative that would allow margin recovery if they inventory it for use in the SSDs they hope to sell in 2H 2014.

So, bottom line, Micron's DRAM business did well but not overwhelmingly great. 3% effective ASP hikes for the business with bit supply and bit costs following the Base Case Q2 forecast feels right to me. If it is, here's the way EPS will fall out.

Scenario #2 - DRAM Upside

Q2

Q3

Chg %

Ratio

Total Revenue

$4,107

$4,066

-1.0%

DRAM

$2,766

$2,835

-2.5%

69.7%

NAND

$1,341

$953

29.0%

23.4%

Total GM$

$1,375

$1,464

6.5%

36.0%

DRAM

$1,044

$1,186

-13.6%

81.0%

NAND

$331

$278

15.9%

19.0%

Expenses

$(733)

$(733)

Pre-Tax Income

$642

$731

-2.1%

Income Tax

9.4%

$(63)

9.4%

$(69)

Equity in Affiliates

$134

$136

1.5%

Consolidated Net Income

$741

$798

-2.1%

Minority Interest Expense

$(10)

$(10)

Net Income

$731

$788

7.8%

19.4%

Non-GAAP Adjustments

$258

$183

-16.3%

Total Non-GAAP Income

$989

$1,004

1.6%

24.7%

Diluted Shares Outstanding

1159

1200

$0.85

$0.81

-5.1%

How does this scenario compare with Scenario #1? Here's Micron's forecast modified only by their historical performance.

Scenario #1 Micron Base Case Forecast

Q2

Q3

Chg %

Ratio

Total Revenue

$4,107

$3,841

-6.5%

DRAM

$2,766

$2,610

-5.6%

68.0%

NAND

$1,341

$1,231

-8.2%

32.0%

Total GM$

$1,375

$1,244

32.4%

DRAM

$1,044

$965

-7.5%

77.6%

NAND

$331

$278

-15.9%

22.4%

Expenses

$(733)

$(733)

Pre-Tax Income

$642

$511

-20.4%

Income Tax

9.4%

$(63)

9.4%

$(48)

Equity in Affiliates

$134

$136

1.5%

Consolidated Net Income

$741

$599

-19.2%

Minority Interest Expense

$(10)

$(10)

Net Income

$731

$589

-19.5%

15.3%

Non-GAAP Adjustments

$258

$183

-16.3%

Total Non-GAAP Income

$989

$805

-18.6%

20.9%

Diluted Shares Outstanding

1159

1200

$0.85

$0.64

-24.6%

Not nearly as pretty, is it?

A note about expense assumptions and non-GAAP adjustments. On the expense line, I basically took last quarter's numbers across because I haven't heard any information that would greatly modify them. Foster projected slightly higher SG&A but nothing material. Income tax was modified to be congruent with earnings. None of the other expenses, including R&D, were called out as materially changing for the upcoming quarter. One contra-entry in the non-GAAP adjustment category was mentioned, and that was the $42M of Elpida inventory write-up that was recorded in Q2 that completed the Elpida acquisition accounting for on-hand inventory. That being the case, Q3 non-GAAP adjustments were reduced by $42M to $216M from $258. $33 million of Elpida asset adjustments from Q2 were also identified, and, absent any reason for assuming there would be additional gains, another $33M was taken from the non-GAAP gains/losses for a total of $183 net. Inotera's profit rose slightly in the last quarter so I bumped the "equity method" gain for them a little. Those of you that have better insights into this category are welcome to set me straight. One category that I could well be material relates to the Dollar-Yen exchange rate, which swung favorably in Q3. I do not know how to measure that impact and that could swing the non-GAAP number a penny or two. Diluted share outstanding were assumed to rise somewhat, so I bumped the number to 1.2M and called it good. Your mileage may vary.

Conclusion:

My take from the numbers and the notions is that an earnings range of $.80 to $.83 cents per share is probable, with the top line down because they had such a miserable quarter for NAND. Could it have been a lot better? You bet, and my take on earnings may prove unduly pessimistic. On the other hand, I really have a high degree of confidence that the number will not be worse than $.80. If it is worse than $.80 then DRAM really was as blah as Micron forecasted in the Q2 call and we'll all be wondering why, which should make for a very interesting Q&A with the analysts after the announcement.

How about Q4? It's looking marvelous! Every 1% of pricing leverage in the DRAM business gives us 3.5 pennies of EPS in my model. I'm not counting on the bleeding in NAND to stop in Q4 so at this point I'm hoping for them to stabilize the business and start building it up in FY '15. Bottom line, high $.80's to low $.90's is my take with the top line growing to $4.1B and a little change. FY '14 will go into the books at $3.30 to $3.40 Non-GAAP, which is about $.25 higher than I would have thought before working through this model.

A final thought regarding the multiple, which has been on the rise of late with the public capitulation of some notable Micron bears, as "Electric Phred" points out in his recent article and with the bumping of price estimates from several others, not to mention the general frothiness of the market. For those of us who have been analyzing the industry over the last year or so in this SA venue and predicting that Micron's earnings and multiple would inevitably rise there can't help but be a little self-satisfied smiling going on. We bulls have won; the big bad bears have been routed and ride in chains behind our triumphal chariot as we hear the cheers of the crowd. Ah, but what's that other sound out there? Somebody is shouting about a rumor of a general war in the Middle East and there's this whisper in our ear from slave that stands behind us in the chariot. What is he saying? Oh yeah.

"Sic Transit Gloria"

Be careful out there - as Phred has so ably pointed out; it's going to be a bumpy ride.

*SWAG - Scientific Wild-Ass Guess

Disclosure: The author is long MU. The author wrote this article themselves, and it expresses their own opinions. The author is not receiving compensation for it (other than from Seeking Alpha). The author has no business relationship with any company whose stock is mentioned in this article.

Additional disclosure: Also long SNDK

Source: Micron: The Best Is Yet To Come