I have been asked to share my projections for Micron (NASDAQ:MU) for its fiscal year ending August 28, 2014. This article is intended to do that. It is also intended to solicit criticism and comments which result in a better set of projections and a better understanding of them. I consider it an ongoing joint project. "Ongoing" because I rely on and borrow extensively from prior work by Russ Fischer, Electric Phred, Bruce Burnworth, Wilson Wang, Jaret Wilson and William Tidwell (as well as others I'm forgetting). "Joint" because the article with the commentary is the message, and we all are writing that.
I began this article on Wednesday, March 5. Then a business emergency came up and I am just now getting back to it. A lot has happened since then. Inotera published February revenue and it was down 16.6% from January sequentially. Jared published an article, Micron and DRAM prices, in which he did the hard work of projecting the supply/demand situation for DRAM and the direction of DRAM pricing and costs of production. I will simply say that I am appropriating his conclusions and skipping that part of this article. Russ published two articles, Micron and Memory Prices and Micron: What's Going On? I will use Russ' mathematical and physical relationships and state my own views where we differ.
First, a little about my methodology. This is not bottom up analysis. I do not know the industry well enough for that. Nor is it top down analysis. This analysis has elements of both methodologies and some approaches I use which are not either. Let's call it left right analysis because, as you'll see, I take some historical numbers and manipulate and adjust them into the future.
What's our starting point? I picked the pro forma Unaudited Condensed Combined Statement Of Operations for the year ended August 29, 2013. My source for this financial statement, and all other data, unless otherwise expressly indicated, will be the SEC files. My sources for management statements and guidance are the recent analyst conferences.
That combined pro forma has twelve months of Micron's operations but only eleven months of Elpida's operations included in it. So I adjusted it, in order to "annualize" it, by multiplying the Elpida numbers by 12/11.
So now I've got an annualized Micron 2013 number in dollars. Let's use it to project 2014. In order to get there, we are going to project bits made and sold in 2013 into bit demand in 2014, both measured in 2013 dollars. The reason for this is that management guidance is stated in "bits" not dollars. Management avoids projecting dollars or prices due to, I assume, antitrust law considerations. I find the bits easy enough to work with, so I am going to project bits first and then convert them into dollars. The bits, however, are measured in 2013 dollars. So essentially I will be projecting 2014 dollars assuming no change in the pricing or the costs of bit production from 2013. We'll deal later with that by plugging in flexible pricing and cost adjustments.
Also, I will be projecting bit demand without regard to whether or not Micron can produce enough bits to supply the demand with the projected growth built into it. We'll deal with what happens if Micron cannot fulfill that demand later. Management has already stated that bit demand from certain sectors will grow beyond the industry's ability to supply it.
Changes in pricing, the costs of production and margins will be estimated at the end. I will use my own range for pricing since I do not believe Micron management's "flat" pricing guidance. I believe that is the "party line" created for use during antitrust settlement negotiations and thereafter, in an oligopolist environment, where it is all too easy to see collusion in public statements about pricing.
We have some management guidance on bit demand growth year-over-year from comments made in various analyst conferences. Unfortunately, management breaks the guidance down according to demand emanating from various industry markets, which is not how Micron breaks down its line of business financial results in its published financials. So we have to find a way to translate one into an approximation of the other in order to be able to work with it.
Here is Micron's revenue and percentage contribution to the total broken down by its five lines of business from the 2014 first quarter SEC filing:
Here are the descriptions of the groupings:
DRAM Solutions Group consists of DRAM products sold to the PC, consumer electronics, networking and server markets.
Wireless Solutions Group consists of DRAM, NAND Flash and NOR Flash products, including multi-chip packages, sold to the mobile device market.
NAND Solutions Group consists of high-volume NAND Flash products sold into data storage, personal music players, and the high-density computing market, as well as NAND Flash products sold to Intel through our IMFT joint venture.
Embedded Solutions Group consists of DRAM, NAND Flash and NOR Flash products sold into automotive and industrial applications, as well as NOR and NAND Flash sold to consumer electronics, networking, PC and server markets.
These groups do not correspond on a one-to-one basis with the groupings for which management gave bit growth annual guidance. Here is a table of that guidance:
As you can see from the descriptions of Micron's lines of business, the DRAM Solutions Group does not cover all the DRAM made and sold. Some of it is included in the Wireless and Embedded solutions groups, which have materially different projected growth in bit demand. Fortunately, we have enough information to create a workable relationship between the two groupings. Let's do that.
We have a breakdown of the DRAM in the DRAM Solutions Group by the markets into which it is sold from SpinyNorman here citing the Raymond James conference. We can use it to break out the DSG numbers by the market into which they are sold. Here is the DSG broken out and the other groupings with percentage of total calculations:
Now we have line of business data from the first quarter 2014 with the DSG broken down into bit numbers by the market into which it is sold. It still totals 100% but it is more defined. We will now use those derived breakdown percentages to break down 2013 combined financials as I adjusted them for Elpida's missing month.
Now that we have a breakdown of 2013 numbers, we will now apply the growth rates for bits made and sold from the management guidance stated above. We are trying to project bits made and sold for fiscal 2014, measured in 2013 dollars. What I mean by "measured in 2013 dollars" is that our projected number is what 2014 sales would be, on our assumptions, if prices did not change at all and if the cost of producing those bits did not change at all.
Here is the data projected out by applying the bit growth rates (I used the median of a range) to 2013 combined financials as I adjusted them:
As you can see, growth rates ranging from 0 to 65%, when applied specifically to the bits they relate to, produce a weighted average growth in bits of 42.6%. By these projections, Micron would sell 42.6% more product in 2014 than in 2013 if it were able to ramp production up to those levels. But management has alluded to a sales increase in the low 20s. That is not consistent with our bit demand growth calculations based on management projections, so I believe that was an allusion to dollars by mistake and no one caught the inconsistency. Let's assume that I'm right.
So if we believe both management and our own calculations, we have demand exceeding supply by about 20% of last year's total production. That is a lot of unfulfilled demand.
Management has already said that Micron and the rest of the industry will be unable to fulfill that level of demand. So what increase in bits sold should we project? Management has put the weighted average bit supply growth in the low 20s. That must mean that Micron will sell more bits than in 2013 by a low 20s percentage increase. Demand, which we have calculated, is higher than that.
Let's reduce the projected bit growth for each sub-market to get an estimate of bits Micron will sell in fiscal 2014, measured in 2013 dollars. Let's be a little more creative than just reducing everything pro-rata to get to the low 20s. Let's use 22.5%, the midpoint. Let's try to anticipate which orders, and therefore which sub-markets, Micron will reject because it is already at capacity. Some of the assumptions I will make may seem arbitrary. Let's hear your suggestions for improving them. Here are mine and my reasoning:
I am holding PC DRAM to down 1% year-over-year.
Lowest margin sales but you have to take care of your long-term customers.
I am allowing mobile DRAM to rise only 31% instead of 65%.
Have to cut somewhere and this is a big somewhere. Most of the growth is in the low end so margins are not likely high at the lowest price level.
I am allowing server DRAM to rise only 40% instead of 45%.
Have to cut but trying to hold on to good customers in good markets.
I am not allowing networking DRAM to rise at all.
Not so lucrative.
I am allowing the Wireless Solutions Group to rise only 27% instead of 65%.
Have to cut somewhere and this is a big somewhere.
I am allowing the NAND Solutions Group to rise the full 37.5%. Management's stated strategy is to achieve SSD attachment rates of greater than the current 15% to 25% in new notebook computers instead of the 10% rate of 2013. Management strategy is to lower pricing in this highly price elastic market and encourage organic growth in order to do so.
I am not allowing the Embedded Solutions Group to rise at all.
Have to cut somewhere and this is less important somewhere.
I took a little off Other. If I don't know what it is, it can't be that important and I am cutting it.
Here is the breakout of the 2014 made, sold and delivered gross bits measured in 2013 dollars. Note that these numbers are consistent with the lower guidance of growth in the low 20s. If we assume that Micron will do better than that in increasing bit production, (and I do think so) you get better numbers.
No account has yet been taken of Average Selling Prices or the Costs of Production. We've only measured the bits. Now we have to project pricing and costs of production to get to gross margins so that we can get dollars out of the bits. But before we do that, just out of curiosity, let's see what the result would be with no change in COGS as a percentage of revenue and with no adjustments to fixed costs. Admittedly, this is a way-down-below calculation:
Admittedly, this $2.95 number is not realistic except as a far lower boundary. But I figured that some of you by now wanted an answer - some answer - any answer, so I calculated it here. Some of you can stop reading here. I could. I already know that if Micron will earn $2.95 as a rock-bottom minimum, I am not selling any stock. But you can keep reading if you'd rather have more realistic estimates.
Now I am going to adjust for cost and pricing differences. Let's start with costs.
Changes in the costs of production are different for DRAM and NAND. We'll do DRAM first.
I recommend that you read Russ' recent article, Micron and Memory Prices, linked above, at least twice. He has done all the hard work for us. He thinks that a DRAM shrink from 30nm to 20nm would take costs per wafer down substantially. If I am understanding Russ correctly, the same cost produces 1,044 chips per wafer at 30nm, instead of 745 at 20nm. That represents a reduction in COGS from 68.3% of sales to 48.7% of sales, raising the gross margin from 31.7% to 51.3%. But hold on, Russ says that not all the parts of the chip shrink, so you don't get the benefit of the full shrink.
I am going to arbitrarily assume that we'll get half of it. Since I will lay out the numbers and the calculations, and provide you with a downloadable and configurable spreadsheet, each of you can do your own calculations. I will redo mine as soon as Russ tells us what he thinks of my arbitrary 50% of the benefit guess. If we cut the difference between 1,044 and 745 in half, that's 149.5. If we subtract that from 1,044, we get a yield of 894.5 20nm chips for the same cost as for 745 previously. That reduces COGS from 68.3% of sales to 56.9% of sales giving us a gross margin of 1.00 minus .569 or 43.1%. Let's leave 2014 pricing flat to 2013, per management guidance, and project 2014 DRAM dollars. Let's project NAND 2014 dollars at down low-teens, and costs down mid-teens. Micron management has stated that they intend to use the price elasticity of the SSD market to achieve greater attachment rates with lower pricing. I will upload configurable spreadsheets for these projections so that each of you can download them and put in your own assumptions to see what comes out.
First, we redo revenue. DRAM revenue is the same as in insert 7 above. NAND revenue comes down 12.5% to take account of management's statements. I am taking NSG revenue down the full 12.5%. I am taking WSG revenue, which includes some NAND revenue, down 3% which is my guesstimate of a weighted average contribution of the NAND revenue to the whole. I am taking ESG revenue down 3% since I don't believe that it is very price sensitive and is also a weighted average.
Now, we redo Costs of Goods Sold using the derived percentages for DRAM and lowering NAND Costs of Goods Sold only 17% from 2013 weighted average COGS. We do this because Micron is already at the leading technology for NAND. That comes out to COGS of 56.7%, almost the same as with DRAM.
Calculating, gross margin comes out to 43.2% and gross profit comes out to $6,514 MM. Taking off my estimates for R&D, SG&A, Interest and Taxes and adding in my estimate for Inotera equity, we get this:
That 2014 EPS estimate uses flat DRAM prices, lower NAND prices by 12.5%, flat NOR and other prices, lowered COGS to take account of our calculations based on management guidance. EPS comes out to $4.37. I think it's on the low side because we kept pricing flat.
There are two extraordinary items that I think should be factored into the results. One is a recent settlement in a civil antitrust suit for $66.1 million. I figure that will reduce earnings, net of taxes, about $.04 per share. The other is if Congress passes, and the president signs, a renewal of the R&D tax credit retroactively, as they have always done before and as I believe they will do know. That will increase earnings, net of taxes, about $.02 per share.
Pricing is usually not projected by Micron management other than in vague directional language. We have conflicting information on DRAM pricing which would allow the reasonable conclusion that no change in ASPs is expected. Some very sophisticated SA authors have suggested that pricing will go down. That doesn't make sense to me. I think prices will go up on balance.
Management has indicated that demand will exceed the available supply in most markets. My own research confirms that. Kip Bedard has recently identified several markets which are expected to grow this year well above the industry's ability to supply it. When OEMs (or anyone) can't buy the stuff they need to make and sell their product, they try harder. They raise the prices they are willing to pay. They offer prepayments. They threaten you with the loss of a good customer, etc.
But ultimately it always comes down to price.
Some higher offers come in and the manufacturer looks around for available supply that it can re-allocate. More higher offers come in and the manufacturer reallocates available supply again and prices go up some more. The theoretical limit is when higher prices cause enough demand to evaporate to create balance. That limit won't be reached since smart manufacturers will always take care of good customers and not gouge them. Prices go up, but not all the way to equilibrium.
This is an oligopoly and it will continue to be an oligopoly for a very long time with an admission price north of $10 billion. Oligopolies don't lower prices when demand substantially exceeds supply, at least not by very much and not for very long.
Russ and others, whose opinions I respect, are of the opinion that memory prices will go down. Russ' reasons are complicated (read his article) and are as good as any. I don't agree. This time, it's different. I mean really, really different. The industry is an oligopoly and the largest member of the oligopoly is Samsung (OTC:SSNLF), a company whose interests lie in higher memory prices so that its rival in the smartphone duopoly, Apple, will have to pay more for component parts. If Samsung were to have its way, mobile memory would double in price leaving Samsung as the only firm selling smartphones at mid-market prices. Micron, as the largest merchant seller in an industry of four, will fall, willingly or unwillingly, into the role of pricing leader.
So let's indulge us both. Let's raise and lower prices 10% and see what happens. Wait, I've uploaded a configurable spreadsheet. You can do it yourself. :-)
The spreadsheet is here (opens to an Excel file).
A word about the future. Micron is the largest merchant seller in an oligopoly of four. That's not likely to change during the next ten years. Demand is growing faster than the industry's ability to supply it. That's not likely to change during the next ten years. Costs are coming down. That's not likely to change during the next ten years. Prices can and will find their equilibrium at the point which maximizes profits for the oligopoly. That's not likely to change during the next ten years.
Result: Micron is now a high growth company in a high growth industry and .....
That's not likely to change during the next ten years.
Disclosure: I am long MU. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.
Additional disclosure: I am long MU stock. I may take one or more positions in Micron call options from time to time for trading purposes.