After listening to many Micron (NASDAQ:MU) presentations, and reading dozens of Micron articles and thousands of comments, I've noticed a few common questions that I'd like to give my take on.
Why is Micron switching its Singapore fab from DRAM to NAND when DRAM has higher margins?
This question is surprisingly common from sell-side analysts.
For mobile, gaming, server, and networking DRAM, demand is inelastic, which means that a change in the price has very little influence on the amount that people want. PC DRAM is somewhat elastic, meaning that if you raise the price a lot, then manufacturers and consumers may opt for computers with less DRAM in order to save money. By contrast, NAND is very elastic. The amount of NAND people buy is directly affected by the cost per bit.
Let me give an example: When Apple's (NASDAQ:AAPL) iPhone 5s was released, the base model carried a bill of materials of $199 and the phone sold for around $649. $11 of this was mobile DRAM, while $9.40 was NAND, so each of the two components were around 5% of the cost and a little under 2% of the price.
Much or all of the iPhone DRAM comes from Micron's Elpida fab in Hiroshima. If Micron called Apple and said, "Hey Apple, we're running low on DRAM, we're going to have to double the price," Apple would simply have either taken a 2% hit to gross margins, or raised the price slightly. An iPhone 5s will not run with less than the 1GB of DRAM that it is normally built with -- use half as much and you're back to the iPhone 4s with cripplingly slow speed or fewer features. The cost of the DRAM is negligible to the final consumer, so even if Apple raised the price of the phone by $11 to protect their margins, the consumer would hardly notice.
Not so with NAND. If SK Hynix, the manufacturer of iPhone 5s NAND, had insisted on double the price, Apple would simply have started making base models with 8GB of storage instead of 16GB. There would be no change in functionality other than a bit less space for apps and media files. NAND demand would drop as Apple ordered less, there would be excess supply, and Hynix would come to their senses and drop the price back down.
This is true across the DRAM world, with the partial exception of PCs. Microsoft's (NASDAQ:MSFT) Xbox can't be built with less than 8GB of DRAM, while SanDisk's (SNDK) SSDs can have any amount of NAND, and consumers will only want them if the cost per bit is below a certain threshold. Micron's Kipp Bedard has reported in recent presentations that consumers are indeed buying SSDs with twice as much NAND now that NAND price has fallen.
This is both a good and bad thing for NAND producers. It means that the system can absorb a really tremendous amount of NAND when the price is low enough, but it also means that it's a lot harder to get really tremendous pricing because people will react by buying less. NAND producers can only increase margins beyond a certain point by decreasing costs.
DRAM producers, however, have another option. They can create an artificial shortage by shifting DRAM production to NAND production. As DRAM demand is inelastic, DRAM prices will skyrocket. Meanwhile, NAND demand is elastic, so as the supply increases, the price decreases, which in turn increases demand, which raises prices back up a bit. The net result of such an artificially induced shortage is that DRAM sells at great prices while NAND sells at ok prices. While the NAND market has higher demand growth, the DRAM market is superior in an oligopoly because of this price control mechanism. Of course, the optimal position is to be in both with a large share of DRAM, as Micron is.
Inelastic demand can have truly dramatic effects -- imagine 10 people and 9 life-saving doses of medicine. A bidding war results, and the medicine sells for at least the amount of money that the 9th richest person has. If instead you have 10 doses of medicine, the medicine can sell for a trivial amount and everyone's happy, except the seller of the medicine.
If you ask Micron why they shifted the Singapore Tech fab from DRAM to NAND, they'll say something about having just acquired a lot of DRAM production so they needed to rebalance. However, I've heard Kipp Bedard allude to price elasticity, so it shows they are at least aware of the principle. I believe that Micron can and will deliberately push up DRAM prices at will, which is a brilliant maneuver.
What are these horrible convertibles I'm hearing about? What effect are they having on Micron's stock?
A convertible bond is a bond that, besides paying interest, can be exchange for a set number of shares during certain time periods. Companies like them because the convertible feature allows them to pay lower interest, since the conversion feature provides the buyer with another benefit. While it's more unpleasant than straight debt to the shareholder, it's less unpleasant than a secondary offering of shares because the convertible bonds are far more likely to be bought back, and the repurchase process is simpler for the company than with shares. However, if the price of the stock goes up a lot, then the cost to buy them back can skyrocket. That's what's happened to Micron's current convertible debt, which is now costing them 2 to 3 times the amount that they got when it was originally issued.
Micron has been steadily buying them back from the market, and when possible, calling them (forcing the buyers to sell.) The next ones that are callable are in early August, but they can buy from the market at any time.
The "converts" currently add nearly 64 million shares to Micron's fully diluted share count, though this number shifts slightly with the price of the stock. Since Micron has around 1.07 billion shares excluding the converts, the converts amount to about a 6% dilution, meaning that earnings are decreased by around 6%. A noticeable, but not huge amount. This is an effect of similar magnitude to the issuance of options to employees, a normal corporate practice.
It's been suggested in comments that Micron is keeping the price of its stock down to make repurchasing the converts cheaper. That's ridiculous -- besides that being illegal, it would also lower the amount that insiders get when they sell their own stock. Insiders are not likely to commit a crime to save their company money, especially when it would simultaneously cost them a lot of money.
It's also been suggested that there's a huge amount of hedging of the converts with shares that have been sold short, and that this amount is steadily growing. That's also wrong, or at least an exaggeration. Buying a convert and then hedging it defeats the purpose of buying a convert. Better to just buy a higher yield bond that's not convertible than to buy the convert and throw away its best benefit. Also, as the converts have been repurchased, the short position has not decreased as would happen if they were actually being hedged.
When will 3D NAND arrive?
This is another favorite question of analysts, and it's the wrong question to ask. The main point of 3D NAND isn't that it's better, although there are some speed advantages, it's that cost can no longer be reduced by building NAND in 2 dimensions, so manufacturers must resort to 3. 3D NAND will arrive when it's about to get cheaper than 2D NAND, although that will take time and it will ramp gradually. The date isn't important, the ability of a particular manufacturer to keep reducing cost is important.
Samsung doesn't have a 16nm NAND process, so it has introduced a relatively high cost 3D NAND process already. Samsung's (OTC:SSNLF) "V-NAND" has been called, "an impressive achievement but not a realistic foundation for the future."
Micron, SanDisk, and Hynix are able to make NAND at 16nm, so they have no motivation to rush to 3D. 3D NAND requires around 50% more clean room space for equipment and also has quite a few manufacturing steps, as well as more complex testing and packaging requirements. It only becomes cost-effective as increasingly more layers are stacked on the die, something that will take more R&D to accomplish.
Samsung's new Xi'an fab provides it the extra space needed for 3D. Hynix and Toshiba/SanDisk are demolishing old fabs to build newer replacements, a time consuming process. Micron has spare space in Singapore for the extra equipment that 3D needs.
How are things right now?
In DRAM, better than anticipated. During the last quarter's conference call, Micron predicted the next quarter would be down low single digits for both production and bit pricing. By contrast, Inotera, Micron's partially owned subsidiary, recently announced that their wafer output will likely increase 5% for the April/May/June quarter when they'd previously predicted 1%. They noted that prices are also rising, and their GM will beat the already record 54% they achieved last quarter. Note that Micron's quarter is March/April/May, so it's unclear how much of that benefit will be seen in this quarter and how much in next quarter.
Spot prices have been continuously hitting new highs in PC DRAM, while mobile DRAM contract prices are stable. In recent presentations Micron representatives have mentioned opportunistically increasing PC DRAM production from 30% to 40% this quarter in order to capitalize on current strong pricing, and Micron's DRAM inventory sits at less than a week, a negligible amount due to shortages. DRAM pricing is likely to remain strong for the rest of the year.
NAND pricing has stabilized at acceptable, but unexciting levels. DRAM makes up approximately 70% of revenue for Micron, while NAND makes up 25%, according to a recent presentation.
The memory industry has typically been one of cyclical losses and capital destruction. How do I really know that this time is different?
This has been answered a dozen different ways. I could talk about the pricing power of oligopolies as the market participants have shrunk from about 40 to 4 (3 in DRAM.) I could talk about the billions of dollars, thousands of patents, and years of experience needed to build even a single fab leaving a very wide moat excluding new entrants. I could talk about Samsung's "crisis culture" (a fascinating read) that has left it willing to buy nearly every kind of electronic component from its competitors even though it already has a dominant position in all of them, as well as Samsung's motivation to keep component costs high in order to hurt the margins of competitors in various electronics industries. Each of those explanations should be (or has already been) an article in itself. The ultimate reason is common sense. Each company is unable to get more than a minimal amount of new market share since there are no weak producers to steal it from. Building new fabs is expensive and corporations are owned by shareholders that expect returns on capital -- oftentimes the executives themselves are shareholders. Going back to my analogy of 10 patients and 9 doses of medicine, if the seller finds a 10th or 11th dose, from a financial perspective he's better off destroying them to create a shortage. He certainly has no motivation to spend a fortune in order to make 20 doses for these 10 people. While an individual might be that irrational, it's that much less likely that a group will, at least for any length of time. A strong pricing environment that lasts indefinitely isn't guaranteed, but it's extremely likely.
Today's Micron bonus bombshell: The dollar-yen is predicted to hit 110 this year, which would add 4 cents per quarter in earnings per share from Elpida cost savings.
Are there other questions I should have included in this list? Let me know in the comments below, and click follow at the top (if you haven't already) to be notified of my next Micron update.
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.
Editor's Note: This article discusses one or more securities that do not trade on a major U.S. exchange. Please be aware of the risks associated with these stocks.