Micron Technology: Jim Handy Is Wrong On The End Of The Cycle

Micron Technology: Jim Handy Is Wrong On The End Of The Cycle Aug. 25, 2017 3:24 PM • MU Summary Jim Handy, "The Memory Guy" is a memory chip analyst who has been predicting the peak of the NAND market for a couple of years. He has long been wrong. The "Greybeards on Storage" podcast #50 interviewed Handy on August 19th in their wrapup of the Flash Memory Summit. Handy and the Greybeards had some interesting factoids in their podcast. Confession: Before becoming a fairly regular listener to the nerdy podcast "Greybeards on Storage," I (almost) thought all there was to advanced memory was soldering NAND chips and a controller down to a board. Who knew what I would learn about mirroring, NVM over fabrics, de-duping, metadata, and other exciting topics. Thanks to the Greybeards, Howard Marks and Ray Lucchesi for my continuing education. On August 19, in their 50th monthly podcast, the Greybeards interviewed Jim Handy, "The Memory Guy," in their Flash Memory Summit wrapup. Handy has been incorrectly predicting peak NAND pricing for some time. Nonetheless, there were some great items bandied about in this interview, per this author's paraphrasing, not exact quotes: The only thing keeping NAND prices up is shortage. [Duh!] 32 layer NAND currently costs about 12 cents per gigabit. At 64 layer, cost will drop to 7-8 cents. [This author's takeaway was that prices might drop more slowly, thus margins should be fine.] Comments on string stacking of 32 layer NAND devices. Samsung (OTC:SSNLF) has a 128Gb die. String stacking can take that to 4 terabytes in a package. 3D XPoint is Phase Change Memory. PCM has been around since 1972 but is still difficult. Semi joke: for 3D XPoint it's not die per wafer, but wafers per useable die. Intel (NASDAQ:INTC) needs to sell Optane, their name for 3D XPoint, for less than DRAM. But to do that they need volume. This is a chicken and egg situation. Intel may lose money for a while on Optane but do well by selling more expensive processors [that can address Optane]. [Pricing armageddon rant beginning at minute 20:50]. Shortages have been caused by two different things. In DRAM, manufacturers didn't add sufficient capacity during a lull in 2014/2015. In 3DNAND, mastering materials never before used in semiconductors is proving difficult. Once mastered, NAND will flip from undersupply to oversupply overnight. Perhaps by mid 2018. NAND prices will drop 60% in two quarters. Planar NAND will no longer be cost effective to produce. Planar manufacturers will then point their capacity toward DRAM and will cause DRAM prices to collapse. DRAM guys will then pursue foundry, which will collapse foundry prices, perhaps by the end of 2018. So what's wrong with Handy's pricing armageddon? Plenty! Handy is too glib on timing changes. Contract pricing has lengthened out per Bernstein and other analysts. Plus the recent price surges are going to cause some caution - buyers are going to be a bit more cautious this time fearing that another price ramp up is around the corner. Prices are going to be a bit stickier this time. Handy would do well to listen to the earnings calls of Apple (AAPL), Cisco (CSCO), and Hewlett Packard (HPE) to get some handle on what pricing looks like with the customer's hat on. Second, despite his wave of the hand, changeovers from planar NAND to DRAM, and from DRAM to foundry take months. Then after the equipment is in, there are months of qualification of first parts down a new line. His cascade of pricing crumbling in different sectors could happen but it's going to take a lot longer than he points out. Perhaps most importantly, the NAND market is very elastic. Once pricing begins to come down, demand is going to surge. Manufacturers will find it tough to keep up with the year plus it takes to build a fab shell, and the months it takes to fill that shell with equipment and get yields up. I don't think manufacturers are going to have the luxury of time to think about changing planar NAND over to DRAM, and DRAM over to foundry. They are going to be hard-pressed to keep putting up more 3DNAND fabs to keep up with demand. And more layers per die, and string stacked dies should allow them to remain profitable even given selling prices, which may indeed decrease rapidly. And what does all this mean for Micron? Missing from Handy's discussion was any evaluation of the different positions of the different manufacturers. #MicronTechnology (MU) has the industry-leading smallest die size in #NAND. Micron has the number one market share position in the fast growing automotive sector (with its very long qualification times.) #Micron and #Intel own the sleeper known as #3DXPoint; perhaps someday it will wake up! And Micron is the cheapest memory company on the planet in terms of valuation multiples. Micron is doing fine and should continue to do well for the next several quarters.
https://seekingalpha.com/article/4102005-micron-technology-jim-handy-wrong-end-cycle