The bad news is that memory suppliers have outdone themselves once again and have stepped off the proverbial cliff [almost on cue]. The good news is that this may be perfect timing for the solid state storage market.
Synoptic Memory Producer Financial Reports
Although Samsung reported a 17 percent net income advance over a year ago on Thursday it cited slowing PC sales as a cautionary on their outlook for the next quarter earnings. The company was quick to add that NAND Flash sales remained robust.
Hynix reported results that were off 1 percent down from second quarter results. Demand for DRAM remained sluggish with Average Selling Price [ASP] dropping 9 percent though bit shipments increased 2 Percent. The company added that NAND flash bit shipments increased 42 percent on ASPs that fell by 23 percent.
Japan?s Elpida Memory Inc. reported strong year-to-year gains for the second quarter and first six months of its fiscal year. The company added that they were experiencing a sales decline due to DRAM price erosion.
Toshiba Corporation reported in October that it would beat its first-half operating profit forecast by more than 43 percent on strong sales of flash memory chips though it said it will be sticking with its full-year outlook. The company made no comments regarding their DRAM memory business.
Taiwan DRAM makers Nanya Technology Corp. and Inotera Memories Inc. recently reported operating losses on declining DRAM ASPs. Nanya?s sales revenue showed a decrease of 5 percent compared to second quarter 2010 and is still 30 percent above year-over-year. The company?s 2-Gbit 50 nm based DDR3 represents 70 percent of the combined production at Nanya and Inotera Memories in the third quarter of 2010.
Micron recently reported a 14 percent decrease in DRAM sales and a 5 percent decrease in NAND sales. An interesting aspect of the report was that the declines were driven primarily by sequential declines in unit sales, while ASPs were much more stable than expected. DRAM ASPs were down 2 percent quarter-to-quarter and NAND flash ASPs were up by about 3 percent quarter-to-quarter. Micron reported that the DRAM decline was mostly driven by a continued capacity volume ramp with Inotera JV in Taiwan which has since been resolved.
PC Sales Fall
The most cited reason for soft memory sales is a shortfall in PC sales worldwide. The question is whether this is a normal end of the year fall off in business [which is normal] or is it a harbinger of darker things to come? It depends upon who you ask though it?s most likely a combination of factors. The crystal ball for 2011 is cloudy at best but the worldwide macroeconomic picture is one of indebted nations learning to live within their means while paying off indebtedness. In two words, slow growth.
The Semiconductor Memory Business Cycle
The semiconductor memory market has historically cycled between excess profits and periods of deep losses. The usual reaction to a down market is to look for ways to fill the fab with products with a higher-than-cost average sales price to survive. It is more a balance sheet argument than anything else. The fab needs to pay for itself and that means wafer starts that sell. What?s on the wafer is somewhat secondary to the fact that a wafer has started. Historically the usual result was to run more DRAM wafers with special tweaks [low power and extended refresh come to mind] along with bin sorts to differentiate the product.
There are several differences in the business environment this time around. The number of NAND flash wafer starts worldwide is larger than its DRAM counterpart by approximately 150 percent [wafer start crossover occurred in 2008-2009]. Flash wafer starts will be approximately twice that of DRAM in 2011 and are forecast to grow to four times by 2015. Manufacturing always loves a high runner and it appears they?ll be greatly motivated by NAND Flash production.
The semiconductor memory industry is entering a period of oligopoly market control where a limited number of suppliers are selling into the market and each is well aware of the actions of his competitors. Decisions by one, influences the others and vice versa. A darker side of oligopoly market control is that manufacturers may tacitly limit risks through synchronized pricing and product development & introduction [i.e. anticompetitive behavior].
There is always a psychological element in market share arrangements where a particular vendor will not be willing to lower ASPs in order to capture market share as doing so would result in a less profitable outcome. In this case it would be best to let your competitor to lose money to gain a dubious amount of market share. It is now best to be fiscally sound long term.
Most large memory consumers now qualify a limited number of top suppliers [usually two]. Tier two need not apply as they are not considered to be financially and technically sound enough to be considered as a long term supplier.
The picture that?s coming into focus is of an industry that has entered an era of very efficient production after a period of chaotic consolidation. There is an air of business confidence despite the DRAM downturn.
Tuned vs. Convertible Fabs
High capacity DRAM fabs are usually "tuned" meaning that they cannot easily be changed to run a different process though some have the ability to run either DRAM or NAND flash memory. Of the top four producers only two are reputed to have the ability to run either DRAM or Flash – but require an indeterminate amount of time to re-establish production. The producer is faced with the dilemma of makin
g the decision to take a line out of DRAM production for flash production. Production shifting has never been that successful and with recent consolidations it?s not considered to be a viable alternative though there will always be those desperate enough to attempt it.
Flash Demand Forecast
The demand for flash is diverging from "appliance" PC demand i.e. rise of smartphones and tablets. Flash target applications requiring low weight, long battery life, instant-on and an attractive form factor represent the fastest growing segment of the consumer "compute" space. Media tablet sales alone are forecast to reach 54 million in 2011 according to Gartner Research. If those sales targets would be achieved, it would mean that 43 percent of the total NAND Flash forecast for 2011… is sold out.
Recent introduction of smart phones, Apple iPad and MacBook Air Gen3 is causing analysts to reassess their recent flash forecasts. The issue [exclusively reported by BSN*] is one of price. The current price of $1.00/GByte is falling to approximately $0.50/GByte in 2011 is popularly thought to depend on a DRAM oversupply situation, driven by a switch to NAND flash production. Of course this may be one of those "imagined" things by the solid-state-disk producers.
Cost Impact Variables
Flash memory producers follow the credo of Moore?s Law of die shrinking to lower cost. Mainline fabs running 300 mm wafers at below 32 nm will increase production. Shrinks play havoc on yield numbers and its unclear how well the technology will yield at the lower geometries [i.e. > 20 nm] though this is not a concern near term.
Historically SLC and MLC designs occupied the same die. The decision to make either Single Level Cell [SLC] or Multi-Level Cell [MLC] was made as a "back-end-of-line" [BEOL] option allowing manufacturing to run one device and marketing the choice of an inventory mix depending on demand. SLC and MLC are now separate designs that allow fine tuning of each.
3-bit MLC now represents better than 96 percent of sales reasserting that cost is always the imperative and identifies the market high runner. Cost is ever-so important in the enterprise as well – we anonymously polled C-level executives from the Solid State Storage industry and the results were quite interesting. It was pointed out that only 4-11% of enterprise customers opt for more expensive SLC memory.
Introduction of Apple?s new products and the viral buzz created on a worldwide basis has many of the traditional PC suppliers implementing crash programs in an effort to field competitive offerings. HP?s slate, which may be the first of a flood of look-a-likes, wasn?t received well indicating that the instant on iPad solution is further out for Wintel pc?s than anyone thinks. The Android-enabled smartphone market is growing faster than anyone can keep track of and promises to be a heavy user of NAND-Flash.
For example, the number of Google Android enabled phones grew by 883% between 2009 and 2010, with triple-digit percentage growth expected to continue in 2011 [est. 300-500%].
In effect, the Apple iPad, MacBook Air and iPhone4 launches introduced an unexpected market expectation for light, long battery life personal computing in an attractive form factor mandating flash as a requirement. When you add products such as Samsung Galaxy, Nintendo 3DS, PlayStation Phone… the overall effect is just not with Apple products but the entire market space of mobile computing.
Outlook & Summary
Though the market for DRAMs has fallen off there?s no reason for panic. It?s well within the normal seasonal bounds of the business. The expectation is for a down market at least till the end of the 2Q 2011 and may see further consolidation of those players not able to produce flash. The main players are lined up and prepared with enough upside to absorb a good amount of unexpected demand so pricing should remain stable.
NAND-Flash seems to be taking a brief breather before the "hustle up" begins post CES next year. The sheer number of flash enabled new products will surprise even the most jaded of optimists. Producers seem to be lined up and ready to supply the demand.
2011 could be a very good year for flash producers?