Memory continues to be in a volatile period showing swings along just about every imaginable variable that could impact a commodity in the semiconductor industry. How has the sector fared thus far and what markets will drive demand in the near and mid-term? While memory is certainly a mature sector, there is growth forecasted over the next few years, though intervening variables will play a critical role.
Memory has had a tough go of it over the past couple of years and, unfortunately, there are few if any green shoots here unlike some other sectors. The DRAM and NAND average selling price (ASP) collapses saw some relief from the dramatic utilization cuts by the fabs. The supply reduction helped drive up NAND ASPs to healthier levels, particularly during 1H09, but with inventory restocking underway, utilization back up to an average 80%, profitability is still a struggle for many in memory.
The road ahead for NAND and DRAM
Most analysts right now see ASP declines back on the docket for NAND and DRAM heading into 3Q09 and beyond (cf., Morgan Stanley, Semiconductor Capital Equipment, 22 May 2009):
NAND contract pricing declined 1.9% for a 16Gb MLC chip and by 6.4% for a 32Gb MLC chip during the second half of May. […] [W]e […] view the ASP decline as a potential portentous sign.
[...] [T]he decline in ASPs was driven by discounting on higher density chips as the market enters the traditional slow season. We believe rising capacity utilization at Toshiba and SanDisk was also factored into the price. (UBS Investment Research, Global Memory Semiconductor Insights, 26 May 2009).
Although the sector is still considered weak, profitability is seen as possible for those chip makers who have reduced costs and improved margins. Figures 1 and 2, below, compiled by Morgan Stanley from internal data and SEMI’s data, present one view to this situation. These Figures track NAND and DRAM pricing in comparison to cost bases for the chip makers, respectively. The data support the claim that profitability can be maintained, albeit not at high levels.
Figure 1. NAND Makers Profitability Forecast for 3Q09 (Source: Morgan Stanley Research, Semiconductor Capital Equipment, 22 May 2009, p.4)
Figure 2. DRAM Pricing Forecast for 2H09 (Source: Morgan Stanley Research, Semiconductor Capital Equipment, 22 May 2009, p.5)
Utilization, CAPEX and ASP effects
Importantly, the CAPEX levels for the past 12 to 18 months have been well below industry averages. Further complicated by the various plays among the Taiwanese companies and the uncertainty around the Taiwan government consolidation, Taiwan Memory Company (TMC), the DRAM sector remains weakened and without significant ramps nor investments on the near horizon (cf. also, Barclays Capital, Semiconductor Handbook 2009, p.195). However, DRAM is not expected to loose significant market share simply based on the wide install base, and revenues are expected to pick up very modestly during 2010, parallel to the modest CAPEX increases. “Looking further ahead, our proprietary Barclays Capital bottom-up supply/demand model suggests global DRAM supply to grow moderately at around 9% YoY in CY09 versus 62% YoY in CY08.” (Barclays Capital, Semiconductor Handbook 2009, p.207)
DRAM is a mature technology with decelerating ASPs but continued unit growth, particularly for mobile DRAM (mDRAM) in all smartphones, and more recently it has experienced a supply reduction due to utilization slow downs and the exiting of some Tier II DRAM suppliers as a result of the economic crisis. Based on these factors in addition to current market variables, the consensus for long-term DRAM CAGR is forecasted as steady in the mid-single digits, post 2H10.
Figure 3. Barclays Capital DRAM and NAND CAPEX Forecast 2004-2010E (Source: Barclays Capital, Semiconductor Handbook 2009, p. 195; compiled from Company Data and Barclays Capital Estimate (CJ Muse, Semiconductor Capital Equipment)).
The best data and forecasts in the memory sector are clearly for NAND in the near- and mid-term. While NOR and pSRAM are destined to loose significant ground in handsets, NAND and mDRAM are gaining significant ground, as highlighted by iSuppli’s data in Figure 4, below, covering revenue by memory type in handsets from 2008 through 2013E:
Figure 4. iSuppli Revenue by Memory Type in Handsets, Forecast 2008-2013 (Source: iSuppli, May 2009)
Underscoring the dramatic shift away from NOR and favoring NAND, iSuppli’s data show that for their teardowns during 2007 – 2008, the percent of NOR based handsets was 53% versus NAND based units at 47%. Whereas one year later, 2008 – 2009, only 19% of all handsets were NOR-based versus 81% that were NAND-based. These are dramatic data and lend undeniable proof of the shift going on in the memory sector for all types of handsets.
Stacking up advantages
Another trend seen in memory for handsets, particularly for the growing smartphone sector, i.e., high-end handsets, is the increase in multi-chip packages (MCP) and Package on Package (PoP) solutions. These architectures not only increase speed, decrease space, reduce heat and energy waste, but they also improve performance. Adding more downward pressure to memory’s rank on BOMs, touch screens are now leading the list of drivers for new MIDs and mobile handsets. Stacking is increasing, particularly for NAND and is providing a solution path for the increased memory without compromising other features that users are demanding for data, music and video storage on their handsets and MIDs.
SSD still a waiting game
While NAND stacking improvements are strengthening the much anticipated future for solid-state drive (SSD) technology, the global economic recession has put the brakes on what was seen as initial momentum. The future of SSDs is almost guaranteed, however, the adoption phase is now being pushed out to 2010 through 2013 by most analysts, but with a healthy CAGR of 17% for unit growth. In the short run, for 2009, growth is forecasted to remain in the low single digits, not good numbers for the NAND flash controller market or for the SSD market.
According to iSuppli SSD reports from May and June (cf. recent summary released at http://www.isuppli.com/MarketWatchDetail.aspx?ID=371), analysts do project a positive stabilizing trend by 2013 for SSD adoption. This is good for Flash NAND, particularly multi-level core (MLC), because MLC NAND makes up the majority of an SSD device.
In short, iSuppli among other industry analysts such as Barclays Capital, UBS, and Morgan Stanley, do see the SSD adoption impact stalled because of the current price barriers that increase the end-cost to consumers (see also http://www.computerworld.com/action/article.do?command=viewArticleBasic&articleId=9134468&source=CTWNLE_nlt_dailyam_2009-06-18). Today’s consumers, both retail and enterprise, are too price sensitive and are favoring speed and battery life over SSD storage technology. However, it is not the end for SSD; momentum is forecasted to take hold in 2011, with 2012 being the more worldwide SSD adoption and move to dominant technology. With these forecasts in hand and backed by solid, triangulated data from across the supply chain by various analysts, SSDs are seen as an important future driver for the market, particularly once the breakthrough ASP for SSD controllers is met, likely in 2011.
The memory sector has been in volatile territory for quite a while. With the added blow of the global economic recession, significant downward pressures have lead to some rather dramatic movements: significant inventory pull-backs, historic utilization lows, severe CAPEX reductions, and significant numbers of consolidations, restructurings and notable insolvencies. With this level of sector ‘shake up’, DRAM in particular, has seen a retrenchment in financial support that has translated into more cautious strategic plans leading to more conservative production and R&D by all but the giants. This retrenchment has contributed to the ASP pressures, but there is also a longer-term easing forecasted with profitability seen as resuming for those who are able to manage through these tough cycles.
NAND has followed a generally similar trajectory to DRAM. Although the fine-grained ASP and revenue movements in NAND differ from DRAM, in part due to the proliferation of smartphones and the market share increases from NOR-based handsets, longer term demand is pegged to consumer, retail and enterprise, demand and therefore seen as volatile (recall Figure 3, above).
The memory sector, having entered a mature phase, is forecasted by most analysts to see modest growth from 2H09 through 2013E, though many intervening variables will impact the level and duration of revenue growth cycles.