You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
TMT<br />
<strong>Memory</strong> Devices Technology<br />
October 2010<br />
<strong>Asia</strong> <strong>Memory</strong><br />
<strong>Bad</strong> <strong>memories</strong> <strong>fading</strong><br />
We believe the market has priced in weak 4Q10/1Q11 numbers; valuations remain<br />
suppressed and close to trough levels in some cases. We expect demand-led recovery<br />
through price elasticity starting at the end of this year.<br />
We expect more sustainable profitability for leading vendors next year on rational<br />
capex, slowing ASP declines, plus process migration-led cost reductions and a better<br />
product mix that leverages into structural growth areas. We identify upside risks from<br />
enterprise DRAM and NAND, plus quantify the impact of tablets and smartphones.<br />
We expect stocks to start outperforming by end-4Q10/1Q11 as expectations rise of<br />
slowing DRAM price declines and firmer NAND prices.<br />
We prefer industry leaders with cost efficiency and leverage into high-growth,<br />
high-margin sub-segments. Our top pick is Samsung Electronics OW,<br />
an HSBC <strong>Asia</strong> Super Ten idea. We lower our TP for Hynix OW(V), Elpida OW(V),<br />
Winbond OW(V), Nanya N(V) and Inotera N(V)<br />
By Nam Park and Carolyn Poon<br />
Disclosures and Disclaimer This report must be read with the disclosures and analyst<br />
certifications in the Disclosure appendix, and with the Disclaimer, which forms part of it
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Summary<br />
We believe the market has priced in weak 4Q10/1Q11 numbers;<br />
valuations remain suppressed and close to trough levels in some<br />
cases. We expect a demand-led recovery through price elasticity<br />
starting at the end of this year. We expect more sustainable<br />
profitability for leading vendors next year on rational competition<br />
plus process migration-led cost reductions and a better product mix<br />
that leverages into structural growth areas. We identify upside risks<br />
from enterprise DRAM and NAND, plus quantify the impact of<br />
tablets and smartphones. We expect valuations to turn more positive<br />
by end-4Q10/1Q11 as expectations rise of slowing DRAM price<br />
declines and firmer NAND prices. Our top pick is Samsung<br />
Electronics, an HSBC <strong>Asia</strong> Super Ten idea. We lower our forecasts<br />
and TPs for Hynix OW(V), Elpida OW(V), Winbond OW(V), Nanya<br />
N(V) and Inotera N(V).<br />
Why bother with memory anyway?<br />
We capitulate to higher rates of ASP decline (see Samsung Electronics: Comprehensive health check, 24<br />
September 2010) and lower our forecasts and target prices other vendors accordingly. For all of our<br />
coverage stocks we forecast y-o-y declining operating profits for 2011e. We forecast Elpida and<br />
Taiwanese vendors to make some improvements off low bases. Looking solely at these y-o-y profit<br />
decline comparisons for next year, there appears little reason to bother with the memory sector.<br />
And yet, despite these revisions, valuations remain low and in some cases close to trough levels; we<br />
believe the market has already priced in a weak 4Q10 and possibly a weak 1Q11 quarter as well. On top<br />
of which, we think macroeconomic demand concerns and persistent fears of the memory cycle turning<br />
conspired to suppress memory stocks’ valuations during recent quarters of growth and relatively high<br />
margins. Those same factors were very swiftly evoked to punish memory stocks as soon as ASPs did start<br />
falling. In fact, recent steep DRAM price declines result from an unusual set of circumstances during<br />
1Q/2Q10, namely PC order pull-in (due to EUR weakening and EU demand concerns) plus delays in<br />
process migration at second-tier players (which tightened supply). These factors shored up ASPs in 2Q10,<br />
setting the market up for a sharp ASP decline starting in 3Q. Still, it is worth remembering that we<br />
forecast DRAM ASP to rise 20% this year, and for NAND ASPs to fall only 20% for the full year.<br />
1
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
We believe next year the industry would stabilise towards more sustainable profitability especially for<br />
leading vendors, albeit at lower levels than this year, on more rational competition, lower costs from<br />
process migration and better product mix. We also think that the market is ignoring some of the upside<br />
risks that we analyse in detail here, including enterprise DRAM and NAND growth, plus the extent of the<br />
impact of tablets and smartphones on NAND.<br />
We expect sentiment towards the industry to turn more positive by end-4Q10 or early 1Q11 as<br />
expectations rise of easing of DRAM ASP pressure by 2Q11; we think price elasticity will restart stunted<br />
growth in DRAM content/system, whilst accelerate higher NAND content/system by end-2010. This<br />
should become evident in PC OEM results and market data. For NAND, key catalysts include firmer than<br />
expected ASPs starting in 4Q10/1Q11e, reflecting strong demand especially from smartphones. ASPs<br />
could also firm as larger density demand rises towards 1Q-2Q11 on a raft of tablet releases.<br />
More sustainable earnings, for leaders<br />
A key argument we made in our previous sector report (<strong>Asia</strong> <strong>Memory</strong>: Seeing is believing, March 2010)<br />
to support a positive view on the memory industry was that more rational competition/capex will prevail<br />
as memory vendors remain wary of overcapacity and focus instead on profitability to strengthen balance<br />
sheets. We think this argument remains largely valid. We believe leading vendors remain acutely wary of<br />
market conditions and may well adjust expansion rates accordingly. A ‘healthy level of profitability’,<br />
coupled with constrained capex/sales remains a key rerating driver for the industry.<br />
DRAM operating profit margins<br />
60%<br />
40%<br />
20%<br />
0%<br />
-20%<br />
1Q09 2Q09 3Q09 4Q09 1Q10 2Q10 3Q10 4Q10 1Q11 2Q11 3Q11 4Q11<br />
-40%<br />
-60%<br />
-80%<br />
-100%<br />
-120%<br />
-140%<br />
SEC Hy nix Elpida<br />
Nany a Tech Inotera Winbond<br />
Source: Company data, HSBC estimates<br />
We believe leading vendors place greater focus on profitability through process migration, rather than<br />
market share gains. We have seen little evidence, for example, of Samsung attempting to crowd out<br />
competition through aggressive pricing of this year. Instead, we think market leaders, able to rapidly<br />
migrate processes and develop new, higher-end products to lower costs and improve product mix to<br />
insulate margins, seek to generate more sustainable returns. We forecast Samsung and Hynix to generate<br />
2
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
20-26% operating margins (higher if we consider just Samsung’s memory operations) during 2011-12,<br />
while managing capex/sales below their respective historical averages.<br />
It is worth stressing that the sharply higher proposed capex by vendors (which to a large extent, the<br />
market anticipated) is not excessive, in our view. For the industry, we forecast 2010-12 capex/sales of 26-<br />
28% for DRAM and 48-51% for NAND. This is well below 2005-08 averages of 54% and 67% for<br />
DRAM and NAND respectively. Additional wafer capacity in 2010-12 is mostly earmarked for NAND<br />
(21-22%/year wafer growth for NAND, 3.7-7.9%/year for DRAM) without which severe capacity<br />
constraints could arise. Other than NAND wafer adds, capex is likely to be funnelled mostly into process<br />
migration. Supporting our thesis, Hynix sees longer-term capex/sales at c30% or lower, plus a normalised<br />
operating margin of 15-20% over the longer term (see Hynix – Raising target price: Confirming our<br />
thesis, 23 April 2010).<br />
Upside risk lurks<br />
Enterprise computing, storage<br />
Retail consumer demand remains uncertain, but corporate IT forges ahead. We think the enterprise server<br />
and storage market provide substantial upside risk for leading vendors that are positioned to cater to this<br />
high-end market. Continued corporate IT spend plus the ongoing growth of cloud computing drives<br />
strong demand for servers. DRAM beneficiaries include Samsung and Hynix. We forecast server unit<br />
growth of 10% y-o-y in 2011, and content/box growth of 38%. Servers account for about 10% of total<br />
global bit demand but blended ASP is far higher for vendors, at 2.5x for a 4GB module up to 6x for a<br />
16GB module. The server storage market also benefits from disproportionately high revenues and<br />
margins. Key vendors include Samsung, IMFT and Hynix. Server storage could account for the<br />
equivalent of 12-17% of our 2012 forecast global NAND revenue, and the equivalent of 20-32% of our<br />
2013 forecast NAND global revenue.<br />
Tablets<br />
We forecast tablet shipments to soar to 73m units next year, accounting for a quarter of our forecast<br />
tablets plus notebooks. We estimate that netbook and notebook sales cannibalisation will lower 2010-12e<br />
global DRAM demand by 1%, 5% and 5%. This will be partially offset for DRAM vendors such as<br />
Samsung and Elpida that supply low-voltage, low-power parts that could be integrated into tablets. We<br />
expect tablets to account for 13% of global NAND demand next year (2010: 5%), mostly driven by a<br />
massive 214% 2010/12e bit growth rate. We think Windows 7 could help drive tablet GB/system. Key<br />
beneficiaries from tablet-related memory growth are Toshiba and Samsung.<br />
Smartphones<br />
Mobile handsets have also emerged as the leading usage of NAND on proliferation of smartphones. Handsets<br />
account for 37% of our 2011 NAND bit demand forecast, growing to 42% of total NAND bit demand by 2012.<br />
We forecast smartphone unit shipments to rise 54%, 39% and 35% during 2010-12, accounting for 35% of<br />
total handsets shipped globally by 2012. Smartphones drive rapid mobile DRAM bit growth through unit<br />
shipment growth and content per box. Mobile DRAM prices tend to be higher on a bit basis, as memory<br />
vendors incorporate custom-made solutions for specific vendors. Samsung, Elpida and Hynix are leaders in<br />
mDRAM. For NAND, we forecast GB/system for smartphones to double each year during 2011-12, driven by<br />
3
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
lower NAND prices, the availability of larger capacity mobile NAND, rapidly rising screen resolutions, higher<br />
resolution video, games content and cannibalisation of media players.<br />
Worth a look? When and what?<br />
Timing – over the next 1-2 quarters<br />
Despite lowering forecasts for our companies, valuations are close to trough levels for many stocks. Has the<br />
market priced in all industry-wide negativity? For commodity DRAM, it is close to doing so – a weak 4Q10<br />
earnings outlook for industry participants, possibly extending out to 1Q11, is the consensus view now.<br />
We expect to see sentiment starting to turn more positive by end-4Q10 or early 1Q11 as potential easing<br />
of DRAM ASP declines by 2Q11 starts pricing in; the PC market would require a couple of quarters to<br />
absorb Taiwanese vendors’ bit growth during 4Q10/1Q11. By the end of this year we think price<br />
elasticity will restart stunted growth in DRAM content/system. This should become evident in PC OEM<br />
results and market data, boosting confidence that future DRAM ASP declines could be more muted.<br />
For NAND, key catalysts would include firmer-than-expected ASPs starting in 4Q10/1Q11e. This would<br />
reflect soaring demand especially from smartphones. Firmer ASPs could also arise from orders of larger<br />
density units towards 1Q-2Q11 as tablet releases accelerate in 2H11. By 3Q11 we could see prices<br />
softening more as Samsung and Toshiba start mass production at new fabs in earnest.<br />
Stock coverage and ratings summary<br />
Company Ticker Rating New TP<br />
(LC)<br />
CP (LC)<br />
Potential<br />
return (%)<br />
Previous<br />
TP<br />
Change<br />
(%)<br />
Previous<br />
rating<br />
Change<br />
Samsung Electronics 005930 KS OW 1,139,000 793,000 44% 1,139,000 0% OW -<br />
Hynix Semiconductor 000660 KS OW(V) 32,000 23,400 37% 36,000 -11% OW(V) -<br />
Elpida <strong>Memory</strong> 6665 JP OW(V) 1,500 928 62% 3000 -50% OW(V) -<br />
Nanya Technology 2408 TT N(V) 20 19.3 4% 25 -20% N(V) -<br />
Inotera Memories 3474 TT N(V) 17 16.2 5% 20 -15% N(V) -<br />
Winbond 2344 TT OW(V) 10.8 8.03 34% 12.8 -16% OW(V) -<br />
Source: HSBC<br />
Go for the leaders<br />
We expect memory vendor prospects and valuations to diverge as leaders accelerate process migration<br />
and improve product mix. Increasingly DRAM and NAND bit growth are driven by mobile, enterprise<br />
applications that demand high-performance parts fabricated on advanced processes.<br />
Our top sector pick is Samsung Electronics (OW, 44% upside potential to TP) based on its unparalleled<br />
overall memory positioning and compelling valuations. It remains the industry leader in cost efficiency<br />
and is highly leveraged into high-growth, high-margin sub-segments in both DRAM and NAND. Its<br />
diversified technology portfolio – especially its telecoms operations – provide portfolio defensiveness,.<br />
We think 1.3x 2011e PB (historic range: 1.1-2.5x) is compelling.<br />
Of the pure plays, we find Hynix (OW(V), 37% upside potential to TP) attractive. Valuations are<br />
undemanding at 1.3x 2011e PB (historic range: 0.5-2.5x) compared to its 20%-plus ROE, driven by<br />
strong positioning in high-end DRAM and rapid catch-up with NAND leaders Toshiba and Samsung with<br />
its 2xnm product (that is customer sampling at the moment). Legacy concerns over its balance sheet are<br />
4
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
<strong>fading</strong> on strong cash flow. Stock catalysts include a creditor stake sale to a major strategic buyer, major<br />
customer wins with new DRAM and NAND products and better than expected quarterly results.<br />
Elpida (OW, 62% upside potential to TP) is a high-beta pure play, in our view, with the most upside to<br />
TP, trading on FY12e (end-March 2012) PB of just 0.5x. It looks oversold on slow process migration and<br />
strong JPY concerns. Valuations also seem to have reflected potential dilution from further capitalraising,<br />
especially as it looks to take strategic stakes in Taiwanese DRAM vendors. We like its strategy of<br />
manufacturing mobile DRAM (in which it has substantial market share) mostly in Japan whilst<br />
completely migrating to a relatively competitive 4xnm process at its Taiwanese fab which focuses on<br />
commodity DRAM.<br />
Of Taiwanese memory vendors we like Winbond (OW(V), 34% upside potential to TP). The stock has<br />
held up reasonably, but trades on 2011e PB of 0.7x (historic range: 0.2-1.2x). We like its diversified niche<br />
portfolio that is heavily exposed to mobile and speciality applications. We expect process migration and<br />
flexible product mix to sustain margins. We expect Nanya and Inotera’s price performance to remain<br />
constrained until signs of improvement in node migration and cycle times become tangible. We remain<br />
Neutral (V) on these two stocks.<br />
5
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Contents<br />
Industry snapshot, themes 7<br />
Demand drivers – DRAM 14<br />
Demand drivers – NAND 25<br />
Supply drivers – DRAM 41<br />
Supply drivers – NAND 47<br />
Valuations, ratings 55<br />
Company profiles 59<br />
Elpida <strong>Memory</strong> (6665) 60<br />
Hynix Semiconductor (000660) 67<br />
Inotera Memories (3474) 74<br />
Nanya Technology (2408) 79<br />
Samsung Electronics<br />
(005930) 84<br />
Winbond (2344) 93<br />
Disclosure appendix 98<br />
Disclaimer 102<br />
6
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Industry snapshot, themes<br />
DRAM set for oversupply until 2Q11, until price elasticity offsets<br />
flood of bit supply. NAND demand-supply looks more positive<br />
Higher-end DRAM and NAND producers to benefit from strong<br />
structural demand growth from mobile, enterprise computing<br />
Relentless process migration and higher-end product mix benefit<br />
leading vendors<br />
Demand & supply forecast (USDm)<br />
2008 2009 2010e 2011e 2012e 1Q10e 2Q10e 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e<br />
<strong>Memory</strong> Revenue (USDm)<br />
DRAM 23,988 42,890 43,497 46,497 44,355 9,856 11,290 10,934 10,811 10,413 10,590 11,118 11,377<br />
NAND 11,857 16,238 21,384 24,834 20,498 3,569 4,066 4,234 4,369 4,545 5,208 5,758 5,872<br />
Total 35,844 59,128 64,881 71,332 64,853 13,424 15,356 15,168 15,180 14,958 15,798 16,876 17,249<br />
Wafer Capacity (12” equ. kwpa)<br />
DRAM 16,985 18,206 18,881 20,366 21,341 1,489 1,489 1,514 1,569 1,569 1,569 1,569 1,569<br />
NAND 9,194 10,476 12,529 15,129 15,849 829 867 878 918 948 978 1,095 1,155<br />
<strong>Memory</strong> Shipment<br />
DRAM (1Gb eq.) 10,413 15,488 23,071 32,990 42,523 3,338 3,557 3,936 4,657 5,005 5,423 6,024 6,619<br />
NAND (16Gb eq.) ,457 2,492 5,335 9,921 12,599 465 575 671 780 924 1,174 1,492 1,744<br />
<strong>Memory</strong> ASP<br />
DRAM (1Gb eq.) 2.30 2.77 1.89 1.41 1.04 2.95 3.17 2.78 2.32 2.08 1.95 1.85 1.72<br />
NAND (16Gb eq.) 4.07 3.26 2.00 1.25 0.81 3.83 3.53 3.16 2.80 2.46 2.22 1.93 1.68<br />
<strong>Memory</strong> sales change (%)<br />
DRAM 0% 79% 1% 7% -5% 7% 15% -3% -1% -4% 2% 5% 2%<br />
NAND -3% 37% 32% 16% -17% -5% 14% 4% 3% 4% 15% 11% 2%<br />
Total -1% 65% 10% 10% -9% 4% 14% -1% 0% -1% 6% 7% 2%<br />
Wafer Capacity change (%)<br />
DRAM -1% 7% 4% 8% 5% 0% 0% 2% 4% 0% 0% 0% 0%<br />
NAND -5% 14% 20% 21% 5% 1% 5% 1% 5% 3% 3% 12% 5%<br />
Bit Shipment change (%)<br />
DRAM 22% 49% 49% 43% 29% 5% 7% 11% 18% 7% 8% 11% 10%<br />
NAND 41% 71% 114% 86% 27% 4% 24% 17% 16% 18% 27% 27% 17%<br />
<strong>Memory</strong> ASP change (%)<br />
DRAM -18% 20% -32% -25% -26% 2% 7% -12% -16% -10% -6% -5% -7%<br />
NAND -32% -20% -38% -38% -35% -9% -8% -11% -11% -12% -10% -13% -13%<br />
Source: IDC, HSBC estimates<br />
7
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
DRAM – price elasticity drives<br />
higher content/box<br />
We expect DRAM price declines to start slowing in<br />
2Q11 as lower ASPs from sharply higher supply<br />
during 4Q10-1Q11 stimulate higher content/system<br />
and as supply increases peak out. Overall, we expect<br />
content/system to rise 33% next year, following a<br />
27% y-o-y rise in 2010; high ASPs this year<br />
depressed content growth this year.<br />
DRAM demand and ASPs<br />
2009 2010e 2011e 2012e<br />
ASP (USD) -18% 20% -32% -25%<br />
Content/system (MB) 27% 27% 33% 32%<br />
Units shipments (m) 2% 14% 7% 9%<br />
Demand (MB bn) 30% 46% 42% 43%<br />
Source: HSBC estimates<br />
Bit growth in the coming quarters could surprise<br />
on the upside. We expect PC OEMs to gradually<br />
start raising content/box starting in 3Q on steep<br />
price declines; 2GB DDR3 modules now cost<br />
around USD36, 23% below the peak of USD46.5<br />
in May. We expect price elasticity to kick in<br />
meaningfully in 4Q10 on persistent DRAM<br />
pricing weakness arising from expanded supply.<br />
We model 18% q-o-q average ASP decline for<br />
DDR3 contract prices in 4Q10, implying an<br />
average price of USD32 per 2GB module for the<br />
quarter. At that level, 4GB content/system on<br />
desktops should accessible to the majority of PC<br />
OEMs. By 1Q11, 2GB modules could have fallen<br />
to around USD28, starting to drive real increase in<br />
content/system above the 4GB level. For higherend<br />
PCs, we expect content/box climbing to 6-<br />
8GB next year to competitively differentiate.<br />
We also expect accelerating adoption of higherend<br />
4GB modules based on 4x-5xnm 2Gb parts.<br />
Leading DRAM vendors such as Samsung and<br />
Hynix that are rapidly shifting production of 2Gb<br />
parts to 4xnm and below from 5xnm will benefit<br />
the most from this trend.<br />
NAND<br />
NAND remains an exciting high-growth segment.<br />
We forecast 2010-12 NAND bit demand CAGR<br />
of 103%, rising 114% next year. Global appetite<br />
remains insatiable for storage capacity that is fast,<br />
robust and low-power.<br />
NAND demand forecast<br />
45,000<br />
120%<br />
100%<br />
30,000<br />
80%<br />
60%<br />
15,000<br />
40%<br />
20%<br />
-<br />
0%<br />
2009 2010f 2011f 2012f<br />
Demand ( GB, m) Change, y-o-y (%)<br />
Source: IDC, HSBC estimates<br />
Foreseeable demand remains strong on rising<br />
affordability, expanding applications in high<br />
potential markets such as smartphones, tablets, PC<br />
storage plus nascent markets such as automotive<br />
applications. We continue to believe the risk is on<br />
the upside, as demand accelerates on more rapid<br />
affordability trends.<br />
Unlike DRAM, NAND demand remains strong<br />
for both commodity parts and embedded solutions.<br />
The main embedded demand driver is handsets,<br />
led by smartphones, which we estimate will<br />
account for 37% of total bit demand next year, up<br />
from 28% in 2009. Another embedded solutions<br />
driver is tablets, a segment which practically did<br />
not exist until April this year. We expect SSD<br />
growth, which had been stalled this year by high<br />
ASPs to see rapid growth due to falling NAND<br />
prices and indirect demand stimulation through<br />
the ‘tablet effect’.<br />
8
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
High-end computing DRAM MB bn, % of total DRAM demand<br />
PC server DRAM MB bn, % of total PC DRAM demand<br />
500<br />
12%<br />
400<br />
20%<br />
400<br />
300<br />
200<br />
100<br />
11%<br />
11%<br />
10%<br />
10%<br />
9%<br />
300<br />
200<br />
100<br />
15%<br />
10%<br />
5%<br />
-<br />
9%<br />
-<br />
0%<br />
2009 2010f 2011f 2012f<br />
2009 2010f 2011f 2012f<br />
High-end com puting<br />
% of total<br />
x 86 serv ers<br />
% of total PC<br />
Source: IDC, HSBC estimates<br />
Source: IDC, HSBC estimates<br />
Enterprise<br />
Enterprise flash drives<br />
We think the enterprise server and storage market<br />
provide tangible upside risk to revenue and<br />
margin forecasts for leading vendors that are<br />
positioned to cater to this high-end market.<br />
Enterprise server DRAM demand<br />
Servers – niche, but very high ASP market<br />
We expect continued corporate spend plus<br />
ongoing growth of cloud computing to drive<br />
strong demand from this segment. Key<br />
beneficiaries include Samsung and Hynix.<br />
We forecast server unit growth of 10% y-o-y in<br />
2011, but higher content/box growth of 38%;<br />
server memory content/system tends to double<br />
every two years. Servers (including PC servers)<br />
account for about 10% of total global bit demand,<br />
but the blended ASP impact is far higher for<br />
vendors, at 2.5x for a 4GB module up to 6x for a<br />
16GB module. Price premiums are driven by<br />
high-density, high performance and lower power<br />
consumption, in addition to higher production<br />
costs. Demand will be further catalysed by<br />
availability of high-density (2Gb and 4Gb very<br />
soon), low-power DRAM modules.<br />
A major upside risk to high-end SSD demand<br />
Similarly, the server storage market offers<br />
disproportionately high revenues and margins.<br />
Key beneficiaries of this market include Samsung,<br />
IMFT and Hynix, to be joined next year by<br />
Toshiba. We do not separately model the server<br />
NAND market in our forecasts but this analysis<br />
illustrates the potential upside risk to our forecasts.<br />
High-end server SSD storage is very expensive.<br />
Despite this, data centres and corporates use SSD<br />
due to 1) extremely fast data input/output,<br />
especially important for cloud computing, 2)<br />
small form factor (space), 3) low power<br />
consumption. As ASPs fall, we expect SSDs will<br />
replace HDDs as ‘front line’ storage, with HDDs<br />
acting as inexpensive backup.<br />
We estimate that server storage could account for<br />
the equivalent of 12-17% of our 2012 forecast<br />
global NAND revenue, and the equivalent of 20-<br />
32% our forecast 2013 NAND global revenue.<br />
These estimates assume SSDs account for 5% and<br />
10% respectively of 2011-12 disk storage systems<br />
and are based on SLC.<br />
9
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Smartphones<br />
Telecoms is becoming a major consumer of<br />
DRAM, accounting for 9 % of DRAM bit demand<br />
this year and growing to 10% in 2012e. Mobile<br />
handsets have also emerged as the leading usage<br />
of NAND on proliferation of smartphones.<br />
Handsets account for 37% of our 2011 NAND bit<br />
demand forecast, growing to 42% of total NAND<br />
bit demand by 2012. We forecast smartphone unit<br />
shipments to rise 54%, 39% and 35% during the<br />
periods, accounting for 35% of total handsets<br />
shipped globally by 2012e (2009: 16%).<br />
Mobile DRAM<br />
Smartphones are driving rapid mobile DRAM bit<br />
growth through both unit shipment growth and<br />
content per box. Development of smartphones<br />
now mirrors that of PCs, driven by resourceintensive<br />
wireless broadband, multimedia and<br />
multitasking. We expect DRAM content to rise in<br />
parallel with surging computing power to support<br />
these demands as well as next-iteration operating<br />
systems such as Android 3.0.<br />
Mobile DRAM prices tend to be higher on a bit<br />
basis, as memory vendors need to work closely<br />
with mobile phone makers to incorporate custommade<br />
solutions for specific handsets. Samsung,<br />
Elpida and Hynix are leaders in mDRAM. The<br />
higher-end mDRAM market (i.e. 128MB<br />
upwards) is dominated by major DRAM players<br />
that are capable of customising MCP (multichip<br />
packages) for handset vendors. We forecast<br />
mDRAM bit growth to rise 102%, 92% and 112%<br />
y-o-y for 2010-12. We assume DRAM<br />
content/system for smartphones to rise 31%, 39%<br />
and 57% for 2010-12, to around 390MB by 2012.<br />
Mobile NAND<br />
We forecast GB/system for smartphones to double<br />
each year during 2011-12 driven by lower NAND<br />
prices, availability of larger capacity mobile<br />
NAND, rapidly rising screen resolutions, higher<br />
resolution video, games content and<br />
cannibalisation of media players. We estimate<br />
NAND on a 16GB Apple iPhone 4 to account for<br />
14% of BOM, a level which we think will be<br />
maintained going forward. We assume most non-<br />
Apple smartphones will use less NAND than<br />
Apple to lower prices, obliging the consumer to<br />
buy flash cards to fulfil extra storage needs.<br />
10
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
% of total NAND bit demand from tablets System unit and GB/system growth for tablets<br />
15<br />
10<br />
5<br />
0<br />
120<br />
100<br />
80<br />
60<br />
40<br />
20<br />
0<br />
60<br />
50<br />
40<br />
30<br />
20<br />
10<br />
0<br />
2009 2010f 2011f 2012f<br />
2009 2010f 2011f 2012f<br />
% of NAND demand<br />
Sy stem units (m)<br />
GB/sy stem<br />
Source: HSBC estimates<br />
Source: HSBC estimates<br />
Tablets<br />
We forecast tablet shipments to soar to 73m units<br />
next year, accounting for a quarter of our forecast<br />
tablets plus notebooks. Tablets represent a major<br />
new NAND application that has explosively<br />
entered the scene.<br />
DRAM<br />
An emerging concern for commodity DRAM is that<br />
tablets will cannibalise netbook and notebook sales,<br />
reducing overall DRAM demand. We estimate 1%,<br />
5% and 5% cannibalisation impact on global DRAM<br />
demand for 2010-12e. DRAM vendors such as<br />
Samsung and Elpida that can supply low-voltage,<br />
low-power DDR (and increasingly DDR2) that<br />
could be integrated into tablets stand to benefit.<br />
The cannibalisation impact will really start to be felt<br />
next year, when we estimate tablets will account for<br />
26% of tablet plus notebook units but only c6% of<br />
tablet plus notebook DRAM bits. We expect DRAM<br />
content in tablets to start rising quickly next year to<br />
support multi-tasking and more resource-intensive<br />
apps. We think this is possible, given DRAM is only<br />
c3% of BOM for current Apple iPads. We think 5-<br />
6% is more likely next year as 512MB becomes<br />
standard. By 2012e we expect tablet DRAM would<br />
reach above 1GB per system, catalysed by Windows<br />
and competitive pressure amongst tablet vendors to<br />
differentiate their products.<br />
NAND<br />
We expect tablets to account for 13% of global<br />
NAND demand next year (2010: 5%), mostly driven<br />
by a massive 214% 2010/12e bit growth rate from a<br />
standing start this year. We expect GB/system to<br />
initially rise slowly as tablets will mostly run<br />
compact Android and iOS operating systems. Key<br />
beneficiaries are Toshiba and Samsung.<br />
For Apple’s iPad NAND as a proportion of<br />
hardware-only BOM varies between 10% and 30%<br />
depending on the model. We expect falling ASPs to<br />
drive content/system; 16GB is about the minimum<br />
acceptable onboard NAND storage for a tablet in our<br />
view. Vendors should have little problem meeting<br />
this requirement this year. Next year, as ASPs fall to<br />
cUSD1/GB, 32GB of NAND would account for<br />
18% of BOM, which we think is a reasonable level.<br />
Interestingly, we think Windows 7 could drive<br />
tablet GB/system. Windows could mount a belated<br />
but serious challenge in the tablet market, as tablet<br />
hardware that could support resources-intensive<br />
Windows becomes available including Intel’s new<br />
Oak Trail processor which is set to appear in<br />
devices starting 1H11. For Windows tablets, we<br />
think there would be a larger NAND requirement as<br />
a full implementation of basic Windows 7 needs a<br />
bare minimum 16GB of storage.<br />
11
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
EBIT margins for memory manufacturers<br />
60%<br />
40%<br />
20%<br />
0%<br />
-20%<br />
-40%<br />
-60%<br />
-80%<br />
-100%<br />
-120%<br />
-140%<br />
-160%<br />
Samsung Hy nix Elpida Toshiba Pow erchip Nany a Inotera ProMOS Micron SanDisk<br />
2006 2007 2008 2009f 2010f 2011f<br />
Source: HSBC estimates, Bloomberg consensus<br />
Revenue, EBIT for memory manufacturers<br />
Revenue (USDm) SEC Hynix Elpida Nanya Tech Inotera Winbond<br />
2007 17,306 8,605 3,877 1,610 1,396 746<br />
2008 22,353 6,818 3,557 1,151 1,190 581<br />
2009 26,808 7,906 3,927 1,284 1,098 594<br />
2010e 37,461 12,429 7,055 1,899 1,450 1,020<br />
2011e 40,639 13,217 6,662 2,538 2,013 1,040<br />
2012e 44,382 14,723 6,969 3,072 2,472 1,141<br />
Revenue changes (%) SEC Hynix Elpida Nanya Tech Inotera Winbond<br />
2008 29% -21% -8% -28% -15% -22%<br />
2009 20% 16% 10% 12% -8% 2%<br />
2010e 40% 57% 80% 48% 32% 72%<br />
2011e 8% 6% -6% 34% 39% 2%<br />
2012e 9% 11% 5% 21% 23% 10%<br />
EBIT margin (%) SEC Hynix Elpida Nanya Tech Inotera Winbond<br />
2007 25% 6% 3% -17% 7% -41%<br />
2008 10% -28% -33% -66% -48% -46%<br />
2009 8% 2% -17% -37% -27% -31%<br />
2010e 28% 30% 21% -6% -8% 11%<br />
2011e 20% 26% 8% 8% 8% 7%<br />
2012e 19% 22% 9% 11% 13% 10%<br />
EBIT (USD) SEC Hynix Elpida Nanya Tech Inotera Winbond<br />
2007 4,340 492 134 (273) 96 (303)<br />
2008 2,143 (1,920) (1,175) (765) (573) (268)<br />
2009 2,059 191 (648) (482) (293) (182)<br />
2010e 10,508 3,741 1,454 (114) (117) 111<br />
2011e 8,328 3,379 563 199 171 75<br />
2012e 8,249 3,296 616 336 315 111<br />
EBIT changes (y-o-y%) SEC Hynix Elpida Nanya Tech Inotera Winbond<br />
2008 -51% nm nm nm nm nm<br />
2009 -4% nm nm nm nm nm<br />
2010e 410% 1858% nm nm nm nm<br />
2011e -21% -10% -61% nm nm -33%<br />
2012e -1% -2 9% 69% 84% 48%<br />
Source: Company, HSBC estimates, note SEC is for consolidated Semiconductor division figures<br />
12
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Samsung DRAM product mix by revenue, 2Q10<br />
Mobile<br />
20%<br />
Hynix DRAM product mix by revenue, 2Q10<br />
Mobile<br />
13%<br />
PC<br />
45%<br />
Serv er<br />
20%<br />
PC<br />
45%<br />
Serv er<br />
19%<br />
Graphic<br />
Specialty<br />
5%<br />
Graphic<br />
10%<br />
Specialty<br />
12%<br />
11%<br />
Source: Company, HSBC estimates<br />
Source: Company, HSBC estimates<br />
Elpida DRAM product mix by revenue, 2Q10<br />
Premier<br />
25%<br />
Nanya DRAM product mix by revenue, 2Q10<br />
Specialty<br />
15%<br />
Serv er<br />
10%<br />
PC<br />
75%<br />
PC<br />
75%<br />
Source: Company, HSBC estimates . Premier includes specialty and mobile DRAM<br />
Source: Company, HSBC estimates<br />
Winbond DRAM product mix by revenue, 2Q10<br />
PC<br />
2%<br />
Mobile<br />
18%<br />
Specialty<br />
40%<br />
NOR<br />
28%<br />
Source: Company, HSBC estimates<br />
Graphic<br />
12%<br />
13
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Demand drivers – DRAM<br />
We forecast 2010/12 bit CAGR of 46% as price elasticity starts<br />
kicking in on ASP declines<br />
PC bit growth from higher per box content; smartphone DRAM set<br />
to grow on higher content<br />
Leading vendors best positioned to harness high-end demand<br />
from low-voltage, high-performance mobile and server application<br />
Time to look at the upside risk<br />
The market now generally expects ASP declines<br />
for the coming 2-3 quarters, driving negative<br />
sentiment and potentially poor share price<br />
performance. Here, instead of focussing on widely<br />
accepted negatives, we focus on upside risks to<br />
forecasts, i.e. factors that could potentially turn<br />
around operating and hence share price<br />
performance. We expect DRAM demand to<br />
remain healthy, growing at a 46% CAGR for<br />
2010/12e. For this year and next year, we expect<br />
48% and 47% y-o-y bit demand growth<br />
respectively. Driving this demand are 1) price<br />
elasticity on sharply lower ASPs, mobile handsets<br />
led by smartphones and 3) strong continued<br />
corporate demand for PCs and servers.<br />
Price elasticity to support DRAM bit growth.<br />
Near term overall PC demand remains relatively<br />
unclear, weighed down by weak consumer<br />
sentiment in developed markets. However, we<br />
expect corporate demand to provide support for<br />
demand as a whole. For the global PC DRAM<br />
market, we forecast content/box rising 33% y-o-y<br />
for 2011, having slowed to c27% this year<br />
(excluding tablet PCs). We expect PC OEMs to<br />
start raising content/box again in 4Q10 as 2GB<br />
modules fall below USD30/2GB. Indeed, 1Gb<br />
DDR3 contract prices for 2H September fell<br />
below USD2 already for the first time since 2H<br />
October 2009. We expect DDR3 contract ASP to<br />
continue falling steeply in 3Q10-1Q11 on rising<br />
supply from vendors’ wafer capacity expansion<br />
and process migration.<br />
Tablets impact commodity, positive for<br />
advanced players. We expect tablets to lower<br />
global DRAM bit demand by c1% this year and<br />
c5% for each of 2011-12e global demand. We<br />
believe memory vendors that can supply the lowvoltage,<br />
low power DRAM for integration with<br />
SoC used in tablets stand to benefit.<br />
Leading vendors best positioned to deliver<br />
higher-end products, and protect margins with<br />
process migration. We think product mix will<br />
become more critical for memory vendors during<br />
the next few quarters as ASP falls steepen. Rapid<br />
process migration by leading DRAM vendors<br />
results not just in better cost structure; from a<br />
competitive perspective, this enables them to meet<br />
demand for low-power parts that advanced nodes<br />
are better suited to deliver, both for mobile<br />
devices space (notebooks, smartphones, tablets)<br />
14
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
and higher end servers. We also expect leading<br />
vendors’ increased production of 2Gb parts to<br />
stimulate demand for larger memory modules.<br />
Here we look at the potential impact of corporate<br />
servers, which account for c10% of global bit<br />
demand, but we believe accounts for a much<br />
larger proportion of global DRAM revenue;<br />
server DRAM is c2.5-6x more expensive<br />
compared to ‘regular’ memory modules.<br />
DRAM demand forecast<br />
40,000<br />
60%<br />
50%<br />
30,000<br />
40%<br />
20,000<br />
30%<br />
10,000<br />
20%<br />
10%<br />
-<br />
0%<br />
2008 2009 2010f 2011f 2012f<br />
Demand (1Gb, m) Change (%)<br />
Source: IDC, HSBC estimates<br />
Upside/downside risks<br />
Key upside risk to our forecasts is a long-term,<br />
controlled growth of the industry, with strong bit<br />
growth and relatively mild ASP declines.<br />
We expect bit growth to accelerate in 2H10, due<br />
to PC build for the holidays, plus corporate PC<br />
recovery. We forecast most of the bit growth to<br />
occur in 2H10 as the impact of node migration by<br />
makers starts to come into play.<br />
We think downside risks are relatively limited for<br />
DRAM demand over the next two years. Risks<br />
include slower than expected increase in DRAM<br />
content per box for PCs on relatively high ASPs.<br />
The key supply downside is irrational competition<br />
leading to rapid ASP deflation. From 2012 onwards,<br />
additional risks include technology migration<br />
difficulties due to newer technologies such as EUV.<br />
Catalysts – short term, long term<br />
Short-term catalysts include strong data<br />
points from various PC OEMs for unit sales<br />
and content per box. Near term, sell through<br />
data from PC OEMs over the holidays in<br />
China and developed markets will be key.<br />
DRAM bit demand, 2009<br />
DRAM bit demand, 2012e<br />
Others<br />
10%<br />
Others<br />
9%<br />
Mobile<br />
handsets<br />
Mobile<br />
handsets<br />
8% 10%<br />
PC<br />
PC<br />
51%<br />
Other data<br />
56%<br />
Other data<br />
processing<br />
26%<br />
processing<br />
27%<br />
Tablets<br />
3%<br />
Source: IDC, HSBC estimates<br />
Source: IDC, HSBC estimates<br />
15
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Global DRAM – demand & supply forecast<br />
2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e<br />
Supply-demand ratio (%) 99.9 100.1 101.5 99.5 98.0 91.9 94.2 115.7 110.1 103.8 94.7 100.5<br />
Supply (1Gb, m) 10,413 15,488 23,071 32,990 3,338 3,557 3,936 4,657 5,005 5,423 6,024 6,619<br />
Change (%) 22% 49% 49% 43% 5% 7% 11% 18% 7% 8% 11% 10%<br />
Demand (1Gb, m) 10,422 15,476 22,721 33,165 3,405 3,869 4,179 4,024 4,544 5,226 6,362 6,589<br />
Change (%) 24% 48% 47% 46% 5% 14% 8% -4% 13% 15% 22% 4%<br />
DRAM sales (USDm) 23,988 42,890 43,497 46,497 9,856 11,290 10,934 10,811 10,413 10,590 11,118 11,377<br />
Change (%) 0% 79% 1% 7% 7% 15% -3% -1% -4% 2% 5% 2%<br />
Blended ASP (USD) 2.30 2.77 1.89 1.41 2.95 3.17 2.78 2.32 2.08 1.95 1.85 1.72<br />
Change (%) -18% 20% -32% -25% 2% 7% -12% -16% -10% -6% -5% -7%<br />
DDR3 contract ASP (USD) 1.84 2.10 1.39 1.05 2.37 2.53 2.15 1.77 1.57 1.46 1.36 1.26<br />
Change (%) 15% -34% -24% 12% 7% -15% -18% -11% -7% -7% -7%<br />
Wafer cap (kwpy, kwpm) 16,985 18,206 18,881 20,366 1,489 1,489 1,514 1,569 1,569 1,569 1,569 1,569<br />
Change (%) -1% 7% 4% 8% -1% 0% 2% 4% 0% 0% 0% 0%<br />
Demand (MB bn) 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e<br />
PC 738 1,076 1,532 2,195 237 269 290 280 306 352 429 444<br />
Tablets - 5 55 115 1 1 1 1 11 13 15 16<br />
Other data processing 346 522 781 1,142 115 130 141 136 156 180 219 226<br />
Mobile handsets 111 177 266 429 39 44 48 46 53 61 74 77<br />
Others 139 202 274 363 44 50 55 52 55 63 77 80<br />
Total 1,334 1,981 2,908 4,245 436 495 535 515 582 669 814 843<br />
Change (%) 24% 48% 47% 46% 5% 14% 8% -4% 13% 15% 22% 4%<br />
Demand mix (%)<br />
PC 55% 54% 53% 52% 54% 54% 54% 54% 53% 53% 53% 53%<br />
Tablets 0% 0% 2% 3% 0% 0% 0% 0% 2% 2% 2% 2%<br />
Other data processing 26% 26% 27% 27% 26% 26% 26% 26% 27% 27% 27% 27%<br />
Mobile handsets 8% 9% 9% 10% 9% 9% 9% 9% 9% 9% 9% 9%<br />
Others 10% 10% 9% 9% 10% 10% 10% 10% 9% 9% 9% 9%<br />
Total 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100%<br />
Device shipment (m units) 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e<br />
PC 299 342 366 397 75 86 92 89 73 84 102 106<br />
Change (%) 2% 14% 7% 9% -19% 14% 8% -4% -18% 15% 22% 4%<br />
DT PC 136 147 156 162 32 37 40 38 31 36 44 45<br />
NB PC 141 171 195 224 38 43 46 44 39 45 55 57<br />
Netbook 22 24 14 11 5 6 7 6 3 3 4 4<br />
Other data processing 43 60 78 97 13 15 16 16 16 18 22 23<br />
Change (%) -11% 12% 13% 18% 43% -7% -8% -8% 44% -7% -8% -8%<br />
Mobile handsets (m) 1,116 1,244 1,350 1,457 284 298 321 342 316 332 346 356<br />
Change (%) -8% 11% 9% 8% -7% 5% 8% 7% -8% 5% 4% 3%<br />
Smartphones (m) 176 271 377 509 55 65 72 79 84 93 98 102<br />
Change (%) 8% 54% 39% 35% 7% 18% 11% 10% 6% 11% 5% 4%<br />
Tablets (m) - 16 73 102 - 3 4 9 12 15 21 25<br />
Change (%) 356% 39% 40% 110% 36% 25% 40% 19%<br />
DRAM MB/system<br />
PC 2,467 3,143 4,187 5,521<br />
Tablets 288 774 1,161<br />
Other data processing 8,111 8,729 10,047 11,760<br />
Telecoms 99 142 197 295<br />
Source: Company, HSBC estimates. Blended ASP and DDR3 contract ASPs are for 1Gb equivalent. Wafer capacity is 12” equivalent<br />
16
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
DRAM shipments for PC, % of total DRAM demand, MB bn<br />
System units and content/system growth for computing<br />
500<br />
56%<br />
500<br />
600%<br />
400<br />
300<br />
200<br />
100<br />
55%<br />
54%<br />
53%<br />
52%<br />
51%<br />
50%<br />
400<br />
300<br />
200<br />
100<br />
500%<br />
400%<br />
300%<br />
200%<br />
100%<br />
-<br />
49%<br />
-<br />
0%<br />
2009 2010f 2011f 2012f<br />
2009 2010f 2011f 2012f<br />
PC<br />
% of total<br />
Unit (mn) MB/system '000<br />
Source: IDC, HSBC estimates<br />
Source: IDC, HSBC estimates<br />
Computing<br />
Relevance: PCs remain the largest demand driver,<br />
accounting for c54% of DRAM bit demand this<br />
year, though falling slowly to c53% by 2012e, with<br />
bit demand CAGR of c44% during the period.<br />
Upside/downside factors. Upside factors include<br />
greater demand arising from price elasticity. This<br />
memory per system to handle more complex tasks<br />
such as high definition video, plus competition<br />
between PC OEMs is driving greater content/box.<br />
PC unit shipments could beat market expectations on<br />
stronger than expected corporate demand recovery.<br />
Downside factors include the industry-wide<br />
downside risk that if DRAM prices rise further,<br />
PC makers may be forced to reduce onboard<br />
content per box, which could cap shorter-term bit<br />
growth. Similarly, high mDRAM ASPs could<br />
suppress inclusion of more mDRAM in handsets.<br />
PC unit shipments could miss market expectations.<br />
Catalysts. Strong bit growth from major makers,<br />
stronger than expected DRAM spot and contract<br />
prices, corporate results reflecting higher<br />
profitability, including that from cost reductions<br />
resulting from process efficiencies and migration.<br />
Positive DRAM demand outlook<br />
Next year, we forecast PC DRAM bit demand to<br />
rise 42.4% y-o-y from this year’s 45.8%. 2010e<br />
growth is relatively high due to the rebound in PC<br />
unit shipments following the global financial<br />
crisis. We expect higher content/system rather<br />
than PC unit shipments to drive 2011e bit growth<br />
for commodity DRAM.<br />
HSBC’s Regional <strong>Asia</strong> Technology Hardware<br />
analyst, Wanli Wang, forecasts global PC<br />
shipment growth (excluding tablets) to slow to<br />
10.9% next year from this year’s estimate of<br />
14.7% (see <strong>Asia</strong>n Hardware – Tablets the swing<br />
factor for 2011 PC market, 16 September 2010).<br />
For 2011 he expects global PC shipments to rise<br />
23% with tablet PCs, 11% without.<br />
Nearer term, Wanli expects 3Q10 Taiwanese<br />
notebook shipments to track in line, though he<br />
expects sequential 3Q10 shipment growth to be<br />
relatively low due to a high 2Q10 base. Unit<br />
shipments for 2Q benefited from order pull-in<br />
ahead of price increases, plus rising seaborne<br />
shipments to lower costs. Channel checks suggest<br />
that there is no sign of weaker demand, even in<br />
Western Europe. Wanli also suggests that 3Q10<br />
PC outlook is sufficiently healthy enough for PC<br />
vendors to maintain component orders and/or<br />
component vendor capacity remains tight, with<br />
little or no inventory build in 2Q10.<br />
17
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
PC OEMs, processor vendors<br />
readacross<br />
Strong server and networking demand…<br />
PC OEMs’ recent financial results and outlooks<br />
reflect strong corporate refresh demand, but<br />
uncertainty in consumer demand. This points to<br />
divergent prospects for DRAM makers, with those<br />
supplying higher-end products for servers and<br />
workstations faring better than those supplying<br />
just commodity parts. We estimate that servers<br />
account for c20% of Hynix and Samsung’s<br />
DRAM revenue. For 3Q10, Samsung guided<br />
blended DRAM ASP decline of high-single digits,<br />
while Hynix guided for ASPs to fall ‘slightly’ this<br />
quarter. We estimate average DDR2 and DDR3<br />
contract prices fell sharply by 13% q-o-q for the<br />
quarter based on DRAMeXchange prices.<br />
Dell (DELL US, not rated) expects corporate<br />
refresh demand in client and enterprise systems to<br />
continue over the next several quarters. The firm<br />
reported strong 2QFY11 results led by corporate<br />
and public sectors. Large enterprise revenue rose<br />
38%, public sector revenue rose 21% and SMB<br />
revenue rose 25%. Note that large, public and<br />
SMB revenues accounted for over 80% of Dell’s<br />
total 2QFY11 revenue. Dell (or HP) do not appear<br />
concerned about component prices, a key negative<br />
point for these firms in the previous quarters.<br />
HP’s (HPQ US, not rated) recent 3QFY10<br />
results saw strong 19% y-o-y growth in enterprise<br />
storage and server (ESS) revenue. HP also saw<br />
Personal Systems revenue rise 17%, within which<br />
commercial client revenue rose 25% led by<br />
desktop and workstation revenue (+27%, 54%<br />
respectively). Overall, commercial hardware<br />
revenue rose 28%, compared to 4% growth in<br />
consumer hardware. HP in particular benefited<br />
from a recent raft of launches of enterprise server,<br />
storage and networking equipment. HP raised its<br />
EPS guidance for 4QFY10 and raised its guidance<br />
for full year revenue (now 9.5% y-o-y).<br />
Lenovo (992 HK, OW(V), HKD4.99, TP<br />
HKD5.70; covered by HSBC analyst Wanli<br />
Wang) FY1Q11 operating margin rose<br />
significantly, which we expect to continue in<br />
FY2Q11 due to lower component prices<br />
(especially DRAM and LCD) plus rising<br />
corporate high-end product demand.<br />
Intel (INTC US, not rated) had a record quarter<br />
in 2QFY10, with revenue buoyed by record<br />
mobile processor and server processor revenues,<br />
which were driven by strong corporate demand. In<br />
particular, the Data Center Group revenue rose<br />
13% q-o-q.<br />
…but uncertain consumer demand<br />
Compared to ongoing strength in the corporate IT<br />
market, uncertainty remains in the consumer side.<br />
For example, Dell’s latest results showed flat<br />
consumer revenue, contrasting sharply with its<br />
enterprise revenues. The picture is more telling from<br />
the processor makers. Intel recently guided lower<br />
3QFY10 revenue due to weaker than expected<br />
demand for consumer PCs in developed markets.<br />
Similarly, AMD (AMD US, not rated) guided down<br />
its 3QFY10 revenue due to weaker than expected<br />
demand especially in the consumer notebook market<br />
in North America and Western Europe.<br />
DDR2, DDR3 contract prices (USD)<br />
3<br />
2.5<br />
2<br />
1.5<br />
1<br />
Jul-09<br />
Oct-09<br />
Source: DRAMeXchange<br />
Jan-10<br />
Apr-10<br />
Jul-10<br />
1Gb DDR2 128Mx 8 667 Mhz contract<br />
1Gb DDR3 128Mx 8 1066MHz contract<br />
18
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Price elasticity = higher content/box<br />
We expect demand downside to be relatively<br />
limited because of the greater affordability of<br />
DRAM modules. We expect DRAM price<br />
declines to start slowing in 2Q11 as lower ASPs<br />
from sharply higher supply during 4Q10-1Q11<br />
stimulate higher content/system and as supply<br />
increases peak out. Overall, we expect content/<br />
system to rise 33% next year, following a 27% y-<br />
o-y rise in 2010; high ASPs this year depressed<br />
content growth this year, though the extent of<br />
DRAM content reduction or freeze appears to<br />
vary according to PC OEMs.<br />
DRAM demand and ASPs<br />
2009 2010e 2011e 2012e<br />
ASP (USD) -18% 20% -32% -25%<br />
Content/system (MB) 27% 27% 33% 32%<br />
Units shipments (m) 2% 14% 7% 9%<br />
Demand (MB bn) 30% 46% 42% 43%<br />
Source: HSBC estimates<br />
Through 1H10, second-tier PC OEMs in<br />
particular have pointed to high DRAM costs as a<br />
proportion of bill of materials (BOM), at about<br />
13-15%. We are aware that some PC vendors<br />
have lowered content to 3GB/box, but vendors’<br />
strategies towards memory content tend to vary.<br />
OEMs more focused on higher-end systems have<br />
been reluctant to reduce content per box on<br />
competitiveness grounds.<br />
Indeed, bit growth in the coming quarters could<br />
surprise on the upside. Before the recent ASP<br />
declines Hynix expected GB/system growth<br />
slowing in 2H10, but at the same time it did not<br />
expect drastic cuts in content growth. Samsung<br />
expected GB/system in 3Q to be flat. Similarly,<br />
Micron saw OEMs continuing to market 4GB<br />
models; it sees limited cases of low-priced<br />
consumer PCs offering lower configurations.<br />
We expect PC OEMs to gradually start raising<br />
content/box starting in 3Q as DRAM prices have<br />
dropped steeply; 2GB DDR3 modules now cost<br />
around USD36, 23% below the peak of USD46.5<br />
in May. We expect price elasticity to kick in<br />
meaningfully in 4Q10 on persistent DRAM<br />
pricing weakness arising from expanded supply.<br />
We model 18% q-o-q average ASP decline for<br />
DDR3 contract prices in 4Q10, implying an<br />
average price of USD32 per 2GB module for the<br />
quarter and a 31% fall from peak prices. At that<br />
level, 4GB content/system on desktops should<br />
become affordable to the majority of PC OEMs.<br />
By 1Q11, 2GB modules could have fallen to<br />
around USD28, starting to drive real increase in<br />
content/system above the 4GB level. For higherend<br />
PCs, we expect content/box climbing to 6-<br />
8GB next year to competitively differentiate.<br />
Even currently, higher-end desktops and<br />
notebooks ship with up to 8GB.<br />
We also expect accelerating adoption of higherend<br />
4GB modules based on 4x-5xnm 2Gb parts to<br />
drive bit demand. As production of the higherdensity<br />
part rises, IDC forecasts 2Gb unit<br />
shipments to almost rival that of 1Gb next year;<br />
by 2012, 2Gb would be the mainstream part. As<br />
such, per Gb ASP differential between 1Gb and<br />
2Gb parts is to close next year. IDC expects 1Gb<br />
and 2Gb ASPs to fall c40% and c50% y-o-y<br />
respectively (IDC forecasts blended DRAM ASPs<br />
to decline c25% next year).<br />
In this light, leading DRAM vendors that are rapidly<br />
shifting production of 2Gb (and to a far lesser extent<br />
for now, 4Gb) parts to 4xnm and below from 5xnm<br />
will benefit the most from this trend. Samsung and<br />
Hynix raised targets for 4xnm proportions to over<br />
60% and 50% respectively by end-2010, we suspect<br />
most of the new process would be geared to 2Gb<br />
parts. We expect 4xnm and below to account for<br />
close to all of Samsung’s DRAM production by the<br />
end of next year. The availability of lower-power,<br />
high performance DRAM should accelerate as<br />
Samsung and Hynix start mass production on the<br />
3xnm process during 2H10.<br />
19
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
High-end computing DRAM MB bn, % of total DRAM demand<br />
PC server DRAM MB bn, % of total PC DRAM demand<br />
500<br />
12%<br />
400<br />
20%<br />
400<br />
300<br />
200<br />
100<br />
11%<br />
11%<br />
10%<br />
10%<br />
9%<br />
300<br />
200<br />
100<br />
15%<br />
10%<br />
5%<br />
-<br />
9%<br />
-<br />
0%<br />
2009 2010f 2011f 2012f<br />
2009 2010f 2011f 2012f<br />
High-end com puting<br />
% of total<br />
x 86 serv ers<br />
% of total PC<br />
Source: IDC, HSBC estimates<br />
Source: IDC, HSBC estimates<br />
Corporate and server demand<br />
Our regional analyst Wanli Wang forecasts 2011<br />
desktop PC unit growth rates to fall to +6% y-o-y<br />
from 8% in 2010; however this is still better than<br />
the annual decline in 2008/09. Corporate PC<br />
replacement demand should continue to support<br />
desktop PC unit growth. In notebooks, we expect<br />
1.35V DRAM to migrate towards notebooks next<br />
year especially as ULV (ultralow voltage)<br />
processors become more ubiquitous. We<br />
understand that some PC makers are asking for<br />
even lower 1.20/1.25V parts already.<br />
Servers – niche, but very high ASP market<br />
More interestingly perhaps in an often overlooked<br />
segment, is demand for higher-end DRAM from<br />
servers. We expect continued corporate spend plus<br />
ongoing growth of cloud computing to drive strong<br />
demand from this segment. Demand will also be<br />
catalysed by availability of high-density (2Gb and<br />
4Gb very soon), low-power DRAM modules.<br />
Micron expects robust demand from major<br />
customers to continue into 2H10 for servers, a<br />
sentiment echoed by Hynix and Samsung. We<br />
forecast server unit growth of c10% y-o-y) in<br />
2011e, but higher content/box growth of c38%.<br />
Price premiums are driven by high-density, high<br />
performance and lower power consumption, in<br />
addition to higher production costs. Server<br />
memory content/system in general tends to double<br />
every two years. High-end servers use DDR3; we<br />
expect migration to lower voltage, i.e. 1.35V,<br />
compared to 1.5V which is standard currently, to<br />
reduce power consumption, driving further highend<br />
DDR3 demand.<br />
Server memory modules are usually ECC RDIMM.<br />
From a cost perspective, these modules are far more<br />
expensive than regular modules because of 1)<br />
cutting-edge capacity memory chips which are more<br />
expensive due to production yields and volumes, 2)<br />
ECC memory uses 9 chips instead of 8 for each<br />
memory bank, 3) lower module volumes, and 4)<br />
additional components required to support buffering.<br />
Samsung started mass production of 32GB server<br />
grade DRAM modules in 2Q10 using 4Gb parts<br />
produced on 4xnm process, though these are likely<br />
to be customer samples until 4Q10. Each server then<br />
could hold up to 384GB with little power<br />
consumption difference or 192GB with 40% lower<br />
consumption than for modules using 2Gb parts.<br />
Servers (including PC servers) account for about<br />
10% of total global bit demand, but the blended<br />
ASP impact is far higher for vendors, at c2.5x for<br />
a 4GB module up to c6x for a 16GB module.<br />
20
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Tablet PC DRAM MB bn, % of total DRAM demand<br />
150<br />
100<br />
50<br />
-<br />
2009 2010f 2011f 2012f<br />
3.0%<br />
2.5%<br />
2.0%<br />
1.5%<br />
1.0%<br />
0.5%<br />
0.0%<br />
PC unit shipments, including tablet<br />
600<br />
500<br />
400<br />
300<br />
200<br />
100<br />
0<br />
2009 2010f 2011f 2012f<br />
Tablets<br />
% of total<br />
DT PC NB PC Tablet<br />
Source: HSBC estimates<br />
Source: IDC, HSBC estimates<br />
Tablets – about 5% global DRAM<br />
cannibalisation<br />
Tablets are a good thing, we concluded for NAND<br />
demand. Unlike netbooks, tablets by their very<br />
nature demand NAND as the primary storage<br />
device. But an emerging concern for commodity<br />
DRAM is that tablets will cannibalise netbook and<br />
notebook sales, reducing overall DRAM demand.<br />
How big is the threat? On the face of it, tablet’s<br />
potential impact on notebook demand could be<br />
substantial; Best Buy estimates that the iPad<br />
cannibalised laptop PC sales by up to 50% in its<br />
last reporting quarter. In our forecast, we calculate<br />
1%, 5% and 5% cannibalisation impact on global<br />
DRAM demand for 2010-12e. This could rise if<br />
cannibalisation reaches the sort of levels that Best<br />
Buy is eluding to. DRAM vendors that can supply<br />
low-voltage, low-power DDR (and increasingly<br />
DDR2) that could be integrated into tablets stand<br />
to benefit, or could offset some of the declines in<br />
commodity DRAM demand.<br />
In terms of units, we estimate tablets to account<br />
for c7.6% of tablets plus notebooks this year. We<br />
forecast tablet shipments to soar to 73m units next<br />
year, accounting for a quarter of our forecast<br />
tablets plus notebooks. We assume netbooks and<br />
CULV notebooks will see negative unit shipment<br />
growth, and that regular notebook growth will<br />
slow to c16.7% y-o-y.<br />
We estimate blended notebook plus netbook<br />
DRAM content/box of 2.7GB this year, with<br />
netbooks averaging a little over 1GB. In contrast,<br />
Apple’s iPad, Samsung’s Galaxy Tab and<br />
Toshiba’s Folio 100 have paltry DRAM content –<br />
256MB, 512MB and 512MB respectively. The<br />
cannibalisation impact will really start to be felt<br />
next year, when we estimate tablets will account<br />
for 26% of tablet plus notebook units but only<br />
c6% of tablet plus notebook DRAM bits. We<br />
estimate the impact of tablet cannibalisation to be<br />
around 1% of total global DRAM demand this<br />
year, but rising to c5% for 2011-12e. In this<br />
estimation we used the following assumptions:<br />
We estimate tablet unit shipments to rise to<br />
73m units next year and 102m units by 2012e.<br />
HSBC’s regional hardware analyst, Wanli<br />
Wang, expects ten vendors to launch tablet<br />
products next year. He expects notebooks to<br />
drive PC demand in 1H and tablets to drive<br />
demand in 2H. ARM estimates there will be<br />
40 tablet devices using ARM-based<br />
processors this year.<br />
We expect DRAM content in tablets to start<br />
rising quickly next year to support multitasking<br />
and more resource-intensive apps. We<br />
think this is doable, given DRAM is only c3%<br />
of BOM for current Apple iPads. We think 5-<br />
6% is more likely next year as 512MB<br />
21
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
becomes standard (as with the Samsung<br />
Galaxy Tab).<br />
By 2012 we expect tablet DRAM would<br />
reach above 1GB per system, catalysed by<br />
Windows and competitive pressure amongst<br />
tablet vendors to differentiate their products.<br />
Intel’s Oak Trail processor – which allows<br />
full Windows 7 to run on tablets (and<br />
netbooks) – was announced in June 2010 and<br />
is set to appear in devices starting 1H11. If we<br />
assume Windows 7 Basic is to run on a tablet,<br />
it would realistically require 1GB to run.<br />
We expect other operating systems to drive<br />
DRAM demand over 2011- 12e as well to<br />
support complex apps as processing power<br />
rapidly increases.<br />
22
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Bit shipments for handsets, % of total DRAM demand, MB bn<br />
Smartphone units and MB/system growth<br />
500<br />
400<br />
300<br />
200<br />
100<br />
-<br />
12%<br />
10%<br />
8%<br />
6%<br />
4%<br />
2%<br />
0%<br />
600<br />
500<br />
400<br />
300<br />
200<br />
100<br />
-<br />
500<br />
400<br />
300<br />
200<br />
100<br />
-<br />
2009 2010f 2011f 2012f<br />
2009 2010f 2011f 2012f<br />
Smart phone units (m)<br />
Telecoms<br />
% of total<br />
Blended MB/sy stem (RHS)<br />
Source: IDC, HSBC estimates<br />
Source: IDC, HSBC estimates<br />
Handsets<br />
Relevance: Telecoms is becoming a major<br />
consumer of DRAM, accounting for c9 % of DRAM<br />
bit demand this year and growing to c10% in 2012e<br />
at a CAGR of c57% during 2010-12e. Within<br />
telecoms, smartphones are a major driver of demand<br />
growth. Mobile DRAM prices tend to be higher on a<br />
bit basis, as memory vendors need to work closely<br />
with mobile phone makers to incorporate custommade<br />
solutions for specific handsets.<br />
Upside/downside factors. Upside factors include<br />
launch of higher-specification smartphones due to<br />
competitive pressures, more demanding apps and<br />
operating systems including Android<br />
‘Gingerbread’. Downside factors include high<br />
mDRAM ASPs suppressing the inclusion of more<br />
content/system.<br />
Catalysts. Better than expected smartphone unit<br />
shipment and product mix guidance/results from<br />
major handset manufacturers.<br />
Mobile DRAM<br />
We think smartphones are driving rapid mobile<br />
DRAM bit growth through both unit shipment<br />
growth and content per box. Mobile phones have<br />
had a very limited amount of DRAM in the past.<br />
In our view, the development of smartphones is<br />
now mirroring that of PCs, driven by resourceintensive<br />
wireless broadband, multimedia and<br />
multitasking. As more demanding apps become<br />
the norm, we expect rapid advances in processors,<br />
screens, volatile and non-volatile storage. Mobile<br />
DRAM, which currently uses mostly DDR1, is<br />
migrating rapidly to DDR2 to support higher<br />
speed processors now found in smartphones.<br />
We forecast mDRAM bit growth to rise c102%,<br />
92% and 112% y-o-y for 2010-12e. We assume<br />
DRAM content/system for smartphones to rise<br />
c31%, c39% and 57% for 2010-12e, to around<br />
390MB by 2012e. We forecast smartphone unit<br />
shipments to rise 54%, 39% and 35% during the<br />
periods, accounting for c38% of total handsets<br />
shipped globally by 2012e (2009: 15%).<br />
At the cusp of an explosion in mDRAM.<br />
Like desktop PCs a decade ago, broadband<br />
connectivity, multitasking, more complex<br />
applications and more powerful graphics are<br />
driving demand for faster processors. We<br />
expect smartphone makers to attempt to<br />
differentiate their products not only through<br />
different operating systems, but ability of<br />
handsets to run applications quickly.<br />
Smartphone DRAM to surge to support<br />
computing power. Qualcomm’s Snapdragon,<br />
Samsung’s Hummingbird, Apple’s A4<br />
processor and other ARM-based SoCs run up<br />
to 1GHz already. We believe smartphones<br />
23
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
processing power will increase rapidly, in part<br />
to support requirements for next-iteration<br />
operating systems such as Android 3.0<br />
(Gingerbread), which will require 1GHz<br />
processors, 512MB DRAM and a 3.5” screen<br />
as minimum hardware requirements. Highend<br />
processors including Qualcomm’s dual<br />
core Snapdragon, NVIDIA’s next generation<br />
Tegra with dual cores and fast GPU (running<br />
DDR2 as opposed to DDR), and Samsung’s<br />
‘Orion’ dual core processor SoC are set to be<br />
become commonplace by end-2011e.<br />
c1GB by end-2011 in higher-end smartphones<br />
is likely, we think. Apple’s iPhone 4G and<br />
Samsung’s Galaxy S, as well as the now<br />
discontinued Google/HTC Nexus One already<br />
have 512MB of onboard mDRAM.<br />
Smartphone DRAM usage, % of total handsets<br />
250<br />
80%<br />
200<br />
60%<br />
150<br />
40%<br />
100<br />
50<br />
20%<br />
-<br />
0%<br />
2009 2010f 2011f 2012f<br />
GB m % of total handsets<br />
Samsung, Elpida and Hynix are leaders in<br />
mDRAM The higher-end mDRAM market (i.e.<br />
128MB upwards) is dominated by major DRAM<br />
players that are capable of customising MCP<br />
(multichip packages) for handset vendors.<br />
DRAMeXchange estimates mobile DRAM rose to<br />
c12% of Hynix’s product mix and c20% for<br />
Samsung.<br />
Several vendors have low power mobile DDR2<br />
(LPDDR2) in mass production at 5xnm process<br />
and above, and some have 4xnm process parts in<br />
engineering or customer sample stages. The latter<br />
development should deliver significant<br />
competitive advantage. With the 4xnm Hynix part,<br />
power consumption was reduced to c30% of<br />
50nm process produced mobile DDR2. Samsung<br />
already ships up to 2Gb LPDDR2 produced at<br />
56nm; its 46nm 2, 4 and 8Gb LPDDR2 parts are<br />
currently at the engineering sample stage; we<br />
believe mass production could start in 2H10.<br />
Hynix completed development of its 2Gb<br />
LPDDR2 using 44nm node technology in January<br />
2010 and is currently at the customer sample stage.<br />
Elapid’s 2Gb LPDDR2 DRAM produced on<br />
4xnm process started sampling in August 2010,<br />
and mass production is slated to start this month<br />
(September). Micron/Nanya is to mass produce its<br />
2Gb LPDDR2 in 3Q10.<br />
Source: IDC, HSBC estimates<br />
24
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Demand drivers – NAND<br />
Remain bullish on NAND; we forecast 2010-12e NAND demand<br />
bit CAGR of 103%<br />
Demand is structural and persistent, as heavy online and offline<br />
growth in media drives portable device storage requirements<br />
Our positive outlook on tablets and smartphones is diluted by<br />
cannibalisation of media players<br />
Summary<br />
Relevance, growth rates<br />
The non-volatile memory (NVM) segment – led<br />
by NAND – remains an exciting high-growth<br />
segment. We forecast 2010-12e NAND bit<br />
demand CAGR of c103% (compared to supply<br />
CAGR of 100%), rising 114% next year. Global<br />
appetite remains insatiable for storage capacity<br />
that is fast, robust and low-power.<br />
NAND demand forecast<br />
45,000<br />
120%<br />
100%<br />
30,000<br />
80%<br />
60%<br />
15,000<br />
40%<br />
20%<br />
-<br />
0%<br />
2009 2010f 2011f 2012f<br />
Demand (GB, m) Change, y -o-y , q-o-q (%)<br />
Source: IDC, HSBC estimates<br />
Foreseeable demand remains strong on rising<br />
affordability, expanding applications in high<br />
potential markets such as smartphones, tablets, PC<br />
storage plus nascent markets such as automotive<br />
applications. We continue to believe the risk is on<br />
the upside, as demand accelerates on more rapid<br />
affordability trends. By 2012e we expect global<br />
NAND revenues to recover to c54% of DRAM’s,<br />
up from this year’s estimated 38%.<br />
Unlike DRAM, NAND demand remains strong<br />
for commodity parts and embedded solutions. The<br />
main embedded demand driver is handsets, led by<br />
smartphones, which we estimate will account for<br />
37% of total bit demand next year, up from 28%<br />
in 2009. Indeed, c60% of Toshiba’s NAND<br />
revenue is from embedded solutions that include<br />
mobile and other consumer electronics. Another<br />
embedded solutions driver is tablets, a segment<br />
which practically did not exist until April this year.<br />
For commodity NAND, we are set to see serious<br />
demand arising from not only SSDs for regular<br />
PCs, but perhaps more importantly, for servers,<br />
which we think is a market that is under-estimated<br />
in terms of its potential. In this report we highlight<br />
the upside risk from corporate server storage<br />
systems. We expect SSD growth, which had been<br />
stalled this year by high ASPs to see rapid growth<br />
due to falling NAND prices and indirect demand<br />
stimulation through the ‘tablet effect’.<br />
25
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Global NAND demand & supply<br />
2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e<br />
Supply-demand ratio (%) 100.4 99.5 99.6 96.2 98.6 105.5 98.1 97.1 91.6 100.7 102.1 101.5<br />
Supply (16Gb equ., m) 2,914 4,984 10,670 19,842 931 1,151 1,342 1,561 1,849 2,348 2,985 3,488<br />
Supply (GB, m) 5,828 9,969 21,341 39,684 1,862 2,302 2,684 3,121 3,697 4,697 5,970 6,977<br />
Change, y-o-y, q-o-q (%) 41.1% 71.0% 114.1% 86.0% 4% 24% 17% 16% 18% 27% 27% 17%<br />
Demand (16Gb equ., m) 2,903 5,011 10,712 20,626 944 1,091 1,368 1,608 2,018 2,332 2,924 3,437<br />
Demand (GB, m) 5,807 10,021 21,423 41,252 1,888 2,182 2,736 3,216 4,036 4,664 5,848 6,875<br />
Change, y-o-y, q-o-q (%) 45.5% 72.6% 113.8% 92.6% 3% 16% 25% 18% 26% 16% 25% 18%<br />
Revenue (USDm) 11,857 16,238 21,384 24,834 3,569 4,066 4,234 4,369 4,545 5,208 5,758 5,872<br />
Change, y-o-y, q-o-q (%) -3.5% 37.0% 31.7% 16.1% -5% 14% 4% 3% 4% 15% 11% 2%<br />
NAND/DRAM revenue (%) 49% 38% 49% 54%<br />
ASP (USD, 16Gb equ.) 4.1 3.3 2.0 1.3 3.8 3.5 3.2 2.8 2.5 2.2 1.9 1.7<br />
Change, y-o-y, q-o-q (%) -32% -20% -38% -38% -9% -8% -11% -11% -12% -10% -13% -13%<br />
Wafer cap. (300mm equ) 9,194 10,476 12,529 15,129 829 867 878 918 948 978 1,095 1,155<br />
Change, y-o-y, q-o-q (%) -5% 14% 20% 21% 1% 5% 1% 5% 3% 3% 12% 5%<br />
Source: Company data, HSBC estimates<br />
Upside/downside risks<br />
Key upside risk to our forecasts is a long-term,<br />
controlled growth of the industry where bit<br />
growth remains strong, and ASP declines are<br />
relatively controlled. Downside demand risks are<br />
limited for NAND, but include slower than<br />
expected adoption due to slow ASP declines<br />
resulting from tight supply-demand; price<br />
elasticity is critical especially for devices such as<br />
SSDs. The key supply downside is irrational<br />
competition leading to rapid ASP deflation.<br />
Catalysts – short term, long term<br />
Strong bit growth from makers, stronger than<br />
expected NAND contract prices.<br />
Better than expected unit shipment and<br />
product mix guidance/results from handset<br />
vendors and USB/flash storage makers such<br />
as SanDisk, A-Data, and PQI.<br />
Short-term catalysts include strong sales of<br />
various Apple iPad devices, launch and strong<br />
sales of similar tablet devices by major vendors<br />
including Samsung, Toshiba and RIMM.<br />
NAND bit demand profile, 2009<br />
NAND bit demand profile, 2012e<br />
DVC<br />
SSD<br />
3%<br />
5%<br />
USB Flash<br />
Driv e<br />
11%<br />
Handset<br />
28%<br />
Others<br />
7%<br />
Digital S till<br />
Camera<br />
20%<br />
Media<br />
Play er<br />
26%<br />
DVC<br />
2%<br />
Tablets<br />
12%<br />
SSD<br />
17%<br />
USB Flash<br />
Driv e<br />
8%<br />
Others<br />
5%<br />
Digital Still<br />
Camera<br />
7%<br />
Media<br />
Play er<br />
6%<br />
Handset<br />
43%<br />
Source: IDC, HSBC estimates<br />
Source: IDC, HSBC forecasts<br />
26
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Global NAND demand breakdown<br />
GB equ. units 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e<br />
Digital Still Camera 1,176 1,520 2,134 3,092 324 355 417 424 469 502 569 594<br />
Media Player 1,504 1,923 2,502 2,560 354 441 534 593 541 588 681 692<br />
Handset 1,638 3,350 7,920 17,472 596 762 898 1,094 1,423 1,824 2,252 2,421<br />
USB Flash Drive 644 1,175 2,063 3,375 221 256 321 377 389 449 563 662<br />
SSD 270 693 2,420 6,714 102 147 199 265 352 476 702 959<br />
Tablets - 496 2,784 4,902 - 93 130 273 458 572 801 953<br />
DVC 175 274 517 886 52 60 75 88 97 112 141 166<br />
Others 400 590 1,084 2,251 111 128 161 189 204 236 296 348<br />
Total 5,807 10,021 21,423 41,252 1,760 2,243 2,735 3,303 3,933 4,760 6,005 6,795<br />
y-o-y, q-o-q (%) 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e<br />
Digital Still Camera 29% 29% 40% 45% -12% 10% 17% 1% 11% 7% 13% 4%<br />
Media Player 23% 28% 30% 2% -20% 25% 21% 11% -9% 9% 16% 2%<br />
Handset 49% 105% 136% 121% 26% 28% 18% 22% 30% 28% 23% 8%<br />
USB Flash Drive 46% 82% 76% 64% 9% 16% 25% 18% 3% 16% 25% 18%<br />
SSD 230% 157% 249% 177% 33% 44% 35% 33% 33% 35% 47% 37%<br />
Tablets na na 461% 76% na na 40% 110% 68% 25% 40% 19%<br />
DVC 150% 57% 88% 71% -6% 16% 25% 18% 11% 16% 25% 18%<br />
Others 138% 47% 84% 108% -12% 16% 25% 18% 8% 16% 25% 18%<br />
Total 46% 73% 114% 93% 1% 27% 22% 21% 19% 21% 26% 13%<br />
Demand mix 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e<br />
Digital Still Camera 20% 15% 10% 7% 18% 16% 15% 13% 12% 11% 9% 9%<br />
Media Player 26% 19% 12% 6% 20% 20% 20% 18% 14% 12% 11% 10%<br />
Handset 28% 33% 37% 42% 34% 34% 33% 33% 36% 38% 38% 36%<br />
USB Flash Drive 11% 12% 10% 8% 13% 11% 12% 11% 10% 9% 9% 10%<br />
SSD 5% 5% 11% 16% 6% 7% 7% 8% 9% 10% 12% 14%<br />
Tablets 0% 5% 13% 12% 0% 4% 5% 8% 12% 12% 13% 14%<br />
DVC 3% 3% 2% 2% 3% 3% 3% 3% 2% 2% 2% 2%<br />
Others 7% 6% 5% 5% 6% 6% 6% 6% 5% 5% 5% 5%<br />
Total 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100%<br />
Unit shipment (m) 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e<br />
Digital Still Camera 130 132 141 149 32 32 34 33 35 35 37 35<br />
Media Player 165 120 100 80 30 29 31 30 25 25 26 25<br />
Handset 1,116 1,244 1,350 1,457 284 298 321 342 316 332 346 356<br />
USB Flash Drive 184 250 275 300 61 61 65 62 68 67 72 68<br />
SSD 14 19 30 51 3 4 5 6 6 7 8 9<br />
Tablets - 16 73 102 - 3 4 9 12 15 21 25<br />
DVC 12 14 18 21 4 4 4 4 4 4 5 4<br />
Others 101 119 155 210 29 29 31 29 38 38 41 38<br />
GB/system 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e<br />
Digital Still Camera 9 12 15 21 10 11 12 13 14 15 15 17<br />
Media Player 9 16 24 32 12 15 17 20 22 24 26 28<br />
Handset 1 3 6 12 2 3 3 3 5 6 7 7<br />
USB Flash Drive 4 5 8 11 6 7 8 9 9 10 13 14<br />
SSD 19 37 81 132 30 35 39 44 55 68 90 110<br />
Tablets - 31 38 48 - 31 31 31 38 38 38 38<br />
DVC 15 19 29 43 15 17 20 25 22 25 30 37<br />
Source: Company data, HSBC estimates<br />
27
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
% of total NAND bit demand from handsets System unit and GB/system growth for handsets<br />
50<br />
40<br />
30<br />
20<br />
10<br />
0<br />
2,000<br />
1,500<br />
1,000<br />
500<br />
0<br />
15<br />
10<br />
5<br />
0<br />
2009 2010f 2011f 2012f<br />
2009 2010f 2011f 2012f<br />
Handset<br />
Sy stem units (m)<br />
GB/sy stem<br />
Source: IDC, HSBC estimates<br />
Source: Company, HSBC estimates<br />
Handsets<br />
Relevance: Mobile handsets have emerged as the<br />
leading usage of NAND thanks to a proliferation<br />
of smartphones. Handsets account for 37% of<br />
2011e total bit demand forecast (2009: 28%).<br />
Other than nascent tablets and SSDs, mobile<br />
handsets is the fastest growing segment also,<br />
growing at a 10/12e CAGR of 128% to account<br />
for c42% of total NAND bit demand by 2012e.<br />
What’s changed in our view? We incorporate<br />
our upwardly revised global handset and<br />
smartphone forecasts.<br />
Upside/downside factors. Watch for Apple – the<br />
leading indicator for mobile device NAND usage<br />
– releasing larger capacity iPhones and iPods.<br />
Upside risks include higher than expected take up<br />
of smartphones in emerging markets especially on<br />
3G network roll out, higher than expected<br />
content/system in smartphones for developed<br />
markets. Downside risks include slowing average<br />
GB growth on smartphones as lower-spec models<br />
(low NAND content) become a higher proportion<br />
of total smartphones.<br />
System units, GB/system both rise<br />
rapidly<br />
Growth in NAND bit demand is driven by rising<br />
GB/system as smartphone proportion of total<br />
handsets rises. Also the amount of storage on<br />
smartphones themselves is rising, driven by<br />
higher resolution content.<br />
Smartphone proportion rises<br />
We forecast smartphones to rise to account for<br />
c38% of total global handset units by 2012e, up<br />
from c15% last year.<br />
Smartphone demand will continue to grow as<br />
model variety expands and functionality<br />
becomes more mature.<br />
We expect greater penetration of smartphones<br />
in emerging markets on better affordability,<br />
and roll out of 3G networks in markets such<br />
as China and India, driving new demand.<br />
28
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Handset and smartphone proportion<br />
2,000<br />
1,500<br />
1,000<br />
40%<br />
30%<br />
20%<br />
Higher resolution video, games content is<br />
becoming widespread and games are being<br />
written for higher-resolution devices. Real<br />
time HD is still highly restricted by<br />
availability and pricing of bandwidth.<br />
500<br />
10%<br />
-<br />
0%<br />
2009 2010f 2011f 2012f<br />
Handsets Smart phones/total handsets<br />
Source: Company, HSBC estimates<br />
Content/system rises<br />
Growth in average GB/system will remain strong<br />
for the foreseeable future. We forecast GB/system<br />
for smartphones to double each year during 2011-<br />
12e driven by:<br />
Lower NAND prices – higher bit supply<br />
growth led by Toshiba and Samsung on<br />
increased wafer capacity, rapid process<br />
migration and adoption of TLC should drive<br />
ASPs to fall by c38%/year for 2011-12e. We<br />
estimate NAND on a 16GB Apple iPhone 4 to<br />
account for 14% of BOM, a level which we<br />
think will be maintained going forward. We<br />
assume most non-Apple smartphones will use<br />
less NAND than Apple to lower prices,<br />
obliging the consumer to buy flash cards to<br />
fulfil extra storage needs.<br />
Availability of larger capacity mobile<br />
NAND. We expect Apple uses 1x 32GB<br />
NAND module for its iPhone 4 and 2x 32GB<br />
NAND modules for the 64GB iPod Touch.<br />
Rapidly rising screen resolutions.<br />
Smartphones are increasingly converging on a<br />
screen size 3.5-4” and WVGA (800x480)-<br />
level screen resolution including Apple<br />
iPhone 4 (960x 640), Samsung Galaxy S and<br />
HTC Desire. HD and gaming support rise far<br />
better graphics performance on latest SoCs).<br />
Cannibalisation of media players. Longerterm<br />
NAND content growth in developed<br />
markets will to a large extent depend on how<br />
far smartphones substitute media players.<br />
Consumers currently buy flash cards to<br />
supplement onboard storage.<br />
Smartphone NAND usage<br />
20,000<br />
15,000<br />
10,000<br />
5,000<br />
-<br />
Source: Company, HSBC estimates<br />
2009 2010f 2011f 2012f<br />
Total smart phone (GB m)<br />
% of total NAND demand<br />
Smart phones/total handset NAND<br />
100%<br />
80%<br />
60%<br />
40%<br />
20%<br />
Low-end smartphones also drive NAND<br />
growth. We expect proliferation of low- to midend<br />
smartphones and rising penetration of<br />
smartphones in emerging markets could<br />
incrementally drive NAND demand in excess of<br />
expectations. In particular, handset vendors such<br />
as Huawei, ZTE, LG Electronics and others are<br />
more than likely to chase greater global market<br />
share with more affordable smartphones.<br />
We believe white box makers are looking at<br />
USD100-120 for very basic smartphones. Huawei<br />
has mentioned previously that USD150 is the<br />
pricing point required for smartphones to drive<br />
potential 3G usage in China. We think a low-end<br />
smartphone with a BOM of cUSD80 could ship<br />
with up to 8GB of internal NAND by next year. In<br />
this case, the NAND cost would equate to 10-16%<br />
0%<br />
29
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
of BOM, which we think is a reasonable level for<br />
such vendors to absorb.<br />
Smartphone NAND content/system growth will<br />
persist for the foreseeable future as smartphone<br />
development mirrors that of PCs. Multi-tasking,<br />
multimedia and wireless broadband (3G in<br />
emerging markets and 4G in developed) are very<br />
powerful drivers for NAND and DRAM demand.<br />
As operating systems and applications become<br />
more demanding on systems requirements (just<br />
look at Android 3.0), advances in hardware<br />
specifications – processor, screen, volatile and<br />
non-volatile storage – are inevitable.<br />
Effects of tablet cannibalisation on<br />
Samsung<br />
Purely from a memory perspective, we estimate<br />
the net-net impact of tablet cannibalisation on<br />
Samsung Electronics’s profits to be minor next<br />
year. In drawing this conclusion we assumed that<br />
1) tablets shave 5% off global DRAM demand for<br />
2011-12; 2) tablets account for 5% of NAND<br />
demand this year rising to c13% next year, most<br />
of which is incremental additional demand. The<br />
minimal net-net effect would obviously not be the<br />
case for other memory vendors, which have<br />
different product mixes.<br />
We also assessed the broader impact on Samsung<br />
Electronics of tablet cannibalisation of netbooks<br />
and notebook PCs. We think the impact varies<br />
widely depending how much of a tablet or<br />
notebook’s components Samsung supplies. The<br />
impact also depends on whether tablets<br />
cannibalise notebooks or netbooks:<br />
Either way, proliferation of tablets should be<br />
positive for Samsung, as tablets increase the<br />
opportunity for the firm to supply more of the<br />
costly parts such as application processors<br />
and NAND that constitute these devices.<br />
likely benefits every time a consumer buys a<br />
tablet instead of a netbook.<br />
If DRAM prices fall to cUSD30/2GB module,<br />
then Samsung's chunk of operating profit<br />
from a notebook falls to cUSD24, close to<br />
that derived from a tablet.<br />
Apple iPad versus notebook and netbook costs, OP for SEC<br />
iPad (32GB) Cost Supplier(s) OPM OP<br />
App processor 15 Samsung 15% 2<br />
NAND 51 Samsung/Toshiba 28% 14<br />
mDRAM/NOR 9 Samsung/Elpida 42% 4<br />
9.7" LED backlit LCD 100 LGD/Samsung 15% 15<br />
Total 175 Total 35<br />
Notebook (4GB) Supplier(s) OPM OP<br />
DRAM (2x2GB DDR3) 72 Samsung et al 38% 27<br />
HDD(250GB, 5400rpm) 30 Samsung et al 5% 2<br />
14" LED backlit LCD 50 Samsung et al 11% 7<br />
Total 152 Total 36<br />
Netbook (2GB) Supplier(s) OPM OP<br />
DRAM (1x2GB DDR3) 36 Samsung et al 40% 14<br />
HDD(160GB, 5400rpm) 28 Samsung et al 5% 1<br />
10" LED backlit LCD 21 Samsung et al 11% 6<br />
Total 85 Total 22<br />
Source: HSBC estimates<br />
Base assumptions:<br />
If we assume that Samsung supplies all of the<br />
apps processor, NAND, mDRAM, NOR and<br />
LCD for a 32GB iPad, we estimate its<br />
operating profit per iPad at cUSD35. This is<br />
similar to our estimate operating profit of<br />
USD36 for a regular 4GB notebook if<br />
Samsung supplied all the DRAM, HDD and<br />
LCD screen. For a netbook, operating profit<br />
possible is much lower at cUSD22.<br />
If we assume that Samsung does not supply<br />
screens but all the other components<br />
mentioned earlier in tablets (SoC, NAND,<br />
mDRAM/NOR) and notebooks (DRAM,<br />
HDD), then Samsung's operating profit on a<br />
tablet could reach cUSD20 compared to<br />
cUSD29 for a 4GB notebook.<br />
If tablets cannibalise netbooks, which appears<br />
the most likely in our view, then Samsung<br />
30
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
% of total NAND bit demand from tablets System unit and GB/system growth for tablets<br />
15<br />
10<br />
5<br />
0<br />
120<br />
100<br />
80<br />
60<br />
40<br />
20<br />
0<br />
60<br />
50<br />
40<br />
30<br />
20<br />
10<br />
0<br />
2009 2010f 2011f 2012f<br />
2009 2010f 2011f 2012f<br />
% of NAND demand<br />
Sy stem units (m)<br />
GB/sy stem<br />
Source: HSBC estimates<br />
Source: HSBC estimates<br />
Tablets<br />
Relevance: Tablets represent a major new NAND<br />
application that has explosively entered the scene.<br />
Tablets by default use flash storage due to<br />
NAND’s inherent advantages of robustness, small<br />
form factor, longer battery life and ability to<br />
support instant-on. We expect tablets to account<br />
for 13% of global NAND demand next year (2010:<br />
5%), mostly driven by a massive 461% 2010/12e<br />
bit growth rate from a standing start this year.<br />
Explosive unit growth<br />
We estimate tablet unit shipments to rise to 73m<br />
units next year and 102m units by 2012e. The<br />
iPad should the biggest bite of this pie, holding<br />
about half of unit market share next year in our<br />
view. Other vendors are set to launch in the near<br />
future, led by Samsung. These include Androidbased<br />
models such as Toshiba’s Folio 100,<br />
RIMM’s ‘Blackpad’ (November 2010), LG<br />
Electronics’ Optimus (end-2010), variations of<br />
Dell’s Streak, Motorola (end-2010). Compal,<br />
Acer, Foxconn, Prowave and MSI also plan<br />
Android-based tablets for the near future. ARM, a<br />
major mobile device application processor vendor,<br />
estimates there will be about forty tablet devices<br />
using its processors by the end of this year. The<br />
flood of tablets will accelerate next year; HSBC’s<br />
regional hardware analyst, Wanli Wang, expects<br />
ten vendors to launch tablet products in 2011.<br />
GB/system to rise slowly at first<br />
We believe GB/system growth would initially be<br />
relatively slow, rising c23% y-o-y next year and<br />
27% in 2012e as tablets will mostly run compact<br />
Android and iOS operating systems; there is more<br />
than a bit of truth to the saying that tablets (for<br />
now at least) are much bigger smartphones. But<br />
the key difference is the larger amount of NAND<br />
that tablets have compared to smartphones, and<br />
hence we need to consider vendors’ attempts to<br />
control BOM. For Apple’s iPad NAND as a<br />
proportion of hardware-only BOM varies between<br />
c10-30% depending on the model. This is a much<br />
higher proportion than hard disk drive (HDD)<br />
storage takes up in PCs and is comparable to<br />
NAND as a proportion of BOM for smartphones.<br />
PC vendors entering the tablet (or smartphone)<br />
market need to completely readjust their<br />
expectations for storage cost in their BOM.<br />
But falling NAND ASPs will continue to support<br />
rising content/system. For non-Apple tablets (and<br />
smartphones), flash cards will be supplementary<br />
storage devices, giving users the option to raise<br />
storage as requirements grow. Even so, 16GB is<br />
about the minimum acceptable onboard NAND<br />
storage for a tablet in our view. Vendors should<br />
have little problem meeting this requirement this<br />
year. NAND price per GB averages USD1.63; a<br />
white box tablet vendor with a BOM of USD200<br />
31
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
could afford installing 16GB of onboard NAND<br />
(13% of BOM, on par with Apple). Next year, the<br />
BOM bar will be higher for tablet vendors, not just<br />
due to ferocious competition between those late to<br />
the party but also due to higher demands from OS<br />
and applications. But as ASPs fall to cUSD1/GB,<br />
32GB of NAND would account for c18% of BOM,<br />
which we think is a reasonable level.<br />
Impact of Windows 7 for tablet content/system<br />
Interestingly, we think Windows 7 could drive<br />
tablet GB/system. We think Windows could<br />
mount a belated but serious challenge in the tablet<br />
market due rising corporate usage of tablets, latent<br />
demand for tablets that run Windows applications,<br />
and development of tablet hardware that could<br />
support resources-intensive Windows. Aimed<br />
precisely at that potential market is Intel’s new<br />
Oak Trail processor, which allows full Windows 7<br />
to run on tablets. This processor is set to appear in<br />
devices starting 1H11. Such tablets could include<br />
MSI Wind Pad and Asus Eee Pad. But we think<br />
the impact of Windows 7 on NAND content/<br />
system would make a substantial impact by 2013.<br />
For Windows tablets, we think:<br />
There would be a larger NAND requirement.<br />
Unlike the much more compact Android and<br />
iOS, a full implementation of basic Windows<br />
7 needs a bare minimum 16GB of storage.<br />
We think 32GB is the minimum. In our view,<br />
SSDs on netbooks failed because the 8-16GB<br />
that was provided was simply insufficient to<br />
support even basic OS and applications.<br />
NAND storage for OS and apps would be<br />
onboard rather than through flash cards due to<br />
the latter's data transfer speed limitations.<br />
For NAND, any cannibalisation impact on regular<br />
notebooks is negligible as penetration of SSDs in<br />
the latter is still very low.<br />
32
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
% of total NAND bit demand from SSD System unit and GB/system growth from SSD<br />
20<br />
15<br />
10<br />
5<br />
0<br />
60<br />
50<br />
40<br />
30<br />
20<br />
10<br />
0<br />
150<br />
100<br />
50<br />
0<br />
2009 2010f 2011f 2012f<br />
2009 2010f 2011f 2012f<br />
% of NAND demand<br />
System units (m)<br />
GB/sy stem<br />
Source: IDC, HSBC estimates<br />
Source: IDC, HSBC estimates<br />
Solid state disk<br />
Relevance: We expect NAND for solid state disks<br />
(SSD) to account for c5% of NAND bit demand<br />
this year, rising to a forecast c11% next year. For<br />
commodity NAND, we see hefty demand arising<br />
from not only SSDs for regular PCs, but also for<br />
servers, which we think is a market that is underestimated<br />
in terms of its potential.<br />
Upside/downside factors. Upside risks include<br />
accelerated adoption by high-end servers (on<br />
improving SSD read/write performance), greater<br />
penetration of ‘regular SSD’ amongst notebook<br />
PCs and adoption by desktops. Another factor that<br />
could start kicking in is SSDs in desktops, an<br />
application we have advocated for sometime. In<br />
fact, Apple appears to be adopting this idea with<br />
the new iMac, which has the option to have both<br />
SSD and HDD in each machine.<br />
Downside risks include high NAND prices<br />
delaying adoption across most applications. SSD<br />
growth could also be limited by substitution from<br />
portable flash drives as the performance gap<br />
between internal and external drives narrow with<br />
eSATA, USB3.0 and FireWire S1600 drives.<br />
Catalysts. Catalysts include a spill-over effect for<br />
SSD demand from tablets into mainstream<br />
notebooks, as consumers increasingly awaken to<br />
the convenience of SSDs. Newer OS – Windows<br />
7 and Mac OS X – offer significantly better start<br />
up/shut down times than before. But tablet users<br />
will increasingly find these times intolerable.<br />
SSD adoption to accelerate<br />
Comparative affordability remains the hurdle<br />
We expect SSDs to start taking off in earnest next<br />
year, accounting for 11% of our total bit demand<br />
forecast as absolute pricing of SSDs becomes<br />
more affordable. In particular, we expect SSD<br />
adoption to accelerate in notebooks and in the<br />
corporate server storage market. We estimate the<br />
cost of implementing 128GB of NAND to reach<br />
USD128 this year, accounting for 20% of BOM<br />
for a mid- to high-end notebook PC; by 2012 we<br />
think this could fall to 13% of BOM, sufficient to<br />
drive widespread adoption. As absolute pricing<br />
becomes more attractive, SSD adoption risk is on<br />
the upside. Consumers are likely to look for the<br />
convenience of SSDs, and distribution channels<br />
are likely to overcome negative experiences in<br />
attempting to sell early (very expensive) SSD<br />
products. Another factor that could start kicking<br />
in, albeit slowly, is SSDs in desktops, an<br />
application we have advocated. Apple recently<br />
adopted this idea with the new iMac, which has<br />
the option of having either SSD or HDD (or both)<br />
in each machine.<br />
33
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
NAND makers generally believe SSDs will take<br />
off next year. Indeed, despite our argument that<br />
absolute pricing of SSDs has now reached the<br />
stage where more widespread adoption in<br />
mainstream computing is possible, we believe<br />
hesitancy in retail channels and per GB<br />
comparisons by consumers has limited SSD<br />
adoption to high-end notebooks, notably from<br />
Sony, Apple, Dell (Adamo), HP (Envy) due to<br />
high SSD prices on a per GB basis compared to<br />
hard disks. SSDs up to 512GB are now available,<br />
using 3xnm NAND, SATA interfaces and fast<br />
data rates by using toggle-mode DDR NAND.<br />
We expect SSD adoption to make faster inroads<br />
starting next year based on falling NAND ASPs<br />
and greater consumer awareness of the advantages<br />
of SSDs, in part driven by soaring awareness of<br />
tablets. We forecast SSDs to account for c16% of<br />
total NAND demand by 2012e. SanDisk expects<br />
SSD adoption to pick up in 2010, reaching c20%<br />
bits shipped by 2012; the vendor expects market<br />
acceleration in 2011e with cost reductions on the<br />
2xnm process, with a ‘tipping point’ achievable<br />
with 1xnm. Gartner forecasts SSD to approach<br />
c20% of bit shipments in 2012.<br />
Indeed, perhaps more than any other NAND<br />
application, due to their very high capacities, SSD<br />
adoption will depend on affordability, which is in<br />
turn driven by technology advances.<br />
Enterprise flash drives – huge upside<br />
risk to NAND revenue forecasts<br />
A major upside risk to our SSD forecasts is the<br />
server storage market, offering disproportionately<br />
high revenues and margins. Hynix sees enterprise<br />
and high-end PCs driving demand for SSDs,<br />
despite its scepticism over prospects for SSDs in<br />
general this year. Key beneficiaries of this market<br />
include Samsung, IMFT and Hynix, to be joined<br />
next year by Toshiba. We admit that we do not<br />
separately model the server NAND market in our<br />
forecasts. But we hope to add value in this<br />
analysis by illustrating the magnitude of the<br />
upside risk to our forecasts.<br />
SSD vendors continue to emphasize that the<br />
enterprise market remains a key demand driver.<br />
We estimate that server storage could account for<br />
the equivalent of 12-17% of our 2012e forecast<br />
global NAND revenue, and the equivalent of 20-<br />
32% our forecast 2013e NAND global revenue.<br />
These estimates assume SSDs account for 5% and<br />
10% respectively of 2011-12e disk storage<br />
systems and are based on SLC.<br />
High-end server SSD storage is very expensive.<br />
Despite this, data centres and corporates use SSD<br />
due to 1) extremely fast data input/output,<br />
especially important for cloud computing, 2)<br />
small form factor (space), 3) low power<br />
consumption. As ASPs fall, we expect SSDs will<br />
replace HDDs as ‘front line’ storage, with HDDs<br />
acting as inexpensive backup.<br />
Server SSDs use SLC as it is generally about<br />
twice as fast as MLC for read applications, and is<br />
considered more reliable due to the fact it has only<br />
one bit per cell. SLC NAND is more expensive at<br />
about 3-3.5x premium compared to equivalent<br />
MLC for 16-32Gb parts. Samsung currently<br />
produces up to 64Gb density SLC.<br />
By the time SLC NAND is incorporated into a<br />
server storage system, prices/GB soar even more.<br />
Server SSDs based on MLC NAND currently cost<br />
about USD3-4/GB. But SLC-based storage<br />
systems (not just simple SSDs) are about 10-25x<br />
more expensive on a per GB basis. The high price<br />
premium is justified by very high cost of using the<br />
best high-density SLC NAND, but also reflects<br />
systems that are optimised for high data transfer<br />
speeds, high reliability and power efficiency, in<br />
addition to close customer service. For example,<br />
Oracle charges systems starting at cUSD46k with<br />
480GB capacity (using 20 x 24GB flash modules)<br />
using SLC, equating to cUSD96/ GB. Texas<br />
34
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
<strong>Memory</strong> charges cUSD220k for a 5TB unit<br />
(cUSD37/GB) for a lower spec system.<br />
Micron produces SLC in 64-128Gb densities. It is<br />
jumped into the enterprise-class SLC-based SSD<br />
market in August this year based on 34nm process,<br />
and is in customer sampling. Samsung and Hynix<br />
have comprehensive SLC line ups, both offering<br />
up to 64Gb densities. We believe Samsung uses<br />
up to 5xnm processes for its SLC. More recently,<br />
Hynix has been validating its 32nm-based SLC<br />
(16Gb) since June 2010. Micron started mass<br />
production of 32Gb<br />
35
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
% of total NAND bit demand from DSC System unit and GB/system growth for DSC<br />
25<br />
20<br />
15<br />
10<br />
5<br />
0<br />
160<br />
150<br />
140<br />
130<br />
120<br />
25<br />
20<br />
15<br />
10<br />
5<br />
0<br />
2009 2010f 2011f 2012f<br />
2009 2010f 2011f 2012f<br />
% of NAND demand<br />
Sy stem units (m)<br />
GB/sy stem<br />
Source: IDC, HSBC estimates<br />
Source: IDC, HSBC estimates<br />
Digital imaging (DSC, DVC)<br />
Demand drivers in digital still cameras<br />
Relevance. As a proportion of total NAND<br />
demand digital imaging (still cameras plus digital<br />
video cameras) is falling, not because of a decline<br />
in usage – we forecast 2010/12e bit demand<br />
CAGR of 49% – but mostly because of slowed<br />
megapixel growth we believe. We forecast digital<br />
imaging to account for c12.4% of total 2011e<br />
NAND demand (2009: 23.3%).<br />
Upside/downside risk. Demand recovery appears<br />
relatively sound, with Nikon and others guiding a<br />
recovery in unit shipments this year. Nikon for<br />
example, guides the global DSLR market to reach<br />
12.2m by end-March 2011, up 12% y-o-y, in part<br />
supported by the emergence of mirrorless DSLRs.<br />
In the compact DSC market Nikon a 7.1% y-o-y<br />
rise in global units. Key upside risk is growth in<br />
memory usage per system as HD video has<br />
become standard in compact DSCs and DSLRs.<br />
We expect 3D still and video and RAW-based<br />
video would further boost demand starting 3-5<br />
years hence.<br />
Catalysts. Better than expected unit shipment and<br />
product mix guidance/results from major device<br />
manufacturers and flash card makers such as<br />
SanDisk.<br />
System unit growth low, but GB/<br />
system to grow steadily<br />
The dynamics of NAND demand from digital<br />
imaging is, in our view, mostly driven rising<br />
content/system, which benefits from convergence<br />
of still and video, migration of video to solid state,<br />
and high definition.<br />
Rapid camera shipments recovery<br />
Camera shipments to continue recovering next<br />
year; we forecast global digital cameras units to<br />
rise 10.7% y-o-y after a c13.8% fall last year.<br />
DSLRs lead growth this year with 13.5%<br />
compared to c10.5% unit shipment growth for<br />
compact DSCs.<br />
Global DSC unit shipments (m)<br />
170<br />
150<br />
130<br />
110<br />
90<br />
70<br />
50<br />
2007 2008 2009e 2010f 2011f 2012f 2013f<br />
Source: CIPA, HSBC estimates<br />
Compact<br />
DSLR<br />
In particular, we estimate mirror-less DSLR to<br />
account for c13% of total 2010 DSLR shipments<br />
36
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
(which is similar to Nikon’s estimate of c1.2m<br />
units to March 2011, excluding Samsung).<br />
GB/system growth diverges<br />
We forecast digital still camera GB/system to rise<br />
c30% and 37% for 2011-12e. In comparison we<br />
expect digital video GB/system to rise 50% per<br />
year for the same period. We believe the biggest<br />
beneficiaries of this trend will be flash card<br />
vendors. Even for digital video the trend is<br />
irreversibly for non-linear, high-speed storage.<br />
For digital still cameras, video convergence<br />
drives demand. In the past six months high<br />
definition video recording (720-1080p) – which<br />
is very data storage-intensive – has become<br />
standard on most compact cameras and DSLRs.<br />
As a rough comparison, a 16GB flash card can<br />
store over 3,300 12MP JPEG pictures, but only<br />
3 hours of 720p AVCHD video.<br />
Tape-less camcorders prevail. Both<br />
consumer and professional level camcorders<br />
have transitioned rapidly to flash-based<br />
recording in the past 12 months as the<br />
convenience of flash wins out and<br />
affordability of flash cards increases.<br />
HD technology maturing. Helping HD video<br />
adoption on both DSCs and camcorders is the<br />
maturity of a key compression technology,<br />
AVCHD (needs less storage space than other<br />
compression technologies).<br />
Slowing camera resolution growth drags<br />
on content/system. Still camera vendors now<br />
focus on image quality, de-emphasising as a<br />
key selling point absolute pixel counts. We<br />
expect steady resolution growth in the next<br />
few years, mostly in larger sensor DSCs as<br />
small sensors are limited by intrusive noise at<br />
higher pixel counts.<br />
Cannibalisation by mobile phones is<br />
limited, for now; megapixel growth of builtin<br />
cameras on mobile phones have slowed<br />
substantially, partially due to handset<br />
vendors’ desire to limit component costs. But<br />
HD video functionality is rising as sufficient<br />
smartphone processor power to support it<br />
becomes available.<br />
We expect adoption of video on DSCs to<br />
accelerate with rising adoption of mirrorless<br />
SLRs that use larger (APS-C sized) sensors.<br />
Indeed, Sony and Panasonic are at the<br />
vanguard, launching high-end amateur/low<br />
end professional video equipment based on<br />
compact camera lens mounts.<br />
Longer term, i.e. starting 3-5 years from now,<br />
we expect 3D and RAW-based video to start<br />
to drive higher GB/system for both still and<br />
video imaging.<br />
37
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
% of total NAND bit demand from USB flash drives System unit and GB/system growth for USB flash drives<br />
12<br />
10<br />
8<br />
6<br />
4<br />
2<br />
0<br />
400<br />
300<br />
200<br />
100<br />
0<br />
12<br />
10<br />
8<br />
6<br />
4<br />
2<br />
0<br />
2009 2010f 2011f 2012f<br />
2009 2010f 2011f 2012f<br />
% of NAND demand<br />
Sy stem units (m)<br />
GB/sy stem<br />
Source: IDC, HSBC estimates<br />
Source: IDC, HSBC estimates<br />
USB flash drive<br />
Relevance: Universal Serial Bus (USB) flash<br />
drives (NAND storage plus integrated interface)<br />
accounted for c11% of NAND bit demand last year,<br />
but growing at a CAGR of 69.5% during the period.<br />
Upside/downside factors. Key upside risk to our<br />
forecast would be faster than expected adoption of<br />
USB3.0, and strong price elasticity should NAND<br />
prices decline. Key downside risks are the<br />
uncertainty over Intel’s adoption rate and high<br />
NAND prices suppressing GB/system growth.<br />
Catalysts. Better than expected unit shipment and<br />
product mix guidance/results from major device<br />
manufacturers and USB storage makers such as<br />
SanDisk, A-Data, and PQI.<br />
USB3.0 – key multi-year driver<br />
starting 2011<br />
Convenience, durability, increasingly higher<br />
capacity drives, and affordability support demand.<br />
Rapid data rates. USB3.0 drives can transfer<br />
data 5-10 times faster than USB2.0 drives; a<br />
600MB movie file takes c5 seconds to download.<br />
Higher-end USB3.0 flash drives start to rival<br />
SSDs in data speeds, enabling external storage of<br />
very large data amounts due to fast transfer times.<br />
Launch of other high-capacity standards for<br />
flash cards will help USB3.0 adoption. Secure<br />
Digital Extended Capacity (SDXC) products<br />
recently launched, addressing up to 2,048GB<br />
(2TB). More recently, CompactFlash Association<br />
released CF5.0 standards, which can address 147<br />
million GB, compared to the current 137GB.<br />
38
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Time taken to read a 600MB movie file<br />
40<br />
30<br />
20<br />
10<br />
0<br />
USB 2.0<br />
drive<br />
Source: Companies, HSBC<br />
USB 3.0<br />
drive<br />
Time taken (s)<br />
SSD driv e<br />
(regular)<br />
SSD drive<br />
(X25)<br />
Product availability. USB3.0 flash drives started<br />
launching at end-2009, including those by Super<br />
Talent and PQI. Vendors including Gigabyte,<br />
Asus and MSI (motherboards), Genesys (device<br />
controllers) VIA (hub controller) have launched<br />
USB3.0-standard products this year.<br />
Data rate comparison for common PC interfaces/storage<br />
Interface/device Launch Data rate (MB/s)<br />
USB<br />
1.0 1Q96 0.2<br />
1.1 3Q98 1.5<br />
2.0 2Q00 60<br />
3.0 4Q09 600<br />
Wireless 4Q06 14-60<br />
But a key factor in the pace of USB3.0 adoption will<br />
be Intel’s push (or lack thereof) for the standard; the<br />
firm’s support is critical in device makers’ adoption<br />
of the standard. Intel delayed roll out of chipsets<br />
supporting USB3.0, but in September 2010 Intel<br />
confirmed it would include support for USB3.0 in its<br />
new chipset that would start shipping in the next 2-3<br />
quarters. AMD is likely to offer chipsets that support<br />
USB 3.0 by 1Q11.<br />
Increase in capacities, affordability. We<br />
forecast NAND ASPs to fall to cUSD1/GB by<br />
2011. This means a 64GB USB3.0 flash drive<br />
could cost cUSD70. Ramp-up of USB3.0 flash<br />
drives demand will be faster than for USB2.0<br />
drives due to better affordability of NAND. Bit<br />
demand from flash drives took c7 years to surge<br />
after USB2.0 was launched in 2Q00; it took that<br />
long for affordability and widespread adoption of<br />
USB2.0. Currently, up to 128GB flash drives are<br />
widely available; these are expensive and data<br />
rates are relatively slow.<br />
IEEE1394<br />
FireWire 400 1995 50<br />
FireWire 800 2Q05 100<br />
FireWire S1600 4Q10 200<br />
FireWire S3200 4Q10 400<br />
eSATA 2004 188<br />
SSD drives<br />
Samsung 256GB MLC 200/160<br />
Intel 160GB MLC 250/100<br />
USB flash drives<br />
Corsair Flash Voyager GTR 34/28<br />
Ironkey secure drive 27/24<br />
Source: Companies, HSBC<br />
39
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
% of total NAND bit demand from media players System unit and GB/system growth from media players<br />
30<br />
25<br />
20<br />
15<br />
10<br />
5<br />
0<br />
200<br />
150<br />
100<br />
50<br />
0<br />
40<br />
30<br />
20<br />
10<br />
0<br />
2009 2010f 2011f 2012f<br />
2009 2010f 2011f 2012f<br />
% of NAND demand<br />
Sy stem units (m)<br />
GB/sy stem<br />
Source: IDC, HSBC estimates<br />
Source: IDC, HSBC estimates<br />
Media players<br />
Relevance: NAND for media players accounted<br />
for c25.9% of NAND bit demand last year; we<br />
expect this proportion to fall to c6% by 2012e.<br />
We think that media players as a segment, despite<br />
launch of newer products by Apple and others is<br />
being gradually absorbed by smartphones and<br />
going forward especially, tablets.<br />
Upside/downside factors. Upside risks include<br />
appearance of dedicated, full HD and 3D media<br />
players. Downside risks include faster than<br />
expected substitution by smartphones and tablets.<br />
Catalysts. Catalysts for NAND demand by media<br />
players include a ‘reinvention’ of media players to<br />
include web surfing.<br />
Substitution by smartphones will accelerate, as:<br />
Smartphone screens become larger and higher<br />
resolution – Google’s Nexus One and has a<br />
800x480 screen, as do others.<br />
Smartphones become faster. The latest<br />
generation of smartphones such as the iPhone<br />
and Samsung Galaxy S runs 1GHz mobile<br />
processors with high-speed graphics<br />
processors.<br />
The segment is set to be cannibalised by tablets<br />
represented by such as Apple’s iPad, and Android<br />
-based such as Samsung’s Galaxy Tab, which<br />
offer the huge flexibility of running a wide range<br />
of apps as well access to wireless web-based<br />
streamed audio and video content.<br />
40
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Supply drivers – DRAM<br />
Oversupply likely to persist through 2Q11 on sharply rising<br />
capacity from second-tier vendors, process migration<br />
Capex increases set to peak this year, with next year set to see<br />
single-digit rises, mostly for continued process migration<br />
Leading vendors shift towards more value-added high-end, mobile<br />
DRAM<br />
In retrospect…<br />
An anomalous state – supply<br />
squeeze in 1H leads to acute<br />
oversupply in 2H<br />
In March, we assumed that ASP decline would be<br />
gradual through the year. Instead, we are set to see<br />
an inverted-V ASP trend through this year. Key<br />
factors that impacted our original assumptions<br />
were the European financial crisis and slower than<br />
expected migration of second tier vendors. We<br />
had expected c8% average DDR3 ASP decline<br />
every quarter starting in 2Q10. But ASPs actually<br />
rose c6.8% during 2Q. In 1Q10 average DDR3<br />
contract prices rose 12% q-o-q. In 2Q10, ASPs<br />
continued to rise on PC demand recovery and<br />
tight DRAM supply, and rose further as PC OEMs<br />
pulled in demand into 2Q10 from 3Q10 to counter<br />
the impact of a weak Euro zone, a situation<br />
aggravated by the Greek debt crisis. In the<br />
meantime, capacity remained relatively<br />
constrained as second-tier vendors struggled to<br />
sort process migrations. For leading vendors the<br />
DRAM industry environment in 1H10 was<br />
optimal. Samsung and Hynix not only enjoyed a<br />
favourable operating environment (from their<br />
point of view especially), but margins also<br />
benefited from early process migration to 5xnm<br />
and 4xnm during the period.<br />
And then the pin dropped. The combination of<br />
stronger demand and weaker supply than expected<br />
in 1H10 was setting the industry up for a weakerthan-expected<br />
2H10. From 3Q onwards DRAM<br />
vendors face low q-o-q PC OEM demand growth<br />
partly due to a high 2Q comparison base and<br />
partly due to PC OEMs taking a more cautious<br />
demand outlook for the rest of the year.<br />
Exacerbating the situation is the prospect of<br />
soaring bit supply starting in 4Q10 as second tier<br />
vendors reach production yields for smaller nodes,<br />
plus Samsung raises wafer capacity. We expect<br />
DDR3 ASPs for 4Q10-1Q11 (averages for the<br />
quarter) to fall 18% and 11% q-o-q respectively.<br />
Nevertheless, when examining next year’s<br />
outlook, we need to take a step back and note that<br />
y-o-y comparisons should take into account that<br />
2010 is a peak year; despite sharp ASP decline in<br />
2H10, average DRAM prices for the full year are<br />
set to rise y-o-y (15% for DDR3), an unusual<br />
situation for the industry.<br />
41
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Short-term oversupply<br />
After a series of years of declining revenue<br />
growth, we estimate global DRAM revenue to rise<br />
79% y-o-y this year, driven by 49% bit growth<br />
and blended ASP rise of c20%. For 2011, the<br />
situation is clearly different; we forecast global<br />
revenue to be flat (+1% y-o-y) as blended ASPs<br />
decline 32% y-o-y to below USD1.9, offsetting<br />
the c49% rise in bit supply. Over the next 2-3<br />
quarters we expect the DRAM industry to suffer<br />
from pricing pressure on greater supply as smaller<br />
players overcome process migration difficulties<br />
and as larger players accelerate their own process<br />
migrations, plus in the case of Samsung, adds<br />
wafer capacity. Nevertheless, we forecast DRAM<br />
industry revenue to reach cUSD46bn by 2012e,<br />
almost doubling from last year’s cUSD24bn.<br />
DRAM revenue, supply bit growth, ASP trends<br />
60,000<br />
50,000<br />
40,000<br />
30,000<br />
20,000<br />
10,000<br />
-<br />
Source: IDC, HSBC estimates<br />
2008 2009 2010f 2011f 2012f<br />
Supply (1Gb, m)<br />
DRAM Sales (USDm)<br />
Blended ASP, 1Gb equ. (USD)<br />
Oversupply likely until 2Q11<br />
3.0<br />
2.5<br />
2.0<br />
1.5<br />
1.0<br />
0.5<br />
-<br />
We forecast supply to continue outstripping; we<br />
estimate supply-demand sufficiency last year was<br />
c99.9%, and expect it to remain relatively tight<br />
until 4Q10, when surging supply is set to<br />
unbalance the market. Next year, we think<br />
oversupply could persist through 2Q11, with a<br />
more balanced supply returning in 2H11.<br />
DRAM supply-demand balance<br />
40,000<br />
30,000<br />
20,000<br />
10,000<br />
-<br />
Source: HSBC forecasts<br />
2008 2009 2010f 2011f 2012f<br />
106<br />
104<br />
102<br />
100<br />
Supply (1Gb, m) Supply -demand ratio (%)<br />
We expect DDR3 average contract prices to fall<br />
18% q-o-q in 4Q10 respectively. This appears<br />
realistic; in 3Q10 average contract prices for<br />
DDR3 and DDR2 have retreated 13% q-o-q,<br />
versus our forecast of 15%. This level of price<br />
erosion is well above Samsung and Hynix’s 3Q10<br />
guidance of blended ASP decline in high single<br />
digits, and a ‘slight’ fall respectively (though the<br />
two firm’s ASP guidance is for blended, not for<br />
commodity DRAM).<br />
DDR2 contract and spot price (USD)<br />
3.5<br />
3.0<br />
2.5<br />
2.0<br />
1.5<br />
1.0<br />
0.5<br />
Jan-08 Jul-08 Jan-09 Jul-09 Jan-10 Jul-10<br />
1Gb DDR2 128Mx 8 667 Mhz contract<br />
1Gb DDR2 128Mx 8 667 Mhz spot<br />
Source: DRAMeXchange<br />
DDR3 enjoyed a 7-15% premium through much<br />
of 9M10, despite DDR2/3 production cross-over<br />
during 1-2Q10 due to strong demand for newer,<br />
faster memory by PC OEMs. However, in<br />
September, the premium narrowed to just 3% as<br />
more vendors ramp up on DDR3.<br />
98<br />
96<br />
94<br />
42
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Nevertheless, we expect demand downside for<br />
commodity DRAM to be relatively limited due to<br />
rising greater affordability of the modules. We<br />
expect DRAM price declines to start slowing in<br />
2Q11 as lower ASPs from sharply higher supply<br />
during 4Q10-1Q11 stimulate higher content/<br />
system and as supply increases peak out. Overall,<br />
we expect content/system to rise 33% next year,<br />
following a 27% y-o-y rise in 2010; high ASPs<br />
this year depressed content growth this year,<br />
though the extent of DRAM content reduction or<br />
freeze appears to vary according to PC OEMs.<br />
DDR3 contract and spot price (USD)<br />
3<br />
2.5<br />
2<br />
1.5<br />
1<br />
Jul-09<br />
Oct-09<br />
Source: DRAMeXchange<br />
Jan-10<br />
Process migration<br />
Apr-10<br />
Jul-10<br />
1Gb DDR3 128Mx 8 1066MHz spot<br />
1Gb DDR3 128Mx 8 1066MHz contract<br />
We believe much of the DRAM bit supply growth<br />
is coming from process migration, with relatively<br />
low actual wafer capacity additions; we forecast<br />
c7% q-o-y wafer capacity increase, but a 49%<br />
supply increase for 2010. The race to improve<br />
processes for cost and product competitiveness is<br />
a key feature of the current spate of bit growth.<br />
DRAM process migration – Samsung Electronics<br />
100%<br />
80%<br />
60%<br />
40%<br />
20%<br />
0%<br />
2007 2008 2009 2010f 2011f 2012f<br />
8xnm 6xnm 5xnm 4xnm 3xnm 2xnm<br />
Source: Company, HSBC estimates<br />
Process migration is increasing the number of dies<br />
per given wafer. Wafer capacity was relatively<br />
stable in the past two years, helped in part by<br />
removal of uneconomic 200mm fab capacity.<br />
DRAM capex fell copiously after peaking in 2007,<br />
hitting a nadir of cUSD4.5bn last year. We<br />
believe most expenditure was on maintenance and<br />
process migration.<br />
DRAM supply<br />
25,000<br />
20%<br />
20,000<br />
15%<br />
15,000<br />
10%<br />
10,000<br />
5%<br />
5,000<br />
0%<br />
-<br />
-5%<br />
2008 2009 2010f 2011f 2012f<br />
Wafer Capacity (12" equ. kw py , kw pm)<br />
Change (%)<br />
Source: Gartner, HSBC estimates<br />
This year, we estimate capex to rise c147% to<br />
cUSD11.2bn as process migration accelerates and<br />
some wafer capacity is added (about 7%, we<br />
estimate), notably by Samsung. Next year, we<br />
expect DRAM capex to rise only 9% y-o-y, with<br />
the bulk of it on process migration; we only<br />
expect wafer capacity we expect wafer capacity to<br />
rise 4% y-o-y. These capex estimates translate<br />
into capex/sales of c26% this year and c28% next<br />
43
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
year. Reassuringly, this is still well below the<br />
2005-08 average of 54%.<br />
Global DRAM capex (USDm)<br />
DRAM contract prices<br />
3<br />
2.5<br />
15,000<br />
200%<br />
2<br />
10,000<br />
150%<br />
100%<br />
1.5<br />
50%<br />
1<br />
5,000<br />
-<br />
2008 2009e 2010f 2011f 2012f<br />
0%<br />
-50%<br />
-100%<br />
Jul-09<br />
Oct-09<br />
Jan-10<br />
Apr-10<br />
Jul-10<br />
1Gb DDR2 128Mx 8 667 Mhz contract<br />
1Gb DDR3 128Mx 8 1066MHz contract<br />
Source: DRAMeXchange<br />
Source: Gartner, HSBC forecasts<br />
Total (HSBC) y-o-y (%)<br />
The off-shoot of this is surging bit supply in<br />
4Q10-1Q11, when the full brunt of extra capacity<br />
by Samsung, plus commercial yields from process<br />
migration by second-tier players such as Nanya<br />
and Inotera, comes into effect, as they bring on<br />
line 5xnm capacity that was held up in the past 2-<br />
3 quarters due to yield issues. We believe Nanya<br />
is now achieving c70% yield for its 5xnm process,<br />
which is sufficient to mass produce. We expect<br />
global supply bit growth in 3Q-4Q10 to rise 11%<br />
and 18% q-o-q respectively.<br />
Interestingly, DDR2 and DDR3 price declines<br />
have converged recently; DDR3 contract prices<br />
plunged 10-11% in 1H September. We believe the<br />
decline in DDR3 prices reflects higher supply<br />
levels following DDR2/3 production cross-over at<br />
leading players in 1Q10 and higher DDR3<br />
production at second-tier players. ASP decline has<br />
also been impacted by demand side; we believe<br />
this ASP decline reflects unwillingness on the part<br />
of PC OEMs to build memory inventory on low<br />
PC unit shipment growth this quarter following<br />
very strong 2Q shipments, plus their anticipation<br />
of even lower prices.<br />
Indeed, DRAM contract ASPs have remained<br />
under pressure after hitting a peak in June. From<br />
their peaks in May/June, DDR3 module prices are<br />
down 27% and DDR2 is down 23% on a periodend<br />
to period-end basis, according to<br />
DRAMeXchange. 2GB DDR3 modules, the<br />
mainstay of PC memory now (DDR2 is almost<br />
defunct as 80%-plus of PC OEM/ODMs have<br />
migrated to DDR3) fell to USD34.<br />
44
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Global DRAM supply by vendor, units (512Mb equ., m)<br />
Units 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e<br />
Samsung 6,581 11,207 17,040 24,735 2,196 2,525 3,029 3,457 3,685 3,980 4,490 4,885<br />
y-o-y (%) 43% 70% 52% 45% 15% 15% 20% 14% 7% 8% 13% 9%<br />
% of total 32% 36% 37% 37% 33% 35% 38% 37% 37% 37% 37% 37%<br />
Hynix 4,382 6,237 9,389 13,564 1,362 1,460 1,567 1,848 2,022 2,174 2,478 2,715<br />
y-o-y (%) 20% 42% 51% 44% 6% 7% 7% 18% 9% 8% 14% 10%<br />
% of total 21% 20% 20% 21% 20% 21% 20% 20% 20% 20% 21% 21%<br />
Qimonda 223<br />
y-o-y (%) -88%<br />
% of total 1%<br />
Elpida 4,114 5,640 7,800 10,489 1,250 1,350 1,415 1,625 1,720 1,899 1,999 2,182<br />
y-o-y (%) 57% 37% 38% 34% -7% 8% 5% 15% 6% 10% 5% 9%<br />
% of total 20% 18% 17% 16% 19% 19% 18% 17% 17% 18% 17% 16%<br />
Micron 2,244 3,568 5,407 8,559 852 859 868 989 1,147 1,272 1,416 1,572<br />
y-o-y (%) 90% 59% 52% 58% 11% 1% 1% 14% 16% 11% 11% 11%<br />
% of total 11% 12% 12% 13% 13% 12% 11% 11% 11% 12% 12% 12%<br />
Others 4,186 4,323 6,507 8,632 1,017 919 993 1,394 1,436 1,522 1,664 1,884<br />
y-o-y (%) -7% 3% 51% 33% -26% -10% 8% 40% 3% 6% 9% 13%<br />
% of total 20% 14% 14% 13% 15% 13% 13% 15% 14% 14% 14% 14%<br />
Total 20,826 30,975 46,142 65,979 6,676 7,114 7,872 9,313 10,010 10,846 12,048 13,238<br />
y-o-y (%) 22% 49% 49% 43% 5% 7% 11% 18% 7% 8% 11% 10%<br />
% of total 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100%<br />
Bit market share (%) 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e<br />
Samsung 32% 36% 37% 37% 33% 35% 38% 37% 37% 37% 37% 37%<br />
Hynix 21% 20% 20% 21% 20% 21% 20% 20% 20% 20% 21% 21%<br />
Qimonda 1% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0%<br />
Elpida 20% 18% 17% 16% 19% 19% 18% 17% 17% 18% 17% 16%<br />
Micron 11% 12% 12% 13% 13% 12% 11% 11% 11% 12% 12% 12%<br />
Others 20% 14% 14% 13% 15% 13% 13% 15% 14% 14% 14% 14%<br />
Bit growth (y-o-y) 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e<br />
Samsung 43% 70% 52% 45% 15% 15% 20% 14% 7% 8% 13% 9%<br />
Hynix 20% 42% 51% 44% 6% 7% 7% 18% 9% 8% 14% 10%<br />
Qimonda -88% na na na na na na na na na na na<br />
Elpida 57% 37% 38% 34% -7% 8% 5% 15% 6% 10% 5% 9%<br />
Micron 90% 59% 52% 58% 11% 1% 1% 14% 16% 11% 11% 11%<br />
Others -7% 3% 51% 33% -26% -10% 8% 40% 3% 6% 9% 13%<br />
Total 22% 49% 49% 43% 5% 7% 11% 18% 7% 8% 11% 10%<br />
Total Shipment, MB bn 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e<br />
MB (bn) 1,333 1,489 1,638 1,795 427 455 504 596 641 694 771 847<br />
Units (512Mb Eq., m) 20,826 30,975 46,142 65,979 6,676 7,114 7,872 9,313 10,010 10,846 12,048 13,238<br />
Source: IDC, HSBC estimates<br />
45
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Global DRAM capacity<br />
Capacity kwpy, kwpm 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e<br />
Samsung 4,500 4,650 5,040 5,850 370 370 390 420 420 420 420 420<br />
Hynix 3,210 3,600 3,600 4,020 300 300 300 300 300 300 300 300<br />
Micron 1,508 2,027 2,027 2,027 169 169 169 169 169 169 169 169<br />
Qimonda 915 – – - - - - - - - - -<br />
Elpida 2,535 3,195 3,480 3,735 260 260 265 290 290 290 290 290<br />
Other 4,318 4,735 4,735 4,735 390 390 390 390 390 390 390 390<br />
Total 16,985 18,206 18,881 20,366 1,489 1,489 1,514 1,569 1,569 1,569 1,569 1,569<br />
Capacity change (%)<br />
Samsung 3% 3% 8% 16% 0% 0% 5% 8% 0% 0% 0% 0%<br />
Hynix -4% 12% 0% 12% 9% 0% 0% 0% 0% 0% 0% 0%<br />
Micron 12% 34% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0%<br />
Qimonda -16% 0% 0% 0%<br />
Elpida 0% 26% 9% 7% 18% 0% 2% 9% 0% 0% 0% 0%<br />
Other -2% 10% 0% 0% 1% 0% 0% 0% 0% 0% 0% 0%<br />
Total -1% 7% 4% 8% 0% 0% 2% 4% 0% 0% 0% 0%<br />
% of total<br />
Samsung 26% 26% 27% 29% 25% 25% 26% 27% 27% 27% 27% 27%<br />
Hynix 19% 20% 19% 20% 20% 20% 20% 19% 19% 19% 19% 19%<br />
Micron 9% 11% 11% 10% 11% 11% 11% 11% 11% 11% 11% 11%<br />
Qimonda 5% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0%<br />
Elpida 15% 18% 18% 18% 17% 17% 18% 18% 18% 18% 18% 18%<br />
Other 25% 26% 25% 23% 26% 26% 26% 25% 25% 25% 25% 25%<br />
Source: IDC, HSBC estimates<br />
46
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Supply drivers – NAND<br />
2010-11e supply-demand profile remains benign, as makers<br />
restrain wafer capacity to maximise profitability, demand soars<br />
Process migration, multi-bits drive bit growth, lowering costs,<br />
product competitiveness; leaders moving rapidly to 2xnm<br />
Capex set to rise 178% and 25% y-o-y in 2010-11e; additional<br />
supply from newcomers limited<br />
Summary<br />
NAND remains an exciting structural growth<br />
area, driven by an explosion in applications of the<br />
robust, lower power storage memory. We expect<br />
demand to continue outstripping supply for the<br />
most part of next year. We estimate supplydemand<br />
sufficiency for 2010 to be c99.5%,<br />
compared to last year’s 100.4%. We forecast<br />
2010/12e CAGR for revenue of 24%, driven by<br />
100% bit supply growth, offset by a c38% fall in<br />
ASP during this period. We forecast NAND<br />
industry revenue to reach cUSD24.8bn by 2012e,<br />
more than doubling from last year’s estimated<br />
cUSD12bn.<br />
We expect ASP to decline c38% next year, after<br />
falling c20% this year as supply starts increasing<br />
especially in 2H11. Indeed, contact NAND prices<br />
have held up well since the start of 4Q09 for the<br />
higher capacities of 32-64GB, though they did<br />
play catch up with the lower densities in 3Q10,<br />
falling c14-16% q-o-q.<br />
NAND ASP (contract)<br />
20<br />
15<br />
10<br />
5<br />
0<br />
Source: DRAMeXchange<br />
Jan-<br />
07<br />
Jul-<br />
07<br />
Jan-<br />
08<br />
Jul-<br />
08<br />
Jan-<br />
09<br />
Jul-<br />
09<br />
Jan-<br />
10<br />
16Gb (2Gbx 8) MLC Flash Spot<br />
16Gb (2Gbx 8) MLC Flash Contract<br />
Jul-<br />
10<br />
Driving bit supply is new wafer capacity and<br />
continued process migration. New wafer capacity<br />
add accelerates, mostly impacting starting 2H11.<br />
This year, we estimate wafer capacity to increase<br />
c14% y-o-y, after falling about 5% last year. Next<br />
year, we forecast c20% increase in wafer capacity,<br />
led by Toshiba and Samsung. Process migration<br />
and multi-level<br />
47
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Margins benefit. Cost reductions from process<br />
migration, multi-bits outpace ASP declines. We<br />
expect cost leaders, namely Toshiba and Samsung<br />
to benefit from aggressive cost reductions. So<br />
despite falling ASP, we expect margins to expand.<br />
NAND revenue, supply bit growth and ASP trends<br />
50,000<br />
8.0<br />
40,000<br />
6.0<br />
30,000<br />
4.0<br />
20,000<br />
10,000<br />
2.0<br />
-<br />
-<br />
2008 2009 2010f 2011f 2012f<br />
Supply (GB, m)<br />
Rev enue (USD mn)<br />
ASP (USD, 16Gb equ.)<br />
Source: Gartner, HSBC forecasts<br />
For example, we estimate Samsung would achieve<br />
operating margin of almost 24% this year.<br />
Toshiba’s 1QFY11 operation margin from its<br />
memory business was 15%, and the company<br />
expects it to rise to achieve c15-20% for 2QFY11<br />
(end-September 2011). Similarly, SanDisk<br />
expects cost reductions/GB to be
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
bring NAND capex/sales to 50%, still well below<br />
the 2005-8 average level of 67%. Other than<br />
accelerated process migration, we expect Samsung<br />
to expand its NAND production capacity probably<br />
at Fab 16, which we expect to be online next year,<br />
to compete with Toshiba’s new NAND fab (Fab 5)<br />
at its facility in Mie Prefecture, Japan; this is<br />
expected to come online in spring 2011.<br />
Global NAND capex (USDm)<br />
14,000<br />
12,000<br />
10,000<br />
8,000<br />
6,000<br />
4,000<br />
2,000<br />
-<br />
Source: Gartner, HSBC forecasts<br />
2009 2010f 2011f 2012f<br />
Total (HSBC) y-o-y (%)<br />
200%<br />
150%<br />
100%<br />
50%<br />
0%<br />
-50%<br />
-100%<br />
We are not alarmed by this increase. Based on our<br />
capacity assumptions for these two major new<br />
fabs and process migrations for the major vendors,<br />
we expect 2011 supply-demand balance to be<br />
similar to this year’s, at 99.6%. We expect waferdriven<br />
NAND capacity additions to hit in earnest<br />
during 2H11, by which time we believe strong<br />
demand growth can absorb the extra production.<br />
Indeed, we expect a major shortage of supply by<br />
next year, if Toshiba’s new Fab 5, and as appears<br />
possible, Samsung’s Fab 16 do not come online.<br />
Furthermore, we believe leading makers –<br />
Toshiba and Samsung – remain cautious on the<br />
rate of capacity ramp up, balancing it against bit<br />
gains from process migration and attempting to<br />
avoid the debilitating overcapacity that plagued<br />
the industry for years.<br />
From a capex/sales perspective, rising capex is<br />
not excessive, either in our view. We estimate<br />
capex/sales for the global NAND industry to<br />
approach and remain at c45-50% during 2010-11e.<br />
This is a steep increase over the anomalous c25%<br />
in 2009, but compares favourably compared to the<br />
over-build situation in 2006-08, when capex/sales<br />
varied between 58% (2008) and 82% (2007). The<br />
sharp forecast capex rise follows two consecutive<br />
years of severe cut backs, during which the vast<br />
majority of less efficient 200mm fab capacity was<br />
decommissioned.<br />
Process migration accelerates…<br />
Despite wafer adds, process migration, multi-level<br />
cells and higher utilisations following migration<br />
to immersion lithography, still drive much of the<br />
NAND bit supply growth. We forecast NAND<br />
supply growth of c114% y-o-y next year, though<br />
wafer capacity is rising c20% y-o-y, as leading<br />
vendors all migrate to 2xnm processes. Already,<br />
3xnm has become mainstream. For Toshiba,<br />
32nm and 24nm scale will account for c90% of its<br />
bit capacity by end of this year. The race to<br />
improve processes between leaders for cost and<br />
product competitiveness is a key feature of the<br />
current spate of bit growth.<br />
NAND process migration – Samsung Electronics<br />
100%<br />
80%<br />
60%<br />
40%<br />
20%<br />
0%<br />
2007 2008 2009 2010f 2011f 2012f<br />
8xnm 6xnm 5xnm 4xnm 3xnm 2xnm<br />
Source: Company, HSBC estimates<br />
A combination of push and pull are emboldening<br />
makers to commit capex to put as many dies on a<br />
wafer as possible, led by node migration:<br />
Strong demand (manufacturers’ consensus is<br />
for c80-90% y-o-y bit growth this year); we<br />
forecast 73% y-o-y bit growth this year.<br />
49
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Stronger balance sheets repaired through<br />
recapitalisation and lower losses/better<br />
profitability, and<br />
Intensifying technology competition (more<br />
advanced processes, multi-bit) to drive better<br />
margins, market positioning.<br />
We expect industry-wide migration to more<br />
advanced processes to lower cost/bit, raising<br />
margins:<br />
Samsung started mass producing 27nm-process<br />
based NAND in April 2010 (development<br />
completed in October 2009) and recently<br />
launched 27nm 32Gb MLC NAND.<br />
Toshiba started mass production of 24nmbased<br />
64Gb MLC NAND in 3Q10, expanding<br />
to include 24nm-based 32Gb NAND and TLC<br />
at a later point. This appears ahead of its<br />
original schedule to achieve 24nm migration<br />
(X2, X3) production by 4Q10 and mass<br />
production in 2011. Like Samsung and Hynix,<br />
Toshiba’s 24nm process products have<br />
Toggle DDR to enhance data transfer speeds.<br />
Intel and Micron (IMFT – Intel Micron Flash<br />
Technologies)’s 25nm-process based 64GB<br />
NAND mass production started in 2Q10.<br />
Hynix is admittedly late to 3xnm, but believes<br />
it has caught up with leading vendors with its<br />
2xnm migration. Hynix plans mass<br />
production in 3Q10 for its 26nm process<br />
based 64GB NAND. We believe Hynix has<br />
customer samples out for this product.<br />
current processes. By 2014-15 EUV lithography is<br />
a likely requirement for the migration to processes<br />
below 1x nm.<br />
Wafer capacity increases<br />
Wafer capacity was relatively stable in the past<br />
two years, helped by removal of uneconomic<br />
200mm fab capacity. This year, we estimate wafer<br />
capacity to increase c14% y-o-y as process<br />
migration accelerates and additional fab capacity<br />
starts to get built out. Next year, we forecast c20%<br />
increase in wafer capacity, led by Toshiba and<br />
possibly Samsung.<br />
SanDisk/Toshiba<br />
SanDisk is focussed on upgrading processes and<br />
implementing X3 in Fab 3 and Fab 4.<br />
Fab 4 expansion. This year, Toshiba has ramped<br />
previously unused space in Fab 4, which should<br />
reach full capacity by mid-2011, i.e. by the<br />
completion of Phase 1 of Fab 5. The original<br />
design wafer capacity of Fab 4 is 210kwpm, but<br />
we expect Toshiba to produce more given<br />
throughput improvements it has made.<br />
Fab 5 finally gets the go ahead. In mid-July<br />
2010 Toshiba started construction of Fab 5 at<br />
Yokkaichi. Notably it is capable of supporting<br />
EUV and 3D flash. Fab 5 will be constructed in<br />
two phases; SanDisk expects the fab to take<br />
several years to ramp to full capacity. Phase 1 will<br />
be producing 24nm process based NAND, whilst<br />
Phase 2, for which timing has not been decided,<br />
will produce 1xnm generation NAND.<br />
One key question that remains unanswered is<br />
capex from 2012e onwards, as vendors ponder the<br />
comparative returns on migration to 1xnm.<br />
Toshiba and Samsung remains undecided over<br />
how to tackle this issue, though both firms are<br />
evaluating EUV. SanDisk expects to be on 1xnm<br />
(X2 and X3) by end 2012/early 2013. But at that<br />
level, the technology is more ‘challenging’ than<br />
50
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Global NAND wafer capacity (‘000 wafers per year)<br />
1,500<br />
30%<br />
20%<br />
1,000<br />
10%<br />
500<br />
0%<br />
-<br />
-10%<br />
2008 2009 2010f 2011f 2012f<br />
Total capacity (kwpm) y-o-y (%)<br />
Source: Gartner, HSBC estimates<br />
We estimate Fab 5 would cost cUSD8-9bn in total.<br />
SanDisk previously estimated that a new NAND<br />
fabs with support for EUV would cost cUSD8bn<br />
for 200kwpm (current SanDisk/ Toshiba capacity<br />
is c270-280kwpm). SanDisk expects the capex to<br />
be spread over 2011-2014.<br />
Fab 5 phase one is to be complete in spring 2011,<br />
initially producing 2xnm NAND; its contribution<br />
to wafer starts will only become meaningful in<br />
2H11 and represent less than 10% of SanDisk’s<br />
total for 2011. As such, SanDisk sees bit growth<br />
next year driven by process migration to 24nm,<br />
and not so much by wafer capacity.<br />
Toshiba signed an agreement with SanDisk to<br />
form a new joint venture to operate the fab.<br />
SanDisk has the option to participate in up to 50%<br />
in the investment and corresponding capacity<br />
allocation at any point.<br />
Samsung and Hynix. We are unaware of specific<br />
NAND fab capacity additions for Samsung.<br />
However, we believe that a potential KRW2trn<br />
additional spend on Fab 16 could be NANDrelated.<br />
Hynix is to raise M11 fab wafer capacity<br />
to c80kwpm from the current c50-60kwpm to<br />
replace the 200mm fabs capacity shut down in<br />
2008-09. Hynix has further room to expand<br />
NAND space in M11 and M12.<br />
Micron is spending about a third of its USD850-<br />
950m capex on NAND; Intel will match the<br />
amount for its 50% stake in the Intel-Micron<br />
NAND JV.<br />
Multi-bit levels – another driver of bit<br />
growth without wafer additions<br />
Unlike DRAM, NAND cost reductions are<br />
progressing in two areas – process migration and<br />
the move to multiple bits per layer. The majority<br />
of current NAND devices are multi-level cell,<br />
using 2 bits per cell. As process migration<br />
becomes more difficult, NAND vendors are<br />
raising the number of bits per cell (which<br />
increases the amount of data that each cell can<br />
store on a given area of the chip), and hence costs<br />
per GB. 3-4-bit cells are becoming important to<br />
increase bit growth.<br />
Whilst there are limitations to performance and<br />
implementing three levels per cell (TLC) at smaller<br />
nodes – hence their current limitation mostly to lowend<br />
storage requirements – the price advantages of<br />
the technology are clear; 32Gb (the minimum<br />
density making commercial sense) TLC is currently<br />
c16-20% cheaper than an equivalent 2-bit cell<br />
NAND to produce on a GB basis.<br />
Toshiba/SanDisk is the leader in multi-cell<br />
technology; SanDisk holds key patents on X3 and<br />
X4 technologies. Toshiba/SanDisk started<br />
producing 3-bit per cell in 2008; SanDisk is<br />
migrating c50% of its NAND bit production to X3<br />
between 2010 and 2013, resulting in c25-35%<br />
annual cost reductions per GB. Samsung started<br />
mass production of 3-bit MLC NAND on 30nm at<br />
end-November 2009. Multi-bit NAND should<br />
increase production of higher densities (32Gbplus)<br />
for demanding content storage.<br />
51
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Technology develops broaden NAND<br />
scope<br />
Bigger modules for mobile devices<br />
We think larger densities are set to accelerate the<br />
take up of NAND in mobile devices. Samsung<br />
started mass production of a 3xnm-based 64GB<br />
embedded NAND (moviNAND) in December<br />
2009, which uses 16 x 32Gb dies and an on-board<br />
controller (total 17-die stack). Similarly,<br />
Toshiba’s 64GB embedded NAND flash module<br />
entered mass production in 1Q10. This 32nmbased<br />
unit has 16 x 32Gb NAND stacked units. In<br />
June 2010, Toshiba announced a 128GB module,<br />
which stacks 16x 8GB (64Gb 30nm) NAND units<br />
and a controller. Mass production is slated for<br />
4Q10. Apple currently uses 1x 32GB NAND<br />
module for its iPhone 4 and 2x 32GB NAND<br />
modules for the 64GB iPod Touch. Apple and<br />
others could use multiple 64GB NAND modules<br />
to deliver capacities approaching that of HDD for<br />
tablets and laptop PCs.<br />
Faster performance, lower power<br />
consumption. This is achieved through process<br />
migration. Samsung started producing 20nm<br />
based NAND (32Gb) in April 2010. SD card<br />
based on 20nm MLC is 30% faster than the 30nm<br />
equivalent, according to Samsung.<br />
Faster performance. Toggle DDR NAND is here.<br />
Samsung and Toshiba have tied up to standardise<br />
faster double data rate (DDR) that would help<br />
SSD performance. This would deliver 3-10x<br />
increase in data throughput over current interfaces.<br />
Toggle DDR NAND is currently available in<br />
16Gb (SLC) and 32Gb (MLC) densities from<br />
Samsung, which started mass production in<br />
December 2009. Samsung launched its first SSD<br />
using toggle-mode DDR NAND in June 2010.<br />
52
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Global NAND supply by vendor<br />
Units (16Gb equ. m) 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e<br />
Samsung 1,039 1,859 3,944 7,135 332 413 511 603 692 853 1,085 1,313<br />
y-o-y (%) 29% 79% 112% 81% 15% 25% 24% 18% 15% 23% 27% 21%<br />
% of total 36% 37% 37% 36% 36% 36% 38% 39% 37% 36% 36% 38%<br />
Toshiba/SanDisk 1,188 2,090 4,884 8,944 394 486 561 649 776 1,053 1,415 1,640<br />
y-o-y (%) 53% 76% 134% 83% 3% 23% 15% 16% 20% 36% 34% 16%<br />
% of total 41% 42% 46% 45% 42% 42% 42% 42% 42% 45% 47% 47%<br />
Hynix 129 251 668 1,319 43 53 67 87 127 156 181 204<br />
y-o-y (%) -30% 94% 167% 97% 0% 22% 27% 29% 47% 22% 16% 13%<br />
% of total 4% 5% 6% 7% 5% 5% 5% 6% 7% 7% 6% 6%<br />
Intel + Micron 539 752 1,120 2,376 157 191 192 211 241 274 290 316<br />
y-o-y (%) 91% 39% 49% 112% -9% 22% 1% 10% 14% 14% 6% 9%<br />
% of total 19% 15% 11% 12% 17% 17% 14% 14% 13% 12% 10% 9%<br />
Others 18 34 54 67 4 7 11 11 12 13 14 15<br />
y-o-y (%) 24% 84% 60% 25% -39% 75% 44% 6% 8% 7% 7% 5%<br />
% of total 1% 1% 1% 0% 0% 1% 1% 1% 1% 1% 0% 0%<br />
Total 2,914 4,984 10,670 19,842 931 1,151 1,342 1,561 1,849 2,348 2,985 3,488<br />
y-o-y (%) 41% 71% 114% 86% 4% 24% 17% 16% 18% 27% 27% 17%<br />
% of total 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100%<br />
Bit market share (%) 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e<br />
Samsung 36% 37% 37% 36% 36% 36% 38% 39% 37% 36% 36% 38%<br />
Toshiba/SanDisk 41% 42% 46% 45% 42% 42% 42% 42% 42% 45% 47% 47%<br />
Hynix 4% 5% 6% 7% 5% 5% 5% 6% 7% 7% 6% 6%<br />
IMFT 19% 15% 11% 12% 17% 17% 14% 14% 13% 12% 10% 9%<br />
Others 1% 1% 1% 0% 0% 1% 1% 1% 1% 1% 0% 0%<br />
Total 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100%<br />
Bit growth (y-o-y) 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e<br />
Samsung 29% 79% 112% 81% 15% 25% 24% 18% 15% 23% 27% 21%<br />
Toshiba 53% 76% 134% 83% 3% 23% 15% 16% 20% 36% 34% 16%<br />
Hynix -30% 94% 167% 97% 0% 22% 27% 29% 47% 22% 16% 13%<br />
IMFT 91% 39% 49% 112% -9% 22% 1% 10% 14% 14% 6% 9%<br />
Others 24% 84% 60% 25% -39% 75% 44% 6% 8% 7% 7% 5%<br />
Total 41% 71% 114% 86% 4% 24% 17% 16% 18% 27% 27% 17%<br />
Source: IDC, HSBC estimates<br />
53
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Global NAND fab capacity<br />
Kwpy, kwpm 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e<br />
Samsung 4,159 4,327 4,860 5,820 362 360 360 360 370 390 420 440<br />
y-o-y (%) 7% 4% 12% 20% -3% -1% 0% 0% 3% 5% 8% 5%<br />
Hynix 907 1,267 1,687 2,187 96 106 106 116 126 136 146 156<br />
y-o-y (%) -42% 40% 33% 30% 26% 10% 0% 9% 9% 8% 7% 7%<br />
Toshiba / SanDisk 2,787 3,360 4,350 5,160 260 270 280 310 320 320 390 420<br />
y-o-y (%) -3% 21% 29% 19% 0% 4% 4% 11% 3% 0% 22% 8%<br />
IMFT 1,260 1,440 1,542 1,962 105 125 125 125 125 125 132 132<br />
y-o-y (%) 16% 14% 7% 27% 0% 19% 0% 0% 0% 0% 6% 0%<br />
Renesas / PSC 11 11 11 11 1 1 1 1 1 1 1 1<br />
y-o-y (%) -81% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0%<br />
Numonyx 71 72 80 80 5 6 6 7 7 7 7 7<br />
y-o-y (%) 2% 2% 11% 0% -25% 8% 8% 7% 0% 0% 0% 0%<br />
8” Total 2,394 1,296 1,164 960 144 95 96 97 97 97 97 97<br />
y-o-y (%) -60% -46% -10% -18% -3% -34% 1% 1% 0% 0% 0% 0%<br />
12” Total 8,130 9,900 12,012 14,702 765 825 835 875 905 935 1,052 1,112<br />
y-o-y (%) 16% 22% 21% 22% 1% 8% 1% 5% 3% 3% 13% 6%<br />
12” total (kwpm) 678 825 1,001 1,225 64 69 70 73 75 78 88 93<br />
12” (8” Eq.) 18,293 22,275 27,027 33,080 1,721 1,856 1,879 1,969 2,036 2,104 2,367 2,502<br />
y-o-y (%) 16% 22% 21% 22% 1% 8% 1% 5% 3% 3% 13% 6%<br />
Total (12” equ. kwpy) 9,194 10,476 12,529 15,129 829 867 878 918 948 978 1,095 1,155<br />
y-o-y (%) -5% 14% 20% 21% 1% 5% 1% 5% 3% 3% 12% 5%<br />
Total (12” eq. kwpm) 766 873 1,044 1,261 829 867 878 918 948 978 1,095 1,155<br />
12” proportion (%) 88.4% 94.5% 95.9% 97.2% 92.3% 95.1% 95.1% 95.3% 95.5% 95.6% 96.1% 96.3%<br />
Source: IDC, HSBC estimates<br />
54
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Valuations, ratings<br />
Maintain TP and OW rating on Samsung. Lower TP on Hynix, but<br />
maintain OW(V)<br />
Lower TP on Elpida, Nanya Tech, Winbond and Elpida. Maintain<br />
OW(V) for Elpida and Winbond, and N(V) for Inotera and Nanya<br />
We expect valuation divergence as product mix and cost structure<br />
drive profitability for leaders<br />
In the past six months, memory stocks have<br />
underperformed their respective markets on<br />
concerns that the memory cycle peaked out in 2Q.<br />
Among our coverage, with the exception of<br />
Winbond, pure memory plays have lagged their<br />
indices by 20-52%. Samsung, being a diversified<br />
tech conglomerate, underperformed KOSPI by<br />
11%, with semiconductor revenue representing<br />
25% of its revenue. The market penalised Nanya<br />
and Inotera (underperformed by c40%), on delays<br />
of their 5xnm migration.<br />
Concerns over memory price correction could<br />
linger (until data points emerge indicating a<br />
deceleration), we expect price performance<br />
divergence between vendors to be more<br />
pronounced going forward; we expect the market to<br />
favour vendors that are capable of accelerating<br />
process migration and improving their product mix.<br />
Our top sector pick is Samsung Electronics based<br />
on its unparalleled overall memory positioning and<br />
compelling valuations, in addition to its diversified<br />
technology portfolio. We also like Hynix and<br />
Elpida for their strong DRAM positioning (inhouse<br />
technology, process migration) and their<br />
leading positions in mobile DRAM. Both Elpida<br />
Stock performances relative to respective markets (%)<br />
10<br />
0<br />
-10<br />
-20<br />
-30<br />
-40<br />
-50<br />
-60<br />
SEC<br />
Source: Bloomberg<br />
Hynix<br />
Elpida<br />
Nanya<br />
Inotera<br />
6 months 3 months 1 month<br />
Winbond<br />
and Winbond are trading below 1x P/B. Trading at<br />
0.5x FY11e P/B, Elpida’s valuations look very<br />
attractive for FY2011e. We think most of the<br />
negative news have been priced in and the stock is<br />
oversold. We also like Winbond’s non-PC DRAM<br />
portfolio and believe it is a defensive portfolio<br />
amid commodity DRAM price correction cycle<br />
given its attractive valuation.<br />
We believe the Taiwanese performance should<br />
still be limited until signs of improvement from<br />
55
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Valuation summary<br />
Valuation<br />
method<br />
Target<br />
multiple<br />
Historical PB<br />
multiple<br />
Samsung Electronics avg of SOTP, 2.2x PB<br />
2011e PB, DCF<br />
Hynix Semiconductor 2011e PB 1.8x 0.8x-3.3x<br />
Elpida <strong>Memory</strong> 2011e PB 0.8x 0.3x-2.4x<br />
Nanya Technology 2011e PB 1.9x 1.6x-4.9x<br />
Inotera Memories 2011e PB 1.5x 0.5x-2.5x<br />
Winbond 2011e PB 1.0x 0.2x-1.2x<br />
EBIT margin comparison<br />
40%<br />
20%<br />
0%<br />
-20%<br />
-40%<br />
-60%<br />
Source: HSBC<br />
SEC<br />
Hynix<br />
Elpida<br />
Nanya<br />
Inotera<br />
Winbond<br />
the node migration and cycle time are becoming<br />
apparent. We remain Neutral (V) on the two stocks.<br />
2009 2010e 2011e<br />
Source: Company data, HSBC estimates, Elpida forecast are calendarised<br />
Stock coverage and ratings summary<br />
Company Ticker Rating New TP<br />
(LC)<br />
CP<br />
(LC)<br />
Potential<br />
return (%)<br />
Previous<br />
TP<br />
Change<br />
(%)<br />
Samsung Electronics 005930 KS OW 1,139,000 793,000 44% 1,139,000 0%<br />
Hynix Semiconductor 000660 KS OW(V) 32,000 23,400 37% 36,000 -11%<br />
Elpida <strong>Memory</strong> 6665 JP OW(V) 1,500 928 62% 3000 -50%<br />
Nanya Technology 2408 TT N(V) 20 19.3 4% 25 -20%<br />
Inotera Memories 3474 TT N(V) 17 16.2 5% 20 -15%<br />
Winbond 2344 TT OW(V) 10.8 8.03 34% 12.8 -16%<br />
Source: HSBC<br />
56
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
10 ROE Vs 10 PB 11 ROE Vs 11 PB<br />
10 ROE<br />
50%<br />
40%<br />
30%<br />
20%<br />
10%<br />
0%<br />
-10%<br />
-20%<br />
Micron<br />
Elpida<br />
Winbond<br />
Hy nix<br />
Pow erchip<br />
SanDisk<br />
Samsung<br />
Inotera<br />
Toshiba<br />
Nanya<br />
0.0 0.5 1.0 1.5 2.0 2.5<br />
10 PB<br />
11 ROE<br />
25%<br />
20%<br />
15%<br />
10%<br />
5%<br />
0%<br />
Hy nix<br />
Micron<br />
Samsung<br />
SanDisk<br />
Elpida<br />
Nanya<br />
Winbond Inotera<br />
Pow erchip<br />
Toshiba<br />
0.0 0.5 1.0 1.5 2.0 2.5<br />
11 PB<br />
Source: HSBC<br />
Source: HSBC<br />
Peer group valuation, profitability comparison<br />
Company Ticker Price (LC) Mkt cap<br />
(USDm)<br />
HSBC<br />
rating<br />
_____________ PE _____________________ EV/EBITDA _________ _____________ PB_____________<br />
2009 2010e 2011e 2009 2010e 2011e 2009 2010e 2011e<br />
Samsung 005930 KS 793,000 104,042 OW 12.2 8.2 8.3 7.7 3.7 3.6 1.8 1.5 1.3<br />
Hynix 000660 KS 23,400 12,303 OW(V) (37.7) 4.7 6.6 10.7 3.3 3.4 2.5 1.6 1.3<br />
Elpida 6665 JP 928 2,269 OW(V) (2.8) 3.4 10.5 85.3 2.6 4.3 0.8 0.8 0.5<br />
Toshiba 6502 JP 423 21,582 NR (14.0) 32.1 13.0 8.5 5.4 15.2 3.1 2.2 2.0<br />
Powerchip 5346 TT 7.89 1,433 NR (3.1) 3.3 7.2 8.6 2.1 2.5 2.4 1.1 1.0<br />
Nanya 2408 TT 19.3 2,145 N(V) (2.2) (13.2) 23.3 (97.6) 11.8 6.5 1.2 2.0 1.8<br />
Inotera 3474 TT 16.2 2,422 N(V) (5.1) (13.6) 23.6 5.5 4.8 3.0 1.2 1.5 1.4<br />
ProMOS 5387 TT 1.47 346 NR (0.5) (0.6) (0.9) (95.5) 11.5 5.3 0.5 4.1 147.0<br />
Winbond 2344 TT 8.03 951 OW(V) (3.4) 8.3 14.6 11.9 2.9 2.7 0.9 0.8 0.7<br />
Micron MU US 6.93 6,890 NR (7.0) 4.3 12.6 3.9 1.7 2.6 1.3 0.9 0.8<br />
SanDisk SNDK US 37.64 8,773 NR 21.0 9.0 9.6 10.5 4.4 4.2 2.2 1.8 1.5<br />
MC WA 3.7 10.2 9.5 7.5 4.0 5.1 2.0 1.6 1.7<br />
Div yield (%) _______ EBIT margin _______ ______ EBITDA margin________ ___________ ROE ____________<br />
Company Ticker 2010e 2009 2010e 2011e 2009 2010e 2011e 2009 2010e 2011e<br />
Samsung 005930 KS 1.5% 7% 12% 11% 16% 20% 19% 15% 20% 17%<br />
Hynix 000660 KS 0.0% -2% 29% 20% 22% 42% 36% -7% 44% 22%<br />
Elpida 6665 JP 0.0% -4% 15% 17% 2% 31% 36% -69% 19% 9%<br />
Toshiba 6502 JP 0.9% 0% 3% 4% 5% 7% 8% -47% 8% 4%<br />
Powerchip 5346 TT 0.0% -59% 17% 1% 47% 56% 39% -57% 40% 3%<br />
Nanya 2408 TT 0.0% -37% -6% 8% -3% 21% 31% -63% -14% 8%<br />
Inotera 3474 TT 0.0% -27% -8% 8% 56% 61% 63% -23% -11% 6%<br />
ProMOS 5387 TT 0.0% -141% -58% -57% -5% 27% 55% -82% -90% -399%<br />
Winbond 2344 TT 0.0% -31% 11% 7% 22% 45% 41% -24% 10% 5%<br />
Micron MU US 0.0% -1% 19% 18% 33% 38% 33% -34% 25% 15%<br />
SanDisk SNDK US 0.0% 15% 30% 26% 21% 34% 30% 11% 24% 18%<br />
MC WA 1.5% 3% 13% 11% 16% 23% 22% -1% 20% 14%<br />
Source: HSBC forecast for SEC, Hynix, Elpida, Nanya, Inotera and Winbond. Bloomberg consensus otherwise. Priced as at 6 Oct 2010. Numbers are calendarised.<br />
57
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Inotera’s relative performance (TWD)<br />
Elpida’s relative performance (JPY)<br />
10,000<br />
50<br />
2,000<br />
8,000<br />
8,000<br />
6,000<br />
4,000<br />
2,000<br />
40<br />
30<br />
20<br />
10<br />
1,500<br />
1,000<br />
6,000<br />
4,000<br />
2,000<br />
-<br />
-<br />
500<br />
-<br />
Jan-06 Jan-07 Jan-08 Jan-09 Jan-10<br />
Jan-06 Jan-07<br />
Jan-08 Jan-09 Jan-10<br />
Taiw an Taiex Index<br />
Inotera (RHS)<br />
Topix Index<br />
Elpida <strong>Memory</strong> (RHS)<br />
Source: Bloomberg<br />
Source: Bloomberg<br />
Samsung Electronics’ relative performance (KRW)<br />
Hynix’s relative performance (KRW)<br />
2,500<br />
1,000,000<br />
3,000<br />
50,000<br />
2,000<br />
800,000<br />
2,500<br />
40,000<br />
1,500<br />
1,000<br />
600,000<br />
400,000<br />
2,000<br />
1,500<br />
1,000<br />
30,000<br />
20,000<br />
500<br />
200,000<br />
500<br />
10,000<br />
-<br />
-<br />
-<br />
-<br />
Jan-06 Jan-07 Jan-08 Jan-09 Jan-10<br />
Jan-06 Jan-07 Jan-08 Jan-09 Jan-10<br />
KOSPI Index<br />
Samsung Electronics (RHS)<br />
KOSPI Index<br />
Hy nix (RHS)<br />
Source: Bloomberg<br />
Source: Bloomberg<br />
Winbond Electronics’ relative performance (TWD)<br />
Nanya Technology’s relative performance (TWD)<br />
10,000<br />
15<br />
10,000<br />
100<br />
8,000<br />
12<br />
8,000<br />
80<br />
6,000<br />
9<br />
6,000<br />
60<br />
4,000<br />
6<br />
4,000<br />
40<br />
2,000<br />
3<br />
2,000<br />
20<br />
-<br />
-<br />
-<br />
-<br />
Jan-06 Jan-07 Jan-08 Jan-09 Jan-10<br />
Jan-06 Jan-07 Jan-08 Jan-09 Jan-10<br />
Taiw an Taiex Index<br />
Winbond Electronics (RHS)<br />
Taiw an Taiex Index<br />
Nany a Tech (RHS)<br />
‘<br />
Source: Bloomberg<br />
Source: Bloomberg<br />
58
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Company profiles<br />
Elpida <strong>Memory</strong> (6665)<br />
Hynix Semiconductor (000660)<br />
Inotera Memories (3474)<br />
Nanya Technology (2408)<br />
Samsung Electronics (005930)<br />
Winbond (2344)<br />
59
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Elpida <strong>Memory</strong> (6665)<br />
Global #3 DRAM maker with high specialist revenue component<br />
of c43% for FY11e, including mobile DRAM<br />
We lower FY10-11 operating profit forecasts by 6% and 27%<br />
respectively on lower ASP assumptions<br />
Maintain OW(V) rating, lower TP to JPY1,500 based on FY11e<br />
target PB of 0.8x. Stock is oversold, in our view<br />
Investment summary<br />
Sensible growth strategy<br />
A pure memory play with mostly DRAM<br />
exposure, Elpida <strong>Memory</strong> was the third largest<br />
DRAM manufacturer on revenues according to<br />
IDC, with c17% share (Samsung: c34%, Hynix<br />
c22%).<br />
Having pursed a strategy that capitalises on its<br />
shrink technology to reduce costs to a 5xnmequivalent<br />
level, Elpida is now migrating to 4xnm<br />
process to remain competitive versus the leaders.<br />
For commodity PC DRAM, it is fully deploying<br />
capacity at Rexchip, its Taiwanese subsidiary to<br />
remain cost competitive. Elpida’s Japan fab<br />
focuses on higher value-add products such as<br />
mobile DRAM and specialist DRAM. To further<br />
enhance its strong positioning in mobile DRAM,<br />
it has co-developed with Spansion NAND that<br />
would enable Elpida to capture the full margins<br />
available by offering MCP (multichip packages).<br />
Though we expect (as we do with all the memory<br />
companies) next year’s earnings to decline y-o-y<br />
from a high base, Elpida has upside risk from its<br />
mobile and specialist DRAM exposure, which we<br />
expect to account for c43% of FY11e revenue<br />
(year-end March 2012). The stock is trading on<br />
FY11e PB of 0.5x. We think it is oversold at<br />
current levels.<br />
Valuations, rating<br />
We value Elpida using PB to derive a 12-month<br />
target price of JPY1,500 by applying a target<br />
FY11 PB multiple of 0.8x. Previously, we valued<br />
Elpida at JPY3,000 using the same methodology,<br />
but using a FY10e PB multiple of 2x. The stock<br />
has historically traded at a PB of 0.3-2.4x.<br />
Though we forecast Elpida’s operating<br />
performance to decline from this year’s peak, we<br />
believe our new target PB sufficiently reflects<br />
downside risks; based on FY11-12e average 8.3%<br />
ROE and 7.5% COE, the stock should in fact, be<br />
valued at a PB ratio of 1.1x. But we believe<br />
negative sentiment on DRAM demand, Elpida’s<br />
weaker market positioning versus leaders,<br />
execution risk of its process migration and JPY<br />
strength limit shorter term valuation upside.<br />
Under our research model, for Japan stocks with a<br />
volatility indicator, the Neutral band is 10ppt<br />
above and below the hurdle rate of 7.5% (risk-free<br />
rate plus risk premium). This translates into a<br />
Neutral band of -2.5% to +17.5% around the<br />
60
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
current share price. Our 12-month target price of<br />
JPY1,500 suggests a potential return of 61.6%,<br />
which is above the Neutral band; thus, we<br />
maintain an Overweight (V) rating on the stock.<br />
Volatile ratings are defined as stocks having<br />
historical volatility (defined as the past month’s<br />
average of the daily 365-day moving average<br />
volatilities) of more than 40%.<br />
HSBC versus consensus – Elpida<br />
JPY bn FY10e FY11e 2QFY10e 3QFY10e<br />
Forward PB chart<br />
7,000<br />
6,000<br />
Price (JPY)<br />
5,000<br />
4,000<br />
3,000<br />
2,000<br />
1,000<br />
-<br />
Jan-06 Jan-07 Jan-08 Jan-09 Jan-10<br />
2.0x<br />
1.5x<br />
1.0x<br />
0.5x<br />
Revenue 647 619 163 157<br />
EBITDA 236 187 65 46<br />
EBIT 109 54 33 16<br />
Net income 72 30 24 7<br />
EPS (JPY) 354 147 118 32<br />
EBITDA margin 36.5% 30.1% 39.9% 29.6%<br />
EBIT margin 16.8% 8.7% 20.3% 10.0%<br />
Net margin 11.1% 4.8% 14.7% 4.2%<br />
HSBC vs consensus<br />
Revenue -1.5% -3.8% -1.1% -1.4%<br />
EBITDA -1.6% -3.3% -4.7% -18.0%<br />
EBIT -3.1% -12.2% -6.2% -0.1%<br />
Net income -4.4% -19.1% -1.7% -33.7%<br />
EPS (JPY) -3.2% -16.0% 8.6% -38.1%<br />
Consensus<br />
Revenue 657 644 165 159<br />
EBITDA 240 193 68 57<br />
EBIT 112 61 35 16<br />
Net income 75 37 24 10<br />
EPS (JPY) 366 175 109 52<br />
EBITDA margin 36.5% 30.0% 41.4% 35.6%<br />
EBIT margin 17.1% 9.5% 21.4% 9.9%<br />
Net margin 11.4% 5.7% 14.8% 6.2%<br />
Source: I/B/E/S, HSBC estimates<br />
Source: I/B/E/S, HSBC estimates<br />
Catalysts and risks<br />
<strong>Memory</strong>-specific catalysts include:<br />
Better-than-expected results and strong<br />
guidance from PC OEMs, such as recent Dell<br />
and HP results.<br />
Launch and strong deliveries of key mobile<br />
devices – driving Elpida’s exposure to mobile<br />
DRAM.<br />
Evidence of strong unit shipments and rising<br />
memory content of smartphones.<br />
Launch of competitive new products,<br />
especially next-generation <strong>memories</strong>.<br />
<strong>Memory</strong>-specific risks include:<br />
Faster-than-expected ASP erosion due to<br />
greater capacity coming online and/or slowerthan-expected<br />
PC unit shipments.<br />
Greater ability of competitors to accelerate<br />
process migration, wafer capacity expansion<br />
through improving financials.<br />
IP disputes. Although these tend to be<br />
ongoing, any potentially major dispute could<br />
be a risk.<br />
61
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Other company-specific catalysts and risks include:<br />
Rebounding global competition following<br />
restructuring and recapitalisation efforts.<br />
Successful proposed IPO of subsidiary Rexchip.<br />
Sharp reduction in debt through internal cash<br />
generation.<br />
Strengthening JPY versus USD, KRW, TWD,<br />
although producing offshore in Taiwan<br />
should offset some of that.<br />
Delayed global economic recovery.<br />
Fundamentals<br />
Operating performance outlook<br />
We forecast Elpida’s FY10 (year-end March<br />
2011) revenue to rise c9% y-o-y, but for FY11,<br />
we forecast revenue to decline c5%. We forecast<br />
Elpida’s FY10e operating profit to rise c310% y-<br />
o-y to JPY110bn but forecast FY11e to see a 51%<br />
decline.<br />
Revenue breakdown (CY)<br />
100%<br />
80%<br />
60%<br />
40%<br />
20%<br />
0%<br />
1Q09 2Q09 3Q09 4Q09 1Q10 2Q10<br />
Computing DRAM Premier DRAM<br />
Source: Company, HSBC estimates<br />
Improving balance sheet, shorter inventory<br />
days.<br />
Highly profitable subsidiary. Rexchip<br />
Electronics has the highest profitability<br />
amongst Taiwanese DRAM makers.<br />
Revenue, OP (JPYbn) and OPM (%), CY<br />
800<br />
30%<br />
600<br />
20%<br />
10%<br />
400<br />
0%<br />
200<br />
-10%<br />
-20%<br />
-<br />
-30%<br />
(200)<br />
2007 2008 2009 2010f 2011f<br />
-40%<br />
Rev enue (JPY bn) OP OPM<br />
Source: Company, HSBC estimates<br />
Revenue, margin outlook<br />
For CY10e we estimate revenue to rise 76% in<br />
JPY terms, on 37% bit growth and blended ASP<br />
increase of 31%. We forecast CY11e revenue<br />
decline of 5% under our current process migration<br />
assumptions; bit growth is stable at c38% y-o-y<br />
and we forecast a 32% fall in ASP on greater<br />
industry-wide supply (supply-demand to be<br />
relatively balanced in 2011).<br />
Historically, Elpida has had a much higher than<br />
industry average proportion of premier (specialist)<br />
DRAM – about a half of revenue – compared to<br />
computing (commodity) DRAM. This proportion<br />
fell dramatically last year as demand premier<br />
DRAM dried up.<br />
Elpida has seen continued recovery in operating<br />
profitability in the past four quarters on higher<br />
revenue mostly on higher computing DRAM ASP<br />
and bit shipments.<br />
Lower costs due to process migration, greater<br />
productivity.<br />
62
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Elpida forecast assumptions<br />
DRAM process migration<br />
y-o-y % changes 2008 2009 2010e 2011e<br />
Bit growth 99% 57% 37% 38%<br />
ASP -53% -31% 31% -32%<br />
FX -11% -12% -2% 0%<br />
USD sales check -8% 10% 80% -6%<br />
USD sales -18% -2% 76% -5%<br />
JPY sales check 99% 57% 37% 38%<br />
JPY sales -53% -31% 31% -32%<br />
OPM -11% -12% -2% 0%<br />
Source: Company, HSBC estimates<br />
100%<br />
80%<br />
60%<br />
40%<br />
20%<br />
0%<br />
2007 2008 2009 2010f 2011f<br />
For FY10, Elpida expects its bit growth will reach<br />
45%; we forecast 34% and 36% respectively for<br />
FY10-11. Our forecasts are lower due to time<br />
required to achieve satisfactory yields in mass<br />
production both at its own fabs and outsourced<br />
production.<br />
CY bit growth, ASP assumptions (%)<br />
150%<br />
100%<br />
50%<br />
0%<br />
-50%<br />
-100%<br />
2008 2009 2010f 2011f 2012f<br />
Bit growth ASP<br />
Source: Company, HSBC estimates<br />
Technology – process migration<br />
Uniquely, Elpida took a dual-pronged R&D<br />
strategy to minimise capex starting in 2008.<br />
Elpida developed ‘shrunk’ versions of its 6xnm<br />
based chips. These were manufactured using ArF<br />
dry scanner equipment, i.e. without having to<br />
invest in expensive immersion lithography<br />
equipment used by competitors like Samsung.<br />
8xnm 6xnm 6x S, XS 4xnm 3xnm 2xnm<br />
Source: Company, HSBC estimates<br />
Elpida originally developed a 65nm S shrink<br />
version in 2008. Based on that version, it then<br />
completed development of a 65nm XS (extrashrink)<br />
version in December 2009. Elpida says that<br />
the XS chip delivers 25% more dies per wafer<br />
compared to the S version and is as costcompetitive<br />
as 5xnm node DRAM. The XS version<br />
1GB DDR3 entered mass production in 1Q10.<br />
This year, though, Elpida (and its subsidiary<br />
Rexchip) is migrating to 4xnm. By going from<br />
65nm-S to 4xnm, the firm estimates dies per<br />
wafer increase c115% (going from 65nm-S to<br />
50nm would be 45% more dies) and increase<br />
shipment volumes by c30%. Elpida’s timeline on<br />
migration to 4xnm and 3xnm has been forced by<br />
competitors’ migration plans:<br />
Hiroshima – mobile and speciality DRAM focus<br />
Elpida started pilot production of 2GB DDR3 chips<br />
on the 4xnm process (6F2) at its 130kwpm<br />
Hiroshima fab, and raised yields sufficiently for<br />
mass production in 2010. By September, it had<br />
migrated c30% of wafer input to 4x-5xnm and<br />
plans to migrate half its capacity to 4xnm by end-<br />
2010. We expect further migration to depend on<br />
product mix, as consumer DRAM uses older<br />
processes. Mobile DRAM is currently produced on<br />
40-50nm processes. Elpida estimates that c60-70%<br />
of Hiroshima capacity is required to produce<br />
mDRAM by end 2011, with the remaining<br />
63
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
comprising. specialty DRAM; 3xnm migration at<br />
Hiroshima is likely to be limited in the shorter term.<br />
Rexchip – the commodity DRAM arm<br />
Over time, Rexchip will manufacture all of<br />
Elpida’s commodity DRAM. Rexchip is to convert<br />
its entire 80kwpm 300mm capacity to Elpida’s<br />
4xnm process by end-2010. As at September, 10%<br />
of wafer input was based on 4xnm, with the<br />
remaining on 65nm XS. Positively, Rexchip should<br />
have received all the tools required for full 4xnm<br />
migration by the end of this November.<br />
Elpida completed development of 3xnm DDR3<br />
(30% smaller, 40% more efficiency compared to<br />
4xnm); engineering samples should be out by<br />
December. Mass production is to start by end-2010,<br />
with production ramp next spring. Production uses<br />
existing equipment with double patterning. Once<br />
satisfactory yields are achieved, the technology will<br />
be transferred to Rexchip, which will completely<br />
migrate to 3xnm in the longer term.<br />
Wafer capacity (kwpy)<br />
4,000<br />
3,000<br />
2,000<br />
1,000<br />
-<br />
2007 2008 2009 2010f 2011f 2012f<br />
Wafer capacity (kwpy)<br />
Source: Company, HSBC estimates<br />
Production capacity, capex<br />
Elpida’s 300mm wafer capacity comprises its<br />
Hiroshima fab plus its 64%-owned Taiwanese<br />
subsidiary Rexchip Electronics. In addition to<br />
Rexchip, Elpida also outsources production to<br />
Powerchip and will start to outsource to ProMOS<br />
(PC DRAM) and Winbond (graphics DRAM).<br />
Elpida’s relationships with Taiwanese makers<br />
Elpida<br />
64%<br />
Rexchip<br />
30.84%<br />
Powerchip<br />
ProMOS<br />
Winbond<br />
Source: Company – arrows indicate outsourcing flow<br />
The relationship with Powerchip is waning as<br />
Rexchip expands. Powerchip has c130kwpm<br />
total capacity; 80kwpm is commodity DRAM,<br />
of which 30kwpm is legacy production. We<br />
expect Elpida to receive c40kwpm equivalent.<br />
We believe Powerchip is also migrating to<br />
4xnm, with one immersion tool expected by<br />
the end of this year followed by another two<br />
next year.<br />
ProMOS is to start DRAM mass production<br />
of 65nm-XS DRAM for Elpida after July<br />
2010, with 35kwpm capacity. But we believe<br />
ProMOS is currently having some yield issues,<br />
so this timeline could be delayed. We believe<br />
for Elpida, ProMOS will act purely as a<br />
foundry.<br />
Winbond started to produce graphics DRAM<br />
in late-1Q10. We believe the number of<br />
wafers it supplies is low, at c5kwpm.<br />
For FY10e (March-2011 end), Elpida expects to<br />
spend JPY115bn capex, of which 30% is<br />
earmarked for Rexchip, which will require<br />
JPY30-35bn to fully migrate its capacity to 4xnm.<br />
60% is earmarked for the Hiroshima fab to<br />
migrated half of its capacity to 4xnm. The<br />
remaining 10% is for subsidiaries to expand wafer<br />
testing and back-end fab facilities.<br />
64
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Forecast revisions<br />
Elpida – forecast revisions<br />
Mar (JPY bn)<br />
___ Revised __ ___ Previous _____ Changes __<br />
2010e 2011e 2010e 2011e 2010e 2011e<br />
Revenue 647.2 619.3 709.9 754.5 -9% -18%<br />
EBITDA 236.0 186.6 290.4 255.1 -19% -27%<br />
EBIT 108.5 53.8 115.4 74.2 -6% -27%<br />
Net income 71.9 29.7 98.3 58.9 -27% -49%<br />
EPS (JPY) 354 146 500 300 -29% -51%<br />
Source: HSBC forecasts and estimates<br />
65
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Financials & valuation: Elpida<br />
Overweight (V)<br />
Financial statements<br />
Year to 03/2010a 03/2011e 03/2012e 03/2013e<br />
Profit & loss summary (JPYb)<br />
Revenue 467 647 619 641<br />
EBITDA 149 236 187 194<br />
Depreciation & amortisation -122 -127 -133 -135<br />
Operating profit/EBIT 27 109 54 59<br />
Net interest -9 -9 -9 -8<br />
PBT 9 85 35 41<br />
HSBC PBT 9 85 35 41<br />
Taxation -2 -13 -5 -8<br />
Net profit 7 72 30 33<br />
HSBC net profit 7 72 30 33<br />
Cash flow summary (JPYb)<br />
Cash flow from operations 56 172 186 179<br />
Capex -60 -60 -60 -64<br />
Cash flow from investment -152 -172 -60 -64<br />
Dividends 0 0 0 0<br />
Change in net debt -52 89 -119 -109<br />
FCF equity -6 100 121 107<br />
Balance sheet summary (JPYb)<br />
Intangible fixed assets 1 1 1 1<br />
Tangible fixed assets 615 539 458 376<br />
Current assets 326 313 420 538<br />
Cash & others 112 23 142 251<br />
Total assets 947 858 884 920<br />
Operating liabilities 324 232 229 232<br />
Gross debt 277 277 277 277<br />
Net debt 164 253 134 25<br />
Shareholders funds 277 349 379 412<br />
Invested capital 506 598 508 433<br />
Ratio, growth and per share analysis<br />
Year to 03/2010a 03/2011e 03/2012e 03/2013e<br />
Y-o-y % change<br />
Valuation data<br />
Year to 03/2010a 03/2011e 03/2012e 03/2013e<br />
EV/sales 0.9 0.7 0.5 0.3<br />
EV/EBITDA 2.9 1.9 1.7 1.1<br />
EV/IC 0.9 0.7 0.6 0.5<br />
PE* 21.9 2.6 6.3 5.6<br />
P/Book value 0.6 0.5 0.5 0.5<br />
FCF yield (%) -2.1 54.2 65.9 58.5<br />
Dividend yield (%) 0.0 0.0 0.0 0.0<br />
Note: * = Based on HSBC EPS (fully diluted)<br />
Issuer information<br />
Share price (JPY) 928 Target price (JPY) 1500 Potent’l return (%) 61.6<br />
Reuters (Equity) 6665.T Bloomberg (Equity) 6665 JP<br />
Market cap (USDm) 2,315 Market cap (JPYb) 192<br />
Free float (%) 89 Enterprise value (JPYb) 441<br />
Country Japan Sector SEMICONDUCTORS<br />
Analyst Nam Park Contact 852 2996 6591<br />
Price relative<br />
4500<br />
4000<br />
3500<br />
3000<br />
2500<br />
2000<br />
1500<br />
1000<br />
500<br />
0<br />
Source: HSBC<br />
2008 2009 2010 2011<br />
Elpida Rel to TOPIX INDEX<br />
Note: price at close of 06 Oct 2010<br />
4500<br />
4000<br />
3500<br />
3000<br />
2500<br />
2000<br />
1500<br />
1000<br />
500<br />
0<br />
Revenue 41.1 38.6 -4.3 3.5<br />
EBITDA 58.7 -20.9 3.7<br />
Operating profit 304.3 -50.4 9.5<br />
PBT 848.3 -58.6 17.9<br />
HSBC EPS 736.4 -58.6 12.3<br />
Ratios (%)<br />
Revenue/IC (x) 0.9 1.2 1.1 1.4<br />
ROIC 4.4 16.7 8.3 10.1<br />
ROE 3.2 23.0 8.2 8.5<br />
ROA 1.6 8.9 4.4 4.6<br />
EBITDA margin 31.8 36.5 30.1 30.2<br />
Operating profit margin 5.7 16.8 8.7 9.2<br />
EBITDA/net interest (x) 15.8 26.4 21.2 25.3<br />
Net debt/equity 45.7 72.7 35.5 6.2<br />
Net debt/EBITDA (x) 1.1 1.1 0.7 0.1<br />
CF from operations/net debt 34.1 68.0 138.7 705.7<br />
Per share data (JPY)<br />
EPS Rep (fully diluted) 42.33 354.02 146.63 164.70<br />
HSBC EPS (fully diluted) 42.33 354.02 146.63 164.70<br />
DPS 0.00 0.00 0.00 0.00<br />
Book value 1628.54 1717.97 1864.60 2029.29<br />
66
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Hynix Semiconductor (000660)<br />
Product portfolio leverages into higher-margin, growth segments<br />
such as servers and mobile; process migration accelerates<br />
Historic bane of debt is shrinking rapidly. Shareholder stake sale<br />
and strong operating metrics are stock catalysts<br />
Remain OW(V), TP lowered to KRW32,000 (-11.1% from previous)<br />
on lower 2011e PB target of 1.8x<br />
Investment summary<br />
Strong, diversified product portfolio<br />
A pure memory play with both DRAM and<br />
NAND exposure, Hynix is one of our preferred<br />
plays in the sector. Hynix has:<br />
Highly competitive technologies and leading<br />
market share leadership in DRAM and a<br />
returning presence in NAND, where it has<br />
caught up with market leaders in process<br />
migration.<br />
Debt is no longer an issue, continued ability<br />
to invest. Hynix is unshackling itself from its<br />
historic bane, high debt. We forecast parent<br />
level operating cash flow of cKRW4.1trn this<br />
year and close to KRW4trn per year for 2011-<br />
12e (2009: KRW735bn). We forecast 2010e<br />
parent net DER at a comfortable c32.5% level,<br />
leaving plenty of scope for Hynix to support<br />
ongoing capex and R&D if required. Though<br />
Hynix could be net cash next year, it is likely<br />
to maintain some debt to optimise ROE.<br />
Determination to expand leadership with<br />
more advanced and cost-effective<br />
technologies, whilst remaining cautious in<br />
terms of capex and competitive positioning.<br />
Valuations, rating<br />
To value Hynix we use a target 2011e PB of 1.8x.<br />
This PB target has been rolled forward from<br />
previous 2010e, and lowered from 2.5x to reflect<br />
memory market uncertainties and lower ROE<br />
expectations. Macro demand concerns and lower<br />
2011e forecast profits are likely to constrain<br />
valuations. Historically, Hynix has traded in a 0.8-<br />
3.3x PB range. Our new target of KRW32,000 is<br />
c11.1% below our previous target of KRW36,000.<br />
Under our research model, for stocks without a<br />
volatility indicator, the Neutral band is five<br />
percentage points above and below our hurdle rate<br />
for Korean stocks of 10.5%, or 5.5-15.5% around<br />
the current share price. Our new 12-month target<br />
price of KRW32,000 suggests a potential return of<br />
36.8% which is above the Neutral band; thus, we<br />
reiterate our Overweight (V) rating on the stock.<br />
Volatile ratings are defined as stocks having<br />
historical volatility (defined as the past month’s<br />
average of the daily 365-day moving average<br />
volatilities) of more than 40%.<br />
67
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
HSBC versus consensus<br />
HSBC 2010e 2011e 3Q10e 4Q10e<br />
Revenue 12,045 12,557 2,968 3,040<br />
EBITDA 5,033 4,510 1,192 1,225<br />
EBIT 3,488 2,546 860 743<br />
EBT 3,223 2,523 852 814<br />
Net profit 3,143 2,220 880 679<br />
EBITDA margin 41.8% 35.9% 40.1% 40.3%<br />
EBIT margin 29.0% 20.3% 29.0% 24.4%<br />
Net margin 26.1% 17.7% 29.7% 22.3%<br />
Versus consensus 2010e 2011e 3Q10e 4Q10e<br />
Revenue -2% 1% -5% 0%<br />
EBITDA -9% -9% -11% -1%<br />
EBIT -1% -9% -15% 4%<br />
EBT -2% -9% -2% 6%<br />
Net profit -3% -13% -4% -11%<br />
Consensus 2010e 2011e 3Q10e 4Q10e<br />
Revenue 12,257 12,424 3,111 3,044<br />
EBITDA 5,510 4,981 1,340 1,234<br />
EBIT 3,533 2,805 1,010 715<br />
EBT 3,290 2,768 871 772<br />
Net profit 3,244 2,565 913 763<br />
EBITDA margin 45.0% 40.1% 43.1% 40.5%<br />
EBIT margin 28.8% 22.6% 32.5% 23.5%<br />
Net margin 26.5% 20.6% 29.4% 25.1%<br />
Source: I/B/E/S, HSBC estimates<br />
Catalysts and risks<br />
Potential catalysts include strong operating<br />
metrics such as ASP, bit growth and cost<br />
reductions; potential transactions involving major<br />
creditor shareholders, including a majority stake<br />
sale; greater exposure to NAND; sharp debt<br />
reduction through internal cash generation, and<br />
greater transparency by K-IFRS implementation.<br />
Downside risks include strengthening KRW<br />
versus the USD, JPY and TWD; potential supplydemand<br />
balance deterioration as players migrate<br />
processes and increase capex; slower-thanexpected<br />
process migration due to technical and<br />
equipment procurement difficulties; and potential<br />
litigation.<br />
Forward PB band<br />
60,000<br />
3.5x<br />
Price (KRW)<br />
50,000<br />
40,000<br />
2.5x<br />
30,000<br />
20,000<br />
1.5x<br />
10,000<br />
0.5x<br />
-<br />
Jan-<br />
05<br />
Jan-<br />
06<br />
Jan-<br />
07<br />
Jan-<br />
08<br />
Jan-<br />
09<br />
Jan-<br />
10<br />
Source: HSBC<br />
Fundamentals<br />
Recent results<br />
Hynix posted consensus-beating 2Q10 results with<br />
OP passing the KRW1trn mark for the first time.<br />
Improved profitability was attributable in part to<br />
higher-end specialty DRAM; non-PC DRAM<br />
revenue is to rise to 60% of total in 3Q (2Q: mid-<br />
50%). Continued demand for higher-end products<br />
(such as server DRAM) and mobile DRAM should<br />
remain robust in 2H10, benefiting Hynix which has<br />
strong exposure to these segments. 3Q10 guidance<br />
is relatively positive; it expects DRAM ASP and<br />
bit growth of a ‘slight’ drop and mid-single digit<br />
increase respectively. For NAND it expects midteen<br />
drop in ASP and +20% bit growth.<br />
Operating performance outlook<br />
This year, Hynix is to see a very sharp recovery in<br />
profitability. We forecast Hynix’s consolidated<br />
revenue to rise c57% y-o-y this year, and for<br />
operating profit to rise c20-fold from a low<br />
KRW191bn base last year. At the parent level, we<br />
forecast revenue to rise c60% this year, and for<br />
the company to turn sharply profitable from a<br />
marginal loss last year.<br />
68
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Hynix consolidated revenue breakdown<br />
100%<br />
80%<br />
industry-wide supply and our assumption of<br />
another 6% strengthening in the KRW.<br />
DRAM revenue (KRWbn), operating margins<br />
60%<br />
12,000<br />
40.0%<br />
40%<br />
20%<br />
10,000<br />
8,000<br />
6,000<br />
20.0%<br />
0.0%<br />
0%<br />
2007 2008 2009 2010f 2011f 2012f<br />
DRAM NAND MCP and others<br />
4,000<br />
2,000<br />
-<br />
2007 2008 2009 2010f 2011f 2012f<br />
-20.0%<br />
-40.0%<br />
Source: Company, HSBC estimates<br />
Revenue (KRW bn)<br />
OPM<br />
Consolidated revenue, OP (KRWbn) and OPM (%)<br />
20,000<br />
40%<br />
15,000<br />
20%<br />
10,000<br />
0%<br />
5,000<br />
-<br />
-20%<br />
(5,000) 2008 2009 2010f 2011f 2012f -40%<br />
Revenue OP OP margin(%)<br />
Source: Company, HSBC estimates<br />
For 2011e, we think the impact of lower ASPs is<br />
unavoidable. We forecast consolidated revenue to<br />
rise c6% y-o-y and consolidated operating profit<br />
to decline 10% y-o-y. At the parent level, this<br />
translates into a 4% increase in revenue and a<br />
25% decline in operating profit. Our 2011e parent<br />
revenue forecast is in line with consensus, but our<br />
operating profit forecast is 9% below consensus.<br />
DRAM<br />
Revenue, margin outlook<br />
This year, we forecast consolidated revenue to<br />
rise 64% in KRW terms, on a 42% bit growth and<br />
blended ASP increase of 27%. We expect the<br />
KRW to strengthen 12% compared to last year.<br />
We forecast 2011e consolidated revenue growth<br />
to fall 4%; bit growth is to remain strong at 51%<br />
y-o-y but is offset by a 32% ASP fall on greater<br />
Source: Company, HSBC estimates<br />
We expect greater cost efficiencies (mostly by<br />
process migration and improved production<br />
techniques) plus operating leverage to expand<br />
operating profitability to 36.1% from 3.7% last<br />
year, reducing to 29.3% in 2011e. For 3Q10,<br />
Hynix guides DRAM ASP and bit growth of a<br />
‘slight’ drop and mid-single digit rise respectively.<br />
Bit growth, ASP assumptions (%)<br />
60%<br />
40%<br />
20%<br />
0%<br />
-20%<br />
-40%<br />
-60%<br />
2008 2009 2010f 2011f 2012f<br />
Bit growth ASP<br />
Source: Company, HSBC estimates<br />
Technology – process migration<br />
Positively, Hynix’s focus remains on profitability<br />
enhancement by 1) increasing higher margin<br />
DDR3 and specialist DRAM proportions and 2)<br />
cost reduction through process migration. DDR3<br />
proportion at end-2009 was c50%; by end-2010<br />
the DDR3 proportion should rise to c70-80%.<br />
69
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Together with Samsung, Hynix is leading the<br />
market in DRAM process migration. As at end-<br />
2009, c45% of DRAM was made on the 54nm<br />
process, plus a low single digit proportion of<br />
44nm process DRAM. Hynix is ramping 44nm<br />
production, initially focussed on the lucrative<br />
server market. Hynix achieved low-teens bit<br />
production on the 44nm process by 2Q10, slightly<br />
slower than previous guidance of mid-teens.<br />
DRAM process migration<br />
100%<br />
80%<br />
60%<br />
40%<br />
20%<br />
0%<br />
2007 2008 2009 2010f 2011f 2012f<br />
8xnm 6xnm 5xnm 4xnm 3xnm 2xnm<br />
Source: Company, HSBC estimates<br />
Nevertheless, the firm expects 44nm migration to<br />
reach 30% of DRAM bit production by end of 3Q<br />
and 50% by end of 4Q, following its revised<br />
capex in June. We expect Hynix to start mass<br />
production of its 3xnm process during 1Q11.<br />
growth areas of mobile, graphics, high-end server<br />
and specialist DRAM (used in consumer<br />
electronics). Hynix plans to raise global mobile<br />
DRAM market share to c30% in 2010 from c7.1%<br />
in 2007. Hynix expects the revenue proportion of<br />
non-PC DRAM to rise to 60% of total in 3Q (2Q:<br />
mid-50%; 2009: c53%). Server, mobile and<br />
specialist DRAM accounted for c20%, c10% and<br />
c10% of revenue in 2Q10.<br />
The main advantage of these segments, other than<br />
high growth rates, is relatively stable margins.<br />
Mobile DRAM prices tend not to be as volatile as<br />
PC DRAM; ASPs tend to fall c5% every quarter<br />
5%. In addition to PC and server use, consumer<br />
electronics devices are now migrating to DDR3,<br />
driving continued recovery of ASPs for specialist<br />
DRAM.<br />
NAND<br />
Revenue, margin outlook<br />
This year, we forecast consolidated revenue to<br />
rise 39% in KRW terms, on 94% bit growth and<br />
ASP decline of 21%. We expect the KRW to<br />
strengthen 12% compared to last year.<br />
NAND revenue growth<br />
4,000<br />
20.0%<br />
Wafer capacity (kwpy), average fab utilisation<br />
5,000<br />
4,000<br />
3,000<br />
2,000<br />
1,000<br />
-<br />
100%<br />
95%<br />
90%<br />
85%<br />
80%<br />
75%<br />
70%<br />
3,000<br />
2,000<br />
1,000<br />
-<br />
2007 2008 2009 2010f 2011f 2012f<br />
Revenue (KRW bn) OPM<br />
10.0%<br />
0.0%<br />
-10.0%<br />
-20.0%<br />
-30.0%<br />
2007 2008 2009 2010f 2011f 2012f<br />
Source: Company, HSBC estimates<br />
Wafer capacity (kwpy) Average utilisation (%)<br />
Source: Company, HSBC estimates<br />
Speciality DRAM<br />
A key Hynix strategy this year has been to<br />
diversify from commodity memory into rapid<br />
We forecast 2011e revenue growth of 53%; bit<br />
growth is to accelerate to 167% y-o-y on 33% rise<br />
in wafer capacity and process migration. We<br />
assume a 35% ASP fall. We also assume another<br />
6% strengthening in the KRW versus the USD.<br />
70
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
We are positive on Hynix’s NAND prospects on<br />
two accounts – rising fab capacity and accelerated<br />
process migration. Hynix lost market share in the<br />
past two years as it shut down uneconomical<br />
200mm fabs. But with strong NAND demand and<br />
removal of excess global capacity, and the<br />
financial ability to build capacity, Hynix is in a<br />
favourable environment to raise NAND<br />
production. This year’s capex includes capacity<br />
expansion of M11 (currently 50-60kwpm at<br />
300mm) to 80kwpm by year-end.<br />
Technology – process migration<br />
Following a couple years of production stagnation,<br />
Hynix’s priority is to narrow the technology gap<br />
with the NAND leaders. By this year-end, Hynix<br />
aims to have over 50% of NAND production on<br />
the 32nm process with another c35% on 41nm.<br />
Still, Hynix’s margins remain lower than rivals’.<br />
The current ramp up to 32nm process should<br />
partially close this gap. Hynix targets >50% of<br />
NAND bit production of 32nm by year-end.<br />
NAND process migration<br />
Production capacity, capex<br />
In June, Hynix raised 2010 capex by 33% to<br />
KRW3.05trn. This increase was not a surprise; as<br />
with Samsung, market expectations of Hynix’s<br />
2010 capex had already been higher. Positively,<br />
Hynix is spending the extra capex to accelerate<br />
process migration, raising its 44nm DRAM bit<br />
proportion from current c15% to c50% by yearend.<br />
More recently in October, Hynix raised its<br />
2010 capex further to KRW3.38trn to accelerate<br />
migration to 3xnm DRAM. Effectively, it brought<br />
forward some capex from next year to compete<br />
with Samsung, which started mass production on<br />
the 3xnm process in July this year.<br />
Consolidated capex (KRWbn)<br />
5,000<br />
4,000<br />
3,000<br />
2,000<br />
1,000<br />
-<br />
2008 2009 2010f 2011f 2012f<br />
100%<br />
80%<br />
60%<br />
Source: Company, HSBC estimates<br />
DRAM<br />
NAND<br />
40%<br />
20%<br />
0%<br />
2007 2008 2009 2010f 2011f 2012f<br />
8xnm 6xnm 5xnm 4xnm 3xnm 2xnm<br />
Source: Company, HSBC estimates<br />
Key to longer term competitiveness is process<br />
migration to 26nm at which level Hynix believes<br />
that it can generate comparable margins. In this<br />
respect, things are looking positive, as Hynix<br />
completed development of its 26nm process based<br />
64Gb NAND in February 2010 and customer<br />
samples are out, with certification within six months.<br />
Creditor shareholding<br />
In early March 2010 Hynix creditors sold down a<br />
6.7% stake (one creditor, Korea Finance Corp.,<br />
did not want to sell) in the firm through a block<br />
sale to mostly domestic institutional investors.<br />
Following the success of this selldown (priced at<br />
zero discount), a follow-through selldown in<br />
2H10 could be successful, especially if<br />
expectations of operating outlook improvements<br />
rise on strong results. Most recently there has<br />
been market speculation over various possible<br />
domestic strategic investors that could acquire the<br />
remaining 15% stake held by creditors.<br />
71
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Hynix consolidated revenue and OP model summary<br />
2008 2009 2010e 2011e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e<br />
Revenue (KRWbn)<br />
DRAM 4,985 6,002 9,831 9,433 2,271 2,623 2,483 2,454 2,250 2,270 2,433 2,479<br />
NAND 1,408 1,336 1,851 2,839 401 466 460 524 641 714 738 746<br />
MCP and others 427 568 746 946 149 190 199 209 219 230 242 254<br />
Total 6,818 7,906 12,429 13,217 2,821 3,279 3,142 3,187 3,110 3,215 3,413 3,479<br />
Revenue breakdown (%)<br />
DRAM 73% 76% 79% 71% 81% 80% 79% 77% 72% 71% 71% 71%<br />
NAND 21% 17% 15% 21% 14% 14% 15% 16% 21% 22% 22% 21%<br />
MCP and others 6% 7% 6% 7% 5% 6% 6% 7% 7% 7% 7% 7%<br />
Total 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100%<br />
Sequential change (%)<br />
DRAM -20% 20% 64% -4% 4% 15% -5% -1% -8% 1% 7% 2%<br />
NAND -23% -5% 39% 53% -10% 16% -1% 14% 22% 11% 3% 1%<br />
MCP and others -23% 33% 31% 27% -11% 27% 5% 5% 5% 5% 5% 5%<br />
Total -21% 16% 57% 6% 1% 16% -4% 1% -2% 3% 6% 2%<br />
EBIT (KRWbn)<br />
DRAM (979) 220 3,561 2,767 740 1,103 897 821 595 607 763 801<br />
NAND (349) (2) 193 409 30 50 55 59 115 159 119 16<br />
MCP and others (135) (27) 20 202 29 (74) 30 36 39 46 53 63<br />
Total (1,920) 191 3,741 3,379 799 1,045 982 915 749 812 936 881<br />
EBIT breakdown (%)<br />
DRAM 51% 115% 95% 82% 93% 106% 91% 90% 79% 75% 82% 91%<br />
NAND 18% -1% 5% 12% 4% 5% 6% 6% 15% 20% 13% 2%<br />
MCP and others 7% -14% 1% 6% 4% -7% 3% 4% 5% 6% 6% 7%<br />
Total 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100%<br />
Sequential change (%)<br />
DRAM -274% -122% 1518% -22% 17% 49% -19% -9% -27% 2% 26% 5%<br />
NAND -216% -100% -11311% 112% -63% 67% 11% 7% 95% 39% -25% -86%<br />
MCP and others 61% -80% -172% 923% -1029% -358% -140% 19% 11% 17% 16% 19%<br />
Total -490% -110% 1858% -10% 13% 31% -6% -7% -18% 8% 15% -6%<br />
EBIT margin (%)<br />
DRAM -20% 4% 36% 29% 33% 42% 36% 33% 26% 27% 31% 32%<br />
NAND -20% 0% 10% 14% 7% 11% 12% 11% 18% 22% 16% 2%<br />
MCP and others -32% -5% 3% 21% 19% -39% 15% 17% 18% 20% 22% 25%<br />
Total -28% 2% 30% 26% 28% 32% 31% 29% 24% 25% 27% 25%<br />
Source: Company, HSBC estimates<br />
Forecast revisions<br />
KRWbn ________________Revised ________________ _______________Previous ________________ _______________Revisions _______________<br />
2010f 2011f 2012f 2010f 2011f 2012f 2010f 2011f 2012f<br />
Revenue 12,045 12,557 13,724 12,061 13,043 14,236 0% -4% -4%<br />
EBIT 3,488 2,546 2,389 3,106 2,590 2,548 12% -2% -6%<br />
Net income 3,143 2,220 2,107 3,050 2,253 2,060 3% -1% 2%<br />
OP margin 29.0% 20.3% 17.4% 25.8% 19.9% 17.9%<br />
NP margin 26.1% 17.7% 15.4% 25.3% 17.3% 14.5%<br />
Source: Company data, HSBC estimates<br />
72
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Financials & valuation: Hynix Semiconductor Inc<br />
Overweight (V)<br />
Financial statements<br />
Year to 12/2009a 12/2010e 12/2011e 12/2012e<br />
Profit & loss summary (KRWb)<br />
Revenue 7,521 12,045 12,557 13,724<br />
EBITDA 1,722 5,033 4,510 4,613<br />
Depreciation & amortisation -1,838 -1,544 -1,964 -2,224<br />
Operating profit/EBIT -116 3,488 2,546 2,389<br />
Net interest -294 -162 -136 -95<br />
PBT -398 3,223 2,523 2,479<br />
HSBC PBT -398 3,223 2,523 2,479<br />
Taxation 50 -80 -303 -372<br />
Net profit -348 3,143 2,220 2,107<br />
HSBC net profit -348 3,143 2,220 2,107<br />
Cash flow summary (KRWb)<br />
Cash flow from operations 735 4,085 3,807 3,859<br />
Capex -1,478 -1,544 -1,960 -2,220<br />
Cash flow from investment -1,461 -2,305 -1,982 -2,169<br />
Dividends 0 0 0 0<br />
Change in net debt -1,012 -558 -1,465 -1,165<br />
FCF equity -744 2,461 1,624 1,571<br />
Balance sheet summary (KRWb)<br />
Intangible fixed assets 461 461 461 461<br />
Tangible fixed assets 6,325 7,861 8,217 8,738<br />
Current assets 3,568 4,611 6,342 7,834<br />
Cash & others 1,329 575 1,529 2,254<br />
Total assets 13,425 16,525 18,612 20,625<br />
Operating liabilities 3,193 4,150 4,528 4,874<br />
Gross debt 4,779 3,467 2,957 2,517<br />
Net debt 3,450 2,893 1,428 263<br />
Shareholders funds 5,452 8,907 11,127 13,234<br />
Invested capital 6,219 8,208 8,963 9,905<br />
Ratio, growth and per share analysis<br />
Year to 12/2009a 12/2010e 12/2011e 12/2012e<br />
Y-o-y % change<br />
Valuation data<br />
Year to 12/2009a 12/2010e 12/2011e 12/2012e<br />
EV/sales 2.0 1.1 1.0 0.8<br />
EV/EBITDA 8.6 2.7 2.7 2.4<br />
EV/IC 2.4 1.7 1.4 1.1<br />
PE* 4.7 6.6 7.0<br />
P/Book value 2.5 1.6 1.3 1.1<br />
FCF yield (%) -6.5 22.7 15.0 14.5<br />
Dividend yield (%) 0.0 0.0 0.0 0.0<br />
Note: * = Based on HSBC EPS (fully diluted)<br />
Issuer information<br />
Share price (KRW) 23,400 Target price (KRW) 32,000 Potent’l return (%) 36.8<br />
Reuters (Equity) 000660.KS Bloomberg (Equity) 000660 KS<br />
Market cap (USDm) 12,221 Market cap (KRWb) 13,812<br />
Free float (%) 81 Enterprise value (KRWb) 13728<br />
Country Korea Sector Semiconductors<br />
Analyst Nam Park Contact 852 2996 6591<br />
Price relative<br />
37507<br />
32507<br />
27507<br />
22507<br />
17507<br />
12507<br />
7507<br />
2507<br />
Source: HSBC<br />
2008 2009 2010 2011<br />
Hynix Semiconductor Inc Rel to KOSPI INDEX<br />
Note: price at close of 06 Oct 2010<br />
37507<br />
32507<br />
27507<br />
22507<br />
17507<br />
12507<br />
7507<br />
2507<br />
Revenue 15.8 60.1 4.2 9.3<br />
EBITDA 192.2 -10.4 2.3<br />
Operating profit -27.0 -6.2<br />
PBT -21.7 -1.7<br />
HSBC EPS -29.4 -5.1<br />
Ratios (%)<br />
Revenue/IC (x) 1.2 1.7 1.5 1.5<br />
ROIC -0.9 47.9 26.6 22.0<br />
ROE -6.6 43.8 22.2 17.3<br />
ROA -0.5 22.2 13.4 11.3<br />
EBITDA margin 22.9 41.8 35.9 33.6<br />
Operating profit margin -1.5 29.0 20.3 17.4<br />
EBITDA/net interest (x) 5.9 31.1 33.2 48.5<br />
Net debt/equity 63.3 32.5 12.8 2.0<br />
Net debt/EBITDA (x) 2.0 0.6 0.3 0.1<br />
CF from operations/net debt 21.3 141.2 266.6 1467.8<br />
Per share data (KRW)<br />
EPS Rep (fully diluted) -589.83 5017.81 3544.80 3364.65<br />
HSBC EPS (fully diluted) -589.83 5017.81 3544.80 3364.65<br />
DPS 0.00 0.00 0.00 0.00<br />
Book value 9246.98 14222.59 17767.39 21132.04<br />
73
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Inotera Memories (3474)<br />
A fully-migrated 50nm producer despite delay; targeting full-42nm<br />
in 1H10<br />
It is crucial how quickly capacity ramp-up and cycle time reduction<br />
can be accomplished, amid falling DRAM prices<br />
Cut 2010-11e revenue by 13% and 8%, cut TP to TWD17<br />
(previously TWD20), reiterate Neutral (V).<br />
Investment summary<br />
Competing with time<br />
As a foundry for Nanya Tech and Micron,<br />
Inotera just became Taiwan’s first fullymigrated<br />
50nm fab in 3Q, using Micron’s<br />
technology.<br />
The migration, however, was behind schedule<br />
in terms of wafer output delivery as cycle<br />
time has been lengthened as node migration<br />
was complicated by the trench to stack<br />
process at the same time.<br />
Meanwhile, we believe Inotera’s yield has<br />
been improving on 50nm despite the twomonth<br />
delay on wafer output.<br />
Still, we believe such problems encountered<br />
on node migration should continue to affect<br />
Inotera’s bit growth (hence Nanya Tech’s)<br />
until the cycle time can be reduced.<br />
With the delay, we now expect Inotera to<br />
achieve bit growth of 27% y-o-y (previously<br />
67%) in 2010, more conservative than Inotera’s<br />
revised guidance of 50% (previously 70-80%).<br />
Bottom line is that we believe Inotera will<br />
still experience a loss for 2010. Whether they<br />
will be able to deliver a profit in 4Q depends<br />
on how quickly they can reduce cycle time,<br />
how rapidly they can improve yield and how<br />
fast they can ramp up their 50nm capacity.<br />
Currently, we expect Inotera to marginally<br />
break even at net level for 4Q10. Inotera<br />
targets to ramp up their capacity to 130k by<br />
the end of 4Q.<br />
Fundamentals<br />
Technology migration<br />
Starting from mid-July, Inotera has successfully<br />
turned itself into a full 50nm fab. The next task<br />
for Inotera is to 1) ramp up current 50nm<br />
capacity and 2) improve cycle time for wafer<br />
input to come through the process.<br />
We believe wafer input for 3Q is around<br />
averaged 80-85k/month. This compares to<br />
Inotera’s target of ramping up capacity to<br />
130k/month by the end of 4Q.<br />
We estimate Inotera’s current wafer cycle<br />
time to be around 70 days. Inotera’s target is<br />
to reduce it to the mid-50s by end-2010.<br />
74
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
We estimate that Inotera’s 50nm technology<br />
proportion represents around 22% of total<br />
shipments in 3Q10, compared to the<br />
scheduled 50%. We believe the delay will<br />
likely affect bit growth in 4Q, where Inotera<br />
had targeted for 100% growth.<br />
On 42nm, Inotera will start its pilot run in late<br />
3Q. It is expected to have mass production by<br />
1H11.<br />
Bit growth, ASP assumptions (%)<br />
150%<br />
100%<br />
50%<br />
0%<br />
-50%<br />
2008 2009 2010f 2011f 2012f<br />
Bit growth ASP<br />
Source: Company data, HSBC estimates<br />
Process migration<br />
100%<br />
80%<br />
60%<br />
40%<br />
20%<br />
0%<br />
2007 2008 2009 2010f 2011f 2012f<br />
9xnm<br />
70nm<br />
50nm<br />
42nm<br />
Source: Company data, HSBC estimates<br />
Forecast revision<br />
On the delay of 50nm migration, lower wafer<br />
output in 3Q as a result of prolonged cycle time,<br />
we revise down Inotera’s 2010 forecast and now<br />
believe significant cost savings from node<br />
migration will only start in 4Q10-1Q11. We now<br />
forecast 4Q10 bit growth of 51% (company<br />
targets 100%) and 2010 bit growth of 27% y-o-y<br />
(previously 67%). We revise down Inotera’s<br />
2010-11e OP margin to -4% and 11%, from 6%<br />
and 18% respectively, and expect the margin<br />
recovery only to happen in 4Q. We now forecast<br />
Inotera’s ASP to go down by 16% in 3Q and<br />
another 10% in 4Q.<br />
Forecast revision for Inotera<br />
TWDm __ Revised ___ __ Previous ___ __Difference _<br />
2010 2011 2010 2011 2010 2011<br />
Sales 46,535 66,433 53,426 72,032 -13% -8%<br />
GP (1,987.1) 7,303 3,049 13,042 nm -44%<br />
OP (3,755.4) 5,642 1,125 11,385 nm -50%<br />
Net income (5,495.7) 3,167 (589.9) 8,100 nm -61%<br />
EPS (TWD) (1.2) 0.7 (0.1) 2.1 nm -68%<br />
GPM (%) -4% 11% 6% 18%<br />
OPM (%) -8% 8% 2% 16%<br />
NPM (%) -12% 5% -1% 11%<br />
Utilization 70% 70% 77% 77%<br />
rate (%)<br />
Bit growth (%) 27% 102% 67% 68%<br />
ASP growth<br />
(%)<br />
4% -31% -5% -28%<br />
Source: HSBC<br />
For 2010, we forecast Inotera to remain at a loss<br />
of TWD5.5bn, but swing to net profit of<br />
TWD3.2bn next year. In 2011, we expect Inotera<br />
to achieve bit growth of 102%, on bit growth that<br />
has been delayed from 3Q-4Q as a result<br />
migration. We also forecast its margin to improve<br />
to 8% next year from -8% this year.<br />
75
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
HSBC vs consensus<br />
HSBC 2010e 2011e 2012e<br />
Revenue 46,535 66,433 79,103<br />
EBITDA 28,578 41,830 49,352<br />
EBIT (3,755) 5,642 10,072<br />
Netincome (5,496) 3,167 7,136<br />
EPS(TWD) (1.2) 0.7 1.5<br />
EBITDAmargin 61.4% 63.0% 62.4%<br />
EBITmargin -8.1% 8.5% 12.7%<br />
Netmargin -11.8% 4.8% 9.0%<br />
HSBC vs consensus 2010e 2011e 2012e<br />
Revenue -12.6% -3.1% 9.6%<br />
EBITDA -14.9% -7.7% 0.4%<br />
EBIT -329.6% -43.9% -33.7%<br />
Net income -2178.4% -60.5% -18.1%<br />
EPS (TWD) -761.2% -70.4% -37.9%<br />
Consensus 2010e 2011e 2012e<br />
Revenue 53,219 68,564 72,195<br />
EBITDA 33,583 45,319 49,158<br />
EBIT 1,636 10,050 15,188<br />
Net income 264 8,010 8,715<br />
EPS (TWD) 0.2 2.3 2.5<br />
EBITDA margin 63.1% 66.1% 68.1%<br />
EBIT margin 3.1% 14.7% 21.0%<br />
Net margin 0.5% 11.7% 12.1%<br />
Source: Company data, I/B/E/S consensus<br />
Valuations, rating<br />
We lower our 12-month target price to TWD17<br />
from TWD20, derived by applying a target FY11<br />
PB multiple of 1.5x, compared to 1.7x before. Our<br />
method for valuation remains unchanged from<br />
before, but we roll over to an FY11e PB multiple<br />
from FY10e previously. We use a mid-point<br />
historical PB range to derive the multiple. The<br />
stock has historically traded at PB of 0.5-2.5x,<br />
with an ROE range of c2% to 25% (prior to the<br />
capacity glut starting in 2007).<br />
Under our research model, for Taiwan stocks with<br />
a volatility indicator, the Neutral band is 10ppt<br />
above and below the hurdle rate of 9.5% (risk-free<br />
rate of 4% plus risk premium of 5.5%). This<br />
translates into a Neutral band of -0.5% to +19.5%<br />
around the current share price. Our 12-month<br />
target price of TWD17 suggests a potential return<br />
of 4.7%, which is within the Neutral band; thus,<br />
we reiterate our Neutral (V) rating.<br />
PB vs ROE<br />
3.0<br />
10%<br />
5%<br />
2.0<br />
0%<br />
-5%<br />
1.0<br />
-10%<br />
-15%<br />
0.0<br />
-20%<br />
Apr-06 Apr-07 Apr-08 Apr-09 Apr-10<br />
PB (x , LHS)<br />
ROE (%, RHS)<br />
Source: HSBC<br />
Forward PB band<br />
60<br />
Price (TWD)<br />
40<br />
2.4x<br />
1.8x<br />
20<br />
1.2x<br />
0.6x<br />
-<br />
Apr-06 Apr-07 Apr-08 Apr-09 Apr-10<br />
Source: Datastream, HSBC<br />
Catalysts and risks<br />
Industry-specific catalysts include:<br />
Favourable DRAM demand and supply in<br />
2010. Tight global DRAM supply given<br />
limited capex spending in 2009.<br />
Better than expected results and strong<br />
guidance from PC OEMs, such as recent Dell<br />
and HP results.<br />
PC demand to pick up from 2Q10 onwards,<br />
boosted by corporate replacement and<br />
upgrade spending.<br />
76
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Company-specific catalysts include:<br />
Inotera’s cost improvement by migrating to<br />
50nm technology.<br />
Micron and Nanya Tech’s ASP premium to<br />
other players.<br />
Industry-specific risks include:<br />
Faster than expected DRAM ASP erosion due<br />
to greater capacity coming online and/or<br />
slower than expected PC unit shipments.<br />
Company-specific risks include:<br />
Delay in 5xnm and 4cnm technology<br />
migration, lower than expected yield achieved.<br />
Potential requirement for further funding to<br />
support capex.<br />
Since Inotera has proprietary technology, the<br />
cost of transferring technology from Micron<br />
may insert pressure on its profitability.<br />
Raw material price hike like silicon wafers.<br />
Raw material price hike like silicon wafers.<br />
Elpida’s possible alliance with other<br />
Taiwanese DRAM makers likely help create a<br />
more competitive environment among<br />
Taiwanese DRAM players.<br />
Quarterly assumptions for Inotera<br />
TWDm 1Q10a 2Q10e 3Q10e 4Q10e 1Q10a 2Q11e 3Q11e 4Q11e<br />
Sales 11,518 11,258 9,963 13,796 16,694 16,414 16,907 16,418<br />
GP (559) (760) (1805) 1137 2,053 1,826 1,972 1,452<br />
OP (1219) (1142) (2164) 769 1,686 1,399 1,533 1,025<br />
Net income (1561) (1810) (2305) 180 1,086 828 948 305<br />
EPS (TWD) (0.4) (0.4) (0.5) 0.0 0.2 0.2 0.2 0.1<br />
GPM (%) -5% -7% -18% 8% 12% 11% 12% 9%<br />
OPM (%) -11% -10% -22% 6% 10% 9% 9% 6%<br />
NPM (%) -14% -16% -23% 1% 7% 5% 6% 2%<br />
Utilization rate (%) 78% 67% 67% 70% 83% 85% 89% 92%<br />
Bit growth (%) -3% -15% 4% 51% 32% 9% 14% 8%<br />
ASP growth (%) -6% 16% -16% -10% -9% -10% -10% -10%<br />
Source: Company data, HSBC estimates<br />
77
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Financials & valuation: Inotera Memories Inc.<br />
Neutral (V)<br />
Financial statements<br />
Year to 12/2009a 12/2010e 12/2011e 12/2012e<br />
Profit & loss summary (TWDm)<br />
Revenue 36,104 46,535 66,433 79,103<br />
EBITDA 20,154 28,578 41,830 49,352<br />
Depreciation & amortisation -29,795 -32,334 -36,187 -39,279<br />
Operating profit/EBIT -9,641 -3,755 5,642 10,072<br />
Net interest -1,549 -2,047 -2,277 -1,582<br />
PBT -11,477 -5,496 3,725 8,920<br />
HSBC PBT -11,477 -5,496 3,725 8,920<br />
Taxation 0 0 -559 -1,784<br />
Net profit -11,477 -5,496 3,167 7,136<br />
HSBC net profit -11,477 -5,496 3,167 7,136<br />
Cash flow summary (TWDm)<br />
Cash flow from operations 11,228 24,158 34,449 44,520<br />
Capex -13,095 -58,000 -22,300 -12,500<br />
Cash flow from investment -23,341 -35,800 -31,429 -31,629<br />
Dividends 0 0 0 0<br />
Change in net debt -14,367 25,826 -13,822 -29,536<br />
FCF equity -1,867 -33,842 11,591 30,236<br />
Balance sheet summary (TWDm)<br />
Intangible fixed assets 1,447 1,037 779 675<br />
Tangible fixed assets 100,032 125,390 110,889 85,343<br />
Current assets 18,917 18,179 28,157 44,251<br />
Cash & others 5,376 442 3,263 14,799<br />
Total assets 120,396 144,606 139,826 130,269<br />
Operating liabilities 29,459 31,873 34,927 36,234<br />
Gross debt 42,109 63,000 52,000 34,000<br />
Net debt 36,733 62,558 48,737 19,201<br />
Shareholders funds 48,828 49,733 52,899 60,035<br />
Invested capital 85,561 112,291 101,636 79,235<br />
Ratio, growth and per share analysis<br />
Year to 12/2009a 12/2010e 12/2011e 12/2012e<br />
Y-o-y % change<br />
Valuation data<br />
Year to 12/2009a 12/2010e 12/2011e 12/2012e<br />
EV/sales 3.1 3.0 1.9 1.2<br />
EV/EBITDA 5.5 4.8 3.0 1.9<br />
EV/IC 1.3 1.2 1.2 1.2<br />
PE* 23.6 10.5<br />
P/Book value 1.2 1.5 1.4 1.2<br />
FCF yield (%) -2.5 -45.2 15.5 40.4<br />
Dividend yield (%) 0.0 0.0 0.0 0.0<br />
Note: * = Based on HSBC EPS (fully diluted)<br />
Issuer information<br />
Share price (TWD) 16.20 Target price (TWD) 17.0 Potent’l return (%) 4.7<br />
Reuters (Equity) 3474.TW Bloomberg (Equity) 3474 TT<br />
Market cap (USDm) 2,407 Market cap (TWDm) 74,804<br />
Free float (%) 46 Enterprise value (TWDm) 137362<br />
Country Taiwan Sector Electronic Equipment<br />
Analyst Carolyn Poon Contact +852 2996 6586<br />
Price relative<br />
32<br />
27<br />
22<br />
17<br />
12<br />
7<br />
2<br />
2008 2009 2010 2011<br />
Inotera Memories Inc. Rel to TAIWAN WEIGHTED INDEX<br />
Source: HSBC<br />
Note: price at close of 06 Oct 2010<br />
32<br />
27<br />
22<br />
17<br />
12<br />
7<br />
2<br />
Revenue -3.8 28.9 42.8 19.1<br />
EBITDA 108.8 41.8 46.4 18.0<br />
Operating profit 78.5<br />
PBT 139.4<br />
HSBC EPS 125.3<br />
Ratios (%)<br />
Revenue/IC (x) 0.4 0.5 0.6 0.9<br />
ROIC -10.1 -3.8 4.4 8.7<br />
ROE -23.2 -11.2 6.2 12.6<br />
ROA -7.8 -2.6 3.6 6.3<br />
EBITDA margin 55.8 61.4 63.0 62.4<br />
Operating profit margin -26.7 -8.1 8.5 12.7<br />
EBITDA/net interest (x) 13.0 14.0 18.4 31.2<br />
Net debt/equity 75.2 125.8 92.1 32.0<br />
Net debt/EBITDA (x) 1.8 2.2 1.2 0.4<br />
CF from operations/net debt 30.6 38.6 70.7 231.9<br />
Per share data (TWD)<br />
EPS Rep (fully diluted) -3.19 -1.19 0.69 1.55<br />
HSBC EPS (fully diluted) -3.19 -1.19 0.69 1.55<br />
DPS 0.00 0.00 0.00 0.00<br />
Book value 13.56 10.77 11.46 13.00<br />
78
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Nanya Technology (2408)<br />
Bit growth limited by Inotera’s 50nm tech migration delay, but<br />
determined to move fast on 42nm node migration<br />
Expectation of more DRAM supply in 2H has dragged DRAM<br />
price exogenously<br />
Cut TP to TWD20 from TWD25, maintain Neutral (V) rating<br />
Investment summary<br />
Are we there yet?<br />
Nanya Technology is the first Taiwanese<br />
DRAM company to start 50nm technology<br />
migration, which is a long-term positive for<br />
the company’s competitiveness. The<br />
execution, however, took longer than<br />
expected time and hence delivering less bit<br />
growth than planned, as foundry Inotera has<br />
delayed its migration process starting from<br />
2Q which impacted wafer output in 2Q-3Q.<br />
The process is complicated by the fact that<br />
Inotera is switching from trench to stack<br />
process at the same time.<br />
Such delay has, unfortunately, coincided with<br />
an ASP decline in 2H, thanks to softer PC<br />
demand and the expectation that more<br />
industry supply (bit growth) expected to be<br />
introduced to the market in 4Q.<br />
The bottom line is that Nanya Tech’s net<br />
profit and margin recovery story is not likely<br />
to materialise this year but will be postponed<br />
to next year: We now forecast net income loss<br />
of TWD5bn for this year, with a net profit of<br />
TWD2.9bn in 2011. Sudden ramp-up of bit<br />
growth, which the market expects to see in<br />
4Q10, should be softer than expected while<br />
postponing to 1H11.<br />
It is also crucial to see whether Nanya Tech<br />
can make its 40nm migration ahead of<br />
schedule and get the yield right quickly amid<br />
an environment where DRAM ASPs are<br />
pointing south.<br />
Fundamentals<br />
Technology migration<br />
Nanya’s bit growth has been slower than planned<br />
in 2Q and 3Q, on its foundry (Inotera)’s 50nm<br />
migration delay. We estimate that Nanya Tech’s<br />
50nm technology proportion represents around<br />
35% of total shipment in 2Q10, compared to the<br />
scheduled 50%. We believe the delay will<br />
continue, resulting in lower-than-expect DRAM<br />
bit growth in 2H, more than offsetting bit growth<br />
output driven by 50nm capacity ramp up in 4Q.<br />
Still, Nanya is accelerating its own fab’s 42nm<br />
migration and expects to have mass production<br />
by 4Q10.<br />
Capacity ramp-up: part of the bit growth in 2H<br />
should also come from its own capacity ramp-up.<br />
Nanya’s own capacity is expected to reach<br />
79
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
50kwpm by 4Q, from 40kwpm at end-2Q, and<br />
further to 60kwpm by 1H11.<br />
Bit growth, ASP assumptions (%)<br />
150%<br />
100%<br />
50%<br />
0%<br />
-50%<br />
-100%<br />
2008 2009 2010f 2011f<br />
Supply bit growth<br />
ASP<br />
Source: Company data, HSBC estimates<br />
2H result<br />
Nanya’s 3Q revenue of TWD14.9bn, was 5% q-oq<br />
weaker. While details of the result will not be<br />
announced until late October, 3Q bit growth was<br />
likely dragged by Inotera’s longer cycle time,<br />
despite Nanya’s own 50nm migration being on<br />
schedule. Together with a DRAM ASP fall, we<br />
forecast Nanya Tech to linger in the red at both<br />
the operating and net profit levels this year,<br />
delivering a loss of TWD5.0bn for 2010. We<br />
expect 4Q to have a recovery of bit growth of<br />
29% as 50nm wafer input in 3Q has started to<br />
materialise, with revenue growth of 7.5% q-o-q<br />
and a net loss of TWD1.0bn.<br />
Process migration<br />
100%<br />
80%<br />
60%<br />
40%<br />
Crucial to get yield right<br />
Unless Inotera can successfully reduce its cycle<br />
time from c70 days to its target of mid-50 days in<br />
4Q10, it appears difficult for Nanya to catch up<br />
with the annual 35% bit growth target for 2010.<br />
We believe it is thus crucial that Nanya can<br />
acceleration its own migration to 40nm ahead of<br />
schedule, which should at least offset some of the<br />
bit growth loss caused by Inotera’s migration<br />
delay. We have revised down Nanya’s annual bit<br />
growth to 10% from 42% for 2010.<br />
Forecast revision<br />
Forecast revision for Nanya Tech<br />
TWDm __ Revised ___ __ Previous ___ __Difference<br />
2010 2011 2010 2011 2010 2011<br />
Sales 60,787 83,749 74,339 101,705 -18% -18%<br />
GP 3,313 15,514 9,846 20,370 -66% -24%<br />
OP (3,660) 6,553 2,635 8,572 nm nm<br />
Net income (4,999) 2,849 1,889 6,152 nm nm<br />
EPS (TWD) (1.46) 0.83 0.55 1.79 nm nm<br />
GPM (%) 5% 19% 13% 20%<br />
OPM (%) -6% 8% 4% 8%<br />
NPM (%) -8% 3% 3% 6%<br />
Utilization 82% 88% 89% 89%<br />
rate (%)<br />
Bit growth (%) 10% 110% 42% 45%<br />
ASP growth<br />
(%)<br />
34% -36% -15% -14%<br />
Source: HSBC estimates<br />
On 1) lower DRAM ASP expectation in 2H10,<br />
and 2) Inotera’s longer-than-expected cycle time<br />
in 50nm node migration, we lower our Nanya’s<br />
2010-11 revenue forecast by 18%. We now<br />
forecast Nanya’s 2010 sales growth of 43% for<br />
2010. We now believe Nanya Tech will achieve a<br />
-6% OPM for 2010 instead of our previous<br />
forecast of 4%.<br />
20%<br />
0%<br />
2007 2008 2009 2010f 2011f 2012f<br />
90nm 70nm 50nm 42nm 3x nm 2x nm<br />
Source: Company data, HSBC estimates<br />
80
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
HSBC vs. consensus<br />
HSBC 2010e 2011e 2012e<br />
Revenue 60,787 83,749 98,318<br />
EBITDA 12,620 25,679 32,510<br />
EBIT (3,660) 6,553 10,748<br />
Net income (4,999) 2,849 6,311<br />
EPS (KRW) (1.5) 0.8 1.9<br />
EBITDA margin 20.8% 30.7% 33.1%<br />
EBIT margin -6.0% 7.8% 10.9%<br />
Net margin -8.2% 3.4% 6.4%<br />
HSBC vs<br />
2010e 2011e 2012e<br />
consensus<br />
Revenue -15.6% -11.9% -11.2%<br />
EBITDA -24.3% 9.2% -1.5%<br />
EBIT 134.8% 21.4% 15.5%<br />
Net income 798.2% -19.6% -14.0%<br />
EPS (KRW) 91.1% -18.4% -14.2%<br />
Consensus 2010e 2011e 2012e<br />
Revenue 72,064 95,109 110,667<br />
EBITDA 16,677 23,510 33,007<br />
EBIT (1,559) 5,397 9,304<br />
Net income (555) 3,586 7,397<br />
EPS (KRW) (0.8) 1.0 2.2<br />
EBITDA margin 23.1% 24.7% 29.8%<br />
EBIT margin -2.2% 5.7% 8.4%<br />
Net margin -0.8% 3.8% 6.7%<br />
Source: Company data, I/B/E/S consensus<br />
Catalysts<br />
Industry-specific catalysts include<br />
DRAM ASP fall less than expected in the<br />
coming quarters<br />
PC demand better than expected on corporate<br />
demand pick up<br />
Higher-than-expected DRAM content per box<br />
on lower DRAM ASP<br />
Company-specific catalysts include:<br />
40nm migration comes ahead of schedule<br />
Diversification of DRAM business to less<br />
volatile and more specialized areas like<br />
mobile DRAM.<br />
Risks<br />
Industry-specific risks include<br />
Sluggish PC demand, delayed corporate<br />
replacement demand<br />
tablet cannibalisation higher than expected<br />
Elpida’s possible alliance with other<br />
Taiwanese DRAM makers likely help create a<br />
more competitive environment among<br />
Taiwanese DRAM players<br />
Company-specific risks include:<br />
Inotera’s 40nm migration comes behind<br />
schedule<br />
Equity funding likely in 2H which should<br />
lead to equity dilution for current<br />
shareholders<br />
Valuations, rating<br />
We lower our 12-month target price to TWD20 from<br />
TWD25, derived by applying a target FY11e PB<br />
multiple of 1.9x (based on 12.2% ROE (11.3%<br />
previously) and 9.5% COE), compared to 2.1x<br />
before. We also roll over our PB target multiple to<br />
2011e from 2010e. The PB multiple of 1.9x is close<br />
to the lower end of Nanya’s historical PB range.<br />
Our method for valuation remains unchanged<br />
from before. The stock has historically traded at a<br />
PB of 1.6-4.9x, with an ROE range of -47% to 9%<br />
(prior to the capacity glut starting in 2007).<br />
Under our research model, for Taiwan stocks with<br />
a volatility indicator, the Neutral band is 10ppt<br />
above and below the hurdle rate of 9.5% (risk-free<br />
rate of 4% plus risk premium of 5.5%). This<br />
translates into a Neutral band of -0.5% to +19.5%<br />
around the current share price. Our 12-month<br />
target price of TWD20 suggests a potential return<br />
of 3.6%, which is within the Neutral band; thus,<br />
we reiterate our Neutral (V) rating.<br />
81
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
PB vs ROE<br />
Forward PB band<br />
6.0<br />
5.0<br />
4.0<br />
3.0<br />
2.0<br />
1.0<br />
50%<br />
30%<br />
10%<br />
-10%<br />
-30%<br />
-50%<br />
80<br />
60<br />
40<br />
20<br />
-<br />
Price (TWD)<br />
4.0x<br />
3.0x<br />
2.0x<br />
1.0x<br />
Jan-05 Jan-06 Jan-07 Jan-08 Jan-09<br />
PB (x , LHS)<br />
ROE (%, RHS)<br />
Jan-<br />
05<br />
Jan-<br />
06<br />
Jan-<br />
07<br />
Jan-<br />
08<br />
Jan-<br />
09<br />
Jan-<br />
10<br />
Source: HSBC<br />
Source: Datastream, HSBC<br />
Assumptions on Nanya Tech<br />
TWDm 1Q10a 2Q10a 3Q10e 4Q10e 1Q10e 2Q10e 3Q10e 4Q10e<br />
Sales 14,121 15,718 14,916 16,033 18,337 20,440 22,437 22,535<br />
GP 189 1,589 705 829 2,141 3,517 4,804 5,051<br />
OP (1225) (516) (933) (986) 198 1,310 2,314 2,731<br />
Net income (1625) (1086) (1277) (997) (719) 393 1,522 1,689<br />
EPS (TWD) (0.48) (0.32) (0.37) (0.29) (0.21) 0.11 0.44 0.49<br />
GPM (%) 1% 10% 5% 5% 12% 17% 21% 22%<br />
OPM (%) -9% -3% -6% -6% 1% 6% 10% 12%<br />
NPM (%) -12% -7% -9% -6% -4% 2% 7% 7%<br />
Utilization rate (%) 80% 87% 80% 83% 85% 88% 89% 90%<br />
Bit growth (%) -9% 3% 12% 29% 26% 21% 19% 9%<br />
ASP growth (%) -6% 9% -16% -18% -10% -8% -8% -8%<br />
Source: Company data, HSBC estimates<br />
82
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Financials & valuation: Nanya Technology<br />
Financial statements<br />
Year to 12/2009a 12/2010e 12/2011e 12/2012e<br />
Profit & loss summary (TWDm)<br />
Revenue 42,456 60,787 83,749 98,318<br />
EBITDA -1,158 12,620 25,679 32,510<br />
Depreciation & amortisation -14,761 -16,280 -19,126 -21,762<br />
Operating profit/EBIT -15,919 -3,660 6,553 10,748<br />
Net interest -1,068 -1,574 -2,492 -2,859<br />
PBT -20,754 -5,434 3,561 7,889<br />
HSBC PBT -20,754 -5,434 3,561 7,889<br />
Taxation 0 435 -712 -1,578<br />
Net profit -20,754 -4,999 2,849 6,311<br />
HSBC net profit -20,754 -4,999 2,849 6,311<br />
Cash flow summary (TWDm)<br />
Cash flow from operations -10,211 1,101 12,735 24,529<br />
Capex -7,533 -30,000 -25,500 -25,500<br />
Cash flow from investment -9,071 -14,490 -11,171 -25,836<br />
Dividends 0 0 0 0<br />
Change in net debt -5,821 35,446 17,648 3,885<br />
FCF equity -17,744 -28,465 -13,477 -2,548<br />
Balance sheet summary (TWDm)<br />
Intangible fixed assets 4,993 10,464 11,754 11,381<br />
Tangible fixed assets 74,691 86,209 94,964 101,411<br />
Current assets 37,296 33,584 45,324 49,316<br />
Cash & others 13,663 1,509 1,861 -1,024<br />
Total assets 132,953 148,230 170,015 180,083<br />
Operating liabilities 35,174 31,158 32,094 34,850<br />
Gross debt 60,405 83,697 101,697 102,697<br />
Net debt 46,742 82,188 99,836 103,721<br />
Shareholders funds 37,375 33,375 36,224 42,536<br />
Invested capital 68,142 97,589 118,086 128,282<br />
Ratio, growth and per share analysis<br />
Year to 12/2009a 12/2010e 12/2011e 12/2012e<br />
Y-o-y % change<br />
Revenue 16.9 43.2 37.8 17.4<br />
EBITDA 103.5 26.6<br />
Operating profit 64.0<br />
PBT 121.5<br />
HSBC EPS 121.5<br />
Ratios (%)<br />
Revenue/IC (x) 1.2 1.2 1.4 1.5<br />
ROIC -39.8 -3.0 12.1 16.8<br />
ROE -63.3 -14.1 8.2 16.0<br />
ROA -29.2 -4.4 5.8 9.6<br />
EBITDA margin -2.7 20.8 30.7 33.1<br />
Operating profit margin -37.5 -6.0 7.8 10.9<br />
EBITDA/net interest (x) 8.0 10.3 11.4<br />
Net debt/equity 125.1 246.3 275.6 243.8<br />
Net debt/EBITDA (x) -40.4 6.5 3.9 3.2<br />
CF from operations/net debt 1.3 12.8 23.6<br />
Per share data (TWD)<br />
EPS Rep (fully diluted) -8.67 -1.46 0.83 1.84<br />
HSBC EPS (fully diluted) -8.67 -1.46 0.83 1.84<br />
DPS 0.00 0.00 0.00 0.00<br />
Book value 15.61 9.72 10.55 12.39<br />
Valuation data<br />
Year to 12/2009a 12/2010e 12/2011e 12/2012e<br />
EV/sales 2.3 2.1 1.8 1.5<br />
EV/EBITDA 10.3 5.8 4.7<br />
EV/IC 1.4 1.3 1.3 1.2<br />
PE* 23.3 10.5<br />
P/Book value 1.2 2.0 1.8 1.6<br />
FCF yield (%) -35.3 -59.0 -27.9 -5.3<br />
Dividend yield (%) 0.0 0.0 0.0 0.0<br />
Note: * = Based on HSBC EPS (fully diluted)<br />
Issuer information<br />
Share price (TWD) 19.30 Target price (TWD) 20.00 Potent’l return (%) 3.6<br />
Reuters (Equity) 2408.TW Bloomberg (Equity) 2408 TT<br />
Market cap (USDm) 2,131 Market cap (TWDm) 66,237<br />
Free float (%) 74 Enterprise value (TWDm 130451<br />
)<br />
Country Taiwan Sector Semiconductors<br />
Analyst Carolyn Poon Contact +852 2996 6586<br />
Price relative<br />
64<br />
54<br />
44<br />
34<br />
24<br />
14<br />
4<br />
2008 2009 2010 2011<br />
Nanya Technologies Rel to TAIWAN WEIGHTED INDEX<br />
Source: HSBC<br />
Note: price at close of 06 Oct 2010<br />
Neutral (V)<br />
64<br />
54<br />
44<br />
34<br />
24<br />
14<br />
4<br />
83
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Samsung Electronics (005930)<br />
Leading DRAM and NAND technologies, market share, strong<br />
balance sheet and vertical integration make SEC our top pick<br />
Samsung’s strong product mix and accelerated process migration<br />
leverages into structural growth areas such as mobile, servers<br />
Remain OW, TP unchanged at KRW1,139,000. An HSBC <strong>Asia</strong><br />
Super Ten idea<br />
Investment summary<br />
For detailed analysis, please refer to our recent<br />
40-page report, Samsung Electronics – OW:<br />
Comprehensive health check, 24 September 2010.<br />
The memory leader<br />
Samsung Electronics is our preferred play in the<br />
memory sector. Samsung has:<br />
Highly competitive technologies and leading<br />
market share leadership in both DRAM and<br />
NAND (which is faster growing).<br />
Leading process migration, having started<br />
35nm DRAM production in 3Q10. We also<br />
expect 2xnm NAND production to start in<br />
4Q10.<br />
Strong presence in promising technologies<br />
such as phase change memory. Only Micron<br />
now has a comparable product portfolio post<br />
its recent acquisition of Numonyx.<br />
A massive balance sheet (cUSD10bn net cash<br />
at the parent level at end-2009) cash flow<br />
backing R&D and capex where necessary.<br />
currently playing out with rapid process<br />
migration and slew of advanced products.<br />
A level of vertical integration that is unique<br />
amongst its peers, giving it access to a vast<br />
internal market of electronics products.<br />
Valuations, rating<br />
Valuations summary<br />
Methodology Value (KRW) Criteria<br />
PB 1,323,318 2.2x FY11e BVPS<br />
Sum of parts 978,350<br />
DCF 1,079,363 12%<br />
Average 1,127,010<br />
Dividend 12,000<br />
Target price 1,139,010<br />
Current price 793,000<br />
Up(downside) 43.6%<br />
Source: HSBC<br />
To value Samsung Electronics, we use an average<br />
of 2011e PB (2.2x target multiple), sum-of-theparts,<br />
and DCF valuations (12% WACC with riskfree<br />
rate 4%, equity risk premium 8%, β=1) to<br />
derive a 12-month target price of KRW1,139,000.<br />
Our price assumes a dividend of KRW12,000.<br />
Determination to expand leadership with more<br />
advanced and cost-effective technologies,<br />
84
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
HSBC vs consensus<br />
(KRWbn) 3Q10e 4Q10e 2010e 2011e<br />
Revenue 41,064 42,884 156,476 169,642<br />
EBITDA 8,083 8,316 31,263 32,792<br />
EBIT 4,942 4,424 18,785 18,072<br />
Net income 4,369 3,781 16,420 16,180<br />
Common EPS (KRW) 29,658 25,669 111,477 109,848<br />
Diluted EPS (KRW) 25,678 22,224 96,516 95,105<br />
EBITDA margin 19.7% 19.4% 20.0% 19.3%<br />
EBIT margin 12.0% 10.3% 12.0% 10.7%<br />
Net margin 10.6% 8.8% 10.5% 9.5%<br />
HSBC vs consensus 3Q10e 4Q10e 2010e 2011e<br />
Revenue 0.0% 1.5% 0.1% 0.6%<br />
EBITDA -7.1% -0.3% 6.3% 12.3%<br />
EBIT -4.6% 6.1% 1.2% 0.1%<br />
Net income -7.1% -2.7% -2.0% 2.2%<br />
EPS (KRW) -11.2% -9.4% -9.1% -4.5%<br />
Consensus 3Q10e 4Q10e 2010e 2011e<br />
Revenue 41,059 42,238 156,312 168,634<br />
EBITDA 8,705 8,341 29,406 29,206<br />
EBIT 5,180 4,169 18,554 18,051<br />
Net income 4,702 3,887 16,750 15,832<br />
EPS (KRW) 28,922 24,523 106,170 99,562<br />
EBITDA margin 21.2% 19.7% 18.8% 17.3%<br />
EBIT margin 12.6% 9.9% 11.9% 10.7%<br />
Net margin 11.5% 9.2% 10.7% 9.4%<br />
Source: I/B/E/S, HSBC forecasts<br />
Under our research model, for stocks without a<br />
volatility indicator, the Neutral band is 5ppt above<br />
and below the hurdle rate of 10.5% for Korean<br />
stocks, or 5.5-15.5% around the current share<br />
Forward PB band<br />
1,400,000<br />
1,300,000<br />
1,200,000<br />
1,100,000<br />
1,000,000<br />
900,000<br />
800,000<br />
700,000<br />
600,000<br />
500,000<br />
400,000<br />
Source: Datastream, HSBC<br />
Price (KRW)<br />
Jan-<br />
05<br />
Jan-<br />
06<br />
Jan-<br />
07<br />
Jan-<br />
08<br />
Jan-<br />
09<br />
Jan-<br />
10<br />
2.5x<br />
2.0x<br />
1.5x<br />
1.0x<br />
price. Our 12-month target price suggests a<br />
potential return of 43.6%, which is above the<br />
Neutral band. Hence we reiterate our Overweight<br />
rating on the stock. Volatile stocks are defined as<br />
those having historical volatility (defined as the<br />
past month’s average of the daily 365-day moving<br />
average volatilities) of more than 40%.<br />
Catalysts and risks<br />
Upside risks to our rating and estimates include<br />
the following:<br />
Stronger-than-expected quarterly results.<br />
New product or new technology launches that<br />
could drive higher revenues and/or margins,<br />
including 4G-related devices, smartphones,<br />
tablet devices.<br />
Business expansion into promising areas, a<br />
pick-up in the global economy and demand.<br />
Downside risks to our rating and estimates include<br />
the following:<br />
Strengthening KRW versus other currencies,<br />
notably the EUR, USD, JPY.<br />
Sharp increases in industry-wide production<br />
capacity in memory and TFT-LCD, leading to<br />
faster-than-expected ASP declines.<br />
Price erosion on key products such as<br />
handsets on intensifying competition.<br />
Rebounding Japanese competition following<br />
restructuring efforts, complete with new<br />
generation products such as 3DTV.<br />
Protracted delay in global economic recovery<br />
dampening demand.<br />
Resolution of any outstanding IP and<br />
regulatory disputes.<br />
Fundamentals<br />
Operating performance outlook<br />
Telecoms and digital media supported Samsung’s<br />
consolidated profits last year; this year will be<br />
memory’s turn. Semiconductors was c20% of<br />
Samsung’s consolidated revenue last year; this<br />
proportion is set to rise to c24% by 2011e.<br />
85
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Samsung consolidated revenue breakdown<br />
Semiconductors revenue, OP (KRWbn) and OPM (%)<br />
100%<br />
80%<br />
60%<br />
40%<br />
20%<br />
0%<br />
2007 2008 2009e 2010f 2011f<br />
50,000<br />
40,000<br />
30,000<br />
20,000<br />
10,000<br />
-<br />
2007 2008 2009e 2010f 2011f<br />
30%<br />
25%<br />
20%<br />
15%<br />
10%<br />
5%<br />
0%<br />
Semiconductor LCD Telecom<br />
Digital Media Others<br />
Source: Company, HSBC estimates<br />
Source: Company, HSBC estimates<br />
Revenue OP OPM<br />
<strong>Memory</strong> is to dramatically impact profitability;<br />
we expect semiconductors to surge from 19% of<br />
consolidated operating profits in 2009 to a<br />
forecast 56% this year.<br />
<strong>Memory</strong> as % of consolidated operating profit<br />
The key driver of this growth is memory, which<br />
accounts for c70% of Semiconductor revenue.<br />
Consolidated revenue of Semiconductor division, 2009<br />
HDD<br />
11%<br />
100%<br />
50%<br />
Sy stems<br />
LSI<br />
15%<br />
DRAM<br />
40%<br />
0%<br />
2007 2008 2009e 2010f 2011f<br />
Semiconductor LCD Telecom<br />
Digital Media Others<br />
Other<br />
memory<br />
5%<br />
Source: HSBC estimates<br />
NAND<br />
29%<br />
Source: Company, HSBC estimates<br />
We forecast Samsung’s consolidated semicon<br />
revenue to rise c40% y-o-y this year, and for<br />
operating profit to rise c411% y-o-y.<br />
Detailed outlook<br />
We expect Samsung to report a 3Q10 operating<br />
profit similar to 2Q10. Revenue should rise 8% q-<br />
o-q to KRW41.1trn, as Telecoms and Digital<br />
Media profitability rebound. TFT LCD margins<br />
are set to contract on softer ASPs, whilst Samsung<br />
should deliver similar Semicon margins on<br />
process migration despite weaker ASPs. 2Q10<br />
was anomalous in that Semicon enjoyed demand<br />
from PC OEMs that brought forward shipments<br />
into 2Q (hence weak 3Q q-o-q growth) and firm<br />
pricing on weaker supply from smaller players.<br />
For Telecoms, strong smartphone sales during<br />
3Q10 drive a margin rebound. We expect<br />
86
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
relatively flat (-1% q-o-q) operating profit at<br />
KRW4.9trn, lower than consensus of KRW5.2trn.<br />
With decent 3Q10 expectations ‘in the bag’ as far<br />
as the market is concerned, the focus is very much<br />
on 4Q10-1Q11. If the traditionally strong 3Q can<br />
only deliver flat earnings q-o-q, and with ASPs<br />
still falling for the major earnings contributors –<br />
memory and TFT LCD – how low can Samsung’s<br />
4Q numbers go? Recent share price performance<br />
suggests that the market is expecting the worst.<br />
Our expectations are slightly more optimistic on<br />
the quarter, with 6% higher operating profit<br />
estimates than consensus. The difference is<br />
probably attributable to our more conservative<br />
outlook on TFT LCD and Digital Media. We<br />
expect operating profit to fall 10% q-o-q,<br />
compared to 20% for consensus. We concede there<br />
is no escape from falling memory ASPs even with<br />
accelerated process migration; we expect Semicon<br />
profits to fall 14% q-o-q. TFT-LCD declines are<br />
slightly lower at 12% q-o-q on cautious panel<br />
procurement. We are also taking a conservative<br />
view on memory ASPs in 1Q11, expecting DDR3<br />
contract prices to fall another 18%.<br />
DRAM<br />
Revenue, margin outlook<br />
For this year we forecast consolidated DRAM<br />
revenue to rise 82% in KRW terms, on a 70% bit<br />
growth and ASP increase of 18%. We expect the<br />
KRW to strengthen 12% compared to last year.<br />
DRAM revenue (KRWbn), operating margins<br />
20,000<br />
15,000<br />
10,000<br />
5,000<br />
-<br />
2007 2008 2009 2010f 2011f 2012f<br />
Source: Company, HSBC estimates<br />
Revenue (KRW bn)<br />
OPM<br />
50.0%<br />
40.0%<br />
30.0%<br />
20.0%<br />
10.0%<br />
0.0%<br />
-10.0%<br />
-20.0%<br />
We forecast 2011e consolidated DRAM revenue<br />
in KRW terms to decline 3% y-o-y; bit growth is<br />
to remain strong at 58% y-o-y but is offset by a<br />
31% blended ASP fall on greater industry-wide<br />
supply and our assumption of another 6%<br />
strengthening in the KRW.<br />
Key themes for Samsung’s business divisions<br />
Division Conclusions, positive catalysts % 2011e OP 2009/12e OP CAGR<br />
Semicon<br />
DRAM<br />
NAND<br />
Non-memory<br />
Telecoms<br />
LCD<br />
Digital Media<br />
Source: HSBC<br />
We expect Samsung’s process migration will support margins, but it also enables it to offer advanced low-power, high<br />
performance devices for rapid growth areas of mobile devices and servers. Samsung has great exposure to the NAND<br />
structural growth story and increasing exposure to mobile processor systems.<br />
We expect DRAM ASP declines to slow in 2Q11 as price elasticity kicks in to drive content/box. Impact of tablets on global<br />
DRAM demand c5% for 2011-12e. Samsung benefits from high-end product mix and process migration.<br />
Strong price elasticity drives greater penetration for smartphones, tablets and SSDs. Technology developments in NAND<br />
such as multi-level cells and toggle DDR, together with rising affordability will drive continued NAND take up.<br />
Samsung is making a strong push into system-on-chip (SoC) for mobile devices, as a high-growth, complementary<br />
business for memory and smartphone operations.<br />
Samsung’s tablet and leverages Samsung’s vertical integration and wireless distribution channels. Samsung is very well<br />
positioned to benefit from supplying 4G handsets going forward.<br />
Industry-wide ASPs could start rebounding in early 2011. We expect Samsung’s high-end product mix to dampen the<br />
impact of ASP declines and forecast Samsung to remain profitable through this ASP downturn.<br />
We are positive on Samsung’s early mover strategy in bringing advanced flat TVs to developed markets. We also believe<br />
other promising areas to watch include digital imaging, PC and printer operations.<br />
46% -11%<br />
28% 20%<br />
14% 4%<br />
12% 19%<br />
87
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Bit growth, ASP assumptions (%)<br />
100%<br />
50%<br />
0%<br />
-50%<br />
-100%<br />
2008 2009 2010f 2011f 2012f<br />
Bit growth<br />
ASP<br />
Source: Company, HSBC estimates<br />
We expect greater cost efficiencies mostly by<br />
process migration and improved production<br />
techniques to expand operating profitability to<br />
42.7% from 12.7% last year, reducing to 30.6% in<br />
2011e.<br />
Samsung has a strong product portfolio; as at<br />
2Q10, by revenue contribution to DRAM, we<br />
estimate c20% each in mDRAM and server<br />
DRAM, c10% in graphics DRAM and c5% in<br />
specialty DRAM. Samsung expects 150% y-o-y<br />
mDRAM bit growth this year as smartphone<br />
content/system rises to 4Gb this year (2009: 2Gb)<br />
and reach c6-8Gb next year.<br />
Technology – process migration<br />
Samsung’s DRAM process migration<br />
100%<br />
80%<br />
60%<br />
40%<br />
20%<br />
0%<br />
2007 2008 2009 2010f 2011f 2012f<br />
8xnm 6xnm 5xnm 4xnm 3xnm 2xnm<br />
Samsung is leading the market in DRAM process<br />
migration. By end-4Q09, it had migrated close to<br />
50% of DRAM output to 56nm and below. By the<br />
end of this year, it aims to achieve more than 80%<br />
of output on 56nm and below. By then, more than<br />
60% of the total is to be on the even more costeffective<br />
46nm process and below; in 2Q09, c35%<br />
of production was on the 46nm process.<br />
Wafer capacity (kwpy), average fab utilisation<br />
8,000<br />
120%<br />
6,000<br />
100%<br />
80%<br />
4,000<br />
60%<br />
2,000<br />
40%<br />
20%<br />
-<br />
0%<br />
2007 2008 2009 2010f 2011f 2012f<br />
Wafer capacity (kwpy) Average utilisation (%)<br />
Source: Company, HSBC estimates<br />
Samsung started mass production on the 35nm<br />
process in July this year. Additional capex cost of<br />
migrating to 35nm – for which mass production is<br />
slated for 2H10 – should not be high, as immersion<br />
lithography equipment remains the same.<br />
NAND<br />
Revenue, margin outlook<br />
We forecast consolidated 2010e NAND revenue to<br />
rise 27% in KRW terms, on a 79% bit growth and<br />
blended ASP decline of 21%. We expect the KRW<br />
to strengthen 12% compared to last year. We<br />
forecast 2011e consolidated revenue growth to slow<br />
to 18%; bit growth is to remain strong at 102% y-o-y<br />
but is offset by a 37% ASP fall and our assumption<br />
of another 6% strengthening in the KRW.<br />
Source: Company, HSBC estimates<br />
88
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
NAND revenue growth<br />
12,000<br />
30.0%<br />
10,000<br />
25.0%<br />
8,000<br />
20.0%<br />
6,000<br />
15.0%<br />
4,000<br />
10.0%<br />
2,000<br />
5.0%<br />
-<br />
0.0%<br />
2007 2008 2009 2010f 2011f 2012f<br />
Revenue (KRW bn) OPM<br />
Source: Company, HSBC estimates<br />
Technology – process migration<br />
By end-4Q09, more than 50% of Samsung’s<br />
NAND output was at 42nm and below. By 2Q10,<br />
about 50% of Samsung’s output was at 32nm. It<br />
aims to achieve over c80% of output at 32nm and<br />
below by end of this year, some of which would<br />
be using 3 bits per cell. Samsung started mass<br />
production of 2xnm 32Gb in April, to be joined<br />
by 64Gb 3-bit products in 2H10.<br />
Process migration<br />
100%<br />
80%<br />
60%<br />
40%<br />
20%<br />
0%<br />
2007 2008 2009 2010f 2011f 2012f<br />
8xnm 6xnm 5xnm 4xnm 3xnm 2xnm<br />
Source: Company, HSBC estimates<br />
Production capacity, capex<br />
In May 2010 Samsung announced a massive<br />
upward revision in capex guidance to KRW18trn,<br />
beating the previous capex high of cKRW13.5trn.<br />
in 2008. The amount includes KRW11trn for<br />
semiconductors (memory KRW9trn, systems LSI<br />
KRW2trn) plus another KRW5trn for LCD. This<br />
compares to KRW5.5trn for semis, KRW3trn for<br />
LCD in the firm’s original guidance. Capex is<br />
focused on capacity expansion in semiconductor<br />
Fabs 15 and 16 and its 8G fab for LCD. Samsung<br />
plans to spend another KRW8trn on R&D this year.<br />
In semiconductors, we expect wafer capacity,<br />
though a limited amount this year. Samsung<br />
earmarked KRW2.5trn for expansion of Fab 15 by<br />
another c30-40kwpm wafer capacity for 30nm<br />
DRAM. Overall, we estimate c40-50kwpm<br />
additional DRAM wafer capacity could be added<br />
by end-2010.<br />
The bulk of memory capex this year relates to<br />
process migration, including double patterning and<br />
other related equipment. If Samsung wants to<br />
accelerate process migration to 3xnm for DRAM<br />
and 2xnm for NAND to enlarge the product costcompetitiveness<br />
gap, ‘double patterning’ is required,<br />
which incurs more costs due to lower throughput.<br />
Wafer capacity increase will mostly come starting<br />
next year, as Samsung plans to spend a total of<br />
KRW12trn on Fab 16; whether this fab will<br />
produce DRAM or NAND has not been decided<br />
yet, but we believe NAND is likely. Fab 16 would<br />
come online next year and ultimately add<br />
c200kwpm capacity.<br />
89
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Samsung semicon capex (KRWbn)<br />
20,000<br />
15,000<br />
10,000<br />
5,000<br />
-<br />
2008 2009 2010f 2011f 2012f<br />
Source: Company, HSBC estimates<br />
An area worth watching going forward is the<br />
systems LSI business, which has great exposure to<br />
mobile application processors for smartphones<br />
and increasingly large exposure to image sensors.<br />
Samsung currently provides some foundry<br />
services, including to Apple and Qualcomm. We<br />
see momentum building in the race to deliver<br />
faster processing in mobile devices, where<br />
Samsung already has a significant presence.<br />
Samsung is investing KRW4.5trn to expand its<br />
Austin fab capacity by 40kwpm by end-2011e on<br />
the 45nm process to produce Systems LSI. We<br />
expect most of this capacity would service<br />
foundry operations for mobile-related devices.<br />
90
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Samsung Electronics: Consolidated P&L summary (K-IFRS)<br />
Revenue (KRWbn) 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e<br />
Semicon 26,808 37,461 40,639 44,382 8,205 9,530 9,869 9,857 9,377 9,836 10,555 10,871<br />
<strong>Memory</strong> 16,770 26,041 27,173 28,796 5,585 6,710 6,937 6,809 6,207 6,539 7,125 7,302<br />
LSI 4,400 5,052 6,160 7,511 1,160 1,235 1,296 1,361 1,429 1,501 1,576 1,654<br />
TFT-LCD 25,837 28,968 30,222 31,280 6,850 7,760 7,302 7,056 6,987 7,434 7,755 8,046<br />
Telecom 37,591 38,952 41,833 42,845 9,182 8,780 10,148 10,842 9,749 9,962 10,896 11,227<br />
Mobile 34,707 35,032 34,593 32,648 8,572 8,050 9,106 9,305 8,457 8,411 8,826 8,898<br />
Digital media 51,262 60,396 66,151 69,841 12,610 14,540 15,968 17,279 15,385 15,867 16,727 18,172<br />
VD 32,834 37,170 40,899 42,757 7,390 8,590 9,998 11,192 9,272 9,571 10,400 11,655<br />
Appliances 9,682 11,887 12,543 12,780 2,473 3,170 3,107 3,138 3,075 3,167 3,104 3,197<br />
Total 136,324 156,476 169,642 178,822 34,638 37,890 41,064 42,884 39,370 40,835 43,571 45,865<br />
Revenue, % of total 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e<br />
Semicon 20% 24% 24% 25% 24% 25% 24% 23% 24% 24% 24% 24%<br />
<strong>Memory</strong> 12% 17% 16% 16% 16% 18% 17% 16% 16% 16% 16% 16%<br />
LSI 3% 3% 4% 4% 3% 3% 3% 3% 4% 4% 4% 4%<br />
TFT-LCD 19% 19% 18% 17% 20% 20% 18% 16% 18% 18% 18% 18%<br />
Telecom 28% 25% 25% 24% 27% 23% 25% 25% 25% 24% 25% 24%<br />
Mobile 25% 22% 20% 18% 25% 21% 22% 22% 21% 21% 20% 19%<br />
Digital media 38% 39% 39% 39% 36% 38% 39% 40% 39% 39% 38% 40%<br />
VD 24% 24% 24% 24% 21% 23% 24% 26% 24% 23% 24% 25%<br />
Appliances 7% 8% 7% 7% 7% 8% 8% 7% 8% 8% 7% 7%<br />
Total 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100%<br />
Revenue, change (%) 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e<br />
Semicon 20% 40% 8% 9% 2% 16% 4% 0% -5% 5% 7% 3%<br />
<strong>Memory</strong> 20% 55% 4% 6% 11% 20% 3% -2% -9% 5% 9% 2%<br />
LSI 5% 15% 22% 22% na 6% 5% 5% 5% 5% 5% 5%<br />
TFT-LCD 20% 12% 4% 4% -6% 13% -6% -3% -1% 6% 4% 4%<br />
Telecom 18% 4% 7% 2% -10% -4% 16% 7% -10% 2% 9% 3%<br />
Mobile 20% 1% -1% -6% -8% -6% 13% 2% -9% -1% 5% 1%<br />
Digital media 15% 18% 10% 6% -19% 15% 10% 8% -11% 3% 5% 9%<br />
VD 13% 13% 10% 5% -28% 16% 16% 12% -17% 3% 9% 12%<br />
Appliances 13% 23% 6% 2% -3% 28% -2% 1% -2% 3% -2% 3%<br />
Total 12% 15% 8% 5% -12% 9% 8% 4% -8% 4% 7% 5%<br />
OP (KRWbn) 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e<br />
Semicon 2,059 10,511 8,328 8,249 1,995 2,903 3,013 2,599 1,781 2,015 2,284 2,248<br />
TFT-LCD 1,707 2,367 2,484 2,572 491 878 529 468 568 584 680 652<br />
Telecom 4,091 3,603 5,013 5,229 1,104 626 948 925 1,060 1,273 1,519 1,161<br />
Digital media 3,060 1,743 2,247 2,476 519 361 431 432 462 555 669 561<br />
Total 10,925 18,785 18,072 18,536 4,406 5,014 4,942 4,424 3,870 4,427 5,153 4,622<br />
OPM (%) 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e<br />
Semicon 7.7% 28.1% 20.5% 18.6% 24.3% 30.5% 30.5% 26.4% 19.0% 20.5% 21.6% 20.7%<br />
TFT-LCD 6.6% 8.2% 8.2% 8.2% 7.2% 11.3% 7.2% 6.6% 8.1% 7.9% 8.8% 8.1%<br />
Telecom 10.9% 9.3% 12.0% 12.2% 12.0% 7.1% 9.3% 8.5% 10.9% 12.8% 13.9% 10.3%<br />
Digital media 6.0% 2.9% 3.4% 3.5% 4.1% 2.5% 2.7% 2.5% 3.0% 3.5% 4.0% 3.1%<br />
Total 8.0% 12.0% 10.7% 10.4% 12.7% 13.2% 12.0% 10.3% 9.8% 10.8% 11.8% 10.1%<br />
OP, % of total 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e<br />
Semicon 19% 56% 46% 45% 45% 58% 61% 59% 46% 46% 44% 49%<br />
TFT-LCD 16% 13% 14% 14% 11% 18% 11% 11% 15% 13% 13% 14%<br />
Telecom 37% 19% 28% 28% 25% 12% 19% 21% 27% 29% 29% 25%<br />
Digital media 28% 9% 12% 13% 12% 7% 9% 10% 12% 13% 13% 12%<br />
Total 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100%<br />
OP, change (%) 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e<br />
Semicon na 411% -21% -1% 50% 46% 4% -14% -31% 13% 13% -2%<br />
TFT-LCD na 39% 5% 4% -8% 79% -40% -12% 21% 3% 17% -4%<br />
Telecom na -12% 39% 4% 6% -43% 51% -2% 15% 20% 19% -24%<br />
Digital media na -43% 29% 10% 11% -30% 20% 0% 7% 20% 20% -16%<br />
Total na 72% -4% 3% 28% 14% -1% -10% -13% 14% 16% -10%<br />
Source: Company data, HSBC estimates<br />
91
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Financials & valuation: Samsung Electronics<br />
Overweight<br />
Financial statements<br />
Year to 12/2009a 12/2010e 12/2011e 12/2012e<br />
Profit & loss summary (KRWb)<br />
Revenue 136,324 156,476 169,642 178,822<br />
EBITDA 22,085 31,263 32,792 34,066<br />
Depreciation & amortisation -11,159 -12,478 -14,720 -15,530<br />
Operating profit/EBIT 10,925 18,785 18,072 18,536<br />
Net interest -205 -323 -460 -528<br />
PBT 12,192 20,413 20,103 19,845<br />
HSBC PBT 12,192 20,413 20,103 19,845<br />
Taxation -2,431 -3,992 -3,923 -3,069<br />
Net profit 9,761 16,420 16,180 16,777<br />
HSBC net profit 9,761 16,420 16,180 16,777<br />
Cash flow summary (KRWb)<br />
Cash flow from operations 20,352 21,212 28,279 31,238<br />
Capex -8,520 -19,131 -22,008 -19,214<br />
Cash flow from investment -10,555 -24,199 -27,947 -33,018<br />
Dividends -871 -947 -1,277 -2,043<br />
Change in net debt -10,104 4,264 1,710 3,823<br />
FCF equity 12,276 2,850 6,151 11,902<br />
Balance sheet summary (KRWb)<br />
Intangible fixed assets 891 941 991 1,042<br />
Tangible fixed assets 39,816 46,419 53,656 57,290<br />
Current assets 62,069 58,683 64,457 74,012<br />
Cash & others 21,570 14,824 18,669 27,069<br />
Total assets 118,281 126,189 142,272 159,090<br />
Operating liabilities 30,816 28,588 30,534 32,618<br />
Gross debt 14,411 9,403 9,403 9,403<br />
Net debt -5,055 -791 919 4,742<br />
Shareholders funds 73,054 88,198 102,336 117,070<br />
Invested capital 50,457 62,631 69,902 72,658<br />
Ratio, growth and per share analysis<br />
Year to 12/2009a 12/2010e 12/2011e 12/2012e<br />
Y-o-y % change<br />
Revenue 12.4 14.8 8.4 5.4<br />
EBITDA 36.9 41.6 4.9 3.9<br />
Operating profit 81.1 71.9 -3.8 2.6<br />
PBT 85.3 67.4 -1.5 -1.3<br />
HSBC EPS 53.6 68.2 -1.5 3.7<br />
Valuation data<br />
Year to 12/2009a 12/2010e 12/2011e 12/2012e<br />
EV/sales 0.8 0.7 0.6 0.6<br />
EV/EBITDA 4.8 3.4 3.3 3.2<br />
EV/IC 2.1 1.7 1.5 1.5<br />
PE* 13.8 8.2 8.3 8.0<br />
P/NAV 1.8 1.5 1.3 1.2<br />
FCF yield (%) 11.1 2.6 5.8 11.4<br />
Dividend yield (%) 1.0 1.5 1.5 1.5<br />
Note: * = Based on HSBC EPS (fully diluted)<br />
Issuer information<br />
Share<br />
price<br />
(KRW) 793,000 Target price (KRW) 1,139,0<br />
00<br />
Potent’l tot rtn (%) 43.6<br />
Reuters (Equity) 005930.KS Bloomberg (Equity) 005930 KS<br />
Market cap (USDm) 103,351 Market cap (KRWb) 116,808<br />
Free float (%) 69 Enterprise value (KRWb) 10704<br />
Country Korea Sector Semiconductors<br />
Analyst Nam Park Contact 852 2996 6591<br />
Price relative<br />
941588<br />
841588<br />
741588<br />
641588<br />
541588<br />
441588<br />
341588<br />
Source: HSBC<br />
2008 2009 2010 2011<br />
Samsung Electronics Rel to KOSPI INDEX<br />
Note: price at close of 06 Oct 2010<br />
941588<br />
841588<br />
741588<br />
641588<br />
541588<br />
441588<br />
341588<br />
Ratios (%)<br />
Revenue/IC (x) 2.6 2.8 2.6 2.5<br />
ROIC 17.1 27.1 22.3 22.4<br />
ROE 14.4 20.4 17.0 15.3<br />
ROA 9.2 13.9 12.5 11.5<br />
EBITDA margin 16.2 20.0 19.3 19.1<br />
Operating profit margin 8.0 12.0 10.7 10.4<br />
EBITDA/net interest (x) 107.9 96.9 71.3 64.5<br />
Net debt/equity -6.9 -0.9 0.9 4.1<br />
Net debt/EBITDA (x) -0.2 0.0 0.0 0.1<br />
CF from operations/net debt 3076.4 658.7<br />
Per share data (KRW)<br />
EPS Rep (fully diluted) 57,370 96,516 95,105 98,610<br />
HSBC EPS (fully diluted) 57,370 96,516 95,105 98,610<br />
DPS 7,500 12,000 12,000 12,000<br />
NAV 429,398 518,410 601,508 688,111<br />
92
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Winbond (2344)<br />
Defensive play in a steep commodity price correction environment<br />
with a successfully restructured, diversified niche portfolio<br />
Technology migration and flexible product mix to maintain margin<br />
Cut TP to TWD10.8 (from TWD12.5) to reflect higher macro risk<br />
using target PB of 1x; maintain OW(V)<br />
Investment summary<br />
Winbond’s memory portfolio restructuring story<br />
has ended on a high note after it successfully<br />
reduced its commodity DRAM portion to less<br />
than 2% (from c30%) of its portfolio in 2Q10.<br />
Financial results in 1H10 were good, thanks to<br />
tight supply of NOR on the back of the industry’s<br />
limited capacity. As the demand/supply in NOR<br />
has slowly approached a balance in 3Q, and while<br />
DRAM prices seem to have softened starting from<br />
3Q, we believe Winbond’s superior portfolio is<br />
the most defensive among our Taiwanese DRAM<br />
coverage amid the memory price correction<br />
outlook in 2H10, given the stock’s current<br />
attractive valuation.<br />
Fundamentals<br />
A superior non-PC memory portfolio<br />
Winbond’s 40% revenue is contributed by<br />
specialty DRAM. The remaining contribution<br />
came from NOR (28%), Mobile DRAM (18%),<br />
graphics DRAM (12%) and commodity DRAM.<br />
(2%). A year ago, around 30% of Winbond’s<br />
revenue was contributed by commodity DRAM, a<br />
portfolio that is highly susceptible to DRAM price<br />
volatility hence impacting margin stability.<br />
More first-tier clients<br />
Most of Winbond’s clients are 3-6 month<br />
contract-based. This enables Winbond to have<br />
relatively stable margins while having high entry<br />
barriers as it takes time and effort to win client<br />
relationships. Winbond has demonstrated its<br />
superior quality by established first-tier clients in<br />
areas like specialty DRAM, mobile DRAM and<br />
NOR flash within the last few years.<br />
Flexible and diversified portfolio<br />
Winbond has the flexibility in adjusting its<br />
memory portfolio: as their product is not<br />
concentrated on a single memory type, capacity<br />
reallocation can be efficiently adjusted. More<br />
importantly, its memory production also supplies<br />
various diversified applications: Within specialty<br />
DRAM, HDD represents 36% of the total,<br />
followed by TV (17%) and networking (14%).<br />
This enables Winbond to focus on or divert to<br />
more lucrative applications when demand for a<br />
certain application is affected by seasonality.<br />
Specialty DRAM has strong growth of 27% q-o-q<br />
growth in 2Q, underpinned by top applications<br />
like HDD and TV growth.<br />
93
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Technology migration<br />
Winbond has successfully migrated its NOR<br />
production from 130nm to 90nm. In 3Q10, all 10k<br />
NOR capacity are in 90nm. For DRAM,<br />
Winbond’s target is to migrate its 16k capacity to<br />
65nm.<br />
Ongoing migrations: Winbond has scheduled to<br />
further migrate 1) NOR capacity to 58nm by<br />
2H11 and 2) DRAM capacity to 46nm also by<br />
2H11.<br />
ASP and bit growth assumptions<br />
80.0%<br />
60.0%<br />
40.0%<br />
20.0%<br />
0.0%<br />
-20.0%<br />
-40.0%<br />
2008 2009 2010f 2011f 2012f<br />
Bit growth ASP<br />
Source: Company data, HSBC estimates<br />
Forecast revision<br />
3Q10 revenue and margins should still be holding<br />
up well, thanks to the long-term contractual<br />
relationships which Winbond has built up with<br />
customers, together with strong demand from<br />
specialty DRAM and mobile demand, even<br />
though NOR supply/demand has come to a<br />
relatively more balanced situation. Winbond will<br />
likely post a gain of TWD0.8bn from the sale of<br />
its Nuvoton stake on the latter’s listing in late<br />
September. Following the sale, Winbond still<br />
holds 62.5% of Nuvoton.<br />
Revenue growth<br />
40,000<br />
30,000<br />
20,000<br />
10,000<br />
-<br />
2007 2008 2009 2010f 2011f 2012f<br />
Source: Company data, HSBC estimates<br />
Revenue (NTD mn)<br />
OPM<br />
20%<br />
10%<br />
0%<br />
-10%<br />
-20%<br />
-30%<br />
-40%<br />
-50%<br />
We lowered Winbond’s bit growth for 2010 to<br />
12% from 16% and revised down blended ASP<br />
for 2011 as memory price softening has come<br />
earlier than we expected. We now forecast<br />
Winbond to achieve 2010-11e revenue growth of<br />
67% and 4% respectively, revised down by 2%<br />
and 4% respectively. We also lowered Winbond’s<br />
2010-11e OP margin to 10.9% and 7.2%<br />
respectively (previously 12.6% and 9.3%)<br />
respectively on lower top-line growth. We are 2%<br />
and 4% below consensus in terms of 2010-11e<br />
revenue.<br />
Forecast revision for Winbond<br />
__Previous___ __ Revised ___ __ Changes___<br />
TWD (m) 2010e 2011e 2010e 2011e 2010e 2011e<br />
Revenue 33,219 35,336 32,596 33,796 -2% -4%<br />
OP 4,172 3,289 3,561 2,431 -15% -26%<br />
Net profit 4,063 2,738 3,532 2,011 -13% -27%<br />
OPM 12.6% 9.3% 10.9% 7.2%<br />
NPM 12.2% 7.7% 10.8% 6.0%<br />
Source: HSBC<br />
Product mix<br />
We expect Winbond to maintain a relatively<br />
balanced portfolio among NOR flash, specialty<br />
DRAM and mobile DRAM in 2010, after<br />
successfully reduced commodity DRAM’s<br />
proportion. Despite NOR flash has come to a<br />
more balanced demand supply situation, we<br />
believe Winbond’s NOR flash bit growth will<br />
continue, underpinned by expanded capacity<br />
94
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
(10k/m by 4Q10) and 90nm full production<br />
capacity, offsetting possible NOR ASP decline.<br />
We’ve modelled NOR ASP to come down by<br />
18% in 4Q and another 10% decline in 1Q11. We<br />
expect NOR flash to represent close to 26% of<br />
Winbond’s revenue by end of 4Q10. We expect<br />
Winbond’s mobile DRAM and specialised<br />
DRAM revenue combined to represent 57% of its<br />
total revenue in 2011, up from 55% this year.<br />
Revenue mix, 2007-11e<br />
100%<br />
80%<br />
60%<br />
40%<br />
20%<br />
0%<br />
2007 2008 2009 2010f 2011f<br />
NOR Flash<br />
Specialty DRAM<br />
Mobile DRAM<br />
Graphic DRAM<br />
Commodity DRAM Others<br />
Source: Company data, HSBC estimates<br />
Revenue growth in 2011 seems minimal, but this<br />
is after a very strong growth year in 2010. We<br />
forecast 2010-11 ASP growth of 53% and -30%<br />
respectively, and 2010-11e bit growth for 12%<br />
and 45%. We expect bit growth to be largely<br />
driven by process migration.<br />
Valuation, ratings<br />
We derive a new 12-month target price of<br />
TWD10.8 (previously TWD12.5) for Winbond by<br />
applying a target FY11e PB multiple of 1.0x<br />
(rolled over to FY11 based on 9.9% ROE and<br />
9.5% COE). Previously, we use a multiple of<br />
1.2x). The lower PB multiple is supported by our<br />
lower forecast in OP and lower ROE, given<br />
Winbond’s NOR growth has come to a<br />
normalized demand and supply situation, in<br />
addition to a softening memory price outlook.<br />
(prior to the capacity glut starting in 2007). With a<br />
less optimistic macro environment, we expect the<br />
stock to reach the lower of its historical PB band<br />
in the next 12 months.<br />
HSBC vs consensus<br />
HSBC 2010e 2011e 2012e<br />
Revenue 32,596 33,796 37,642<br />
EBITDA 14,594 13,729 15,477<br />
EBIT 3,561 2,431 3,662<br />
Net income 3,532 2,011 3,201<br />
EPS (KRW) 1.0 0.6 0.9<br />
EBITDA margin 44.8% 40.6% 41.1%<br />
EBIT margin 10.9% 7.2% 9.7%<br />
Net margin 10.8% 6.0% 8.5%<br />
HSBC vs<br />
2010e 2011e 2012e<br />
consensus<br />
Revenue -2.0% -4.0% -0.6%<br />
EBITDA -5.0% -11.5% 9.8%<br />
EBIT -14.6% -26.1% 1.4%<br />
Net income -15.2% -53.1% -12.3%<br />
EPS (KRW) 9.7% -54.7% 7.6%<br />
Consensus 2010e 2011e 2012e<br />
Revenue 33,255 35,209 37,867<br />
EBITDA 15,365 15,518 14,097<br />
EBIT 4,172 3,289 3,610<br />
Net income 4,165 4,292 3,650<br />
EPS (KRW) 0.9 1.2 0.8<br />
EBITDA margin 46.2% 44.1% 37.2%<br />
EBIT margin 12.5% 9.3% 9.5%<br />
Net margin 12.5% 12.2% 9.6%<br />
Source: HSBC, I/B/E/S consensus<br />
Under our research model, for Taiwan stocks with<br />
a volatility indicator, the Neutral band is 10ppt<br />
above and below the hurdle rate of 9.5% (risk-free<br />
rate of 4% plus risk premium of 5.5%). This<br />
translates into a Neutral band of -0.5% to +19.5%<br />
around the current share price. Our 12-month<br />
target price of TWD10.8 suggests a potential<br />
return of 34.5%, which is in the Overweight band;<br />
thus, we maintain our Overweight (V) rating.<br />
The stock has historically traded at a PB of 0.2-<br />
1.2x, with a high ROE of 2-11% when profitable<br />
95
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Quarterly assumptions for Winbond<br />
TWDm 1Q10 2Q10 3Q10e 4Q10e 2010e 2011e<br />
Revenue 7,013 8,531 8,772 8,280 32,596 33,796<br />
OP 484 1,290 987 800 3,561 2,431<br />
Net profit 384 1,236 1,361 551 3,532 2,011<br />
OPM 6.9% 15.1% 11.3% 9.7% 10.9% 7.2%<br />
NPM 5.5% 14.5% 15.5% 6.7% 10.8% 6.0%<br />
Blended ASP 16% 10% -6% -14% 53% -30%<br />
Bit growth (%) -6% 11% 9% 9% 12% 45%<br />
Source: Company data, HSBC estimates<br />
Forward PB band<br />
20<br />
Price (TWD)<br />
15<br />
1.2x<br />
10<br />
0.9x<br />
5<br />
0.6x<br />
0.3x<br />
-<br />
Jan-<br />
05<br />
Jan-<br />
06<br />
Jan-<br />
07<br />
Jan-<br />
08<br />
Jan-<br />
09<br />
Jan-<br />
10<br />
Source: HSBC<br />
Catalysts and risks<br />
<strong>Memory</strong> industry-specific catalysts:<br />
Evidence of strong unit shipments and rising<br />
memory content of smartphones, LCD TV,<br />
HDD, PC<br />
Faster than expected ASP erosion due to<br />
greater capacity coming online and/or slower<br />
than expected demand from applications like<br />
TV, HDD, LCD TV<br />
More consolidation of players, such as<br />
Micron’s acquisition of Numonyx,<br />
strengthening competition.<br />
Greater ability of competitors to accelerate<br />
process migration, wafer capacity expansion<br />
through improving financials.<br />
Delayed global economic recovery.<br />
Company-specific risks include:<br />
Delay in process migration.<br />
Delay in graphic DRAM execution<br />
Evidence of rising memory content of NOR<br />
flash required in integrated chip platform in<br />
PC<br />
Company-specific catalysts include:<br />
Technology migration to reduce cost from 1)<br />
moving NOR flash production to 58nm by<br />
2H11, 2) 65nm DRAM capacity to 16kwpm;<br />
Capacity ramp up to 10k from 7k for NOR<br />
Further DRAM migration to 46nm in 2H11.<br />
<strong>Memory</strong> industry-specific risks:<br />
96
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Financials & valuation: Winbond Electronics<br />
Overweight (V)<br />
Financial statements<br />
Year to 12/2009a 12/2010e 12/2011e 12/2012e<br />
Profit & loss summary (TWDm)<br />
Revenue 19,533 32,596 33,796 37,642<br />
EBITDA 4,246 14,594 13,729 15,477<br />
Depreciation & amortisation -10,241 -11,033 -11,297 -11,815<br />
Operating profit/EBIT -5,995 3,561 2,431 3,662<br />
Net interest -640 -629 -390 -258<br />
PBT -8,613 3,532 2,366 3,904<br />
HSBC PBT -8,613 3,532 2,366 3,904<br />
Taxation 0 0 -355 -703<br />
Net profit -8,613 3,532 2,011 3,201<br />
HSBC net profit -8,613 3,532 2,011 3,201<br />
Cash flow summary (TWDm)<br />
Cash flow from operations 4,768 11,837 13,143 14,593<br />
Capex -4,200 -8,000 -7,500 -7,500<br />
Cash flow from investment -15,535 -746 -7,540 -7,540<br />
Dividends 0 0 0 0<br />
Change in net debt -2,453 -7,744 -5,573 -6,850<br />
FCF equity 568 3,837 5,288 6,390<br />
Balance sheet summary (TWDm)<br />
Intangible fixed assets 1,531 1,106 813 517<br />
Tangible fixed assets 45,866 43,297 39,833 35,854<br />
Current assets 12,834 29,317 35,329 43,584<br />
Cash & others 2,859 14,570 20,143 26,994<br />
Total assets 65,163 78,653 80,907 84,887<br />
Operating liabilities 7,250 13,223 13,466 14,244<br />
Gross debt 24,058 28,025 28,025 28,025<br />
Net debt 21,199 13,455 7,882 1,032<br />
Shareholders funds 33,854 37,405 39,416 42,617<br />
Invested capital 50,121 45,928 42,366 38,717<br />
Ratio, growth and per share analysis<br />
Year to 12/2009a 12/2010e 12/2011e 12/2012e<br />
Y-o-y % change<br />
Revenue -10.5 66.9 3.7 11.4<br />
EBITDA 36.0 243.7 -5.9 12.7<br />
Operating profit -31.7 50.6<br />
PBT -33.0 65.0<br />
HSBC EPS -43.1 59.2<br />
Ratios (%)<br />
Revenue/IC (x) 0.4 0.7 0.8 0.9<br />
ROIC -10.7 8.4 5.3 8.1<br />
ROE -23.9 9.9 5.2 7.8<br />
ROA -11.5 5.8 3.3 4.5<br />
EBITDA margin 21.7 44.8 40.6 41.1<br />
Operating profit margin -30.7 10.9 7.2 9.7<br />
EBITDA/net interest (x) 6.6 23.2 35.2 60.1<br />
Net debt/equity 62.6 36.0 20.0 2.4<br />
Net debt/EBITDA (x) 5.0 0.9 0.6 0.1<br />
CF from operations/net debt 22.5 88.0 166.7 1414.7<br />
Per share data (TWD)<br />
EPS Rep (fully diluted) -2.36 0.97 0.55 0.88<br />
HSBC EPS (fully diluted) -2.36 0.97 0.55 0.88<br />
DPS 0.00 0.00 0.00 0.00<br />
Book value 9.28 10.25 10.80 11.68<br />
Valuation data<br />
Year to 12/2009a 12/2010e 12/2011e 12/2012e<br />
EV/sales 2.3 1.2 1.0 0.7<br />
EV/EBITDA 10.7 2.6 2.4 1.6<br />
EV/IC 0.9 0.8 0.8 0.7<br />
PE* 8.3 14.6 9.2<br />
P/Book value 0.9 0.8 0.7 0.7<br />
FCF yield (%) 2.3 15.7 21.6 26.2<br />
Dividend yield (%) 0.0 0.0 0.0 0.0<br />
Note: * = Based on HSBC EPS (fully diluted)<br />
Issuer information<br />
Share price (TWD) 8.03 Target price (TWD) 10.80 Potent’l return (%) 34.5<br />
Reuters (Equity) 2344.TW Bloomberg (Equity) 2344 TT<br />
Market cap (USDm) 945 Market cap (TWDm) 29,362<br />
Free float (%) 74 Enterprise value (TWDm 37885<br />
)<br />
Country Taiwan Sector Semiconductors<br />
Analyst Carolyn Poon Contact +852 2996 6586<br />
Price relative<br />
13<br />
11<br />
9<br />
7<br />
5<br />
3<br />
1<br />
2008 2009 2010 2011<br />
Winbond Electronics Rel to TAIWAN WEIGHTED INDEX<br />
Source: HSBC<br />
Note: price at close of 06 Oct 2010<br />
13<br />
11<br />
9<br />
7<br />
5<br />
3<br />
1<br />
97
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Notes<br />
98
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Disclosure appendix<br />
Analyst Certification<br />
The following analyst(s), economist(s), and/or strategist(s) who is(are) primarily responsible for this report, certifies(y) that the<br />
opinion(s) on the subject security(ies) or issuer(s) and/or any other views or forecasts expressed herein accurately reflect their<br />
personal view(s) and that no part of their compensation was, is or will be directly or indirectly related to the specific<br />
recommendation(s) or views contained in this research report: Nam Park and Carolyn Poon<br />
Important disclosures<br />
Stock ratings and basis for financial analysis<br />
HSBC believes that investors utilise various disciplines and investment horizons when making investment decisions, which<br />
depend largely on individual circumstances such as the investor's existing holdings, risk tolerance and other considerations.<br />
Given these differences, HSBC has two principal aims in its equity research: 1) to identify long-term investment opportunities<br />
based on particular themes or ideas that may affect the future earnings or cash flows of companies on a 12 month time horizon;<br />
and 2) from time to time to identify short-term investment opportunities that are derived from fundamental, quantitative,<br />
technical or event-driven techniques on a 0-3 month time horizon and which may differ from our long-term investment rating.<br />
HSBC has assigned ratings for its long-term investment opportunities as described below.<br />
This report addresses only the long-term investment opportunities of the companies referred to in the report. As and when<br />
HSBC publishes a short-term trading idea the stocks to which these relate are identified on the website at<br />
www.hsbcnet.com/research. Details of these short-term investment opportunities can be found under the Reports section of this<br />
website.<br />
HSBC believes an investor's decision to buy or sell a stock should depend on individual circumstances such as the investor's<br />
existing holdings and other considerations. Different securities firms use a variety of ratings terms as well as different rating<br />
systems to describe their recommendations. Investors should carefully read the definitions of the ratings used in each research<br />
report. In addition, because research reports contain more complete information concerning the analysts' views, investors<br />
should carefully read the entire research report and should not infer its contents from the rating. In any case, ratings should not<br />
be used or relied on in isolation as investment advice.<br />
Rating definitions for long-term investment opportunities<br />
Stock ratings<br />
HSBC assigns ratings to its stocks in this sector on the following basis:<br />
For each stock we set a required rate of return calculated from the risk free rate for that stock's domestic, or as appropriate,<br />
regional market and the relevant equity risk premium established by our strategy team. The price target for a stock represents<br />
the value the analyst expects the stock to reach over our performance horizon. The performance horizon is 12 months. For a<br />
stock to be classified as Overweight, the implied return must exceed the required return by at least 5 percentage points over the<br />
next 12 months (or 10 percentage points for a stock classified as Volatile*). For a stock to be classified as Underweight, the<br />
stock must be expected to underperform its required return by at least 5 percentage points over the next 12 months (or 10<br />
percentage points for a stock classified as Volatile*). Stocks between these bands are classified as Neutral.<br />
Our ratings are re-calibrated against these bands at the time of any 'material change' (initiation of coverage, change of volatility<br />
status or change in price target). Notwithstanding this, and although ratings are subject to ongoing management review,<br />
expected returns will be permitted to move outside the bands as a result of normal share price fluctuations without necessarily<br />
triggering a rating change.<br />
*A stock will be classified as volatile if its historical volatility has exceeded 40%, if the stock has been listed for less than 12<br />
months (unless it is in an industry or sector where volatility is low) or if the analyst expects significant volatility. However,<br />
99
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
stocks which we do not consider volatile may in fact also behave in such a way. Historical volatility is defined as the past<br />
month's average of the daily 365-day moving average volatilities. In order to avoid misleadingly frequent changes in rating,<br />
however, volatility has to move 2.5 percentage points past the 40% benchmark in either direction for a stock's status to change.<br />
Rating distribution for long-term investment opportunities<br />
As of 08 October 2010, the distribution of all ratings published is as follows:<br />
Overweight (Buy) 50% (21% of these provided with Investment Banking Services)<br />
Neutral (Hold) 36% (18% of these provided with Investment Banking Services)<br />
Underweight (Sell) 14% (19% of these provided with Investment Banking Services)<br />
Information regarding company share price performance and history of HSBC ratings and price targets in respect of its longterm<br />
investment opportunities for the companies the subject of this report,is available from www.hsbcnet.com/research.<br />
HSBC & Analyst disclosures<br />
Disclosure checklist<br />
Company Ticker Recent price Price Date Disclosure<br />
HYNIX SEMICONDUCTOR INC 000660.KS 23200.00 07-Oct-2010 11<br />
INOTERA MEMORIES INC. 3474.TW 16.45 07-Oct-2010 1, 2, 5<br />
SAMSUNG ELECTRONICS 005930.KS 770000.00 07-Oct-2010 6, 11<br />
Source: HSBC<br />
1 HSBC* has managed or co-managed a public offering of securities for this company within the past 12 months.<br />
2 HSBC expects to receive or intends to seek compensation for investment banking services from this company in the next<br />
3 months.<br />
3 At the time of publication of this report, HSBC Securities (USA) Inc. is a Market Maker in securities issued by this<br />
company.<br />
4 As of 31 August 2010 HSBC beneficially owned 1% or more of a class of common equity securities of this company.<br />
5 As of 31 August 2010, this company was a client of HSBC or had during the preceding 12 month period been a client of<br />
and/or paid compensation to HSBC in respect of investment banking services.<br />
6 As of 31 August 2010, this company was a client of HSBC or had during the preceding 12 month period been a client of<br />
and/or paid compensation to HSBC in respect of non-investment banking-securities related services.<br />
7 As of 31 August 2010, this company was a client of HSBC or had during the preceding 12 month period been a client of<br />
and/or paid compensation to HSBC in respect of non-securities services.<br />
8 A covering analyst/s has received compensation from this company in the past 12 months.<br />
9 A covering analyst/s or a member of his/her household has a financial interest in the securities of this company, as<br />
detailed below.<br />
10 A covering analyst/s or a member of his/her household is an officer, director or supervisory board member of this<br />
company, as detailed below.<br />
11 At the time of publication of this report, HSBC is a non-US Market Maker in securities issued by this company and/or in<br />
securities in respect of this company<br />
Analysts, economists, and strategists are paid in part by reference to the profitability of HSBC which includes investment<br />
banking revenues.<br />
For disclosures in respect of any company mentioned in this report, please see the most recently published report on that<br />
company available at www.hsbcnet.com/research.<br />
* HSBC Legal Entities are listed in the Disclaimer below.<br />
100
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Additional disclosures<br />
1 This report is dated as at 08 October 2010.<br />
2 All market data included in this report are dated as at close 06 October 2010, unless otherwise indicated in the report.<br />
3 HSBC has procedures in place to identify and manage any potential conflicts of interest that arise in connection with its<br />
Research business. HSBC's analysts and its other staff who are involved in the preparation and dissemination of Research<br />
operate and have a management reporting line independent of HSBC's Investment Banking business. Information Barrier<br />
procedures are in place between the Investment Banking and Research businesses to ensure that any confidential and/or<br />
price sensitive information is handled in an appropriate manner.<br />
101
TMT<br />
<strong>Memory</strong> Devices Technology<br />
8 October 2010<br />
abc<br />
Disclaimer<br />
* Legal entities as at 31 January 2010<br />
'UAE' HSBC Bank Middle East Limited, Dubai; 'HK' The Hongkong and Shanghai Banking<br />
Corporation Limited, Hong Kong; 'TW' HSBC Securities (Taiwan) Corporation Limited; 'CA'<br />
HSBC Securities (Canada) Inc, Toronto; HSBC Bank, Paris branch; HSBC France; 'DE' HSBC<br />
Trinkaus & Burkhardt AG, Dusseldorf; 000 HSBC Bank (RR), Moscow; 'IN' HSBC Securities<br />
and Capital Markets (India) Private Limited, Mumbai; 'JP' HSBC Securities (Japan) Limited,<br />
Tokyo; 'EG' HSBC Securities Egypt S.A.E., Cairo; 'CN' HSBC Investment Bank <strong>Asia</strong> Limited,<br />
Beijing Representative Office; The Hongkong and Shanghai Banking Corporation Limited,<br />
Singapore branch; The Hongkong and Shanghai Banking Corporation Limited, Seoul Securities<br />
Branch; The Hongkong and Shanghai Banking Corporation Limited, Seoul Branch; HSBC<br />
Securities (South Africa) (Pty) Ltd, Johannesburg; 'GR' HSBC Pantelakis Securities S.A., Athens;<br />
HSBC Bank plc, London, Madrid, Milan, Stockholm, Tel Aviv, 'US' HSBC Securities (USA) Inc,<br />
New York; HSBC Yatirim Menkul Degerler A.S., Istanbul; HSBC México, S.A., Institución de<br />
Banca Múltiple, Grupo Financiero HSBC, HSBC Bank Brasil S.A. - Banco Múltiplo, HSBC Bank<br />
Australia Limited, HSBC Bank Argentina S.A., HSBC Saudi Arabia Limited.<br />
Issuer of report<br />
The Hongkong and Shanghai<br />
Banking Corporation Limited<br />
Level 19, 1 Queen’s Road Central<br />
Hong Kong SAR<br />
Telephone: +852 2843 9111<br />
Telex: 75100 CAPEL HX<br />
Fax: +852 2596 0200<br />
Website: www.research.hsbc.com<br />
This document has been issued by The Hongkong and Shanghai Banking Corporation Limited (“HSBC”) in the conduct of its Hong Kong<br />
regulated business for the information of its institutional and professional customers; it is not intended for and should not be distributed to retail<br />
customers in Hong Kong. The Hongkong and Shanghai Banking Corporation Limited is regulated by the Securities and Futures Commission. All<br />
enquires by recipients in Hong Kong must be directed to your HSBC contact in Hong Kong. If it is received by a customer of an affiliate of<br />
HSBC, its provision to the recipient is subject to the terms of business in place between the recipient and such affiliate. This document is not and<br />
should not be construed as an offer to sell or the solicitation of an offer to purchase or subscribe for any investment. HSBC has based this<br />
document on information obtained from sources it believes to be reliable but which it has not independently verified; HSBC makes no guarantee,<br />
representation or warranty and accepts no responsibility or liability as to its accuracy or completeness. Expressions of opinion are those of the<br />
Research Division of HSBC only and are subject to change without notice. HSBC and its affiliates and/or their officers, directors and employees<br />
may have positions in any securities mentioned in this document (or in any related investment) and may from time to time add to or dispose of<br />
any such securities (or investment). HSBC and its affiliates may act as market maker or have assumed an underwriting commitment in the<br />
securities of companies discussed in this document (or in related investments), may sell them to or buy them from customers on a principal basis<br />
and may also perform or seek to perform investment banking or underwriting services for or relating to those companies.<br />
HSBC Securities (USA) Inc. accepts responsibility for the content of this research report prepared by its non-US foreign affiliate. All U.S.<br />
persons receiving and/or accessing this report and wishing to effect transactions in any security discussed herein should do so with HSBC<br />
Securities (USA) Inc. in the United States and not with its non-US foreign affiliate, the issuer of this report.<br />
In the UK this report may only be distributed to persons of a kind described in Article 19(5) of the Financial Services and Markets Act 2000<br />
(Financial Promotion) Order 2001. The protections afforded by the UK regulatory regime are available only to those dealing with a<br />
representative of HSBC Bank plc in the UK. In Singapore, this publication is distributed by The Hongkong and Shanghai Banking<br />
Corporation Limited, Singapore Branch for the general information of institutional investors or other persons specified in Sections 274 and<br />
304 of the Securities and Futures Act (Chapter 289) (“SFA”) and accredited investors and other persons in accordance with the conditions<br />
specified in Sections 275 and 305 of the SFA. This publication is not a prospectus as defined in the SFA. It may not be further distributed in<br />
whole or in part for any purpose. The Hongkong and Shanghai Banking Corporation Limited Singapore Branch is regulated by the Monetary<br />
Authority of Singapore. Recipients in Singapore should contact a "Hongkong and Shanghai Banking Corporation Limited, Singapore Branch"<br />
representative in respect of any matters arising from, or in connection with this report. In Australia, this publication has been distributed by<br />
The Hongkong and Shanghai Banking Corporation Limited (ABN 65 117 925 970, AFSL 301737) for the general information of its<br />
“wholesale” customers (as defined in the Corporations Act 2001). Where distributed to retail customers, this research is distributed by HSBC<br />
Bank Australia Limited (AFSL No. 232595). These respective entities make no representations that the products or services mentioned in this<br />
document are available to persons in Australia or are necessarily suitable for any particular person or appropriate in accordance with local<br />
law. No consideration has been given to the particular investment objectives, financial situation or particular needs of any recipient.<br />
In Japan, this publication has been distributed by HSBC Securities (Japan) Limited. It may not be further distributed in whole or in part for<br />
any purpose. In Korea, this publication is distributed by The Hongkong and Shanghai Banking Corporation Limited, Seoul Securities Branch<br />
("HBAP SLS") for the general information of professional investors specified in Article 9 of the Financial Investment Services and Capital<br />
Markets Act (“FSCMA”). This publication is not a prospectus as defined in the FSCMA. It may not be further distributed in whole or in part<br />
for any purpose. HBAP SLS is regulated by the Financial Services Commission and the Financial Supervisory Service of Korea.<br />
© Copyright. The Hongkong and Shanghai Banking Corporation Limited 2010, ALL RIGHTS RESERVED. No part of this publication may<br />
be reproduced, stored in a retrieval system, or transmitted, on any form or by any means, electronic, mechanical, photocopying, recording, or<br />
otherwise, without the prior written permission of The Hongkong and Shanghai Banking Corporation Limited. MICA (P) 142/06/2010 and<br />
MICA (P) 193/04/2010<br />
102
Nam Park*<br />
Analyst<br />
The Hongkong and Shanghai Banking Corporation Limited (HK)<br />
+852 2996 6591<br />
nampark@hsbc.com.hk<br />
Nam was appointed Head of <strong>Memory</strong> Devices Technology Research, <strong>Asia</strong> Pacific, in October 2008. Since joining HSBC in 2000, his roles<br />
have included Regional Head of Telecoms and Internet Research, Head of Korea Small & Mid-cap Research and Regional Head of<br />
Healthcare Research. Nam has extensive equity research experience in <strong>Asia</strong> since 1996. He has covered most sectors, including<br />
conglomerates, autos, media, alternative energy and defence, as well as writing Korea strategy. Nam holds a Ph.D. in Biophysics from<br />
Imperial College of Science, Technology and Medicine.<br />
Carolyn Poon*<br />
Analyst<br />
The Hongkong and Shanghai Banking Corporation Limited (HK)<br />
+852 2996 6586<br />
carolynpoon@hsbc.com.hk<br />
Carolyn covers memory devices technology for <strong>Asia</strong> Pacific. She joined HSBC as part of the <strong>Asia</strong>n Telecoms team, and subsequently<br />
was an analyst in the Korean Mid-caps and <strong>Asia</strong>n Healthcare teams. Before joining HSBC, she worked as an economist covering<br />
Greater China.<br />
* Employed by a non-US affiliate of HSBC Securities (USA) Inc, and is not registered/qualified pursuant to FINRA regulations.