Asia Memory-Bad memories fading

TMT
Memory Devices Technology
October 2010
Asia Memory
Bad memories fading
We believe the market has priced in weak 4Q10/1Q11 numbers; valuations remain
suppressed and close to trough levels in some cases. We expect demand-led recovery
through price elasticity starting at the end of this year.
We expect more sustainable profitability for leading vendors next year on rational
capex, slowing ASP declines, plus process migration-led cost reductions and a better
product mix that leverages into structural growth areas. We identify upside risks from
enterprise DRAM and NAND, plus quantify the impact of tablets and smartphones.
We expect stocks to start outperforming by end-4Q10/1Q11 as expectations rise of
slowing DRAM price declines and firmer NAND prices.
We prefer industry leaders with cost efficiency and leverage into high-growth,
high-margin sub-segments. Our top pick is Samsung Electronics OW,
an HSBC Asia Super Ten idea. We lower our TP for Hynix OW(V), Elpida OW(V),
Winbond OW(V), Nanya N(V) and Inotera N(V)
By Nam Park and Carolyn Poon
Disclosures and Disclaimer This report must be read with the disclosures and analyst
certifications in the Disclosure appendix, and with the Disclaimer, which forms part of it

TMT
Memory Devices Technology
8 October 2010
abc
Summary
We believe the market has priced in weak 4Q10/1Q11 numbers;
valuations remain suppressed and close to trough levels in some
cases. We expect a demand-led recovery through price elasticity
starting at the end of this year. We expect more sustainable
profitability for leading vendors next year on rational competition
plus process migration-led cost reductions and a better product mix
that leverages into structural growth areas. We identify upside risks
from enterprise DRAM and NAND, plus quantify the impact of
tablets and smartphones. We expect valuations to turn more positive
by end-4Q10/1Q11 as expectations rise of slowing DRAM price
declines and firmer NAND prices. Our top pick is Samsung
Electronics, an HSBC Asia Super Ten idea. We lower our forecasts
and TPs for Hynix OW(V), Elpida OW(V), Winbond OW(V), Nanya
N(V) and Inotera N(V).
Why bother with memory anyway?
We capitulate to higher rates of ASP decline (see Samsung Electronics: Comprehensive health check, 24
September 2010) and lower our forecasts and target prices other vendors accordingly. For all of our
coverage stocks we forecast y-o-y declining operating profits for 2011e. We forecast Elpida and
Taiwanese vendors to make some improvements off low bases. Looking solely at these y-o-y profit
decline comparisons for next year, there appears little reason to bother with the memory sector.
And yet, despite these revisions, valuations remain low and in some cases close to trough levels; we
believe the market has already priced in a weak 4Q10 and possibly a weak 1Q11 quarter as well. On top
of which, we think macroeconomic demand concerns and persistent fears of the memory cycle turning
conspired to suppress memory stocks’ valuations during recent quarters of growth and relatively high
margins. Those same factors were very swiftly evoked to punish memory stocks as soon as ASPs did start
falling. In fact, recent steep DRAM price declines result from an unusual set of circumstances during
1Q/2Q10, namely PC order pull-in (due to EUR weakening and EU demand concerns) plus delays in
process migration at second-tier players (which tightened supply). These factors shored up ASPs in 2Q10,
setting the market up for a sharp ASP decline starting in 3Q. Still, it is worth remembering that we
forecast DRAM ASP to rise 20% this year, and for NAND ASPs to fall only 20% for the full year.
1

TMT
Memory Devices Technology
8 October 2010
abc
We believe next year the industry would stabilise towards more sustainable profitability especially for
leading vendors, albeit at lower levels than this year, on more rational competition, lower costs from
process migration and better product mix. We also think that the market is ignoring some of the upside
risks that we analyse in detail here, including enterprise DRAM and NAND growth, plus the extent of the
impact of tablets and smartphones on NAND.
We expect sentiment towards the industry to turn more positive by end-4Q10 or early 1Q11 as
expectations rise of easing of DRAM ASP pressure by 2Q11; we think price elasticity will restart stunted
growth in DRAM content/system, whilst accelerate higher NAND content/system by end-2010. This
should become evident in PC OEM results and market data. For NAND, key catalysts include firmer than
expected ASPs starting in 4Q10/1Q11e, reflecting strong demand especially from smartphones. ASPs
could also firm as larger density demand rises towards 1Q-2Q11 on a raft of tablet releases.
More sustainable earnings, for leaders
A key argument we made in our previous sector report (Asia Memory: Seeing is believing, March 2010)
to support a positive view on the memory industry was that more rational competition/capex will prevail
as memory vendors remain wary of overcapacity and focus instead on profitability to strengthen balance
sheets. We think this argument remains largely valid. We believe leading vendors remain acutely wary of
market conditions and may well adjust expansion rates accordingly. A ‘healthy level of profitability’,
coupled with constrained capex/sales remains a key rerating driver for the industry.
DRAM operating profit margins
60%
40%
20%
0%
-20%
1Q09 2Q09 3Q09 4Q09 1Q10 2Q10 3Q10 4Q10 1Q11 2Q11 3Q11 4Q11
-40%
-60%
-80%
-100%
-120%
-140%
SEC Hy nix Elpida
Nany a Tech Inotera Winbond
Source: Company data, HSBC estimates
We believe leading vendors place greater focus on profitability through process migration, rather than
market share gains. We have seen little evidence, for example, of Samsung attempting to crowd out
competition through aggressive pricing of this year. Instead, we think market leaders, able to rapidly
migrate processes and develop new, higher-end products to lower costs and improve product mix to
insulate margins, seek to generate more sustainable returns. We forecast Samsung and Hynix to generate
2

TMT
Memory Devices Technology
8 October 2010
abc
20-26% operating margins (higher if we consider just Samsung’s memory operations) during 2011-12,
while managing capex/sales below their respective historical averages.
It is worth stressing that the sharply higher proposed capex by vendors (which to a large extent, the
market anticipated) is not excessive, in our view. For the industry, we forecast 2010-12 capex/sales of 26-
28% for DRAM and 48-51% for NAND. This is well below 2005-08 averages of 54% and 67% for
DRAM and NAND respectively. Additional wafer capacity in 2010-12 is mostly earmarked for NAND
(21-22%/year wafer growth for NAND, 3.7-7.9%/year for DRAM) without which severe capacity
constraints could arise. Other than NAND wafer adds, capex is likely to be funnelled mostly into process
migration. Supporting our thesis, Hynix sees longer-term capex/sales at c30% or lower, plus a normalised
operating margin of 15-20% over the longer term (see Hynix – Raising target price: Confirming our
thesis, 23 April 2010).
Upside risk lurks
Enterprise computing, storage
Retail consumer demand remains uncertain, but corporate IT forges ahead. We think the enterprise server
and storage market provide substantial upside risk for leading vendors that are positioned to cater to this
high-end market. Continued corporate IT spend plus the ongoing growth of cloud computing drives
strong demand for servers. DRAM beneficiaries include Samsung and Hynix. We forecast server unit
growth of 10% y-o-y in 2011, and content/box growth of 38%. Servers account for about 10% of total
global bit demand but blended ASP is far higher for vendors, at 2.5x for a 4GB module up to 6x for a
16GB module. The server storage market also benefits from disproportionately high revenues and
margins. Key vendors include Samsung, IMFT and Hynix. Server storage could account for the
equivalent of 12-17% of our 2012 forecast global NAND revenue, and the equivalent of 20-32% of our
2013 forecast NAND global revenue.
Tablets
We forecast tablet shipments to soar to 73m units next year, accounting for a quarter of our forecast
tablets plus notebooks. We estimate that netbook and notebook sales cannibalisation will lower 2010-12e
global DRAM demand by 1%, 5% and 5%. This will be partially offset for DRAM vendors such as
Samsung and Elpida that supply low-voltage, low-power parts that could be integrated into tablets. We
expect tablets to account for 13% of global NAND demand next year (2010: 5%), mostly driven by a
massive 214% 2010/12e bit growth rate. We think Windows 7 could help drive tablet GB/system. Key
beneficiaries from tablet-related memory growth are Toshiba and Samsung.
Smartphones
Mobile handsets have also emerged as the leading usage of NAND on proliferation of smartphones. Handsets
account for 37% of our 2011 NAND bit demand forecast, growing to 42% of total NAND bit demand by 2012.
We forecast smartphone unit shipments to rise 54%, 39% and 35% during 2010-12, accounting for 35% of
total handsets shipped globally by 2012. Smartphones drive rapid mobile DRAM bit growth through unit
shipment growth and content per box. Mobile DRAM prices tend to be higher on a bit basis, as memory
vendors incorporate custom-made solutions for specific vendors. Samsung, Elpida and Hynix are leaders in
mDRAM. For NAND, we forecast GB/system for smartphones to double each year during 2011-12, driven by
3

TMT
Memory Devices Technology
8 October 2010
abc
lower NAND prices, the availability of larger capacity mobile NAND, rapidly rising screen resolutions, higher
resolution video, games content and cannibalisation of media players.
Worth a look? When and what?
Timing – over the next 1-2 quarters
Despite lowering forecasts for our companies, valuations are close to trough levels for many stocks. Has the
market priced in all industry-wide negativity? For commodity DRAM, it is close to doing so – a weak 4Q10
earnings outlook for industry participants, possibly extending out to 1Q11, is the consensus view now.
We expect to see sentiment starting to turn more positive by end-4Q10 or early 1Q11 as potential easing
of DRAM ASP declines by 2Q11 starts pricing in; the PC market would require a couple of quarters to
absorb Taiwanese vendors’ bit growth during 4Q10/1Q11. By the end of this year we think price
elasticity will restart stunted growth in DRAM content/system. This should become evident in PC OEM
results and market data, boosting confidence that future DRAM ASP declines could be more muted.
For NAND, key catalysts would include firmer-than-expected ASPs starting in 4Q10/1Q11e. This would
reflect soaring demand especially from smartphones. Firmer ASPs could also arise from orders of larger
density units towards 1Q-2Q11 as tablet releases accelerate in 2H11. By 3Q11 we could see prices
softening more as Samsung and Toshiba start mass production at new fabs in earnest.
Stock coverage and ratings summary
Company Ticker Rating New TP
(LC)
CP (LC)
Potential
return (%)
Previous
TP
Change
(%)
Previous
rating
Change
Samsung Electronics 005930 KS OW 1,139,000 793,000 44% 1,139,000 0% OW -
Hynix Semiconductor 000660 KS OW(V) 32,000 23,400 37% 36,000 -11% OW(V) -
Elpida Memory 6665 JP OW(V) 1,500 928 62% 3000 -50% OW(V) -
Nanya Technology 2408 TT N(V) 20 19.3 4% 25 -20% N(V) -
Inotera Memories 3474 TT N(V) 17 16.2 5% 20 -15% N(V) -
Winbond 2344 TT OW(V) 10.8 8.03 34% 12.8 -16% OW(V) -
Source: HSBC
Go for the leaders
We expect memory vendor prospects and valuations to diverge as leaders accelerate process migration
and improve product mix. Increasingly DRAM and NAND bit growth are driven by mobile, enterprise
applications that demand high-performance parts fabricated on advanced processes.
Our top sector pick is Samsung Electronics (OW, 44% upside potential to TP) based on its unparalleled
overall memory positioning and compelling valuations. It remains the industry leader in cost efficiency
and is highly leveraged into high-growth, high-margin sub-segments in both DRAM and NAND. Its
diversified technology portfolio – especially its telecoms operations – provide portfolio defensiveness,.
We think 1.3x 2011e PB (historic range: 1.1-2.5x) is compelling.
Of the pure plays, we find Hynix (OW(V), 37% upside potential to TP) attractive. Valuations are
undemanding at 1.3x 2011e PB (historic range: 0.5-2.5x) compared to its 20%-plus ROE, driven by
strong positioning in high-end DRAM and rapid catch-up with NAND leaders Toshiba and Samsung with
its 2xnm product (that is customer sampling at the moment). Legacy concerns over its balance sheet are
4

TMT
Memory Devices Technology
8 October 2010
abc
fading on strong cash flow. Stock catalysts include a creditor stake sale to a major strategic buyer, major
customer wins with new DRAM and NAND products and better than expected quarterly results.
Elpida (OW, 62% upside potential to TP) is a high-beta pure play, in our view, with the most upside to
TP, trading on FY12e (end-March 2012) PB of just 0.5x. It looks oversold on slow process migration and
strong JPY concerns. Valuations also seem to have reflected potential dilution from further capitalraising,
especially as it looks to take strategic stakes in Taiwanese DRAM vendors. We like its strategy of
manufacturing mobile DRAM (in which it has substantial market share) mostly in Japan whilst
completely migrating to a relatively competitive 4xnm process at its Taiwanese fab which focuses on
commodity DRAM.
Of Taiwanese memory vendors we like Winbond (OW(V), 34% upside potential to TP). The stock has
held up reasonably, but trades on 2011e PB of 0.7x (historic range: 0.2-1.2x). We like its diversified niche
portfolio that is heavily exposed to mobile and speciality applications. We expect process migration and
flexible product mix to sustain margins. We expect Nanya and Inotera’s price performance to remain
constrained until signs of improvement in node migration and cycle times become tangible. We remain
Neutral (V) on these two stocks.
5

TMT
Memory Devices Technology
8 October 2010
abc
Contents
Industry snapshot, themes 7
Demand drivers – DRAM 14
Demand drivers – NAND 25
Supply drivers – DRAM 41
Supply drivers – NAND 47
Valuations, ratings 55
Company profiles 59
Elpida Memory (6665) 60
Hynix Semiconductor (000660) 67
Inotera Memories (3474) 74
Nanya Technology (2408) 79
Samsung Electronics
(005930) 84
Winbond (2344) 93
Disclosure appendix 98
Disclaimer 102
6

TMT
Memory Devices Technology
8 October 2010
abc
Industry snapshot, themes
DRAM set for oversupply until 2Q11, until price elasticity offsets
flood of bit supply. NAND demand-supply looks more positive
Higher-end DRAM and NAND producers to benefit from strong
structural demand growth from mobile, enterprise computing
Relentless process migration and higher-end product mix benefit
leading vendors
Demand & supply forecast (USDm)
2008 2009 2010e 2011e 2012e 1Q10e 2Q10e 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e
Memory Revenue (USDm)
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
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
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
Wafer Capacity (12” equ. kwpa)
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
NAND 9,194 10,476 12,529 15,129 15,849 829 867 878 918 948 978 1,095 1,155
Memory Shipment
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
NAND (16Gb eq.) ,457 2,492 5,335 9,921 12,599 465 575 671 780 924 1,174 1,492 1,744
Memory ASP
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
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
Memory sales change (%)
DRAM 0% 79% 1% 7% -5% 7% 15% -3% -1% -4% 2% 5% 2%
NAND -3% 37% 32% 16% -17% -5% 14% 4% 3% 4% 15% 11% 2%
Total -1% 65% 10% 10% -9% 4% 14% -1% 0% -1% 6% 7% 2%
Wafer Capacity change (%)
DRAM -1% 7% 4% 8% 5% 0% 0% 2% 4% 0% 0% 0% 0%
NAND -5% 14% 20% 21% 5% 1% 5% 1% 5% 3% 3% 12% 5%
Bit Shipment change (%)
DRAM 22% 49% 49% 43% 29% 5% 7% 11% 18% 7% 8% 11% 10%
NAND 41% 71% 114% 86% 27% 4% 24% 17% 16% 18% 27% 27% 17%
Memory ASP change (%)
DRAM -18% 20% -32% -25% -26% 2% 7% -12% -16% -10% -6% -5% -7%
NAND -32% -20% -38% -38% -35% -9% -8% -11% -11% -12% -10% -13% -13%
Source: IDC, HSBC estimates
7

TMT
Memory Devices Technology
8 October 2010
abc
DRAM – price elasticity drives
higher content/box
We expect DRAM price declines to start slowing in
2Q11 as lower ASPs from sharply higher supply
during 4Q10-1Q11 stimulate higher content/system
and as supply increases peak out. Overall, we expect
content/system to rise 33% next year, following a
27% y-o-y rise in 2010; high ASPs this year
depressed content growth this year.
DRAM demand and ASPs
2009 2010e 2011e 2012e
ASP (USD) -18% 20% -32% -25%
Content/system (MB) 27% 27% 33% 32%
Units shipments (m) 2% 14% 7% 9%
Demand (MB bn) 30% 46% 42% 43%
Source: HSBC estimates
Bit growth in the coming quarters could surprise
on the upside. We expect PC OEMs to gradually
start raising content/box starting in 3Q on steep
price declines; 2GB DDR3 modules now cost
around USD36, 23% below the peak of USD46.5
in May. We expect price elasticity to kick in
meaningfully in 4Q10 on persistent DRAM
pricing weakness arising from expanded supply.
We model 18% q-o-q average ASP decline for
DDR3 contract prices in 4Q10, implying an
average price of USD32 per 2GB module for the
quarter. At that level, 4GB content/system on
desktops should accessible to the majority of PC
OEMs. By 1Q11, 2GB modules could have fallen
to around USD28, starting to drive real increase in
content/system above the 4GB level. For higherend
PCs, we expect content/box climbing to 6-
8GB next year to competitively differentiate.
We also expect accelerating adoption of higherend
4GB modules based on 4x-5xnm 2Gb parts.
Leading DRAM vendors such as Samsung and
Hynix that are rapidly shifting production of 2Gb
parts to 4xnm and below from 5xnm will benefit
the most from this trend.
NAND
NAND remains an exciting high-growth segment.
We forecast 2010-12 NAND bit demand CAGR
of 103%, rising 114% next year. Global appetite
remains insatiable for storage capacity that is fast,
robust and low-power.
NAND demand forecast
45,000
120%
100%
30,000
80%
60%
15,000
40%
20%
-
0%
2009 2010f 2011f 2012f
Demand ( GB, m) Change, y-o-y (%)
Source: IDC, HSBC estimates
Foreseeable demand remains strong on rising
affordability, expanding applications in high
potential markets such as smartphones, tablets, PC
storage plus nascent markets such as automotive
applications. We continue to believe the risk is on
the upside, as demand accelerates on more rapid
affordability trends.
Unlike DRAM, NAND demand remains strong
for both commodity parts and embedded solutions.
The main embedded demand driver is handsets,
led by smartphones, which we estimate will
account for 37% of total bit demand next year, up
from 28% in 2009. Another embedded solutions
driver is tablets, a segment which practically did
not exist until April this year. We expect SSD
growth, which had been stalled this year by high
ASPs to see rapid growth due to falling NAND
prices and indirect demand stimulation through
the ‘tablet effect’.
8

TMT
Memory Devices Technology
8 October 2010
abc
High-end computing DRAM MB bn, % of total DRAM demand
PC server DRAM MB bn, % of total PC DRAM demand
500
12%
400
20%
400
300
200
100
11%
11%
10%
10%
9%
300
200
100
15%
10%
5%
-
9%
-
0%
2009 2010f 2011f 2012f
2009 2010f 2011f 2012f
High-end com puting
% of total
x 86 serv ers
% of total PC
Source: IDC, HSBC estimates
Source: IDC, HSBC estimates
Enterprise
Enterprise flash drives
We think the enterprise server and storage market
provide tangible upside risk to revenue and
margin forecasts for leading vendors that are
positioned to cater to this high-end market.
Enterprise server DRAM demand
Servers – niche, but very high ASP market
We expect continued corporate spend plus
ongoing growth of cloud computing to drive
strong demand from this segment. Key
beneficiaries include Samsung and Hynix.
We forecast server unit growth of 10% y-o-y in
2011, but higher content/box growth of 38%;
server memory content/system tends to double
every two years. Servers (including PC servers)
account for about 10% of total global bit demand,
but the blended ASP impact is far higher for
vendors, at 2.5x for a 4GB module up to 6x for a
16GB module. Price premiums are driven by
high-density, high performance and lower power
consumption, in addition to higher production
costs. Demand will be further catalysed by
availability of high-density (2Gb and 4Gb very
soon), low-power DRAM modules.
A major upside risk to high-end SSD demand
Similarly, the server storage market offers
disproportionately high revenues and margins.
Key beneficiaries of this market include Samsung,
IMFT and Hynix, to be joined next year by
Toshiba. We do not separately model the server
NAND market in our forecasts but this analysis
illustrates the potential upside risk to our forecasts.
High-end server SSD storage is very expensive.
Despite this, data centres and corporates use SSD
due to 1) extremely fast data input/output,
especially important for cloud computing, 2)
small form factor (space), 3) low power
consumption. As ASPs fall, we expect SSDs will
replace HDDs as ‘front line’ storage, with HDDs
acting as inexpensive backup.
We estimate that server storage could account for
the equivalent of 12-17% of our 2012 forecast
global NAND revenue, and the equivalent of 20-
32% our forecast 2013 NAND global revenue.
These estimates assume SSDs account for 5% and
10% respectively of 2011-12 disk storage systems
and are based on SLC.
9

TMT
Memory Devices Technology
8 October 2010
abc
Smartphones
Telecoms is becoming a major consumer of
DRAM, accounting for 9 % of DRAM bit demand
this year and growing to 10% in 2012e. Mobile
handsets have also emerged as the leading usage
of NAND on proliferation of smartphones.
Handsets account for 37% of our 2011 NAND bit
demand forecast, growing to 42% of total NAND
bit demand by 2012. We forecast smartphone unit
shipments to rise 54%, 39% and 35% during the
periods, accounting for 35% of total handsets
shipped globally by 2012e (2009: 16%).
Mobile DRAM
Smartphones are driving rapid mobile DRAM bit
growth through both unit shipment growth and
content per box. Development of smartphones
now mirrors that of PCs, driven by resourceintensive
wireless broadband, multimedia and
multitasking. We expect DRAM content to rise in
parallel with surging computing power to support
these demands as well as next-iteration operating
systems such as Android 3.0.
Mobile DRAM prices tend to be higher on a bit
basis, as memory vendors need to work closely
with mobile phone makers to incorporate custommade
solutions for specific handsets. Samsung,
Elpida and Hynix are leaders in mDRAM. The
higher-end mDRAM market (i.e. 128MB
upwards) is dominated by major DRAM players
that are capable of customising MCP (multichip
packages) for handset vendors. We forecast
mDRAM bit growth to rise 102%, 92% and 112%
y-o-y for 2010-12. We assume DRAM
content/system for smartphones to rise 31%, 39%
and 57% for 2010-12, to around 390MB by 2012.
Mobile NAND
We forecast GB/system for smartphones to double
each year during 2011-12 driven by lower NAND
prices, availability of larger capacity mobile
NAND, rapidly rising screen resolutions, higher
resolution video, games content and
cannibalisation of media players. We estimate
NAND on a 16GB Apple iPhone 4 to account for
14% of BOM, a level which we think will be
maintained going forward. We assume most non-
Apple smartphones will use less NAND than
Apple to lower prices, obliging the consumer to
buy flash cards to fulfil extra storage needs.
10

TMT
Memory Devices Technology
8 October 2010
abc
% of total NAND bit demand from tablets System unit and GB/system growth for tablets
15
10
5
0
120
100
80
60
40
20
0
60
50
40
30
20
10
0
2009 2010f 2011f 2012f
2009 2010f 2011f 2012f
% of NAND demand
Sy stem units (m)
GB/sy stem
Source: HSBC estimates
Source: HSBC estimates
Tablets
We forecast tablet shipments to soar to 73m units
next year, accounting for a quarter of our forecast
tablets plus notebooks. Tablets represent a major
new NAND application that has explosively
entered the scene.
DRAM
An emerging concern for commodity DRAM is that
tablets will cannibalise netbook and notebook sales,
reducing overall DRAM demand. We estimate 1%,
5% and 5% cannibalisation impact on global DRAM
demand for 2010-12e. DRAM vendors such as
Samsung and Elpida that can supply low-voltage,
low-power DDR (and increasingly DDR2) that
could be integrated into tablets stand to benefit.
The cannibalisation impact will really start to be felt
next year, when we estimate tablets will account for
26% of tablet plus notebook units but only c6% of
tablet plus notebook DRAM bits. We expect DRAM
content in tablets to start rising quickly next year to
support multi-tasking and more resource-intensive
apps. We think this is possible, given DRAM is only
c3% of BOM for current Apple iPads. We think 5-
6% is more likely next year as 512MB becomes
standard. By 2012e we expect tablet DRAM would
reach above 1GB per system, catalysed by Windows
and competitive pressure amongst tablet vendors to
differentiate their products.
NAND
We expect tablets to account for 13% of global
NAND demand next year (2010: 5%), mostly driven
by a massive 214% 2010/12e bit growth rate from a
standing start this year. We expect GB/system to
initially rise slowly as tablets will mostly run
compact Android and iOS operating systems. Key
beneficiaries are Toshiba and Samsung.
For Apple’s iPad NAND as a proportion of
hardware-only BOM varies between 10% and 30%
depending on the model. We expect falling ASPs to
drive content/system; 16GB is about the minimum
acceptable onboard NAND storage for a tablet in our
view. Vendors should have little problem meeting
this requirement this year. Next year, as ASPs fall to
cUSD1/GB, 32GB of NAND would account for
18% of BOM, which we think is a reasonable level.
Interestingly, we think Windows 7 could drive
tablet GB/system. Windows could mount a belated
but serious challenge in the tablet market, as tablet
hardware that could support resources-intensive
Windows becomes available including Intel’s new
Oak Trail processor which is set to appear in
devices starting 1H11. For Windows tablets, we
think there would be a larger NAND requirement as
a full implementation of basic Windows 7 needs a
bare minimum 16GB of storage.
11

TMT
Memory Devices Technology
8 October 2010
abc
EBIT margins for memory manufacturers
60%
40%
20%
0%
-20%
-40%
-60%
-80%
-100%
-120%
-140%
-160%
Samsung Hy nix Elpida Toshiba Pow erchip Nany a Inotera ProMOS Micron SanDisk
2006 2007 2008 2009f 2010f 2011f
Source: HSBC estimates, Bloomberg consensus
Revenue, EBIT for memory manufacturers
Revenue (USDm) SEC Hynix Elpida Nanya Tech Inotera Winbond
2007 17,306 8,605 3,877 1,610 1,396 746
2008 22,353 6,818 3,557 1,151 1,190 581
2009 26,808 7,906 3,927 1,284 1,098 594
2010e 37,461 12,429 7,055 1,899 1,450 1,020
2011e 40,639 13,217 6,662 2,538 2,013 1,040
2012e 44,382 14,723 6,969 3,072 2,472 1,141
Revenue changes (%) SEC Hynix Elpida Nanya Tech Inotera Winbond
2008 29% -21% -8% -28% -15% -22%
2009 20% 16% 10% 12% -8% 2%
2010e 40% 57% 80% 48% 32% 72%
2011e 8% 6% -6% 34% 39% 2%
2012e 9% 11% 5% 21% 23% 10%
EBIT margin (%) SEC Hynix Elpida Nanya Tech Inotera Winbond
2007 25% 6% 3% -17% 7% -41%
2008 10% -28% -33% -66% -48% -46%
2009 8% 2% -17% -37% -27% -31%
2010e 28% 30% 21% -6% -8% 11%
2011e 20% 26% 8% 8% 8% 7%
2012e 19% 22% 9% 11% 13% 10%
EBIT (USD) SEC Hynix Elpida Nanya Tech Inotera Winbond
2007 4,340 492 134 (273) 96 (303)
2008 2,143 (1,920) (1,175) (765) (573) (268)
2009 2,059 191 (648) (482) (293) (182)
2010e 10,508 3,741 1,454 (114) (117) 111
2011e 8,328 3,379 563 199 171 75
2012e 8,249 3,296 616 336 315 111
EBIT changes (y-o-y%) SEC Hynix Elpida Nanya Tech Inotera Winbond
2008 -51% nm nm nm nm nm
2009 -4% nm nm nm nm nm
2010e 410% 1858% nm nm nm nm
2011e -21% -10% -61% nm nm -33%
2012e -1% -2 9% 69% 84% 48%
Source: Company, HSBC estimates, note SEC is for consolidated Semiconductor division figures
12

TMT
Memory Devices Technology
8 October 2010
abc
Samsung DRAM product mix by revenue, 2Q10
Mobile
20%
Hynix DRAM product mix by revenue, 2Q10
Mobile
13%
PC
45%
Serv er
20%
PC
45%
Serv er
19%
Graphic
Specialty
5%
Graphic
10%
Specialty
12%
11%
Source: Company, HSBC estimates
Source: Company, HSBC estimates
Elpida DRAM product mix by revenue, 2Q10
Premier
25%
Nanya DRAM product mix by revenue, 2Q10
Specialty
15%
Serv er
10%
PC
75%
PC
75%
Source: Company, HSBC estimates . Premier includes specialty and mobile DRAM
Source: Company, HSBC estimates
Winbond DRAM product mix by revenue, 2Q10
PC
2%
Mobile
18%
Specialty
40%
NOR
28%
Source: Company, HSBC estimates
Graphic
12%
13

TMT
Memory Devices Technology
8 October 2010
abc
Demand drivers – DRAM
We forecast 2010/12 bit CAGR of 46% as price elasticity starts
kicking in on ASP declines
PC bit growth from higher per box content; smartphone DRAM set
to grow on higher content
Leading vendors best positioned to harness high-end demand
from low-voltage, high-performance mobile and server application
Time to look at the upside risk
The market now generally expects ASP declines
for the coming 2-3 quarters, driving negative
sentiment and potentially poor share price
performance. Here, instead of focussing on widely
accepted negatives, we focus on upside risks to
forecasts, i.e. factors that could potentially turn
around operating and hence share price
performance. We expect DRAM demand to
remain healthy, growing at a 46% CAGR for
2010/12e. For this year and next year, we expect
48% and 47% y-o-y bit demand growth
respectively. Driving this demand are 1) price
elasticity on sharply lower ASPs, mobile handsets
led by smartphones and 3) strong continued
corporate demand for PCs and servers.
Price elasticity to support DRAM bit growth.
Near term overall PC demand remains relatively
unclear, weighed down by weak consumer
sentiment in developed markets. However, we
expect corporate demand to provide support for
demand as a whole. For the global PC DRAM
market, we forecast content/box rising 33% y-o-y
for 2011, having slowed to c27% this year
(excluding tablet PCs). We expect PC OEMs to
start raising content/box again in 4Q10 as 2GB
modules fall below USD30/2GB. Indeed, 1Gb
DDR3 contract prices for 2H September fell
below USD2 already for the first time since 2H
October 2009. We expect DDR3 contract ASP to
continue falling steeply in 3Q10-1Q11 on rising
supply from vendors’ wafer capacity expansion
and process migration.
Tablets impact commodity, positive for
advanced players. We expect tablets to lower
global DRAM bit demand by c1% this year and
c5% for each of 2011-12e global demand. We
believe memory vendors that can supply the lowvoltage,
low power DRAM for integration with
SoC used in tablets stand to benefit.
Leading vendors best positioned to deliver
higher-end products, and protect margins with
process migration. We think product mix will
become more critical for memory vendors during
the next few quarters as ASP falls steepen. Rapid
process migration by leading DRAM vendors
results not just in better cost structure; from a
competitive perspective, this enables them to meet
demand for low-power parts that advanced nodes
are better suited to deliver, both for mobile
devices space (notebooks, smartphones, tablets)
14

TMT
Memory Devices Technology
8 October 2010
abc
and higher end servers. We also expect leading
vendors’ increased production of 2Gb parts to
stimulate demand for larger memory modules.
Here we look at the potential impact of corporate
servers, which account for c10% of global bit
demand, but we believe accounts for a much
larger proportion of global DRAM revenue;
server DRAM is c2.5-6x more expensive
compared to ‘regular’ memory modules.
DRAM demand forecast
40,000
60%
50%
30,000
40%
20,000
30%
10,000
20%
10%
-
0%
2008 2009 2010f 2011f 2012f
Demand (1Gb, m) Change (%)
Source: IDC, HSBC estimates
Upside/downside risks
Key upside risk to our forecasts is a long-term,
controlled growth of the industry, with strong bit
growth and relatively mild ASP declines.
We expect bit growth to accelerate in 2H10, due
to PC build for the holidays, plus corporate PC
recovery. We forecast most of the bit growth to
occur in 2H10 as the impact of node migration by
makers starts to come into play.
We think downside risks are relatively limited for
DRAM demand over the next two years. Risks
include slower than expected increase in DRAM
content per box for PCs on relatively high ASPs.
The key supply downside is irrational competition
leading to rapid ASP deflation. From 2012 onwards,
additional risks include technology migration
difficulties due to newer technologies such as EUV.
Catalysts – short term, long term
Short-term catalysts include strong data
points from various PC OEMs for unit sales
and content per box. Near term, sell through
data from PC OEMs over the holidays in
China and developed markets will be key.
DRAM bit demand, 2009
DRAM bit demand, 2012e
Others
10%
Others
9%
Mobile
handsets
Mobile
handsets
8% 10%
PC
PC
51%
Other data
56%
Other data
processing
26%
processing
27%
Tablets
3%
Source: IDC, HSBC estimates
Source: IDC, HSBC estimates
15

TMT
Memory Devices Technology
8 October 2010
abc
Global DRAM – demand & supply forecast
2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e
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
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
Change (%) 22% 49% 49% 43% 5% 7% 11% 18% 7% 8% 11% 10%
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
Change (%) 24% 48% 47% 46% 5% 14% 8% -4% 13% 15% 22% 4%
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
Change (%) 0% 79% 1% 7% 7% 15% -3% -1% -4% 2% 5% 2%
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
Change (%) -18% 20% -32% -25% 2% 7% -12% -16% -10% -6% -5% -7%
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
Change (%) 15% -34% -24% 12% 7% -15% -18% -11% -7% -7% -7%
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
Change (%) -1% 7% 4% 8% -1% 0% 2% 4% 0% 0% 0% 0%
Demand (MB bn) 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e
PC 738 1,076 1,532 2,195 237 269 290 280 306 352 429 444
Tablets - 5 55 115 1 1 1 1 11 13 15 16
Other data processing 346 522 781 1,142 115 130 141 136 156 180 219 226
Mobile handsets 111 177 266 429 39 44 48 46 53 61 74 77
Others 139 202 274 363 44 50 55 52 55 63 77 80
Total 1,334 1,981 2,908 4,245 436 495 535 515 582 669 814 843
Change (%) 24% 48% 47% 46% 5% 14% 8% -4% 13% 15% 22% 4%
Demand mix (%)
PC 55% 54% 53% 52% 54% 54% 54% 54% 53% 53% 53% 53%
Tablets 0% 0% 2% 3% 0% 0% 0% 0% 2% 2% 2% 2%
Other data processing 26% 26% 27% 27% 26% 26% 26% 26% 27% 27% 27% 27%
Mobile handsets 8% 9% 9% 10% 9% 9% 9% 9% 9% 9% 9% 9%
Others 10% 10% 9% 9% 10% 10% 10% 10% 9% 9% 9% 9%
Total 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100%
Device shipment (m units) 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e
PC 299 342 366 397 75 86 92 89 73 84 102 106
Change (%) 2% 14% 7% 9% -19% 14% 8% -4% -18% 15% 22% 4%
DT PC 136 147 156 162 32 37 40 38 31 36 44 45
NB PC 141 171 195 224 38 43 46 44 39 45 55 57
Netbook 22 24 14 11 5 6 7 6 3 3 4 4
Other data processing 43 60 78 97 13 15 16 16 16 18 22 23
Change (%) -11% 12% 13% 18% 43% -7% -8% -8% 44% -7% -8% -8%
Mobile handsets (m) 1,116 1,244 1,350 1,457 284 298 321 342 316 332 346 356
Change (%) -8% 11% 9% 8% -7% 5% 8% 7% -8% 5% 4% 3%
Smartphones (m) 176 271 377 509 55 65 72 79 84 93 98 102
Change (%) 8% 54% 39% 35% 7% 18% 11% 10% 6% 11% 5% 4%
Tablets (m) - 16 73 102 - 3 4 9 12 15 21 25
Change (%) 356% 39% 40% 110% 36% 25% 40% 19%
DRAM MB/system
PC 2,467 3,143 4,187 5,521
Tablets 288 774 1,161
Other data processing 8,111 8,729 10,047 11,760
Telecoms 99 142 197 295
Source: Company, HSBC estimates. Blended ASP and DDR3 contract ASPs are for 1Gb equivalent. Wafer capacity is 12” equivalent
16

TMT
Memory Devices Technology
8 October 2010
abc
DRAM shipments for PC, % of total DRAM demand, MB bn
System units and content/system growth for computing
500
56%
500
600%
400
300
200
100
55%
54%
53%
52%
51%
50%
400
300
200
100
500%
400%
300%
200%
100%
-
49%
-
0%
2009 2010f 2011f 2012f
2009 2010f 2011f 2012f
PC
% of total
Unit (mn) MB/system '000
Source: IDC, HSBC estimates
Source: IDC, HSBC estimates
Computing
Relevance: PCs remain the largest demand driver,
accounting for c54% of DRAM bit demand this
year, though falling slowly to c53% by 2012e, with
bit demand CAGR of c44% during the period.
Upside/downside factors. Upside factors include
greater demand arising from price elasticity. This
memory per system to handle more complex tasks
such as high definition video, plus competition
between PC OEMs is driving greater content/box.
PC unit shipments could beat market expectations on
stronger than expected corporate demand recovery.
Downside factors include the industry-wide
downside risk that if DRAM prices rise further,
PC makers may be forced to reduce onboard
content per box, which could cap shorter-term bit
growth. Similarly, high mDRAM ASPs could
suppress inclusion of more mDRAM in handsets.
PC unit shipments could miss market expectations.
Catalysts. Strong bit growth from major makers,
stronger than expected DRAM spot and contract
prices, corporate results reflecting higher
profitability, including that from cost reductions
resulting from process efficiencies and migration.
Positive DRAM demand outlook
Next year, we forecast PC DRAM bit demand to
rise 42.4% y-o-y from this year’s 45.8%. 2010e
growth is relatively high due to the rebound in PC
unit shipments following the global financial
crisis. We expect higher content/system rather
than PC unit shipments to drive 2011e bit growth
for commodity DRAM.
HSBC’s Regional Asia Technology Hardware
analyst, Wanli Wang, forecasts global PC
shipment growth (excluding tablets) to slow to
10.9% next year from this year’s estimate of
14.7% (see Asian Hardware – Tablets the swing
factor for 2011 PC market, 16 September 2010).
For 2011 he expects global PC shipments to rise
23% with tablet PCs, 11% without.
Nearer term, Wanli expects 3Q10 Taiwanese
notebook shipments to track in line, though he
expects sequential 3Q10 shipment growth to be
relatively low due to a high 2Q10 base. Unit
shipments for 2Q benefited from order pull-in
ahead of price increases, plus rising seaborne
shipments to lower costs. Channel checks suggest
that there is no sign of weaker demand, even in
Western Europe. Wanli also suggests that 3Q10
PC outlook is sufficiently healthy enough for PC
vendors to maintain component orders and/or
component vendor capacity remains tight, with
little or no inventory build in 2Q10.
17

TMT
Memory Devices Technology
8 October 2010
abc
PC OEMs, processor vendors
readacross
Strong server and networking demand…
PC OEMs’ recent financial results and outlooks
reflect strong corporate refresh demand, but
uncertainty in consumer demand. This points to
divergent prospects for DRAM makers, with those
supplying higher-end products for servers and
workstations faring better than those supplying
just commodity parts. We estimate that servers
account for c20% of Hynix and Samsung’s
DRAM revenue. For 3Q10, Samsung guided
blended DRAM ASP decline of high-single digits,
while Hynix guided for ASPs to fall ‘slightly’ this
quarter. We estimate average DDR2 and DDR3
contract prices fell sharply by 13% q-o-q for the
quarter based on DRAMeXchange prices.
Dell (DELL US, not rated) expects corporate
refresh demand in client and enterprise systems to
continue over the next several quarters. The firm
reported strong 2QFY11 results led by corporate
and public sectors. Large enterprise revenue rose
38%, public sector revenue rose 21% and SMB
revenue rose 25%. Note that large, public and
SMB revenues accounted for over 80% of Dell’s
total 2QFY11 revenue. Dell (or HP) do not appear
concerned about component prices, a key negative
point for these firms in the previous quarters.
HP’s (HPQ US, not rated) recent 3QFY10
results saw strong 19% y-o-y growth in enterprise
storage and server (ESS) revenue. HP also saw
Personal Systems revenue rise 17%, within which
commercial client revenue rose 25% led by
desktop and workstation revenue (+27%, 54%
respectively). Overall, commercial hardware
revenue rose 28%, compared to 4% growth in
consumer hardware. HP in particular benefited
from a recent raft of launches of enterprise server,
storage and networking equipment. HP raised its
EPS guidance for 4QFY10 and raised its guidance
for full year revenue (now 9.5% y-o-y).
Lenovo (992 HK, OW(V), HKD4.99, TP
HKD5.70; covered by HSBC analyst Wanli
Wang) FY1Q11 operating margin rose
significantly, which we expect to continue in
FY2Q11 due to lower component prices
(especially DRAM and LCD) plus rising
corporate high-end product demand.
Intel (INTC US, not rated) had a record quarter
in 2QFY10, with revenue buoyed by record
mobile processor and server processor revenues,
which were driven by strong corporate demand. In
particular, the Data Center Group revenue rose
13% q-o-q.
…but uncertain consumer demand
Compared to ongoing strength in the corporate IT
market, uncertainty remains in the consumer side.
For example, Dell’s latest results showed flat
consumer revenue, contrasting sharply with its
enterprise revenues. The picture is more telling from
the processor makers. Intel recently guided lower
3QFY10 revenue due to weaker than expected
demand for consumer PCs in developed markets.
Similarly, AMD (AMD US, not rated) guided down
its 3QFY10 revenue due to weaker than expected
demand especially in the consumer notebook market
in North America and Western Europe.
DDR2, DDR3 contract prices (USD)
3
2.5
2
1.5
1
Jul-09
Oct-09
Source: DRAMeXchange
Jan-10
Apr-10
Jul-10
1Gb DDR2 128Mx 8 667 Mhz contract
1Gb DDR3 128Mx 8 1066MHz contract
18

TMT
Memory Devices Technology
8 October 2010
abc
Price elasticity = higher content/box
We expect demand downside to be relatively
limited because of the greater affordability of
DRAM modules. We expect DRAM price
declines to start slowing in 2Q11 as lower ASPs
from sharply higher supply during 4Q10-1Q11
stimulate higher content/system and as supply
increases peak out. Overall, we expect content/
system to rise 33% next year, following a 27% y-
o-y rise in 2010; high ASPs this year depressed
content growth this year, though the extent of
DRAM content reduction or freeze appears to
vary according to PC OEMs.
DRAM demand and ASPs
2009 2010e 2011e 2012e
ASP (USD) -18% 20% -32% -25%
Content/system (MB) 27% 27% 33% 32%
Units shipments (m) 2% 14% 7% 9%
Demand (MB bn) 30% 46% 42% 43%
Source: HSBC estimates
Through 1H10, second-tier PC OEMs in
particular have pointed to high DRAM costs as a
proportion of bill of materials (BOM), at about
13-15%. We are aware that some PC vendors
have lowered content to 3GB/box, but vendors’
strategies towards memory content tend to vary.
OEMs more focused on higher-end systems have
been reluctant to reduce content per box on
competitiveness grounds.
Indeed, bit growth in the coming quarters could
surprise on the upside. Before the recent ASP
declines Hynix expected GB/system growth
slowing in 2H10, but at the same time it did not
expect drastic cuts in content growth. Samsung
expected GB/system in 3Q to be flat. Similarly,
Micron saw OEMs continuing to market 4GB
models; it sees limited cases of low-priced
consumer PCs offering lower configurations.
We expect PC OEMs to gradually start raising
content/box starting in 3Q as DRAM prices have
dropped steeply; 2GB DDR3 modules now cost
around USD36, 23% below the peak of USD46.5
in May. We expect price elasticity to kick in
meaningfully in 4Q10 on persistent DRAM
pricing weakness arising from expanded supply.
We model 18% q-o-q average ASP decline for
DDR3 contract prices in 4Q10, implying an
average price of USD32 per 2GB module for the
quarter and a 31% fall from peak prices. At that
level, 4GB content/system on desktops should
become affordable to the majority of PC OEMs.
By 1Q11, 2GB modules could have fallen to
around USD28, starting to drive real increase in
content/system above the 4GB level. For higherend
PCs, we expect content/box climbing to 6-
8GB next year to competitively differentiate.
Even currently, higher-end desktops and
notebooks ship with up to 8GB.
We also expect accelerating adoption of higherend
4GB modules based on 4x-5xnm 2Gb parts to
drive bit demand. As production of the higherdensity
part rises, IDC forecasts 2Gb unit
shipments to almost rival that of 1Gb next year;
by 2012, 2Gb would be the mainstream part. As
such, per Gb ASP differential between 1Gb and
2Gb parts is to close next year. IDC expects 1Gb
and 2Gb ASPs to fall c40% and c50% y-o-y
respectively (IDC forecasts blended DRAM ASPs
to decline c25% next year).
In this light, leading DRAM vendors that are rapidly
shifting production of 2Gb (and to a far lesser extent
for now, 4Gb) parts to 4xnm and below from 5xnm
will benefit the most from this trend. Samsung and
Hynix raised targets for 4xnm proportions to over
60% and 50% respectively by end-2010, we suspect
most of the new process would be geared to 2Gb
parts. We expect 4xnm and below to account for
close to all of Samsung’s DRAM production by the
end of next year. The availability of lower-power,
high performance DRAM should accelerate as
Samsung and Hynix start mass production on the
3xnm process during 2H10.
19

TMT
Memory Devices Technology
8 October 2010
abc
High-end computing DRAM MB bn, % of total DRAM demand
PC server DRAM MB bn, % of total PC DRAM demand
500
12%
400
20%
400
300
200
100
11%
11%
10%
10%
9%
300
200
100
15%
10%
5%
-
9%
-
0%
2009 2010f 2011f 2012f
2009 2010f 2011f 2012f
High-end com puting
% of total
x 86 serv ers
% of total PC
Source: IDC, HSBC estimates
Source: IDC, HSBC estimates
Corporate and server demand
Our regional analyst Wanli Wang forecasts 2011
desktop PC unit growth rates to fall to +6% y-o-y
from 8% in 2010; however this is still better than
the annual decline in 2008/09. Corporate PC
replacement demand should continue to support
desktop PC unit growth. In notebooks, we expect
1.35V DRAM to migrate towards notebooks next
year especially as ULV (ultralow voltage)
processors become more ubiquitous. We
understand that some PC makers are asking for
even lower 1.20/1.25V parts already.
Servers – niche, but very high ASP market
More interestingly perhaps in an often overlooked
segment, is demand for higher-end DRAM from
servers. We expect continued corporate spend plus
ongoing growth of cloud computing to drive strong
demand from this segment. Demand will also be
catalysed by availability of high-density (2Gb and
4Gb very soon), low-power DRAM modules.
Micron expects robust demand from major
customers to continue into 2H10 for servers, a
sentiment echoed by Hynix and Samsung. We
forecast server unit growth of c10% y-o-y) in
2011e, but higher content/box growth of c38%.
Price premiums are driven by high-density, high
performance and lower power consumption, in
addition to higher production costs. Server
memory content/system in general tends to double
every two years. High-end servers use DDR3; we
expect migration to lower voltage, i.e. 1.35V,
compared to 1.5V which is standard currently, to
reduce power consumption, driving further highend
DDR3 demand.
Server memory modules are usually ECC RDIMM.
From a cost perspective, these modules are far more
expensive than regular modules because of 1)
cutting-edge capacity memory chips which are more
expensive due to production yields and volumes, 2)
ECC memory uses 9 chips instead of 8 for each
memory bank, 3) lower module volumes, and 4)
additional components required to support buffering.
Samsung started mass production of 32GB server
grade DRAM modules in 2Q10 using 4Gb parts
produced on 4xnm process, though these are likely
to be customer samples until 4Q10. Each server then
could hold up to 384GB with little power
consumption difference or 192GB with 40% lower
consumption than for modules using 2Gb parts.
Servers (including PC servers) account for about
10% of total global bit demand, but the blended
ASP impact is far higher for vendors, at c2.5x for
a 4GB module up to c6x for a 16GB module.
20

TMT
Memory Devices Technology
8 October 2010
abc
Tablet PC DRAM MB bn, % of total DRAM demand
150
100
50
-
2009 2010f 2011f 2012f
3.0%
2.5%
2.0%
1.5%
1.0%
0.5%
0.0%
PC unit shipments, including tablet
600
500
400
300
200
100
0
2009 2010f 2011f 2012f
Tablets
% of total
DT PC NB PC Tablet
Source: HSBC estimates
Source: IDC, HSBC estimates
Tablets – about 5% global DRAM
cannibalisation
Tablets are a good thing, we concluded for NAND
demand. Unlike netbooks, tablets by their very
nature demand NAND as the primary storage
device. But an emerging concern for commodity
DRAM is that tablets will cannibalise netbook and
notebook sales, reducing overall DRAM demand.
How big is the threat? On the face of it, tablet’s
potential impact on notebook demand could be
substantial; Best Buy estimates that the iPad
cannibalised laptop PC sales by up to 50% in its
last reporting quarter. In our forecast, we calculate
1%, 5% and 5% cannibalisation impact on global
DRAM demand for 2010-12e. This could rise if
cannibalisation reaches the sort of levels that Best
Buy is eluding to. DRAM vendors that can supply
low-voltage, low-power DDR (and increasingly
DDR2) that could be integrated into tablets stand
to benefit, or could offset some of the declines in
commodity DRAM demand.
In terms of units, we estimate tablets to account
for c7.6% of tablets plus notebooks this year. We
forecast tablet shipments to soar to 73m units next
year, accounting for a quarter of our forecast
tablets plus notebooks. We assume netbooks and
CULV notebooks will see negative unit shipment
growth, and that regular notebook growth will
slow to c16.7% y-o-y.
We estimate blended notebook plus netbook
DRAM content/box of 2.7GB this year, with
netbooks averaging a little over 1GB. In contrast,
Apple’s iPad, Samsung’s Galaxy Tab and
Toshiba’s Folio 100 have paltry DRAM content –
256MB, 512MB and 512MB respectively. The
cannibalisation impact will really start to be felt
next year, when we estimate tablets will account
for 26% of tablet plus notebook units but only
c6% of tablet plus notebook DRAM bits. We
estimate the impact of tablet cannibalisation to be
around 1% of total global DRAM demand this
year, but rising to c5% for 2011-12e. In this
estimation we used the following assumptions:
We estimate tablet unit shipments to rise to
73m units next year and 102m units by 2012e.
HSBC’s regional hardware analyst, Wanli
Wang, expects ten vendors to launch tablet
products next year. He expects notebooks to
drive PC demand in 1H and tablets to drive
demand in 2H. ARM estimates there will be
40 tablet devices using ARM-based
processors this year.
We expect DRAM content in tablets to start
rising quickly next year to support multitasking
and more resource-intensive apps. We
think this is doable, given DRAM is only c3%
of BOM for current Apple iPads. We think 5-
6% is more likely next year as 512MB
21

TMT
Memory Devices Technology
8 October 2010
abc
becomes standard (as with the Samsung
Galaxy Tab).
By 2012 we expect tablet DRAM would
reach above 1GB per system, catalysed by
Windows and competitive pressure amongst
tablet vendors to differentiate their products.
Intel’s Oak Trail processor – which allows
full Windows 7 to run on tablets (and
netbooks) – was announced in June 2010 and
is set to appear in devices starting 1H11. If we
assume Windows 7 Basic is to run on a tablet,
it would realistically require 1GB to run.
We expect other operating systems to drive
DRAM demand over 2011- 12e as well to
support complex apps as processing power
rapidly increases.
22

TMT
Memory Devices Technology
8 October 2010
abc
Bit shipments for handsets, % of total DRAM demand, MB bn
Smartphone units and MB/system growth
500
400
300
200
100
-
12%
10%
8%
6%
4%
2%
0%
600
500
400
300
200
100
-
500
400
300
200
100
-
2009 2010f 2011f 2012f
2009 2010f 2011f 2012f
Smart phone units (m)
Telecoms
% of total
Blended MB/sy stem (RHS)
Source: IDC, HSBC estimates
Source: IDC, HSBC estimates
Handsets
Relevance: Telecoms is becoming a major
consumer of DRAM, accounting for c9 % of DRAM
bit demand this year and growing to c10% in 2012e
at a CAGR of c57% during 2010-12e. Within
telecoms, smartphones are a major driver of demand
growth. Mobile DRAM prices tend to be higher on a
bit basis, as memory vendors need to work closely
with mobile phone makers to incorporate custommade
solutions for specific handsets.
Upside/downside factors. Upside factors include
launch of higher-specification smartphones due to
competitive pressures, more demanding apps and
operating systems including Android
‘Gingerbread’. Downside factors include high
mDRAM ASPs suppressing the inclusion of more
content/system.
Catalysts. Better than expected smartphone unit
shipment and product mix guidance/results from
major handset manufacturers.
Mobile DRAM
We think smartphones are driving rapid mobile
DRAM bit growth through both unit shipment
growth and content per box. Mobile phones have
had a very limited amount of DRAM in the past.
In our view, the development of smartphones is
now mirroring that of PCs, driven by resourceintensive
wireless broadband, multimedia and
multitasking. As more demanding apps become
the norm, we expect rapid advances in processors,
screens, volatile and non-volatile storage. Mobile
DRAM, which currently uses mostly DDR1, is
migrating rapidly to DDR2 to support higher
speed processors now found in smartphones.
We forecast mDRAM bit growth to rise c102%,
92% and 112% y-o-y for 2010-12e. We assume
DRAM content/system for smartphones to rise
c31%, c39% and 57% for 2010-12e, to around
390MB by 2012e. We forecast smartphone unit
shipments to rise 54%, 39% and 35% during the
periods, accounting for c38% of total handsets
shipped globally by 2012e (2009: 15%).
At the cusp of an explosion in mDRAM.
Like desktop PCs a decade ago, broadband
connectivity, multitasking, more complex
applications and more powerful graphics are
driving demand for faster processors. We
expect smartphone makers to attempt to
differentiate their products not only through
different operating systems, but ability of
handsets to run applications quickly.
Smartphone DRAM to surge to support
computing power. Qualcomm’s Snapdragon,
Samsung’s Hummingbird, Apple’s A4
processor and other ARM-based SoCs run up
to 1GHz already. We believe smartphones
23

TMT
Memory Devices Technology
8 October 2010
abc
processing power will increase rapidly, in part
to support requirements for next-iteration
operating systems such as Android 3.0
(Gingerbread), which will require 1GHz
processors, 512MB DRAM and a 3.5” screen
as minimum hardware requirements. Highend
processors including Qualcomm’s dual
core Snapdragon, NVIDIA’s next generation
Tegra with dual cores and fast GPU (running
DDR2 as opposed to DDR), and Samsung’s
‘Orion’ dual core processor SoC are set to be
become commonplace by end-2011e.
c1GB by end-2011 in higher-end smartphones
is likely, we think. Apple’s iPhone 4G and
Samsung’s Galaxy S, as well as the now
discontinued Google/HTC Nexus One already
have 512MB of onboard mDRAM.
Smartphone DRAM usage, % of total handsets
250
80%
200
60%
150
40%
100
50
20%
-
0%
2009 2010f 2011f 2012f
GB m % of total handsets
Samsung, Elpida and Hynix are leaders in
mDRAM The higher-end mDRAM market (i.e.
128MB upwards) is dominated by major DRAM
players that are capable of customising MCP
(multichip packages) for handset vendors.
DRAMeXchange estimates mobile DRAM rose to
c12% of Hynix’s product mix and c20% for
Samsung.
Several vendors have low power mobile DDR2
(LPDDR2) in mass production at 5xnm process
and above, and some have 4xnm process parts in
engineering or customer sample stages. The latter
development should deliver significant
competitive advantage. With the 4xnm Hynix part,
power consumption was reduced to c30% of
50nm process produced mobile DDR2. Samsung
already ships up to 2Gb LPDDR2 produced at
56nm; its 46nm 2, 4 and 8Gb LPDDR2 parts are
currently at the engineering sample stage; we
believe mass production could start in 2H10.
Hynix completed development of its 2Gb
LPDDR2 using 44nm node technology in January
2010 and is currently at the customer sample stage.
Elapid’s 2Gb LPDDR2 DRAM produced on
4xnm process started sampling in August 2010,
and mass production is slated to start this month
(September). Micron/Nanya is to mass produce its
2Gb LPDDR2 in 3Q10.
Source: IDC, HSBC estimates
24

TMT
Memory Devices Technology
8 October 2010
abc
Demand drivers – NAND
Remain bullish on NAND; we forecast 2010-12e NAND demand
bit CAGR of 103%
Demand is structural and persistent, as heavy online and offline
growth in media drives portable device storage requirements
Our positive outlook on tablets and smartphones is diluted by
cannibalisation of media players
Summary
Relevance, growth rates
The non-volatile memory (NVM) segment – led
by NAND – remains an exciting high-growth
segment. We forecast 2010-12e NAND bit
demand CAGR of c103% (compared to supply
CAGR of 100%), rising 114% next year. Global
appetite remains insatiable for storage capacity
that is fast, robust and low-power.
NAND demand forecast
45,000
120%
100%
30,000
80%
60%
15,000
40%
20%
-
0%
2009 2010f 2011f 2012f
Demand (GB, m) Change, y -o-y , q-o-q (%)
Source: IDC, HSBC estimates
Foreseeable demand remains strong on rising
affordability, expanding applications in high
potential markets such as smartphones, tablets, PC
storage plus nascent markets such as automotive
applications. We continue to believe the risk is on
the upside, as demand accelerates on more rapid
affordability trends. By 2012e we expect global
NAND revenues to recover to c54% of DRAM’s,
up from this year’s estimated 38%.
Unlike DRAM, NAND demand remains strong
for commodity parts and embedded solutions. The
main embedded demand driver is handsets, led by
smartphones, which we estimate will account for
37% of total bit demand next year, up from 28%
in 2009. Indeed, c60% of Toshiba’s NAND
revenue is from embedded solutions that include
mobile and other consumer electronics. Another
embedded solutions driver is tablets, a segment
which practically did not exist until April this year.
For commodity NAND, we are set to see serious
demand arising from not only SSDs for regular
PCs, but perhaps more importantly, for servers,
which we think is a market that is under-estimated
in terms of its potential. In this report we highlight
the upside risk from corporate server storage
systems. We expect SSD growth, which had been
stalled this year by high ASPs to see rapid growth
due to falling NAND prices and indirect demand
stimulation through the ‘tablet effect’.
25

TMT
Memory Devices Technology
8 October 2010
abc
Global NAND demand & supply
2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e
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
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
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
Change, y-o-y, q-o-q (%) 41.1% 71.0% 114.1% 86.0% 4% 24% 17% 16% 18% 27% 27% 17%
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
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
Change, y-o-y, q-o-q (%) 45.5% 72.6% 113.8% 92.6% 3% 16% 25% 18% 26% 16% 25% 18%
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
Change, y-o-y, q-o-q (%) -3.5% 37.0% 31.7% 16.1% -5% 14% 4% 3% 4% 15% 11% 2%
NAND/DRAM revenue (%) 49% 38% 49% 54%
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
Change, y-o-y, q-o-q (%) -32% -20% -38% -38% -9% -8% -11% -11% -12% -10% -13% -13%
Wafer cap. (300mm equ) 9,194 10,476 12,529 15,129 829 867 878 918 948 978 1,095 1,155
Change, y-o-y, q-o-q (%) -5% 14% 20% 21% 1% 5% 1% 5% 3% 3% 12% 5%
Source: Company data, HSBC estimates
Upside/downside risks
Key upside risk to our forecasts is a long-term,
controlled growth of the industry where bit
growth remains strong, and ASP declines are
relatively controlled. Downside demand risks are
limited for NAND, but include slower than
expected adoption due to slow ASP declines
resulting from tight supply-demand; price
elasticity is critical especially for devices such as
SSDs. The key supply downside is irrational
competition leading to rapid ASP deflation.
Catalysts – short term, long term
Strong bit growth from makers, stronger than
expected NAND contract prices.
Better than expected unit shipment and
product mix guidance/results from handset
vendors and USB/flash storage makers such
as SanDisk, A-Data, and PQI.
Short-term catalysts include strong sales of
various Apple iPad devices, launch and strong
sales of similar tablet devices by major vendors
including Samsung, Toshiba and RIMM.
NAND bit demand profile, 2009
NAND bit demand profile, 2012e
DVC
SSD
3%
5%
USB Flash
Driv e
11%
Handset
28%
Others
7%
Digital S till
Camera
20%
Media
Play er
26%
DVC
2%
Tablets
12%
SSD
17%
USB Flash
Driv e
8%
Others
5%
Digital Still
Camera
7%
Media
Play er
6%
Handset
43%
Source: IDC, HSBC estimates
Source: IDC, HSBC forecasts
26

TMT
Memory Devices Technology
8 October 2010
abc
Global NAND demand breakdown
GB equ. units 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e
Digital Still Camera 1,176 1,520 2,134 3,092 324 355 417 424 469 502 569 594
Media Player 1,504 1,923 2,502 2,560 354 441 534 593 541 588 681 692
Handset 1,638 3,350 7,920 17,472 596 762 898 1,094 1,423 1,824 2,252 2,421
USB Flash Drive 644 1,175 2,063 3,375 221 256 321 377 389 449 563 662
SSD 270 693 2,420 6,714 102 147 199 265 352 476 702 959
Tablets - 496 2,784 4,902 - 93 130 273 458 572 801 953
DVC 175 274 517 886 52 60 75 88 97 112 141 166
Others 400 590 1,084 2,251 111 128 161 189 204 236 296 348
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
y-o-y, q-o-q (%) 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e
Digital Still Camera 29% 29% 40% 45% -12% 10% 17% 1% 11% 7% 13% 4%
Media Player 23% 28% 30% 2% -20% 25% 21% 11% -9% 9% 16% 2%
Handset 49% 105% 136% 121% 26% 28% 18% 22% 30% 28% 23% 8%
USB Flash Drive 46% 82% 76% 64% 9% 16% 25% 18% 3% 16% 25% 18%
SSD 230% 157% 249% 177% 33% 44% 35% 33% 33% 35% 47% 37%
Tablets na na 461% 76% na na 40% 110% 68% 25% 40% 19%
DVC 150% 57% 88% 71% -6% 16% 25% 18% 11% 16% 25% 18%
Others 138% 47% 84% 108% -12% 16% 25% 18% 8% 16% 25% 18%
Total 46% 73% 114% 93% 1% 27% 22% 21% 19% 21% 26% 13%
Demand mix 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e
Digital Still Camera 20% 15% 10% 7% 18% 16% 15% 13% 12% 11% 9% 9%
Media Player 26% 19% 12% 6% 20% 20% 20% 18% 14% 12% 11% 10%
Handset 28% 33% 37% 42% 34% 34% 33% 33% 36% 38% 38% 36%
USB Flash Drive 11% 12% 10% 8% 13% 11% 12% 11% 10% 9% 9% 10%
SSD 5% 5% 11% 16% 6% 7% 7% 8% 9% 10% 12% 14%
Tablets 0% 5% 13% 12% 0% 4% 5% 8% 12% 12% 13% 14%
DVC 3% 3% 2% 2% 3% 3% 3% 3% 2% 2% 2% 2%
Others 7% 6% 5% 5% 6% 6% 6% 6% 5% 5% 5% 5%
Total 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100%
Unit shipment (m) 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e
Digital Still Camera 130 132 141 149 32 32 34 33 35 35 37 35
Media Player 165 120 100 80 30 29 31 30 25 25 26 25
Handset 1,116 1,244 1,350 1,457 284 298 321 342 316 332 346 356
USB Flash Drive 184 250 275 300 61 61 65 62 68 67 72 68
SSD 14 19 30 51 3 4 5 6 6 7 8 9
Tablets - 16 73 102 - 3 4 9 12 15 21 25
DVC 12 14 18 21 4 4 4 4 4 4 5 4
Others 101 119 155 210 29 29 31 29 38 38 41 38
GB/system 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e
Digital Still Camera 9 12 15 21 10 11 12 13 14 15 15 17
Media Player 9 16 24 32 12 15 17 20 22 24 26 28
Handset 1 3 6 12 2 3 3 3 5 6 7 7
USB Flash Drive 4 5 8 11 6 7 8 9 9 10 13 14
SSD 19 37 81 132 30 35 39 44 55 68 90 110
Tablets - 31 38 48 - 31 31 31 38 38 38 38
DVC 15 19 29 43 15 17 20 25 22 25 30 37
Source: Company data, HSBC estimates
27

TMT
Memory Devices Technology
8 October 2010
abc
% of total NAND bit demand from handsets System unit and GB/system growth for handsets
50
40
30
20
10
0
2,000
1,500
1,000
500
0
15
10
5
0
2009 2010f 2011f 2012f
2009 2010f 2011f 2012f
Handset
Sy stem units (m)
GB/sy stem
Source: IDC, HSBC estimates
Source: Company, HSBC estimates
Handsets
Relevance: Mobile handsets have emerged as the
leading usage of NAND thanks to a proliferation
of smartphones. Handsets account for 37% of
2011e total bit demand forecast (2009: 28%).
Other than nascent tablets and SSDs, mobile
handsets is the fastest growing segment also,
growing at a 10/12e CAGR of 128% to account
for c42% of total NAND bit demand by 2012e.
What’s changed in our view? We incorporate
our upwardly revised global handset and
smartphone forecasts.
Upside/downside factors. Watch for Apple – the
leading indicator for mobile device NAND usage
– releasing larger capacity iPhones and iPods.
Upside risks include higher than expected take up
of smartphones in emerging markets especially on
3G network roll out, higher than expected
content/system in smartphones for developed
markets. Downside risks include slowing average
GB growth on smartphones as lower-spec models
(low NAND content) become a higher proportion
of total smartphones.
System units, GB/system both rise
rapidly
Growth in NAND bit demand is driven by rising
GB/system as smartphone proportion of total
handsets rises. Also the amount of storage on
smartphones themselves is rising, driven by
higher resolution content.
Smartphone proportion rises
We forecast smartphones to rise to account for
c38% of total global handset units by 2012e, up
from c15% last year.
Smartphone demand will continue to grow as
model variety expands and functionality
becomes more mature.
We expect greater penetration of smartphones
in emerging markets on better affordability,
and roll out of 3G networks in markets such
as China and India, driving new demand.
28

TMT
Memory Devices Technology
8 October 2010
abc
Handset and smartphone proportion
2,000
1,500
1,000
40%
30%
20%
Higher resolution video, games content is
becoming widespread and games are being
written for higher-resolution devices. Real
time HD is still highly restricted by
availability and pricing of bandwidth.
500
10%
-
0%
2009 2010f 2011f 2012f
Handsets Smart phones/total handsets
Source: Company, HSBC estimates
Content/system rises
Growth in average GB/system will remain strong
for the foreseeable future. We forecast GB/system
for smartphones to double each year during 2011-
12e driven by:
Lower NAND prices – higher bit supply
growth led by Toshiba and Samsung on
increased wafer capacity, rapid process
migration and adoption of TLC should drive
ASPs to fall by c38%/year for 2011-12e. We
estimate NAND on a 16GB Apple iPhone 4 to
account for 14% of BOM, a level which we
think will be maintained going forward. We
assume most non-Apple smartphones will use
less NAND than Apple to lower prices,
obliging the consumer to buy flash cards to
fulfil extra storage needs.
Availability of larger capacity mobile
NAND. We expect Apple uses 1x 32GB
NAND module for its iPhone 4 and 2x 32GB
NAND modules for the 64GB iPod Touch.
Rapidly rising screen resolutions.
Smartphones are increasingly converging on a
screen size 3.5-4” and WVGA (800x480)-
level screen resolution including Apple
iPhone 4 (960x 640), Samsung Galaxy S and
HTC Desire. HD and gaming support rise far
better graphics performance on latest SoCs).
Cannibalisation of media players. Longerterm
NAND content growth in developed
markets will to a large extent depend on how
far smartphones substitute media players.
Consumers currently buy flash cards to
supplement onboard storage.
Smartphone NAND usage
20,000
15,000
10,000
5,000
-
Source: Company, HSBC estimates
2009 2010f 2011f 2012f
Total smart phone (GB m)
% of total NAND demand
Smart phones/total handset NAND
100%
80%
60%
40%
20%
Low-end smartphones also drive NAND
growth. We expect proliferation of low- to midend
smartphones and rising penetration of
smartphones in emerging markets could
incrementally drive NAND demand in excess of
expectations. In particular, handset vendors such
as Huawei, ZTE, LG Electronics and others are
more than likely to chase greater global market
share with more affordable smartphones.
We believe white box makers are looking at
USD100-120 for very basic smartphones. Huawei
has mentioned previously that USD150 is the
pricing point required for smartphones to drive
potential 3G usage in China. We think a low-end
smartphone with a BOM of cUSD80 could ship
with up to 8GB of internal NAND by next year. In
this case, the NAND cost would equate to 10-16%
0%
29

TMT
Memory Devices Technology
8 October 2010
abc
of BOM, which we think is a reasonable level for
such vendors to absorb.
Smartphone NAND content/system growth will
persist for the foreseeable future as smartphone
development mirrors that of PCs. Multi-tasking,
multimedia and wireless broadband (3G in
emerging markets and 4G in developed) are very
powerful drivers for NAND and DRAM demand.
As operating systems and applications become
more demanding on systems requirements (just
look at Android 3.0), advances in hardware
specifications – processor, screen, volatile and
non-volatile storage – are inevitable.
Effects of tablet cannibalisation on
Samsung
Purely from a memory perspective, we estimate
the net-net impact of tablet cannibalisation on
Samsung Electronics’s profits to be minor next
year. In drawing this conclusion we assumed that
1) tablets shave 5% off global DRAM demand for
2011-12; 2) tablets account for 5% of NAND
demand this year rising to c13% next year, most
of which is incremental additional demand. The
minimal net-net effect would obviously not be the
case for other memory vendors, which have
different product mixes.
We also assessed the broader impact on Samsung
Electronics of tablet cannibalisation of netbooks
and notebook PCs. We think the impact varies
widely depending how much of a tablet or
notebook’s components Samsung supplies. The
impact also depends on whether tablets
cannibalise notebooks or netbooks:
Either way, proliferation of tablets should be
positive for Samsung, as tablets increase the
opportunity for the firm to supply more of the
costly parts such as application processors
and NAND that constitute these devices.
likely benefits every time a consumer buys a
tablet instead of a netbook.
If DRAM prices fall to cUSD30/2GB module,
then Samsung's chunk of operating profit
from a notebook falls to cUSD24, close to
that derived from a tablet.
Apple iPad versus notebook and netbook costs, OP for SEC
iPad (32GB) Cost Supplier(s) OPM OP
App processor 15 Samsung 15% 2
NAND 51 Samsung/Toshiba 28% 14
mDRAM/NOR 9 Samsung/Elpida 42% 4
9.7" LED backlit LCD 100 LGD/Samsung 15% 15
Total 175 Total 35
Notebook (4GB) Supplier(s) OPM OP
DRAM (2x2GB DDR3) 72 Samsung et al 38% 27
HDD(250GB, 5400rpm) 30 Samsung et al 5% 2
14" LED backlit LCD 50 Samsung et al 11% 7
Total 152 Total 36
Netbook (2GB) Supplier(s) OPM OP
DRAM (1x2GB DDR3) 36 Samsung et al 40% 14
HDD(160GB, 5400rpm) 28 Samsung et al 5% 1
10" LED backlit LCD 21 Samsung et al 11% 6
Total 85 Total 22
Source: HSBC estimates
Base assumptions:
If we assume that Samsung supplies all of the
apps processor, NAND, mDRAM, NOR and
LCD for a 32GB iPad, we estimate its
operating profit per iPad at cUSD35. This is
similar to our estimate operating profit of
USD36 for a regular 4GB notebook if
Samsung supplied all the DRAM, HDD and
LCD screen. For a netbook, operating profit
possible is much lower at cUSD22.
If we assume that Samsung does not supply
screens but all the other components
mentioned earlier in tablets (SoC, NAND,
mDRAM/NOR) and notebooks (DRAM,
HDD), then Samsung's operating profit on a
tablet could reach cUSD20 compared to
cUSD29 for a 4GB notebook.
If tablets cannibalise netbooks, which appears
the most likely in our view, then Samsung
30

TMT
Memory Devices Technology
8 October 2010
abc
% of total NAND bit demand from tablets System unit and GB/system growth for tablets
15
10
5
0
120
100
80
60
40
20
0
60
50
40
30
20
10
0
2009 2010f 2011f 2012f
2009 2010f 2011f 2012f
% of NAND demand
Sy stem units (m)
GB/sy stem
Source: HSBC estimates
Source: HSBC estimates
Tablets
Relevance: Tablets represent a major new NAND
application that has explosively entered the scene.
Tablets by default use flash storage due to
NAND’s inherent advantages of robustness, small
form factor, longer battery life and ability to
support instant-on. We expect tablets to account
for 13% of global NAND demand next year (2010:
5%), mostly driven by a massive 461% 2010/12e
bit growth rate from a standing start this year.
Explosive unit growth
We estimate tablet unit shipments to rise to 73m
units next year and 102m units by 2012e. The
iPad should the biggest bite of this pie, holding
about half of unit market share next year in our
view. Other vendors are set to launch in the near
future, led by Samsung. These include Androidbased
models such as Toshiba’s Folio 100,
RIMM’s ‘Blackpad’ (November 2010), LG
Electronics’ Optimus (end-2010), variations of
Dell’s Streak, Motorola (end-2010). Compal,
Acer, Foxconn, Prowave and MSI also plan
Android-based tablets for the near future. ARM, a
major mobile device application processor vendor,
estimates there will be about forty tablet devices
using its processors by the end of this year. The
flood of tablets will accelerate next year; HSBC’s
regional hardware analyst, Wanli Wang, expects
ten vendors to launch tablet products in 2011.
GB/system to rise slowly at first
We believe GB/system growth would initially be
relatively slow, rising c23% y-o-y next year and
27% in 2012e as tablets will mostly run compact
Android and iOS operating systems; there is more
than a bit of truth to the saying that tablets (for
now at least) are much bigger smartphones. But
the key difference is the larger amount of NAND
that tablets have compared to smartphones, and
hence we need to consider vendors’ attempts to
control BOM. For Apple’s iPad NAND as a
proportion of hardware-only BOM varies between
c10-30% depending on the model. This is a much
higher proportion than hard disk drive (HDD)
storage takes up in PCs and is comparable to
NAND as a proportion of BOM for smartphones.
PC vendors entering the tablet (or smartphone)
market need to completely readjust their
expectations for storage cost in their BOM.
But falling NAND ASPs will continue to support
rising content/system. For non-Apple tablets (and
smartphones), flash cards will be supplementary
storage devices, giving users the option to raise
storage as requirements grow. Even so, 16GB is
about the minimum acceptable onboard NAND
storage for a tablet in our view. Vendors should
have little problem meeting this requirement this
year. NAND price per GB averages USD1.63; a
white box tablet vendor with a BOM of USD200
31

TMT
Memory Devices Technology
8 October 2010
abc
could afford installing 16GB of onboard NAND
(13% of BOM, on par with Apple). Next year, the
BOM bar will be higher for tablet vendors, not just
due to ferocious competition between those late to
the party but also due to higher demands from OS
and applications. But as ASPs fall to cUSD1/GB,
32GB of NAND would account for c18% of BOM,
which we think is a reasonable level.
Impact of Windows 7 for tablet content/system
Interestingly, we think Windows 7 could drive
tablet GB/system. We think Windows could
mount a belated but serious challenge in the tablet
market due rising corporate usage of tablets, latent
demand for tablets that run Windows applications,
and development of tablet hardware that could
support resources-intensive Windows. Aimed
precisely at that potential market is Intel’s new
Oak Trail processor, which allows full Windows 7
to run on tablets. This processor is set to appear in
devices starting 1H11. Such tablets could include
MSI Wind Pad and Asus Eee Pad. But we think
the impact of Windows 7 on NAND content/
system would make a substantial impact by 2013.
For Windows tablets, we think:
There would be a larger NAND requirement.
Unlike the much more compact Android and
iOS, a full implementation of basic Windows
7 needs a bare minimum 16GB of storage.
We think 32GB is the minimum. In our view,
SSDs on netbooks failed because the 8-16GB
that was provided was simply insufficient to
support even basic OS and applications.
NAND storage for OS and apps would be
onboard rather than through flash cards due to
the latter's data transfer speed limitations.
For NAND, any cannibalisation impact on regular
notebooks is negligible as penetration of SSDs in
the latter is still very low.
32

TMT
Memory Devices Technology
8 October 2010
abc
% of total NAND bit demand from SSD System unit and GB/system growth from SSD
20
15
10
5
0
60
50
40
30
20
10
0
150
100
50
0
2009 2010f 2011f 2012f
2009 2010f 2011f 2012f
% of NAND demand
System units (m)
GB/sy stem
Source: IDC, HSBC estimates
Source: IDC, HSBC estimates
Solid state disk
Relevance: We expect NAND for solid state disks
(SSD) to account for c5% of NAND bit demand
this year, rising to a forecast c11% next year. For
commodity NAND, we see hefty demand arising
from not only SSDs for regular PCs, but also for
servers, which we think is a market that is underestimated
in terms of its potential.
Upside/downside factors. Upside risks include
accelerated adoption by high-end servers (on
improving SSD read/write performance), greater
penetration of ‘regular SSD’ amongst notebook
PCs and adoption by desktops. Another factor that
could start kicking in is SSDs in desktops, an
application we have advocated for sometime. In
fact, Apple appears to be adopting this idea with
the new iMac, which has the option to have both
SSD and HDD in each machine.
Downside risks include high NAND prices
delaying adoption across most applications. SSD
growth could also be limited by substitution from
portable flash drives as the performance gap
between internal and external drives narrow with
eSATA, USB3.0 and FireWire S1600 drives.
Catalysts. Catalysts include a spill-over effect for
SSD demand from tablets into mainstream
notebooks, as consumers increasingly awaken to
the convenience of SSDs. Newer OS – Windows
7 and Mac OS X – offer significantly better start
up/shut down times than before. But tablet users
will increasingly find these times intolerable.
SSD adoption to accelerate
Comparative affordability remains the hurdle
We expect SSDs to start taking off in earnest next
year, accounting for 11% of our total bit demand
forecast as absolute pricing of SSDs becomes
more affordable. In particular, we expect SSD
adoption to accelerate in notebooks and in the
corporate server storage market. We estimate the
cost of implementing 128GB of NAND to reach
USD128 this year, accounting for 20% of BOM
for a mid- to high-end notebook PC; by 2012 we
think this could fall to 13% of BOM, sufficient to
drive widespread adoption. As absolute pricing
becomes more attractive, SSD adoption risk is on
the upside. Consumers are likely to look for the
convenience of SSDs, and distribution channels
are likely to overcome negative experiences in
attempting to sell early (very expensive) SSD
products. Another factor that could start kicking
in, albeit slowly, is SSDs in desktops, an
application we have advocated. Apple recently
adopted this idea with the new iMac, which has
the option of having either SSD or HDD (or both)
in each machine.
33

TMT
Memory Devices Technology
8 October 2010
abc
NAND makers generally believe SSDs will take
off next year. Indeed, despite our argument that
absolute pricing of SSDs has now reached the
stage where more widespread adoption in
mainstream computing is possible, we believe
hesitancy in retail channels and per GB
comparisons by consumers has limited SSD
adoption to high-end notebooks, notably from
Sony, Apple, Dell (Adamo), HP (Envy) due to
high SSD prices on a per GB basis compared to
hard disks. SSDs up to 512GB are now available,
using 3xnm NAND, SATA interfaces and fast
data rates by using toggle-mode DDR NAND.
We expect SSD adoption to make faster inroads
starting next year based on falling NAND ASPs
and greater consumer awareness of the advantages
of SSDs, in part driven by soaring awareness of
tablets. We forecast SSDs to account for c16% of
total NAND demand by 2012e. SanDisk expects
SSD adoption to pick up in 2010, reaching c20%
bits shipped by 2012; the vendor expects market
acceleration in 2011e with cost reductions on the
2xnm process, with a ‘tipping point’ achievable
with 1xnm. Gartner forecasts SSD to approach
c20% of bit shipments in 2012.
Indeed, perhaps more than any other NAND
application, due to their very high capacities, SSD
adoption will depend on affordability, which is in
turn driven by technology advances.
Enterprise flash drives – huge upside
risk to NAND revenue forecasts
A major upside risk to our SSD forecasts is the
server storage market, offering disproportionately
high revenues and margins. Hynix sees enterprise
and high-end PCs driving demand for SSDs,
despite its scepticism over prospects for SSDs in
general this year. Key beneficiaries of this market
include Samsung, IMFT and Hynix, to be joined
next year by Toshiba. We admit that we do not
separately model the server NAND market in our
forecasts. But we hope to add value in this
analysis by illustrating the magnitude of the
upside risk to our forecasts.
SSD vendors continue to emphasize that the
enterprise market remains a key demand driver.
We estimate that server storage could account for
the equivalent of 12-17% of our 2012e forecast
global NAND revenue, and the equivalent of 20-
32% our forecast 2013e NAND global revenue.
These estimates assume SSDs account for 5% and
10% respectively of 2011-12e disk storage
systems and are based on SLC.
High-end server SSD storage is very expensive.
Despite this, data centres and corporates use SSD
due to 1) extremely fast data input/output,
especially important for cloud computing, 2)
small form factor (space), 3) low power
consumption. As ASPs fall, we expect SSDs will
replace HDDs as ‘front line’ storage, with HDDs
acting as inexpensive backup.
Server SSDs use SLC as it is generally about
twice as fast as MLC for read applications, and is
considered more reliable due to the fact it has only
one bit per cell. SLC NAND is more expensive at
about 3-3.5x premium compared to equivalent
MLC for 16-32Gb parts. Samsung currently
produces up to 64Gb density SLC.
By the time SLC NAND is incorporated into a
server storage system, prices/GB soar even more.
Server SSDs based on MLC NAND currently cost
about USD3-4/GB. But SLC-based storage
systems (not just simple SSDs) are about 10-25x
more expensive on a per GB basis. The high price
premium is justified by very high cost of using the
best high-density SLC NAND, but also reflects
systems that are optimised for high data transfer
speeds, high reliability and power efficiency, in
addition to close customer service. For example,
Oracle charges systems starting at cUSD46k with
480GB capacity (using 20 x 24GB flash modules)
using SLC, equating to cUSD96/ GB. Texas
34

TMT
Memory Devices Technology
8 October 2010
abc
Memory charges cUSD220k for a 5TB unit
(cUSD37/GB) for a lower spec system.
Micron produces SLC in 64-128Gb densities. It is
jumped into the enterprise-class SLC-based SSD
market in August this year based on 34nm process,
and is in customer sampling. Samsung and Hynix
have comprehensive SLC line ups, both offering
up to 64Gb densities. We believe Samsung uses
up to 5xnm processes for its SLC. More recently,
Hynix has been validating its 32nm-based SLC
(16Gb) since June 2010. Micron started mass
production of 32Gb
35

TMT
Memory Devices Technology
8 October 2010
abc
% of total NAND bit demand from DSC System unit and GB/system growth for DSC
25
20
15
10
5
0
160
150
140
130
120
25
20
15
10
5
0
2009 2010f 2011f 2012f
2009 2010f 2011f 2012f
% of NAND demand
Sy stem units (m)
GB/sy stem
Source: IDC, HSBC estimates
Source: IDC, HSBC estimates
Digital imaging (DSC, DVC)
Demand drivers in digital still cameras
Relevance. As a proportion of total NAND
demand digital imaging (still cameras plus digital
video cameras) is falling, not because of a decline
in usage – we forecast 2010/12e bit demand
CAGR of 49% – but mostly because of slowed
megapixel growth we believe. We forecast digital
imaging to account for c12.4% of total 2011e
NAND demand (2009: 23.3%).
Upside/downside risk. Demand recovery appears
relatively sound, with Nikon and others guiding a
recovery in unit shipments this year. Nikon for
example, guides the global DSLR market to reach
12.2m by end-March 2011, up 12% y-o-y, in part
supported by the emergence of mirrorless DSLRs.
In the compact DSC market Nikon a 7.1% y-o-y
rise in global units. Key upside risk is growth in
memory usage per system as HD video has
become standard in compact DSCs and DSLRs.
We expect 3D still and video and RAW-based
video would further boost demand starting 3-5
years hence.
Catalysts. Better than expected unit shipment and
product mix guidance/results from major device
manufacturers and flash card makers such as
SanDisk.
System unit growth low, but GB/
system to grow steadily
The dynamics of NAND demand from digital
imaging is, in our view, mostly driven rising
content/system, which benefits from convergence
of still and video, migration of video to solid state,
and high definition.
Rapid camera shipments recovery
Camera shipments to continue recovering next
year; we forecast global digital cameras units to
rise 10.7% y-o-y after a c13.8% fall last year.
DSLRs lead growth this year with 13.5%
compared to c10.5% unit shipment growth for
compact DSCs.
Global DSC unit shipments (m)
170
150
130
110
90
70
50
2007 2008 2009e 2010f 2011f 2012f 2013f
Source: CIPA, HSBC estimates
Compact
DSLR
In particular, we estimate mirror-less DSLR to
account for c13% of total 2010 DSLR shipments
36

TMT
Memory Devices Technology
8 October 2010
abc
(which is similar to Nikon’s estimate of c1.2m
units to March 2011, excluding Samsung).
GB/system growth diverges
We forecast digital still camera GB/system to rise
c30% and 37% for 2011-12e. In comparison we
expect digital video GB/system to rise 50% per
year for the same period. We believe the biggest
beneficiaries of this trend will be flash card
vendors. Even for digital video the trend is
irreversibly for non-linear, high-speed storage.
For digital still cameras, video convergence
drives demand. In the past six months high
definition video recording (720-1080p) – which
is very data storage-intensive – has become
standard on most compact cameras and DSLRs.
As a rough comparison, a 16GB flash card can
store over 3,300 12MP JPEG pictures, but only
3 hours of 720p AVCHD video.
Tape-less camcorders prevail. Both
consumer and professional level camcorders
have transitioned rapidly to flash-based
recording in the past 12 months as the
convenience of flash wins out and
affordability of flash cards increases.
HD technology maturing. Helping HD video
adoption on both DSCs and camcorders is the
maturity of a key compression technology,
AVCHD (needs less storage space than other
compression technologies).
Slowing camera resolution growth drags
on content/system. Still camera vendors now
focus on image quality, de-emphasising as a
key selling point absolute pixel counts. We
expect steady resolution growth in the next
few years, mostly in larger sensor DSCs as
small sensors are limited by intrusive noise at
higher pixel counts.
Cannibalisation by mobile phones is
limited, for now; megapixel growth of builtin
cameras on mobile phones have slowed
substantially, partially due to handset
vendors’ desire to limit component costs. But
HD video functionality is rising as sufficient
smartphone processor power to support it
becomes available.
We expect adoption of video on DSCs to
accelerate with rising adoption of mirrorless
SLRs that use larger (APS-C sized) sensors.
Indeed, Sony and Panasonic are at the
vanguard, launching high-end amateur/low
end professional video equipment based on
compact camera lens mounts.
Longer term, i.e. starting 3-5 years from now,
we expect 3D and RAW-based video to start
to drive higher GB/system for both still and
video imaging.
37

TMT
Memory Devices Technology
8 October 2010
abc
% of total NAND bit demand from USB flash drives System unit and GB/system growth for USB flash drives
12
10
8
6
4
2
0
400
300
200
100
0
12
10
8
6
4
2
0
2009 2010f 2011f 2012f
2009 2010f 2011f 2012f
% of NAND demand
Sy stem units (m)
GB/sy stem
Source: IDC, HSBC estimates
Source: IDC, HSBC estimates
USB flash drive
Relevance: Universal Serial Bus (USB) flash
drives (NAND storage plus integrated interface)
accounted for c11% of NAND bit demand last year,
but growing at a CAGR of 69.5% during the period.
Upside/downside factors. Key upside risk to our
forecast would be faster than expected adoption of
USB3.0, and strong price elasticity should NAND
prices decline. Key downside risks are the
uncertainty over Intel’s adoption rate and high
NAND prices suppressing GB/system growth.
Catalysts. Better than expected unit shipment and
product mix guidance/results from major device
manufacturers and USB storage makers such as
SanDisk, A-Data, and PQI.
USB3.0 – key multi-year driver
starting 2011
Convenience, durability, increasingly higher
capacity drives, and affordability support demand.
Rapid data rates. USB3.0 drives can transfer
data 5-10 times faster than USB2.0 drives; a
600MB movie file takes c5 seconds to download.
Higher-end USB3.0 flash drives start to rival
SSDs in data speeds, enabling external storage of
very large data amounts due to fast transfer times.
Launch of other high-capacity standards for
flash cards will help USB3.0 adoption. Secure
Digital Extended Capacity (SDXC) products
recently launched, addressing up to 2,048GB
(2TB). More recently, CompactFlash Association
released CF5.0 standards, which can address 147
million GB, compared to the current 137GB.
38

TMT
Memory Devices Technology
8 October 2010
abc
Time taken to read a 600MB movie file
40
30
20
10
0
USB 2.0
drive
Source: Companies, HSBC
USB 3.0
drive
Time taken (s)
SSD driv e
(regular)
SSD drive
(X25)
Product availability. USB3.0 flash drives started
launching at end-2009, including those by Super
Talent and PQI. Vendors including Gigabyte,
Asus and MSI (motherboards), Genesys (device
controllers) VIA (hub controller) have launched
USB3.0-standard products this year.
Data rate comparison for common PC interfaces/storage
Interface/device Launch Data rate (MB/s)
USB
1.0 1Q96 0.2
1.1 3Q98 1.5
2.0 2Q00 60
3.0 4Q09 600
Wireless 4Q06 14-60
But a key factor in the pace of USB3.0 adoption will
be Intel’s push (or lack thereof) for the standard; the
firm’s support is critical in device makers’ adoption
of the standard. Intel delayed roll out of chipsets
supporting USB3.0, but in September 2010 Intel
confirmed it would include support for USB3.0 in its
new chipset that would start shipping in the next 2-3
quarters. AMD is likely to offer chipsets that support
USB 3.0 by 1Q11.
Increase in capacities, affordability. We
forecast NAND ASPs to fall to cUSD1/GB by
2011. This means a 64GB USB3.0 flash drive
could cost cUSD70. Ramp-up of USB3.0 flash
drives demand will be faster than for USB2.0
drives due to better affordability of NAND. Bit
demand from flash drives took c7 years to surge
after USB2.0 was launched in 2Q00; it took that
long for affordability and widespread adoption of
USB2.0. Currently, up to 128GB flash drives are
widely available; these are expensive and data
rates are relatively slow.
IEEE1394
FireWire 400 1995 50
FireWire 800 2Q05 100
FireWire S1600 4Q10 200
FireWire S3200 4Q10 400
eSATA 2004 188
SSD drives
Samsung 256GB MLC 200/160
Intel 160GB MLC 250/100
USB flash drives
Corsair Flash Voyager GTR 34/28
Ironkey secure drive 27/24
Source: Companies, HSBC
39

TMT
Memory Devices Technology
8 October 2010
abc
% of total NAND bit demand from media players System unit and GB/system growth from media players
30
25
20
15
10
5
0
200
150
100
50
0
40
30
20
10
0
2009 2010f 2011f 2012f
2009 2010f 2011f 2012f
% of NAND demand
Sy stem units (m)
GB/sy stem
Source: IDC, HSBC estimates
Source: IDC, HSBC estimates
Media players
Relevance: NAND for media players accounted
for c25.9% of NAND bit demand last year; we
expect this proportion to fall to c6% by 2012e.
We think that media players as a segment, despite
launch of newer products by Apple and others is
being gradually absorbed by smartphones and
going forward especially, tablets.
Upside/downside factors. Upside risks include
appearance of dedicated, full HD and 3D media
players. Downside risks include faster than
expected substitution by smartphones and tablets.
Catalysts. Catalysts for NAND demand by media
players include a ‘reinvention’ of media players to
include web surfing.
Substitution by smartphones will accelerate, as:
Smartphone screens become larger and higher
resolution – Google’s Nexus One and has a
800x480 screen, as do others.
Smartphones become faster. The latest
generation of smartphones such as the iPhone
and Samsung Galaxy S runs 1GHz mobile
processors with high-speed graphics
processors.
The segment is set to be cannibalised by tablets
represented by such as Apple’s iPad, and Android
-based such as Samsung’s Galaxy Tab, which
offer the huge flexibility of running a wide range
of apps as well access to wireless web-based
streamed audio and video content.
40

TMT
Memory Devices Technology
8 October 2010
abc
Supply drivers – DRAM
Oversupply likely to persist through 2Q11 on sharply rising
capacity from second-tier vendors, process migration
Capex increases set to peak this year, with next year set to see
single-digit rises, mostly for continued process migration
Leading vendors shift towards more value-added high-end, mobile
DRAM
In retrospect…
An anomalous state – supply
squeeze in 1H leads to acute
oversupply in 2H
In March, we assumed that ASP decline would be
gradual through the year. Instead, we are set to see
an inverted-V ASP trend through this year. Key
factors that impacted our original assumptions
were the European financial crisis and slower than
expected migration of second tier vendors. We
had expected c8% average DDR3 ASP decline
every quarter starting in 2Q10. But ASPs actually
rose c6.8% during 2Q. In 1Q10 average DDR3
contract prices rose 12% q-o-q. In 2Q10, ASPs
continued to rise on PC demand recovery and
tight DRAM supply, and rose further as PC OEMs
pulled in demand into 2Q10 from 3Q10 to counter
the impact of a weak Euro zone, a situation
aggravated by the Greek debt crisis. In the
meantime, capacity remained relatively
constrained as second-tier vendors struggled to
sort process migrations. For leading vendors the
DRAM industry environment in 1H10 was
optimal. Samsung and Hynix not only enjoyed a
favourable operating environment (from their
point of view especially), but margins also
benefited from early process migration to 5xnm
and 4xnm during the period.
And then the pin dropped. The combination of
stronger demand and weaker supply than expected
in 1H10 was setting the industry up for a weakerthan-expected
2H10. From 3Q onwards DRAM
vendors face low q-o-q PC OEM demand growth
partly due to a high 2Q comparison base and
partly due to PC OEMs taking a more cautious
demand outlook for the rest of the year.
Exacerbating the situation is the prospect of
soaring bit supply starting in 4Q10 as second tier
vendors reach production yields for smaller nodes,
plus Samsung raises wafer capacity. We expect
DDR3 ASPs for 4Q10-1Q11 (averages for the
quarter) to fall 18% and 11% q-o-q respectively.
Nevertheless, when examining next year’s
outlook, we need to take a step back and note that
y-o-y comparisons should take into account that
2010 is a peak year; despite sharp ASP decline in
2H10, average DRAM prices for the full year are
set to rise y-o-y (15% for DDR3), an unusual
situation for the industry.
41

TMT
Memory Devices Technology
8 October 2010
abc
Short-term oversupply
After a series of years of declining revenue
growth, we estimate global DRAM revenue to rise
79% y-o-y this year, driven by 49% bit growth
and blended ASP rise of c20%. For 2011, the
situation is clearly different; we forecast global
revenue to be flat (+1% y-o-y) as blended ASPs
decline 32% y-o-y to below USD1.9, offsetting
the c49% rise in bit supply. Over the next 2-3
quarters we expect the DRAM industry to suffer
from pricing pressure on greater supply as smaller
players overcome process migration difficulties
and as larger players accelerate their own process
migrations, plus in the case of Samsung, adds
wafer capacity. Nevertheless, we forecast DRAM
industry revenue to reach cUSD46bn by 2012e,
almost doubling from last year’s cUSD24bn.
DRAM revenue, supply bit growth, ASP trends
60,000
50,000
40,000
30,000
20,000
10,000
-
Source: IDC, HSBC estimates
2008 2009 2010f 2011f 2012f
Supply (1Gb, m)
DRAM Sales (USDm)
Blended ASP, 1Gb equ. (USD)
Oversupply likely until 2Q11
3.0
2.5
2.0
1.5
1.0
0.5
-
We forecast supply to continue outstripping; we
estimate supply-demand sufficiency last year was
c99.9%, and expect it to remain relatively tight
until 4Q10, when surging supply is set to
unbalance the market. Next year, we think
oversupply could persist through 2Q11, with a
more balanced supply returning in 2H11.
DRAM supply-demand balance
40,000
30,000
20,000
10,000
-
Source: HSBC forecasts
2008 2009 2010f 2011f 2012f
106
104
102
100
Supply (1Gb, m) Supply -demand ratio (%)
We expect DDR3 average contract prices to fall
18% q-o-q in 4Q10 respectively. This appears
realistic; in 3Q10 average contract prices for
DDR3 and DDR2 have retreated 13% q-o-q,
versus our forecast of 15%. This level of price
erosion is well above Samsung and Hynix’s 3Q10
guidance of blended ASP decline in high single
digits, and a ‘slight’ fall respectively (though the
two firm’s ASP guidance is for blended, not for
commodity DRAM).
DDR2 contract and spot price (USD)
3.5
3.0
2.5
2.0
1.5
1.0
0.5
Jan-08 Jul-08 Jan-09 Jul-09 Jan-10 Jul-10
1Gb DDR2 128Mx 8 667 Mhz contract
1Gb DDR2 128Mx 8 667 Mhz spot
Source: DRAMeXchange
DDR3 enjoyed a 7-15% premium through much
of 9M10, despite DDR2/3 production cross-over
during 1-2Q10 due to strong demand for newer,
faster memory by PC OEMs. However, in
September, the premium narrowed to just 3% as
more vendors ramp up on DDR3.
98
96
94
42

TMT
Memory Devices Technology
8 October 2010
abc
Nevertheless, we expect demand downside for
commodity DRAM to be relatively limited due to
rising greater affordability of the modules. We
expect DRAM price declines to start slowing in
2Q11 as lower ASPs from sharply higher supply
during 4Q10-1Q11 stimulate higher content/
system and as supply increases peak out. Overall,
we expect content/system to rise 33% next year,
following a 27% y-o-y rise in 2010; high ASPs
this year depressed content growth this year,
though the extent of DRAM content reduction or
freeze appears to vary according to PC OEMs.
DDR3 contract and spot price (USD)
3
2.5
2
1.5
1
Jul-09
Oct-09
Source: DRAMeXchange
Jan-10
Process migration
Apr-10
Jul-10
1Gb DDR3 128Mx 8 1066MHz spot
1Gb DDR3 128Mx 8 1066MHz contract
We believe much of the DRAM bit supply growth
is coming from process migration, with relatively
low actual wafer capacity additions; we forecast
c7% q-o-y wafer capacity increase, but a 49%
supply increase for 2010. The race to improve
processes for cost and product competitiveness is
a key feature of the current spate of bit growth.
DRAM process migration – Samsung Electronics
100%
80%
60%
40%
20%
0%
2007 2008 2009 2010f 2011f 2012f
8xnm 6xnm 5xnm 4xnm 3xnm 2xnm
Source: Company, HSBC estimates
Process migration is increasing the number of dies
per given wafer. Wafer capacity was relatively
stable in the past two years, helped in part by
removal of uneconomic 200mm fab capacity.
DRAM capex fell copiously after peaking in 2007,
hitting a nadir of cUSD4.5bn last year. We
believe most expenditure was on maintenance and
process migration.
DRAM supply
25,000
20%
20,000
15%
15,000
10%
10,000
5%
5,000
0%
-
-5%
2008 2009 2010f 2011f 2012f
Wafer Capacity (12" equ. kw py , kw pm)
Change (%)
Source: Gartner, HSBC estimates
This year, we estimate capex to rise c147% to
cUSD11.2bn as process migration accelerates and
some wafer capacity is added (about 7%, we
estimate), notably by Samsung. Next year, we
expect DRAM capex to rise only 9% y-o-y, with
the bulk of it on process migration; we only
expect wafer capacity we expect wafer capacity to
rise 4% y-o-y. These capex estimates translate
into capex/sales of c26% this year and c28% next
43

TMT
Memory Devices Technology
8 October 2010
abc
year. Reassuringly, this is still well below the
2005-08 average of 54%.
Global DRAM capex (USDm)
DRAM contract prices
3
2.5
15,000
200%
2
10,000
150%
100%
1.5
50%
1
5,000
-
2008 2009e 2010f 2011f 2012f
0%
-50%
-100%
Jul-09
Oct-09
Jan-10
Apr-10
Jul-10
1Gb DDR2 128Mx 8 667 Mhz contract
1Gb DDR3 128Mx 8 1066MHz contract
Source: DRAMeXchange
Source: Gartner, HSBC forecasts
Total (HSBC) y-o-y (%)
The off-shoot of this is surging bit supply in
4Q10-1Q11, when the full brunt of extra capacity
by Samsung, plus commercial yields from process
migration by second-tier players such as Nanya
and Inotera, comes into effect, as they bring on
line 5xnm capacity that was held up in the past 2-
3 quarters due to yield issues. We believe Nanya
is now achieving c70% yield for its 5xnm process,
which is sufficient to mass produce. We expect
global supply bit growth in 3Q-4Q10 to rise 11%
and 18% q-o-q respectively.
Interestingly, DDR2 and DDR3 price declines
have converged recently; DDR3 contract prices
plunged 10-11% in 1H September. We believe the
decline in DDR3 prices reflects higher supply
levels following DDR2/3 production cross-over at
leading players in 1Q10 and higher DDR3
production at second-tier players. ASP decline has
also been impacted by demand side; we believe
this ASP decline reflects unwillingness on the part
of PC OEMs to build memory inventory on low
PC unit shipment growth this quarter following
very strong 2Q shipments, plus their anticipation
of even lower prices.
Indeed, DRAM contract ASPs have remained
under pressure after hitting a peak in June. From
their peaks in May/June, DDR3 module prices are
down 27% and DDR2 is down 23% on a periodend
to period-end basis, according to
DRAMeXchange. 2GB DDR3 modules, the
mainstay of PC memory now (DDR2 is almost
defunct as 80%-plus of PC OEM/ODMs have
migrated to DDR3) fell to USD34.
44

TMT
Memory Devices Technology
8 October 2010
abc
Global DRAM supply by vendor, units (512Mb equ., m)
Units 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e
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
y-o-y (%) 43% 70% 52% 45% 15% 15% 20% 14% 7% 8% 13% 9%
% of total 32% 36% 37% 37% 33% 35% 38% 37% 37% 37% 37% 37%
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
y-o-y (%) 20% 42% 51% 44% 6% 7% 7% 18% 9% 8% 14% 10%
% of total 21% 20% 20% 21% 20% 21% 20% 20% 20% 20% 21% 21%
Qimonda 223
y-o-y (%) -88%
% of total 1%
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
y-o-y (%) 57% 37% 38% 34% -7% 8% 5% 15% 6% 10% 5% 9%
% of total 20% 18% 17% 16% 19% 19% 18% 17% 17% 18% 17% 16%
Micron 2,244 3,568 5,407 8,559 852 859 868 989 1,147 1,272 1,416 1,572
y-o-y (%) 90% 59% 52% 58% 11% 1% 1% 14% 16% 11% 11% 11%
% of total 11% 12% 12% 13% 13% 12% 11% 11% 11% 12% 12% 12%
Others 4,186 4,323 6,507 8,632 1,017 919 993 1,394 1,436 1,522 1,664 1,884
y-o-y (%) -7% 3% 51% 33% -26% -10% 8% 40% 3% 6% 9% 13%
% of total 20% 14% 14% 13% 15% 13% 13% 15% 14% 14% 14% 14%
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
y-o-y (%) 22% 49% 49% 43% 5% 7% 11% 18% 7% 8% 11% 10%
% of total 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100%
Bit market share (%) 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e
Samsung 32% 36% 37% 37% 33% 35% 38% 37% 37% 37% 37% 37%
Hynix 21% 20% 20% 21% 20% 21% 20% 20% 20% 20% 21% 21%
Qimonda 1% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0%
Elpida 20% 18% 17% 16% 19% 19% 18% 17% 17% 18% 17% 16%
Micron 11% 12% 12% 13% 13% 12% 11% 11% 11% 12% 12% 12%
Others 20% 14% 14% 13% 15% 13% 13% 15% 14% 14% 14% 14%
Bit growth (y-o-y) 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e
Samsung 43% 70% 52% 45% 15% 15% 20% 14% 7% 8% 13% 9%
Hynix 20% 42% 51% 44% 6% 7% 7% 18% 9% 8% 14% 10%
Qimonda -88% na na na na na na na na na na na
Elpida 57% 37% 38% 34% -7% 8% 5% 15% 6% 10% 5% 9%
Micron 90% 59% 52% 58% 11% 1% 1% 14% 16% 11% 11% 11%
Others -7% 3% 51% 33% -26% -10% 8% 40% 3% 6% 9% 13%
Total 22% 49% 49% 43% 5% 7% 11% 18% 7% 8% 11% 10%
Total Shipment, MB bn 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e
MB (bn) 1,333 1,489 1,638 1,795 427 455 504 596 641 694 771 847
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
Source: IDC, HSBC estimates
45

TMT
Memory Devices Technology
8 October 2010
abc
Global DRAM capacity
Capacity kwpy, kwpm 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e
Samsung 4,500 4,650 5,040 5,850 370 370 390 420 420 420 420 420
Hynix 3,210 3,600 3,600 4,020 300 300 300 300 300 300 300 300
Micron 1,508 2,027 2,027 2,027 169 169 169 169 169 169 169 169
Qimonda 915 – – - - - - - - - - -
Elpida 2,535 3,195 3,480 3,735 260 260 265 290 290 290 290 290
Other 4,318 4,735 4,735 4,735 390 390 390 390 390 390 390 390
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
Capacity change (%)
Samsung 3% 3% 8% 16% 0% 0% 5% 8% 0% 0% 0% 0%
Hynix -4% 12% 0% 12% 9% 0% 0% 0% 0% 0% 0% 0%
Micron 12% 34% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0%
Qimonda -16% 0% 0% 0%
Elpida 0% 26% 9% 7% 18% 0% 2% 9% 0% 0% 0% 0%
Other -2% 10% 0% 0% 1% 0% 0% 0% 0% 0% 0% 0%
Total -1% 7% 4% 8% 0% 0% 2% 4% 0% 0% 0% 0%
% of total
Samsung 26% 26% 27% 29% 25% 25% 26% 27% 27% 27% 27% 27%
Hynix 19% 20% 19% 20% 20% 20% 20% 19% 19% 19% 19% 19%
Micron 9% 11% 11% 10% 11% 11% 11% 11% 11% 11% 11% 11%
Qimonda 5% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0%
Elpida 15% 18% 18% 18% 17% 17% 18% 18% 18% 18% 18% 18%
Other 25% 26% 25% 23% 26% 26% 26% 25% 25% 25% 25% 25%
Source: IDC, HSBC estimates
46

TMT
Memory Devices Technology
8 October 2010
abc
Supply drivers – NAND
2010-11e supply-demand profile remains benign, as makers
restrain wafer capacity to maximise profitability, demand soars
Process migration, multi-bits drive bit growth, lowering costs,
product competitiveness; leaders moving rapidly to 2xnm
Capex set to rise 178% and 25% y-o-y in 2010-11e; additional
supply from newcomers limited
Summary
NAND remains an exciting structural growth
area, driven by an explosion in applications of the
robust, lower power storage memory. We expect
demand to continue outstripping supply for the
most part of next year. We estimate supplydemand
sufficiency for 2010 to be c99.5%,
compared to last year’s 100.4%. We forecast
2010/12e CAGR for revenue of 24%, driven by
100% bit supply growth, offset by a c38% fall in
ASP during this period. We forecast NAND
industry revenue to reach cUSD24.8bn by 2012e,
more than doubling from last year’s estimated
cUSD12bn.
We expect ASP to decline c38% next year, after
falling c20% this year as supply starts increasing
especially in 2H11. Indeed, contact NAND prices
have held up well since the start of 4Q09 for the
higher capacities of 32-64GB, though they did
play catch up with the lower densities in 3Q10,
falling c14-16% q-o-q.
NAND ASP (contract)
20
15
10
5
0
Source: DRAMeXchange
Jan-
07
Jul-
07
Jan-
08
Jul-
08
Jan-
09
Jul-
09
Jan-
10
16Gb (2Gbx 8) MLC Flash Spot
16Gb (2Gbx 8) MLC Flash Contract
Jul-
10
Driving bit supply is new wafer capacity and
continued process migration. New wafer capacity
add accelerates, mostly impacting starting 2H11.
This year, we estimate wafer capacity to increase
c14% y-o-y, after falling about 5% last year. Next
year, we forecast c20% increase in wafer capacity,
led by Toshiba and Samsung. Process migration
and multi-level
47

TMT
Memory Devices Technology
8 October 2010
abc
Margins benefit. Cost reductions from process
migration, multi-bits outpace ASP declines. We
expect cost leaders, namely Toshiba and Samsung
to benefit from aggressive cost reductions. So
despite falling ASP, we expect margins to expand.
NAND revenue, supply bit growth and ASP trends
50,000
8.0
40,000
6.0
30,000
4.0
20,000
10,000
2.0
-
-
2008 2009 2010f 2011f 2012f
Supply (GB, m)
Rev enue (USD mn)
ASP (USD, 16Gb equ.)
Source: Gartner, HSBC forecasts
For example, we estimate Samsung would achieve
operating margin of almost 24% this year.
Toshiba’s 1QFY11 operation margin from its
memory business was 15%, and the company
expects it to rise to achieve c15-20% for 2QFY11
(end-September 2011). Similarly, SanDisk
expects cost reductions/GB to be

TMT
Memory Devices Technology
8 October 2010
abc
bring NAND capex/sales to 50%, still well below
the 2005-8 average level of 67%. Other than
accelerated process migration, we expect Samsung
to expand its NAND production capacity probably
at Fab 16, which we expect to be online next year,
to compete with Toshiba’s new NAND fab (Fab 5)
at its facility in Mie Prefecture, Japan; this is
expected to come online in spring 2011.
Global NAND capex (USDm)
14,000
12,000
10,000
8,000
6,000
4,000
2,000
-
Source: Gartner, HSBC forecasts
2009 2010f 2011f 2012f
Total (HSBC) y-o-y (%)
200%
150%
100%
50%
0%
-50%
-100%
We are not alarmed by this increase. Based on our
capacity assumptions for these two major new
fabs and process migrations for the major vendors,
we expect 2011 supply-demand balance to be
similar to this year’s, at 99.6%. We expect waferdriven
NAND capacity additions to hit in earnest
during 2H11, by which time we believe strong
demand growth can absorb the extra production.
Indeed, we expect a major shortage of supply by
next year, if Toshiba’s new Fab 5, and as appears
possible, Samsung’s Fab 16 do not come online.
Furthermore, we believe leading makers –
Toshiba and Samsung – remain cautious on the
rate of capacity ramp up, balancing it against bit
gains from process migration and attempting to
avoid the debilitating overcapacity that plagued
the industry for years.
From a capex/sales perspective, rising capex is
not excessive, either in our view. We estimate
capex/sales for the global NAND industry to
approach and remain at c45-50% during 2010-11e.
This is a steep increase over the anomalous c25%
in 2009, but compares favourably compared to the
over-build situation in 2006-08, when capex/sales
varied between 58% (2008) and 82% (2007). The
sharp forecast capex rise follows two consecutive
years of severe cut backs, during which the vast
majority of less efficient 200mm fab capacity was
decommissioned.
Process migration accelerates…
Despite wafer adds, process migration, multi-level
cells and higher utilisations following migration
to immersion lithography, still drive much of the
NAND bit supply growth. We forecast NAND
supply growth of c114% y-o-y next year, though
wafer capacity is rising c20% y-o-y, as leading
vendors all migrate to 2xnm processes. Already,
3xnm has become mainstream. For Toshiba,
32nm and 24nm scale will account for c90% of its
bit capacity by end of this year. The race to
improve processes between leaders for cost and
product competitiveness is a key feature of the
current spate of bit growth.
NAND process migration – Samsung Electronics
100%
80%
60%
40%
20%
0%
2007 2008 2009 2010f 2011f 2012f
8xnm 6xnm 5xnm 4xnm 3xnm 2xnm
Source: Company, HSBC estimates
A combination of push and pull are emboldening
makers to commit capex to put as many dies on a
wafer as possible, led by node migration:
Strong demand (manufacturers’ consensus is
for c80-90% y-o-y bit growth this year); we
forecast 73% y-o-y bit growth this year.
49

TMT
Memory Devices Technology
8 October 2010
abc
Stronger balance sheets repaired through
recapitalisation and lower losses/better
profitability, and
Intensifying technology competition (more
advanced processes, multi-bit) to drive better
margins, market positioning.
We expect industry-wide migration to more
advanced processes to lower cost/bit, raising
margins:
Samsung started mass producing 27nm-process
based NAND in April 2010 (development
completed in October 2009) and recently
launched 27nm 32Gb MLC NAND.
Toshiba started mass production of 24nmbased
64Gb MLC NAND in 3Q10, expanding
to include 24nm-based 32Gb NAND and TLC
at a later point. This appears ahead of its
original schedule to achieve 24nm migration
(X2, X3) production by 4Q10 and mass
production in 2011. Like Samsung and Hynix,
Toshiba’s 24nm process products have
Toggle DDR to enhance data transfer speeds.
Intel and Micron (IMFT – Intel Micron Flash
Technologies)’s 25nm-process based 64GB
NAND mass production started in 2Q10.
Hynix is admittedly late to 3xnm, but believes
it has caught up with leading vendors with its
2xnm migration. Hynix plans mass
production in 3Q10 for its 26nm process
based 64GB NAND. We believe Hynix has
customer samples out for this product.
current processes. By 2014-15 EUV lithography is
a likely requirement for the migration to processes
below 1x nm.
Wafer capacity increases
Wafer capacity was relatively stable in the past
two years, helped by removal of uneconomic
200mm fab capacity. This year, we estimate wafer
capacity to increase c14% y-o-y as process
migration accelerates and additional fab capacity
starts to get built out. Next year, we forecast c20%
increase in wafer capacity, led by Toshiba and
possibly Samsung.
SanDisk/Toshiba
SanDisk is focussed on upgrading processes and
implementing X3 in Fab 3 and Fab 4.
Fab 4 expansion. This year, Toshiba has ramped
previously unused space in Fab 4, which should
reach full capacity by mid-2011, i.e. by the
completion of Phase 1 of Fab 5. The original
design wafer capacity of Fab 4 is 210kwpm, but
we expect Toshiba to produce more given
throughput improvements it has made.
Fab 5 finally gets the go ahead. In mid-July
2010 Toshiba started construction of Fab 5 at
Yokkaichi. Notably it is capable of supporting
EUV and 3D flash. Fab 5 will be constructed in
two phases; SanDisk expects the fab to take
several years to ramp to full capacity. Phase 1 will
be producing 24nm process based NAND, whilst
Phase 2, for which timing has not been decided,
will produce 1xnm generation NAND.
One key question that remains unanswered is
capex from 2012e onwards, as vendors ponder the
comparative returns on migration to 1xnm.
Toshiba and Samsung remains undecided over
how to tackle this issue, though both firms are
evaluating EUV. SanDisk expects to be on 1xnm
(X2 and X3) by end 2012/early 2013. But at that
level, the technology is more ‘challenging’ than
50

TMT
Memory Devices Technology
8 October 2010
abc
Global NAND wafer capacity (‘000 wafers per year)
1,500
30%
20%
1,000
10%
500
0%
-
-10%
2008 2009 2010f 2011f 2012f
Total capacity (kwpm) y-o-y (%)
Source: Gartner, HSBC estimates
We estimate Fab 5 would cost cUSD8-9bn in total.
SanDisk previously estimated that a new NAND
fabs with support for EUV would cost cUSD8bn
for 200kwpm (current SanDisk/ Toshiba capacity
is c270-280kwpm). SanDisk expects the capex to
be spread over 2011-2014.
Fab 5 phase one is to be complete in spring 2011,
initially producing 2xnm NAND; its contribution
to wafer starts will only become meaningful in
2H11 and represent less than 10% of SanDisk’s
total for 2011. As such, SanDisk sees bit growth
next year driven by process migration to 24nm,
and not so much by wafer capacity.
Toshiba signed an agreement with SanDisk to
form a new joint venture to operate the fab.
SanDisk has the option to participate in up to 50%
in the investment and corresponding capacity
allocation at any point.
Samsung and Hynix. We are unaware of specific
NAND fab capacity additions for Samsung.
However, we believe that a potential KRW2trn
additional spend on Fab 16 could be NANDrelated.
Hynix is to raise M11 fab wafer capacity
to c80kwpm from the current c50-60kwpm to
replace the 200mm fabs capacity shut down in
2008-09. Hynix has further room to expand
NAND space in M11 and M12.
Micron is spending about a third of its USD850-
950m capex on NAND; Intel will match the
amount for its 50% stake in the Intel-Micron
NAND JV.
Multi-bit levels – another driver of bit
growth without wafer additions
Unlike DRAM, NAND cost reductions are
progressing in two areas – process migration and
the move to multiple bits per layer. The majority
of current NAND devices are multi-level cell,
using 2 bits per cell. As process migration
becomes more difficult, NAND vendors are
raising the number of bits per cell (which
increases the amount of data that each cell can
store on a given area of the chip), and hence costs
per GB. 3-4-bit cells are becoming important to
increase bit growth.
Whilst there are limitations to performance and
implementing three levels per cell (TLC) at smaller
nodes – hence their current limitation mostly to lowend
storage requirements – the price advantages of
the technology are clear; 32Gb (the minimum
density making commercial sense) TLC is currently
c16-20% cheaper than an equivalent 2-bit cell
NAND to produce on a GB basis.
Toshiba/SanDisk is the leader in multi-cell
technology; SanDisk holds key patents on X3 and
X4 technologies. Toshiba/SanDisk started
producing 3-bit per cell in 2008; SanDisk is
migrating c50% of its NAND bit production to X3
between 2010 and 2013, resulting in c25-35%
annual cost reductions per GB. Samsung started
mass production of 3-bit MLC NAND on 30nm at
end-November 2009. Multi-bit NAND should
increase production of higher densities (32Gbplus)
for demanding content storage.
51

TMT
Memory Devices Technology
8 October 2010
abc
Technology develops broaden NAND
scope
Bigger modules for mobile devices
We think larger densities are set to accelerate the
take up of NAND in mobile devices. Samsung
started mass production of a 3xnm-based 64GB
embedded NAND (moviNAND) in December
2009, which uses 16 x 32Gb dies and an on-board
controller (total 17-die stack). Similarly,
Toshiba’s 64GB embedded NAND flash module
entered mass production in 1Q10. This 32nmbased
unit has 16 x 32Gb NAND stacked units. In
June 2010, Toshiba announced a 128GB module,
which stacks 16x 8GB (64Gb 30nm) NAND units
and a controller. Mass production is slated for
4Q10. Apple currently uses 1x 32GB NAND
module for its iPhone 4 and 2x 32GB NAND
modules for the 64GB iPod Touch. Apple and
others could use multiple 64GB NAND modules
to deliver capacities approaching that of HDD for
tablets and laptop PCs.
Faster performance, lower power
consumption. This is achieved through process
migration. Samsung started producing 20nm
based NAND (32Gb) in April 2010. SD card
based on 20nm MLC is 30% faster than the 30nm
equivalent, according to Samsung.
Faster performance. Toggle DDR NAND is here.
Samsung and Toshiba have tied up to standardise
faster double data rate (DDR) that would help
SSD performance. This would deliver 3-10x
increase in data throughput over current interfaces.
Toggle DDR NAND is currently available in
16Gb (SLC) and 32Gb (MLC) densities from
Samsung, which started mass production in
December 2009. Samsung launched its first SSD
using toggle-mode DDR NAND in June 2010.
52

TMT
Memory Devices Technology
8 October 2010
abc
Global NAND supply by vendor
Units (16Gb equ. m) 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e
Samsung 1,039 1,859 3,944 7,135 332 413 511 603 692 853 1,085 1,313
y-o-y (%) 29% 79% 112% 81% 15% 25% 24% 18% 15% 23% 27% 21%
% of total 36% 37% 37% 36% 36% 36% 38% 39% 37% 36% 36% 38%
Toshiba/SanDisk 1,188 2,090 4,884 8,944 394 486 561 649 776 1,053 1,415 1,640
y-o-y (%) 53% 76% 134% 83% 3% 23% 15% 16% 20% 36% 34% 16%
% of total 41% 42% 46% 45% 42% 42% 42% 42% 42% 45% 47% 47%
Hynix 129 251 668 1,319 43 53 67 87 127 156 181 204
y-o-y (%) -30% 94% 167% 97% 0% 22% 27% 29% 47% 22% 16% 13%
% of total 4% 5% 6% 7% 5% 5% 5% 6% 7% 7% 6% 6%
Intel + Micron 539 752 1,120 2,376 157 191 192 211 241 274 290 316
y-o-y (%) 91% 39% 49% 112% -9% 22% 1% 10% 14% 14% 6% 9%
% of total 19% 15% 11% 12% 17% 17% 14% 14% 13% 12% 10% 9%
Others 18 34 54 67 4 7 11 11 12 13 14 15
y-o-y (%) 24% 84% 60% 25% -39% 75% 44% 6% 8% 7% 7% 5%
% of total 1% 1% 1% 0% 0% 1% 1% 1% 1% 1% 0% 0%
Total 2,914 4,984 10,670 19,842 931 1,151 1,342 1,561 1,849 2,348 2,985 3,488
y-o-y (%) 41% 71% 114% 86% 4% 24% 17% 16% 18% 27% 27% 17%
% of total 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100%
Bit market share (%) 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e
Samsung 36% 37% 37% 36% 36% 36% 38% 39% 37% 36% 36% 38%
Toshiba/SanDisk 41% 42% 46% 45% 42% 42% 42% 42% 42% 45% 47% 47%
Hynix 4% 5% 6% 7% 5% 5% 5% 6% 7% 7% 6% 6%
IMFT 19% 15% 11% 12% 17% 17% 14% 14% 13% 12% 10% 9%
Others 1% 1% 1% 0% 0% 1% 1% 1% 1% 1% 0% 0%
Total 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100%
Bit growth (y-o-y) 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e
Samsung 29% 79% 112% 81% 15% 25% 24% 18% 15% 23% 27% 21%
Toshiba 53% 76% 134% 83% 3% 23% 15% 16% 20% 36% 34% 16%
Hynix -30% 94% 167% 97% 0% 22% 27% 29% 47% 22% 16% 13%
IMFT 91% 39% 49% 112% -9% 22% 1% 10% 14% 14% 6% 9%
Others 24% 84% 60% 25% -39% 75% 44% 6% 8% 7% 7% 5%
Total 41% 71% 114% 86% 4% 24% 17% 16% 18% 27% 27% 17%
Source: IDC, HSBC estimates
53

TMT
Memory Devices Technology
8 October 2010
abc
Global NAND fab capacity
Kwpy, kwpm 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e
Samsung 4,159 4,327 4,860 5,820 362 360 360 360 370 390 420 440
y-o-y (%) 7% 4% 12% 20% -3% -1% 0% 0% 3% 5% 8% 5%
Hynix 907 1,267 1,687 2,187 96 106 106 116 126 136 146 156
y-o-y (%) -42% 40% 33% 30% 26% 10% 0% 9% 9% 8% 7% 7%
Toshiba / SanDisk 2,787 3,360 4,350 5,160 260 270 280 310 320 320 390 420
y-o-y (%) -3% 21% 29% 19% 0% 4% 4% 11% 3% 0% 22% 8%
IMFT 1,260 1,440 1,542 1,962 105 125 125 125 125 125 132 132
y-o-y (%) 16% 14% 7% 27% 0% 19% 0% 0% 0% 0% 6% 0%
Renesas / PSC 11 11 11 11 1 1 1 1 1 1 1 1
y-o-y (%) -81% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0%
Numonyx 71 72 80 80 5 6 6 7 7 7 7 7
y-o-y (%) 2% 2% 11% 0% -25% 8% 8% 7% 0% 0% 0% 0%
8” Total 2,394 1,296 1,164 960 144 95 96 97 97 97 97 97
y-o-y (%) -60% -46% -10% -18% -3% -34% 1% 1% 0% 0% 0% 0%
12” Total 8,130 9,900 12,012 14,702 765 825 835 875 905 935 1,052 1,112
y-o-y (%) 16% 22% 21% 22% 1% 8% 1% 5% 3% 3% 13% 6%
12” total (kwpm) 678 825 1,001 1,225 64 69 70 73 75 78 88 93
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
y-o-y (%) 16% 22% 21% 22% 1% 8% 1% 5% 3% 3% 13% 6%
Total (12” equ. kwpy) 9,194 10,476 12,529 15,129 829 867 878 918 948 978 1,095 1,155
y-o-y (%) -5% 14% 20% 21% 1% 5% 1% 5% 3% 3% 12% 5%
Total (12” eq. kwpm) 766 873 1,044 1,261 829 867 878 918 948 978 1,095 1,155
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%
Source: IDC, HSBC estimates
54

TMT
Memory Devices Technology
8 October 2010
abc
Valuations, ratings
Maintain TP and OW rating on Samsung. Lower TP on Hynix, but
maintain OW(V)
Lower TP on Elpida, Nanya Tech, Winbond and Elpida. Maintain
OW(V) for Elpida and Winbond, and N(V) for Inotera and Nanya
We expect valuation divergence as product mix and cost structure
drive profitability for leaders
In the past six months, memory stocks have
underperformed their respective markets on
concerns that the memory cycle peaked out in 2Q.
Among our coverage, with the exception of
Winbond, pure memory plays have lagged their
indices by 20-52%. Samsung, being a diversified
tech conglomerate, underperformed KOSPI by
11%, with semiconductor revenue representing
25% of its revenue. The market penalised Nanya
and Inotera (underperformed by c40%), on delays
of their 5xnm migration.
Concerns over memory price correction could
linger (until data points emerge indicating a
deceleration), we expect price performance
divergence between vendors to be more
pronounced going forward; we expect the market to
favour vendors that are capable of accelerating
process migration and improving their product mix.
Our top sector pick is Samsung Electronics based
on its unparalleled overall memory positioning and
compelling valuations, in addition to its diversified
technology portfolio. We also like Hynix and
Elpida for their strong DRAM positioning (inhouse
technology, process migration) and their
leading positions in mobile DRAM. Both Elpida
Stock performances relative to respective markets (%)
10
0
-10
-20
-30
-40
-50
-60
SEC
Source: Bloomberg
Hynix
Elpida
Nanya
Inotera
6 months 3 months 1 month
Winbond
and Winbond are trading below 1x P/B. Trading at
0.5x FY11e P/B, Elpida’s valuations look very
attractive for FY2011e. We think most of the
negative news have been priced in and the stock is
oversold. We also like Winbond’s non-PC DRAM
portfolio and believe it is a defensive portfolio
amid commodity DRAM price correction cycle
given its attractive valuation.
We believe the Taiwanese performance should
still be limited until signs of improvement from
55

TMT
Memory Devices Technology
8 October 2010
abc
Valuation summary
Valuation
method
Target
multiple
Historical PB
multiple
Samsung Electronics avg of SOTP, 2.2x PB
2011e PB, DCF
Hynix Semiconductor 2011e PB 1.8x 0.8x-3.3x
Elpida Memory 2011e PB 0.8x 0.3x-2.4x
Nanya Technology 2011e PB 1.9x 1.6x-4.9x
Inotera Memories 2011e PB 1.5x 0.5x-2.5x
Winbond 2011e PB 1.0x 0.2x-1.2x
EBIT margin comparison
40%
20%
0%
-20%
-40%
-60%
Source: HSBC
SEC
Hynix
Elpida
Nanya
Inotera
Winbond
the node migration and cycle time are becoming
apparent. We remain Neutral (V) on the two stocks.
2009 2010e 2011e
Source: Company data, HSBC estimates, Elpida forecast are calendarised
Stock coverage and ratings summary
Company Ticker Rating New TP
(LC)
CP
(LC)
Potential
return (%)
Previous
TP
Change
(%)
Samsung Electronics 005930 KS OW 1,139,000 793,000 44% 1,139,000 0%
Hynix Semiconductor 000660 KS OW(V) 32,000 23,400 37% 36,000 -11%
Elpida Memory 6665 JP OW(V) 1,500 928 62% 3000 -50%
Nanya Technology 2408 TT N(V) 20 19.3 4% 25 -20%
Inotera Memories 3474 TT N(V) 17 16.2 5% 20 -15%
Winbond 2344 TT OW(V) 10.8 8.03 34% 12.8 -16%
Source: HSBC
56

TMT
Memory Devices Technology
8 October 2010
abc
10 ROE Vs 10 PB 11 ROE Vs 11 PB
10 ROE
50%
40%
30%
20%
10%
0%
-10%
-20%
Micron
Elpida
Winbond
Hy nix
Pow erchip
SanDisk
Samsung
Inotera
Toshiba
Nanya
0.0 0.5 1.0 1.5 2.0 2.5
10 PB
11 ROE
25%
20%
15%
10%
5%
0%
Hy nix
Micron
Samsung
SanDisk
Elpida
Nanya
Winbond Inotera
Pow erchip
Toshiba
0.0 0.5 1.0 1.5 2.0 2.5
11 PB
Source: HSBC
Source: HSBC
Peer group valuation, profitability comparison
Company Ticker Price (LC) Mkt cap
(USDm)
HSBC
rating
_____________ PE _____________________ EV/EBITDA _________ _____________ PB_____________
2009 2010e 2011e 2009 2010e 2011e 2009 2010e 2011e
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
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
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
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
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
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
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
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
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
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
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
MC WA 3.7 10.2 9.5 7.5 4.0 5.1 2.0 1.6 1.7
Div yield (%) _______ EBIT margin _______ ______ EBITDA margin________ ___________ ROE ____________
Company Ticker 2010e 2009 2010e 2011e 2009 2010e 2011e 2009 2010e 2011e
Samsung 005930 KS 1.5% 7% 12% 11% 16% 20% 19% 15% 20% 17%
Hynix 000660 KS 0.0% -2% 29% 20% 22% 42% 36% -7% 44% 22%
Elpida 6665 JP 0.0% -4% 15% 17% 2% 31% 36% -69% 19% 9%
Toshiba 6502 JP 0.9% 0% 3% 4% 5% 7% 8% -47% 8% 4%
Powerchip 5346 TT 0.0% -59% 17% 1% 47% 56% 39% -57% 40% 3%
Nanya 2408 TT 0.0% -37% -6% 8% -3% 21% 31% -63% -14% 8%
Inotera 3474 TT 0.0% -27% -8% 8% 56% 61% 63% -23% -11% 6%
ProMOS 5387 TT 0.0% -141% -58% -57% -5% 27% 55% -82% -90% -399%
Winbond 2344 TT 0.0% -31% 11% 7% 22% 45% 41% -24% 10% 5%
Micron MU US 0.0% -1% 19% 18% 33% 38% 33% -34% 25% 15%
SanDisk SNDK US 0.0% 15% 30% 26% 21% 34% 30% 11% 24% 18%
MC WA 1.5% 3% 13% 11% 16% 23% 22% -1% 20% 14%
Source: HSBC forecast for SEC, Hynix, Elpida, Nanya, Inotera and Winbond. Bloomberg consensus otherwise. Priced as at 6 Oct 2010. Numbers are calendarised.
57

TMT
Memory Devices Technology
8 October 2010
abc
Inotera’s relative performance (TWD)
Elpida’s relative performance (JPY)
10,000
50
2,000
8,000
8,000
6,000
4,000
2,000
40
30
20
10
1,500
1,000
6,000
4,000
2,000
-
-
500
-
Jan-06 Jan-07 Jan-08 Jan-09 Jan-10
Jan-06 Jan-07
Jan-08 Jan-09 Jan-10
Taiw an Taiex Index
Inotera (RHS)
Topix Index
Elpida Memory (RHS)
Source: Bloomberg
Source: Bloomberg
Samsung Electronics’ relative performance (KRW)
Hynix’s relative performance (KRW)
2,500
1,000,000
3,000
50,000
2,000
800,000
2,500
40,000
1,500
1,000
600,000
400,000
2,000
1,500
1,000
30,000
20,000
500
200,000
500
10,000
-
-
-
-
Jan-06 Jan-07 Jan-08 Jan-09 Jan-10
Jan-06 Jan-07 Jan-08 Jan-09 Jan-10
KOSPI Index
Samsung Electronics (RHS)
KOSPI Index
Hy nix (RHS)
Source: Bloomberg
Source: Bloomberg
Winbond Electronics’ relative performance (TWD)
Nanya Technology’s relative performance (TWD)
10,000
15
10,000
100
8,000
12
8,000
80
6,000
9
6,000
60
4,000
6
4,000
40
2,000
3
2,000
20
-
-
-
-
Jan-06 Jan-07 Jan-08 Jan-09 Jan-10
Jan-06 Jan-07 Jan-08 Jan-09 Jan-10
Taiw an Taiex Index
Winbond Electronics (RHS)
Taiw an Taiex Index
Nany a Tech (RHS)
‘
Source: Bloomberg
Source: Bloomberg
58

TMT
Memory Devices Technology
8 October 2010
abc
Company profiles
Elpida Memory (6665)
Hynix Semiconductor (000660)
Inotera Memories (3474)
Nanya Technology (2408)
Samsung Electronics (005930)
Winbond (2344)
59

TMT
Memory Devices Technology
8 October 2010
abc
Elpida Memory (6665)
Global #3 DRAM maker with high specialist revenue component
of c43% for FY11e, including mobile DRAM
We lower FY10-11 operating profit forecasts by 6% and 27%
respectively on lower ASP assumptions
Maintain OW(V) rating, lower TP to JPY1,500 based on FY11e
target PB of 0.8x. Stock is oversold, in our view
Investment summary
Sensible growth strategy
A pure memory play with mostly DRAM
exposure, Elpida Memory was the third largest
DRAM manufacturer on revenues according to
IDC, with c17% share (Samsung: c34%, Hynix
c22%).
Having pursed a strategy that capitalises on its
shrink technology to reduce costs to a 5xnmequivalent
level, Elpida is now migrating to 4xnm
process to remain competitive versus the leaders.
For commodity PC DRAM, it is fully deploying
capacity at Rexchip, its Taiwanese subsidiary to
remain cost competitive. Elpida’s Japan fab
focuses on higher value-add products such as
mobile DRAM and specialist DRAM. To further
enhance its strong positioning in mobile DRAM,
it has co-developed with Spansion NAND that
would enable Elpida to capture the full margins
available by offering MCP (multichip packages).
Though we expect (as we do with all the memory
companies) next year’s earnings to decline y-o-y
from a high base, Elpida has upside risk from its
mobile and specialist DRAM exposure, which we
expect to account for c43% of FY11e revenue
(year-end March 2012). The stock is trading on
FY11e PB of 0.5x. We think it is oversold at
current levels.
Valuations, rating
We value Elpida using PB to derive a 12-month
target price of JPY1,500 by applying a target
FY11 PB multiple of 0.8x. Previously, we valued
Elpida at JPY3,000 using the same methodology,
but using a FY10e PB multiple of 2x. The stock
has historically traded at a PB of 0.3-2.4x.
Though we forecast Elpida’s operating
performance to decline from this year’s peak, we
believe our new target PB sufficiently reflects
downside risks; based on FY11-12e average 8.3%
ROE and 7.5% COE, the stock should in fact, be
valued at a PB ratio of 1.1x. But we believe
negative sentiment on DRAM demand, Elpida’s
weaker market positioning versus leaders,
execution risk of its process migration and JPY
strength limit shorter term valuation upside.
Under our research model, for Japan stocks with a
volatility indicator, the Neutral band is 10ppt
above and below the hurdle rate of 7.5% (risk-free
rate plus risk premium). This translates into a
Neutral band of -2.5% to +17.5% around the
60

TMT
Memory Devices Technology
8 October 2010
abc
current share price. Our 12-month target price of
JPY1,500 suggests a potential return of 61.6%,
which is above the Neutral band; thus, we
maintain an Overweight (V) rating on the stock.
Volatile ratings are defined as stocks having
historical volatility (defined as the past month’s
average of the daily 365-day moving average
volatilities) of more than 40%.
HSBC versus consensus – Elpida
JPY bn FY10e FY11e 2QFY10e 3QFY10e
Forward PB chart
7,000
6,000
Price (JPY)
5,000
4,000
3,000
2,000
1,000
-
Jan-06 Jan-07 Jan-08 Jan-09 Jan-10
2.0x
1.5x
1.0x
0.5x
Revenue 647 619 163 157
EBITDA 236 187 65 46
EBIT 109 54 33 16
Net income 72 30 24 7
EPS (JPY) 354 147 118 32
EBITDA margin 36.5% 30.1% 39.9% 29.6%
EBIT margin 16.8% 8.7% 20.3% 10.0%
Net margin 11.1% 4.8% 14.7% 4.2%
HSBC vs consensus
Revenue -1.5% -3.8% -1.1% -1.4%
EBITDA -1.6% -3.3% -4.7% -18.0%
EBIT -3.1% -12.2% -6.2% -0.1%
Net income -4.4% -19.1% -1.7% -33.7%
EPS (JPY) -3.2% -16.0% 8.6% -38.1%
Consensus
Revenue 657 644 165 159
EBITDA 240 193 68 57
EBIT 112 61 35 16
Net income 75 37 24 10
EPS (JPY) 366 175 109 52
EBITDA margin 36.5% 30.0% 41.4% 35.6%
EBIT margin 17.1% 9.5% 21.4% 9.9%
Net margin 11.4% 5.7% 14.8% 6.2%
Source: I/B/E/S, HSBC estimates
Source: I/B/E/S, HSBC estimates
Catalysts and risks
Memory-specific catalysts include:
Better-than-expected results and strong
guidance from PC OEMs, such as recent Dell
and HP results.
Launch and strong deliveries of key mobile
devices – driving Elpida’s exposure to mobile
DRAM.
Evidence of strong unit shipments and rising
memory content of smartphones.
Launch of competitive new products,
especially next-generation memories.
Memory-specific risks include:
Faster-than-expected ASP erosion due to
greater capacity coming online and/or slowerthan-expected
PC unit shipments.
Greater ability of competitors to accelerate
process migration, wafer capacity expansion
through improving financials.
IP disputes. Although these tend to be
ongoing, any potentially major dispute could
be a risk.
61

TMT
Memory Devices Technology
8 October 2010
abc
Other company-specific catalysts and risks include:
Rebounding global competition following
restructuring and recapitalisation efforts.
Successful proposed IPO of subsidiary Rexchip.
Sharp reduction in debt through internal cash
generation.
Strengthening JPY versus USD, KRW, TWD,
although producing offshore in Taiwan
should offset some of that.
Delayed global economic recovery.
Fundamentals
Operating performance outlook
We forecast Elpida’s FY10 (year-end March
2011) revenue to rise c9% y-o-y, but for FY11,
we forecast revenue to decline c5%. We forecast
Elpida’s FY10e operating profit to rise c310% y-
o-y to JPY110bn but forecast FY11e to see a 51%
decline.
Revenue breakdown (CY)
100%
80%
60%
40%
20%
0%
1Q09 2Q09 3Q09 4Q09 1Q10 2Q10
Computing DRAM Premier DRAM
Source: Company, HSBC estimates
Improving balance sheet, shorter inventory
days.
Highly profitable subsidiary. Rexchip
Electronics has the highest profitability
amongst Taiwanese DRAM makers.
Revenue, OP (JPYbn) and OPM (%), CY
800
30%
600
20%
10%
400
0%
200
-10%
-20%
-
-30%
(200)
2007 2008 2009 2010f 2011f
-40%
Rev enue (JPY bn) OP OPM
Source: Company, HSBC estimates
Revenue, margin outlook
For CY10e we estimate revenue to rise 76% in
JPY terms, on 37% bit growth and blended ASP
increase of 31%. We forecast CY11e revenue
decline of 5% under our current process migration
assumptions; bit growth is stable at c38% y-o-y
and we forecast a 32% fall in ASP on greater
industry-wide supply (supply-demand to be
relatively balanced in 2011).
Historically, Elpida has had a much higher than
industry average proportion of premier (specialist)
DRAM – about a half of revenue – compared to
computing (commodity) DRAM. This proportion
fell dramatically last year as demand premier
DRAM dried up.
Elpida has seen continued recovery in operating
profitability in the past four quarters on higher
revenue mostly on higher computing DRAM ASP
and bit shipments.
Lower costs due to process migration, greater
productivity.
62

TMT
Memory Devices Technology
8 October 2010
abc
Elpida forecast assumptions
DRAM process migration
y-o-y % changes 2008 2009 2010e 2011e
Bit growth 99% 57% 37% 38%
ASP -53% -31% 31% -32%
FX -11% -12% -2% 0%
USD sales check -8% 10% 80% -6%
USD sales -18% -2% 76% -5%
JPY sales check 99% 57% 37% 38%
JPY sales -53% -31% 31% -32%
OPM -11% -12% -2% 0%
Source: Company, HSBC estimates
100%
80%
60%
40%
20%
0%
2007 2008 2009 2010f 2011f
For FY10, Elpida expects its bit growth will reach
45%; we forecast 34% and 36% respectively for
FY10-11. Our forecasts are lower due to time
required to achieve satisfactory yields in mass
production both at its own fabs and outsourced
production.
CY bit growth, ASP assumptions (%)
150%
100%
50%
0%
-50%
-100%
2008 2009 2010f 2011f 2012f
Bit growth ASP
Source: Company, HSBC estimates
Technology – process migration
Uniquely, Elpida took a dual-pronged R&D
strategy to minimise capex starting in 2008.
Elpida developed ‘shrunk’ versions of its 6xnm
based chips. These were manufactured using ArF
dry scanner equipment, i.e. without having to
invest in expensive immersion lithography
equipment used by competitors like Samsung.
8xnm 6xnm 6x S, XS 4xnm 3xnm 2xnm
Source: Company, HSBC estimates
Elpida originally developed a 65nm S shrink
version in 2008. Based on that version, it then
completed development of a 65nm XS (extrashrink)
version in December 2009. Elpida says that
the XS chip delivers 25% more dies per wafer
compared to the S version and is as costcompetitive
as 5xnm node DRAM. The XS version
1GB DDR3 entered mass production in 1Q10.
This year, though, Elpida (and its subsidiary
Rexchip) is migrating to 4xnm. By going from
65nm-S to 4xnm, the firm estimates dies per
wafer increase c115% (going from 65nm-S to
50nm would be 45% more dies) and increase
shipment volumes by c30%. Elpida’s timeline on
migration to 4xnm and 3xnm has been forced by
competitors’ migration plans:
Hiroshima – mobile and speciality DRAM focus
Elpida started pilot production of 2GB DDR3 chips
on the 4xnm process (6F2) at its 130kwpm
Hiroshima fab, and raised yields sufficiently for
mass production in 2010. By September, it had
migrated c30% of wafer input to 4x-5xnm and
plans to migrate half its capacity to 4xnm by end-
2010. We expect further migration to depend on
product mix, as consumer DRAM uses older
processes. Mobile DRAM is currently produced on
40-50nm processes. Elpida estimates that c60-70%
of Hiroshima capacity is required to produce
mDRAM by end 2011, with the remaining
63

TMT
Memory Devices Technology
8 October 2010
abc
comprising. specialty DRAM; 3xnm migration at
Hiroshima is likely to be limited in the shorter term.
Rexchip – the commodity DRAM arm
Over time, Rexchip will manufacture all of
Elpida’s commodity DRAM. Rexchip is to convert
its entire 80kwpm 300mm capacity to Elpida’s
4xnm process by end-2010. As at September, 10%
of wafer input was based on 4xnm, with the
remaining on 65nm XS. Positively, Rexchip should
have received all the tools required for full 4xnm
migration by the end of this November.
Elpida completed development of 3xnm DDR3
(30% smaller, 40% more efficiency compared to
4xnm); engineering samples should be out by
December. Mass production is to start by end-2010,
with production ramp next spring. Production uses
existing equipment with double patterning. Once
satisfactory yields are achieved, the technology will
be transferred to Rexchip, which will completely
migrate to 3xnm in the longer term.
Wafer capacity (kwpy)
4,000
3,000
2,000
1,000
-
2007 2008 2009 2010f 2011f 2012f
Wafer capacity (kwpy)
Source: Company, HSBC estimates
Production capacity, capex
Elpida’s 300mm wafer capacity comprises its
Hiroshima fab plus its 64%-owned Taiwanese
subsidiary Rexchip Electronics. In addition to
Rexchip, Elpida also outsources production to
Powerchip and will start to outsource to ProMOS
(PC DRAM) and Winbond (graphics DRAM).
Elpida’s relationships with Taiwanese makers
Elpida
64%
Rexchip
30.84%
Powerchip
ProMOS
Winbond
Source: Company – arrows indicate outsourcing flow
The relationship with Powerchip is waning as
Rexchip expands. Powerchip has c130kwpm
total capacity; 80kwpm is commodity DRAM,
of which 30kwpm is legacy production. We
expect Elpida to receive c40kwpm equivalent.
We believe Powerchip is also migrating to
4xnm, with one immersion tool expected by
the end of this year followed by another two
next year.
ProMOS is to start DRAM mass production
of 65nm-XS DRAM for Elpida after July
2010, with 35kwpm capacity. But we believe
ProMOS is currently having some yield issues,
so this timeline could be delayed. We believe
for Elpida, ProMOS will act purely as a
foundry.
Winbond started to produce graphics DRAM
in late-1Q10. We believe the number of
wafers it supplies is low, at c5kwpm.
For FY10e (March-2011 end), Elpida expects to
spend JPY115bn capex, of which 30% is
earmarked for Rexchip, which will require
JPY30-35bn to fully migrate its capacity to 4xnm.
60% is earmarked for the Hiroshima fab to
migrated half of its capacity to 4xnm. The
remaining 10% is for subsidiaries to expand wafer
testing and back-end fab facilities.
64

TMT
Memory Devices Technology
8 October 2010
abc
Forecast revisions
Elpida – forecast revisions
Mar (JPY bn)
___ Revised __ ___ Previous _____ Changes __
2010e 2011e 2010e 2011e 2010e 2011e
Revenue 647.2 619.3 709.9 754.5 -9% -18%
EBITDA 236.0 186.6 290.4 255.1 -19% -27%
EBIT 108.5 53.8 115.4 74.2 -6% -27%
Net income 71.9 29.7 98.3 58.9 -27% -49%
EPS (JPY) 354 146 500 300 -29% -51%
Source: HSBC forecasts and estimates
65

TMT
Memory Devices Technology
8 October 2010
abc
Financials & valuation: Elpida
Overweight (V)
Financial statements
Year to 03/2010a 03/2011e 03/2012e 03/2013e
Profit & loss summary (JPYb)
Revenue 467 647 619 641
EBITDA 149 236 187 194
Depreciation & amortisation -122 -127 -133 -135
Operating profit/EBIT 27 109 54 59
Net interest -9 -9 -9 -8
PBT 9 85 35 41
HSBC PBT 9 85 35 41
Taxation -2 -13 -5 -8
Net profit 7 72 30 33
HSBC net profit 7 72 30 33
Cash flow summary (JPYb)
Cash flow from operations 56 172 186 179
Capex -60 -60 -60 -64
Cash flow from investment -152 -172 -60 -64
Dividends 0 0 0 0
Change in net debt -52 89 -119 -109
FCF equity -6 100 121 107
Balance sheet summary (JPYb)
Intangible fixed assets 1 1 1 1
Tangible fixed assets 615 539 458 376
Current assets 326 313 420 538
Cash & others 112 23 142 251
Total assets 947 858 884 920
Operating liabilities 324 232 229 232
Gross debt 277 277 277 277
Net debt 164 253 134 25
Shareholders funds 277 349 379 412
Invested capital 506 598 508 433
Ratio, growth and per share analysis
Year to 03/2010a 03/2011e 03/2012e 03/2013e
Y-o-y % change
Valuation data
Year to 03/2010a 03/2011e 03/2012e 03/2013e
EV/sales 0.9 0.7 0.5 0.3
EV/EBITDA 2.9 1.9 1.7 1.1
EV/IC 0.9 0.7 0.6 0.5
PE* 21.9 2.6 6.3 5.6
P/Book value 0.6 0.5 0.5 0.5
FCF yield (%) -2.1 54.2 65.9 58.5
Dividend yield (%) 0.0 0.0 0.0 0.0
Note: * = Based on HSBC EPS (fully diluted)
Issuer information
Share price (JPY) 928 Target price (JPY) 1500 Potent’l return (%) 61.6
Reuters (Equity) 6665.T Bloomberg (Equity) 6665 JP
Market cap (USDm) 2,315 Market cap (JPYb) 192
Free float (%) 89 Enterprise value (JPYb) 441
Country Japan Sector SEMICONDUCTORS
Analyst Nam Park Contact 852 2996 6591
Price relative
4500
4000
3500
3000
2500
2000
1500
1000
500
0
Source: HSBC
2008 2009 2010 2011
Elpida Rel to TOPIX INDEX
Note: price at close of 06 Oct 2010
4500
4000
3500
3000
2500
2000
1500
1000
500
0
Revenue 41.1 38.6 -4.3 3.5
EBITDA 58.7 -20.9 3.7
Operating profit 304.3 -50.4 9.5
PBT 848.3 -58.6 17.9
HSBC EPS 736.4 -58.6 12.3
Ratios (%)
Revenue/IC (x) 0.9 1.2 1.1 1.4
ROIC 4.4 16.7 8.3 10.1
ROE 3.2 23.0 8.2 8.5
ROA 1.6 8.9 4.4 4.6
EBITDA margin 31.8 36.5 30.1 30.2
Operating profit margin 5.7 16.8 8.7 9.2
EBITDA/net interest (x) 15.8 26.4 21.2 25.3
Net debt/equity 45.7 72.7 35.5 6.2
Net debt/EBITDA (x) 1.1 1.1 0.7 0.1
CF from operations/net debt 34.1 68.0 138.7 705.7
Per share data (JPY)
EPS Rep (fully diluted) 42.33 354.02 146.63 164.70
HSBC EPS (fully diluted) 42.33 354.02 146.63 164.70
DPS 0.00 0.00 0.00 0.00
Book value 1628.54 1717.97 1864.60 2029.29
66

TMT
Memory Devices Technology
8 October 2010
abc
Hynix Semiconductor (000660)
Product portfolio leverages into higher-margin, growth segments
such as servers and mobile; process migration accelerates
Historic bane of debt is shrinking rapidly. Shareholder stake sale
and strong operating metrics are stock catalysts
Remain OW(V), TP lowered to KRW32,000 (-11.1% from previous)
on lower 2011e PB target of 1.8x
Investment summary
Strong, diversified product portfolio
A pure memory play with both DRAM and
NAND exposure, Hynix is one of our preferred
plays in the sector. Hynix has:
Highly competitive technologies and leading
market share leadership in DRAM and a
returning presence in NAND, where it has
caught up with market leaders in process
migration.
Debt is no longer an issue, continued ability
to invest. Hynix is unshackling itself from its
historic bane, high debt. We forecast parent
level operating cash flow of cKRW4.1trn this
year and close to KRW4trn per year for 2011-
12e (2009: KRW735bn). We forecast 2010e
parent net DER at a comfortable c32.5% level,
leaving plenty of scope for Hynix to support
ongoing capex and R&D if required. Though
Hynix could be net cash next year, it is likely
to maintain some debt to optimise ROE.
Determination to expand leadership with
more advanced and cost-effective
technologies, whilst remaining cautious in
terms of capex and competitive positioning.
Valuations, rating
To value Hynix we use a target 2011e PB of 1.8x.
This PB target has been rolled forward from
previous 2010e, and lowered from 2.5x to reflect
memory market uncertainties and lower ROE
expectations. Macro demand concerns and lower
2011e forecast profits are likely to constrain
valuations. Historically, Hynix has traded in a 0.8-
3.3x PB range. Our new target of KRW32,000 is
c11.1% below our previous target of KRW36,000.
Under our research model, for stocks without a
volatility indicator, the Neutral band is five
percentage points above and below our hurdle rate
for Korean stocks of 10.5%, or 5.5-15.5% around
the current share price. Our new 12-month target
price of KRW32,000 suggests a potential return of
36.8% which is above the Neutral band; thus, we
reiterate our Overweight (V) rating on the stock.
Volatile ratings are defined as stocks having
historical volatility (defined as the past month’s
average of the daily 365-day moving average
volatilities) of more than 40%.
67

TMT
Memory Devices Technology
8 October 2010
abc
HSBC versus consensus
HSBC 2010e 2011e 3Q10e 4Q10e
Revenue 12,045 12,557 2,968 3,040
EBITDA 5,033 4,510 1,192 1,225
EBIT 3,488 2,546 860 743
EBT 3,223 2,523 852 814
Net profit 3,143 2,220 880 679
EBITDA margin 41.8% 35.9% 40.1% 40.3%
EBIT margin 29.0% 20.3% 29.0% 24.4%
Net margin 26.1% 17.7% 29.7% 22.3%
Versus consensus 2010e 2011e 3Q10e 4Q10e
Revenue -2% 1% -5% 0%
EBITDA -9% -9% -11% -1%
EBIT -1% -9% -15% 4%
EBT -2% -9% -2% 6%
Net profit -3% -13% -4% -11%
Consensus 2010e 2011e 3Q10e 4Q10e
Revenue 12,257 12,424 3,111 3,044
EBITDA 5,510 4,981 1,340 1,234
EBIT 3,533 2,805 1,010 715
EBT 3,290 2,768 871 772
Net profit 3,244 2,565 913 763
EBITDA margin 45.0% 40.1% 43.1% 40.5%
EBIT margin 28.8% 22.6% 32.5% 23.5%
Net margin 26.5% 20.6% 29.4% 25.1%
Source: I/B/E/S, HSBC estimates
Catalysts and risks
Potential catalysts include strong operating
metrics such as ASP, bit growth and cost
reductions; potential transactions involving major
creditor shareholders, including a majority stake
sale; greater exposure to NAND; sharp debt
reduction through internal cash generation, and
greater transparency by K-IFRS implementation.
Downside risks include strengthening KRW
versus the USD, JPY and TWD; potential supplydemand
balance deterioration as players migrate
processes and increase capex; slower-thanexpected
process migration due to technical and
equipment procurement difficulties; and potential
litigation.
Forward PB band
60,000
3.5x
Price (KRW)
50,000
40,000
2.5x
30,000
20,000
1.5x
10,000
0.5x
-
Jan-
05
Jan-
06
Jan-
07
Jan-
08
Jan-
09
Jan-
10
Source: HSBC
Fundamentals
Recent results
Hynix posted consensus-beating 2Q10 results with
OP passing the KRW1trn mark for the first time.
Improved profitability was attributable in part to
higher-end specialty DRAM; non-PC DRAM
revenue is to rise to 60% of total in 3Q (2Q: mid-
50%). Continued demand for higher-end products
(such as server DRAM) and mobile DRAM should
remain robust in 2H10, benefiting Hynix which has
strong exposure to these segments. 3Q10 guidance
is relatively positive; it expects DRAM ASP and
bit growth of a ‘slight’ drop and mid-single digit
increase respectively. For NAND it expects midteen
drop in ASP and +20% bit growth.
Operating performance outlook
This year, Hynix is to see a very sharp recovery in
profitability. We forecast Hynix’s consolidated
revenue to rise c57% y-o-y this year, and for
operating profit to rise c20-fold from a low
KRW191bn base last year. At the parent level, we
forecast revenue to rise c60% this year, and for
the company to turn sharply profitable from a
marginal loss last year.
68

TMT
Memory Devices Technology
8 October 2010
abc
Hynix consolidated revenue breakdown
100%
80%
industry-wide supply and our assumption of
another 6% strengthening in the KRW.
DRAM revenue (KRWbn), operating margins
60%
12,000
40.0%
40%
20%
10,000
8,000
6,000
20.0%
0.0%
0%
2007 2008 2009 2010f 2011f 2012f
DRAM NAND MCP and others
4,000
2,000
-
2007 2008 2009 2010f 2011f 2012f
-20.0%
-40.0%
Source: Company, HSBC estimates
Revenue (KRW bn)
OPM
Consolidated revenue, OP (KRWbn) and OPM (%)
20,000
40%
15,000
20%
10,000
0%
5,000
-
-20%
(5,000) 2008 2009 2010f 2011f 2012f -40%
Revenue OP OP margin(%)
Source: Company, HSBC estimates
For 2011e, we think the impact of lower ASPs is
unavoidable. We forecast consolidated revenue to
rise c6% y-o-y and consolidated operating profit
to decline 10% y-o-y. At the parent level, this
translates into a 4% increase in revenue and a
25% decline in operating profit. Our 2011e parent
revenue forecast is in line with consensus, but our
operating profit forecast is 9% below consensus.
DRAM
Revenue, margin outlook
This year, we forecast consolidated revenue to
rise 64% in KRW terms, on a 42% bit growth and
blended ASP increase of 27%. We expect the
KRW to strengthen 12% compared to last year.
We forecast 2011e consolidated revenue growth
to fall 4%; bit growth is to remain strong at 51%
y-o-y but is offset by a 32% ASP fall on greater
Source: Company, HSBC estimates
We expect greater cost efficiencies (mostly by
process migration and improved production
techniques) plus operating leverage to expand
operating profitability to 36.1% from 3.7% last
year, reducing to 29.3% in 2011e. For 3Q10,
Hynix guides DRAM ASP and bit growth of a
‘slight’ drop and mid-single digit rise respectively.
Bit growth, ASP assumptions (%)
60%
40%
20%
0%
-20%
-40%
-60%
2008 2009 2010f 2011f 2012f
Bit growth ASP
Source: Company, HSBC estimates
Technology – process migration
Positively, Hynix’s focus remains on profitability
enhancement by 1) increasing higher margin
DDR3 and specialist DRAM proportions and 2)
cost reduction through process migration. DDR3
proportion at end-2009 was c50%; by end-2010
the DDR3 proportion should rise to c70-80%.
69

TMT
Memory Devices Technology
8 October 2010
abc
Together with Samsung, Hynix is leading the
market in DRAM process migration. As at end-
2009, c45% of DRAM was made on the 54nm
process, plus a low single digit proportion of
44nm process DRAM. Hynix is ramping 44nm
production, initially focussed on the lucrative
server market. Hynix achieved low-teens bit
production on the 44nm process by 2Q10, slightly
slower than previous guidance of mid-teens.
DRAM process migration
100%
80%
60%
40%
20%
0%
2007 2008 2009 2010f 2011f 2012f
8xnm 6xnm 5xnm 4xnm 3xnm 2xnm
Source: Company, HSBC estimates
Nevertheless, the firm expects 44nm migration to
reach 30% of DRAM bit production by end of 3Q
and 50% by end of 4Q, following its revised
capex in June. We expect Hynix to start mass
production of its 3xnm process during 1Q11.
growth areas of mobile, graphics, high-end server
and specialist DRAM (used in consumer
electronics). Hynix plans to raise global mobile
DRAM market share to c30% in 2010 from c7.1%
in 2007. Hynix expects the revenue proportion of
non-PC DRAM to rise to 60% of total in 3Q (2Q:
mid-50%; 2009: c53%). Server, mobile and
specialist DRAM accounted for c20%, c10% and
c10% of revenue in 2Q10.
The main advantage of these segments, other than
high growth rates, is relatively stable margins.
Mobile DRAM prices tend not to be as volatile as
PC DRAM; ASPs tend to fall c5% every quarter
5%. In addition to PC and server use, consumer
electronics devices are now migrating to DDR3,
driving continued recovery of ASPs for specialist
DRAM.
NAND
Revenue, margin outlook
This year, we forecast consolidated revenue to
rise 39% in KRW terms, on 94% bit growth and
ASP decline of 21%. We expect the KRW to
strengthen 12% compared to last year.
NAND revenue growth
4,000
20.0%
Wafer capacity (kwpy), average fab utilisation
5,000
4,000
3,000
2,000
1,000
-
100%
95%
90%
85%
80%
75%
70%
3,000
2,000
1,000
-
2007 2008 2009 2010f 2011f 2012f
Revenue (KRW bn) OPM
10.0%
0.0%
-10.0%
-20.0%
-30.0%
2007 2008 2009 2010f 2011f 2012f
Source: Company, HSBC estimates
Wafer capacity (kwpy) Average utilisation (%)
Source: Company, HSBC estimates
Speciality DRAM
A key Hynix strategy this year has been to
diversify from commodity memory into rapid
We forecast 2011e revenue growth of 53%; bit
growth is to accelerate to 167% y-o-y on 33% rise
in wafer capacity and process migration. We
assume a 35% ASP fall. We also assume another
6% strengthening in the KRW versus the USD.
70

TMT
Memory Devices Technology
8 October 2010
abc
We are positive on Hynix’s NAND prospects on
two accounts – rising fab capacity and accelerated
process migration. Hynix lost market share in the
past two years as it shut down uneconomical
200mm fabs. But with strong NAND demand and
removal of excess global capacity, and the
financial ability to build capacity, Hynix is in a
favourable environment to raise NAND
production. This year’s capex includes capacity
expansion of M11 (currently 50-60kwpm at
300mm) to 80kwpm by year-end.
Technology – process migration
Following a couple years of production stagnation,
Hynix’s priority is to narrow the technology gap
with the NAND leaders. By this year-end, Hynix
aims to have over 50% of NAND production on
the 32nm process with another c35% on 41nm.
Still, Hynix’s margins remain lower than rivals’.
The current ramp up to 32nm process should
partially close this gap. Hynix targets >50% of
NAND bit production of 32nm by year-end.
NAND process migration
Production capacity, capex
In June, Hynix raised 2010 capex by 33% to
KRW3.05trn. This increase was not a surprise; as
with Samsung, market expectations of Hynix’s
2010 capex had already been higher. Positively,
Hynix is spending the extra capex to accelerate
process migration, raising its 44nm DRAM bit
proportion from current c15% to c50% by yearend.
More recently in October, Hynix raised its
2010 capex further to KRW3.38trn to accelerate
migration to 3xnm DRAM. Effectively, it brought
forward some capex from next year to compete
with Samsung, which started mass production on
the 3xnm process in July this year.
Consolidated capex (KRWbn)
5,000
4,000
3,000
2,000
1,000
-
2008 2009 2010f 2011f 2012f
100%
80%
60%
Source: Company, HSBC estimates
DRAM
NAND
40%
20%
0%
2007 2008 2009 2010f 2011f 2012f
8xnm 6xnm 5xnm 4xnm 3xnm 2xnm
Source: Company, HSBC estimates
Key to longer term competitiveness is process
migration to 26nm at which level Hynix believes
that it can generate comparable margins. In this
respect, things are looking positive, as Hynix
completed development of its 26nm process based
64Gb NAND in February 2010 and customer
samples are out, with certification within six months.
Creditor shareholding
In early March 2010 Hynix creditors sold down a
6.7% stake (one creditor, Korea Finance Corp.,
did not want to sell) in the firm through a block
sale to mostly domestic institutional investors.
Following the success of this selldown (priced at
zero discount), a follow-through selldown in
2H10 could be successful, especially if
expectations of operating outlook improvements
rise on strong results. Most recently there has
been market speculation over various possible
domestic strategic investors that could acquire the
remaining 15% stake held by creditors.
71

TMT
Memory Devices Technology
8 October 2010
abc
Hynix consolidated revenue and OP model summary
2008 2009 2010e 2011e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e
Revenue (KRWbn)
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
NAND 1,408 1,336 1,851 2,839 401 466 460 524 641 714 738 746
MCP and others 427 568 746 946 149 190 199 209 219 230 242 254
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
Revenue breakdown (%)
DRAM 73% 76% 79% 71% 81% 80% 79% 77% 72% 71% 71% 71%
NAND 21% 17% 15% 21% 14% 14% 15% 16% 21% 22% 22% 21%
MCP and others 6% 7% 6% 7% 5% 6% 6% 7% 7% 7% 7% 7%
Total 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100%
Sequential change (%)
DRAM -20% 20% 64% -4% 4% 15% -5% -1% -8% 1% 7% 2%
NAND -23% -5% 39% 53% -10% 16% -1% 14% 22% 11% 3% 1%
MCP and others -23% 33% 31% 27% -11% 27% 5% 5% 5% 5% 5% 5%
Total -21% 16% 57% 6% 1% 16% -4% 1% -2% 3% 6% 2%
EBIT (KRWbn)
DRAM (979) 220 3,561 2,767 740 1,103 897 821 595 607 763 801
NAND (349) (2) 193 409 30 50 55 59 115 159 119 16
MCP and others (135) (27) 20 202 29 (74) 30 36 39 46 53 63
Total (1,920) 191 3,741 3,379 799 1,045 982 915 749 812 936 881
EBIT breakdown (%)
DRAM 51% 115% 95% 82% 93% 106% 91% 90% 79% 75% 82% 91%
NAND 18% -1% 5% 12% 4% 5% 6% 6% 15% 20% 13% 2%
MCP and others 7% -14% 1% 6% 4% -7% 3% 4% 5% 6% 6% 7%
Total 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100%
Sequential change (%)
DRAM -274% -122% 1518% -22% 17% 49% -19% -9% -27% 2% 26% 5%
NAND -216% -100% -11311% 112% -63% 67% 11% 7% 95% 39% -25% -86%
MCP and others 61% -80% -172% 923% -1029% -358% -140% 19% 11% 17% 16% 19%
Total -490% -110% 1858% -10% 13% 31% -6% -7% -18% 8% 15% -6%
EBIT margin (%)
DRAM -20% 4% 36% 29% 33% 42% 36% 33% 26% 27% 31% 32%
NAND -20% 0% 10% 14% 7% 11% 12% 11% 18% 22% 16% 2%
MCP and others -32% -5% 3% 21% 19% -39% 15% 17% 18% 20% 22% 25%
Total -28% 2% 30% 26% 28% 32% 31% 29% 24% 25% 27% 25%
Source: Company, HSBC estimates
Forecast revisions
KRWbn ________________Revised ________________ _______________Previous ________________ _______________Revisions _______________
2010f 2011f 2012f 2010f 2011f 2012f 2010f 2011f 2012f
Revenue 12,045 12,557 13,724 12,061 13,043 14,236 0% -4% -4%
EBIT 3,488 2,546 2,389 3,106 2,590 2,548 12% -2% -6%
Net income 3,143 2,220 2,107 3,050 2,253 2,060 3% -1% 2%
OP margin 29.0% 20.3% 17.4% 25.8% 19.9% 17.9%
NP margin 26.1% 17.7% 15.4% 25.3% 17.3% 14.5%
Source: Company data, HSBC estimates
72

TMT
Memory Devices Technology
8 October 2010
abc
Financials & valuation: Hynix Semiconductor Inc
Overweight (V)
Financial statements
Year to 12/2009a 12/2010e 12/2011e 12/2012e
Profit & loss summary (KRWb)
Revenue 7,521 12,045 12,557 13,724
EBITDA 1,722 5,033 4,510 4,613
Depreciation & amortisation -1,838 -1,544 -1,964 -2,224
Operating profit/EBIT -116 3,488 2,546 2,389
Net interest -294 -162 -136 -95
PBT -398 3,223 2,523 2,479
HSBC PBT -398 3,223 2,523 2,479
Taxation 50 -80 -303 -372
Net profit -348 3,143 2,220 2,107
HSBC net profit -348 3,143 2,220 2,107
Cash flow summary (KRWb)
Cash flow from operations 735 4,085 3,807 3,859
Capex -1,478 -1,544 -1,960 -2,220
Cash flow from investment -1,461 -2,305 -1,982 -2,169
Dividends 0 0 0 0
Change in net debt -1,012 -558 -1,465 -1,165
FCF equity -744 2,461 1,624 1,571
Balance sheet summary (KRWb)
Intangible fixed assets 461 461 461 461
Tangible fixed assets 6,325 7,861 8,217 8,738
Current assets 3,568 4,611 6,342 7,834
Cash & others 1,329 575 1,529 2,254
Total assets 13,425 16,525 18,612 20,625
Operating liabilities 3,193 4,150 4,528 4,874
Gross debt 4,779 3,467 2,957 2,517
Net debt 3,450 2,893 1,428 263
Shareholders funds 5,452 8,907 11,127 13,234
Invested capital 6,219 8,208 8,963 9,905
Ratio, growth and per share analysis
Year to 12/2009a 12/2010e 12/2011e 12/2012e
Y-o-y % change
Valuation data
Year to 12/2009a 12/2010e 12/2011e 12/2012e
EV/sales 2.0 1.1 1.0 0.8
EV/EBITDA 8.6 2.7 2.7 2.4
EV/IC 2.4 1.7 1.4 1.1
PE* 4.7 6.6 7.0
P/Book value 2.5 1.6 1.3 1.1
FCF yield (%) -6.5 22.7 15.0 14.5
Dividend yield (%) 0.0 0.0 0.0 0.0
Note: * = Based on HSBC EPS (fully diluted)
Issuer information
Share price (KRW) 23,400 Target price (KRW) 32,000 Potent’l return (%) 36.8
Reuters (Equity) 000660.KS Bloomberg (Equity) 000660 KS
Market cap (USDm) 12,221 Market cap (KRWb) 13,812
Free float (%) 81 Enterprise value (KRWb) 13728
Country Korea Sector Semiconductors
Analyst Nam Park Contact 852 2996 6591
Price relative
37507
32507
27507
22507
17507
12507
7507
2507
Source: HSBC
2008 2009 2010 2011
Hynix Semiconductor Inc Rel to KOSPI INDEX
Note: price at close of 06 Oct 2010
37507
32507
27507
22507
17507
12507
7507
2507
Revenue 15.8 60.1 4.2 9.3
EBITDA 192.2 -10.4 2.3
Operating profit -27.0 -6.2
PBT -21.7 -1.7
HSBC EPS -29.4 -5.1
Ratios (%)
Revenue/IC (x) 1.2 1.7 1.5 1.5
ROIC -0.9 47.9 26.6 22.0
ROE -6.6 43.8 22.2 17.3
ROA -0.5 22.2 13.4 11.3
EBITDA margin 22.9 41.8 35.9 33.6
Operating profit margin -1.5 29.0 20.3 17.4
EBITDA/net interest (x) 5.9 31.1 33.2 48.5
Net debt/equity 63.3 32.5 12.8 2.0
Net debt/EBITDA (x) 2.0 0.6 0.3 0.1
CF from operations/net debt 21.3 141.2 266.6 1467.8
Per share data (KRW)
EPS Rep (fully diluted) -589.83 5017.81 3544.80 3364.65
HSBC EPS (fully diluted) -589.83 5017.81 3544.80 3364.65
DPS 0.00 0.00 0.00 0.00
Book value 9246.98 14222.59 17767.39 21132.04
73

TMT
Memory Devices Technology
8 October 2010
abc
Inotera Memories (3474)
A fully-migrated 50nm producer despite delay; targeting full-42nm
in 1H10
It is crucial how quickly capacity ramp-up and cycle time reduction
can be accomplished, amid falling DRAM prices
Cut 2010-11e revenue by 13% and 8%, cut TP to TWD17
(previously TWD20), reiterate Neutral (V).
Investment summary
Competing with time
As a foundry for Nanya Tech and Micron,
Inotera just became Taiwan’s first fullymigrated
50nm fab in 3Q, using Micron’s
technology.
The migration, however, was behind schedule
in terms of wafer output delivery as cycle
time has been lengthened as node migration
was complicated by the trench to stack
process at the same time.
Meanwhile, we believe Inotera’s yield has
been improving on 50nm despite the twomonth
delay on wafer output.
Still, we believe such problems encountered
on node migration should continue to affect
Inotera’s bit growth (hence Nanya Tech’s)
until the cycle time can be reduced.
With the delay, we now expect Inotera to
achieve bit growth of 27% y-o-y (previously
67%) in 2010, more conservative than Inotera’s
revised guidance of 50% (previously 70-80%).
Bottom line is that we believe Inotera will
still experience a loss for 2010. Whether they
will be able to deliver a profit in 4Q depends
on how quickly they can reduce cycle time,
how rapidly they can improve yield and how
fast they can ramp up their 50nm capacity.
Currently, we expect Inotera to marginally
break even at net level for 4Q10. Inotera
targets to ramp up their capacity to 130k by
the end of 4Q.
Fundamentals
Technology migration
Starting from mid-July, Inotera has successfully
turned itself into a full 50nm fab. The next task
for Inotera is to 1) ramp up current 50nm
capacity and 2) improve cycle time for wafer
input to come through the process.
We believe wafer input for 3Q is around
averaged 80-85k/month. This compares to
Inotera’s target of ramping up capacity to
130k/month by the end of 4Q.
We estimate Inotera’s current wafer cycle
time to be around 70 days. Inotera’s target is
to reduce it to the mid-50s by end-2010.
74

TMT
Memory Devices Technology
8 October 2010
abc
We estimate that Inotera’s 50nm technology
proportion represents around 22% of total
shipments in 3Q10, compared to the
scheduled 50%. We believe the delay will
likely affect bit growth in 4Q, where Inotera
had targeted for 100% growth.
On 42nm, Inotera will start its pilot run in late
3Q. It is expected to have mass production by
1H11.
Bit growth, ASP assumptions (%)
150%
100%
50%
0%
-50%
2008 2009 2010f 2011f 2012f
Bit growth ASP
Source: Company data, HSBC estimates
Process migration
100%
80%
60%
40%
20%
0%
2007 2008 2009 2010f 2011f 2012f
9xnm
70nm
50nm
42nm
Source: Company data, HSBC estimates
Forecast revision
On the delay of 50nm migration, lower wafer
output in 3Q as a result of prolonged cycle time,
we revise down Inotera’s 2010 forecast and now
believe significant cost savings from node
migration will only start in 4Q10-1Q11. We now
forecast 4Q10 bit growth of 51% (company
targets 100%) and 2010 bit growth of 27% y-o-y
(previously 67%). We revise down Inotera’s
2010-11e OP margin to -4% and 11%, from 6%
and 18% respectively, and expect the margin
recovery only to happen in 4Q. We now forecast
Inotera’s ASP to go down by 16% in 3Q and
another 10% in 4Q.
Forecast revision for Inotera
TWDm __ Revised ___ __ Previous ___ __Difference _
2010 2011 2010 2011 2010 2011
Sales 46,535 66,433 53,426 72,032 -13% -8%
GP (1,987.1) 7,303 3,049 13,042 nm -44%
OP (3,755.4) 5,642 1,125 11,385 nm -50%
Net income (5,495.7) 3,167 (589.9) 8,100 nm -61%
EPS (TWD) (1.2) 0.7 (0.1) 2.1 nm -68%
GPM (%) -4% 11% 6% 18%
OPM (%) -8% 8% 2% 16%
NPM (%) -12% 5% -1% 11%
Utilization 70% 70% 77% 77%
rate (%)
Bit growth (%) 27% 102% 67% 68%
ASP growth
(%)
4% -31% -5% -28%
Source: HSBC
For 2010, we forecast Inotera to remain at a loss
of TWD5.5bn, but swing to net profit of
TWD3.2bn next year. In 2011, we expect Inotera
to achieve bit growth of 102%, on bit growth that
has been delayed from 3Q-4Q as a result
migration. We also forecast its margin to improve
to 8% next year from -8% this year.
75

TMT
Memory Devices Technology
8 October 2010
abc
HSBC vs consensus
HSBC 2010e 2011e 2012e
Revenue 46,535 66,433 79,103
EBITDA 28,578 41,830 49,352
EBIT (3,755) 5,642 10,072
Netincome (5,496) 3,167 7,136
EPS(TWD) (1.2) 0.7 1.5
EBITDAmargin 61.4% 63.0% 62.4%
EBITmargin -8.1% 8.5% 12.7%
Netmargin -11.8% 4.8% 9.0%
HSBC vs consensus 2010e 2011e 2012e
Revenue -12.6% -3.1% 9.6%
EBITDA -14.9% -7.7% 0.4%
EBIT -329.6% -43.9% -33.7%
Net income -2178.4% -60.5% -18.1%
EPS (TWD) -761.2% -70.4% -37.9%
Consensus 2010e 2011e 2012e
Revenue 53,219 68,564 72,195
EBITDA 33,583 45,319 49,158
EBIT 1,636 10,050 15,188
Net income 264 8,010 8,715
EPS (TWD) 0.2 2.3 2.5
EBITDA margin 63.1% 66.1% 68.1%
EBIT margin 3.1% 14.7% 21.0%
Net margin 0.5% 11.7% 12.1%
Source: Company data, I/B/E/S consensus
Valuations, rating
We lower our 12-month target price to TWD17
from TWD20, derived by applying a target FY11
PB multiple of 1.5x, compared to 1.7x before. Our
method for valuation remains unchanged from
before, but we roll over to an FY11e PB multiple
from FY10e previously. We use a mid-point
historical PB range to derive the multiple. The
stock has historically traded at PB of 0.5-2.5x,
with an ROE range of c2% to 25% (prior to the
capacity glut starting in 2007).
Under our research model, for Taiwan stocks with
a volatility indicator, the Neutral band is 10ppt
above and below the hurdle rate of 9.5% (risk-free
rate of 4% plus risk premium of 5.5%). This
translates into a Neutral band of -0.5% to +19.5%
around the current share price. Our 12-month
target price of TWD17 suggests a potential return
of 4.7%, which is within the Neutral band; thus,
we reiterate our Neutral (V) rating.
PB vs ROE
3.0
10%
5%
2.0
0%
-5%
1.0
-10%
-15%
0.0
-20%
Apr-06 Apr-07 Apr-08 Apr-09 Apr-10
PB (x , LHS)
ROE (%, RHS)
Source: HSBC
Forward PB band
60
Price (TWD)
40
2.4x
1.8x
20
1.2x
0.6x
-
Apr-06 Apr-07 Apr-08 Apr-09 Apr-10
Source: Datastream, HSBC
Catalysts and risks
Industry-specific catalysts include:
Favourable DRAM demand and supply in
2010. Tight global DRAM supply given
limited capex spending in 2009.
Better than expected results and strong
guidance from PC OEMs, such as recent Dell
and HP results.
PC demand to pick up from 2Q10 onwards,
boosted by corporate replacement and
upgrade spending.
76

TMT
Memory Devices Technology
8 October 2010
abc
Company-specific catalysts include:
Inotera’s cost improvement by migrating to
50nm technology.
Micron and Nanya Tech’s ASP premium to
other players.
Industry-specific risks include:
Faster than expected DRAM ASP erosion due
to greater capacity coming online and/or
slower than expected PC unit shipments.
Company-specific risks include:
Delay in 5xnm and 4cnm technology
migration, lower than expected yield achieved.
Potential requirement for further funding to
support capex.
Since Inotera has proprietary technology, the
cost of transferring technology from Micron
may insert pressure on its profitability.
Raw material price hike like silicon wafers.
Raw material price hike like silicon wafers.
Elpida’s possible alliance with other
Taiwanese DRAM makers likely help create a
more competitive environment among
Taiwanese DRAM players.
Quarterly assumptions for Inotera
TWDm 1Q10a 2Q10e 3Q10e 4Q10e 1Q10a 2Q11e 3Q11e 4Q11e
Sales 11,518 11,258 9,963 13,796 16,694 16,414 16,907 16,418
GP (559) (760) (1805) 1137 2,053 1,826 1,972 1,452
OP (1219) (1142) (2164) 769 1,686 1,399 1,533 1,025
Net income (1561) (1810) (2305) 180 1,086 828 948 305
EPS (TWD) (0.4) (0.4) (0.5) 0.0 0.2 0.2 0.2 0.1
GPM (%) -5% -7% -18% 8% 12% 11% 12% 9%
OPM (%) -11% -10% -22% 6% 10% 9% 9% 6%
NPM (%) -14% -16% -23% 1% 7% 5% 6% 2%
Utilization rate (%) 78% 67% 67% 70% 83% 85% 89% 92%
Bit growth (%) -3% -15% 4% 51% 32% 9% 14% 8%
ASP growth (%) -6% 16% -16% -10% -9% -10% -10% -10%
Source: Company data, HSBC estimates
77

TMT
Memory Devices Technology
8 October 2010
abc
Financials & valuation: Inotera Memories Inc.
Neutral (V)
Financial statements
Year to 12/2009a 12/2010e 12/2011e 12/2012e
Profit & loss summary (TWDm)
Revenue 36,104 46,535 66,433 79,103
EBITDA 20,154 28,578 41,830 49,352
Depreciation & amortisation -29,795 -32,334 -36,187 -39,279
Operating profit/EBIT -9,641 -3,755 5,642 10,072
Net interest -1,549 -2,047 -2,277 -1,582
PBT -11,477 -5,496 3,725 8,920
HSBC PBT -11,477 -5,496 3,725 8,920
Taxation 0 0 -559 -1,784
Net profit -11,477 -5,496 3,167 7,136
HSBC net profit -11,477 -5,496 3,167 7,136
Cash flow summary (TWDm)
Cash flow from operations 11,228 24,158 34,449 44,520
Capex -13,095 -58,000 -22,300 -12,500
Cash flow from investment -23,341 -35,800 -31,429 -31,629
Dividends 0 0 0 0
Change in net debt -14,367 25,826 -13,822 -29,536
FCF equity -1,867 -33,842 11,591 30,236
Balance sheet summary (TWDm)
Intangible fixed assets 1,447 1,037 779 675
Tangible fixed assets 100,032 125,390 110,889 85,343
Current assets 18,917 18,179 28,157 44,251
Cash & others 5,376 442 3,263 14,799
Total assets 120,396 144,606 139,826 130,269
Operating liabilities 29,459 31,873 34,927 36,234
Gross debt 42,109 63,000 52,000 34,000
Net debt 36,733 62,558 48,737 19,201
Shareholders funds 48,828 49,733 52,899 60,035
Invested capital 85,561 112,291 101,636 79,235
Ratio, growth and per share analysis
Year to 12/2009a 12/2010e 12/2011e 12/2012e
Y-o-y % change
Valuation data
Year to 12/2009a 12/2010e 12/2011e 12/2012e
EV/sales 3.1 3.0 1.9 1.2
EV/EBITDA 5.5 4.8 3.0 1.9
EV/IC 1.3 1.2 1.2 1.2
PE* 23.6 10.5
P/Book value 1.2 1.5 1.4 1.2
FCF yield (%) -2.5 -45.2 15.5 40.4
Dividend yield (%) 0.0 0.0 0.0 0.0
Note: * = Based on HSBC EPS (fully diluted)
Issuer information
Share price (TWD) 16.20 Target price (TWD) 17.0 Potent’l return (%) 4.7
Reuters (Equity) 3474.TW Bloomberg (Equity) 3474 TT
Market cap (USDm) 2,407 Market cap (TWDm) 74,804
Free float (%) 46 Enterprise value (TWDm) 137362
Country Taiwan Sector Electronic Equipment
Analyst Carolyn Poon Contact +852 2996 6586
Price relative
32
27
22
17
12
7
2
2008 2009 2010 2011
Inotera Memories Inc. Rel to TAIWAN WEIGHTED INDEX
Source: HSBC
Note: price at close of 06 Oct 2010
32
27
22
17
12
7
2
Revenue -3.8 28.9 42.8 19.1
EBITDA 108.8 41.8 46.4 18.0
Operating profit 78.5
PBT 139.4
HSBC EPS 125.3
Ratios (%)
Revenue/IC (x) 0.4 0.5 0.6 0.9
ROIC -10.1 -3.8 4.4 8.7
ROE -23.2 -11.2 6.2 12.6
ROA -7.8 -2.6 3.6 6.3
EBITDA margin 55.8 61.4 63.0 62.4
Operating profit margin -26.7 -8.1 8.5 12.7
EBITDA/net interest (x) 13.0 14.0 18.4 31.2
Net debt/equity 75.2 125.8 92.1 32.0
Net debt/EBITDA (x) 1.8 2.2 1.2 0.4
CF from operations/net debt 30.6 38.6 70.7 231.9
Per share data (TWD)
EPS Rep (fully diluted) -3.19 -1.19 0.69 1.55
HSBC EPS (fully diluted) -3.19 -1.19 0.69 1.55
DPS 0.00 0.00 0.00 0.00
Book value 13.56 10.77 11.46 13.00
78

TMT
Memory Devices Technology
8 October 2010
abc
Nanya Technology (2408)
Bit growth limited by Inotera’s 50nm tech migration delay, but
determined to move fast on 42nm node migration
Expectation of more DRAM supply in 2H has dragged DRAM
price exogenously
Cut TP to TWD20 from TWD25, maintain Neutral (V) rating
Investment summary
Are we there yet?
Nanya Technology is the first Taiwanese
DRAM company to start 50nm technology
migration, which is a long-term positive for
the company’s competitiveness. The
execution, however, took longer than
expected time and hence delivering less bit
growth than planned, as foundry Inotera has
delayed its migration process starting from
2Q which impacted wafer output in 2Q-3Q.
The process is complicated by the fact that
Inotera is switching from trench to stack
process at the same time.
Such delay has, unfortunately, coincided with
an ASP decline in 2H, thanks to softer PC
demand and the expectation that more
industry supply (bit growth) expected to be
introduced to the market in 4Q.
The bottom line is that Nanya Tech’s net
profit and margin recovery story is not likely
to materialise this year but will be postponed
to next year: We now forecast net income loss
of TWD5bn for this year, with a net profit of
TWD2.9bn in 2011. Sudden ramp-up of bit
growth, which the market expects to see in
4Q10, should be softer than expected while
postponing to 1H11.
It is also crucial to see whether Nanya Tech
can make its 40nm migration ahead of
schedule and get the yield right quickly amid
an environment where DRAM ASPs are
pointing south.
Fundamentals
Technology migration
Nanya’s bit growth has been slower than planned
in 2Q and 3Q, on its foundry (Inotera)’s 50nm
migration delay. We estimate that Nanya Tech’s
50nm technology proportion represents around
35% of total shipment in 2Q10, compared to the
scheduled 50%. We believe the delay will
continue, resulting in lower-than-expect DRAM
bit growth in 2H, more than offsetting bit growth
output driven by 50nm capacity ramp up in 4Q.
Still, Nanya is accelerating its own fab’s 42nm
migration and expects to have mass production
by 4Q10.
Capacity ramp-up: part of the bit growth in 2H
should also come from its own capacity ramp-up.
Nanya’s own capacity is expected to reach
79

TMT
Memory Devices Technology
8 October 2010
abc
50kwpm by 4Q, from 40kwpm at end-2Q, and
further to 60kwpm by 1H11.
Bit growth, ASP assumptions (%)
150%
100%
50%
0%
-50%
-100%
2008 2009 2010f 2011f
Supply bit growth
ASP
Source: Company data, HSBC estimates
2H result
Nanya’s 3Q revenue of TWD14.9bn, was 5% q-oq
weaker. While details of the result will not be
announced until late October, 3Q bit growth was
likely dragged by Inotera’s longer cycle time,
despite Nanya’s own 50nm migration being on
schedule. Together with a DRAM ASP fall, we
forecast Nanya Tech to linger in the red at both
the operating and net profit levels this year,
delivering a loss of TWD5.0bn for 2010. We
expect 4Q to have a recovery of bit growth of
29% as 50nm wafer input in 3Q has started to
materialise, with revenue growth of 7.5% q-o-q
and a net loss of TWD1.0bn.
Process migration
100%
80%
60%
40%
Crucial to get yield right
Unless Inotera can successfully reduce its cycle
time from c70 days to its target of mid-50 days in
4Q10, it appears difficult for Nanya to catch up
with the annual 35% bit growth target for 2010.
We believe it is thus crucial that Nanya can
acceleration its own migration to 40nm ahead of
schedule, which should at least offset some of the
bit growth loss caused by Inotera’s migration
delay. We have revised down Nanya’s annual bit
growth to 10% from 42% for 2010.
Forecast revision
Forecast revision for Nanya Tech
TWDm __ Revised ___ __ Previous ___ __Difference
2010 2011 2010 2011 2010 2011
Sales 60,787 83,749 74,339 101,705 -18% -18%
GP 3,313 15,514 9,846 20,370 -66% -24%
OP (3,660) 6,553 2,635 8,572 nm nm
Net income (4,999) 2,849 1,889 6,152 nm nm
EPS (TWD) (1.46) 0.83 0.55 1.79 nm nm
GPM (%) 5% 19% 13% 20%
OPM (%) -6% 8% 4% 8%
NPM (%) -8% 3% 3% 6%
Utilization 82% 88% 89% 89%
rate (%)
Bit growth (%) 10% 110% 42% 45%
ASP growth
(%)
34% -36% -15% -14%
Source: HSBC estimates
On 1) lower DRAM ASP expectation in 2H10,
and 2) Inotera’s longer-than-expected cycle time
in 50nm node migration, we lower our Nanya’s
2010-11 revenue forecast by 18%. We now
forecast Nanya’s 2010 sales growth of 43% for
2010. We now believe Nanya Tech will achieve a
-6% OPM for 2010 instead of our previous
forecast of 4%.
20%
0%
2007 2008 2009 2010f 2011f 2012f
90nm 70nm 50nm 42nm 3x nm 2x nm
Source: Company data, HSBC estimates
80

TMT
Memory Devices Technology
8 October 2010
abc
HSBC vs. consensus
HSBC 2010e 2011e 2012e
Revenue 60,787 83,749 98,318
EBITDA 12,620 25,679 32,510
EBIT (3,660) 6,553 10,748
Net income (4,999) 2,849 6,311
EPS (KRW) (1.5) 0.8 1.9
EBITDA margin 20.8% 30.7% 33.1%
EBIT margin -6.0% 7.8% 10.9%
Net margin -8.2% 3.4% 6.4%
HSBC vs
2010e 2011e 2012e
consensus
Revenue -15.6% -11.9% -11.2%
EBITDA -24.3% 9.2% -1.5%
EBIT 134.8% 21.4% 15.5%
Net income 798.2% -19.6% -14.0%
EPS (KRW) 91.1% -18.4% -14.2%
Consensus 2010e 2011e 2012e
Revenue 72,064 95,109 110,667
EBITDA 16,677 23,510 33,007
EBIT (1,559) 5,397 9,304
Net income (555) 3,586 7,397
EPS (KRW) (0.8) 1.0 2.2
EBITDA margin 23.1% 24.7% 29.8%
EBIT margin -2.2% 5.7% 8.4%
Net margin -0.8% 3.8% 6.7%
Source: Company data, I/B/E/S consensus
Catalysts
Industry-specific catalysts include
DRAM ASP fall less than expected in the
coming quarters
PC demand better than expected on corporate
demand pick up
Higher-than-expected DRAM content per box
on lower DRAM ASP
Company-specific catalysts include:
40nm migration comes ahead of schedule
Diversification of DRAM business to less
volatile and more specialized areas like
mobile DRAM.
Risks
Industry-specific risks include
Sluggish PC demand, delayed corporate
replacement demand
tablet cannibalisation higher than expected
Elpida’s possible alliance with other
Taiwanese DRAM makers likely help create a
more competitive environment among
Taiwanese DRAM players
Company-specific risks include:
Inotera’s 40nm migration comes behind
schedule
Equity funding likely in 2H which should
lead to equity dilution for current
shareholders
Valuations, rating
We lower our 12-month target price to TWD20 from
TWD25, derived by applying a target FY11e PB
multiple of 1.9x (based on 12.2% ROE (11.3%
previously) and 9.5% COE), compared to 2.1x
before. We also roll over our PB target multiple to
2011e from 2010e. The PB multiple of 1.9x is close
to the lower end of Nanya’s historical PB range.
Our method for valuation remains unchanged
from before. The stock has historically traded at a
PB of 1.6-4.9x, with an ROE range of -47% to 9%
(prior to the capacity glut starting in 2007).
Under our research model, for Taiwan stocks with
a volatility indicator, the Neutral band is 10ppt
above and below the hurdle rate of 9.5% (risk-free
rate of 4% plus risk premium of 5.5%). This
translates into a Neutral band of -0.5% to +19.5%
around the current share price. Our 12-month
target price of TWD20 suggests a potential return
of 3.6%, which is within the Neutral band; thus,
we reiterate our Neutral (V) rating.
81

TMT
Memory Devices Technology
8 October 2010
abc
PB vs ROE
Forward PB band
6.0
5.0
4.0
3.0
2.0
1.0
50%
30%
10%
-10%
-30%
-50%
80
60
40
20
-
Price (TWD)
4.0x
3.0x
2.0x
1.0x
Jan-05 Jan-06 Jan-07 Jan-08 Jan-09
PB (x , LHS)
ROE (%, RHS)
Jan-
05
Jan-
06
Jan-
07
Jan-
08
Jan-
09
Jan-
10
Source: HSBC
Source: Datastream, HSBC
Assumptions on Nanya Tech
TWDm 1Q10a 2Q10a 3Q10e 4Q10e 1Q10e 2Q10e 3Q10e 4Q10e
Sales 14,121 15,718 14,916 16,033 18,337 20,440 22,437 22,535
GP 189 1,589 705 829 2,141 3,517 4,804 5,051
OP (1225) (516) (933) (986) 198 1,310 2,314 2,731
Net income (1625) (1086) (1277) (997) (719) 393 1,522 1,689
EPS (TWD) (0.48) (0.32) (0.37) (0.29) (0.21) 0.11 0.44 0.49
GPM (%) 1% 10% 5% 5% 12% 17% 21% 22%
OPM (%) -9% -3% -6% -6% 1% 6% 10% 12%
NPM (%) -12% -7% -9% -6% -4% 2% 7% 7%
Utilization rate (%) 80% 87% 80% 83% 85% 88% 89% 90%
Bit growth (%) -9% 3% 12% 29% 26% 21% 19% 9%
ASP growth (%) -6% 9% -16% -18% -10% -8% -8% -8%
Source: Company data, HSBC estimates
82

TMT
Memory Devices Technology
8 October 2010
abc
Financials & valuation: Nanya Technology
Financial statements
Year to 12/2009a 12/2010e 12/2011e 12/2012e
Profit & loss summary (TWDm)
Revenue 42,456 60,787 83,749 98,318
EBITDA -1,158 12,620 25,679 32,510
Depreciation & amortisation -14,761 -16,280 -19,126 -21,762
Operating profit/EBIT -15,919 -3,660 6,553 10,748
Net interest -1,068 -1,574 -2,492 -2,859
PBT -20,754 -5,434 3,561 7,889
HSBC PBT -20,754 -5,434 3,561 7,889
Taxation 0 435 -712 -1,578
Net profit -20,754 -4,999 2,849 6,311
HSBC net profit -20,754 -4,999 2,849 6,311
Cash flow summary (TWDm)
Cash flow from operations -10,211 1,101 12,735 24,529
Capex -7,533 -30,000 -25,500 -25,500
Cash flow from investment -9,071 -14,490 -11,171 -25,836
Dividends 0 0 0 0
Change in net debt -5,821 35,446 17,648 3,885
FCF equity -17,744 -28,465 -13,477 -2,548
Balance sheet summary (TWDm)
Intangible fixed assets 4,993 10,464 11,754 11,381
Tangible fixed assets 74,691 86,209 94,964 101,411
Current assets 37,296 33,584 45,324 49,316
Cash & others 13,663 1,509 1,861 -1,024
Total assets 132,953 148,230 170,015 180,083
Operating liabilities 35,174 31,158 32,094 34,850
Gross debt 60,405 83,697 101,697 102,697
Net debt 46,742 82,188 99,836 103,721
Shareholders funds 37,375 33,375 36,224 42,536
Invested capital 68,142 97,589 118,086 128,282
Ratio, growth and per share analysis
Year to 12/2009a 12/2010e 12/2011e 12/2012e
Y-o-y % change
Revenue 16.9 43.2 37.8 17.4
EBITDA 103.5 26.6
Operating profit 64.0
PBT 121.5
HSBC EPS 121.5
Ratios (%)
Revenue/IC (x) 1.2 1.2 1.4 1.5
ROIC -39.8 -3.0 12.1 16.8
ROE -63.3 -14.1 8.2 16.0
ROA -29.2 -4.4 5.8 9.6
EBITDA margin -2.7 20.8 30.7 33.1
Operating profit margin -37.5 -6.0 7.8 10.9
EBITDA/net interest (x) 8.0 10.3 11.4
Net debt/equity 125.1 246.3 275.6 243.8
Net debt/EBITDA (x) -40.4 6.5 3.9 3.2
CF from operations/net debt 1.3 12.8 23.6
Per share data (TWD)
EPS Rep (fully diluted) -8.67 -1.46 0.83 1.84
HSBC EPS (fully diluted) -8.67 -1.46 0.83 1.84
DPS 0.00 0.00 0.00 0.00
Book value 15.61 9.72 10.55 12.39
Valuation data
Year to 12/2009a 12/2010e 12/2011e 12/2012e
EV/sales 2.3 2.1 1.8 1.5
EV/EBITDA 10.3 5.8 4.7
EV/IC 1.4 1.3 1.3 1.2
PE* 23.3 10.5
P/Book value 1.2 2.0 1.8 1.6
FCF yield (%) -35.3 -59.0 -27.9 -5.3
Dividend yield (%) 0.0 0.0 0.0 0.0
Note: * = Based on HSBC EPS (fully diluted)
Issuer information
Share price (TWD) 19.30 Target price (TWD) 20.00 Potent’l return (%) 3.6
Reuters (Equity) 2408.TW Bloomberg (Equity) 2408 TT
Market cap (USDm) 2,131 Market cap (TWDm) 66,237
Free float (%) 74 Enterprise value (TWDm 130451
)
Country Taiwan Sector Semiconductors
Analyst Carolyn Poon Contact +852 2996 6586
Price relative
64
54
44
34
24
14
4
2008 2009 2010 2011
Nanya Technologies Rel to TAIWAN WEIGHTED INDEX
Source: HSBC
Note: price at close of 06 Oct 2010
Neutral (V)
64
54
44
34
24
14
4
83

TMT
Memory Devices Technology
8 October 2010
abc
Samsung Electronics (005930)
Leading DRAM and NAND technologies, market share, strong
balance sheet and vertical integration make SEC our top pick
Samsung’s strong product mix and accelerated process migration
leverages into structural growth areas such as mobile, servers
Remain OW, TP unchanged at KRW1,139,000. An HSBC Asia
Super Ten idea
Investment summary
For detailed analysis, please refer to our recent
40-page report, Samsung Electronics – OW:
Comprehensive health check, 24 September 2010.
The memory leader
Samsung Electronics is our preferred play in the
memory sector. Samsung has:
Highly competitive technologies and leading
market share leadership in both DRAM and
NAND (which is faster growing).
Leading process migration, having started
35nm DRAM production in 3Q10. We also
expect 2xnm NAND production to start in
4Q10.
Strong presence in promising technologies
such as phase change memory. Only Micron
now has a comparable product portfolio post
its recent acquisition of Numonyx.
A massive balance sheet (cUSD10bn net cash
at the parent level at end-2009) cash flow
backing R&D and capex where necessary.
currently playing out with rapid process
migration and slew of advanced products.
A level of vertical integration that is unique
amongst its peers, giving it access to a vast
internal market of electronics products.
Valuations, rating
Valuations summary
Methodology Value (KRW) Criteria
PB 1,323,318 2.2x FY11e BVPS
Sum of parts 978,350
DCF 1,079,363 12%
Average 1,127,010
Dividend 12,000
Target price 1,139,010
Current price 793,000
Up(downside) 43.6%
Source: HSBC
To value Samsung Electronics, we use an average
of 2011e PB (2.2x target multiple), sum-of-theparts,
and DCF valuations (12% WACC with riskfree
rate 4%, equity risk premium 8%, β=1) to
derive a 12-month target price of KRW1,139,000.
Our price assumes a dividend of KRW12,000.
Determination to expand leadership with more
advanced and cost-effective technologies,
84

TMT
Memory Devices Technology
8 October 2010
abc
HSBC vs consensus
(KRWbn) 3Q10e 4Q10e 2010e 2011e
Revenue 41,064 42,884 156,476 169,642
EBITDA 8,083 8,316 31,263 32,792
EBIT 4,942 4,424 18,785 18,072
Net income 4,369 3,781 16,420 16,180
Common EPS (KRW) 29,658 25,669 111,477 109,848
Diluted EPS (KRW) 25,678 22,224 96,516 95,105
EBITDA margin 19.7% 19.4% 20.0% 19.3%
EBIT margin 12.0% 10.3% 12.0% 10.7%
Net margin 10.6% 8.8% 10.5% 9.5%
HSBC vs consensus 3Q10e 4Q10e 2010e 2011e
Revenue 0.0% 1.5% 0.1% 0.6%
EBITDA -7.1% -0.3% 6.3% 12.3%
EBIT -4.6% 6.1% 1.2% 0.1%
Net income -7.1% -2.7% -2.0% 2.2%
EPS (KRW) -11.2% -9.4% -9.1% -4.5%
Consensus 3Q10e 4Q10e 2010e 2011e
Revenue 41,059 42,238 156,312 168,634
EBITDA 8,705 8,341 29,406 29,206
EBIT 5,180 4,169 18,554 18,051
Net income 4,702 3,887 16,750 15,832
EPS (KRW) 28,922 24,523 106,170 99,562
EBITDA margin 21.2% 19.7% 18.8% 17.3%
EBIT margin 12.6% 9.9% 11.9% 10.7%
Net margin 11.5% 9.2% 10.7% 9.4%
Source: I/B/E/S, HSBC forecasts
Under our research model, for stocks without a
volatility indicator, the Neutral band is 5ppt above
and below the hurdle rate of 10.5% for Korean
stocks, or 5.5-15.5% around the current share
Forward PB band
1,400,000
1,300,000
1,200,000
1,100,000
1,000,000
900,000
800,000
700,000
600,000
500,000
400,000
Source: Datastream, HSBC
Price (KRW)
Jan-
05
Jan-
06
Jan-
07
Jan-
08
Jan-
09
Jan-
10
2.5x
2.0x
1.5x
1.0x
price. Our 12-month target price suggests a
potential return of 43.6%, which is above the
Neutral band. Hence we reiterate our Overweight
rating on the stock. Volatile stocks are defined as
those having historical volatility (defined as the
past month’s average of the daily 365-day moving
average volatilities) of more than 40%.
Catalysts and risks
Upside risks to our rating and estimates include
the following:
Stronger-than-expected quarterly results.
New product or new technology launches that
could drive higher revenues and/or margins,
including 4G-related devices, smartphones,
tablet devices.
Business expansion into promising areas, a
pick-up in the global economy and demand.
Downside risks to our rating and estimates include
the following:
Strengthening KRW versus other currencies,
notably the EUR, USD, JPY.
Sharp increases in industry-wide production
capacity in memory and TFT-LCD, leading to
faster-than-expected ASP declines.
Price erosion on key products such as
handsets on intensifying competition.
Rebounding Japanese competition following
restructuring efforts, complete with new
generation products such as 3DTV.
Protracted delay in global economic recovery
dampening demand.
Resolution of any outstanding IP and
regulatory disputes.
Fundamentals
Operating performance outlook
Telecoms and digital media supported Samsung’s
consolidated profits last year; this year will be
memory’s turn. Semiconductors was c20% of
Samsung’s consolidated revenue last year; this
proportion is set to rise to c24% by 2011e.
85

TMT
Memory Devices Technology
8 October 2010
abc
Samsung consolidated revenue breakdown
Semiconductors revenue, OP (KRWbn) and OPM (%)
100%
80%
60%
40%
20%
0%
2007 2008 2009e 2010f 2011f
50,000
40,000
30,000
20,000
10,000
-
2007 2008 2009e 2010f 2011f
30%
25%
20%
15%
10%
5%
0%
Semiconductor LCD Telecom
Digital Media Others
Source: Company, HSBC estimates
Source: Company, HSBC estimates
Revenue OP OPM
Memory is to dramatically impact profitability;
we expect semiconductors to surge from 19% of
consolidated operating profits in 2009 to a
forecast 56% this year.
Memory as % of consolidated operating profit
The key driver of this growth is memory, which
accounts for c70% of Semiconductor revenue.
Consolidated revenue of Semiconductor division, 2009
HDD
11%
100%
50%
Sy stems
LSI
15%
DRAM
40%
0%
2007 2008 2009e 2010f 2011f
Semiconductor LCD Telecom
Digital Media Others
Other
memory
5%
Source: HSBC estimates
NAND
29%
Source: Company, HSBC estimates
We forecast Samsung’s consolidated semicon
revenue to rise c40% y-o-y this year, and for
operating profit to rise c411% y-o-y.
Detailed outlook
We expect Samsung to report a 3Q10 operating
profit similar to 2Q10. Revenue should rise 8% q-
o-q to KRW41.1trn, as Telecoms and Digital
Media profitability rebound. TFT LCD margins
are set to contract on softer ASPs, whilst Samsung
should deliver similar Semicon margins on
process migration despite weaker ASPs. 2Q10
was anomalous in that Semicon enjoyed demand
from PC OEMs that brought forward shipments
into 2Q (hence weak 3Q q-o-q growth) and firm
pricing on weaker supply from smaller players.
For Telecoms, strong smartphone sales during
3Q10 drive a margin rebound. We expect
86

TMT
Memory Devices Technology
8 October 2010
abc
relatively flat (-1% q-o-q) operating profit at
KRW4.9trn, lower than consensus of KRW5.2trn.
With decent 3Q10 expectations ‘in the bag’ as far
as the market is concerned, the focus is very much
on 4Q10-1Q11. If the traditionally strong 3Q can
only deliver flat earnings q-o-q, and with ASPs
still falling for the major earnings contributors –
memory and TFT LCD – how low can Samsung’s
4Q numbers go? Recent share price performance
suggests that the market is expecting the worst.
Our expectations are slightly more optimistic on
the quarter, with 6% higher operating profit
estimates than consensus. The difference is
probably attributable to our more conservative
outlook on TFT LCD and Digital Media. We
expect operating profit to fall 10% q-o-q,
compared to 20% for consensus. We concede there
is no escape from falling memory ASPs even with
accelerated process migration; we expect Semicon
profits to fall 14% q-o-q. TFT-LCD declines are
slightly lower at 12% q-o-q on cautious panel
procurement. We are also taking a conservative
view on memory ASPs in 1Q11, expecting DDR3
contract prices to fall another 18%.
DRAM
Revenue, margin outlook
For this year we forecast consolidated DRAM
revenue to rise 82% in KRW terms, on a 70% bit
growth and ASP increase of 18%. We expect the
KRW to strengthen 12% compared to last year.
DRAM revenue (KRWbn), operating margins
20,000
15,000
10,000
5,000
-
2007 2008 2009 2010f 2011f 2012f
Source: Company, HSBC estimates
Revenue (KRW bn)
OPM
50.0%
40.0%
30.0%
20.0%
10.0%
0.0%
-10.0%
-20.0%
We forecast 2011e consolidated DRAM revenue
in KRW terms to decline 3% y-o-y; bit growth is
to remain strong at 58% y-o-y but is offset by a
31% blended ASP fall on greater industry-wide
supply and our assumption of another 6%
strengthening in the KRW.
Key themes for Samsung’s business divisions
Division Conclusions, positive catalysts % 2011e OP 2009/12e OP CAGR
Semicon
DRAM
NAND
Non-memory
Telecoms
LCD
Digital Media
Source: HSBC
We expect Samsung’s process migration will support margins, but it also enables it to offer advanced low-power, high
performance devices for rapid growth areas of mobile devices and servers. Samsung has great exposure to the NAND
structural growth story and increasing exposure to mobile processor systems.
We expect DRAM ASP declines to slow in 2Q11 as price elasticity kicks in to drive content/box. Impact of tablets on global
DRAM demand c5% for 2011-12e. Samsung benefits from high-end product mix and process migration.
Strong price elasticity drives greater penetration for smartphones, tablets and SSDs. Technology developments in NAND
such as multi-level cells and toggle DDR, together with rising affordability will drive continued NAND take up.
Samsung is making a strong push into system-on-chip (SoC) for mobile devices, as a high-growth, complementary
business for memory and smartphone operations.
Samsung’s tablet and leverages Samsung’s vertical integration and wireless distribution channels. Samsung is very well
positioned to benefit from supplying 4G handsets going forward.
Industry-wide ASPs could start rebounding in early 2011. We expect Samsung’s high-end product mix to dampen the
impact of ASP declines and forecast Samsung to remain profitable through this ASP downturn.
We are positive on Samsung’s early mover strategy in bringing advanced flat TVs to developed markets. We also believe
other promising areas to watch include digital imaging, PC and printer operations.
46% -11%
28% 20%
14% 4%
12% 19%
87

TMT
Memory Devices Technology
8 October 2010
abc
Bit growth, ASP assumptions (%)
100%
50%
0%
-50%
-100%
2008 2009 2010f 2011f 2012f
Bit growth
ASP
Source: Company, HSBC estimates
We expect greater cost efficiencies mostly by
process migration and improved production
techniques to expand operating profitability to
42.7% from 12.7% last year, reducing to 30.6% in
2011e.
Samsung has a strong product portfolio; as at
2Q10, by revenue contribution to DRAM, we
estimate c20% each in mDRAM and server
DRAM, c10% in graphics DRAM and c5% in
specialty DRAM. Samsung expects 150% y-o-y
mDRAM bit growth this year as smartphone
content/system rises to 4Gb this year (2009: 2Gb)
and reach c6-8Gb next year.
Technology – process migration
Samsung’s DRAM process migration
100%
80%
60%
40%
20%
0%
2007 2008 2009 2010f 2011f 2012f
8xnm 6xnm 5xnm 4xnm 3xnm 2xnm
Samsung is leading the market in DRAM process
migration. By end-4Q09, it had migrated close to
50% of DRAM output to 56nm and below. By the
end of this year, it aims to achieve more than 80%
of output on 56nm and below. By then, more than
60% of the total is to be on the even more costeffective
46nm process and below; in 2Q09, c35%
of production was on the 46nm process.
Wafer capacity (kwpy), average fab utilisation
8,000
120%
6,000
100%
80%
4,000
60%
2,000
40%
20%
-
0%
2007 2008 2009 2010f 2011f 2012f
Wafer capacity (kwpy) Average utilisation (%)
Source: Company, HSBC estimates
Samsung started mass production on the 35nm
process in July this year. Additional capex cost of
migrating to 35nm – for which mass production is
slated for 2H10 – should not be high, as immersion
lithography equipment remains the same.
NAND
Revenue, margin outlook
We forecast consolidated 2010e NAND revenue to
rise 27% in KRW terms, on a 79% bit growth and
blended ASP decline of 21%. We expect the KRW
to strengthen 12% compared to last year. We
forecast 2011e consolidated revenue growth to slow
to 18%; bit growth is to remain strong at 102% y-o-y
but is offset by a 37% ASP fall and our assumption
of another 6% strengthening in the KRW.
Source: Company, HSBC estimates
88

TMT
Memory Devices Technology
8 October 2010
abc
NAND revenue growth
12,000
30.0%
10,000
25.0%
8,000
20.0%
6,000
15.0%
4,000
10.0%
2,000
5.0%
-
0.0%
2007 2008 2009 2010f 2011f 2012f
Revenue (KRW bn) OPM
Source: Company, HSBC estimates
Technology – process migration
By end-4Q09, more than 50% of Samsung’s
NAND output was at 42nm and below. By 2Q10,
about 50% of Samsung’s output was at 32nm. It
aims to achieve over c80% of output at 32nm and
below by end of this year, some of which would
be using 3 bits per cell. Samsung started mass
production of 2xnm 32Gb in April, to be joined
by 64Gb 3-bit products in 2H10.
Process migration
100%
80%
60%
40%
20%
0%
2007 2008 2009 2010f 2011f 2012f
8xnm 6xnm 5xnm 4xnm 3xnm 2xnm
Source: Company, HSBC estimates
Production capacity, capex
In May 2010 Samsung announced a massive
upward revision in capex guidance to KRW18trn,
beating the previous capex high of cKRW13.5trn.
in 2008. The amount includes KRW11trn for
semiconductors (memory KRW9trn, systems LSI
KRW2trn) plus another KRW5trn for LCD. This
compares to KRW5.5trn for semis, KRW3trn for
LCD in the firm’s original guidance. Capex is
focused on capacity expansion in semiconductor
Fabs 15 and 16 and its 8G fab for LCD. Samsung
plans to spend another KRW8trn on R&D this year.
In semiconductors, we expect wafer capacity,
though a limited amount this year. Samsung
earmarked KRW2.5trn for expansion of Fab 15 by
another c30-40kwpm wafer capacity for 30nm
DRAM. Overall, we estimate c40-50kwpm
additional DRAM wafer capacity could be added
by end-2010.
The bulk of memory capex this year relates to
process migration, including double patterning and
other related equipment. If Samsung wants to
accelerate process migration to 3xnm for DRAM
and 2xnm for NAND to enlarge the product costcompetitiveness
gap, ‘double patterning’ is required,
which incurs more costs due to lower throughput.
Wafer capacity increase will mostly come starting
next year, as Samsung plans to spend a total of
KRW12trn on Fab 16; whether this fab will
produce DRAM or NAND has not been decided
yet, but we believe NAND is likely. Fab 16 would
come online next year and ultimately add
c200kwpm capacity.
89

TMT
Memory Devices Technology
8 October 2010
abc
Samsung semicon capex (KRWbn)
20,000
15,000
10,000
5,000
-
2008 2009 2010f 2011f 2012f
Source: Company, HSBC estimates
An area worth watching going forward is the
systems LSI business, which has great exposure to
mobile application processors for smartphones
and increasingly large exposure to image sensors.
Samsung currently provides some foundry
services, including to Apple and Qualcomm. We
see momentum building in the race to deliver
faster processing in mobile devices, where
Samsung already has a significant presence.
Samsung is investing KRW4.5trn to expand its
Austin fab capacity by 40kwpm by end-2011e on
the 45nm process to produce Systems LSI. We
expect most of this capacity would service
foundry operations for mobile-related devices.
90

TMT
Memory Devices Technology
8 October 2010
abc
Samsung Electronics: Consolidated P&L summary (K-IFRS)
Revenue (KRWbn) 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e
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
Memory 16,770 26,041 27,173 28,796 5,585 6,710 6,937 6,809 6,207 6,539 7,125 7,302
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
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
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
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
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
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
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
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
Revenue, % of total 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e
Semicon 20% 24% 24% 25% 24% 25% 24% 23% 24% 24% 24% 24%
Memory 12% 17% 16% 16% 16% 18% 17% 16% 16% 16% 16% 16%
LSI 3% 3% 4% 4% 3% 3% 3% 3% 4% 4% 4% 4%
TFT-LCD 19% 19% 18% 17% 20% 20% 18% 16% 18% 18% 18% 18%
Telecom 28% 25% 25% 24% 27% 23% 25% 25% 25% 24% 25% 24%
Mobile 25% 22% 20% 18% 25% 21% 22% 22% 21% 21% 20% 19%
Digital media 38% 39% 39% 39% 36% 38% 39% 40% 39% 39% 38% 40%
VD 24% 24% 24% 24% 21% 23% 24% 26% 24% 23% 24% 25%
Appliances 7% 8% 7% 7% 7% 8% 8% 7% 8% 8% 7% 7%
Total 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100%
Revenue, change (%) 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e
Semicon 20% 40% 8% 9% 2% 16% 4% 0% -5% 5% 7% 3%
Memory 20% 55% 4% 6% 11% 20% 3% -2% -9% 5% 9% 2%
LSI 5% 15% 22% 22% na 6% 5% 5% 5% 5% 5% 5%
TFT-LCD 20% 12% 4% 4% -6% 13% -6% -3% -1% 6% 4% 4%
Telecom 18% 4% 7% 2% -10% -4% 16% 7% -10% 2% 9% 3%
Mobile 20% 1% -1% -6% -8% -6% 13% 2% -9% -1% 5% 1%
Digital media 15% 18% 10% 6% -19% 15% 10% 8% -11% 3% 5% 9%
VD 13% 13% 10% 5% -28% 16% 16% 12% -17% 3% 9% 12%
Appliances 13% 23% 6% 2% -3% 28% -2% 1% -2% 3% -2% 3%
Total 12% 15% 8% 5% -12% 9% 8% 4% -8% 4% 7% 5%
OP (KRWbn) 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e
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
TFT-LCD 1,707 2,367 2,484 2,572 491 878 529 468 568 584 680 652
Telecom 4,091 3,603 5,013 5,229 1,104 626 948 925 1,060 1,273 1,519 1,161
Digital media 3,060 1,743 2,247 2,476 519 361 431 432 462 555 669 561
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
OPM (%) 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e
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%
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%
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%
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%
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%
OP, % of total 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e
Semicon 19% 56% 46% 45% 45% 58% 61% 59% 46% 46% 44% 49%
TFT-LCD 16% 13% 14% 14% 11% 18% 11% 11% 15% 13% 13% 14%
Telecom 37% 19% 28% 28% 25% 12% 19% 21% 27% 29% 29% 25%
Digital media 28% 9% 12% 13% 12% 7% 9% 10% 12% 13% 13% 12%
Total 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100%
OP, change (%) 2009 2010e 2011e 2012e 1Q10 2Q10 3Q10e 4Q10e 1Q11e 2Q11e 3Q11e 4Q11e
Semicon na 411% -21% -1% 50% 46% 4% -14% -31% 13% 13% -2%
TFT-LCD na 39% 5% 4% -8% 79% -40% -12% 21% 3% 17% -4%
Telecom na -12% 39% 4% 6% -43% 51% -2% 15% 20% 19% -24%
Digital media na -43% 29% 10% 11% -30% 20% 0% 7% 20% 20% -16%
Total na 72% -4% 3% 28% 14% -1% -10% -13% 14% 16% -10%
Source: Company data, HSBC estimates
91

TMT
Memory Devices Technology
8 October 2010
abc
Financials & valuation: Samsung Electronics
Overweight
Financial statements
Year to 12/2009a 12/2010e 12/2011e 12/2012e
Profit & loss summary (KRWb)
Revenue 136,324 156,476 169,642 178,822
EBITDA 22,085 31,263 32,792 34,066
Depreciation & amortisation -11,159 -12,478 -14,720 -15,530
Operating profit/EBIT 10,925 18,785 18,072 18,536
Net interest -205 -323 -460 -528
PBT 12,192 20,413 20,103 19,845
HSBC PBT 12,192 20,413 20,103 19,845
Taxation -2,431 -3,992 -3,923 -3,069
Net profit 9,761 16,420 16,180 16,777
HSBC net profit 9,761 16,420 16,180 16,777
Cash flow summary (KRWb)
Cash flow from operations 20,352 21,212 28,279 31,238
Capex -8,520 -19,131 -22,008 -19,214
Cash flow from investment -10,555 -24,199 -27,947 -33,018
Dividends -871 -947 -1,277 -2,043
Change in net debt -10,104 4,264 1,710 3,823
FCF equity 12,276 2,850 6,151 11,902
Balance sheet summary (KRWb)
Intangible fixed assets 891 941 991 1,042
Tangible fixed assets 39,816 46,419 53,656 57,290
Current assets 62,069 58,683 64,457 74,012
Cash & others 21,570 14,824 18,669 27,069
Total assets 118,281 126,189 142,272 159,090
Operating liabilities 30,816 28,588 30,534 32,618
Gross debt 14,411 9,403 9,403 9,403
Net debt -5,055 -791 919 4,742
Shareholders funds 73,054 88,198 102,336 117,070
Invested capital 50,457 62,631 69,902 72,658
Ratio, growth and per share analysis
Year to 12/2009a 12/2010e 12/2011e 12/2012e
Y-o-y % change
Revenue 12.4 14.8 8.4 5.4
EBITDA 36.9 41.6 4.9 3.9
Operating profit 81.1 71.9 -3.8 2.6
PBT 85.3 67.4 -1.5 -1.3
HSBC EPS 53.6 68.2 -1.5 3.7
Valuation data
Year to 12/2009a 12/2010e 12/2011e 12/2012e
EV/sales 0.8 0.7 0.6 0.6
EV/EBITDA 4.8 3.4 3.3 3.2
EV/IC 2.1 1.7 1.5 1.5
PE* 13.8 8.2 8.3 8.0
P/NAV 1.8 1.5 1.3 1.2
FCF yield (%) 11.1 2.6 5.8 11.4
Dividend yield (%) 1.0 1.5 1.5 1.5
Note: * = Based on HSBC EPS (fully diluted)
Issuer information
Share
price
(KRW) 793,000 Target price (KRW) 1,139,0
00
Potent’l tot rtn (%) 43.6
Reuters (Equity) 005930.KS Bloomberg (Equity) 005930 KS
Market cap (USDm) 103,351 Market cap (KRWb) 116,808
Free float (%) 69 Enterprise value (KRWb) 10704
Country Korea Sector Semiconductors
Analyst Nam Park Contact 852 2996 6591
Price relative
941588
841588
741588
641588
541588
441588
341588
Source: HSBC
2008 2009 2010 2011
Samsung Electronics Rel to KOSPI INDEX
Note: price at close of 06 Oct 2010
941588
841588
741588
641588
541588
441588
341588
Ratios (%)
Revenue/IC (x) 2.6 2.8 2.6 2.5
ROIC 17.1 27.1 22.3 22.4
ROE 14.4 20.4 17.0 15.3
ROA 9.2 13.9 12.5 11.5
EBITDA margin 16.2 20.0 19.3 19.1
Operating profit margin 8.0 12.0 10.7 10.4
EBITDA/net interest (x) 107.9 96.9 71.3 64.5
Net debt/equity -6.9 -0.9 0.9 4.1
Net debt/EBITDA (x) -0.2 0.0 0.0 0.1
CF from operations/net debt 3076.4 658.7
Per share data (KRW)
EPS Rep (fully diluted) 57,370 96,516 95,105 98,610
HSBC EPS (fully diluted) 57,370 96,516 95,105 98,610
DPS 7,500 12,000 12,000 12,000
NAV 429,398 518,410 601,508 688,111
92

TMT
Memory Devices Technology
8 October 2010
abc
Winbond (2344)
Defensive play in a steep commodity price correction environment
with a successfully restructured, diversified niche portfolio
Technology migration and flexible product mix to maintain margin
Cut TP to TWD10.8 (from TWD12.5) to reflect higher macro risk
using target PB of 1x; maintain OW(V)
Investment summary
Winbond’s memory portfolio restructuring story
has ended on a high note after it successfully
reduced its commodity DRAM portion to less
than 2% (from c30%) of its portfolio in 2Q10.
Financial results in 1H10 were good, thanks to
tight supply of NOR on the back of the industry’s
limited capacity. As the demand/supply in NOR
has slowly approached a balance in 3Q, and while
DRAM prices seem to have softened starting from
3Q, we believe Winbond’s superior portfolio is
the most defensive among our Taiwanese DRAM
coverage amid the memory price correction
outlook in 2H10, given the stock’s current
attractive valuation.
Fundamentals
A superior non-PC memory portfolio
Winbond’s 40% revenue is contributed by
specialty DRAM. The remaining contribution
came from NOR (28%), Mobile DRAM (18%),
graphics DRAM (12%) and commodity DRAM.
(2%). A year ago, around 30% of Winbond’s
revenue was contributed by commodity DRAM, a
portfolio that is highly susceptible to DRAM price
volatility hence impacting margin stability.
More first-tier clients
Most of Winbond’s clients are 3-6 month
contract-based. This enables Winbond to have
relatively stable margins while having high entry
barriers as it takes time and effort to win client
relationships. Winbond has demonstrated its
superior quality by established first-tier clients in
areas like specialty DRAM, mobile DRAM and
NOR flash within the last few years.
Flexible and diversified portfolio
Winbond has the flexibility in adjusting its
memory portfolio: as their product is not
concentrated on a single memory type, capacity
reallocation can be efficiently adjusted. More
importantly, its memory production also supplies
various diversified applications: Within specialty
DRAM, HDD represents 36% of the total,
followed by TV (17%) and networking (14%).
This enables Winbond to focus on or divert to
more lucrative applications when demand for a
certain application is affected by seasonality.
Specialty DRAM has strong growth of 27% q-o-q
growth in 2Q, underpinned by top applications
like HDD and TV growth.
93

TMT
Memory Devices Technology
8 October 2010
abc
Technology migration
Winbond has successfully migrated its NOR
production from 130nm to 90nm. In 3Q10, all 10k
NOR capacity are in 90nm. For DRAM,
Winbond’s target is to migrate its 16k capacity to
65nm.
Ongoing migrations: Winbond has scheduled to
further migrate 1) NOR capacity to 58nm by
2H11 and 2) DRAM capacity to 46nm also by
2H11.
ASP and bit growth assumptions
80.0%
60.0%
40.0%
20.0%
0.0%
-20.0%
-40.0%
2008 2009 2010f 2011f 2012f
Bit growth ASP
Source: Company data, HSBC estimates
Forecast revision
3Q10 revenue and margins should still be holding
up well, thanks to the long-term contractual
relationships which Winbond has built up with
customers, together with strong demand from
specialty DRAM and mobile demand, even
though NOR supply/demand has come to a
relatively more balanced situation. Winbond will
likely post a gain of TWD0.8bn from the sale of
its Nuvoton stake on the latter’s listing in late
September. Following the sale, Winbond still
holds 62.5% of Nuvoton.
Revenue growth
40,000
30,000
20,000
10,000
-
2007 2008 2009 2010f 2011f 2012f
Source: Company data, HSBC estimates
Revenue (NTD mn)
OPM
20%
10%
0%
-10%
-20%
-30%
-40%
-50%
We lowered Winbond’s bit growth for 2010 to
12% from 16% and revised down blended ASP
for 2011 as memory price softening has come
earlier than we expected. We now forecast
Winbond to achieve 2010-11e revenue growth of
67% and 4% respectively, revised down by 2%
and 4% respectively. We also lowered Winbond’s
2010-11e OP margin to 10.9% and 7.2%
respectively (previously 12.6% and 9.3%)
respectively on lower top-line growth. We are 2%
and 4% below consensus in terms of 2010-11e
revenue.
Forecast revision for Winbond
__Previous___ __ Revised ___ __ Changes___
TWD (m) 2010e 2011e 2010e 2011e 2010e 2011e
Revenue 33,219 35,336 32,596 33,796 -2% -4%
OP 4,172 3,289 3,561 2,431 -15% -26%
Net profit 4,063 2,738 3,532 2,011 -13% -27%
OPM 12.6% 9.3% 10.9% 7.2%
NPM 12.2% 7.7% 10.8% 6.0%
Source: HSBC
Product mix
We expect Winbond to maintain a relatively
balanced portfolio among NOR flash, specialty
DRAM and mobile DRAM in 2010, after
successfully reduced commodity DRAM’s
proportion. Despite NOR flash has come to a
more balanced demand supply situation, we
believe Winbond’s NOR flash bit growth will
continue, underpinned by expanded capacity
94

TMT
Memory Devices Technology
8 October 2010
abc
(10k/m by 4Q10) and 90nm full production
capacity, offsetting possible NOR ASP decline.
We’ve modelled NOR ASP to come down by
18% in 4Q and another 10% decline in 1Q11. We
expect NOR flash to represent close to 26% of
Winbond’s revenue by end of 4Q10. We expect
Winbond’s mobile DRAM and specialised
DRAM revenue combined to represent 57% of its
total revenue in 2011, up from 55% this year.
Revenue mix, 2007-11e
100%
80%
60%
40%
20%
0%
2007 2008 2009 2010f 2011f
NOR Flash
Specialty DRAM
Mobile DRAM
Graphic DRAM
Commodity DRAM Others
Source: Company data, HSBC estimates
Revenue growth in 2011 seems minimal, but this
is after a very strong growth year in 2010. We
forecast 2010-11 ASP growth of 53% and -30%
respectively, and 2010-11e bit growth for 12%
and 45%. We expect bit growth to be largely
driven by process migration.
Valuation, ratings
We derive a new 12-month target price of
TWD10.8 (previously TWD12.5) for Winbond by
applying a target FY11e PB multiple of 1.0x
(rolled over to FY11 based on 9.9% ROE and
9.5% COE). Previously, we use a multiple of
1.2x). The lower PB multiple is supported by our
lower forecast in OP and lower ROE, given
Winbond’s NOR growth has come to a
normalized demand and supply situation, in
addition to a softening memory price outlook.
(prior to the capacity glut starting in 2007). With a
less optimistic macro environment, we expect the
stock to reach the lower of its historical PB band
in the next 12 months.
HSBC vs consensus
HSBC 2010e 2011e 2012e
Revenue 32,596 33,796 37,642
EBITDA 14,594 13,729 15,477
EBIT 3,561 2,431 3,662
Net income 3,532 2,011 3,201
EPS (KRW) 1.0 0.6 0.9
EBITDA margin 44.8% 40.6% 41.1%
EBIT margin 10.9% 7.2% 9.7%
Net margin 10.8% 6.0% 8.5%
HSBC vs
2010e 2011e 2012e
consensus
Revenue -2.0% -4.0% -0.6%
EBITDA -5.0% -11.5% 9.8%
EBIT -14.6% -26.1% 1.4%
Net income -15.2% -53.1% -12.3%
EPS (KRW) 9.7% -54.7% 7.6%
Consensus 2010e 2011e 2012e
Revenue 33,255 35,209 37,867
EBITDA 15,365 15,518 14,097
EBIT 4,172 3,289 3,610
Net income 4,165 4,292 3,650
EPS (KRW) 0.9 1.2 0.8
EBITDA margin 46.2% 44.1% 37.2%
EBIT margin 12.5% 9.3% 9.5%
Net margin 12.5% 12.2% 9.6%
Source: HSBC, I/B/E/S consensus
Under our research model, for Taiwan stocks with
a volatility indicator, the Neutral band is 10ppt
above and below the hurdle rate of 9.5% (risk-free
rate of 4% plus risk premium of 5.5%). This
translates into a Neutral band of -0.5% to +19.5%
around the current share price. Our 12-month
target price of TWD10.8 suggests a potential
return of 34.5%, which is in the Overweight band;
thus, we maintain our Overweight (V) rating.
The stock has historically traded at a PB of 0.2-
1.2x, with a high ROE of 2-11% when profitable
95

TMT
Memory Devices Technology
8 October 2010
abc
Quarterly assumptions for Winbond
TWDm 1Q10 2Q10 3Q10e 4Q10e 2010e 2011e
Revenue 7,013 8,531 8,772 8,280 32,596 33,796
OP 484 1,290 987 800 3,561 2,431
Net profit 384 1,236 1,361 551 3,532 2,011
OPM 6.9% 15.1% 11.3% 9.7% 10.9% 7.2%
NPM 5.5% 14.5% 15.5% 6.7% 10.8% 6.0%
Blended ASP 16% 10% -6% -14% 53% -30%
Bit growth (%) -6% 11% 9% 9% 12% 45%
Source: Company data, HSBC estimates
Forward PB band
20
Price (TWD)
15
1.2x
10
0.9x
5
0.6x
0.3x
-
Jan-
05
Jan-
06
Jan-
07
Jan-
08
Jan-
09
Jan-
10
Source: HSBC
Catalysts and risks
Memory industry-specific catalysts:
Evidence of strong unit shipments and rising
memory content of smartphones, LCD TV,
HDD, PC
Faster than expected ASP erosion due to
greater capacity coming online and/or slower
than expected demand from applications like
TV, HDD, LCD TV
More consolidation of players, such as
Micron’s acquisition of Numonyx,
strengthening competition.
Greater ability of competitors to accelerate
process migration, wafer capacity expansion
through improving financials.
Delayed global economic recovery.
Company-specific risks include:
Delay in process migration.
Delay in graphic DRAM execution
Evidence of rising memory content of NOR
flash required in integrated chip platform in
PC
Company-specific catalysts include:
Technology migration to reduce cost from 1)
moving NOR flash production to 58nm by
2H11, 2) 65nm DRAM capacity to 16kwpm;
Capacity ramp up to 10k from 7k for NOR
Further DRAM migration to 46nm in 2H11.
Memory industry-specific risks:
96

TMT
Memory Devices Technology
8 October 2010
abc
Financials & valuation: Winbond Electronics
Overweight (V)
Financial statements
Year to 12/2009a 12/2010e 12/2011e 12/2012e
Profit & loss summary (TWDm)
Revenue 19,533 32,596 33,796 37,642
EBITDA 4,246 14,594 13,729 15,477
Depreciation & amortisation -10,241 -11,033 -11,297 -11,815
Operating profit/EBIT -5,995 3,561 2,431 3,662
Net interest -640 -629 -390 -258
PBT -8,613 3,532 2,366 3,904
HSBC PBT -8,613 3,532 2,366 3,904
Taxation 0 0 -355 -703
Net profit -8,613 3,532 2,011 3,201
HSBC net profit -8,613 3,532 2,011 3,201
Cash flow summary (TWDm)
Cash flow from operations 4,768 11,837 13,143 14,593
Capex -4,200 -8,000 -7,500 -7,500
Cash flow from investment -15,535 -746 -7,540 -7,540
Dividends 0 0 0 0
Change in net debt -2,453 -7,744 -5,573 -6,850
FCF equity 568 3,837 5,288 6,390
Balance sheet summary (TWDm)
Intangible fixed assets 1,531 1,106 813 517
Tangible fixed assets 45,866 43,297 39,833 35,854
Current assets 12,834 29,317 35,329 43,584
Cash & others 2,859 14,570 20,143 26,994
Total assets 65,163 78,653 80,907 84,887
Operating liabilities 7,250 13,223 13,466 14,244
Gross debt 24,058 28,025 28,025 28,025
Net debt 21,199 13,455 7,882 1,032
Shareholders funds 33,854 37,405 39,416 42,617
Invested capital 50,121 45,928 42,366 38,717
Ratio, growth and per share analysis
Year to 12/2009a 12/2010e 12/2011e 12/2012e
Y-o-y % change
Revenue -10.5 66.9 3.7 11.4
EBITDA 36.0 243.7 -5.9 12.7
Operating profit -31.7 50.6
PBT -33.0 65.0
HSBC EPS -43.1 59.2
Ratios (%)
Revenue/IC (x) 0.4 0.7 0.8 0.9
ROIC -10.7 8.4 5.3 8.1
ROE -23.9 9.9 5.2 7.8
ROA -11.5 5.8 3.3 4.5
EBITDA margin 21.7 44.8 40.6 41.1
Operating profit margin -30.7 10.9 7.2 9.7
EBITDA/net interest (x) 6.6 23.2 35.2 60.1
Net debt/equity 62.6 36.0 20.0 2.4
Net debt/EBITDA (x) 5.0 0.9 0.6 0.1
CF from operations/net debt 22.5 88.0 166.7 1414.7
Per share data (TWD)
EPS Rep (fully diluted) -2.36 0.97 0.55 0.88
HSBC EPS (fully diluted) -2.36 0.97 0.55 0.88
DPS 0.00 0.00 0.00 0.00
Book value 9.28 10.25 10.80 11.68
Valuation data
Year to 12/2009a 12/2010e 12/2011e 12/2012e
EV/sales 2.3 1.2 1.0 0.7
EV/EBITDA 10.7 2.6 2.4 1.6
EV/IC 0.9 0.8 0.8 0.7
PE* 8.3 14.6 9.2
P/Book value 0.9 0.8 0.7 0.7
FCF yield (%) 2.3 15.7 21.6 26.2
Dividend yield (%) 0.0 0.0 0.0 0.0
Note: * = Based on HSBC EPS (fully diluted)
Issuer information
Share price (TWD) 8.03 Target price (TWD) 10.80 Potent’l return (%) 34.5
Reuters (Equity) 2344.TW Bloomberg (Equity) 2344 TT
Market cap (USDm) 945 Market cap (TWDm) 29,362
Free float (%) 74 Enterprise value (TWDm 37885
)
Country Taiwan Sector Semiconductors
Analyst Carolyn Poon Contact +852 2996 6586
Price relative
13
11
9
7
5
3
1
2008 2009 2010 2011
Winbond Electronics Rel to TAIWAN WEIGHTED INDEX
Source: HSBC
Note: price at close of 06 Oct 2010
13
11
9
7
5
3
1
97

TMT
Memory Devices Technology
8 October 2010
abc
Notes
98

TMT
Memory Devices Technology
8 October 2010
abc
Disclosure appendix
Analyst Certification
The following analyst(s), economist(s), and/or strategist(s) who is(are) primarily responsible for this report, certifies(y) that the
opinion(s) on the subject security(ies) or issuer(s) and/or any other views or forecasts expressed herein accurately reflect their
personal view(s) and that no part of their compensation was, is or will be directly or indirectly related to the specific
recommendation(s) or views contained in this research report: Nam Park and Carolyn Poon
Important disclosures
Stock ratings and basis for financial analysis
HSBC believes that investors utilise various disciplines and investment horizons when making investment decisions, which
depend largely on individual circumstances such as the investor's existing holdings, risk tolerance and other considerations.
Given these differences, HSBC has two principal aims in its equity research: 1) to identify long-term investment opportunities
based on particular themes or ideas that may affect the future earnings or cash flows of companies on a 12 month time horizon;
and 2) from time to time to identify short-term investment opportunities that are derived from fundamental, quantitative,
technical or event-driven techniques on a 0-3 month time horizon and which may differ from our long-term investment rating.
HSBC has assigned ratings for its long-term investment opportunities as described below.
This report addresses only the long-term investment opportunities of the companies referred to in the report. As and when
HSBC publishes a short-term trading idea the stocks to which these relate are identified on the website at
www.hsbcnet.com/research. Details of these short-term investment opportunities can be found under the Reports section of this
website.
HSBC believes an investor's decision to buy or sell a stock should depend on individual circumstances such as the investor's
existing holdings and other considerations. Different securities firms use a variety of ratings terms as well as different rating
systems to describe their recommendations. Investors should carefully read the definitions of the ratings used in each research
report. In addition, because research reports contain more complete information concerning the analysts' views, investors
should carefully read the entire research report and should not infer its contents from the rating. In any case, ratings should not
be used or relied on in isolation as investment advice.
Rating definitions for long-term investment opportunities
Stock ratings
HSBC assigns ratings to its stocks in this sector on the following basis:
For each stock we set a required rate of return calculated from the risk free rate for that stock's domestic, or as appropriate,
regional market and the relevant equity risk premium established by our strategy team. The price target for a stock represents
the value the analyst expects the stock to reach over our performance horizon. The performance horizon is 12 months. For a
stock to be classified as Overweight, the implied return must exceed the required return by at least 5 percentage points over the
next 12 months (or 10 percentage points for a stock classified as Volatile*). For a stock to be classified as Underweight, the
stock must be expected to underperform its required return by at least 5 percentage points over the next 12 months (or 10
percentage points for a stock classified as Volatile*). Stocks between these bands are classified as Neutral.
Our ratings are re-calibrated against these bands at the time of any 'material change' (initiation of coverage, change of volatility
status or change in price target). Notwithstanding this, and although ratings are subject to ongoing management review,
expected returns will be permitted to move outside the bands as a result of normal share price fluctuations without necessarily
triggering a rating change.
*A stock will be classified as volatile if its historical volatility has exceeded 40%, if the stock has been listed for less than 12
months (unless it is in an industry or sector where volatility is low) or if the analyst expects significant volatility. However,
99

TMT
Memory Devices Technology
8 October 2010
abc
stocks which we do not consider volatile may in fact also behave in such a way. Historical volatility is defined as the past
month's average of the daily 365-day moving average volatilities. In order to avoid misleadingly frequent changes in rating,
however, volatility has to move 2.5 percentage points past the 40% benchmark in either direction for a stock's status to change.
Rating distribution for long-term investment opportunities
As of 08 October 2010, the distribution of all ratings published is as follows:
Overweight (Buy) 50% (21% of these provided with Investment Banking Services)
Neutral (Hold) 36% (18% of these provided with Investment Banking Services)
Underweight (Sell) 14% (19% of these provided with Investment Banking Services)
Information regarding company share price performance and history of HSBC ratings and price targets in respect of its longterm
investment opportunities for the companies the subject of this report,is available from www.hsbcnet.com/research.
HSBC & Analyst disclosures
Disclosure checklist
Company Ticker Recent price Price Date Disclosure
HYNIX SEMICONDUCTOR INC 000660.KS 23200.00 07-Oct-2010 11
INOTERA MEMORIES INC. 3474.TW 16.45 07-Oct-2010 1, 2, 5
SAMSUNG ELECTRONICS 005930.KS 770000.00 07-Oct-2010 6, 11
Source: HSBC
1 HSBC* has managed or co-managed a public offering of securities for this company within the past 12 months.
2 HSBC expects to receive or intends to seek compensation for investment banking services from this company in the next
3 months.
3 At the time of publication of this report, HSBC Securities (USA) Inc. is a Market Maker in securities issued by this
company.
4 As of 31 August 2010 HSBC beneficially owned 1% or more of a class of common equity securities of this company.
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
and/or paid compensation to HSBC in respect of investment banking services.
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
and/or paid compensation to HSBC in respect of non-investment banking-securities related services.
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
and/or paid compensation to HSBC in respect of non-securities services.
8 A covering analyst/s has received compensation from this company in the past 12 months.
9 A covering analyst/s or a member of his/her household has a financial interest in the securities of this company, as
detailed below.
10 A covering analyst/s or a member of his/her household is an officer, director or supervisory board member of this
company, as detailed below.
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
securities in respect of this company
Analysts, economists, and strategists are paid in part by reference to the profitability of HSBC which includes investment
banking revenues.
For disclosures in respect of any company mentioned in this report, please see the most recently published report on that
company available at www.hsbcnet.com/research.
* HSBC Legal Entities are listed in the Disclaimer below.
100

TMT
Memory Devices Technology
8 October 2010
abc
Additional disclosures
1 This report is dated as at 08 October 2010.
2 All market data included in this report are dated as at close 06 October 2010, unless otherwise indicated in the report.
3 HSBC has procedures in place to identify and manage any potential conflicts of interest that arise in connection with its
Research business. HSBC's analysts and its other staff who are involved in the preparation and dissemination of Research
operate and have a management reporting line independent of HSBC's Investment Banking business. Information Barrier
procedures are in place between the Investment Banking and Research businesses to ensure that any confidential and/or
price sensitive information is handled in an appropriate manner.
101

TMT
Memory Devices Technology
8 October 2010
abc
Disclaimer
* Legal entities as at 31 January 2010
'UAE' HSBC Bank Middle East Limited, Dubai; 'HK' The Hongkong and Shanghai Banking
Corporation Limited, Hong Kong; 'TW' HSBC Securities (Taiwan) Corporation Limited; 'CA'
HSBC Securities (Canada) Inc, Toronto; HSBC Bank, Paris branch; HSBC France; 'DE' HSBC
Trinkaus & Burkhardt AG, Dusseldorf; 000 HSBC Bank (RR), Moscow; 'IN' HSBC Securities
and Capital Markets (India) Private Limited, Mumbai; 'JP' HSBC Securities (Japan) Limited,
Tokyo; 'EG' HSBC Securities Egypt S.A.E., Cairo; 'CN' HSBC Investment Bank Asia Limited,
Beijing Representative Office; The Hongkong and Shanghai Banking Corporation Limited,
Singapore branch; The Hongkong and Shanghai Banking Corporation Limited, Seoul Securities
Branch; The Hongkong and Shanghai Banking Corporation Limited, Seoul Branch; HSBC
Securities (South Africa) (Pty) Ltd, Johannesburg; 'GR' HSBC Pantelakis Securities S.A., Athens;
HSBC Bank plc, London, Madrid, Milan, Stockholm, Tel Aviv, 'US' HSBC Securities (USA) Inc,
New York; HSBC Yatirim Menkul Degerler A.S., Istanbul; HSBC México, S.A., Institución de
Banca Múltiple, Grupo Financiero HSBC, HSBC Bank Brasil S.A. - Banco Múltiplo, HSBC Bank
Australia Limited, HSBC Bank Argentina S.A., HSBC Saudi Arabia Limited.
Issuer of report
The Hongkong and Shanghai
Banking Corporation Limited
Level 19, 1 Queen’s Road Central
Hong Kong SAR
Telephone: +852 2843 9111
Telex: 75100 CAPEL HX
Fax: +852 2596 0200
Website: www.research.hsbc.com
This document has been issued by The Hongkong and Shanghai Banking Corporation Limited (“HSBC”) in the conduct of its Hong Kong
regulated business for the information of its institutional and professional customers; it is not intended for and should not be distributed to retail
customers in Hong Kong. The Hongkong and Shanghai Banking Corporation Limited is regulated by the Securities and Futures Commission. All
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
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
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
document on information obtained from sources it believes to be reliable but which it has not independently verified; HSBC makes no guarantee,
representation or warranty and accepts no responsibility or liability as to its accuracy or completeness. Expressions of opinion are those of the
Research Division of HSBC only and are subject to change without notice. HSBC and its affiliates and/or their officers, directors and employees
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
any such securities (or investment). HSBC and its affiliates may act as market maker or have assumed an underwriting commitment in the
securities of companies discussed in this document (or in related investments), may sell them to or buy them from customers on a principal basis
and may also perform or seek to perform investment banking or underwriting services for or relating to those companies.
HSBC Securities (USA) Inc. accepts responsibility for the content of this research report prepared by its non-US foreign affiliate. All U.S.
persons receiving and/or accessing this report and wishing to effect transactions in any security discussed herein should do so with HSBC
Securities (USA) Inc. in the United States and not with its non-US foreign affiliate, the issuer of this report.
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
(Financial Promotion) Order 2001. The protections afforded by the UK regulatory regime are available only to those dealing with a
representative of HSBC Bank plc in the UK. In Singapore, this publication is distributed by The Hongkong and Shanghai Banking
Corporation Limited, Singapore Branch for the general information of institutional investors or other persons specified in Sections 274 and
304 of the Securities and Futures Act (Chapter 289) (“SFA”) and accredited investors and other persons in accordance with the conditions
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
whole or in part for any purpose. The Hongkong and Shanghai Banking Corporation Limited Singapore Branch is regulated by the Monetary
Authority of Singapore. Recipients in Singapore should contact a "Hongkong and Shanghai Banking Corporation Limited, Singapore Branch"
representative in respect of any matters arising from, or in connection with this report. In Australia, this publication has been distributed by
The Hongkong and Shanghai Banking Corporation Limited (ABN 65 117 925 970, AFSL 301737) for the general information of its
“wholesale” customers (as defined in the Corporations Act 2001). Where distributed to retail customers, this research is distributed by HSBC
Bank Australia Limited (AFSL No. 232595). These respective entities make no representations that the products or services mentioned in this
document are available to persons in Australia or are necessarily suitable for any particular person or appropriate in accordance with local
law. No consideration has been given to the particular investment objectives, financial situation or particular needs of any recipient.
In Japan, this publication has been distributed by HSBC Securities (Japan) Limited. It may not be further distributed in whole or in part for
any purpose. In Korea, this publication is distributed by The Hongkong and Shanghai Banking Corporation Limited, Seoul Securities Branch
("HBAP SLS") for the general information of professional investors specified in Article 9 of the Financial Investment Services and Capital
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
for any purpose. HBAP SLS is regulated by the Financial Services Commission and the Financial Supervisory Service of Korea.
© Copyright. The Hongkong and Shanghai Banking Corporation Limited 2010, ALL RIGHTS RESERVED. No part of this publication may
be reproduced, stored in a retrieval system, or transmitted, on any form or by any means, electronic, mechanical, photocopying, recording, or
otherwise, without the prior written permission of The Hongkong and Shanghai Banking Corporation Limited. MICA (P) 142/06/2010 and
MICA (P) 193/04/2010
102

Nam Park*
Analyst
The Hongkong and Shanghai Banking Corporation Limited (HK)
+852 2996 6591
nampark@hsbc.com.hk
Nam was appointed Head of Memory Devices Technology Research, Asia Pacific, in October 2008. Since joining HSBC in 2000, his roles
have included Regional Head of Telecoms and Internet Research, Head of Korea Small & Mid-cap Research and Regional Head of
Healthcare Research. Nam has extensive equity research experience in Asia since 1996. He has covered most sectors, including
conglomerates, autos, media, alternative energy and defence, as well as writing Korea strategy. Nam holds a Ph.D. in Biophysics from
Imperial College of Science, Technology and Medicine.
Carolyn Poon*
Analyst
The Hongkong and Shanghai Banking Corporation Limited (HK)
+852 2996 6586
carolynpoon@hsbc.com.hk
Carolyn covers memory devices technology for Asia Pacific. She joined HSBC as part of the Asian Telecoms team, and subsequently
was an analyst in the Korean Mid-caps and Asian Healthcare teams. Before joining HSBC, she worked as an economist covering
Greater China.
* Employed by a non-US affiliate of HSBC Securities (USA) Inc, and is not registered/qualified pursuant to FINRA regulations.