GS US Semi & SPE Asia Tour Day One -
Handset demand remains great but some
signs of choppiness in PCs and LCD-TVs;
too early to draw definitive conclusions
On Day 1 of our Asia trip, handset demand seems extremely robust but several of the
companies closer to the PC and LCD-TV end-market highlighted choppy conditions in
Aug./Sept., consistent with the recent sell-off in many Taiwanese stocks. While it?s FAR
too early (both on our trip and relative to the calendar) to draw any definitive conclusions,
what we do know is that the semi stocks are not priced for any demand weakness, so
choppy demand datapoints warrant attention. In particular, we picked up somewhat
cautious datapoints on LCD TV sales, which support what our Asia team highlighted last
week regarding Samsung?s more muted tone on 32-inch TVs. Similarly on PCs, several
sources suggested that DRAM and monitor orders from PC OEMs are a bit below
expectations. We don?t want investors to extrapolate our end-demand comments because
it?s far too early but we believe they are worth noting.
KEY TAKEAWAYS FROM THE FIRST DAY OF OUR ASIA TOUR. During the first
day of our week-long Asia tour we met with several companies across the supply chain in
Taiwan, including ASE, Nanya, TPV, Compal Communications, DRAMeXchange, Arima
Computer, BenQ, and Amtran. Our key takeaways from our meetings in Taiwan include:
(1) MEETING WITH PC MONITOR-MAKERS AND DRAM SUPPLIERS SUGGESTS
THAT PC DEMAND MAY BE A BIT LESS ROBUST THAN EXPECTED IN CQ3,
BUT IT'S TOO EARLY TO MAKE A CALL ON CQ4. As we highlighted in our weekly
note, our global team's checks indicate that PC OEMs have recently cancelled some orders
at their DRAM customers and that these order cancellations may be driven, in part, by a
work-down of DRAM inventories and, in part, by slower than expected PC demand.
These cancellations were confirmed on Monday during several of our visits. We spoke
with TPV, a PC monitor and LCD TV maker. TPV indicated that it is still too early to tell
how robust PC demand will be during the seasonally strong fourth-quarter, but that the
September quarter isn't likely as robust as some had expected earlier in the year, perhaps
driven by hurricane Katrina and high oil prices impacting consumer demand. TPV's
biggest customers include Dell, HPQ, and Lenovo, with the company's top 5 customers
representing about 60% of revenues and no individual customer representing more than
30% of revenues. Other TPV customers on the PC side include Sony, NEC, Siemens, and
Acer.
In addition to a slightly softer demand read on PCs from TPV, BenQ, which also makes
LCD monitors for PCs, indicated during our meeting that its PC LCD monitor business
was weak in August. From a seasonal perspective, the company noted that August can go
either way with August sales good in a good year and softer in less robust years. While the
company has heard in the market that some PC OEMs have recently implemented some
component cancellations, despite the softer August, BenQ suggested that it has yet to see a
material drop in business from its PC OEM customers (which include Dell).
DRAMeXchange (which serves as a DRAM market research organization as well as a
DRAM trading house) noted that it has learned of cancellations to some of the DRAM
companies by a major PC OEM, which has recently revised its forecast downward due to:
1) demand that was pulled forward earlier this year through aggressive pricing, 2) some
softness in the U.S. consumer due to higher oil prices and the effects of hurricane Katrina, and 3)
the negative effect of Intel chipset shortages. We would note that the concerns about the U.S. PC
market are consistent with those suggested by Best Buy on its earnings calls last week, especially on
the desktop side.
(2) Consistent with our Asia team's downgrade of LCD last week, it appears that LCD TV
momentum has slowed and that it may take another price cut on mainstream 32-inch LCD TVs to
again kickstart the market heading into the holiday season. Several different contacts with whom we
met on Monday suggested that, while overall LCD TV demand remains at reasonably high levels,
the demand environment today isn't as robust as several months ago with TPV, Amtran, and BenQ
all suggesting that unit growth was choppy in either August or September. There seems to be a view
amongst these players that another price cut heading into the holiday season will stimulate
additional growth, so that is a situation that needs to be watched closely. We would note that the
view expressed by TPV, Amtran and BenQ is consistent with what our Asia team suggested last
week, that Samsung's tone on the 32-inch LCD TV demand is a bit more muted than it had been
several months ago.
3) According to our meetings, the handset market seems to be by far the healthiest end market with
none of the incremental signs of slowing or choppiness seen in the other end-markets. Interestingly,
one issue that we have wrestled significantly with lately is why Motorola's handset business is doing
so well (which was reconfirmed again by Compal, MOT's OMD for handsets), but Freescale's
baseband handset business doesn't seem to have the same momentum (which was also confirmed by
a contact on Monday, who suggested that FSL's baseband business was "not doing that well"
although Freescale's RF business was doing better). One likely explanation for this is that MOT's
ODM handsets mostly use TI baseband chipsets. So, the fact that TI supplies the baseband chipsets
for MOT's ODM handsets and that the ODM handsets are driving some of MOT's handset growth
likely explains why TI's baseband business is seemingly so much stronger than Freescale's baseband
business currently.
Tomorrow we will meet with several more companies in Taiwan including UMC, Winbond,
Gemtek, and Realtek. Below we provide greater detail from our individual company meetings.
BUSINESS AT ASE TEST REMAINS SOLID AND THE COMPANY CONTINUES TO ADD
CAPACITY RATIONALLY. We also met with ASE, an outsourced package and test house, in
Taiwan on Monday. The company said that customer indications have not changed in recent weeks
and customer forecasts have been very accurate for several months. Customers also continue to be
cautious although optimistic about Q4, inventory levels remain under control, and business is
growing in-line with seasonality. ASE began to raise its prices about 2-3 months ago and those price
increases were pretty broadly accepted by customers. The company is considering raising prices
again. Customer segments that are strong for ASE include consumer (driven by gaming chips),
low-end chipsets (as Intel's lack of low-end chipset supply is driving growth at VIA, SIS, and ATI),
and wireless (FSL, RFMD, SLAB, Philips, and Infineon are the company's biggest wireless
customers).
A large focus of our meeting was ASE's move to adjust its business model to have more stable
margin performance over the course of the full cycle. Management would prefer to have
approximately 25% gross margins that fluctuate +/- 5% over the course of the cycle as opposed to
30% peak gross margins and gross margins of about -5% at the trough. ASE is attempting to drive a
more stable business model by: (1) Adding capacity more rationally and requiring a residual value
guarantee from the tester companies in an effort to better understand and lower its costs. The
company is asking customers to meet certain volume specifications before adding significant new
capacity as ASE hopes to reduce the amount of idle capacity it is holding during downturns. While
this likely limits the significant growth that the back-end semi test companies tend to experience
during the heat of an upturn, it may also help to make their businesses less volatile. As the
outsourced package and test model has gained prominence in the past few years, the semi test
companies have seen increased volatility as outsourced package and test houses have quickly turned
on and off their capacity requirements. Management indicated that it remains very cautious in terms
of adding capacity and the company expects its capex in 2006 to be about flat with 2005 capex of
US$300 million. And (2) ASE no longer wants to provide services (including R&D) to its
customers for free.
In regards to the tester companies, ASE's largest tester suppliers are Agilent and Teradyne. The
company indicated that its tester supplier's cost structures remain too high, as the tester companies
are manufacturing in expensive regions (i.e. Munich for Agilent). ASE indicated that it has offered
the tester companies the option to work with ASE, using its facilities in Asia to manufacture lower
cost testers, although ASE does not know if the tester companies will accept this offer.
On the topic of tester company consolidation, particularly in light of Agilent's intended divestiture
of its semiconductor test business, ASE believes that if Agilent's business were purchased by
another tester company it would create a short-term disruption as platforms consolidate, although it
would be a long-term positive for the industry. This is because ASE believes there needs to be more
tester standardization, with testers more modularized and more easily upgradeable. That said, ASE
has not seen Advantest's T2000 tester gain significant traction as customers are not yet interested in
using the tester despite its open, modular design. Of interesting note, ASE has begun making its
own wirebonders, which we view as a negative for Kulicke & Soffa, which derives a significant
percentage of its revenues from ASE. Importantly, we would note that ASE is not sure of how
aggressive it will be with its own wire bonder build out so investors shouldn't extrapolate too much
from this initiative.
In 2006, ASE expects to be able to achieve a minimum of 10% Y/Y top-line growth. Management
expects to exit this year at a gross margin run-rate in excess of about 20% and is trying to manage to
a gross margin of about 15% in the seasonally weaker first-quarter of 2006. For full-year 2006, the
company is targeting a blended gross margin of about 20%. ASE is also aiming to reduce its
operating expenses to 8% from the current approximate 10% level. Management also noted that
testing gross margins were near about 30% in August.
BENQ PROVIDED AN UPDATE ON THE SIEMENS WIRELESS ACQUISITION AND
CURRENT BUSINESS TRENDS IN EACH OF ITS BUSINESS SEGMENTS. Our meeting with
BenQ focused on the company's recent acquisition of Siemens' wireless handset business and
implications for respective incumbent baseband suppliers (Texas Instruments for BenQ and
Infineon for Siemens) as well as end market trends related to wireless handsets, optical disk drives,
LCD TV, and DLP projectors.
Regarding BenQ's acquisition of Siemens handset business, BenQ indicated that it believes the
acquisition and integration of the wireless handset business of Siemens is going as planned,
including the EU anti-trust approval, organizational appointments, rationalization of combined
product roadmap, and identification of cost discrepancies on various components. Regarding cost
discrepancies, the company estimates approximately 10-15% on average across three categories: (1)
ICs (30-40% of COGS), (2) customized components, including keypad, casing, camera module, and
LCD module (40-50% of COGS), and (3) commodity components, such as capacitors/MLCCs
(20-30%). BenQ indicated that it held an advantage on ICs and customized components, while
Siemens had an advantage in commodity components due to scale. Siemens shipped 51M units in
FY04 vs. BenQ's 16M in CY04, however, both businesses have been impacted negatively in 2005
by respective issues. BenQ will work to rationalize the product roadmap and component suppliers,
looking to move ideally to no more than two suppliers for primary components. Baseband has been
speculated as an area for potential rationalization given Texas Instruments' incumbent position
related to BenQ's wireless handsets and Infineon's incumbent position related to Siemens' wireless
handsets. BenQ noted that both suppliers will continue as strategic partners, considering high
switching costs and the advanced stage of current product development, with product roadmap
pruned. BenQ hopes to work with incumbent suppliers in developing "win-win" relationships going
forward, with a focus on price/performance (in particular related to Infineon), diversity of the
supplier base, and driving economies of scale. We would note that BenQ also highlighted Mediatek
as having a competitive solution with the possibility to become a baseband supplier going forward.
The company indicated that business trends across its product segments are: (1) Wireless handsets -
BenQ's handset business trend has been weak in 2005, as the company has transitioned its business
model from ODM to its own brand and going forward by acquisition of Siemens' wireless handset
business. (2) LCD monitors - BenQ expects its LCD monitor business to recover in the month of
September from a decline in the month of August, with September shipments expected to be better
than August and July; the company attributes the dip in August to customers' product rotation. (3)
DLP projectors - BenQ noted mixed trends for its projector business thus far in 2005 with its own
branded business up 30%+ Y/Y, but ODM-related sales down significantly due to the loss of the HP
account, which accounted for quite a bit of the company's volume in 2005. In addition, industry
trends have been weak, with the industry full year shipment forecast downticked to the 4.0M - 4.2M
range from the original projection at the start of the year of 5.2M. Industry shipments have been
impacted by inventory, which BenQ expects has been rectified. The company is also encouraged by
new products, including wireless and handheld (2 lbs) projectors. (4) Optical storage - BenQ noted
that its optical storage business had a tough month in August and a tough Q2, but the company
remains optimistic going into Q3 due to seasonality and the ramp of the latest generation DVD-RW
product. The company noted pricing pressures, in particular driven by Korean and smaller
Taiwanese competitors. (5) LCD TV - BenQ highlighted stiff competition, in particular from a
competitor in Europe and pricing pressure in the Asia-Pacific region, as well as, vendors looking to
clear out of whitebox goods. BenQ shipped 70K units in Q1, with Q2 down, and Q3 expected to
start to recover, with the pace of decline slowing, but unit shipments still likely flat to down Q/Q.
The company is hopeful that unit growth will occur in Q4.
TPV MEETING SUGGESTS THAT SEPTEMBER PC MONITOR TRENDS MAY NOT BE AS
STRONG AS EXPECTED. Recall that TPV is a maker of PC monitors and LCD TVs. On the PC
monitor side, as we noted earlier, TPV indicated that it is still too early to tell how robust PC
demand will be during the seasonally strong fourth-quarter although September may not have been
as robust as some had expected earlier in the year, perhaps driven by hurricane Katrina and high oil
prices. TPV's biggest customers include Dell, HPQ, and Lenovo, with the company's top 5
customers representing less than 60% of revenues and no individual customer representing more
than 30% of revenues. Other TPV customers on the PC side include Sony, NEC, Siemens, and
Acer.
TPV recently purchased Philips' monitor and entry-level LCD businesses with Philips becoming a
15% shareholder of TPV (potentially 30% depending on its convertible bond) post the deal. TPV
expects Philips-related volume in the first year post the acquisition to reach 7M-8M units for
monitors and 1M-2M units for flat-panel TVs. The businesses that TPV acquired from Philips are
currently operating at a profit of about 0.5%, and TPV's goal is to bring this profit to about 3% by
realizing synergies that include: (1) procuring from less expensive regions, (2) changing the
manufacturing process from outsourced to in-house, and (3) redesigning monitors to save at least
1%, which takes about 6-9 months. The company believes that it can double its profit from US$100
million by 2007/2008 due to these synergies.
Management indicated that the company usually realizes about 55% of it sales in the second-half of
the year. Without integrating the Philips business, the company expects to ship 17M LCD monitors
and 10M-11M CRT monitors. Including the Philips business, the company expects to ship 25M
LCD monitors and 11M CRT monitors (which management believes is a conservative estimate, due
to the fact that the market is declining). The company indicated that it is quite concerned about
excess supply, but it believes there are many factors involved to determine if excess supply is a
problem, including end-demand during the seasonally strong fourth-quarter. In terms of the ICs that
TPV uses in its products, for LCD TV ICs the company uses Philips, Trident, and Genesis.
Management will likely consider another 1-2 Taiwanese suppliers. For monitor ICs, TPV primarily
uses Genesis as well as Novatek.
In regards to PC monitor unit demand, TPV indicated that flat panel monitors continue to grow at
the expense of CRT monitors. The company quoted third party unit growth estimates, stating that
LCD monitors are expected to grow to over 100 million units in 2005 from 70 million units in 2004.
Units are expected to reach 115-120 million units in 2006 with growth expected to slow in
2007/2008 (at which point in time TPV expects TV demand to drive the company's top-line
growth). TPV also indicated that PC customers are moving to replace 15-inch monitors with 17- and
18-inch monitors and the company expects 17-inch monitors to have a more dominant market share
than 15-inch monitors over time. TPV also said that some customers have expressed the intention of
moving to a wider monitor format (20 - 21 inches) upon the Microsoft Longhorn launch in 2006.
AMTRAN ALSO SUGGESTED THAT LCD TV ORDER TRENDS IN SEPTEMBER ARE A BIT
SOFTER THAN EXPECTED, BUT IT IS HOPING THAT ANOTHER PRICE REDUCTION
WILL DRIVE HIGHER SALES DURING THE HOLIDAY SEASON. Our meeting with Amtran,
the Taiwanese display OEM/ODM company for both PC Monitors and Flat Panel TVs (Plasma and
LCD), also suggested that LCD TV order trends in September were a bit softer than expected. The
company still expects a global LCD TV market of 18M-20M units in 2005 (in-line with the GS
estimate of 19.2mn units), but mentioned that the Q2 halt in LCD panel price declines may put a lid
on elasticity of demand. In the company's opinion, 32" LCD TV retail prices will have to come
down to the $899 mark (from the $999 Amtran is seeing currently) to drive incremental adoption,
which will only be achievable if 32" panel prices drop to around $500 by the middle of Q4. Amtran
currently uses components from Genesis (42" LCD, 50" Plasma TV as well as Monitors), MStar
(monitors), Trident (LCD TV) and Mediatek (for its 32" LCD product, which is expected to be the
mainstream for this Christmas season). The company currently does not employ chips from
Micronas. Amtran's primary panel source tends to be LG Philips and Samsung, while the company
also sources from the Taiwanese Panel makers.
NANYA AND DRAMEXCHANGE MEETINGS HIGHLIGHT THAT TRENDS IN DRAM
CONTINUE TO BE SLUGGISH, UNDERSCORING OUR LESS CONSTRUCTIVE VIEW ON
SHARES OF MICRON (U/N) AND INFINEON (IL/N). Nanya indicated that, in general, it expects
H2 September DRAM contract prices to be down slightly (-1% to -2%), which is a cause for
concern in what is normally a seasonally stronger period. DDR1 contract prices are expected to be
slightly up in H2 September, while DDR2 contract prices are expected to trend down as the
mismatch between DDR1 and DDR2 supply/demand that has existed for several months continues.
Nanya also indicated that the industry is likely to see the DDR1/DDR2 cross-over in Q4'05. Nanya
has not seen any cancellations from the likes of Dell and HP as suggested by others on our trip
although the company did suggest that orders from the US market were weaker than expected in
August. September bit shipments are expected to grow 5% M/M, in-line with previous expectations,
and the company is sticking by its output plan of +27% Q/Q and +10% Q/Q in Q3 and Q4,
respectively.
Nanya's comments with respect to regional bit-per-box configurations would suggest that the
company does not necessarily believe in a substantial increase in content per PC through year-end,
with China currently running at approximately 400 MB/box, and India at about 300MB/box,
compared to an estimated 600MB/box in the U.S. In other words, the move to lower-end emerging
market driven PCs is hurting the DRAM industry as it takes 1.5 to 2 PCs sold in those emerging
markets to equal 1 PC sold in the U.S.
Early indications on 2006 suggest between 40% and 45% Y/Y bit shipment growth, in-line with
what the company expects to be the broader industry bit growth rate, although we believe that that
industry bit growth will be significantly greater than 40-45% in 2006 given that DRAM makers'
capex is at an 8-year high in 2005. On current output plans, Nanya is expecting 2006 capex to be
flat to up 10% Y/Y (2005 capex is about $120 million). That said, this capex number could be
significantly higher if Nanya goes ahead with its tentative plans to build its own 300mm fab. The
bottom-line on this is that if Infineon decides to exit the DRAM business and sell its stake in Inotera
(the joint DRAM venture between Infineon and Nanya), then Nanya would go ahead and build its
own fab as it would likely not feel comfortable sharing output with another partner. Our
conversations and other checks suggest that Nanya is waiting for Infineon's management board to
give a presentation to Infineon's supervisory board in Q4 regarding Infineon's potential strategic
options for its DRAM business. Separately, we believe that Inotera's 2006 capex (which is funded
equally by Nanya and Infineon) is likely to be flat to up 10% from its 2005 level of about $900m.
Our discussion with DRAMeXchange focused on its business methodology and trends in the
DRAM industry. DRAMeXchange tracks and reports DRAM market data, including spot market
and contract pricing trends based on surveys with spot brokers, memory module makers, and PC
vendors. The company also operates a relatively small trading business, which averages
approximately $10M/month and a volume of 250-500K units (DRAM chips)/day. The company is
focused on market pricing information, rather than becoming a significant trader.
DRAMeXchange expects that spot pricing could rebound moderately in the month of September,
though the company is more cautious than it had been previously as many DRAM manufacturers
have brought up 12-in capacity ahead of schedule and DRAMeXchange now estimates that CQ4
supply bit growth (10%) will outpace demand bit growth (4-5%) quite substantially. H2 September
contracts are expected to be flattish to down for DDR-2 and likely flattish for DDR-1; DDR-2
remains tepid with some demand seen from PC OEMs, but little from the whitebox market due to
little incremental benefit from DDR-2. DDR-2 demand is expected to improve in the spot market
only after the 667MHz chipset is brought to market, as this speed will allow for improved
performance in PCs. DRAMeXchange doesn't expect DDR1 vs, DDR2 crossover until Q1 2006
(note that is later than Nanya) as that is when notebook is expected to approach 50% (vs. 40%
currently). Notebooks are more likely to adopt DDR2 because there is a power consumption benefit
to DDR2 compared to DDR1. DRAMeXchange noted that there is a 30% power savings recognized
in the memory module by moving to DDR2, but since memory consumes only 10% of the overall
power in a notebook, the power savings for the entire system are about 3% (10% of 30%), still
significant enough to drive DDR2 adoption.
ARIMA'S SHIPMENTS TO SONYERICSSON CONTINUE POSITIVE MOMENTUM. Our three
main takeaways from our visit with Arima are: (1) no change to trends are evident from HP's orders
to Arima, though Arima is a smaller ODM for PCs and is benefiting from the recovery of market
share at this account, so should not be used to make broader industry implications, (2) Arima's
shipments to SonyEricsson show continued positive momentum, and (3) Arima and SonyEricsson
have been evaluating the use of other chipset vendors (current suppliers include Broadcom); the
strongest contender is Ericsson Mobile Platforms (EMP).
For PCs, Arima expects stronger-than-market growth of up 20-25% H/H in H2'05, vs. its estimate of
up 10-15% H/H for the industry, due to a ramp-up with HP upon the re-initiation of their
relationship, which was terminated a few years ago. Arima's other ODM customer, Gateway,
continues to struggle and the loss of momentum is evident at Arima in the form of
lower-than-expected Gateway business. While Arima has seen no recent change to HP's order
forecast, we would highlight that the company has a very small business in this segment and hence
we would be careful to draw broader market conclusions from its commentary.
For handsets, Arima's customer concentration is heavily weighted towards SonyEricsson (~75% of
CY'05E sales), for which it is the sole ODM; SonyEricsson also leverages internal manufacturing
and OEM manufacturing. Arima's other handset customers include NEC (~18%) and trailing
revenues from China ODM handsets (~7%), which Arima is exiting. For SonyEricsson, Arima has
seen relatively good business trends, with stable unit volumes for one model and
better-than-expected volumes for another; we note that Arima estimates that SonyEricsson uses
Arima for only about ~16-17% of its total handset business, so this may not be indicative for
SonyEricsson's business as a whole. Arima noted that chipset vendors have been presenting
competitive solutions and Arima/SonyEricsson may consider leveraging in future handsets; current
chipset suppliers to Arima include Broadcom, however, EMP appears to be the strongest
competitor. On components broadly, Arima is not seeing any shortages; the company expects that
component pricing, broadly, is continuing to trend downward.
6 Goldman Sachs Global Investment Research
September 19, 2005 Analyst Comment
COMPAL COMMUNICATIONS MEETING HIGHLIGHTS THAT MOTORLA ODM HANDSET
UNITS REMAIN STRONG. Our main takeaway from our visit with Compal is that Motorola ODM
handset units remain strong, which benefits chipset supplier TI, though likely embedded in TI's
healthy forecast. Compal is the primary ODM for Motorola, with ~90% of its CY'05E revenue from
Motorola; its other two customers are Panasonic (~8-9%) and Alcatel (very small). Additionally,
Compal expects to announce another top-10 ODM customer soon, and will gain CDMA handset,
PDA, and a Japanese customer's 3G handset business when PMCC (a Compal Electronics
subsidiary) is transferred to Compal in Jan-06.
Compal has benefited significantly from Motorola's strength in 2005, as its projected CY'05E unit
volume in the mid-30M's far exceeds its Nov-04 target for the year of ~15M. Compal sees its solid
execution to timelines, ability to support Motorola despite the significant unit surprise, high quality,
and cost-saving design as key factors in its strong relationship with Motorola. Compal shipped 6.8M
units in Q1'05 and 8.7M units in Q2'05, vs. its initial estimate of 5-6M which was based on
Motorola's 6-month rolling forecast. Its H2'05 unit expectation is currently for 9M-10M unit
shipments in each of Q3'05 and Q4'05. Compal continues to experience upside surprises to the
forecasts received from Motorola, suggesting Motorola unit strength continues to be solid for
low-end handsets; this strength would be an incremental positive for TI, which provides the chipset
for most Motorola ODM handsets. On components broadly, Compal continues to work with its
suppliers to improve efficiencies and continuously lower costs.
Each of the analysts named below hereby certifies that, with respect to |
