GS US Semi & SPE weekly - Takeaways
from west coast company meetings and
SanDisk patent call
1) We spent last week on the west coast meeting with a diverse group of semi and SPE
companies. According to our meetings, October was a good month for the semi industry,
but many semi companies are cautious on business for the rest of quarter, both because of
what the semi companies are hearing from their OEM customers, and the fact that
business momentum has faded at the end of the quarter for the majority of the last 6
quarters. This dynamic is what caused the earnings disappointments last quarter. 2) There
are several events this week including AMAT, MRVL and ADI earnings and analyst
meetings from AMD, KLAC and TER. MRVL continues to be our favorite idea in the
semi space due to product cycle driven opportunities. 3) Our research on SanDisk?s patent
litigation suggests that Samsung will continue to pay SanDisk royalties, but royalties from
other producers of NAND are less likely.
SNDK (IL/N): SANDISK MEETING WITH MANAGEMENT FOCUSED ON
GROWTH PROSPECTS, CAPACITY, AND PATENT LITIGATION. We met with
SanDisk management last week at the company's headquarters. Our meeting focused on
the growth drivers of the business as well as recent patent litigation and its potential
impact on the company's royalty revenue stream. Regarding potential growth drivers,
management is most excited about growth opportunities in the mobile phone market.
Mobile phones grew as a percentage of SanDisk's revenues from 6% in 2004 to 15% in
Q3'05. Management estimates that there will be about 150M mobile phones with flash
card slots in 2005 and the company expects this number to grow to about 350M in 2007.
On the debate regarding embedded flash memory vs. removable flash memory in mobile
phones, while SanDisk offers both solutions, management believes that removable flash
will likely be the preferred storage means in mobile phones given that carriers are unlikely
to want to subsidize the cost of storage. Additionally, management believes that the
consumer prefers the portability aspect of removable flash storage.
Regarding SanDisk's capacity situation, the company is currently ramping 70-nm at its
200mm facility and 90-nm at its 300mm facility. In Q4, management expects its 70-nm
product to be about 25% of the company's captive bit supply and its 300mm product is
expected to be about 30% of captive bit supply. At the company's joint-venture 300mm
facility with Toshiba, Fab 3, the company is ramping wafer starts to an average of 10k
wafer starts per month (WSPM) in Q4'05 and intends to reach about 30k WSPM by the
end of Q1'06. SanDisk then expects to ramp capacity at Fab 3 to 49k WSPM by Q1'07.
The company doesn't have a specific capacity ramp plan after Q1'07, but does aim to
eventually get to 62.5k WSPM in capacity at Fab 3. Note that the Fab has enough
cleanroom capacity to eventually ramp to 100k WSPM, although this is not currently
anticipated in the company's capital spending plans.
Another key area of focus during our meeting was recent litigation with STMicro. On
Friday last week, we hosted a conference call with an experienced patent attorney during
which the attorney outlined key areas of focus in order to better understand the potential
ramifications of the patent infringement suit, particularly as it relates to the royalty stream
Samsung pays SanDisk. Please contact our office if you would like the replay information
for the conference call.
SanDisk brought a patent infringement suit against STMicro before the International Trade
Commission (ITC) in 2004, under which SanDisk alleged the STMicro was infringing upon its '338
patent (which relates to flash memory). The ITC issued an initial determination in the suit a few
weeks ago, indicating that SanDisk's '338 patent is enforceable and valid, but that STMicro is not
infringing upon the patent. SanDisk management indicated during our meeting that they have filed
an appeal with the ITC requesting a review of the initial determination and the ITC has 30 days to
answer whether or not it will consider reviewing the initial determination. The patent case with
STMicro is important because it is regarding the same patent for which Samsung pays SanDisk a
royalty. In 1997, the ITC ruled that Samsung was indeed infringing upon the '338 patent and as a
consequence, Samsung and SanDisk entered into a licensing agreement under which Samsung pays
SanDisk a royalty rate in exchange for the right to sell its NAND flash product in the US.
The attorney hosting our conference call posited several interesting views, including: (1) The ITC's
initial determination that STMicro is not infringing upon SanDisk's '338 patent is unlikely to have
any impact on the contract between Samsung and SanDisk, and in turn, the royalty stream SanDisk
receives from Samsung. Note that the contract expires around fiscal year-end 2009 and we
understand that it covers SLC and MLC technology. (2) In the attorney's experience (although he
has not seen the contract between Samsung and SanDisk and is only offering his view) typically a
way in which a contract of this nature is broken is if the patent is found to be invalid or
unenforceable, which the ITC did NOT conclude in its initial determination in the STMicro case.
And (3) a patent usually consists of several "claims" that define the underlying characteristics of the
technology being protected by the patent, with each claim consisting of a number of elements. A
company is found to infringe on a patent if the device being accused of infringement, infringes on
one or more claims covered by the patent. Further, the characteristics of the device in question need
to match all of the elements in a given claim in order to infringe on that claim, and in turn the
patent. If one or more elements of a claim are missing from the allegedly infringing device, then
there is no infringement.
Aside from the Samsung royalty stream which we believe is very solid, it is still too early to make
any firm conclusions regarding SanDisk's royalty stream from other producers of NAND flash for
the next several years, both because the suit with ST isn't final, and because there are a number of
complications regarding royalty payments from many of the potential incremental
producers/entrants into the space. For example, Intel has a license from SanDisk but may not have
to pay royalties because of technology that Intel has licensed to SanDisk. Hynix may not have to pay
royalties to SanDisk because they are in a joint venture with ST Micro to produce NAND flash and
may be covered by the recent initial finding from the ITC.
Our bottom-line conclusion is,
recognizing that a lot could change over the coming months, we believe it is prudent for investors to
only model royalty income from Samsung because it is possible that royalty payments from other
producers of NAND will be limited. (Goldman Sachs & Co., and or one of its affiliates, is acting as
financial advisor to STMicroelectronics NV in a potential transaction with Hynix Semiconductor
Inc. Goldman Sachs & Co., and or one of its affiliates will receive a fee for its financial advisor
role).
XLNX (IL/N): MEETING WITH MANAGEMENT FOCUSED ON L/T GROWTH DRIVERS,
GROSS MARGIN TRENDS, MANUFACTURING. Our discussion with Xilinx CFO Jon Olson
centered on company and industry growth drivers, traction of new products, gross margin trends,
and manufacturing. 1) We believe near-term business trends are tracking fairly in-line with Xilinx's
CQ4 guidance for 1-5% Q/Q growth with a sales rebound in the month of October, but still too early
to become overly ambitious due to high turns requirement and sales volatility in recent months. We
remain comfortable with our CQ4 estimates of $407.1M (+2.0% Q/Q)/EPS of $0.23, versus street at
$412.9M (+3.5% Q/Q)/$0.24. 2) Xilinx views communications, consumer, and automotive markets
as its most likely growth drivers over the intermediate term and the company has an optimistic
outlook in these areas given design win success of newer products, in particular Virtex-4 and
Spartan-3. The company is also encouraged by traction of embedded cores, including high-speed
transceivers, DSPs, and embedded processor. 3) Gross margin trends have benefited from improved
yields on advanced products, which are now back on track with expected defect density/cost curve
2 Goldman Sachs Global Investment Research
November 13, 2005 Analyst Comment
following the resolution of prior yield weakness related to difficulty at 90nm and conversion to
300mm. We expect gross margins will likely remain fairly predictable over the next few quarters as
advanced product yields continue to mature and advanced products (and respective costs/margins)
are now flowing through normal supply chain/inventories, rather than being expedited. 4) We
expect Xilinx remains in relatively good shape for both foundry and backend capacity, and that the
company is generally able to fill needs within lead times. We believe Xilinx is cognizant of working
capital and i
ts high inventory, and expect the company will continue to work with suppliers to gently thin
production over time. There is no change to our IL/N rating.
NVDA (IL/N): MEETING WITH MANAGEMENT SUGGESTS AGGRESSIVE PRODUCT
LINE-UP, AFFIRMING OUR PREFERENCE OVER ATYT; REVENUE GUIDANCE REFLECTS
SOME CONCERN OVER MACRO ENVIRONMENT. We met with Nvidia management last
Thursday and have three main takeaways. (1) It appears that Nvidia's somewhat disappointing FQ4
revenue guidance was largely driven by management's concern following Dell's preannouncement
and cautious comments on the US consumer. Nvidia has limited near-term visibility into its sales
given the desktop discrete graphics business is split 60-40 between the add-in-board channel and
OEMs, with channel order patterns being notoriously volatile. (2) In the intermediate term, we
maintain our view that Nvidia is better positioned competitively than rival ATI. Nvidia's current
desktop graphics processor (GPU) line-up is at least on par with ATI's even though Nvidia has more
headroom for improvement since its GPUs are on 110nm vs 90nm for ATI. Nvidia is set to launch a
top-to-bottom family on 90nm this quarter and next, which will likely exceed ATI's performance as
the smaller geometry allows for faster speeds. In addition, Nvidia is readying its GeForce 8xxx
series of GPUs for 2H'06, representing another architectural leap. On the notebook side, Nvidia has
won a number of design wins with OEMs, including Dell, for the spring'06 refresh window, which
will likely meaningfully increase its market share. (3) We remained unconvinced that Nvidia would
reach its 45% gross margin target over the next 1-2 years, since much of its growth is coming from
products that have below-average margins: notebooks, chipsets, handsets, and lower-end desktop
GPUs. We believe that growth in workstations (a business with $400M in annual sales) which has
above-average margins is not enough to offset that. Further, Nvidia already derives 80% of its GPU
sales from its higher-margin 6/7 series products than the older FX architecture, thus there is limited
room for additional mix improvement. We are modeling improvement to only 42% over the next
year and view that as a more achievable target.
MCRL (U/N): MEETING WITH MANAGEMENT POINTS TO FURTHER GROSS MARGIN
IMPROVEMENT; LOW INVENTORY PRESENTS A CHALLENGE TO MEETING HIGHER
HANDSET DEMAND FROM A KOREAN CUSTOMER. Our meeting with Micrel's management
last Friday focused on the drivers for gross margin expansion and return to topline growth.
However, near-term the handset and inventory issues from last quarter persist. (1) Micrel's gross
margins will likely continue to trend upward as the company ramps utilization from the mid-50%s
range, lowers capex from $34M average in the last 5 years to $20M average over the next 5 years,
and lowers depreciation/sales from 7.8% to under 6%. (2) We remain unconvinced that the
company can grow sales above our analog industry forecast of 8%, as its analog design team is
smaller than competitors on a sales-adjusted basis, and pricing discipline on legacy products (70%
of sales) will necessarily hinder growth. Micrel expects most of its growth to come from the
higher-margin wireline segment (20% of sales), where it is gaining share in high bandwidth
products from On Semiconductor and also selling "drag-along" analog content. In addition, it
expects to see growth from the lower-margin handset segment (27% of sales), where it has
penetrated Nokia and SonyEricsson in addition to long-standing customers Samsung and LG. Micrel
expects its CDMA/GSM mix to go from 80/20 to 40/60 over the next couple of years. We expect
industrial (26%) growth to be more challenging given the need for a large distribution network, and
computing (23%) to continue facing pricing pressure. On the consumer side ( <3%), Micrel should
see some growth in 2006 as it is single-sourced in the PS3 with a $0.75 LDO part, and also sells the
same LDO into a Sony HDTV. Finally, Micrel has engaged two new higher-margin customers for
its foundry business, which should grow from ~2% to ~4-5% of total sales. (3) The current quarter
continues to present challenges of inventory management in meeting handset demand, similar to last
quarter. Micrel noted it is once again facing much higher than forecasted last-minute demand from
one of its Korean handset customers, without having the inventory in place to meet it. If Micrel can
execute to meet some of this excess demand, it is possible it can show better-than-expected Q4
results. However, significant upside is unlikely given that Micrel entered the quarter with too low an
inventory level, and already premised its guidance on a higher turns-fill requirement than Q3. As a
result, we believe Micrel may have to make some difficult trade-offs with regards to which
customers' demand it can meet. A higher-than-expected handset mix would likely result in lower
gross margins.
MPWR (IL/N): WE ARE STILL POSITIVE ON GROWTH OPPORTUNITIES AND
INCREMENTALLY LESS CONCERNED ABOUT LITIGATION RISKS AFTER LAST WEEK'S
JUDGMENT AND OUR MEETINGS WITH MANAGEMENT. We attended meetings with
Monolithic Power Systems (MPS) last week, and remain positive on its growth opportunities in
DC/DC converters, and longer-term into other areas such as LDOs, op-amps, and voltage regulators.
We also have three takeaways on the litigation matters: (1) We learned that MPS has recently sued
Sumida for breach of contract. Recall that Sumida, an MPS customer, is being sued by MPS' rival
O2Micro for patent infringement in a trial that starts today. MPS has been paying Sumida's legal
expenses in this case, but is now suing it for reimbursement. This suggests that MPS' litigation
expenses could run a bit higher and/or longer than previously expected, though if successful MPS
could receive a couple of million dollars back. (2) On Friday, the judge in the Oakland O2Micro vs.
MPS case entered a judgment setting aside the $12M jury award to O2Micro, instead awarding
O2Micro $2.7M plus costs-a net benefit to MPS of $9.3M or $0.30 per share (the stock was up
$2.12 on Friday). (3) The MicroSemi and Micrel trials will likely commence in H1'06. Based on our
meeting with Micrel's management last week, it is our sense the company intends to pursue a trial
rather than a settlement; however, it first needs to overcome MPS' motion to dismiss based on the
statute of limitations. We believe that even with an adverse outcome to MPS, a monetary award is
more likely than a business-crippling injunction. We maintain our IL/N rating despite our positive
view on MPS' growth opportunities given the headline risk from the ongoing litigation. In our view,
the risk to estimates from adverse litigation outcomes is fairly reflected in MPWR's 23x CY'06 P/E
multiple.
SEVERAL ANALYST MEETINGS TAKING PLACE THIS WEEK INCLUDING AMD,
TERADYNE, AND KLA-TENCOR. This week there will be several analyst meetings across the
Semiconductor and SPE sectors including AMD, Teradyne, and KLA-Tencor. We expect AMD's
analyst meeting on 11/15 to focus on the company's technology/product roadmap, share gains in
dual core processors, as well as the company's manufacturing strategy. We believe that KLA's
analyst meeting on 11/16 is likely to focus on widely speculated management changes within the
company. We expect Teradyne's analyst meeting on 11/17 to focus on the company's new business
model post the TCS divestiture, the company's FLEX platform and what traction it is gaining with
customers, and progress that the company is making in its assembly and test division. (Goldman
Sachs & Co., and or one of its affiliates, is acting as advisor to Teradyne, Inc in the proposed sale
and or partial sale of its Connection Systems division to Amphenol Corporation. Goldman Sachs &
Co., and or one of its affiliates, will receive a fee for this advisory role).
MRVL (OP/N): CONTINUE TO VIEW AS THE BEST GROWTH STORY IN OUR GROUP;
ORGANIC BUSINESS TRENDS H
EALTHY AND ACQUISITION CONTRIBUTION SOLIDLY ACCRETIVE. We believe Marvell's
business tracked well during its FQ3 (Oct) and would expect solid results and guidance. We
maintain our Outperform rating on Marvell in the context of our Neutral semiconductor devices
sector view, as we continue to view Marvell as the best growth story in our coverage group. We
expect continued solid execution to drive sales and earnings momentum. MRVL trades at 33x our
CY'06E EPS, ahead of the group median at 23x, but fair in our opinion given Marvell's
higher-than-industry growth on solid product cycle-driven opportunities. We expect stock
outperformance will continue to be driven by continued estimate momentum, which we believe
remains biased to the upside. MRVL is our top pick in the semiconductor space and we believe
investors should continue to buy the stock.
We expect FQ3 will be in-line to modestly ahead of our estimates of $420.7M (+7.7% Q/Q)/$0.33,
similar to the Street at $420.8M (+7.8% Q/Q)/$0.33, with potential for EPS upside on top-line
strength, operational execution, and leverage. We believe Marvell experienced broad strength
including: 1) acceleration of WiFi with holiday build for Playstation Portable (PSP; Sony recently
revised PSP production to 14M from 13M for its fiscal year ending March) and beginnings of ramp
into broader range of consumer devices, 2) pickup in enterprise networks transition to gigabit
Ethernet, and 3) in-line growth in data storage with strong seasonality in desktop HDDs (including
good CQ3 results and CQ4 outlook from Marvell's largest customer Western Digital), offset by
choppy enterprise HDDs. We remain comfortable with our FQ4 (Jan) expectation of $451.2 (+7.2%
Q/Q)/$0.35, vs. the Street at $453.9M (+7.9% Q/Q)/$0.36, as we believe Marvell will benefit from
continued product cycle ramps related to WiFi and gigabit Ethernet. Our current estimates do not
yet factor in contribution from acquisition of QLogic's hard disk and tape drive controller business,
which closed on 11/4. We believe the acquisition will be solidly accretive as it had sales of
approximately $40M/quarter with gross margin around 60% and operating margin near 50%. While
incremental EPS of about $0.05/quarter is possible, we would expect on-going contribution to be in
the $0.02-0.03/quarter range factoring in lower interest income and higher shares (Marvell paid
$180M in cash and $45M in MRVL shares), likely increased investment for product development,
and natural conservativism on acquisitions.
ADI (IL/N): EXPECT IN-LINE OCT QTR RESULTS AND MORE DETAILS ON RECENTLY
ANNOUNCED RESTRUCTURING. We expect ADI to report FQ4 results in-line with our
estimates of $617.4M (+6.0% Q/Q)/$0.35, versus Street at $615.3M (+5.7% Q/Q)/$0.34, and in-line
with revised guidance issued on 10/19 for +6% Q/Q. We expect Analog has participated in strong
seasonal demand trends for consumer, computing and handsets, having had disappointing results for
the previous past 4 quarters. Also, we will look for more detail around the recently announced fab
closure and organizational cost reductions. We remain comfortable with our FQ1'06 estimates of
$625.6M (+1.3% Q/Q)/$0.36, in-line with +1% Q/Q seasonal growth. We rate Analog Devices
IL/N. While we view improved near-term sales momentum and cost-reduction actions
constructively, we believe that ADI is fairly valued and expect shares to remain rangebound, as
industry trends and ADI's own business pattern remain choppy. ADI trades at 23x CY'06 P/E,
in-line with the analog median (excluding ESO expense) of 23x, a reasonable multiple given Analog
Devices' competitive technology positions, broad customer base, and solid margins. Our checks
suggest that the sale of ADI's baseband business could be somewhat imminent so if there is a major
surprise on the call, it could be an announcement of the previously speculated sale of this business.
WEEKLY MEMORY MONITOR: NAND flash memory retail prices declined 1% W/W and
increased 4% M/M, with spot prices down 4% W/W and up 10% M/M. We would also note that
both DRAM and NAND contract prices were released last week, with DRAM contract prices
declining ~4.5% on average in 1H November (please refer to Tables 2 and 3 for more detail
regarding the DRAM and NAND contract prices).
Each of the analysts named below hereby certifies that,... |
