Saturday, January 22, 2011

The Stance of Semiconductors in 2011

Even if cyclical metrics (units, inventory, capex) are normalizing in 2011 (versus being under trend in 2010),I remain bullish on the semiconductor cycle as I believe we are in the second year of a multi-year method of revaluing the group from “cyclical” back to“growth”.In 2011 the stock returns will be more heavily dependent on the demand trends.
The semiconductor industry grew revenue 32.9% in 2010.The units were up 30.2%, ASPs were up 2.1%., I believe the semiconductor industry is on track to raise 2% in 2011, units up 11.4% with ASPs down 2.0% approximately.
The last time I experienced double digit growth in two consecutive years was 2003 & 2004, 1999 & 2000 and 1993, 1994 & 1995. According to segment, I estimate that Micro Controller Units, Communications, and Analog will outgrow and the overall semiconductor industry increases in 2011 while PCs Semis and Memory will stumpy grow.

The most important possibility to our growth expectations in 2011 is the possible for a very optimistic 1Quarter and 2Quarter manufacture of tablets that can or can not be doing well in the market place. I believe in 2011 investors will be forced to identify that increased semiconductor content in the economy. Tops down we are on the cusp of acceleration in the growth of the global middle class, bottoms up we have multiple product areas of increasing content: Green, Smartphone, Tablets, Autos, and Industrial, et cetera.

The core drivers of the forecast
There are two main components to our bullish PC view: (1) A continuation of cyclical recovery in Corporate PC demand, which while well understood by investors, may still be being underestimated and (2) A structural acceleration of global PC penetration especially in the consumer market, which is still being under-appreciated by investors.
Semiconductor Cyclical
There are 3 elements to the cyclical argument to growth in 2011.
Corporate PC Refresh rate could return to pre-crisis levels. In 2009, we estimate that the PC refresh rate fell to 18% (i.e. average life span of 5½ years) and recovered to approximately 21% in 2010. However, we would note that an aging installed base of PCs as a result of the Windows Vista refresh skip in 2006-07 could accelerate the corporate refresh rate back to the 2006-2008 average of 24% (or 4¼ years). Our current forecast of 14% PC unit growth assumes a corporate PC refresh rate of 22% (or 5 years); if the refresh rate were to return to 24% in 2011 we could see upside to our estimate – specifically, on average every 100 bps increase in refresh equates to 190 bps increase in unit growth from the current base. Accelerated Consumer PC Unit Growth. Weaker consumer PC demand in 2H10 (specifically in developed markets) caused consensus 2010 PC unit growth estimates to be revised down from ~19% in 2Q10 to ~14% by year’s end. The easy compares resulting from weaker 2H shipments coupled with continued Windows 7 adoption could drive upside to our consumer unit growth estimate of 15% for 2011. If consumer units grew 16% y/y in 2011 and the corporate refresh rate increased to the 2006-2008 average of 24%, we could see approximately 300bps of upside to our 2011 PC unit shipment estimate Servers. In 2010, server units increased by 26% y/y and we forecast 14% unit growth for 2011. We believe servers remain well positioned to benefit from the continued acceleration of IT spending due to the continued adoption of the power efficient Nehalem architecture based Xeon’s which offer significant ROI improvements. Our analysis suggests a 3-6 month ROI on for servers > 2 years old.

What is not so well understood by Street is that structurally PC’s (especially notebooks) are entering the accelerated part of the “S” curve of adoption. We believe the market continues to underestimate price elasticity and the fact that PCs are entering the accelerated or fat part of the “s” curve in adoption. The introduction and widespread popularity of net books at $300 and notebooks at $400 in 2008 and 2009 spurred a shift from single PC/households in developed markets to a PC/person model similar to the transition that occurred from single land line/household to cellular phone/person earlier this decade. In addition, the low prices drove rapid adoption in developing markets where first time buyers found low-end PCs within their budgets. Our analysis indicates that in the US, notebook consumer PC penetration has tripled in the last 4 years – from 10% in 2006 to 30% in 2010. We believe this is because notebook PCs are entering the accelerated part of the adoption “S” curve. Similarly, we expect to see consumer notebook penetration rates in the US double and hit 60% by 2013. Interestingly, globally PCs are 21% penetrated today; approximately where cell phones were in 2002. Over the next 4 years, largely as a function of handsets entering the accelerated part of the “S” curve, the cellular phone penetration rate hit 50% -- handset units grew at a 22% CAGR from 2002-07. We believe that most investors are missing the structural growth in PCs over the next 3-5 years from PCs entering the sweet-spot of the global penetration curve.
Microprocessors (MPU) or the “Central Processing Unit” (CPU) is a general purpose computing engine (commonly referred to as the brains of the PC) which executes instructions on the computer and accounts for 79% of PC semiconductor revenues and 12% of semi industry revenues. Within MPUs, the largest segment is the x86 instruction standards used in PCs which accounts for 91% of the industry revenues. The other 9% represents embedded MPUs used in retail point of sale terminals, Industrial, Military etc.

2011 MPU Outlook
We forecast MPU revenues to increase 9% y/y in 2011 driven by unit growth of +11% y/y and an ASP decrease of -2% y/y. As discussed in the prior section we expect unit growth to be driven by both cyclical (PC refresh rate, servers, Windows 7) and structural (increasing PC penetration from entering the accelerated part of the “S” curve) drivers. We are forecasting notebook units to grow the most (+21%), followed by servers (+14%) and lastly by desktops (-2%) in 2011. We expect the ASP decline to be better than normal (~-6%) primarily due to higher MPU$ content as graphics functionality moves into the MPU, resulting in somewhat modest GPU growth (+7% y/y).
In 2011, we expect continued strength in enterprise IT spending, a Windows 7 PC refresh, and an aging corporate desktop/notebook to drive corporate PC unit growth of 13%. On the consumer front, we expect renewed strength after a weaker 2H10, driven by increasing penetration of notebook PCs and strong emerging markets and project 15% growth in 2011.
In 2010, notebooks MPUs showed fairly robust unit growth of 22% y/y, as a pickup in enterprise IT spending and a corporate refresh offset weaker 2H consumer demand in developed markets. In 2011, we expect notebooks MPU units to remain strong, growing 21% y/y. In desktops, post an 8% increase in units in 2010, we expect units to decline 2% y/y in 2011.

MPU Market Shares Should Remain Relatively Unchanged in 2010
The x86 architecture is the standard used in PCs and is essentially an evolution of the first microprocessor, the 8086 introduced by Intel in 1978. There are two primary players: Intel (currently 81% market share on a unit basis) and AMD (18% share); with one other minor participant, VIA accounting for < 1% of the market. AMD began the ramp of its 40nm Brazo’s platform single chip fusion (MPU + GPU) products in late 2010 and expects 32nm Llano shipments for mainstream NBs in mid-2011. Despite Intel’s official launch of Sandybridge (32nm “Tock”) featuring single chip MPU and Graphics at CES, we expect AMD’s ramp to result in a more competitive NB line-up and could drive modest upside to its current NB share. In desktops, we believe INTC is likely to hold it’s current market share of 73%. MPU Mix Will Continue to Shift To Notebooks Beginning in 2004, driven by Intel’s launch of the Centrino notebook platform which featured integrated WiFi and the ubiquity of 802.11 wireless technology, the PC mix began to shift from desktops to notebooks. In 2004, notebooks represented 20% of the PC mix and by 2008-09, notebooks grew to represent 50% of the MPU client mix (i.e. ex-servers). In 2011, we expect the mix shift from desktops to notebooks to continue, with notebooks to hit 64% of the mix exiting 2011. Server Outlook
In 2010, as a result of the cyclical recovery in enterprise/IT spending, x86 server units are likely to grow by 26% (peak-to-trough server units contracted 34% during this downturn). Server attach rates to clients averaged 5.0% from 2005-08 and dropped to 3.8% in 2009 before improving to 4.1% in 2010. In 2011, we expect the continued cyclical recovery in IT spending to drive relative server strength with units +14% y/y. In addition, Intel’s Nehalem, a new MPU architecture that was introduced in early 2009 specifically for the server market, is gaining significant traction on two fronts – (1) TAM expansion into network and storage servers at CSCO/EMC as x86 grows its share of server industry revenues, (2) Compelling server upgrade value proposition – Nehalem offers significant ROI improvements with recent channel checks indicating a 3 month ROI on servers > 2 years old. After peaking at 23% in 2Q06, AMD’s market share in 2P servers has declined steadily and was only 7% in 3Q10. While AMD expected to gain market share in 2010 with the 8/12 core Maranello in high end and 4/6 core San Marino in low end, INTC’s Westmere EP continued to hold share. In 2011, we believe AMD’s introduction of the Bulldozer 32nm SOI processor with a Nehalem-like modular approach and 4/8/12/16 core variants could be a more competitive product. We expect AMD’s position in servers to remain challenged until the 2H11 introduction of Bulldozer, but ultimately exit 2011 with 13% market share in DP Servers. In 1Q10, INTC introduced its Nehalem EX geared towards higher-end (i.e. $10,000+) Multi-Processor servers. MP server MPUs have ASPs of $940 with higher gross margins relative to the server average of $512. In 2011, we expect INTC to improve its MP market share from current 85% to ~100% exiting the year.

Discrete Graphics
2011 Outlook
After a 20% increase in discrete GPU revenues in 2010, we expect a slow down and are forecasting revenue growth of +2% in 2011. We expect revenue growth of +3% in notebooks and 2% in desktops with strong unit growth in notebooks (+7%) and desktops (+7%) and ASPs down -4% and -5%, respectively. In 2011, we believe that GPUs are likely to benefit from i) a continued rebound in professional workstations from higher IT spending and ii) growth in software applications that use GPUs for non-graphics computing tasks After 7 years of a steady decline in discrete graphics attach rates in PCs, we expect a small uptick in discrete graphics in 2011 for three reasons Consumer PC units, with higher discrete GPU attach rates, will out grow Corporate PC units, as Windows 7 drives consumer PC sales.
11 January 2011
Atom/Netbooks continue to decline as a percent of the notebook mix due to tablet cannibalization. The increasing prevalence of software that showcases GPU functionality.
2011 Outlook
Chipset revenues grew 13% y/y in 2010, on units +18% and ASPs -4%. In 2011, we expect units to grow in-line with MPUs (+11% y/y); however we expect a 4% decline in ASPs as core logic components continue to move into the MPU with Intel’s ramp; we forecast chipset revenue growth of 17%.

Chipsets Begin To Move To Leading Edge in 2010
Beginning with the 32nm Westmere product generation – Arrandale (notebooks) and Clarkdale (desktops) in January 2010, Intel began to move the graphics logic (controllers, drivers) from the Northbridge chipset into the MPU as it chose to use the higher transistor densities at 32nm for greater product integration instead of higher performance. Intel believes that these more integrated products will be better suited to target the low-end and mainstream PC market from a more cost appropriate basis. By combining these two changes, Intel was able to move from a 3 chip solution for PCs to a 2 chip solution, which i) Increased performance by lowering the latency to access DRAM memory and graphics, ii) decreased motherboard area to target thinner PC form factors and iii) lowered PC build costs by reducing motherboard complexity. Westmere products actually combine two dies – a 32nm MPU and a 45nm graphics logic (controller/drivers) in a single chip package. With Westmere, graphics are still one process generation behind MPUs and do not yield all the area and cost savings of a fully integrated 32nm die. However, Sandy Bridge, which launched last week at CES, combines the two chips into a single monolithic die. AMD plans to introduce the first 32nm fusion products combining CPU and GPU into a single chip in the middle of 2011.

We estimate that memory revenues will increase by 5.9% in 2011, below the semiconductor industry’s 9.2%. In DRAM we expect to see a weaker pricing environment as our company-by-company bottom’s up 55% bit growth for 2011 outpaces our bit demand growth estimate of 48%, causing ASPs to fall 35% y/y (historic norm is -30%). Relative to consensus our view on DRAM remains positive as we believe supply growth estimates could be challenged as we see very little wafer growth and shrinks to the next process node could prove to be more difficult than anticipated. Furthermore there are three key demand drivers still intact, i) higher mix of dual channel processors which lend themselves to 4GB+ per box, ii) PC refresh, and iii) Nehalem servers driving higher memory content in servers. Relative to NAND, we see strong application demand growth driven by Tablets and Smartphone pushing bit shipment growth of +90% vs. our bottom up shipment growth estimate of 87% in 2011. MU remains our top pick in memory with Outperform rating and $10 price target. We believe MU’s near-term headwinds resulting from inventory write-ups related to the Numonyx acquisition, incremental start-up costs as MU moves forward with IMFS without INTC, and a slower Inotera ramp will soon become tailwinds driving upside potential to consensus estimates. The added benefit that Inotera DRAM will likely become Sever qualified by MayQ could driver further upside.

Revenue Model
The below table details our bottom-up DRAM supply and demand assumptions. For 2010 we expect 55% bit supply growth, driven by 47% improvement in bits/wafer from 4xnm Semiconductors 11 January 2011 shrinks and 5% higher wafer starts. Following a 3% capacity increase in 2010, we expect another modest DRAM capacity growth of 3% in 2011 as the majority of the $7.2bn in DRAM capex (-38% y/y) is expected to be directed to 4xnm shrinks.The below exhibit details our bottom-up NAND supply and demand assumptions. For 2010we expect 87% bit supply growth, driven by 44% improvement in bits/wafer from 2xnm shrinks and 30% higher wafer starts, primarily driven by a capacity increase of 29%. We model NAND capex of $12.4bn (+51% y/y) following 78% growth in 2010 as suppliers work to add capacity and move to the 2xnm simultaneously.
Key forecast drivers and trends Cyclical Bit Growth Still Below Trendline in 2011 While the 2011 bit supply growth may appear high in absolute terms, we would note that it is still below historical trends – the average DRAM bit growth was 57% y/y from 2001-08 period, or 2 percentage points above our estimate of 55% (see below two exhibits). In NAND, bit growth has come down from 250% in 2005 as the industry has increased scale. Our estimates suggest 87% bit supply growth in 2011, which is still significantly lower than the 182% average for 2001-08.

DRAM Demand Upside from PC Cycle
Our base case DRAM bit demand growth estimate is 55% in 2011 but we believe the proliferation of dual channel processors and a better than expected PC cycle could drive upside to our published estimates. Total DRAM bit demand would grow 62% y/y in 2011 if growth in PC units and MB/system is 18%/40% y/y vs. our base case estimates of 14%/26%, respectively and growth in server units and MB/system is 18%/90% y/y vs. our estimate of 8%/44%. Note that our upside scenario estimates assume PC unit growth at 100-200bps below seasonal and server unit growth inline with seasonal.
We believe the upside in DRAM demand can be driven by the following:
PC Unit Growth Drivers. We see several key drivers for PC units in 2011 i) continued rebound in enterprise IT spending, ii) an aging corporate desktop (~5yrs) and notebook (~4yrs) installed base and iii) accelerating growth in the global middle class. Note, we typically see an inflection point in PC sales when the price hits 8 weeks of income, a point that China recently surpassed and India is quickly approaching. Re-acceleration of PC MB/system. In 2009 and 2010 we saw a deceleration in MB/system for PCs as DRAM amounted to 10-12% of the total BOM. In 2011, we expect to see a re-acceleration of MB/system because 1) recent DRAM price declines will allow price elasticity to kick in and 2) the mix of dual channel processors as a percent of PCs will increase with the ramp of INTC’s Sandybridge and AMD’s Fusion products. We believe that dual channel processors will drive MB/System over 4000MB, as 3000MB does not lend itself to the dual channel and 2000MB is not sufficient. Server upgrades and higher MB/system facilitated by Nehalem CPUs. The continued adoption of INTC’s NehalemX, which has memory bandwidth of up to 64 GB/second - a 3x improvement from prior generation - Xeon 5500 CPUs could drive server DRAM content from current 18 GB/system to 24 GB/system, still leaving large headroom to the 144 GB/system limit. Note that servers account for 12% of DRAM.
Strong NAND Demand Drivers
We expect NAND demand bit growth of 66% in 2010 driven by Smartphones and model bit demand growth of 90% in 2011 as the continued strength of Smartphones is complimented by the proliferation of Tablets which we believe will represent more than 10% of total NAND demand in 2011. Optically, our growth estimates appear to be high, but we would note that secular drivers are strong and our 2011 estimate is still well below the 2001-2008 average of 18% y/y growth.
We estimate NAND consumption in handsets will grow 118% y/y in 2011, as total NAND demand grows by 90%. This assumes 17% handset unit growth in 2011 and 33% Smartphone unit growth. Additionally, these estimates reflect only attached content at the time of shipment and do not include aftermarket upgrades through flash cards, which represent 30-35% of total NAND consumption. Consequently, we believe our estimate of NAND consumption by handsets (12% in 2008, 15% in 2009, 21% in 2010 and 24% in 2011) understate total impact of handsets on NAND demand SSDs. We believe SSDs represent the strongest long term NAND demand driver. Tablets should drive near term strength, as Notebook SSDs suffer from a lack of price competitiveness vs. HDDs. In the long-term we expect thin form factors and decreasing NAND ASPs to drive Notebook SSD strength. We estimate NAND consumption in SSDs will grow 151% y/y in 2011. This assumes 51m Tablet unit shipments with 55GB/Tablet.

Our top-down analysis suggests that we will see limited wafer start growth for DRAM in 2011 as suppliers direct more capex towards shrinks to the next process node. We model installed capacity growth of only 3% y/y, as there is little room for increased utilization as rates are already at 95%+. Relative to NAND, we see a more balanced mix of capex driving installed capacity growth of 29% y/y in 2011 driving wafer starts up 30% y/y.New DRAM Capacity Capex to Remain Depressed in 2011. We expect DRAM capex to decline by 38% y/y in 2011 to $7.2bn, with the majority of the depressed
2011 spend dedicated to 4xnm shrinks, as shrinks tend to offer significantly higher return per dollar of capex. Conversely, we expect NAND capex to increase by 51% y/y in 2011 to $12.4bn. In NAND we expect utilization rates of 100% to drive a more even capex split between installed capacity growth and shrinks to the 2xnm process node.
Communication ICs Bottom-line
We expect Comm IC revenue growth of 10.5% in 2011, modestly ahead of the 9.2% growth in the overall semiconductor market, on increasing Comms and smartphone semiconductor content (CS estimates 32% smartphone unit growth y/y), and continued growth in emerging market wireless capex. We model wireless comm IC revenues +12.6% y/y, and Wireline revenues +4.4% y/y. Our top picks in Comms remain Outperform-rated BRCM, MRVL and AVGO, while we have Neutral ratings on MLNX, ENTR and ANAD.
Revenue Model
We define comm ICs to include wireline comm ICs (25% of revenues in 2010) and wireless comm ICs (75% of revenues). In 2010, total comm IC revenues grew 25% (after declining 9.4% in 2009), with wireline comm ICs gained 29.7% while wireless grew 23.5% y/y.
2010 Top’s Down vs. Bottom’s Up
The below exhibit outlines our bottom up comm IC estimates, which indicates 11.4% y/y growth in 2011 and 9.5% y/y in 2012, vs. our tops down estimates of 10.5% and 11.1%.
Key forecast drivers and trends
Our tops down comm IC estimates include the following key components:
Smartphone Growth. We continue to see handset unit growth supported by smartphones as well as increasing handset penetration in emerging economies. We model 32% unit growth in smartphones in 2011 as we see this category in the sweet spot of the penetration S-curve, rising from ~20% penetration in 2010 to 41% in 2012. In addition, we believe device-level integration trends are driving strong adoption of combo chips, favoring vendors with broad technology portfolios and potentially facilitating M&A in the supplier landscape. We expect combinations of single-product companies such as CSR/SIRF, ATHR/ITLN and SIGM/Coppergate to remain a key theme but we do not expect a meaningful impact on competition until 2011-2012 at the earliest. In combo chips, we expect market traction to double the handset attach rate from 23% in 2009 to 50% in 2012, creating a $2bn market opportunity in 2011. Product Cycles. We have identified twelve product cycles in comm ICs, such as FTTH, MoCA and 10G Ethernet, which collectively represent $18bn of revenues, growing at 18% CAGR. However, we believe there are few clear winners among companies who benefit from these product cycles and we favor BRCM with its broad positioning in virtually all key product cycles Semiconductors 11 January 2011 Handset Penetration S-Curve Revisited – Smartphones & Emerging Markets Wireless comm ICs grew at 15.8% CAGR from 2002 to 2008, 7 percentage points above the semiconductor industry CAGR of 8.8%, led primarily by the increasing penetration of wireless handsets, particularly in developed markets. With handset penetration in developed markets reaching beyond 70%, the slope of the penetration S-curve has declined, indicating deceleration of handset unit growth. However, we expect renewed Growth momentum in handsets, driven by smartphones as well as increasing penetration in developing countries. As shown in the exhibits below, smartphones are entering the sweet spot of the Penetration S-curve and we expect global smartphone penetration to double from 20% of addressable market in 2010 to 41% in 2012. We estimate that this high rate of smartphone penetration drive 25%-35% smartphone unit growth and expect to see 32% unit growth in 2011. Handset penetration in developing markets is also increasing, and we estimate penetration rates to increase by 35-40% in India and China, for example, over the period from 2009 to 2012. We currently model 15% y/y growth for handset units in 2011.
In 2011 we expect Wireless Comm IC revenues to grow by 12.6%, above the 9.2% tops down growth of the semiconductor industry. For Wireless Handsets we estimate 13.4% revenue growth in 2011, +3.3% for Wireless LAN and +3.0% for Wireless Infrastructure. Beyond 2011, we expect handsets to remain a key growth engine for the semiconductor industry and expect that in 2012 Wireless Comm ICs will grow 13.0%, slightly above the 10.7% growth we estimate for the semiconductor industry. Note that the 2002-08 CAGR for Wireless Comm ICs is 15.8%, although it was helped in great part by the increasing penetration of handsets through the sweet-spot of the S-curve.

We expect PLD revenues to grow by +9.2% in 2011, in-line with overall semiconductor revenue growth of 9.2% y/y. We continue to see signs of potential acceleration of PLD share gains vs. ASICs during the upturn, that could add 300bps of growth to our 2010-13 PLD revenue CAGR of 9.5%. From competitive perspective, we see ALTR improving its market share from current 41% (up from 36.6% in 2009) towards its target 60% over the next few years.

Bottom Line. We are currently estimating Analog revenue growth in-line with to modestly above the overall Semiconductor industry for both 2011E and 2012E, with 2011 revenues +9.5% y/y, 30bps above overall Semiconductor revenues +9.2% y/y, and 2011 revenues +10.8% y/y, in-line with overall Semiconductor revenues +10.7%. Our revenue estimates assume an increasingly competitive pricing environment in 2011 (pricing down 2.4% y/y, vs. -1.3% in 2010, where pricing benefited from capacity constraints) as TXN begins to ramp its 300mm analog capacity, offset by a shift towards higher dollar content applications.
A bottom-up aggregation of our analog coverage universe would suggest revenue growth of +5.0% in 2010 and +4.0% In 2011, suggesting upside to current revenue estimates. Our top picks in Analog include TXN, MXIM, and LLTC, while we continue avoid mid to lower - tier names which we continue to view as vulnerable to TXN’s 300mm ramp, includng NSM, ISIL, ONNN, FCS.
Semiconductor Content Growth
We also believe that analog is well-positioned to benefit from increasing semiconductor content, particularly in handsets, base stations, autos, industrials, and green which are seeing significant content increase. Our content increase analysis suggests cyclical upside, with units potentially still 15% below a new structural trend line. We also believe Analog companies with higher exposure to Comms, Autos, and Industrials will see a disproportionate benefit from increasing content exposure.

Analog Stock Picks
Our top picks in Analog include TXN, MXIM, and LLTC, while we continue avoid mid to lower-tier names which we continue to view as vulnerable to TXN’s 300mm ramp, including NSM, ISIL, ONNN, FCS. We would also note that ADI, LLTC and MXIM screen well for end markets exposed to increasing content, which include Comms, Auto, and Industrials within Analog.
Analog Market Outlook and Competitive Environment
The analog semiconductor industry is a $42.2bn industry (estimated 2010 revenues), representing 14.0% of the total semiconductor industry at $300.8bn in 2010. Analog industry revenues have grown at a 5.9% CAGR post bubble (2003-2008), versus overall semis which grew at a 8.4% CAGR over the same period, and we estimate that analog revenues grew 31.9% y/y in 2010 vs. overall Semis +32.9%, recovering from its -10.2% y/y decline in 2009 vs. overall semis down -9.0%.
Analog industry revenues generally grow with close correlation to overall semis (96% correlation over the last 10 years, and 90% correlation over the last 20 years). We highlight analog revenues relative to total semiconductor revenues below; since 1990, analog revenues have represented 14.5% of total semiconductor revenues, roughly in-line with 2010 at 14.0% of total analog revenues.

Analog companies are characterized by high gross margins averaging 60%, 10 points above the average for the overall semiconductor industry at ~50%, attributable to diverse end markets and a fragmented customer base typically with tens of thousands of customers, and little depreciation expense. Analog engineering talent is scarce tends to be scarce relative to digital engineering talent, and design expertise, proprietary process technologies, and large IP portfolios are vital to the success of analog companies. We model analog industry revenues broken down by similar categories as published by WSTS – general purpose analog, and application specific analog, as we describe in further detail below.
General Purpose Analog
General purpose analog represents 42.2% of total analog at an estimated $17.8bn in revenues in 2010. General purpose analog is typically segmented by product group - converters, power management, amplifiers, and interface products, with different share dynamics within each product category. We estimate general purpose analog revenues will grow 9.3% in 2011 and 10.7% in 2012, after growing 37.2% in 2010.Below, we illustrate market share trends according to Gartner estimates over the past five years; as evident, TXN continues to dominate the market, growing its leading share from 14.0 by 176bps to 15.7% (+64bps y/y). We also highlight the fragmented nature of the market, with the top 5 companies representing 47.8% of the market, down from 51.8% in 2005. Other notable share shifts include NSM, which lost 336bps of market share from 2005-2009, and 133bps y/y.
Data converters represented 17% of general purpose analog revenues in 2010 at $3.0bn, and are used to change analog to digital signals or vice versa quickly and accurately. Data converters are a difficult market to penetrate, as they require a wide range of products for varying speed, performance and accuracy, and are sold in small lots to thousands of customers across most end-markets. It takes years to develop a broad catalog, and incumbents have the advantage that product developers are comfortable with their products. We estimate that converters will grow 7.1% and 10.7% in 2011 and 2012 after growing 29% in 2010 (versus overall general purpose analog growth of 37.2%) with growth driven by communications infrastructure buildouts and industrials-related demand.
Analog Devices has historically been the market share leader in this segment, holding 40.6% of the market in 2009 (vs. share peak of 44.7% in 2004 and share trough of 27.5% in 2007), more than double its closest competitor TI at 17%. However, TI has been steadily gaining share in converters, increasing its share by 410bps since 2003.
Power Management
Power management analog represents the largest sub-segment of general purpose analog, at an estimated $9.3bn in revenue in 2010, or 52.2% of total general purpose revenues. Power management semiconductors deliver power and control the flow of electrical energy among the various power loads and energy sources in a product or system. This category of product consists primarily of linear regulators, which drops voltage (also called an LDO – low drop-out regulator) and switch-mode regulators, which convert voltage up or down within an electronic system for power management and battery charging. Voltage references also serve as electronic benchmarks providing a constant voltage output independent of input, temperature changes or time. We estimate power management to be the fastest growing segment within general purpose analog with revenues up 10.9% in 2011 and 11.3% in 2012, following 2010’s estimated 40.9% growth, better than overall general purpose revenue growth of 37.2% - driven by continued demands for precise and efficient energy solutions across end Markets offset by pricing declines in consumer applications.TI is the dominant player in power management, steadily increasing its share of the market to 14.7% in 2009 from 11.6% in 2003, overtaking prior leader NSM in 2006 as NSM moved away from handset related power management and other higher volume lower margin segments. MXIM has also since exceeded NSM’s share of the market, growing its share from 8.1% to 8.7% in 2009 y/y, while NSM fell from 9.9% of the market to 8.2%.
Amplifier & Comparators
In 2009, SIA adjusted its general purpose categories to include amplifiers & comparators (roughly 1% of general purpose revenues) within the same category. We estimate amplifiers and converters represented $3.1bn in revenues or 17.4% of the general purpose analog market in 2010. Amplifiers are devices that increase the power of a signal, and include operational amplifiers, which amplify signals for applications such as video, fast data acquisition and communications, and audio amplifier which reproduce audio frequencies and can be used to drive speakers. We estimate am plifier growth of 9.7% in 2011, and 10.7% in 2012, vs. overall General Purpose analog growth of 9.3% and 10.7%. This follows growth of 40.5% in 2010, vs. overall General Purpose growth of 37.2% According to Gartner, the general purpose amplifier market is also led by TXN at 18.9%, which took the number one position in this market from ADI in 2004 (ADI is now a close second at 17.3%). As with power management, NSM has also seen its share of the amplifier market decrease steadily to 11.1% in 2009 from 15.1% in 2003.InterfaceInterface products represented the smallest sub-segment of the general purpose market, representing an estimated $2.4bn in revenues in 2010, or 13.4% of the general purpose market. Interfaces switch an electrical signal between devices or communications channels. For example, an interface product can switch an external connector on an appliance between a USB interface function and an audio input/output function. Typically, interface devices require high-speed, high-current drive and low-noise attributes. These types of products are mixed-signal in nature and require a high level of analog wave shaping techniques on the output structures, and so are difficult for new entrants without deep engineering resources to penetrate.
We estimate that interface will represent the slowest growing sub-segment of general purpose analog, growing at 5.3% and 8.1% in 2011 and 2012, versus overall general purpose at 9.3% and 10.7% respectively. In 2010, we estimate the segment grew 30.7%, below overall general purpose +37.2%.
According to Gartner, TXN holds 26.2% of the interface market, with MXIM 10 points behind with 16.6% share. NXP has grown its share of this market from zero in 2005 to the third largest player at 12.3% in 2009.
Application Specific
The second broad analog category is application specific analog which we estimate generated $24.4bn in revenues in 2010, or 57.8 of total analog. Application specific analog products are designed for specific end-market applications and may combine multiple functions on a single device. This market can be characterized by higher volumes and lower margins relative to general purpose analog devices. SIA breaks application-specific analog devices into computing (estimated 14.3% of application specific analog market), comms (45.1%), consumer (12.5%), automotive (18.9%), and industrials and other (9.2%). We estimate application specific analog revenues will grow 9.6% in 2011 and 10.9% in 2012, after growing 28.3% in 2010 (vs. overall Analog +31.9%).
We forecast Microcontrollers (MCU) to exhibit growth approximately in-line with the semiconductor industry in both 2010E and 2011E. Within MCUs we project 32-bit (use of industry architecture standards vs. proprietary standards is resulting in faster ramp, better platform performance and capabilities in hardware and software, large competitors) to out grow the industry in 2010 and 2011, while 8-bit (share gains from 16-bit, move up the performance stack, touch screen) grows in-line with the industry and 16-bit to under grows (share losses to both 8- and 32-bit) the industry. We would be selective and focus on companies with strong cyclical exposure to autos and industrials (MCHP), companies with new product cycles (CY) and companies with a broad and well diversified product portfolio (TXN).
Post a -21% decline in 2009, we expect MCU revenues to grow by +34% in 2010 to $14.6bn. We expect growth to be driven by a strong cyclical rebound in Automobiles (32% of MCU revenues), industrials (18% of revs) and by several product cycles, the most prominent of which is touch screens, albeit its impact (as % of revenues) will remain relatively small in 2010-11.

Segments Growth Rates
Microcontrollers are small computers-on-chip consisting of a relatively simple microprocessor combined with support functions, such as timers, analog, I/O, and memory (DRAM and NAND) and are typically segmented by the complexity of the microprocessor which is the core of the chip. 8-bit MCUs (which sell for $0.70 to $1.00 on average) are Semiconductors
11 January 2011 used in products with relatively simple performance requirements such as consumer white goods (toasters, refrigerators, microwaves), automobile control systems etc. and 32-bit MCUs (which sell for $3-$4 on average) are used in more complex products such as automobile infotainment systems, GPS, retail point-of-sale terminals etc, and 16-bit tends to occupy the middle ground between the 8-bit in the low-end and 32-bit in the high-end. 8-Bit (38% of MCU revenues). Project growth of +10% y/y in 2011, approximately in- line with the market from increasing adoption of TouchScreen in handsets, Tablets, medical and industrial devices, higher logic performance and peripheral (especially
analog) capabilities. 16-Bit (26% of MCU revenues). Expect it to under perform the industry with +9% y/y
growth in 2011 as it gets squeezed between the higher performance 32-bit MCU with (ASPs $3-4) and lower-end 8-bit MCU (ASPs ~$1). 32-Bit (35% of MCU revenues). Expect modest outperformance relative to the industry with +14% y/y growth in 2011 from the continued adoption of industry architecture standards such as ARM and MIPS vs. largely proprietary standards in 8- bit and 16-bit and from better platform performance and capabilities in hardware and software. In particular, ARM has been fairly aggressive in attempting to establish its 32-bit architecture standard in embedded applications, similar to its success in the handset market. Several leading semi companies – Texas Instruments, Atmel, Renesas, Infineon, S TMicro offer ARM based MCUs.

Market Share to Continue To Change in 2011
Microcontrollers were a $14.6bn industry and represented 5% of semi industry revenues in 2010. Similar to the analog industry, MCUs are also characterized by a highly fragmented market with very gradual market share shifts that play out over an extended period of time. MCUs have fairly long design cycles (1-2 years), so when customers prototype an MCU it is typically followed by relatively long product-life cycles which range from 1 to 8 years with an avg. of 3-4 years vs. the industry average of < 2 years in PCs, handsets and consumer electronics end-markets which represent 80% of semi industry revenues.
However because several players were distracted with concerns around improving weakened balance sheets as a result of the credit crisis (Freescale, NXP in this camp), restructuring (STMicro, Infineon) and mergers (Renesas-NEC), we believe 2010 was characterized by relatively significant share shifts and that 2011 play out in much the same way.
Within our coverage, we expect CHP to solidify its leading position in 8-bit and increase its market share in 16-bit; CY to continue to gain share in 8-bit from its TrueTouch.
Touch Screen MicrocontrollersWhile touch screen controllers, which typically use an 8-bit MCU core, will represent only a small portion of total MCU revenues, the segment has garnered significant investor attention ever since Apple introduced the Technology with the iPhone in 2007. Our analysis suggests that the adoption of touch screens in mobile handsets continues to increase rapidly and while it is a feature primarily found in high-end smartphones today, it is beginning to become more popular in mid-range handsets. The touch screen use in handsets today stands at 260mn units or ~16% attach rate. By 2011 we project this to increase to 393mn units with a ~21% attach rate. The increasing penetration of touch in handsets, and the introduction of tablets (~100% attach rate) in 2010 and their expected proliferation in 2011 (we estimate 55mn units vs. 19mn in 2010) could drive an incremental 266mn (+40% y/y) Touch units in 2011 Prior to 2009, the touch screen controller market was module driven – where a single vertically integrated supplier (notably Synaptics) integrated the panel, the ASIC controller, software, firmware, testing and validation into a single module which sold for $20-$30.
While this integrated module approach offered time-to-market advantages, it resulted in both higher costs (particularly important now that touch technology is beginning to enter into mainstream handsets) and reduced ability the ability of handset OEM to customize products. Cypress entered the touch screen market in late 2008 with TrueTouch, a product that married two existing products – 8-bit PSoC microcontrollers with CapSense Plus, which was used as a mechanical button replacement & slider in GPS systems, PoS terminals, automobiles & PC mice. More importantly, instead of adopting a modular approach, CY began to sells just the touch screen controller chip integrated with PSoC for $2-$4. This allowed the O EM the flexibility to customize the touch screen panel (screen size, resolution, features), and software (O/S, applications, gesture recognition, firmware). Cypress also solely focused on providing capacitive touch sensing which improved Semiconductors 11 January 2011 durability, finger touch (vs. stylus), multi-touch functionality (vs. single touch prior) and better sensing clarity over prior generation resistive touch screen solutions. Since Cypress entered the touch screen market, 3 other US based competitors – Atmel, Silicon Labs, and Microchip have all announced capacitive touch screen solutions with touch screen controllers using MCUs, similar to the Cypress approach. However, based on reported design-wins at the top handset OEMs we believe Cypress took over the leading position in handset ICs from Synaptics in 2010 and will maintain that position in 2011.

General Expectations for 2011 Our current estimate for 2010 revenue growth for our coverage universe is 33.2% y/y, inline with the consensus estimate of +33.2%. Ex-Intel, our revenue growth expectation is +40.4%, in-line with consensus at 40.6%. For 2011, our estimates imply a revenue increase of 5.3% (5.5% ex-Intel), consensus is looking for revenue growth of 4.0% (4.3% ex-Intel).


TECHSEMIGURU is a blog on technology investing written by Farhan Khan. The blog presents news and innovative reporting on events vital to investors in Semiconductor linked fields. Remarks and tips can be sent to