VITESSE SEMICONDUCTOR CORP(VTSS)*
Rating: 1H
05/05/1999
Salomon Smith Barney ~ April 30, 1999
04/30/99 ( $,-,Tgt $) Clark Westmont --SUMMARY:----Semiconductors
*This represents the first in a series of reports analyzing the forces of change in the semiconductor
industry. Alternating between issues facing traditional semiconductor suppliers and trends
influencing the more-focused communications component companies, we will explore topical
subjects of interest to companies and investors alike.*Our 'Shift Happens' thesis is two-fold:1) the
focus of the semiconductor industry is shifting from computer applications to communications
applications, and 2) the communications industry is transforming from a vertically integrated
structure to a horizontally specialized form.*One of the clearest illustrations of this thesis is in the
Internet IC segment. By our definition, Internet ICs are chips that can be used only in broadband
communications applications (e.g., SONET, ATM, DWDM, T/E, and Ethernet).
04/30/99 ( $,-,Tgt $) Clark Westmont
--OPINION-------------------------------------------------------------------
A specialized group of chip suppliers defied the 1998 semiconductor industry recession, posting
growth of 100% (on average and collectively). This group, the Internet IC suppliers, are uniquely
positioned to benefit from the build-out of the Internet infrastructure. In many ways, they may
represent the future of the semiconductor industry. The 'Shift Happens' thesis in action Two years
ago, we authored a report titled "Shift Happens - The Structural Realignment of the
Communications Industry". In it, we described a fundamental secular shift in the communications
industry away from vertical integration and toward horizontal specialization. This was a fancy way
of saying that the trend toward out-sourcing (which had already transformed the computer
industry) was beginning to manifest itself in the communications industry. This trend posed unique
opportunities for component suppliers, in our view. Last year, we updated our thesis to focus on
the Internet, describing a distinctive set of chip suppliers we dubbed the 'Internet IC' companies.
The Internet IC suppliers benefit from both the 'Shift Happens' thesis and the extraordinary growth
in the Internet. As such, they have attracted significant attention and may be a portent of the future
of the semiconductor industry. Now, virtually every chip supplier on the planet is scrambling to
target Internet applications. As the semiconductor industry becomes increasingly focused upon
communications, the industry itself will likely morph into a form unrecognizable to its fathers. This
transformation creates challenges (and opportunities) for companies and investors alike. To
analyze these forces of change, we will publish an on-going series of reports under the 'Shift
Happens' moniker. In this, the first of the series, we review the dynamics of the Internet IC group.
Chips.com? Introduction Now that it is evident that the Web is not CB radio with more typing,
investors have heatedly sought to mine the Internet mother lode. One area that is still dawning
upon the collective investor consciousness is the Internet IC group. While the valuations reflect
much credit for the tremendous growth delivered by the group, the actual fundamentals are
somewhat obscured by a mind-bending alphabet soup of acronyms like ATM, SONET, DWDM,
T/E, xDSL and GbE. In concept, however, the Internet IC concept is very straightforward: chips
are part of the Internet food chain, which spans content, carriers, equipment, contract
manufacturers, and component companies. Nothin' but 'Net The Internet IC companies supply
chips, but have largely escaped much of the turmoil sweeping the semiconductor industry over the
last year. Specifically, sales of the seven publicly-traded Internet IC suppliers doubled last year,
while the semiconductor industry at large shrank by roughly 10%. The key reasons for this
difference are that the Internet IC suppliers have essentially no PC exposure and are generally not
capital intensive (in contrast to microprocessor and memory companies). Most importantly, these
suppliers are focused upon broadband data applications that are driven by the explosion in
Internet traffic. By our definition, Internet ICs are chips that can be used only in broadband
communications applications (e.g., SONET, ATM, DWDM, T/E, and Ethernet). The Internet IC
name is intentionally provocative, but the reality is that companies supplying components for
broadband data communications are enjoying a demand environment that can only be explained
by the growth in the Internet. Figure 1 lists the seven companies we incorporate in this category,
including recent sales growth trends and projections for the year. There are several things worth
noting in this table. First, the sales growth rates for 1997 and 1998 by any measure (average,
median, or group total) were remarkable - all the more so since the semiconductor industry (of
which these companies nominally are a part) contracted last year by 9-10%. The most important
investment consideration is not history, of course, but forward projections. As such, there are two
dimensions to the outlook -- one obvious, one subtle. Most obviously, the 53% median sales
growth outlook for 1999 ranks among the best in technology, driven by the on-going deployment
of broadband data networks. The subtle dimension can be discerned by comparing the sales
growth estimates for 1999 versus the two prior years. Specifically, the 53% median sales growth
hurdle set for 1999 is impressive, but less than the pace set in 1997 and 1998. Realistically, we
see no fundamental reason for any deceleration this year, with the implication that there remains
good upside potential to Street expectations for the Internet IC suppliers in 1999. In fact, six
months ago our projection was for 41% median growth, which was raised three months ago to
48% after the companies reported December 1998 results. After the March earnings season, the
median was bumped to 53%. We would expect the median to continue to step up as we progress
across the year, likely meeting or exce eding last year's pace of 63%. Attractive margin structure
The margins within the Internet IC group are above the semiconductor industry average. The high
gross margins represent both the high value-added they bring to customers as well as the low
capital intensity of the communications IC business. R&D spending also tends to be above
semiconductor industry norms, reflecting the essential systems and device expertise required for
leadership in the marketplace. Name Association Game One way to most easily appreciate the
Internet IC group is to analyze their customer base. As might be anticipated, the customer list
comprises the most recognized names in communications equipment, including Cisco, Lucent,
Tellabs, Nortel, 3Com and Alcatel. Figure 3 cross-references each component company's
exposure to each of these major accounts. Of course, virtually every chip supplier can claim Cisco
and Lucent as customers. In general, there are three broad categories of chips sold into
communications equipment: custom, general purpose, and specialized standard products. Custom
chips (also called ASICs for application-specific integrate circuits) are designed by the original
equipment manufacturer (OEM) and are manufactured exclusively for them by ASIC suppliers
such as LSI Logic, IBM, or NEC. Although the ASIC supplier may supply 'cores' or other
functional building blocks, most of the intellectual property in the chip is contributed by the OEM.
General purpose chips include programmable logic devices (PLDs), memories (e.g., SRAMs and
DRAMs), and RISC microprocessors. For the most part, the same general purpose chips that are
used in communications applications can be used in other markets equally well. As such,
producers of general purpose chips often have little deep understanding of the systems in which
they are used. Again, it is the OEM that provides the critical intellectual property in stitching
general purpose chips together into a finished product. The third category, often labeled
application-specific standard products (ASSPs), are chips developed to implement a particular
system function for sale to any customer. In other words, ASSPs take the functionality that would
otherwise be done with custom ASICs and/or general purpose chips and makes it available
off-the-shelf. When used for broadband communications equipment (i.e., anything transmitting
more than a million bits per second), we define these as Internet ICs. In general, ASSPs save
OEMs' time, engineering resources, opportunity cost, and potentially unit cost. All three categories
of chips are typically used in a communications system, but historically ASICs and general
purpose chips dominated the mix. As the communications market has grown in importance and as
more systems knowledge has been absorbed by chip suppliers, more ASSPs have become
available. In addition, as the time-to-market pressures have become more intense, OEMs have
become increasingly interested in using more off-the-shelf product to shorten their design cycles.
In effect, OEMs are 'out-sourcing' part of the system design. In cases where the differentiation
between competing systems is becoming negligible (such as low-end Ethernet switches), virtually
all of the component content may move toward ASSPs. It is this increased adoption of ASSPs by
communications OEMs that has accelerated the growth rates of the Internet IC suppliers,
propelling them well ahead of the growth of the OEMs themselves. Given that we are still relatively
early in the ASSP adoption cycle, we believe this relative out-performance may persist for several
years. Noveau Niche The communications chip market is not a monolithic entity. Indeed, the
Internet IC suppliers often have little market overlap with each other, which increases the amount
of homework that needs to be done on the group. We believe appreciating the variety within the
market niches is important to understanding the Internet IC group. Internet IC applications can be
loosely segmented into three categories: local area networking (LAN), wide area networking
(WAN), and broadband remote access. Since each category is highly fragmented, we will explore
each in turn. Although the nomenclature might at times be intimidating, we recommend that the
reader simply see the acronyms as labels to put on pigeonholes. The dynamics within each
pigeonhole can be left for deeper analysis later. Category #1: Local Area Networking (LAN)
Typical LAN equipment includes repeaters, switches, routers, and network interface cards.
ASSPs used in LAN gear include interface chips (dubbed 'transceivers', short for
transmitter-receivers) and control ICs such as switch fabrics and media access control (MAC)
chips -- these are typically digital chips that determine the flow of traffic through the network. The
dominant customers for LAN components are Cisco, Nortel (Bay), and 3Com, although
Cabletron, Lucent, Hewlett-Packard, Extreme, Fore Systems (GE), and Xylan (Alcatel) each
have some market presence. Taiwanese-based manufacturers also play an important role in
low-end, high-volume commodity segments such as repeater hubs and workgroup switches.
Similar to the evolution of the PC market, we see the importance of Taiwan increasing as the
private label market gains ground. As the U.S. dollar is to the currency market, Ethernet is to the
LAN market. It follows, then, that the most important LAN components are building blocks for
Ethernet equipment. Up until now, Gigabit Ethernet (GbE, which sends data at a billion bits per
second) has been a very small niche market, but it is growing rapidly. Nevertheless, Fast Ethernet
(i.e., 100 million bits per second) should remain the mainstream for the next few years. Table 1
displays the most prominent players in each of these product segments. A few observations: 1
Qualitatively, the LAN market is much more homogenous than the WAN market, and changes in
LAN architecture are tending to be more incremental than revolutionary. This inherently leads to
less differentiation and more commoditization than the WAN market at both the chip and system
level. 2 WAN component suppliers such as AMCC and Vitesse have little exposure in the LAN.
This helps to illustrate the dramatically different skill sets (e.g., systems knowledge, modulation
expertise) required in each segment. 3 Galileo and MMC are the only companies that currently
derive much revenue from switch fabrics since most systems are still ASIC-based. In contrast,
transceivers are rarely ASIC-based. As systems increasingly use off-the-shelf silicon, the switch
opportunity is potentially explosive > which is why so many companies are pursuing this market. 4
Largely unnoticed, Intel already had a distant #4 slot in Fast Ethernet transceivers prior to its Level
One acquisition. Intel has steadily lost share in the market over the last year, however, and was on
the brink of disappearing from the merchant market entirely before it bought Level One. Category
#2: Wide Area Networking (WAN) Typical WAN equipment includes SONET multiplexers,
integrated access devices, frame relay and ATM switches, routers, and dense wave division
multiplexing (DWDM) gear. ASSPs used in WAN equipment include interface chips ('PHYs'),
framers, mappers, overhead processors, cross-point switches, switch fabrics, and segmentation
and re-assembly (SAR) chips. The dominant customers for WAN components are Lucent, Cisco,
Nortel, Alcatel, and Tellabs, although Ericsson, Fujitsu, Ciena, Nokia and Siemens each have
strength in certain segments. A number of entrepreneurs such as Juniper Networks, Nexabit, Avici
Systems, and NetCore are also interesting customers, since component companies are finding that
start-ups can be an attractive back-door into a top-tier customer through acquisition (witness
Torrent, Shasta, Fibex, Cambrian, et. al.). As a 'network of networks', part of the power of the
Internet is its heterogeneity. This is evident in Table 2, where each column can be either
complementary, competing, or neither depending on the particular application. For example, ATM
cells can run over a SONET transport layer, or SONET can carry a pure packet payload. Other
observations: 1 The WAN market is marked by long gestation periods. In fact, it often can take a
year to win a design with a major customer, another year or two to implement the project and
debug software, and then another year or more for carriers to test the equipment and begin
deployment. The upside is that a major program may remain in production for many years. This is
also why WAN component companies often feel as though they have better-than-average
long-term visibility. 2 The LAN space is further down the ASSP road than the WAN, if only
because the LAN market is less fragmented. As time-to-market pressures mount in the WAN
equipment space and component suppliers become more adept at filling essential functional
blocks, the use of ASSPs in the WAN could rise significantly. 3 Although we have included ATM
under the WAN segment, there remains a fair amount of ATM usage in the LAN (maybe 20-30%
of the total ATM market). ATM's future prospects largely hinge upon the WAN, however, since
10/100/1000 Ethernet has essentially won the LAN protocol battle. 4 Similarly, serial
backplane/fibre channel transceivers are used in chassis equipment that may be either LAN or
WAN oriented. Fibre channel is also a rising force in mass storage applications like redundant
array of independent disks (RAID) and storage area networks. Category #3: Broadband Access
Currently, the access market is dominated by narrowband communications like analog modems or
ISDN. After years of promises, however, both cable modems and xDSL show signs of meaningful
deployment in 1999. Broadcom is the only Internet IC company with meaningful broadband
access exposure, and in its case it is quite meaningful. The company dominates the market for
cable modem interface chips, and also is uniquely positioned in very high-speed digital subscriber
line (VDSL) chips. Barriers to entry (Speed Kills - the competition) We believe there are three
primary barriers to entry: systems knowledge, analog/mixed-signal design expertise, and speed.
Each successful supplier has these elements in varying combinations, and in some cases even just
two of the three are necessary. However, in every case speed is absolutely essential. Another
barrier in the WAN market is the long time-to-money period of WAN-oriented designs.
Start-ups, for example, have a burn-rate constraint and need to hit pay dirt before the money runs
out or before the venture capitalists grow restless. Large companies often have 'attention drift',
where commitment to a particular niche may fade due to reorganizations, paralysis by analysis, or
politicking. This is one reason why the LAN and broadband access markets have attracted more
'announced' interest, since the niches tend to be more monolithic and the chances of (relatively)
instant gratification are better. Note that corporate size has not proven to be a success factor in
any of these markets. For example, Lucent, Intel and TI have each actively pursued various
Internet IC markets without much sustained impact. It's not that the big will eat the small - the fast
will eat the slow. This is one reason why we believe Intel's impact upon the industry will be
relatively negligible, despite its ambitious plans. Historically, Intel has had limited success when it
tries to spread out beyond the PC motherboard. We believe the most likely new entrants into the
Internet IC space will be start-ups with 'patient' capital backing, as well as best-in-class analog IC
specialists such as Maxim and Micrel. Risk factors The running joke in Silicon Valley is that you
can recognize pioneers by the arrows in their backs. While we believe the explosive growth of the
last few years has already helped to separate the wheat from the chaff in the Internet IC market,
the continuous evolution of the market poses an on-going set of challenges for the incumbents (and
opportunities for challengers). We would characterize three general risk factors for the Internet IC
group: customer concentration, valuation, and product obsolescence. The first risk factor makes
the Internet IC suppliers vulnerable to inventory corrections and customer product cycles. The
valuation risk is inherent in the high P/Es of the stocks, but generally the 1999 P/E ratios are
roughly equal to the companies growth prospects for the year. Product obsolescence is a general
risk factor for technology companies, but the fast pace of the communications market means that
none of the Internet IC companies can rest on their laurels. Investment Opinion The seven stocks
in the Internet IC group collectively represent roughly $17 billion in market capitalization, so there
is a range of stock sizes and liquidity suitable for a variety of portfolios. We continue to favor the
WAN-oriented suppliers, and Applied Micro Circuits and PMC-Sierra remain our two favorite
picks in the group. |
