Mike,
Interesting read. Strong CSCO favor with interesting arguments for ATM/non-ATM..
Carrier-Class Switch/Routers
Investing For The Internet Of Tomorrow
Communications Technology
January 14, 1999 Industry Overview
We believe a new product class of switch/routers is emerging targeted
exclusively at carriers and Internet service providers (ISPs). According to our
estimates, carrier-class switch/router sales are expected to grow at CAGR of 20%, creating a total market opportunity of over $2 billion by 2002.
In our view, third-party market forecasts of the overall router market, focused
on enterprise sales, underestimate the addressable market because they do not
fully account for increasing average selling prices at the high end, and the
potential for carrier-class switch/routers to divert revenue from adjacent
products. Over the 5-year period from 1998–2002, independent forecasts underestimate the aggregate market for carrier switch/routers by $1.2 to $3 billion in absolute terms, in our opinion.
Based on the most critical problems facing the Internet, next-generation carrier-class switch/routers will likely target the following key areas for improvement over current offerings: (1) system performance and throughput; (2) software richness and robustness; (3) scalability and port density; (4) packet latency; (5) reliability and fault-tolerance; and (6) traffic engineering and quality-of-service features.
For investors seeking immediate exposure to this emerging space, we
believe the best public company play is Cisco Systems. While other public
companies compete in the router market, Cisco is the current leader, with roughly 70% market share and, we believe, will likely retain the top position in carrier-class switch/routers. There are numerous, privately held start-ups that are rapidly entering this market, which could provide future investment opportunities through IPOs or acquisitions.
Increased merger and acquisition activity is anticipated in this space as incumbent telecom equipment vendors look to bolster their data product portfolio or to protect existing revenue streams. Potential acquirers include Alcatel, Ascend, Ciena, Ericsson, Fujitsu, Lucent, Nokia, Northern Telecom, Siemens, and Tellabs, and the right acquisition could make these companies an attractive play in carrier-class switch/routers.
Investment Thesis and Ideas
Our enthusiasm for the opportunity presented by carrier-class switch/routers is predicated on the following tenets of our investment thesis.
Robust demand generating CAGR of 20% over the next five years. We expectthe carrier-class switch/router market to grow from an estimated $890 million in 1998 to $2.1 billion in 2002. We believe revenue growth will come from both strong unit demand and a stable-to-upward trend in average selling prices.
Fundamentally underestimated potential market opportunity. Independent
market forecasts have the aggregate router market growing from $6 billion in 1997 to
$8.1 billion in 2002, representing a compound annual growth of only 6%. Because
these forecasts primarily focus on the enterprise market, we do not believe they
indicate the true potential for carrier-class products and fundamentally understate the
market opportunity. In our opinion, market forecasts do not adequately reflect the trend toward higher average selling prices (ASPs) in carrier-class switch/routers, which runs counter to the overall trend in router ASPs.
Furthermore, we do not believe market forecasts anticipate the potential for
carrier-class switch/routers to encroach on other equipment revenue streams as
overlapping functionality is absorbed by routing platforms. The chart below
shows the relative difference between our internal estimates and independent market forecasts. The sum of the yearly differences yields an aggregate underestimation of $1.2-3 billion over the next five years.
Faster growth in the core, but larger opportunity at the edge. Switch/routersfocused on the Internet core will likely be the fastest-growing segment, with a CAGR of nearly 45% through 2002, but edge routers are more numerous and comprise roughly 75% of the addressable market.
Carrier-Class Switch/Routers
Router processing performance lags behind Internet and fiber capacity
growth, creating demand-pull. The first and foremost revenue driver is the growth of data traffic, most of which is directly related to the Internet. Routers are vital components of the Internet's infrastructure, but in many circumstances do not currently possess ample performance to meet its accelerating demands. Router inadequacies are becoming increasingly acute as raw transmission capacity expands through fiber deployment and wave-division multiplexing.
Voice and data convergence requires new class of product. Service providers are strategically maneuvering toward a single, integrated voice and data infrastructure. Routers will need to incorporate circuit switching-like features to enable differentiated service offerings over a packet-based network architecture.
Unhealthy reliance on Cisco products motivates customers to look for second sources. Everyone agrees that Cisco is the dominant router provider across all product segments, and we expect Cisco will continue to be the market leader. However, there currently is no clear number two in the market. Many service providers feel their current reliance on Cisco routers is unhealthy and desire a credible second source. Service providers historically have operated with a 60/30/10 or 70/20/10 vendor distribution for a given type of telecom gear. The largest percentage is allocated to the primary vendors, 20-30% to a secondary vendor and the remaining 10% reserved for testing new equipment in lab or field trials. We believe the market could support 2-3 additional vendors. We would focus investors on the following companies that presently offer products in the high-end router market, and the chart below, which outlines key statistics for their router sales.
Cisco Systems (CSCO – “Strong Buy”): We believe Cisco is presently the best wayfor investors to gain exposure the emerging carrier-class switch/router market. Cisco has long been the router market leader, with roughly 70% market share. Last year, the Company introduced its GSR 12000 platform for the Internet core, which was rapidly accepted by numerous ISPs. The 12000 is a solid addition to its 7500 series products, in our view, which have been Internet mainstays for several years. Despite a host of competitors lurking on the horizon, we believe Cisco remains the best-positioned company in this space.
Ascend (ASND – “Buy”): At one time, Ascend looked to be a serious threat in the carrier-class router market, but encountered difficulties in scaling that portion of its business. While our overall investment thesis is relatively positive and the company possesses some decent router technology, Ascend is not our top choice in carrier-class switch/routers. We plan to revisit our opinion on this topic, all else being equal, when the Company comes to market with next-generation products, which at the earliest, is probably in late 1999.
Lucent (LU – “Market Perform”): Although Lucent has been focusing increasinglyon data networking, and made a splash with the launch of its PacketStar product line, the company has yet to make a measurable impact on the market. Until Lucent begins to demonstrate customer wins and forward momentum, we cannot recommend the stock as an attractive play in carrier-class switch/routers all else being equal.
Northern Telecom (NT – “Market Perform”): Northern Telecom is technicallysecond in the router market, as a result of its Bay Networks acquisition. However, router sales are enterprise-focused, rather than carrier-focused, and Bay's market share has been essentially flat for some time. We believe the company needs to enhance its carrier product offering, potentially through acquisition, before it can compete effectively against Cisco.
Merger and Acquisition Activity
We believe several companies, both public and private, are maneuvering to assume the secondary and tertiary roles within the switch/router segment. Among public companies, the short list is Ascend, Ciena, Lucent, and Northern Telecom and Tellabs. On the private side, prospective companies include Argon, Avici, Juniper, Neo Networks, Netcore, Nexabit, Pluris, Redstone, Shasta and Torrent. We would also note that traditional European telecom powerhouses, such as Alcatel, Ericsson, Siemens and Nokia, have expressed considerable interest in gaining greater IP expertise and could move to develop products internally or, more likely, acquire that expertise from external sources. We believe that most potential acquirers have ample currency to fund a transaction either through stock pooling or by utilizing strong cash positions on their balance sheets. Since many of the private companies have yet to ship product and lack an existing revenue stream, the acquisition valuation likely would be based on the sales potential of the acquired products leveraged across the acquirers' sales channels and customer base. Another motivating factor in merger and acquisition activity is time-to-market. In our view, acquirers will pay a premium to be early to market, which is important to establish a beachhead against competitors, gain market share momentum, and demonstrate innovation and leadership in an emerging market.
We believe acquisition activity will be motivated by both offensive and defensive
rationales. Incumbent providers of SONET add/drop multiplexing (Alcatel, Fujitsu,
Lucent, Northern Telecom) and digital access cross-connects (Alcatel, Tellabs) are most
threatened by carrier-class switch/routers, in our view. Many of the new carrier-class
switch/router platforms possess at least a subset of the functionality traditionally provided
by these products, typically at a fraction of the cost. Consequently, as switch/routers increasingly are connected with the optical transmission layer, wave-division multiplexing and switch/router providers are likely to engage in joint development and collaborative engineering activities, akin to the relationship between Cisco and Ciena. Several switch/routers also possess extensive ATM switching capabilities, and in certain circumstances could replace the need for a separate ATM switch. However, due to some carriers' strategic commitment to ATM, nontrivial product differences between routers and stand-alone switches, and somewhat less compelling cost justification for integration we believe the displacement of ATM switches will take longer to develop. Indeed, we expect the ATM switching market will remain strong for the near-to-midterm, but could weaken longer term as switch/routers narrow the performance gap and further incorporate ATM-like features.
From a more forward-looking perspective, leading telecom equipment providers are keenly aware of the trend toward data-centricity in public networks. Furthermore, the Internet Protocol (IP) is gaining momentum as the integration layer for multiple services. While some telecom providers have made forays into data communications equipment, Lucent and Northern Telecom most aggressively in our view, most still lack credible IP offerings. The emerging crop of switch/router start-ups could be seen as a quick and cost-effective way to gain a stronger foothold in IP.
Continuation further into article…
Our projections show carrier-class switch/routers growing at a compound annual
growth rate of 20% through 2002. We have broken the market into two product
classes, edge and core. Edge routers typically are optimized for high-density access, OC-3
to OC-12 trunk speeds and an overall capacity of 10-40 Gbps. Core routers generally
are optimized for OC-48 and above, have switch capacities ranging from 60 Gbps to
multiterabits, and sacrifice density for scalability and fault-tolerance. These requirements
are likely to shift over time as costs for high-speed local access decline. Edge routers
are more numerous than core products by a factor of 10:1-15:1, and we expect this
ratio to hold relatively constant. However, edge router average selling prices are
considerably less than core ASPs. Although edge router ASPs may edge upward in the
near term, we anticipate a diverging trend in ASPs longer term as edge router prices
experience modest declines while core routers become more expensive. As a result, we
project core routers could experience the fastest growth of any router category, but edge routers will continue to dominate the product mix for the foreseeable future.
Several years ago, the most significant issue plaguing Internet bandwidth was long-haul transmission capacity. In the past three years, we have witnessed an unprecedented coalescence of events that ultimately are expected to increase long-haul fiber-optic capacity by several orders of magnitude. The Telecom Act of 1996, in conjunction with favorable capital markets, has enabled several new entrants, such as Qwest, Level3 and Frontier, to begin massive build-outs of nationwide fiber-optic networks. In addition, the emergence of dense wave-division multiplexing (DWDM) technology is enabling service providers to enhance the capacity of both new and existing fiber networks. Unfortunately, the routers attached to this infrastructure providing the intelligent overlay on the raw fiber capacity have not kept pace with advances in transmission throughput. Presently, router interfaces are available that allow physical connectivity to high-speed optical transport networks; however, the processing capabilities of current routers are insufficient to exploit the available capacity. The current situation is akin to attaching a fire hose to a garden spigot. Physically adapting the spigot is not a problem, but it does not have the necessary water pressure to fill the fire hose.
We also believe it is a foregone conclusion that switch/routers will play a prominent role in the converged public network of the future, as voice and data are integrated over a packet-based infrastructure, further punctuating the need for greater fault tolerance and resiliency. The existing public switched telephone network (PSTN) enjoys a far greater level of reliability than do present-day data networks. In addition, routers will need to provide differentiated and guaranteed levels of service.
Other excerpts…
Another interesting development, in our view, will be automatic protection switching or APS. APS is a critical function, typically provided in the SONET/SDH transmission layer, that provides extremely fast rerouting around failed circuits, frequently invoked when service technicians inadvertently sever fiber-optic cable. The rerouting is so quick with SONET/SDH that the upper-layer protocols often do not detect a failure. However, with several new service providers advocating direct integration of IP over DWDM, effectively eliminating the underlying SONET/SDH infrastructure, APS will need to find an alternate home. We expect carrier-class switch/router vendors, among others, to incorporate at least some subset of APS functionality in their core platforms.
Traffic Engineering and Quality-of-Service
Among engineers and industry watchers there are few technology issues ensnarled in greater uncertainty than quality-of-service (QoS). Solutions to traffic engineering problems have been in a constant state of change for most of the decade. Until recently, the consensus opinion was that ATM would solve most issues; therefore, technologists focused on ways to integrate existing technologies, such as IP and voice, over an ATM infrastructure. Although substantial work continues in this area, an opposing opinion has emerged advocating IP, rather than ATM, as the convergence layer. The fundamental argument for ATM emphasizes its circuit-like characteristics, and ATM was designed from its inception to support multiple traffic types. To that end, ATM possesses a rich set of explicit quality-of-service capabilities, congestion control and traffic management mechanisms, resource reservation protocols, and connection control services. Service providers' fondness for ATM stems not only from it voice and data duality, but also because it could allow carriers to offer multiple service levels and provide service level guarantees to their customers, at a higher price, of course. Ironically, ATM has found its way into datacentric networks, like the Internet, largely because ATM switches have a significant advantage over present-day routers with respect to raw performance.
The anti-ATM argument and, thus, implicitly pro-IP, is that if IP were to gain QoS capabilities similar to ATM and the performance gap between IP routers and ATM switches could be closed, then IP would be the preferential convergence layer. IP puritans believe that because ATM supports multiple traffic types, it sacrifices data efficiency. This sacrifice is often referred to as the “cell tax,” which is the 10-20% capacity loss incurred when utilizing ATM versus pure IP. Moreover, the argument goes, since data is expected to ultimately dominate the traffic mix, converged networks should be built from a technology best suited for data, which is IP.
In our view, IP is still significantly behind ATM when it comes to quality-of-service features. Currently, carriers desiring first-rate traffic engineering capabilities are likely to rely on ATM. Nevertheless, carrier-class switch/routers are approaching and, in some circumstances, exceeding ATM switch performance benchmarks. Over time, we expect IP routers to incorporate more ATM-like features, while ATM switches concurrently integrate greater IP functionality, blurring the line between the two product categories. In addition, several standards are under development to add enhanced QoS capabilities to IP, which is inherently a best effort protocol. The standard that has received the greatest attention is the multi-protocol label switching (MPLS) protocol. MPLS has a rather convoluted history, with several competitors trying to “add value” to the standards process. Nonetheless, consensus is building behind the protocol, and a draft standard is expected some time in 1999. MPLS appends small labels, or tags, to IP packets, which communicate routing information. MPLS is expected to solve many of the difficulties associated with deploying IP over an ATM network, by providing a mapping mechanism that enables routers to assign individual packets to specific paths. Another promising standard is the IP Differentiated Services protocol, or DiffServ.
DiffServ utilizes an existing field within IP headers, called the Type of Service (TOS)
bits, to designate multiple service levels. Based on the TOS field, routers could treat
packets with different precedence, essentially creating coach, business and first classes
for IP traffic. Ultimately, given the amount of resources devoted to enhancing IP, it is difficult to imagine IP switch/routers not achieving quality-of-service capabilities comparable to those of ATM, but it could take several years before mainstream carriers are comfortable with an all-IP approach.
Risks..
Capital Spending: We believe the overall capital spending outlook for telecommunication equipment in 1999 is relatively grim. However, we are expecting a shift in spending to data-and packet-oriented gear, which would include carrier-class switch/routers. If this shift does not occur or if capital spending is lower than anticipated, sales growth could be negatively impacted.
Shifting Capacity Constraints: As carrier-class/switch routers increase the total data processing power of the Internet, other bottlenecks could be exposed, not unlike the way in which WDM revealed the inadequacies of present-day routers. It is possible that service providers would curtail spending on switch/routers if other elements of the network were found to be more critical to increasing capacity. |
