Last week, Cisco (NASDAQ:CSCO) unveiled its new ASR 9000 edge router. The new product, with an advertised capacity of 400 gig per slot and system capacity of 6.4 terabytes has six times the capacity of competing products from Juniper (NYSE:JNPR)(MX-series), Alcatel-Lucent (NYSE:ALU) (7750) and Ericsson (NASDAQ:ERIC) - Redback (RBAK) (SmartEdge series).
While Cisco continues to make investments in its widely deployed but aging Series 7600 router, the ASR 9000 appears to have been built as the successor to this product. The 7600 router starting shipping in 2001 and was built primarily for routing latency-tolerant data traffic over data networks, while the ASR 9000 was purpose built for handling video traffic and to accommodate today’s multi-service (voice, video, and data) IP telecom networks. The ASR 9000 has the raw throughput capacity, the content delivery capability, and the packet processing intelligence to, in Cisco’s words, “deliver high definition video stream to every household in Los Angeles (1.2 million) simultaneously.”
The industry press attendant with the release of the ASR 9000 was equal parts awe at the sheer capacity of the new router, and skepticism that Cisco could actually deliver the advertised capacity and features at any time in the foreseeable future. After all the CRS-1, Cisco’s core router, only delivers 40 gig per slot and the Series 7600, upgraded just two months ago with the new ES40 line cards, also only delivers 40 gig per slot. The skepticism that Cisco could actually deliver the purported 400 gig per slot performance of the ASR 9000 has been fueled in part by Cisco’s reluctance to discuss the technology at the heart of new router.
As an example, Jim Duffy, Managing Editor at NetworkWorld.com wrote:
…Cisco did not discuss how or when it can deliver those 400Gbps per slot, nor did it detail the actual interface line cards and port densities to be housed in the ASR 9000 chassis… After three follow-up attempts to gather these details, all Cisco would say is that the ASR 9000 will support, but is not limited to, Fast Ethernet, 1 Gigabit Ethernet, 10 Gigabit Ethernet and 100 Gigabit Ethernet. Therein lies the problem with the new router, competitors say: It's big on hype, small on reality. If it were a cowboy, it'd be all hat and no cattle.
Further on in the comments section of the article Mr. Duffy wrote with a bit of frustration, “Why won’t Cisco divulge details on when and how it gets to 400gig per slot? Why does it take me 3 failed attempts to get this information and other details? What are they hiding and why?”
What is the company hiding and why? We think we know what the answer might be and it is EZchip Semiconductor (NASDAQ:EZCH). EZchip develops network processors for carrier ethernet equipment and counts among its earliest customers “two of the three largest CESR vendors.” While management is very tight-lipped about who exactly its customers are or what products its processors are going into, we have come to learn through our independent research efforts that EZchip’s NP-3c processor is being used in the initial line cards for the ASR 9000 and EZchip’s 100 gig NP-4 processors are likely to be included in future line cards for this product.
So in one sense the skeptics are right: the pro forma capacity and functionality of the ASR 9000 will not be available in 1Q 2009 when the product begins shipping, but rather this functionality will be available when the NP-4 based line cards are available which we estimate will be in 2010. The important consideration though is that the capacity and functionality of the ASR 9000 isn’t hype, but rather it is already in development.
In reviewing EZchip’s NP-4 processor at DominoAnalytics.com we wrote:
The NP-4 is a 100-gigabit network processor that is software compatible with the NP-2 and NP-3 and provides a significant advancement over these earlier processors in terms of integration, performance, and functionality. The NP-4 provides layer-2 switching and layer-3 routing functions, but in addition also provides higher layer applications aware processing. The NP-4 has integrated support for the enhanced transport of video applications which moves the video processing functionality in switches and routers from centralized services card onto each line card. The system’s services-card processor, typically a multi-core general-purpose processor, will offload critical tasks, in particular those related to video transport, to the NP-4. In order to maintain quality of service, network availability and reliability it is becoming imperative that the packet processing be done at the port level (i.e. on the line card) so that the service card processors do not become bottlenecked. Because there can be as many as 64 network processors in a switch/router chassis versus only one or two multi-core services processors, the offload of video transport can be very significant and greatly enhance the switch/router's performance. The net result is a significant boost in the switch/router’s performance and its support for video applications, while freeing the services card processor for other tasks.
So why is Cisco tight-lipped about EZchip as a technology supplier? After all, Cisco relies on lots of merchant suppliers for its products – Finisar (NASDAQ:FNSR) for optical transceivers, Netlogic (NASDAQ:NETL) for memory, Xilinx (NASDAQ:XLNX) for glue logic, etc. For sure, part of the reason is simply that it is Cisco’s M.O., it never highlights its suppliers; but in this case there may be a bit more sensitivity on the part of Cisco.
EZchip is not just supplying a low-margin off-the-shelf commodity part to Cisco for the ASR 9000; rather EZchip’s network processor is critical technology that is displacing the Cisco silicon (custom ASICs) that has traditionally been central to its routers. So while the ASR 9000 represents a significant technology leap forward for Cisco, perhaps for the first time in the company’s history a third party supplier is integral to much of that technology progress. Viewed in this light, Cisco’s reluctance to disclose the technology behind the ASR 9000 makes a little more sense.
Since EZchip’s network processors provide the performance capabilities of an ASIC but are also flexible and programmable, they are turning up across the board in designs for next generation CESR gear and in the process displacing increasingly expensive in-house custom ASICs. As EZchip becomes ubiquitous in virtually all next generation networking equipment (including the aforementioned recently announced ES40 line cards for the Cisco Series 7600 router and Juniper’s MX-series products, to name but two) over the next couple of years, all vendors will likely be sensitive to the fact that their products increasingly rely on third party intellectual property and as a result providers of merchant silicon like EZchip are positioned to carve out a portion of networking system profits for themselves.
EZchip under the Radar. Despite the critical role that EZchip’s processors are already playing in next generation networking equipment the company is still virtually unknown. It has no formal Wall Street coverage, little institutional sponsorship, and hosts sparsely attended quarterly conference calls. Even Wall St. industry analysts whose business it is to know the networking supply chain of the companies they cover only seem to have a glancing familiarity with the company.
As best we can tell, Tai Liani, Merrill Lynch’s networking analyst was the only sell-side analyst to correctly identify EZchip as having a role in the ASR 9000. He wrote, “The ASR also utilizes Cisco’s QuantumFlow Processor and is powered by E-Z Chip’s MP4 processor." Aside from the small matter of getting the name of the company somewhat wrong ("EZchip") and the name of the chip wrong ("NP-4") he also erred as to which EZchip processor is being used by the ASR 9000. As mentioned previously, the first deployments will use the NP-3c since, as per management’s update two weeks ago, the NP-4 will only begin to sample in 1H'09 and implies that new line cards based on the NP-4 will not be available until 2010. That a company whose technology lies at the heart of most next generation networking equipment can remain virtually unknown to the analysts that follow both networking stocks and comm IC stocks is frankly amazing.
We think it’s only a matter of time before EZchip’s ascendance in the networking sector becomes realized and the stock price trades materially higher. In fact, if analysts stay true to form it will be stock price momentum that attracts their attention and not the business and financial momentum that is already clearly visible. At the end of the third quarter, the company had 166 design wins all of which will be either in production or close to production by the end of next year. This total includes approximately 50 design wins from Cisco and Juniper alone.
While it is still very early in the company’s revenue ramp (3Q’08 revenue of just $9.0 million though up an impressive 72 percent y/y), EZchip is already generating positive cash flow and an impressive 27% non-GAAP net income margins. Juniper currently accounts for 50 percent of revenue off the strength of its MX-series products while Cisco’s first production revenue won’t be realized until 1Q 2009 when both the ES40 line cards and ASR 9000 are scheduled to begin commercial shipments.
While short-term revenue visibility at EZchip remains low—owing primarily to the uncertain timing of design wins going into production and the pace at which in-production design wins ramp—and prevents management from providing guidance, we believe EZchip is on track to generate $0.30 of EPS for FY 2008 on sales of $34 million and could generate earnings in the neighborhood of $1.00 per share for FY 2009 on sales of approximately $60 million. For a company with critical intellectual property in one of the hottest areas of technology spending (carrier ethernet), key design wins at the major OEMs, a deep development pipeline, and a financial model that will easily support net income margins above 40 percent, the shares—currently valued at approximately 9.0x forward earnings—are an absolute gift.