Juniper Networks is grabbing market share in Internet routers from networking giant Cisco, even as CEO Scott Kriens tries hard to downplay the fight
By Russ Arensman, photograph by Christopher Springmann
Scott Kriens wants us to believe his company is misunderstood. Kriens, the 43-year-old CEO of Sunnyvale, CA-based Juniper Networks Inc. particularly wants to debunk the notion that his company is the biggest competitive threat to networking giant Cisco Systems Inc., San Jose, CA. After all, as he tells it, the two have fundamentally different approaches to the computer networking business.
Cisco, he notes, offers networking systems for nearly everyone, from home-office users up to major telecommunications carriers. It sells hundreds of products and services, including more than two dozen kinds of routers—the devices that monitor and direct the flow of data across computer networks. Juniper, on the other hand, produces just a handful of ultra-fast router types, all of them aimed at moving data over the Internet.
Still, headline writers continue to describe Juniper in terms like “Cisco Killer,” “The Next Cisco” and “the David figure playing to Cisco’s Goliath.”
Flattering as those descriptions may be, Kriens maintains that to view his company and Cisco as head-on competitors is to make the same mistake as those who, in the early days of Intel Corp., thought the Santa Clara, CA-based chip company was trying to compete with Armonk, NY-based IBM Corp. in computers. “In fact, as it unfolded, what Intel was building was the microprocessor business,” Kriens says.
He contends that the networking segment Juniper focuses on, similarly, will soon be an industry in its own right and that, as the first company “purpose built” for this emerging market, Juniper will inevitably become the dominant player.
“Right now it appears as though there’s one networking industry, when in fact, as with the computer industry, it will segment into many industries,” he says. “And so we think we’re simply repeating history. When a new problem surfaces in a very large, fast-moving market, a company that is built exactly for that requirement becomes the market leader.”
Juniper, though not yet the market leader, has made impressive strides in the two years since introducing its first products. Despite Kriens’ protestations, it clearly has established itself as a strong competitor to Cisco, at least in the Internet routing market. Its sales and stock price have soared dramatically, as has its customer list that now numbers more than 130. The next question, however, is whether Juniper can withstand a renewed challenge from Cisco, as well as from a host of new rivals. It also must gear up to run a much larger business while keeping up the pace of technology development and deciding whether, and where, to expand its product line.
Driving the market
Juniper’s long-term success is still far from a sure bet, but Chris Nicoll, vice president of telecom infrastructure for Current Analysis Inc., a Sterling, VA-based market research firm, says the company is off to a promising start. “What makes Juniper most impressive is that they’re doing what a lot of other start-ups say they can’t do,” he says. “They’re really positioning themselves to drive the market, rather than respond to the market.”
“Right now it appears as though there’s one networking industry, when in fact, as with the computer industry, it will segment into many industries.” —Scott Kriens, CEO, Juniper Networks
Juniper’s target market— high-performance routers used mainly at the core of Internet service provider (ISP) networks—barely existed five years ago. But it has already become a market worth more than $2 billion and is doubling every year as ISPs scramble to handle the torrent of new Internet traffic, which itself is doubling every four to six months. Market researcher Ryan Hankin Kent Inc. (RHK), South San Francisco, CA, predicts that by 2002 the market should reach nearly $9 billion (see chart).
RHK puts Juniper’s share of the high-end Internet routing market at 20%, while Redwood City, CA-based market researcher Dell’Oro Group Inc. puts it at nearly 30%, double its market share of a year ago and up nearly 7% in the past quarter alone. Dell’Oro estimates that Cisco’s market share has fallen from more than 80% to 69% during the past year (see chart).
Dell’Oro’s numbers for Juniper are higher than RHK’s, for a few reasons, but, those differences aside, “the trend remains the same,” says Tam Dell’Oro, founder of the research firm. No matter how you slice the numbers, Juniper is rapidly gaining market share at the expense of Cisco.
Kriens prefers to downplay the competition with Cisco, pointing out that there’s plenty of growth available for both companies. Indeed, Dell’Oro estimates that while Juniper’s quarterly sales quadrupled during the nine months ended Sept. 30, 2000, Cisco’s much larger quarterly sales in that category more than doubled—still an enviable growth rate. It’s also worth noting that Cisco controls more than 90% of the overall $10-billion-plus router market. “By no stretch of the imagination is Cisco in jeopardy of not capitalizing on much of the opportunities in this new networking world,” says Kriens.
Still, Juniper clearly is making inroads into Cisco territory. And Cisco isn’t taking the challenge lying down. For starters, it is devoting more high-level executive attention to its service-provider business line, which competes directly with Juniper. Cisco board member Ed Kozel, the company’s original chief technical officer (CTO), has returned from a hiatus as a venture capitalist to become CTO for the service provider business (although Cisco insists the move is unrelated to Juniper).
On the product front, Cisco introduced the 10000 ESR edge router in April, several months before Juniper unveiled its M5 and M10 routers, also aimed at the network’s edge. As of year-end, however, Cisco still hadn’t introduced its long-awaited OC192 interface cards to link its top-of-the-line 12016 GSR routers to high-speed optical networks at a blazing 10 gigabits, or 10 billion bits of data per second. Juniper has been offering an OC192 interface for its top-of-the-line M160 router for almost a year.
“Juniper has a six-month lead in time to market, but that doesn’t mean Cisco can’t catch up,” says Muayyad Al-Chalabi, RHK core switching and routing analyst. “[Cisco has] a lot of good people; they’ve got the will; they’ve got the deep pockets; and they’re putting a lot of energy and focus behind the service provider business.”
Certainly Cisco, with $21 billion in sales during the 12 months ended last Oct. 28, nearly 40,000 employees and more than $6 billion in cash, has far greater resources than Juniper, which has about 1,000 employees, $1.6 billion in cash and sales of $673.5 million in 2000. Yet despite the mismatch in size, Kriens is convinced Juniper can stay ahead by focusing tightly on its target market and by constantly learning in what he calls “the Internet classroom.”
The strategy, in short, is to build increasingly powerful networks, then apply the lessons learned along the way to improve subsequent products. “As we execute and deliver systems at the cutting edge, we learn a great deal that can only be known from having done that,” Kriens says. He argues that the company with the lead inevitably discovers things ahead of its rivals and, if it applies those lessons diligently, it “just continually knows 12 months more about the problem than the people 12 months behind.”
Hard to catch
Juniper founder Pradeep Sindhu set out quite deliberately in 1996, after 12 years with Stamford, CT-based Xerox Corp., to create a company that would be difficult for competitors to catch. He took a two-month vacation from his job as a principal scientist at Xerox’s Palo Alto Research Center to think about the problem. Though his doctorate from Carnegie Mellon University was in computer science, he felt the computer field was saturated with products that were in many cases becoming commodities. “In the networking field, on the other hand, things were wide open,” he recalls.
Sindhu knew that the transmission capacity of optical fiber had been doubling almost every year, and he also saw that Internet use was expanding exponentially, thanks to the development of the World Wide Web and user-friendly browser software. What stood between this seemingly boundless confluence of data networking supply and demand were routers—the essential traffic-cop-like devices that manage the flow of Internet data.
“The companies building routers back then were using technology that, to my taste, was about four to five years behind the technology curve if you compared it to what was being used to build advanced computers,” says Sindhu.
He also noticed that routers, at that time, were aimed more at translating between proprietary networking protocols such as IBM’s Systems Network Architecture and Digital Equipment Corp.’s DECnet than at maximizing data throughput. It was clear to him, however, that the future belonged to Internet protocol (IP), the first truly universal language allowing any computer to talk to any other.
“The companies building routers back then were using technology that, to my taste, was about four to five years behind the technology curve.” —Pradeep Sindhu, Juniper founder, vice-chairman and chief technology officer
Sindhu reckoned he could build a better router by optimizing it to handle IP data, while using the latest techniques from computing and semiconductors. “I convinced myself relatively quickly that it was possible to get a 50x to 100X improvement in price/performance, and literally a 10x to 20X improvement in performance,” he says.
“At that point, it was almost like an emergency,” he recalls. “If I realized these things, there were other people that must realize them also. And so it was absolutely imperative to actually go do something about it.”
Sindhu took his idea to several venture capitalists before Vinod Khosla, a partner with Menlo Park, CA-based Kleiner, Perkins, Caufield & Byers agreed to provide the seed money for Juniper. The company was launched in February 1996 with $2 million, enough to hire 15 engineers, including key designers from Cisco, Sun Microsystems Inc., and MCI Communications (now part of MCI Worldcom Inc.). Later that year, a second round of venture funding allowed the company to recruit more staff, including CEO Kriens, who had previously co-founded and headed both sales and operations for StrataCom Inc., a San Jose, CA-based network equipment company that was acquired by Cisco in July 1996 for $4.5 billion.
Close customers
While Sindhu, now Juniper’s chief technical officer, oversaw the development of the company’s crucial Internet processing chips and software, Kriens turned his attention to lining up customers. Borrowing a strategy from StrataCom, he enlisted the help of key potential customers, several of which even agreed to help fund the company. Lucent Technologies, Nortel Networks, LM Ericsson, 3Com, MCI Worldcom’s UUNet Technologies subsidiary and a joint venture of Siemens and Newbridge Networks combined to invest $40 million in the young company in late 1997 in return for 4% ownership stakes. Other early investors included AT&T and numerous well-known venture capitalists.
Doug Junkins, vice president for IP engineering at Verio Inc., an Englewood, CO-based ISP serving many large U.S. corporations, says his company and others worked closely with Juniper to test and fine-tune its routers. “We worked with their engineers for over a year before Juniper shipped their first production routers,” he says. As a result, Verio and other key customers were so familiar with Juniper’s products they were ready to begin ordering them as soon as they hit the market.
Juniper’s initial success came partly from being the first legitimate second source for Cisco-compatible routers. Others, such as Ascend Communications Inc., which was acquired by Lucent Technologies Inc., Murray Hill, N.J., for $20 billion in 1999, developed competing routers for the network core, but had limited success because their products weren’t inter-operable with Cisco’s.
“Juniper came in at a time when people were looking for alternate providers and suppliers,” says RHK analyst Al-Chalabi. “And it happened they were the only game in town as an alternative to Cisco.”
“Juniper has a six-month lead in time to market, but that doesn’t mean Cisco can’t catch up.” —Muayyad Al-Chalabi, RHK routing analyst
Verio’s Junkins says his company was eager to see an alternative supplier for core routers. “Cisco had such a dominance in the market that we thought some diversity would really help, both to push Cisco to move ahead from a reliability point of view and also from a cost perspective.” he says. As far as he’s concerned, the competition has been good for everyone. Says Junkins: “Juniper has definitely pushed Cisco ahead.”
Juniper, for instance, helped bring the issue of software reliability to the fore by making it one of its key selling points. “Juniper brought in better reliability and stability,” agrees Current Analysis analyst Nicoll. He cites a University of Michigan study that found more than half of the Internet’s downtime a few years ago was due to Cisco routers being shut down either by failures or for routine maintenance. “Cisco’s come a long way since then,” he says, adding that its software reliability is comparable to Juniper’s now.
One reason for Juniper’s initial edge in both reliability and performance was its ability to design its routers starting with a clean slate. Sindhu says Juniper implemented many processor-intensive routing functions like packet forwarding in hardware, rather than in its JUNOS operating software. It also designed its software in modular form, with separate programs running in protected memory mode. That allows maintenance and software changes to be done one module at a time, without affecting the others. Cisco’s vast and venerable IOS software, on the other hand, offers more features than JUNOS but uses a “monolithic” structure, with interconnected programs. Juniper officials claim that IOS’s complexity makes it harder to upgrade and more prone to failure.
Juniper also has been scrupulous about testing its software extensively with multiple customers before introducing it commercially. “The thing that’s made them successful is they’ve spent a lot of time debugging their software before it came out,” says Al-Chalabi.
On the hardware side, Juniper’s routers are designed around a handful of key application-specific integrated circuits (ASICs) fabricated by IBM. The company’s router chips are divided into two key functions: a routing engine based on an Intel Pentium processor, and a packet-forwarding engine that uses Juniper’s own Internet Processor ASIC . Last April, the company introduced an improved version of that chip, the Internet Processor II , with additional processing functions including traffic management and load-balancing to shift heavy traffic away from overworked parts of a network.
Kriens concedes that developing custom chips is a time-consuming, difficult process, but says that no off-the-shelf network processors currently meet Juniper’s needs. “It’s only from necessity that we will develop our own silicon,” he says. “At the level of technology we’re pushing, there are no standard parts.” Juniper does buy standard chips for the 70-some varieties of interface cards that connect its routers to users’ networks, via Ethernet, asynchronous transfer mode (ATM) and other standard protocols.
Juniper is more than happy, however, to outsource its entire manufacturing process to its contract-manufacturing partners, Milpitas, CA-based Solectron Corp. and Celestica Inc., Toronto, Canada. The contract manufacturers are responsible for everything from procuring materials to final assembly, test and shipment to customers. That arrangement not only saves Juniper from devoting employees, capital and building space to manufacturing, it also has helped the company increase its sales very quickly.
Growing it alone
From the moment Juniper introduced its first product, the M40 router, in late 1998, its growth trajectory has mirrored the hyper-growth of the Internet itself. The company’s revenues hit $103 million in 1999, and by the end of 2000 it was selling three times that much every quarter. The company eked out its first profit of $9 million in the fourth quarter of 1999. A year later in the fourth quarter, its profit hit $62 million on sales of $295 million.
Not surprisingly, Wall Street and growth-hungry investors have climbed on the Juniper bandwagon in droves. Juniper’s June 1999 IPO raised $65 million, and its shares tripled to nearly $100 on their first day of trading. A secondary stock offering in October 1999 raised another $324 million and the stock has split twice since then. At the peak of nearly $250 last October, the price of a Juniper share was up more than 40-fold from its split-adjusted IPO price of $5.66 and the company’s market capitalization was more than $70 billion. Even after the market’s subsequent sell-off of technology stocks, Juniper’s shares in January were still up close to 25-fold at $136 from their split-adjusted IPO price and its market cap was more than $40 billion.
“We’re not exactly sure what Juniper’s next move is going to be…They’re not really a company of talkers, they’re a company of doers.” —Chris Nicoll, vice president of telecom infrastructure, Current Analysis Inc.
Although Juniper has been rumored to be a takeover candidate on several occasions, its hefty valuation has no doubt helped it to remain independent. The Wall Street Journal recently reported that Lucent considered buying Juniper in late 1998. More recently, Canada’s Nortel Networks Corp., Brampton, Ontario, was rumored as a possible acquirer. But as Dick Moley, StrataCom’s former CEO and Kriens’ former boss, notes, “the best poison pill is a very high market value.”
Some have suggested Juniper might do well to join in a “merger of equals,” with one or more other young networking players such as Palo Alto, CA-based Copper Mountain Networks Inc., Sunnyvale, CA-based Redback Networks Inc. or Chelmsford, MA-based Sycamore Networks Inc. But while much of the networking industry has resembled a 1970s singles bar in recent years, with start-ups and even established companies pairing up with wild abandon, Juniper seems content to go it alone.
“Most of the great companies I know of were built by hand, through innovation,” says Kriens. While not ruling out future acquisitions, he vows that “we’ll never become dependent on them.”
Cisco, by contrast, has made acquisitions central to its growth strategy. During the past three years, it has spent more than $23 billion, mostly in stock, to buy dozens of companies, while making equity investments in scores of others. Its stated goal is to acquire 25 companies annually in order to sustain its 50%-plus annual sales growth.
Kriens concedes that Cisco has been remarkably successful as a serial acquirer of new companies and technology, but calls it the exception, rather than the rule. He contends that a better example is Lucent, which acquired close to 40 companies in a two-year stretch before stumbling into an unexpected growth slowdown last year.
Juniper has made a few relatively small acquisitions. In December, it agreed to pay $260 million in cash and stock for privately held Micro Magic Inc., a Sunnyvale, CA-based chip design company. Back in January 2000, it bought Hong Kong-based Pacific Advantage Ltd. for $4 million, adding 15 new sales and marketing staff to Juniper’s seven Asian offices. In November 1999, it spent $19 million on Layer Five, a Palo Alto, CA-based developer of network-processing algorithms.
Juniper also has made strategic investments in several companies, including San Jose, CA-based optical networking start-up Zaffire Inc., and network processor developer SiByte Inc., which was later acquired by Broadcom Corp., Irvine, CA, for about $2 billion. Yet none of these deals have put much of a dent in Juniper’s $1.6-billion cash hoard, nor its huge stock valuation. Juniper clearly could afford a sizable acquisition, but Kriens is convinced it would be a mistake for his company to wander far from its expertise in core routers.
Despite the company’s disdain for growth through acquisitions, Juniper has been anything but reluctant to introduce new products. After taking three years to complete its first product, the company took less than two years to launch four additional routers, which shared the same ASICs and other components, but offer different levels of performance. In November 1999, Juniper rolled out the M20 , a compact, lower-power and lower-cost router for smaller ISPs. In March 2000, it introduced the powerful M160 , with four times the throughput and performance of the original M40. Then last September, it introduced two new smaller routers, the M5 and the M10 , aimed at connecting users to the network’s edge.
On the marketing front, Nortel has agreed to resell Juniper’s routers in a deal that Juniper officials hope will boost their sales to telephone companies in Europe and Asia. Juniper also agreed in December to invest $50 million in a joint venture with Sweden’s LM Ericsson Telephone Co., Stockholm, to sell mobile Internet routing products to ISPs and mobile network operators.
Juniper officials aren’t saying what comes next. They’re notorious, in fact, for refusing to offer any clues about future products. “We’re not exactly sure what Juniper’s next move is going to be,” says Current Analysis’ Nicoll. He admits that the lack of advance information can be frustrating, especially in an industry where most companies announce products months or even years in advance. But it’s only added to Juniper’s credibility. “They’re not really a company of talkers, they’re a company of doers,” he says.
Encore?
Nonetheless, questions linger about what Juniper will do for an encore. Having produced essentially five variations of its original design, analysts doubt the company’s ASIC architecture can stretch much further. “The M40 architecture, now extended into the M160, has reached the limits of its scalability,” says Nicoll.
RHK analyst Al-Chalabi says both Juniper and Cisco will have to fend off challenges from a new generation of router companies, including Avici Systems Inc., North Billerica, MA; Pluris Inc., Cupertino, CA; IronBridge Networks Inc., Lexington, MA; Charlotte’s Web Networks Ltd., Andover, MA, and Procket Networks Inc., Milpitas, CA. All are developing routers with more processing power and scalability, often by expanding the number of available network connections. Most claim to be developing routers that can process at least a “terabit,” or 1 trillion bits, of information per second—roughly six times more than Juniper’s M160.
Of those, Avici’s product appears to be the closest to market-ready. The company has completed field trials of its routers with AT&T and Denver, CO-based Qwest Communications International Inc., which has agreed to deploy Avici routers in its next-generation optical network. IronBridge, whose routers are being tested by several potential customers, offers “virtual routing,” the ability to provide private network services to multiple customers on the same router. And Procket, while younger than some of the others, is notable for having hired Tony Li, the celebrated router designer who played a key role in developing Juniper’s original M40, as well as Cisco’s GSR core routers.
“Most of the great companies I know of were built by hand, through innovation.” —Scott Kriens, CEO, Juniper Networks
Al-Chalabi isn’t worried about either Juniper’s or Cisco’s ability to survive the router upstarts’ challenge. But he definitely expects one or more of the newcomers to carve out a spot. “I think the market should be able to support multiple players,” he says. “Juniper would lead you to believe there is room for only two, but I don’t believe that.”
Lauri Vickers, senior networking analyst with Scottsdale, AZ-based Cahners In-Stat Group, notes that while several of the new router players look promising, their much-hyped terabit routers are still more marketing ploy than reality. “True terabit throughput doesn’t exist at this time,” she says. For now, she says, the Internet router market remains “a two-horse race” between Juniper and Cisco.
As for Kriens, he says it’s not the competition, new or old, that keeps him up at night—it’s whether or not Juniper lives up to its full potential. “I don’t want to look back on all of this and say ‘we only did 80% of what we were capable of,’ ” he says. “The market opportunity right now, the potential of changing the way that people connect and communicate with each other in this global language, is a completely undeveloped opportunity—and to not take full advantage of that would be criminal.”
eb-mag.com |
