Smartmoney article on Corvis and other optical switch companies:
June 10, 1999
Does Cisco Have a Problem?
By Tiernan Ray
DOES CISCO SYSTEMS (CSCO) have a problem?
A gaggle of startup companies backed by a handful
of competitors have been itching to take the
company down for a few years now. It started in
1996 with Juniper Networks, a Silicon Valley-based
outfit backed by the troika of Lucent Technologies
(LU), Ericsson (ERICY) and 3Com (COMS), with a
little help from German computer giant Siemans AG
and some venture capitalists.
And then they really piled on. Cisco's Internet
switches top out at 40 gigabits per second, but
privately held Avici Systems of North Billerica,
Mass., which got $72 million last year from Nortel
Networks (NT), among others, can rout multiple
connections at speeds of 10 gigabits per second, for
a total of 5.6 terabits, or trillion bits per second.
Privately held Pluris of Cupertino, Calif., advertises a
whopping 184 terabits per second. Cisco has to be mildly concerned.
Enter Dr. David Huber, fiber optics pioneer, who took Ciena (CIEN) public in 1997 in the most
expensive IPO ever at the time, valuing it at $3.4 billion, only to bail out last year after a fight with
the board and with CEO Patrick Nettles. Huber speaks in a patient West Coast drawl about
transforming fiber optics and, maybe, saving Cisco's bacon. In the past year, his former employer,
Ciena, has been rushing to keep up with large competitors such as Nortel and Lucent, building
ever faster wavelength division multiplexing (WDM) equipment to improve fiber optics transmission.
Huber, now with a new venture called Corvis that is backed by Cisco and venture capitalists
Kleiner Perkins, thinks Ciena and others are sadly mistaken.
"I don't think it's any secret that some of the second generation WDM actually has a higher cost
per bit than the first generation, and of course that's wrong," Huber observed as I talked with him
from the floor of Supercomm, the networking trade show in Atlanta this week. "What you want is
the cheapest cost of transmission."
That's a nice way to pull the legs out from his former employer, which is at the show this week
talking about WDM systems that can move two trillion bits, or two terabits, in a second. How did
Ciena go wrong? It all started simply enough. A few years back, straining from exploding Internet
use, the long-distance phone companies wanted a way to carry more traffic without digging more
trenches for fiber-optic cables. Ciena and Huber arrived with the first commercial offering of WDM
technology. Huber's fiber optic equipment split a beam of light into eight colors, or channels, each
of which carried the same amount of information, effectively multiplying the bandwidth eightfold.
Then some phone companies decided they would capitalize on the growth of the Internet by doing
nothing but digging up the streets and laying more fiber. Qwest Communications (QWST) is
completing construction of an 18,500 mile network coast-to-coast this summer, some parts of
which will run as fast as 2.5 gigabits per second. Frontier Communications (FRO), Williams
Companies (WMB), Global Crossing (GBLX), etc. -- all are building huge new fiber networks. To
aid them, Ciena et al. have upped the offerings in WDM. From systems running at 100 gigabits
per second, Ciena is moving up to two terabit capacity. Nortel currently offers equipment that lets
fiber move up to 1.6 terabits per second. The arms race is ceaseless.
Huber has gone in a different direction. Conventional WDM is like a big funnel, moving bits from
one point to another. It still needs conventional routers to make choices, to decide whether to
send bits to New York or to Texas. So Corvis is also offering an optical switch, a fiber optic
system that doesn't need a router. It will decide how to send the bits not by examining packets,
as routers do, but by redirecting light waves. Just as Cisco's routers added intelligence to
conventional networking earlier in the decade, allowing the company to surpass 3Com and Bay
Networks (now a division of Nortel Networks), Huber's box leapfrogs existing optical systems. The
immediate effect, however, is to remove conventional routing from fiber optics, which Huber says
will save some money on equipment.
That may be true, but it's clear as well that Cisco's hand is all over this thing. An optical router
would allow Cisco to leapfrog the Avici's of the world by making light waves, not Internet packets,
the intelligence of the Internet. After years of talking about Internet Protocol (IP), Cisco could have
the last laugh by changing the rules of the game. In fact, Cisco has made private investments in
two competitors of Corvis, Monterey Networks of Richardson, Texas, and Optical Networks of San
Jose, Calif. It seems no amount of optical routing is too much for Cisco.
The punch line is, it's not clear phone companies will want optical switches costing hundreds of
thousands or even millions a pop. Exact technical details are still under wraps, and Corvis is soon
to move from its present Maryland headquarters to a larger facility to gear up for manufacturing.
But the arguments for Corvis' equipment are frighteningly similar to those made in Ciena's defense
shortly before the company collapsed from the high 70s to $8.00 per share last fall. Huber says,
"You've got exponential growth in the information age, and to fuel that you've got to have an
exponential decrease in the cost per bit."
Well, one analyst I highly respect remarked of Ciena last year just before its fall, "I'd be a buyer of
this stock at just about any price, because [the technology] lowers the cost to send a bit." He
may have been right, but it didn't help once WDM became a fiercely competitive business and
Lucent and Nortel quickly overtook Ciena. The company lost a major contract from AT&T (T) once
it became clear there were cheaper and cheaper ways to send a bit. The problem is it became
just too easy for everyone and his brother to churn out WDM equipment. Already, the dogs are on
the trail of optical switching. Ciena issued $463.5 million worth of stock earlier this year to buy
another light switching company called Lightera, and is talking up its own optical switch this
week.
Corvis has some other advantages, though. Its WDM lasers run at only 400 gigabits per second,
but they stretch farther: between two optical tin cans, it can send bits up to 3,200 kilometers, or a
little over 2,000 miles, two-thirds of the way from New York to California. Conventional systems
run only 500-600 kilometers. Again, that should save on equipment costs. "What you have to look
at is how far you can send information before you convert it back to electronics," Huber says.
"Because as long as you stay in the optical domain your costs are cheap."
Lucent and Ciena may catch up in short order. Lucent is demonstrating at Supercomm this week
a system that uses a special kind of laser amplifier, called a Raman amplifier, that can multiply
transmission distances by several times over conventional fiber.
Time will tell if Cisco's money and Huber's daring can bring about a transformation equal to WDM.
Already it smells as if optical switching may be the next arms race. That would be good for Ciena,
for it might breathe new life into the company. At 30 3/8 on Thursday morning, Ciena has been
slowly working its way back from last fall's nadir. If optical switches catch on, it probably means
there's enough demand for bandwidth to feed Cisco and all the Avicis and Plurises out there, at
least in the near future. Probably, too, it signals a scary turning point for companies lower down
on the food chain, like Tellabs (TLAB) and ADC Telecom (ADCT). It is no exaggeration in this
case to say that technology is moving at the speed of light.
|
