x
Comms chips

Net Processors Brace for Shakeout

SAN JOSE, Calif. -- In a way, 2002 has been a great year for network processors, as equipment vendors are softening in their resistance to using off-the-shelf parts. But presenters at this week's Network Processors Conference aren't averting their eyes from the market reality -- that network processors are an overpopulated sector facing a bleak economy.

The result: Consolidation will continue, as small chip makers vanish and larger ones drop network-processing projects.

"There's an awful lot left to go," says Doug Spreng, retired president of Applied Micro Circuits Corp. (AMCC) (Nasdaq: AMCC). "2003 could actually be a worse year for this market than 2002 was."

Roughly 30 vendors are producing network processors, according to Linley Gwennap, president of consultancy The Linley Group.

Spreng said he expects that number to shrink to one or two players per market category by 2005, implying that the final number will be more than two or three.

That jibes with the optimistic predictions of panelists at a kickoff session the evening before, who noted that this won't be a Pentium-like, winner-take-nearly-all race.

"This is more like the DSP [digital signal processor] market," where several suppliers have emerged from the glut of startups that competed in the early 90s, says Jeff Cashen, vice president and general manager of AMCC. The key, he believes, is that more than one or two companies emerged in that shakeout. "There are still several suppliers out there who are all making a good living."

Panelists didn't necessarily believe that the culling will be quick and brutal. Doug Davis, general manager of the network processor division at Intel Corp. (Nasdaq: INTC), said companies will drop out more because of the bad economy than because of the overcrowded market. "You'll have maybe a handful of major players out there, but even that might take a while."

Spreng pointed out one life-or-death factor that's in play already, however: that the major equipment players will pick their network processor partners early, in order to have some input into their suppliers' future products. "If you're not on the short list in the next year or so, that's going to be tough."

Spreng believes most of the survivors will have to offer network processors as well as ancillary functions, traffic management in particular. But he did see some hope for the co-processor startups that are specializing in high-end parts -- security processors and deep-classification chips in particular. "If these companies persevere they'll be rewarded with a nice big market in the next couple of years."

Overall, panelists insisted that equipment vendors increasingly are choosing network processors over home-grown ASICs, the "make vs. buy" decision that often accompanies new categories of semiconductors.

Panel members and Spreng both point to ASIC economics as a driving force for network processors. Put simply, it's becoming too expensive to develop chips; Spreng estimates a $5 million to $10 million cost behind creating an ASIC. The difficulty of designing an ASIC has increased as well, a function of the increased amount of circuitry that can be packed onto a chip. "If you're making four ASICs and you screw one of them up, you're going to miss a window," Cashen says.

It's been difficult to track how well network processors are being accepted, however. Network processor vendors insist that design activity is robust, but it's been rare for any systems vendors to admit they're using the chips.

Many systems vendors are afraid that if competitors know which network processor they use, they will lose an edge.

In addition, some network processors are waiting to be included in newer systems products whose launches have been delayed by the economy. "You will start seeing a lot of announcements in the very near term," Cashen said.

The Asian market has been a bright spot for networking in general, and panelists agreed that Asia likes network processors. Compared with European and North American equipment vendors, Asian companies have been faster to complete the build-vs-buy decision and are falling on the "buy" side, Cashen says.

But some panelists had doubts about the Asian market. "In Asia, people are willing to give you a lot more time and talk about your products. How much of that turns into business in the long term, I don't know," says Perry Constantine, CEO of Silicon Access Networks Inc.

— Craig Matsumoto, Senior Editor, Light Reading
www.lightreading.com

Want to know more? The big cheeses of the optical networking industry will be discussing network processors at Lightspeed Europe. Check it out at Lightspeed Europe 02.

Be the first to post a comment regarding this story.
HOME
Sign In
SEARCH
CLOSE
MORE
CLOSE