Market Watch (Winter 2005)

It has been a long, hard winter for network processing. The dot-com meltdown and US economic recession throttled network expansion, reducing demand for components. Add to that the over-promise and under-delivery that seem to characterize new technologies, and network processing struggled to prove its worth to telecom customers. Spring is on its way, however, according to major market research firms.

One aspect of the earlier slow market for network processing units (NPUs) has been a partial consolidation. Startup Internet Machines ceased operations without having reached production with its designs. IBM sold its product line to Hifn (Los Gatos, CA). Freescale Semiconductor (a Motorola spin-off; Austin, TX) discontinued development of its C-Port product line, effectively taking itself out of the market for new designs. Similarly, Vitesse Semiconductor (Camarillo, CA) terminated new NPU product development. Its existing design wins have maintained its market share, but it is unlikely to gain new ground without a roadmap for future development.

The companies that survived the market winter have started seeing revenue growth. Intel (Santa Clara, CA) in particular has benefited with its two new network processor product line introductions, pulling close to lead competitor AMCC (San Diego, CA). Similarly, startups EZchip (Campbell, CA) and Bay Microsystems (Santa Clara, CA) have risen to a combined 5% of market share, according to the Linley Group (Mountain View, CA), publishers of the Linley Wire newsletter on networking as well as several market studies.


Market share for network processors has shifted with consolidation that the downturn stimulated, but the market still has growth potential and the competitive landscape is still subject to change.

Analysts expect that the Spring thaw will be slow, however. A correction of current inventory will offset new-product ramps somewhat, according to the Linley Group’s 2004 A Guide to Network Processors, Sixth Edition. The report estimates that the NPU market will only reach about $150 million in 2005.

Still, this is up from $120 million in 2004, however, and there is still room to grow. According to research firm In-Stat (Scottsdale, AZ), the market penetration for NPUs as ASIC replacements is only 10%, leaving significant opportunities to pursue. In-Stat expects the market to approach a half-billion dollars by 2008.

Despite the eventual size, however, the market is likely to consolidate more before it is through. The slow pace of growth means that new startups would have to survive for several years before achieving significant revenue, and their chances of taking a large enough market share to recover that investment are small. Similarly, companies with marginal revenue streams will continue to pull out, especially as communications processors nibble away at the low-end of the NPU market.

High-End NPUs (10 Gpbs and up) are the bright lights in this market. “Most types of information protocols can fit into harmonics of 10 Gbps,” says Eric Mantion, an In-Stat analyst. “This means that High-End NPUs are very valuable in places where there are many different types of protocols such as between Local Area Networks (LANs) and storage facilities.”

Yet, not all of the competitors are targeting this high-end market. Intel and AMCC offer products that span a wide performance range to target access, enterprise, edge, and metro applications. The other companies have more focused targeting. EZchip is working at the 10 Gbps metro and high-end enterprise markets. Agere (Allentown, PA) is concentrating on the access market. Hifn is going after security and service applications. More 10 Gbps, and even 40 Gbps, products are due out in 2005, however, from several competitors. The race is still on for the high end.

Network processors are programmable devices, designed for easy application. Still, the effort needed to develop end products is significant enough that customers are looking for more. They want completed solutions. During the coming consolidation, it may ultimately not be the technology that wins, but the support. As with general-purpose processors, the availability of tools, software, and applications libraries may dictate the final bloom of network processing and determine which competitors bear fruit.