Back to GetFilings.com

Table of Contents

UNITED STATES

SECURITIES AND EXCHANGE COMMISSION

Washington, D.C. 20549

FORM 10-K

(Mark One)

For the fiscal year ended December 31, 2004

or

For the transition period from              to             

COMMISSION FILE NO.: 000-50838

NETLOGIC MICROSYSTEMS, INC.

(Exact name of Registrant as specified in its charter)

1875 Charleston Road

Mountain View, California 94043

(650) 961-6676

(Address, including zip code, and telephone number, including area code, of the registrant’s principal executive offices)

SECURITIES REGISTERED PURSUANT TO SECTION 12(b) OF THE ACT:

None

SECURITIES REGISTERED PURSUANT TO SECTION 12(g) OF THE ACT:

COMMON STOCK, par value $.01 per share

Indicate by check mark whether the registrant: (1) has filed all reports required to be filed by Section 13 or 15(d) of the Securities Exchange Act of 1934 during the preceding 12 months (or for such shorter period that the registrant was required to file such reports), and (2) has been subject to such filing requirements for the past 90 days. Yes  x    No  ¨

Indicate by check mark if disclosure of delinquent filers pursuant to Item 405 of Regulation S-K is not contained herein, and will not be contained, to the best of the registrant’s knowledge, in definitive proxy or information statements incorporated by reference in Part III of this Form 10-K or any amendment to this Form 10-K.  x

Indicate by check mark whether the registrant is an accelerated filer (as defined in Rule 12b-2 of the Exchange Act). Yes  ¨     No  x

The aggregate market value of the registrant’s common stock held by non-affiliates of the registrant, based upon the closing sale price of the Common Stock on February 28, 2005 as reported on the Nasdaq National Market, was $123,857,841. This calculation does not reflect a determination that certain persons are affiliates of the Registrant for any other purpose.

As of February 28, 2005, registrant had outstanding 17,679,857 shares of common stock, its only class of voting or non-voting common equity.

DOCUMENTS INCORPORATED BY REFERENCE

Portions of the Registrant’s proxy statement to be delivered to the stockholders in connection with registrant’s 2005 Annual Meeting of Stockholders to be held on or about May 18, 2005, are incorporated by reference into Part III of this Form 10-K. The registrant intends to file its proxy statement within 120 days after its fiscal year end.

Table of Contents

NETLOGIC MICROSYSTEMS, INC.

FISCAL 2004 FORM 10-K

TABLE OF CONTENTS

Table of Contents

PART I

Forward-looking Statements

This report contains forward-looking statements within the meaning of Section 27A of the Securities Act of 1933, as amended, and Section 21E of the Securities Exchange Act of 1934, as amended, which include, without limitation, statements about the market for our technology, our strategy and competition. Such statements are based upon current expectations that involve risks and uncertainties. Any statements contained herein that are not statements of historical fact may be deemed forward-looking statements. For example, the words “believes”, “anticipates”, “plans”, “expects”, “intends” and similar expressions are intended to identify forward-looking statements. Our actual results and the timing of certain events may differ significantly from the results discussed in the forward-looking statements. Factors that might cause such a discrepancy include, but are not limited to, those discussed in “Overview”, “Results of Operations,” “Liquidity and Capital Resources” and “Risks Factors” below. All forward-looking statements in this report are based on information available to us as of the date hereof and we assume no obligation to update any such forward-looking statements. The information contained in this report should be read in conjunction with our condensed financial statements and the accompanying notes contained herein. Unless expressly stated or the context otherwise requires, the terms “we”, “our”, “us” and “NetLogic Microsystems” refer to NetLogic Microsystems, Inc.

ITEM 1. BUSINESS.

Overview

We are a semiconductor company that designs, develops and markets high performance knowledge-based processors for a variety of advanced Internet, corporate and other networking systems, such as routers, switches, network access equipment and networked storage devices. Knowledge-based processors are integrated circuits that employ an advanced processor architecture and a large knowledge database containing network and network user information to make complex decisions about individual packets of information travelling through the network. Our knowledge-based processors significantly enhance the ability of networking original equipment manufacturers, or OEMs, to supply network service providers with systems offering more advanced functionality for the Internet, such as voice transmission over the Internet, or VoIP, virtual private networks, or VPNs, and streaming video and audio.

Prior to our development of knowledge-based processors, we developed integrated circuits to address basic forwarding functions used in networking systems for the core and enterprise networking markets. We introduced our first product in July 1997, which was sold in limited quantities. To respond to evolving networking requirements, we developed our next generation of products, our network search engines, which featured more advanced processing capabilities. From 1998 to 2001, we introduced several of these network search engine products. During this time, our revenue from these products was low, and we experienced significant net operating losses. In 2000, in response to the dramatic growth in and greater complexity of Internet traffic, we recognized the need to develop more advanced processors to enable higher performance for a variety of advanced networking systems. By 2001, we were able to broaden our customer base to include networking OEMs such as Cisco Systems, Inc., Huawei Technologies Co., Ltd. and Nortel Networks Corporation. In 2002, we introduced our knowledge-based processors, and began substantial production in the second half of 2003, resulting in the majority of our revenue in 2003 and 2004.

Our knowledge-based processors incorporate advanced technologies that enable rapid processing, such as a superscalar architecture, which uses parallel-processing techniques, and deep pipelining, which segments processing tasks into smaller sub-tasks, for higher decision throughput. These technologies enable networking systems to perform a broad range of network-aware processing functions, such as access control for network security, prioritization of traffic flow to maintain quality of service, or QoS, and statistical measurement of Internet traffic for transaction billing.

3

Table of Contents

We design our products at the transistor level and use a full-custom layout flow to define how circuits are constructed in silicon. This allows us to optimize circuit design, minimize chip size and reduce power dissipation of our integrated circuits. By minimizing chip size, we are able to optimize the cost of our knowledge-based processors and facilitate the design of our customers’ products within smaller enclosures, or form factors.

We provide complete, systems-level solutions that include interface designs and firmware, device driver, packet-processing and knowledge database management application software, design tools and environments and reference designs. By providing a comprehensive systems-level solution, we help networking OEMs reliably introduce next generation networking systems and significantly enhance their time-to-market. These systems-level solutions are provided free-of-charge to our OEM customers to encourage sales of our products.

Our products are designed into systems offered by leading networking OEMs, including Alcatel, ARRIS Group, Inc., Atrica, Inc., Cisco, CloudShield Technologies, Inc., Extreme Networks, Inc., Fujitsu Limited, Hitachi, Ltd., Huawei, Juniper Networks, Inc. and Nortel Networks. We organized our business in 1995 as a California limited liability company and, incorporated in Delaware in 2000.

Industry Overview

Networking Market Overview

The Internet has experienced dramatic growth and evolved significantly due to a sharp increase in the level of worldwide voice, video and data traffic. According to International Data Corporation (IDC, 2003), total worldwide Internet traffic is expected to increase from 180 petabits per day in 2002 to 5,175 petabits per day in 2007. This represents a compound annual growth rate of 95.7%. This growth has been driven primarily by a wider variety of uses for the Internet, an increased amount of digital media content available through the Internet, and more advanced Internet applications. These applications include:

Due to the rapid growth of voice, video and data traffic, as well as the greater complexity created by the convergence of these types of traffic, there has been significant expansion of the global networking infrastructure using advanced packet-switching protocols, which are the data formats that enable communication among the systems within the network. These networking systems, based upon packet-switching protocols, transport packets of information through the network. The most common packet-switching protocol is the Internet Protocol, or IP.

The Internet infrastructure consists of various networking systems that handle the processing of IP packets. These systems include routers, switches, network access equipment and networked storage devices. An IP packet that is sent from one user’s device to another typically travels through a variety of networks that comprise the Internet infrastructure. These types of networks include:

4

Table of Contents

The following diagram depicts typical network connections within the Internet infrastructure:

IP packets are transferred from one networking system to another through these network connections. Each system within the network and each connected end-user device, such as a computer, is assigned a unique identifier, known as an IP address, which allows these systems and devices to communicate with each other. Decisions on how to handle IP packets are made using the data that is contained in the packet header. The packet header information consists of key data regarding the packet, including the IP address of the system that generated the packet, referred to as the source IP address, and the IP address of the device to which the packet is to be transmitted, referred to as the destination IP address. When a packet arrives at a networking system such as a switch or a router, the packet is processed and decisions about the packet header are made. For example, an IP packet traveling from New York to San Francisco might travel through as many as 15 routers or switches and be processed a number of times by each router or switch. For many networking applications, packet processing must be performed without slowing down the overall flow of communication. Keeping pace with the rate of communication flow is referred to as wire-speed performance.

Transporting a packet from its source to a destination involves a basic class of packet processing commonly known as forwarding. For example, to forward a packet, a switch or router would use a packet processor to extract the header information from an incoming packet and store the information to be transported temporarily in an area known as the buffer. Next, the packet header information, in particular

5

Table of Contents

the destination IP address, would be analyzed to establish the networking system that the packet should be forwarded to in order to move the packet one step closer to its final destination. Networking OEMs implement packet processor functionality either by developing their own custom integrated circuit solutions, or by using network processors, or NPUs, developed by third parties.

Due to the increased usage of the Internet, as well as the greater complexity of Internet-based applications, the amount of processing required for packets is increasing significantly. These more complex applications require multiple classes of packet processing that depend on both the type of content being transported and the information, or knowledge, of the overall network.

Trend Towards Network-aware Processing

Rapid growth of voice, video and data traffic, as well as the greater complexity created by the convergence of these types of traffic, increasingly challenges OEMs to offer systems that enable network service providers to introduce new services over the Internet, such as VoIP, VPNs, video on demand, streaming video and audio and music file downloading. In particular, networking OEM systems must increasingly use knowledge about the overall network, which includes the method and manner in which networking systems are interconnected as well as traffic patterns and congestion points, connection availability, user-based privileges, priorities and other attributes. Using this knowledge to make complex decisions about individual packets of information involves network awareness, which includes the following:

Network awareness in advanced systems requires multiple classes of packet processing, in addition to forwarding. These additional classes of processing include access control for network security, prioritization of packets to maintain QoS and statistical measurement of Internet traffic for transaction billing. Compared to the basic processing task of forwarding, these additional classes of packet processing require a significantly higher degree of processing of IP packets to enable network awareness, or network-aware processing. To maintain wire-speed performance in a network-aware environment, major networking OEMs require hundreds of millions of packet decisions each second, while also updating the knowledge database up to 100,000 times per second.

Several powerful trends are driving greater demand for network-aware processing:

6

Table of Contents

The multiplicative effect of these trends leads to a significantly greater need for advanced processing that utilizes overall knowledge of the network to enable network awareness within switches, routers and other networking systems. Higher levels of performance are required to enable advanced processing for a greater variety of packet processing, such as access control for network security, prioritization of packets to maintain QoS and statistical measurement of Internet traffic for transaction billing, in addition to the forwarding functions.

Networking OEMs have used several approaches to enable network awareness in their systems. One approach involves the use of internally designed custom integrated circuit solutions. Other OEMs have chosen to outsource this requirement to merchant integrated circuit suppliers. Networking OEMs use these integrated circuits to analyze and make decisions about an IP packet based on the packet’s header information, which is extracted by the OEMs’ packet processors. The packet processor inspects the specific implementations of packet-switching protocols and executes specific instructions needed to move the packet through the networking system.

The custom integrated circuit and merchant approaches have both been adequate for the basic decision-making required for forwarding, particularly at lower speeds. However, as the demand for bandwidth and the need to support more advanced Internet applications increases, these approaches are increasingly unable to scale at the pace demanded by advanced applications because of their slower and less efficient processing capability. This creates a bottleneck in the information flow and limits overall system performance. Further, in designing high performance systems, networking OEMs need to address other performance issues, such as power dissipation. Minimizing the power dissipated by integrated circuits is becoming more important for networking systems such as routers and switches, which are increasingly designed in smaller form factors.

Networking OEMs face growing pressure to rapidly introduce new products, reduce their design and manufacturing costs and respond to the growing demand from network service providers for new and advanced services. These OEMs choose to focus on their core competencies in the design and development of certain functionalities within their networking systems, as well as systems-level design and integration. As a result, networking OEMs increasingly seek third party providers of advanced processing solutions that complement their core competencies to enable network awareness within their systems and meet their escalating performance requirements for rapid processing speeds, complex decision-processing capabilities, low power dissipation, small form factor and rapid time-to-market.

7

Table of Contents

Our Solution

To enable network awareness for a variety of advanced networking systems, such as routers, switches, network access equipment and networked storage devices, we offer high performance knowledge-based processors. Our knowledge-based processors use an advanced processor architecture and a large knowledge database containing network and network user information to make complex decisions about individual packets of information travelling through the network. These features enable advanced processing across a variety of classes of packet processing, including access control for network security, prioritization of packets to maintain QoS and statistical measurement of Internet traffic for transaction billing. In addition, we design our products by connecting individual transistors and we use a full-custom layout flow to define precisely how circuits are constructed in silicon, enabling us to optimize circuit design, minimize chip size and reduce power dissipation of our integrated circuits.

Key features of our solution include:

Advanced Architecture for High-Speed Performance. Our knowledge-based processors enable networking OEMs to offer products that process packets at wire-speed performance. Our knowledge-based processors are designed with a superscalar architecture that enables multiple decisions to be processed in parallel. In addition, our knowledge-based processors employ deep pipelining, which segments processing tasks into smaller sub-tasks for higher decision throughput. We use these advanced technologies to enable faster decision throughput in the network. In addition, our knowledge-based processors include features that give access to and support multiple NPUs, allowing more than one NPU to handle packet processing simultaneously. By incorporating our products, networking OEMs are able to process packets more rapidly.

Expandable Processing Resources. We offer knowledge-based processors that can process packets using knowledge databases containing up to approximately 512,000 records on a single integrated circuit. Additionally, our customers can interconnect multiple knowledge-based processors, which extends the usable knowledge database to up to approximately four million records. This allows our OEM customers’ products to support a range of decision-making capacities that scales with end-user requirements. This feature becomes more critical as the number of devices connected to the Internet increases and networking OEMs deploy IPv6, creating the need for additional processing resources and larger knowledge databases to support longer IP addresses.

Full-Custom Integrated Circuit Design for Reduced Cost and Low Power Dissipation. We design our products using full-custom methodologies that allow us to optimize circuit area to implement specific functionality and accommodate larger knowledge databases. Our use of a full-custom layout flow allows for enhanced control of transistor characteristics as needed for optimized circuit design and enables us to minimize chip size and reduce power dissipation of our integrated circuits. By minimizing chip size, we are able to optimize the cost of our knowledge-based processors and facilitate the design of our OEM customers’ products within smaller form factors.

Systems-Level Solutions for Enhanced Design Flexibility and Rapid Time-to-Market. To encourage our customers to design into their products our knowledge-based processors and to assist their design efforts, we offer various systems-level solutions. These include designs for programmable products that interface a customer’s custom integrated circuits with our knowledge-based processors, software and firmware to program our knowledge-based processors and products that interface with our knowledge-based processors, and design tools and environments and reference designs that facilitate the incorporation of our knowledge-based processors into a customer’s system. We do not charge our customers for providing these system-level solutions. We work with NPU providers to validate our reference hardware and software, so that networking OEMs using our reference hardware and software can design their products with our knowledge-based processors more reliably and move to production more quickly. We also provide without charge dedicated applications support to enhance the product time-to-market for our OEM customers who choose to develop their own interfaces to our products.

Transistor-Level Circuit Design for Enhanced Performance. In order to meet the stringent demands of our knowledge-based processors for high speed, low power dissipation and small form factors, we use a highly

8

Table of Contents

customized design approach using transistor-level circuit designs. By using a highly customized design flow, we are able to control precisely how the processing elements are constructed in silicon, leading to higher levels of integrated circuit performance. Designing integrated circuits at the transistor level requires a deep understanding of device physics to maximize transistor device performance. We employ simulation tools that are commonly used in the transistor-level design of analog integrated circuits. We complement these tools with our proprietary techniques to meet the complex design requirements of our knowledge-based processors.

Our Strategy

Our objectives are to be the leading provider of network-aware processing solutions to networking OEMs and to expand into new markets and applications. To achieve these goals, we are pursuing the following strategies:

Maintain and Extend our Market and Technology Leadership Positions. We were the first supplier of knowledge-based processors with approximately 512,000 records, the first supplier to achieve 1.0 Volt operation of knowledge-based processors for lower power dissipation, and the first supplier to achieve operating frequencies of over 300 MHz. We intend to expand our market and technology leadership positions by continuing to invest in the development of successive generations of our knowledge-based processors to meet the increasingly high performance needs of networking OEMs. We intend to leverage our engineering capabilities and continue to invest significant resources in recruiting and developing additional expertise in the area of high performance circuit design, custom circuit layout, high performance I/O interfaces, and applications engineering. By utilizing our proprietary design methodologies, we intend to continue to target the most demanding, advanced applications for our knowledge-based processors.

Focus on Long-Term Relationships with Industry-Leading OEM Customers. The design and product life cycles of our OEM customers’ products have traditionally been lengthy, and we work with our OEM customers at the pre-design and design stages. As a result, our sales process typically requires us to maintain a long-term commitment and close working relationship with our existing and potential OEM customers. This process involves significant collaboration between our engineering team and the engineering and design teams of our OEM customers, and typically involves the concurrent development of our knowledge-based processors and the internally-designed packet processors of our OEM customers. We intend to continue to focus on building long-term relationships with industry-leading networking OEMs to facilitate the adoption of our products and to gain greater insight into the needs of our OEM customers.

Leverage Technologies to Create New Products and Pursue New Market Opportunities. We intend to leverage our core design expertise to develop our knowledge-based processors for a broader range of applications to further expand our market opportunities. We plan to address new market segments that are increasingly adopting network-aware processing, such as corporate storage networks, which increasingly use IP-based packet-switching networking protocols.

Capitalize on Highly Focused Business Model. We are a fabless semiconductor company, utilizing third parties to manufacture, assemble and test our products. This approach reduces our capital and operating requirements and enables us to focus greater resources on product development. We work closely with our wafer foundries to incorporate advanced process technologies in our solutions to achieve higher levels of performance and reduced cost. These technologies include advanced complementary metal oxide semiconductor, or CMOS, implemented in a 0.13 micron logic process flow, up to eight layers of copper interconnect and 300 millimeter wafer sizes. Our business model allows us to benefit from the large manufacturing investment of our wafer foundries who are able to leverage their investment across many markets.

Expand International Presence. We sell our products on a worldwide basis and utilize a network of direct sales and independent sales representatives in the U.S., Europe and Asia. We intend to continue to expand our sales and technical support organization to broaden our customer reach in new markets. We believe that Asia, in

9

Table of Contents

particular China, where we have already established customer relationships, provides the potential for significant additional long-term growth for our products. Given the continued globalization of OEM supply chains, particularly with respect to design and manufacturing, we believe that having a global presence will become increasingly important to securing new customers and design wins and to support OEMs in bringing their products to markets.

Our Markets and Products

Our products are incorporated in a broad variety of networking systems that handle the processing of IP packets. These systems are used throughout multiple types of networks that comprise the global Internet infrastructure, including the enterprise, metro, access, edge and core networking markets. These networks vary in their requirements for bandwidth, number of users to support and complexity of IP packet processing. For example, the core networking market has very high bandwidth requirements, as it typically handles traffic from many individual users, to enable Internet traffic over distances that typically span hundreds or thousands of miles. Our OEM customers’ networking systems in the core network typically incorporate several of our knowledge-based processors to provide very large knowledge databases to accommodate large numbers of users. Due to the increased usage of the Internet, as well as the higher complexity of Internet-based applications, we expect network-aware processing to increasingly become a more essential component of networking systems throughout the global Internet infrastructure.

Key characteristics of our knowledge-based processors include:

We offer a broad range of our knowledge-based processors in two main product families.

Proprietary Interface Knowledge-based Processors—NL5000 Family

Our proprietary interface knowledge-based processors are used primarily by networking OEMs developing their own packet processors. Our products operate in conjunction with an OEM-developed custom integrated circuit or a programmable logic device, such as a field programmable gate array, and feature a proprietary interface that provides advanced interface technology to enable networking OEMs to meet their demanding system performance requirements.

Networking OEMs typically require solutions at different prices in order to target different market segments with the same design. To satisfy this demand, our proprietary interface knowledge-based processor family incorporates product offerings with a range of knowledge database sizes, and all of our knowledge-based processors are designed to be connected in groups to increase the knowledge database available for processing.

We introduced our proprietary interface knowledge-based processors, which are designed in a 0.13 micron TSMC logic process, to the market in the second quarter of 2002. These processors operate from a 1.0 Volt power supply for reduced power consumption and support a knowledge database of up to approximately 512,000 records with performance of up to 500 million decisions per second. These processors also support advanced features for improved fault tolerance that help maintain the data integrity of the knowledge database by providing built-in circuitry to detect faults in the knowledge database.

10

Table of Contents

We also provide versions of our proprietary interface knowledge-based processors that work with proprietary custom integrated circuits and application software developed by Cisco.

NPU Interface Knowledge-based Processors—NL5000GLQ Family

Our NPU interface knowledge-based processors are designed to interface directly to NPUs, such as those from Intel Corporation. They incorporate architectural features that simultaneously support multiple NPUs and NPU-based designs, resulting in more rapid packet processing. These features enable a single knowledge-based processor to make network-aware decisions for both incoming and outgoing communications line channels.

We introduced our NPU interface knowledge-based processors, which are designed in a 0.13 micron TSMC logic process, to the market in the first quarter of 2004. These processors operate from a 1.0 Volt power supply for reduced power consumption and support a knowledge database of up to approximately 512,000 records with performance of up to 125 million decisions per second.

The following table summarizes our current knowledge-based processor offerings:

Knowledge-based Processors Under Development

We are actively developing proprietary interface and NPU interface knowledge-based processors using CMOS logic manufacturing processes with geometries of 0.90 microns and higher with up to eight layers of copper interconnect. These new designs will enable us to offer knowledge-based processors that feature higher levels of performance, including additional functionality developed in close cooperation with our customers to improve application-specific performance.

NETLite^™ Processors

Our NETLite^™ NL3100 processor product family is specifically designed for cost-sensitive, high-volume applications such as entry-level switches, routers and access equipment. The NETLite processor family leverages

11

Table of Contents

circuit techniques developed and refined during the design of our knowledge-based processor families, and benefits from die size optimization, lower power dissipation and redundant computing techniques. In addition, the NETLite processor’s simplified pipeline architecture allows for lower cost manufacturing and assembly in less expensive packages than our knowledge-based processors, and allows for lower cost system designs. As such, the NETLite processors are ideal for entry-level systems that do not require the advanced parallel processing and deep pipelining performance of our high-end knowledge-based processors.

For rapid time to market, our customers can use our software development kit, or NLSDK, to develop and verify hardware and software using the NETLite processors. The NLSDK allows customers to run cycle-accurate patterns at varying operating speeds to exercise the functionality of the NETLite processors and confirm compatibility with target applications. The NETLite processor family is also supported by a suite of production qualified firmware and software drivers and system reference designs, which will enable the growing entry-level system segment to more quickly ramp production with new designs supporting next-generation Internet features such as QoS, security and Layer 3 routing.

The following table summarizes our current NETLite processor offerings:

Legacy Products

We continue to support our legacy network search engines, which include the NL1000 through NL4000 network search engine families and the NL3128GLM network search engines, a device that interfaces directly to certain NPUs from Applied Micro Circuits Corporation. We introduced our network search engine products between 1998 and 2001. These products are fabricated by UMC or TSMC using a range of process technologies from 0.35 micron to 0.15 micron.

We also continue to support a legacy classification and forwarding processor, or CFP, product, that provides certain advantages over NSEs for particular classes of packet processing commonly used in networking systems. We introduced the CFP, which is fabricated by UMC using a 0.25 micron process, to the market in the second quarter of 2000. We continue to research CFP technology and may incorporate it into a future knowledge-based processor product.

Customers

The markets for networking systems utilizing our products and services are mainly served by large networking OEMs, such as Alcatel, ARRIS, Atrica, Cisco, Cloudshield, Extreme Networks, Force 10 Networks, Foundry Networks, Inc., Fujitsu, Hitachi, Huawei, Juniper Networks and Nortel Networks. We work with these and other networking OEMs to understand their requirements, and provide them with solutions that they then qualify and, in some cases, specify for use within their systems. While we sell directly to some networking OEMs, we also provide our products and services indirectly to other networking OEMs through their contract manufacturers, who in turn assemble our products into systems for delivery to our OEM customers. Sales to contract manufacturers accounted for 78%, 38% and 27% of total revenue in 2004, 2003 and 2002, respectively. Sales of our products are made under short-term, cancelable purchase orders. As a result, our ability to predict future sales in any given period is limited and subject to change based on demand for our OEM customers’ systems and their supply chain decisions.

12

Table of Contents

We also provide our products and services indirectly to our OEM customers through our international stocking sales representatives. Our stocking sales representatives are independent entities that assist us in identifying and servicing foreign networking OEMs and generally purchase our products directly from us for resale to OEMs or contract manufacturers located outside the U.S. Our international stocking sales representatives generally exclusively service a particular foreign region or customer base, and purchase our products pursuant to cancellable and reschedulable purchase orders containing our standard warranty provisions for defects in materials, workmanship and product performance. At our option, defective products may be returned for their purchase price or for replacement. To date, our international stocking sales representatives have returned a small number of defective products to us. Our international stocking sales representatives may also act as a sales representative and receive commissions on sales of our products. Our international stocking sales representatives include Bussan Microelectronics Corporation/Mitsui Comtek Corporation and Lestina International Limited. Sales through our international stocking sales representatives accounted for 12%, 22% and 19% of total revenue in 2004, 2003 and 2002, respectively. While we have purchase agreements with our international stocking sales representatives, our international stocking sales representatives do not have long-term contracts with any of our OEM customers that use our products and services.

In 2004, Cisco, including its contract manufacturers, accounted for 73% of our total revenue. In 2003, Solectron (as Cisco’s contract manufacturer), Micron Technology, Inc. and Bussan Microelectronics /Mitsui Comtek accounted for 27.4%, 25.9% and 15.5% of our total revenue, respectively. In addition, in 2003, Cisco, including its contract manufacturers, accounted for 33.5% of our total revenue. In 2002, Cisco, including its contract manufacturers, accounted for 21.6% of our total revenue.

Sales and Marketing

Our sales and marketing strategy is to achieve design wins with leaders and emerging participants in the networking systems market and to maintain these design wins primarily through leading-edge products and superior customer service. We focus our marketing and sales efforts at a high organizational level of our potential customers to access key decision makers. In addition, as many networking OEMs design custom integrated circuits to interface to our products, we believe that applications support at the early stages of design is critical to reducing time-to-market and minimizing costly redesigns for our customers.

Our product sales cycles can take up to 24 months to complete, requiring a significant investment in time, resources and engineering before realization of income from product sales, if at all. Such long sales cycles mean that OEM customers’ vendor selections, once made, are normally difficult to change. As a result, a design loss to the competition can negatively impact our financial results for several years. Similarly, design wins can result in an extended period of revenue opportunities with that customer.

We market and sell our products through our direct sales force and through approximately 17 independent sales representatives throughout the world. Our direct sales force is dedicated to enhancing relationships with our customers. We supplement our direct sales force with independent sales representatives, who have been selected based on their understanding of the networking systems market and their level of penetration at our target OEM customers. We also use application engineers to provide technical support and design assistance to existing and potential customers.

Our marketing group is responsible for market and competitive analyses and defining our product roadmaps and specifications to take advantage of market opportunities. This group works closely with our research and development group to align development programs and product launches with our OEM customers’ schedules. Additionally, this group develops and maintains marketing materials, training programs and our web site to convey our benefits to networking OEMs.

Research and Development

We devote substantial resources to the development of new products, improvement of existing products and support of the emerging requirements of networking OEMs. We have assembled a team of product designers

13

Table of Contents

possessing extensive experience in system architecture, analog and digital circuit design, hardware reference board design, software architecture and driver design and advanced fabrication process technologies. As of December 31, 2004, we had 55 full-time employees engaged in research and development. Our research and development expense was $17.3 million, $18.3 million and $17.1 million for the years ended December 31, 2004, 2003 and 2002, respectively.

We use a number of standard design tools in the design, manufacture and verification of our products. Due to the highly complex design requirements of our products, we typically supplement these standard tools with our own tools to create a proprietary design methodology that allows us to optimize the circuit-level performance of our products.

Technology

We have technological core competencies in the design of integrated circuits to enable network-aware processing using very large knowledge databases. Our products integrate in a single integrated circuit high performance processing, storage circuitry, control functionality and advanced I/O interfaces. Due to the highly specialized nature of our design process, we implement almost all portions of our product design without third party technology, with the exception of readily available intellectual property to implement standard functions, such as memory and timing control circuits.

We have assembled a research and development team with extensive expertise in the following areas:

Transistor-level Circuit Design. A common approach to application specific processor design is to use pre-defined logic functions. This approach is used extensively to shorten the development cycle by allowing an automated process for mapping a product’s logical definition to its construction in silicon. In order to provide knowledge-based processors which feature high speed, low power dissipation and small form factors, we use a more fundamental approach using transistor-level circuit design. With this highly-customized design flow, we are able to implement processing elements that are defined at the most fundamental transistor level and therefore provide higher levels of performance. We employ standard simulation tools that are commonly used in the transistor-level design of analog products. We complement these tools with unique and proprietary methods to meet the complex design requirements of our knowledge-based processors.

Full-custom Layout. In order to implement a transistor-level circuit design, we use a full-custom layout flow to define how circuits are constructed in silicon. This flow enables us to control transistor characteristics to optimize circuit design and minimize chip size. By minimizing chip size, we are able to reduce the cost of our knowledge-based processors. This flow also enables us to control the precise layout of transistors and the connections between them in order to reduce power dissipation. Minimizing the power dissipated by integrated circuits becomes increasingly important for networking systems, which are increasingly designed in small form factors.

Advanced Design Architecture. By working closely with the engineering and design teams of our OEM customers, we utilize our design architecture skills to help ensure that our knowledge-based processors are deployed within their systems in a manner that best addresses their target applications. This product architecture task involves effective partitioning of our knowledge-based processors’ resources to multiple network decision processes, optimized timing to ensure efficient interfaces to other devices and determination of instruction sequences to allow for unique applications. We have acquired our advanced design architecture skills and application knowledge through close collaboration with networking OEMs during the development of successive generations of our products.

Device Physics. We possess a comprehensive understanding of device physics, which is important to the development of knowledge-based processors. This understanding includes not only the desired transistor characteristics to be implemented but also the way in which process variations can affect the operation of an

14

Table of Contents

integrated circuit. To mitigate these effects, we utilize our extensive knowledge of device physics and skills in conjunction with standard tools to make circuit-level design modifications or manufacturing process changes to improve the performance of our products.

Software Product Code and Development Tools. Our knowledge-based processors are delivered to our OEM customers with a suite of supporting software that is intended to accelerate the integration of our solution in their overall system environment. This product code includes knowledge database management software to assist in the initialization and management of records retained on our knowledge-based processors, as well as software used to communicate with our knowledge-based processor. In addition, we provide our OEM customers with emulation and modeling software for the design and verification of their software and hardware. We develop software packages using a team of engineers that possess advanced system knowledge and device modeling skills.

High-speed I/O Interface. Our products interface with high performance packet processors that utilize our knowledge-based processors to decide what action to take on an incoming packet of information. Due to the nature of this functional partitioning, a very high bandwidth connection is required between the packet processor and our knowledge-based processor. To meet the complex requirements of this interface, we develop custom high-speed I/O interfaces. We develop these circuits with advanced technology to support integrated circuit-to-integrated circuit communications.

Manufacturing

We design and develop our products and electronically transfer our proprietary designs to third party wafer foundries to manufacture our products. Wafers processed by these foundries are shipped to our subcontractors, where they are assembled into finished products and electronically tested before delivery to our customers. We believe that this manufacturing model significantly reduces our capital requirements and allows us to focus our resources on the design, development and marketing of our products.

Our principal wafer foundry is TSMC in Taiwan, and we also use UMC in Taiwan. We are actively involved with product development on next-generation processes, and are designing products on TSMC’s 90-nanometer process geometries and higher. The latest generation of our products employs up to eight layers of copper interconnect and 300 millimeter wafer sizes.

Our products are designed to use industry standard packages and be tested using widely available automatic test equipment. We develop and control product test programs used by our subcontractors based on our product specifications. We currently rely on ASAT Holdings Limited in Hong Kong, Amkor Technology, Inc., Advanced Semiconductor Engineering, Inc. in Taiwan, King Yuan Electronics Co., Ltd. in Taiwan, ISE Labs, Inc. and Viko Test Lab in the U.S. to assemble and test our products. In February 2005, we established a representative office in Taiwan to employ local personnel to work directly with our Asian wafer manufacturers and assembly and test houses to facilitate manufacturing operations.

We have designed and implemented an ISO9001-certified quality management system that provides the framework for continual improvement of our products, processes and customer service. We apply well-established design rules and practices for CMOS devices through standard design, layout and test processes. We also rely on in-depth simulation studies, testing and practical application testing to validate and verify our products. We emphasize a strong supplier quality management practice in which our manufacturing suppliers are pre-qualified by our operations and quality teams. To ensure consistent product quality, reliability and yield, we closely monitor the production cycle by reviewing electrical, parametric and manufacturing process data from each of our wafer foundries and assembly subcontractors. We currently do not have long-term supply contracts with any of our significant third party manufacturing service providers. We generally place purchase orders with these providers according to terms and conditions of sale which specify price and 30-day payment terms and which limit the providers’ liability.

15

Table of Contents

Competition

The markets for our products are highly competitive. We believe that the principal bases of competition are:

We believe that we compete favorably with respect to each of the bases identified above. However, some of our larger competitors have greater financial resources and a longer track record as a semiconductor supplier than we do. We anticipate that the market for our products will be subject to rapid technological change. As we enter new markets and pursue additional applications for our products, we expect to face competition from a larger number of competitors. Within our target market, we primarily compete with certain divisions of Cypress Semiconductor Corporation and Integrated Device Technology, Inc. or IDT. We expect to face competition in the future from our current competitors, other manufacturers and designers of semiconductors, and innovative start-up semiconductor design companies.

Intellectual Property

Our success and future growth will depend, in part, on our ability to protect our intellectual property. We rely primarily on patent, copyright, trademark and trade secret laws to protect our intellectual property. We also attempt to protect our trade secrets and other proprietary information through agreements with our customers, suppliers, employees and consultants and through security protection of our computer network and physical premises. However, these measures may not provide meaningful protection for our intellectual property.

As of December 31, 2004, we held 69 issued U.S. patents and 7 issued foreign patents. In addition, as of December 31, 2004, we had 73 patent applications pending in the U.S. We may not receive any additional patents as a result of these applications or future applications. Our U.S. patents have expiration dates from 2017 through 2023. Nonetheless, we continue to pursue the filing of additional patent applications. Any rights granted under any of our existing or future patents may not provide meaningful protection or any commercial advantage to us.

While our patents and other intellectual property rights are important, we believe that our technical expertise and ability to introduce new products in a timely manner will also be important factors in maintaining our competitive position.

Many participants in the semiconductor industry have a significant number of patents and have frequently demonstrated a willingness to commence litigation based on allegations of patent and other intellectual property infringement. From time to time, we have received, and expect to continue to receive, notices of claims of infringement or misappropriation of other parties’ proprietary rights. We cannot assure you that we will prevail in these actions, or that other actions alleging infringement by us of third party intellectual property rights, misappropriation or misuse by us of third party trade secrets, or invalidity or unenforceability of our patents will not be asserted against us or that any assertions of infringement, misappropriation, misuse, invalidity or unenforceability will not materially and adversely affect our business, financial condition and results of operations.

16

Table of Contents

We intend to protect our rights vigorously, but there can be no assurance that our efforts will be successful. Thus, despite our precautions, a third party may copy or otherwise obtain and use our products, services or technology without authorization, develop similar technology independently or design around our patents. In addition, effective patent, copyright, trademark and trade secret protection may be unavailable or limited in certain foreign countries. Moreover, we often incorporate the intellectual property of third parties into our designs, which is subject to certain obligations with respect to the non-use and non-disclosure of such intellectual property. We cannot assure you that the steps we have taken to prevent infringement, misappropriation or misuse of our intellectual property or the intellectual property of third parties will be successful. Furthermore, enforcement of our intellectual property rights may divert the efforts and attention of our management team and may be costly to us.