CRAY INC - 10-K Annual Report

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 past 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 o

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 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. o

Indicate by check mark whether the Registrant is an accelerated filer (as defined in Exchange Act Rule 12b-2): Yes x No o

The aggregate market value of the Common Stock held by non-affiliates of the Registrant as of June 30, 2003, was approximately $542,000,000, based upon the last sale price of $7.90 reported for such date on the Nasdaq National Market System.

As of March 1, 2004, there were 73,168,483 shares of Common Stock issued and outstanding.

Portions of the Proxy Statement to be delivered to shareholders in connection with the Registrant’s Annual Meeting of Shareholders to be held on May 12, 2004, are incorporated by reference into Part III.

Cray, Cray-1, UNICOS and UNICOS/mk are federally registered trademarks of Cray Inc., and Cray Y-MP, Cray C90, Cray J90, Cray T90, Cray T3E, Cray SV1, Cray SV1ex, Cray SX-6, Cray MTA, Cray MTA-2, Cray MTX, Cray X1 and Cray X1E are trademarks of Cray Inc. Other trademarks used in this report are the property of their respective owners.

This Annual Report on Form 10-K contains forward-looking statements that involve risks and uncertainties, as well as assumptions that, if they never materialize or prove incorrect, could cause our results to differ materially from those expressed or implied by such forward-looking statements. All statements other than statements of historical fact are statements that could be deemed forward-looking statements, including any projections of earnings, revenues or other financial items; any statements of the plans, strategies and objectives of management for future operations; any statements concerning proposed new products, services or developments; any statements regarding future economic conditions or performance; statements of belief and any statement of assumptions underlying any of the foregoing.

The risks, uncertainties and assumptions referred to above include fluctuating quarterly results; the possibility of quarterly net losses; the timing of product orders, deliveries and customer acceptances; the timely development, production and acceptance of products and services and their features; the timing and level of governmental support for supercomputers; a volatile market price for our common stock; our dependency on third-party suppliers to build and deliver necessary components; the challenge of managing asset levels, including inventory; the difficulty of keeping expense growth at modest levels while increasing revenue; our ability to retain and motivate key employees; and other risks that are described from time to time in our Securities and Exchange Commission reports, including but not limited to the items discussed in “Factors That Could Affect Future Results” set forth in “Business” in Item 1 below in this report, and in subsequently filed reports. We assume no obligation to update these forward-looking statements.

In this report, we rely on and refer to information and statistics regarding the markets for various products. We obtained this information from third party sources, discussions with our customers and our own internal estimates. We believe that these third-party sources are reliable, but we have not independently verified them and there can be no assurance that they are accurate.

We design, develop, market and service high performance computer systems, commonly known as supercomputers. These systems provide capability and capacity far beyond typical mainframe computer systems and address the world’s most challenging computing problems for government, industry and academia. In 2002 we completed the hardware development of and began selling our Cray X1 system, an “extreme performance” supercomputer designed for the high end of the supercomputer market. We are developing enhancements to this system that will increase significantly processor speed and capability, which we call the Cray X1E system, with first shipments scheduled for the second half of 2004. In mid-2002 we began a development project with Sandia National Laboratories to design and deliver in 2004 a new, high bandwidth, massively parallel processing supercomputer system called Red Storm. In October 2003 we announced that we would develop a product line based on the Red Storm system, targeting the need for highly scalable microprocessor-based Linux supercomputers with high bandwidth. This product is scheduled for shipment in the second half of 2004. In mid-2003 we began work under a contract with the Defense Advanced Research Projects Agency (“DARPA”) that supports our program to develop a commercially available system capable of sustained performance in excess of one petaflops, which we call our Cascade program. We expect that most of our 2004 product revenue will come from sales of our Cray X1 and Cray X1E systems, with additional contributions from the completion of the Red Storm project, continued work on the Cascade project and sales of the commercial version of the Red Storm system. We provide maintenance services to the worldwide installed base of Cray computers. We also offer high performance computing services that leverage our industry technical knowledge. See “— Product Offerings and Projects” below.

In February 2004 we announced that we had signed a definitive agreement to acquire OctigaBay Systems Corporation, a privately-held development-stage company located in Vancouver, B.C. OctigaBay is developing a balanced high bandwidth system designed to be highly reliable and easy-to-use and targeted for the

midrange market. The acquisition is subject to customary approvals and expected to close before the end of April 2004. For further information, see “— Agreement to Acquire OctigaBay” below.

We were incorporated under the laws of the State of Washington in December 1987. Our headquarters offices are located at 411 First Avenue South, Suite 600, Seattle, Washington, 98104-2860, our telephone number is (206) 701-2000 and our web site address is: www.cray.com.

In many ways our current history began on April 1, 2000, when we, as Tera Computer Company, acquired the operating assets of the Cray Research division from Silicon Graphics, Inc. (“SGI”), and renamed ourselves Cray Inc.

Tera Computer Company was founded in 1987 with the purpose of developing a new supercomputer system based on multithreaded architecture. We had an initial public offering in 1995. In 2000, we were still in the development stage with limited revenue from sales to one customer, and approximately 125 employees, almost all of whom were located in our Seattle office.

Cray Research was founded in 1972 by Seymour Cray and introduced its first product, the Cray-1, in 1976. Cray Research pioneered the use of vector systems in a variety of market sectors and dominated the supercomputer market in the late 1970’s and 1980’s. Cray Research introduced a series of vector-based systems, including the Cray Y-MP, C90, J90, T90 and SV1 systems. Cray Research also developed leading high bandwidth massively parallel systems, notably the Cray T3D and T3E systems, using Alpha microprocessors from Digital Equipment and later Compaq Computer. In 1996, SGI acquired Cray Research and cancelled the development of the successors to the only two U.S. produced capability-class supercomputers at the time, the Cray T90 and T3E systems. In 1997, at the instigation of Cray Research, the U.S. government imposed extensive anti-dumping duties on Japanese vector supercomputers, effectively preventing them from entering in the U.S. market. SGI also moved a substantial number of the established Cray Research customers from Cray Research products to the SGI Origin line of products. In 1998, SGI and the Department of Defense entered into a cost-sharing contract for the development of the Cray X1 system (then code-named the SV2). In 1999, SGI announced that it would consider offers to purchase the Cray Research division.

On April 1, 2000, we acquired the operating assets of the Cray Research business unit from SGI and changed our corporate name to Cray Inc. In that transaction, we acquired the Cray T90, SV1, T3E and other product lines, the Cray X1 development project and related cost-sharing contract, a service organization supporting Cray supercomputers installed in about 200 sites worldwide, integration and final assembly operations, software products and related experience and expertise, approximately 775 employees, product and service inventory, real property located in Chippewa Falls, Wisconsin, and the Cray brand name. Pursuant to a technology agreement, SGI assigned to us various patents and other intellectual property and licensed to us the rights to other patents and intellectual property. We paid SGI $50.3 million in cash and issued SGI 1,000,000 shares of our common stock.

As part of the acquisition, we assumed responsibility for the cost of servicing the Cray T90 vector computers. We agreed with SGI that we would not utilize specified technology to develop specific successor products to the T3E product line, and we agreed to limit our use of SGI’s IRIX operating system to the Cray X1 product family.

Following the acquisition, we integrated our approximately 900 employees into one company, established company-wide financial, communication and other networks, moved employees out of SGI facilities into new offices, established over 20 subsidiaries for our foreign sales and service operations, either had service, sales and other contracts assigned to us or entered into new contracts with customers and vendors, continued the development of the Cray X1 system and continued to sell the then-existing Cray products, principally the Cray T3E and SV1 systems.

In May 2001 the U.S. anti-dumping order against Japanese vector supercomputers was lifted, NEC Corporation invested $25 million in us and we became a distributor of the NEC SX series of supercomputers, re-branded under the Cray name, with exclusive rights in North America and non-exclusive rights outside of North America. In 2003 NEC sold its investment in us, cancelled our exclusive rights and we became a non-exclusive distributor in North America.

In 2002 we completed the hardware development of and began selling our Cray X1 system, an “extreme performance” supercomputer designed for the high end of the supercomputer market. We are developing enhancements to this system that will increase significantly processor speed and capability, which we call the Cray X1E system, with first shipments scheduled for the second half of 2004. In mid-2002 we began a development project with Sandia National Laboratories to design and deliver in 2004 a new, high bandwidth, massively parallel processing supercomputer system called Red Storm. In October 2003 we announced that we would develop a product line based on the Red Storm system, targeting the need for highly scalable microprocessor-based Linux supercomputers with high bandwidth. This product is scheduled for shipment in the second half of 2004. In mid-2003 we began work under a contract with DARPA that supports our program to develop a commercially available system capable of sustained performance in excess of one petaflops, which we call our Cascade program. See “— Product Offerings and Projects” below. In February 2004 we announced a definitive agreement to acquire OctigaBay Systems Corporation, a privately-held development-stage company located in Vancouver, B.C. OctigaBay is developing a balanced high bandwidth system designed to be highly reliable and easy-to-use that is targeted for the midrange market. Initial commercial shipments of the OctigaBay product are not expected until late 2004, with full production ramp in 2005. See “— Agreement to Acquire OctigaBay” below.

Discussions that relate to periods prior to April 1, 2000, refer to our operations as Tera Computer Company, and discussions that relate to periods after April 1, 2000, refer to our combined operations as Cray Inc.

Since the pioneering Cray-1 system arrived in 1976, supercomputers — defined simply as the most powerful class of computers at any time — have contributed substantially to the advancement of knowledge and the quality of human life. Problems of major economic, scientific and strategic importance typically are addressed by supercomputers, which usually sell for several millions of dollars each, years before becoming tractable with less capable systems. For scientific applications, the increased need for computing power has been driven by highly challenging problems that can be solved only through numerically intensive computation. For engineering applications, high performance computers boost productivity and decrease risk and the time to market for companies and products in a broad range of industries. The U.S. government has recognized that the continued development of high performance computer systems is of critical importance to the national defense and the economic, scientific and strategic competitiveness of the United States.

Applications promising future competitive and scientific advantage demand 10 to 1,000 times more supercomputer power than anything available today, including current low bandwidth systems and existing enterprise-class and mainframe servers. There are three principal drivers to the predicted substantial growth in the high performance computing market: the continuing demand for advanced design capability, increased focus on national security issues and the recognized need for more powerful scientific research tools.

The demand for design capabilities grows seemingly without limit. Automotive companies are targeting increased passenger cabin comfort, better fuel mileage and improved safety and handling. Aerospace firms envision more efficient planes and space vehicles. Using genomic and proteomic technologies for drug development are areas of intensive research and substantial spending by research centers and biotechnology and pharmaceutical companies.

Governments have a wide range of unmet security needs, heightened by the recent emphasis on anti-terrorism. These needs primarily relate to burgeoning cryptanalysis requirements arising from a more diverse and growing number of sources and requirements for rapid and accurate analysis and fusion of information from many disparate sources. In addition, governments need better simulation and modeling of a wide range of weapons and battlefield scenarios and the computational ability to address various classified applications.

In 2002 the Japanese government announced the completion of the Japanese Earth Simulator project. This high bandwidth, vector-based system remains acknowledged as the world’s most powerful installed computer system, with a peak speed of approximately 40 teraflops and high sustained operating performance on real applications. The Japanese Earth Simulator validates our proposition that high bandwidth and sustained performance are critical, and provides Japan with the opportunity to lead in scientific research in fields such as weather and climate, geophysics, nanotechnology and metallurgy.

International Data Corporation (“IDC”), a leading industry market research firm, provides information regarding the high performance computing technical systems market, including projections. IDC segments the technical systems market based on prices, complexity and intended use, with classes for capability, enterprise, divisional and departmental systems. The capability segment is made up of systems targeted to solve the largest most demanding problems. The enterprise, divisional and departmental segments support technical applications in throughput environments and are further segmented by price: $1,000,000 and up for enterprise, $250,000 to $999,000 for divisional, and under $250,000 for departmental.

Traditionally, we have focused on the capability segment where the features we are known for — high speed processors coupled with extreme communication speed — are widely recognized as necessary to solve the world’s most difficult computing problems. There has been an increasing need for high performance supercomputing performance in the enterprise, divisional and departmental segments. With the October 2003 announcement of our plans to create a product line based on the Red Storm system we are developing for Sandia National Laboratories, our addressable market will expand into the technical enterprise segment. In addition, our recently announced proposed acquisition of OctigaBay, if completed, will further extend our reach into the divisional and departmental segments. We expect these two developments, when completed, will effectively quadruple our addressable market by 2005.

According to IDC, the 2002 capability market totaled approximately $1.0 billion and is projected to grow at an annual rate of about 2.4% through 2007. In this market segment, we grew our market share significantly, from 4% in 2001 to almost 14% in 2002. Leading the capability segment was IBM with approximately 40% market share, followed by Hewlett-Packard with 26%. The annual revenue for capability class systems historically has fluctuated as much as 25% due to new product introductions, large system procurements and government funding cycles.

The technical enterprise market totaled approximately $785 million in 2001 and, according to IDC, is expected to grow at an annual rate of 8.2% through 2007. The combined divisional and departmental markets, which we expect to enter in 2005, was estimated to total $2.9 billion in 2002 with an annual growth rate of approximately 6.5% through 2007.

According to 2002 data from IDC, the overall technical systems market is dominated by usage-based segments that have been our traditional targets. Approximately 98% of the total addressable market for technical systems is in scientific research, classified/defense, design/engineering, life sciences, geoscience, geo-engineering and simulation.

Scientific Research. This sector includes government laboratories and research centers that may also collaborate with university consortia to reach their objectives. These centers investigate computational modeling of a broad range of physical phenomena in such fields as astrophysics, chemistry, materials science, nuclear fusion and particle physics. Weather forecasting and climate modeling comprise about one-fourth of this market. The scientific research sector requires supercomputers with increasing levels of throughput and faster turn-around time, system robustness and the ability to process large volumes of data. With the success of the Japanese Earth Simulator, the U.S. Department of Energy has indicated its intention to support a competitive U.S. response. If this initiative is funded, the revenue in the scientific research segment should increase accordingly.

Classified/ Defense. According to IDC forecasts, the long-term spending on national defense and homeland security is expected to increase as a result of the events of September 2001 and related anti-terrorism initiatives. The major effect will be an increase in both the number and size of systems purchased for computational uses in the classified and anti-terrorism arenas.

Design Engineering. Simulation of new products before they are built is an invaluable industrial tool. The automotive sector uses simulation to design lighter, safer and more durable vehicles. In the aerospace sector, software running on supercomputers simulates flight dynamics as well as aspects similar to those of the automotive sector. Government agencies such as NASA and the Department of Defense employ these techniques to improve design effectiveness, improve product quality and decrease the time to deployment.

Life Sciences. Since the mapping of the human genome, there has been an explosion in the volume of genomic and proteomic data available. High performance computers are used to predict molecular structure at various levels of detail based on these data and to search genomic and proteomic data for structural similarities among and across individuals and species.

Other. A small number of customers in scientific industries, such as geosciences, which includes petroleum, geoengineering and other engineering functions, have objectives and application needs not addressed by widely used application programs and require the use of supercomputers.

Ironically, despite the demand for increased supercomputer power, supercomputers capable of exploiting these new opportunities have become rare. Today’s supercomputer market is replete with low bandwidth cluster systems that loosely link together multiple commodity servers or personal computers by means of commercially available interconnect products. Because these systems are measured and priced based upon the number of transistors they contain, they are sometimes referred to as “Type T” systems. In Type T systems, each processor typically is directly connected to its own private (“local”) memory and the programmer must manage the movement of data among memory units and processors. As a result, computer systems relying on this architecture can be difficult to program. Given their low bandwidth, these systems are best suited for applications that can be partitioned easily into discrete tasks that do not need to communicate often with each other. Vendors of low bandwidth Type T systems, such as IBM, design and build their processors and systems to meet the requirements of their larger, more commercial computer markets — for servers and personal computers rather than for the benefit of supercomputer users. These vendors’ processors and memory systems do not have the internal bandwidth to communicate and process data at the speeds necessary to address today’s most challenging supercomputer problems. Low bandwidth Type T systems can offer greater performance and price/performance advantages on small problems and larger problems lacking communications complexity, but are inefficient for the most demanding and important challenges.

Nevertheless, the supercomputer market largely filled with Type T systems for several reasons. Type T systems handle less challenging problems well. Secondly, the U.S. scientific, engineering and government users have had to turn to these systems in recent years for their more difficult problems primarily because they had no alternative. The SGI acquisition of Cray Research in 1996 and the imposition by the U.S. government in 1997 of anti-dumping duties on Japanese vector supercomputer vendors combined to eliminate the availability of high bandwidth vector supercomputers to U.S. users. The SGI acquisition also resulted in the

cancellation of the successor to the Cray T3E, the only commercially available high bandwidth, non-vector product. With no competitor planning to offer next-generation high bandwidth systems in the United States, U.S. interest in investing in these systems diminished substantially.

The gap between need and availability for high bandwidth systems did not go unrecognized. In a report to the President’s Information Technology Committee, a leading industrial supercomputer user observed in 1998 that, “The high performance computing industry in the United States today appears almost as if someone hit the pause button. We’re seeing a reduction in innovation.” A December 2000 report from the U.S. climate researchers to the White House Office of Science and Technology Policy noted that, “Parallel computers manufactured in the U.S., often with distributed memory [i.e., Type T systems], are difficult to use — There are intrinsic limitations to the ability of climate-research algorithms to achieve high levels of performance on these computers.” Other scientists noted that using tens of thousands of commodity chips may provide adequate capacity (peak flop rates), but not adequate capability, because of lack of memory bandwidth.

We are dedicated solely to the high performance computer market. We believe that by concentrating our product roadmap on high bandwidth interconnect systems and highly capable processors (whether developed by ourselves or others), we are in the best position to provide supercomputer systems with high sustained operating performance that meet the market’s most demanding needs.

The greatest differentiator between our systems and Type T systems, such as clusters, is bandwidth. When we speak of “bandwidth,” we mean the ability of processors to communicate with the system’s memory, with other processors and with input/output (“I/ O”) connections. Because our systems employ more connections, or wires, we package these connections more densely than our major competitors, and we transfer data through these connections at very high rates; our supercomputers are able to handle more data at higher speeds. As our systems are optimized for bandwidth and internal communications, they are sometimes referred to as “Type C” systems because they emphasize communication capabilities rather than transistors.

Type C systems are important because the world’s most challenging scientific and technical computing problems require many processors to communicate with each other frequently during computation. These processors need to have fast access to large memory and quantities of data. Low bandwidth microprocessor-based Type T systems are not designed for these demanding requirements. They do not support high bandwidth communications and therefore cannot deliver the performance necessary for these critical applications.

Our high performance computer systems are designed to provide high actual sustained performance on difficult computational problems. Theoretical peak performance is the highest possible speed at which a computer system can operate (obtained simply by multiplying the number of processors by the designed rated speed of each processor), and is always a theoretical number. Sustained performance, always lower than peak, is the actual speed at which a supercomputer system operates running an application program. Many Type T systems offer high theoretical peak performance due to their low internal bandwidth and distributed memory; however, their performance on complex applications frequently is a small fraction of their theoretical peak performance. While sustained performance may vary widely on different applications, our Cray X1 system generally operates on a sustained basis from 3 to 10 times that of competitive systems. Large cluster, or Type T, systems generally operate at less than 10% of their theoretical peak performance and, as these systems become larger, their efficiency declines even further.

We expect our systems to provide price/performance advantages over low bandwidth cluster systems when performance on real applications used at supercomputer sites is taken into account. In addition, our systems typically use far less electric power and occupy less space than cluster systems and, as a result, our systems have significantly lower costs of operation. And since our systems offer greater capability — they run application programs faster — they provide greater operating efficiency to the user.

Our mission is to become the premier provider of supercomputer solutions for our customers. Key elements of our strategy include:

Focus on high performance computer systems with high bandwidth that run customer applications at high sustained speeds. Our systems are designed to process very large quantities of data quickly and to provide high actual performance on the most difficult computational problems.

Leverage our strong brand, reputation and pioneering position to increase our market share. Cray Research introduced the first supercomputers more than 25 years ago, and we have remained focused solely on the high performance computer market. We intend to leverage our strong Cray brand and reputation to increase our share of the government, industrial and academic markets for supercomputers.

Pursue an aggressive research and development plan to implement our product roadmap. We plan to continue to devote a substantial portion of our resources to research and development activities that lead to supercomputers with higher speed and increased usability characteristics. We currently participate in government research and development programs that co-fund our Cray X1, Cray X1E and Black Widow programs and our Red Storm and Cascade projects. We expect that these and future activities will create technologies that we can use to meet the needs of our customers.

Build relationships with key researchers to penetrate emerging government and industrial markets. The most challenging problems require far more computing power than is currently available. We are developing relationships with government and industrial researchers and users to understand their needs for increased speed and for other supercomputer characteristics that would allow them to solve these problems.

Our target markets for 2004 and 2005 principally include the government/classified, scientific research, weather/environmental, automotive and aerospace, and life sciences markets. In certain of our targeted markets, such as the government/classified and scientific research markets, customers have their own application programs and are accustomed to using new, less proven systems. Other target customers, such as automotive and aerospace firms and some governmental agencies, require third-party application programs in production environments. We continue to devote significant resources to porting widely used third-party application programs to the Cray X1 and X1E systems to expand their market.

Government agencies have represented a significant segment for Cray Research and ourselves for many years. Certain governmental departments continue to provide partial funding support for our research and development efforts to meet their objectives. We expect long-term spending on national security and defense to increase. Current and target customers include Department of Defense classified customers and the Department of Energy, which funds the Sandia National Laboratories, Los Alamos National Laboratory and Lawrence Livermore National Laboratory, and certain foreign counterparts.

This segment includes both unclassified governmental and academic research laboratories and centers. The success of the Japanese Earth Simulator has spurred increased interest in Type C supercomputers in basic research in areas such as climate and physics. The Department of Defense, through its Defense Modernization Program, funds a number of research organizations. Network Computing Services, Inc., the system integrator for the Army High Performance Computing Research Center in Minneapolis, and the Arctic Region Supercomputing Center in Fairbanks, for example, were early purchasers of our Cray X1 system. The Office of Science in the Department of Energy, which funds the Oak Ridge National Laboratory, Argonne National Laboratory and National Energy Research Scientific Computing Center, is a key target customer as is the National Aeronautics and Space Administration. Oak Ridge National Laboratory is a significant customer for Cray X1 and X1E systems and related services.

While short-term weather forecasting has largely moved to low bandwidth cluster systems, more challenging climate modeling applications require increasing speed and larger volumes of data and thus are targets for our Type C systems. The success of the Japanese Earth Simulator has spurred interest in high bandwidth systems in this segment. Cray supercomputers are used in weather centers worldwide, from the United Kingdom to China. We have announced a sale of a Cray X1 system to the Spanish National Institute of Meteorology, and we are pursuing proposals at weather and climate centers in the United States and other countries. Some customers have achieved significant weather forecast improvements through the ability to run higher resolution models (rather than the standard 10-kilometer) in practical timeframes. Using a Cray X1 system, a 5-kilometer resolution model can now be run for the entire continental United States in about one and one-half hours and a 2.5-kilometer resolution model in about three and one-half hours.

These industries, a subset of the “design engineering” market segment, use supercomputers to design lighter, safer and more durable vehicles as well as to study wind noise and airflow around the vehicle. Several of the major automobile companies and aerospace companies are Cray customers. We have installed a Cray X1 system at The Boeing Company, which will use the system primarily to run structural analysis and computational fluid dynamics codes. We are pursuing proposals with other customers in this market. The Army High Performance Computing Research Center has achieved sustained performance of a Cray X1 system of over one teraflops (one trillion floating point operations per second) on an unstructured finite element method fluid dynamics problem. This computation was performed on a mesh containing 2.1 billion tetrahedral elements and calculated the fluid flow around an unmanned aerial vehicle.

While we do not expect this to be a significant market for us in the near term, we believe this emerging segment will contribute to our long-term growth. We currently have a system used for computational drug design at a drug manufacturer and ongoing life sciences collaborative efforts with various laboratories. In addition, the Arctic Region Supercomputing Center and Oak Ridge National Laboratory are using the Cray X1 systems for life sciences projects as will Sandia National Laboratories with the Red Storm system.

Our high performance computer products provide high bandwidth and other capabilities needed for exploiting new and existing market opportunities. Among supercomputer vendors, we offer the largest variety of products and services in order to address the broadest range of customer requirements and market segments. The decisions to commercialize the Red Storm system and to acquire OctigaBay further this strategy. Our goal is to bring major enhancements and/or new projects to market every eighteen to twenty-four months.

With the Cray X1 system as the cornerstone, we now have developed a product roadmap of high performance computer systems that stretches past 2010, with a goal of then delivering systems capable of running a variety of challenging applications at sustained speeds in excess of one petaflops (1,000 trillion floating point operations per second).

In late 2002, we completed hardware development of the new Cray X1 system, which incorporates in its design both vector processing capabilities from the long line of Cray Research vector systems and massively parallel capabilities analogous to those of our T3E system. The Cray X1 system is an “extreme performance” supercomputer aimed at the high end of the vector processing market and the high end of the market for massively parallel systems. We commenced delivering production systems late in the fourth quarter of 2002. In 2003, we enhanced the Cray X1 system hardware and software, ported application programs to provide the features and stability required in a production environment by governmental and industrial users, and

delivered ever larger integrated systems. Our selling focus for the Cray X1 system covers a range of peak performance from 200 gigaflops to multiple tens of teraflops. Various U.S. and foreign governmental agencies were early customers of the Cray X1 system.

We are developing enhancements to the Cray X1 system — the Cray X1E system — which will significantly increase processor speed and capability. We will be able to add these enhancements to Cray X1 systems in the field. We currently plan to begin shipping Cray X1E systems in the second half of 2004.

Following the Cray X1 product family will be the product family code-named Black Widow, planned to be introduced as an initial system followed by two major upgrades. Black Widow systems will have an instruction set compatible with the Cray X1 and X1E systems. We expect that the initial Black Widow systems will have a peak performance of several hundred teraflops that, with two enhancements, will grow to a peak performance in excess of one petaflops.

By 2010 our goal is to have high performance computer systems operating applications at sustained speeds in excess of several petaflops. We expect three major programs or projects will influence these future systems in addition to our planned products: the Red Storm project with Sandia National Laboratories, our multithreaded technology represented by the Cray MTA-2 and our Cascade project. We will utilize advancements in operating systems, programming tools, interconnect systems and other features from these programs and projects into the products on our product roadmap.

In mid-2002 we contracted with Sandia National Laboratories to design and deliver a new massively parallel 40-teraop processing system, called Red Storm, that will use 10,000 Opteron^TM processors from Advanced Micro Devices connected via our proprietary low-latency, high bandwidth, three-dimensional interconnect network based on HyperTransport^TM technology, coupled with our custom interconnect technology. The Red Storm project involves critical network and Linux-based operating system development.

In October 2003 we announced that we would develop a product line based on the Red Storm system that targets the need for highly scalable microprocessor-based Linux supercomputers with high bandwidth in the capability and enterprise market segments. This product line will be designed to be more efficient and cost-effective for challenging problems and workloads than cluster systems now available in the marketplace. We currently plan to begin shipments of early versions of this new product in the second half of 2004.

We were formed originally under the name Tera Computer Company to pursue a significant breakthrough in high performance computing by developing a scalable uniform shared memory system that utilizes a multithreaded architecture and a high bandwidth interconnection network. In 2002 we delivered a 40-processor MTA-2 system to the Naval Research Laboratories (“NRL”), which makes this system available for investigative purposes by its own researchers and to the Department of Defense national research community. The Cray MTA-2 is aimed at new applications not well served by vector or cluster systems, such as dynamically adaptive meshes, data sorting and problems benefiting from advanced scalability, large uniform shared memory and easier parallel programming. We are pursuing further development of this system and architecture with NRL.

In mid-2002 we signed an agreement with DARPA to initiate phase 1 of an advanced research program leading to the development of a commercially available system capable of running with sustained performance in excess of one petaflops by 2010. In addition to having high sustained performance, the resulting system is to

be designed to be much easier to program, more broadly applicable, and more robust than current designs. In mid-2003 we signed a phase 2 agreement with DARPA that will provide us and our research partners, Stanford University, California Institute of Technology/Jet Propulsion Laboratories and the University of Notre Dame, $49.9 million over three years to investigate advanced design concepts for the petaflops system. IBM and Hewlett-Packard also received similar awards. In mid-2006 DARPA plans to select up to two vendors for the final full-scale development phase with initial prototype deliveries scheduled for 2010.

We also market one classic vector system, the Cray SX-6. Pursuant to our distribution agreement with NEC, we currently market on a non-exclusive basis the NEC SX-6 system, rebranded as the Cray SX-6, to industrial, academic and governmental customers requiring intense computing power, very large high performance memory and high I/O rates on a vector platform. These systems offer high reliability in a balanced, commercial quality system. The selling focus for the Cray SX-6 supercomputers is from 16 to 64 gigaflops, with selling prices ranging from $1.0 million to $3 million. We have sold several Cray SX-6 systems to Canadian customers.

Our high performance computing services organization supports our emphasis on providing solutions rather than just computer systems to our customers. Our high-performance computing services team provides consulting, integration of Cray products and cluster solutions, custom hardware and software engineering, advanced computer training, site engineering, data center operation and computing-on-demand services. These services leverage our reputation and skills for services and industry technical leadership.

Our leadership in the high performance computer industry depends on successful development and introduction of new products and enhancements to existing products. Our research and development activities are focused on system architecture, hardware and software necessary to implement our product roadmap.

We are the only company in the world to provide systems that use or combine all three of the basic high performance computer architectures — vectors, massively parallel and multithreading.

Cray Research pioneered the use of vector systems, from the Cray-1 to the Cray C90 and T90 systems. These systems typically use a moderate number (one to 32) of very fast custom processors in connection with a shared memory. Vector processing has proven to be highly effective for many scientific and engineering application programs which over the years have been written to maximize the number of long vectors. Traditional vector systems do not scale effectively (that is, increase performance by increasing the number of processors) past a limited number of processors. We currently market one classic vector supercomputer, the Cray SX-6 system.

Massively parallel processing architectures typically link tens, hundreds or thousands of standard or commodity processors to act either on multiple tasks at the same time or together in concert on a single computationally-intensive task. Type T systems connect each processor directly to its own private memory and the programmer must manage the movement of data among memory units and processors. Consequently these systems can be difficult to program. Type C massively parallel systems, unlike low bandwidth clusters, have high bandwidth and low latency interconnect systems and are said to be “tightly coupled” — the Cray T3E, Red Storm and the OctigaBay product are examples of balanced high bandwidth purpose built systems that employ standard microprocessors.

The Cray X1 system is revolutionary in that it is the first supercomputer that combines the attributes of both vector and high bandwidth massively parallel systems. The Cray X1 system has up to 64 processors per

cabinet and a shared memory. The Cray X1 system can run small problems as a vector processor would or, by focusing many processors on a task, the Cray X1 system operates as a massively parallel system with a system-wide shared memory and a single-system image. The Cray X1 system is designed to provide efficient scalability and high bandwidth to run complex applications at high sustained speeds. The Cray X1E system furthers this architectural design with increased processor speed and capability.

Our MTA-2 project for NRL is designed to have sustainable high speed, be broadly applicable and easy to program, provide scalability as systems increase in size and have balanced I/O capability. The multithreading processors make the MTA-2 system latency tolerant and, with the system’s flat shared memory, able to address data anywhere in the system.

We have extensive experience in designing all of the components of high performance computer systems — the processors, the interconnect system and controls, the I/O system and the supporting cooling infrastructure — to operate together. Our hardware research and development experience includes:


	•	Integrated circuit design — we have experience in designing custom and standard cell integrated circuits. Our processors and other integrated circuits have special features that let them use the high available memory bandwidth efficiently. We work closely with our suppliers to take advantage of the latest advances in high speed, high density integrated circuit technology.

	•	High speed interconnect systems — we design high speed interconnect systems using a combination of conventional and microwave circuits, high density connectors and carefully chosen transmission media together with complex memory and cache controls to operate with our network protocols and highly optimized logic design. We are investigating the use of optical interconnects for future systems.

	•	Printed circuit board design — our printed circuit boards are some of the most sophisticated in the world, often more than 40 layers packed with wires and inter-layer connections.

	•	System I/O — we design high performance I/O interfaces that deliver high bandwidth transfer rates and large capacity storage capabilities using low cost devices in highly reliable configurations.

	•	Packaging and cooling — we use very dense packaging in order to produce systems with the necessary bandwidth at reasonable costs. This generates more heat per unit volume. We use specialized cooling techniques to address this issue, including immersion, conductive and spray cooling using various liquids and high volume air cooling.

	•	Fault tolerance — we design our systems to be tolerant of component failure. As individual components fail, our systems operate with minimal adverse performance impact due to designed alternative circuits and paths. We closely coordinate our hardware and operating system design with field service requirements for fast repair with minimal impact to users.

We design and maintain our system software internally. We support multiple operating systems, although all are based on UNIX. The Cray X1 operating system is UNIX-based with common UNICOS extensions. We offer a UNIX-based system called Cray MTX for the Cray MTA-2 system. The Cray SX-6 system and successors use NEC’s SUPER-UX operating system, also based on UNIX. Our Red Storm systems and the OctigaBay product will use Linux-based operating systems.

We continue to design and build highly optimized programming environments and performance management diagnostic software products that allow our customers to obtain maximum benefit from our systems. In addition to supporting third-party applications, we develop advanced algorithms and other approaches to improving application performance. We also purchase or license software technologies from third parties when necessary to provide appropriate support to our customers, while focusing on our own resources where we add the highest value.

Our extensive worldwide maintenance and support systems provide us with a competitive advantage and a predictable flow of revenue and cash. Support services are provided under separate maintenance contracts with our customers. These contracts generally provide for support services on an annual basis, although some cover multiple years. While most customers pay for support monthly, others pay on a quarterly or annual basis.

Our employees providing these services include field service engineers, product and applications specialists and product support engineers. They are supported by a central support services group located in Chippewa Falls, Wisconsin. On December 31, 2003, we had 92 field support personnel in the United States and Canada, another 60 support personnel in other countries and 68 employees providing central support services. Most of our support engineers are based at customer sites and thus have knowledge of the customer’s requirements for system and application program performance.

We primarily sell our products through a direct sales force that operates throughout the United States and in Europe, Canada, Japan and Asia-Pacific. We serve smaller foreign markets through sales representatives.

As of December 31, 2003, we had 50 sales staff, including sales representatives, sales managers, pre-sale analysts and administrative personnel located in the United States and Canada and 46 sales staff located overseas.

If the OctigaBay acquisition is consummated, we will explore the use of additional sales channels and additional sales personnel for the OctigaBay product.

Our marketing staff has a strategic focus on our target markets and those solutions that will facilitate our customers’ success in solving their most challenging scientific and engineering problems. On December 31, 2003, we had 26 employees in our marketing group.

In 2003, one customer, Oak Ridge National Laboratories, accounted for 11% of our total revenue. No single end-user customer accounted for 10% or more of our revenue in 2002 and 2001. Agencies of the United States government, both directly and indirectly through system integrators and other resellers, accounted for approximately 74% of our 2003 revenue, 79% of our 2002 revenue and 85% of our 2001 revenue. Information with respect to our international operations and export sales is set forth in Note 15 of the Notes to the Consolidated Financial Statements.

While we design many of the hardware components for all of our products, we subcontract the manufacture of these components, including integrated circuits, printed circuit boards, flex circuits, memory modules, machined enclosures and support structures, cooling systems, high performance cables and other items to third-party suppliers. Our strategy is to avoid the large capital commitment and overhead associated with establishing full-scale manufacturing facilities and to maintain the flexibility to adopt new technologies as they become available without the risk of equipment obsolescence. We perform final system integration and testing of our hardware systems.

Our manufacturing facilities are located in Chippewa Falls, Wisconsin. At December 31, 2003, we had 110 full-time employees in manufacturing.

Our systems incorporate some components that are available from one or limited sources. Key components that are sole-sourced include our integrated circuits and processors, interconnect systems and memory products. We obtain integrated circuits for our Cray X1 systems from IBM and for the Red Storm project from Advanced Micro Devices, Inc. IBM also provides packaging for our Cray X1 systems and Red Storm project. We obtain custom interconnect components for our Cray X1 from InterCon Systems, Inc., and we obtain I/ O systems for our Cray X1 systems from Sun Microsystems, Inc. We obtain custom memory products for our Cray X1 systems from Samsung Semiconductor, Inc. We acquire power modules and spray

cap cooling systems for the Cray X1 from SAE Power Incorporated and Parker Hannifin Corporation, respectively. We use Celestica, Inc., to assemble our Cray X1 systems and for repair of components for our vector and Cray X1 systems.

Our procurements from these vendors are primarily through purchase orders. We have chosen to deal with sole sources in these cases because of the availability of specific technologies, economic advantages and other factors. We also have sole or limited sources for less critical components, such as peripherals, power supplies, cooling and chassis hardware. Reliance on single or limited source vendors involves several risks, including the possibility of shortages of key components, long lead times, reduced control over delivery schedules and changes in direction by vendors.

The high performance computer market is intensely competitive. The barriers to entry are high, as is the cost of remaining competitive. We compete by offering systems that have superior sustained performance, price/performance based on sustained performance and lower cost of operation coupled with our excellent post-sale service capabilities and established customer relationships.

IBM, SGI, Hewlett-Packard and Sun Microsystems offer low bandwidth massively parallel systems for the high performance market. These systems offer greater performance and price/performance on small problems and larger problems lacking complexity and offer higher theoretical peak performance.

Internationally we compete primarily with IBM, Hewlett-Packard, SGI and NEC. While the first three companies offer low bandwidth massively parallel systems, NEC offers high bandwidth vector-based systems with a large suite of ported application programs. We have non-exclusive rights to market NEC vector processing supercomputers throughout the world. Competition with NEC outside of North America is difficult due to NEC’s aggressive pricing strategies. See “Factors That Could Affect Future Results — The change by NEC Corporation of our distribution rights for the Cray SX-6 system may increase competition.”

Each of our competitors named above has substantially greater engineering, manufacturing, marketing and financial resources than we do.

We attempt to protect our trade secrets and other proprietary rights through formal agreements with our employees, customers, suppliers and consultants, and through patent protection. Although we intend to protect our rights vigorously, there can be no assurance that our contractual and other security arrangements will be successful. There can be no assurance that such arrangements will not be terminated or that we will be able to enter into similar arrangements on favorable terms if required in the future. In addition, if such agreements were breached, there can be no assurance that we would have adequate remedies for any breach.

We have a number of patents relating to our hardware and software systems. We license certain patents and other intellectual property from SGI as part of our acquisition of the Cray Research operations. These licenses contain restrictions on our use of the underlying technology, generally limiting the use to historic Cray products, vector processor computers and the Cray X1 systems. Our general policy is to seek patent protection for those inventions and improvements likely to be incorporated into our products and services or to give us a competitive advantage. While we believe our patents and applications have value, no single patent is in itself essential to us as a whole or to any of our key products. Any of our proprietary rights could be challenged, invalidated or circumvented and may not provide significant competitive advantage.

There can be no assurance that the steps we take will be adequate to protect or prevent the misappropriation of our intellectual property. We may infringe or be subject to claims that we infringe the intellectual property rights of others. Litigation may be necessary in the future to enforce patents we obtain, and to protect copyrights, trademarks, trade secrets and know-how we own, or to defend infringement claims from others. Such litigation could result in substantial expense to us and a diversion of our efforts.

As of December 31, 2003, we employed 905 employees, of whom 335 were in development and engineering, 110 were in manufacturing, 96 were in sales, 26 in marketing, 220 in service, 54 were in information systems, and 64 were in administration. We also employed 12 individuals on a temporary basis or as interns. We have no collective bargaining agreement with our employees. We have never experienced a work stoppage and believe that our employee relations are excellent.

On February 25, 2004, we signed a definitive agreement to acquire OctigaBay Systems Corporation, a privately-held development-stage company located in Vancouver, B.C. We will acquire all of the outstanding stock of OctigaBay in return for $14,925,000 in cash and the issuance of 12,733,786 shares of our common stock. We also will assume outstanding options exercisable for 408,253 shares of our common stock. The purchase price will be paid out of current funds. Upon completion of the acquisition, we expect to operate OctigaBay as a separate subsidiary under the Cray name and market its products and services under the Cray brand. OctigaBay currently has approximately 65 employees and we expect that all employees will continue with the company following the acquisition. OctigaBay is developing a balanced high bandwidth computing system designed to be highly reliable and easy to use and targeted for the midrange market. OctigaBay’s current development schedule contemplates first production units available in late 2004 with full production ramp in 2005. OctigaBay has no sales force or servicing capability, and we plan to use our existing sales and servicing resources, with potential additional sales personnel and sales channels, for these tasks.

The acquisition of OctigaBay is subject to customary approvals, a vote of its shareholders and other standard closing conditions. The transaction is expected to close before the end of April 2004. Upon consummation, we will file appropriate disclosure documents with the Securities and Exchange Commission.

The following factors should be considered in evaluating our business, operations and prospects and may affect our future results and financial condition.

Our quarterly operating results may fluctuate significantly. Our operating results are subject to significant fluctuations due to many factors. One or a few system sales may account for a substantial percentage of our quarterly revenue, and thus revenue, net income or loss and cash flow are likely to fluctuate significantly from quarter to quarter. This is due to the high average sales prices of our products and limited number of sales per quarter, the timing of purchase orders and product delivery, and our general policy of not recognizing product revenue until customers accept our products. These factors make forecasting revenue and earnings in any short-term period very difficult. While we were profitable in every quarter in 2003, profitability often depended upon customers accepting systems at the end of the quarter. Any delay in an acceptance of a system at the end of a quarter, which would have moved the associated revenue into a subsequent quarter, could have resulted in a loss for the quarter. We were able to book most of our 2003 sales of the Cray X1 systems through a few very large contracts entered into during the first quarter of 2003. We expect that sales of Cray X1 systems in 2004 will be pursuant to more numerous and smaller contracts, often in very competitive situations. We have not yet signed sales contracts covering most of our anticipated 2004 revenue from sales of Cray X1 and X1E systems. If we do not obtain such contracts, our operating results in any particular period will be adversely affected and may produce a loss. Red Storm revenue and margin may fluctuate from quarter to quarter due to our level of contract activity, including purchases of materials and potential changes in the estimates of the cost to complete. Because a number of our prospective customers receive funding from the U.S. or foreign governments, the timing of orders from our government customers may be subject to the funding schedules for the relevant government agencies as well as delays that may be experienced in competitive procurements. The timing of orders and shipments also could be affected by other events outside our control, such as:


	•	the timely availability of acceptable components in sufficient quantities to meet customer delivery schedules;


	•	changes in levels of customer capital spending;

	•	the introduction or announcement of competitive products;

	•	the receipt and timing of necessary export licenses; and

	•	currency fluctuations and international conflicts or economic crises.

We have experienced annual losses from operations prior to 2002, and we may not achieve quarterly or annual net income on a consistent basis. We experienced net losses in each full year of our development-stage operations prior to 2002. We incurred net losses of approximately $35.2 million in 2001, $25.4 million in 2000, and $34.5 million in 1999. For 2002, we had net income of $5.4 million and for 2003 we had net income of $24.7 million (before non-recurring items added another $38.5 million to net income). Whether we will achieve net income on a consistent quarterly and annual basis will depend on a number of factors, including:


	•	successfully selling the Cray X1 system, the Cray X1E system, Red Storm and other products, and the timing and funding of government purchases, especially in the United States;

	•	completing the development of the Red Storm project, the Cray X1E system and the commercial version of the Red Storm system in time for deliveries and customer acceptances in 2004 and subsequently maintaining our other development projects on schedule and within budgetary limitations;

	•	the level of revenue in any given period, including the timing of product acceptances by customers;

	•	our expense levels, particularly for research and development and manufacturing and service costs; and

	•	the terms and conditions of sale or lease for our products.

Because of the numerous factors affecting our results of operations, there can be no assurance that we will have consistent net income on a quarterly and annual basis in the future.

If the U.S. government purchases fewer supercomputers, our revenue would be reduced and our profitability would be adversely affected. Historically, sales to the U.S. government and customers primarily serving the U.S. government have represented a significant market for supercomputers. From January 1, 2001, through December 31, 2002, approximately 79% of our product revenue was derived from sales to various agencies of the U.S. government; in 2003, approximately 83% of our product revenue was derived from such sales. Our sales of Cray X1 systems to date have been largely to government agencies in the United States and other countries, and we expect that will continue throughout 2004. To date, however, we have not entered into any significant new contracts for sales of the Cray X1 and Cray X1E systems. Sales to government agencies may be affected by factors outside our control, such as changes in procurement policies, budget considerations and international political developments. If the United States or other governments were to stop, reduce or delay their use and purchases of supercomputers, our revenue would be reduced which could lead to reduced profitability or a loss in future periods.

We may not be successful in completing the Red Storm project on time and on budget, which would adversely affect our earnings. Our efforts to complete the development and delivery of the Red Storm project for Sandia National Laboratories in 2004 on time and on budget are subject to significant risks. Our work is pursuant to a fixed-price contract with payment against significant monthly milestones setting out a tight development schedule and technically challenging performance requirements. Our success depends on third-party software development, some of which is to be supplied by Sandia National Laboratories. The contract is incrementally funded and is subject to future federal government appropriations. This project is lengthy and technically challenging, and requires a significant investment of engineering and other resources. Falling behind schedule or incurring cost overruns would adversely affect our capital resources and earnings.

We face significant pressure on the pricing of our products, which may result in lower margins and earnings. In 2004, our product margins will be negatively impacted by the low margins recognized on the Red Storm and Cascade development contracts. We anticipate that the product line based on the Red Storm system, which is targeted for markets now largely served by clustered systems, generally will have lower

margins than our vector-based products due to a more competitive market. If our acquisition of OctigaBay Systems Corporation is consummated, we expect that the OctigaBay system will sell at similarly lower margins. We also face margin pressure for our Cray X1 systems in 2004, particularly as we near introduction of the Cray X1E system. We may grant favorable pricing for large multi-system contracts and to obtain strategic accounts. We may not be able to sell sufficient additional systems and produce more revenue to offset lower gross margins; if not, our earnings would be reduced.

If application programs were not successfully ported to the Cray X1 system, we would have difficulty selling these systems to a number of customers. To make sales of the Cray X1 system, including the planned Cray X1E upgrade system, in the automotive, aerospace, chemistry and other engineering and technical markets, including certain governmental users, we must have application programs ported to the Cray X1 system and tuned so that they will achieve high performance. The Cray X1 system has a new architecture that makes porting and tuning of application programs difficult. These application programs are owned in some instances by independent software vendors and in others by potential customers. We must induce these vendors and customers to undertake this activity. The relatively low volume of supercomputer sales makes it difficult for us to attract independent software vendors to make this investment. We also modify and rewrite thirty-party and customer specific application programs to run on the Cray X1 and X1E systems. There can be no assurance that we will be able to induce the third-party vendors and customers to rewrite their applications or that we will rewrite successfully third-party and customer specific applications for use on those systems.

Our inability to overcome the technical challenges of completing the development of our high performance computer systems would adversely affect our revenue and earnings in 2004 and beyond. We expect that our success in 2004 and in the following years depends on completing the Red Storm project; adapting the Red Storm concept to a highly scalable microprocessor-based high bandwidth Linux system for the governmental, industrial and academic markets; developing the Cray X1E system as a significant enhancement to the Cray X1 system and completing the development of the OctigaBay system and successfully selling it in the midrange market. In subsequent years we must develop further enhancements to the Red Storm system for the high end market, and to the product adapted from it, along with the OctigaBay product, for the midrange market, and develop the Black Widow system as a successor to the Cray X1 and X1E systems. We also must increasingly integrate our product lines so that we use as many common components and systems as possible. These development efforts are lengthy and technically challenging processes, and require a significant investment of capital, engineering and other resources. Our engineering and technical personnel resources are limited. Difficulties in delivering the initial Cray X1 systems and integrating large configurations of the Cray X1 system have introduced delays in the development schedule for the Cray X1E and Black Widow systems; we may not be successful in shortening these development schedules. Delays in completing the design of the hardware components, of software for the systems, or in integrating the full systems would make it difficult for us to develop and market these systems successfully. If we were unable to market and sell the Cray X1E and Red Storm systems in the second half of 2004, our revenue and earnings would be adversely affected. We are dependent on our outsourced vendors to manufacture components of these systems, and few companies can meet our design requirements. The failure of vendors to manufacture our components to our design specifications would delay the completion of our products. Redesign work may be costly and cause delays in the development of these systems.

If we were unable to continue to improve our Cray X1 system software, our revenue and profits would be reduced. We need to improve the reliability of system software to sell Cray X1 systems to production environment governmental and industrial customers. We continue to fix reported software problems and anticipate additional software problems to be reported in the future.

Our stock price is volatile. The stock market has been and is subject to price and volume fluctuations that particularly affect the market prices for small capitalization, high technology companies like us. The trading price of our common stock is subject to significant fluctuations in response to many factors, including our quarterly operating results, changes in analysts’ estimates, our future capital raising activities, announcements of technological innovations by us or our competitors and general conditions in our industry.

If we lose government support for supercomputer systems, our capital requirements would increase and our ability to conduct research and development would decrease. A few government agencies and research laboratories fund a significant portion of our development efforts. Agencies of the U.S. government historically have facilitated the development of, and have constituted a market for, new and enhanced very high performance computer systems, including the current Cray X1 system and our planned Cray X1E system, Black Widow, Cascade and Red Storm development projects. U.S. government agencies may delay or decrease funding of these development efforts due to change of priorities, international political developments or for any other reason. Any such decrease or delay may cause an increased need for capital, increase significantly our research and development expenditures and adversely impact our ability to implement our product roadmap.

The failure to integrate the planned acquisition of OctigaBay Systems Corporation could adversely affect our business. If we complete the acquisition of OctigaBay Systems Corporation, we will add an additional product line, 65 employees and a fourth major office location, our first outside of the United States. We will need to increase our sales force and develop new sales channels to handle the OctigaBay product, develop a different approach for providing customer servicing of the OctigaBay product, integrate our financial and information systems and over time integrate our development programs. These changes may place a significant strain on our management resources. The failure to retain the current OctigaBay engineers and employees would adversely affect the development schedule and delay introduction of the OctigaBay system. Difficulties in integrating our operations will divert our management’s time and resources. Failure to complete this integration successfully could cause us to increase expenditures and adversely affect our revenue and results of operations.

Procurement proposals based on theoretical peak performance reduce our ability to market our systems. Our high performance computer systems are designed to provide high actual sustained performance on difficult computational problems. Some of our competitors offer systems with higher theoretical peak performance at lower prices, although their actual sustained performance on real applications frequently is a small fraction of their theoretical peak performance. Nevertheless, a number of requests for proposals, primarily from governmental agencies in the United States and elsewhere, continue to have criteria based wholly or significantly on theoretical peak performance. Under such criteria, the price/peak performance ratio of our products compares unfavorably to the price/peak performance ratio of our competitors’ products. Unless these criteria are changed to favor actual performance, we will continue to be disadvantaged in these instances by being unable to submit competitive bids, which limits our revenue potential.

The change by NEC Corporation of our distribution rights for the Cray SX-6 system may increase competition. We market a rebranded product known as the Cray SX-6 system, which was developed and is built in Japan by NEC Corporation. This product first became available for delivery in North America in the first quarter of 2002, and we became the exclusive distributor of NEC vector supercomputer systems in North America and a non-exclusive distributor outside North America. Effective August 1, 2003, our North American distribution rights for this product became non-exclusive. Supercomputer customers in the United States have been reluctant to purchase supercomputers from non-U.S. sources, and domestic demand for the Cray SX-6 systems has been far less than we anticipated. NEC may decide to compete directly with us in North America, which could adversely affect our revenue. Outside of North America, NEC has competed aggressively based on price, and promised deliveries of its NEC SX-8 system, which it has not yet announced formally.

Lower than anticipated sales of new supercomputers would further reduce our service revenue from maintenance service contracts. High performance computer systems are typically sold with maintenance service contracts. These contracts generally are for annual periods, although some are for multi-year periods, and provide a predictable revenue base. Our revenue from maintenance service contracts has declined from approximately $125 million in 1999 to approximately $57 million in 2003. This revenue is expected to decline further, as our older systems are withdrawn from service, until a sufficient number of our new computer systems are placed in service to balance or exceed the withdrawal of our older systems.

Development of a new operating system for our Black Widow and other future products is a difficult process, and may delay the availability of our Black Widow System. We plan to develop our own UNIX-based operating system based on Linux for successor systems to the Cray X1 product family, starting with the Black Widow system. Developing a new operating system is a lengthy and difficult process, and might delay the availability of the Black Widow system.

Our reliance on third-party suppliers poses significant risks to our business and prospects. We subcontract the manufacture of substantially all of our hardware components for all of our products, including integrated circuits (processor and memory), printed circuit boards, flex circuits and power supplies, on a sole or limited source basis to third-party suppliers. We use a contract manufacturer to assemble our components for the Cray X1 and other systems. We are subject to substantial risks because of our reliance on these and other limited or sole source suppliers. For example:


	•	if a supplier did not provide components that meet our specifications in sufficient quantities, then production and sale of our systems would be delayed;

	•	if a reduction or an interruption of supply of our components occurred, either because of a significant problem by a supplier or a single-source supplier deciding to no longer provide those components to us, it could take us a considerable period of time to identify and qualify alternative suppliers to redesign our products as necessary and to begin manufacture of the redesigned components;

	•	if we were ever unable to locate a supplier for a key component, we would be unable to deliver our products;

	•	one or more suppliers could make strategic changes in their product offerings, which might delay, suspend manufacture or increase the cost of our components or systems; and

	•	some of our key suppliers are small companies with limited financial and other resources, and consequently may be more likely to experience financial and operational difficulties than larger, well-established companies.

From time to time we have experienced delays in obtaining manufactured components and completed assemblies on a timely basis and in sufficient quantities from our suppliers, which have resulted in delays in the development and production of our products.

Additional financings may be dilutive to our shareholders. We may need to raise additional equity or debt capital if we experience lower than anticipated product sales due to delays in availability of Cray X1E or Red Storm systems for delivery to customers or general economic conditions, or if we fail to receive sufficient governmental support for our products and research activities. If we are successful in our product developments and market conditions are favorable, we may consider financings to enhance our cash and working capital positions. Financings may not be available to us when needed or, if available, may not be available on satisfactory terms and may be dilutive to our shareholders.

If we are unable to compete successfully against larger, more established companies in the high performance computer market, our revenue will decline. The performance of our products may not be competitive with the computer systems offered by our competitors. Many of our competitors are established companies that are well known in the high performance computer market, including IBM, NEC, Hewlett-Packard, SGI, Dell and Sun Microsystems. Each of these competitors has broader product lines and substantially greater research, engineering, manufacturing, marketing and financial resources than we do. Periodic announcements by our competitors of new high performance computer systems (or plans for future systems) and price adjustments may reduce customer demand for our products. Most of our potential customers already own or lease very high performance computer systems. Some of our competitors offer trade-in allowances or substantial discounts to potential customers, and engage in other aggressive pricing tactics, and we have not always been able to match these sales incentives. We may be required to provide discounts to make sales or to provide lease financing for our products, which would result in a deferral of our receipt of cash for these systems. These developments would limit our revenue and resources and would reduce our ability to be profitable.

We may not compete successfully against innovative competitors or new entrants. Our market is characterized by rapidly changing technology, accelerated product obsolescence and continuously evolving industry standards. Our success will depend upon our ability to sell our current products, and to develop successor systems. We will need to introduce new products and features in a timely manner to meet evolving customer requirements. We may not succeed in these efforts. Even if we succeed, products or technologies developed by others may render our products or technologies noncompetitive or obsolete. New companies have capitalized on developments in parallel processing and increased computer performance through networking and cluster systems. Currently, these products are limited in applicability and scalability and can be difficult to program. A breakthrough in architecture or software technology could make cluster systems more attractive to our existing and potential customers. Such a breakthrough would impair our ability to sell our products and reduce our revenue.

General economic and market conditions could decrease our revenue, increase our need for cash and adversely affect our profitability. While much of our business is related to the government sector, which is less affected by short-term economic cycles, a slow-down in the overall U.S. and global economy and resultant decreases in capital expenditures have affected sales to our industrial customers and may continue to do so. Cancellations or delays in purchases would decrease our revenue, increase our need for working capital and adversely affect our profitability.

We may infringe or be subject to claims that we infringe the intellectual property rights of others. Third parties may assert intellectual property infringement claims against us, and such claims, if proved, could require us to pay substantial damages or to redesign our existing products. Regardless of the merits, any claim of infringement requires management attention and causes us to incur significant expense to defend.

If we cannot attract, retain and motivate key personnel, we may be unable to implement effectively our business plan. Our success also depends in large part upon our ability to attract, retain and motivate highly skilled management, technical and marketing and sales personnel. Recruitment for highly skilled management, technical, marketing and sales personnel is very competitive, and we may not be successful in attracting and retaining such personnel.

We may not be able to protect our proprietary information and rights adequately. We rely on a combination of patent, copyright and trade secret protection, nondisclosure agreements and licensing arrangements to establish, protect and enforce our proprietary information and rights. We have a number of patents and have additional applications pending. There can be no assurance, however, that patents will be issued from the pending applications or that any issued patents will protect adequately those aspects of our technology to which such patents will relate. Despite our efforts to safeguard and maintain our proprietary rights, we cannot be certain that we will succeed in doing so or that our competitors will not independently develop or patent technologies that are substantially equivalent or superior to our technologies. The laws of some countries do not protect intellectual property rights to the same extent or in the same manner as do the laws of the United States. Although we continue to implement protective measures and intend to defend our proprietary rights vigorously, these efforts may not be successful.

U.S. export controls could hinder our ability to make sales to foreign customers and our futur


Washington		93-0962605
(State or Other Jurisdiction of Incorporation or Organization)		(I.R.S. Employer Identification No.)

411 First Avenue South, Suite 600 Seattle, Washington (Address of Principal Executive Office)		98104-2860 (Zip Code)


		Page

PART I
Item 1.	Business		3
Item 2.	Properties		23
Item 3.	Legal Proceedings		23
Item 4.	Submission of Matters to a Vote of Security Holders		23
Item E.O	Executive Officers of the Company		23

PART II
Item 5.	Market for the Company’s Common Equity and Related Stockholder Matters		26
Item 6.	Selected Financial Data		26
Item 7.	Management’s Discussion and Analysis of Financial Condition and Results of Operations		27
Item 7A	Quantitative and Qualitative Disclosures About Market Risk		35
Item 8.	Financial Statements and Supplementary Data		35
Item 9.	Changes in and Disagreements with Accountants on Accounting and Financial Disclosure		36
Item 9A	Controls and Procedures		36

PART III
Item 10.	Directors and Executive Officers of the Company		37
Item 11.	Executive Compensation		37
Item 12.	Security Ownership of Certain Beneficial Owners and Management and Related Stockholder Matters		37
Item 13.	Certain Relationships and Related Transactions		37
Item 14.	Principal Accountant Fees and Services		37

PART IV
Item 15.	Exhibits, Financial Statement Schedules, and Reports on Form 8-K		37


	o	TRANSITION REPORT PURSUANT TO SECTION 13 OR 15(d) OF THE


Item 1.	*Business*


	*Tera Computer*


	*Cray Research*


	*Cray Research Acquisition*