SECURITIES AND EXCHANGE COMMISSION

Washington, D.C. 20549

FORM 10-K

Commission File Number: 0-26820

CRAY INC.

(Exact name of registrant as specified in its charter)

Registrant’s Telephone Number, Including Area Code: (206) 701-2000

Securities Registered Pursuant to Section 12(b) of the Exchange Act: NONE

Securities Registered Pursuant to Section 12(g) of the Exchange Act:

Common Stock, $.01 par value

      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 þ     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 þ     No o

      The aggregate market value of the Common Stock held by non-affiliates of the Registrant as of June 30, 2004, was approximately $553,000,000, based upon the closing price of $6.62 reported for such date on the Nasdaq National Market System.

      As of March 1, 2005, there were 87,638,651 shares of Common Stock issued and outstanding.

DOCUMENTS INCORPORATED BY REFERENCE

      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 11, 2005, are incorporated by reference into Part III.

CRAY INC.

FORM 10-K

For Fiscal Year Ended December 31, 2004

INDEX

      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 MTA, Cray MTA-2, Cray MTX, Cray X1, Cray X1E, Cray XT3 and Cray XD1 are trademarks of Cray Inc. Other trademarks used in this report are the property of their respective owners.

Forward-Looking Statements

      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, revenue 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 and annual net losses; uneven and possibly negative cash flow from operations; the timing of product orders, deliveries and customer acceptances; the timely development, production and acceptance of products and services and their features, including stable system software for our Cray XT3 systems; the timing and level of governmental support for supercomputers; the market impact of a conclusion that our internal control over financial reporting is ineffective; 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 reports filed with the Securities and Exchange Commission (“SEC” or “Commission”), including but not limited to the items discussed in “Factors That Could Affect Future Results” set forth 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.

PART I

General

      We design, develop, market and service high performance computer systems, commonly known as supercomputers. These systems provide capability and capacity far beyond typical server-based computer systems and address challenging scientific and engineering computing problems for government, industry and academia.

      We are dedicated solely to the high performance computing market. We have concentrated our product roadmap on building purpose-built, balanced systems combining highly capable processors (whether developed by ourselves or others) with rapid interconnect and communications capabilities throughout the entire computing system, not solely processor-to-processor. We believe we are in the best position to meet the high performance computer market’s demanding needs by providing superior supercomputer systems with performance and cost advantages over low-bandwidth and cluster systems when sustained performance on challenging applications and workloads and total cost of ownership are taken into account.

      Our 2004 product revenue primarily came from sales of our Cray X1 system and government funding for our Red Storm and Cascade development projects. In the second half of 2004, we were in transition from offering one product, the Cray X1 system, to the three products that we currently offer: the Cray X1E, XT3 and XD1 systems. We also derive revenue from providing maintenance and support services to the worldwide installed base of Cray computers and professional services that leverage our technical knowledge. See “Product Offerings, Projects and Services” below.

      Our revenue, net income or loss and cash balances are likely to fluctuate significantly from quarter to quarter and within a quarter due to the high average sales prices and limited number of sales of our larger products, the timing of purchase orders and product deliveries, our general policy of not recognizing product revenue for our larger systems until customer acceptance and other contractual provisions have been fulfilled, and the uncertain timing of payments for product sales, maintenance services, government research and development funding, and inventory.

      We were incorporated under the laws of the State of Washington in December 1987. Our corporate headquarter 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. The contents of our web site are not incorporated by reference into this Annual Report on Form 10-K or our other SEC reports and filings.

Our History

      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 completed an initial public offering in 1995. In 2000 we were still in the development stage with limited revenue 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 T3E system, 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 the U.S. market. These developments combined to eliminate the availability of high-bandwidth computer systems in the United States high performance computing market, greatly diminishing the U.S. market’s access to these systems. 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 Cray SV2). In 1999, having moved a substantial number of established Cray Research customers to its Origin product line, 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 worldwide service organization supporting Cray supercomputers installed at customer sites, 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 Cray 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, 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 world-wide.

      In 2001 and 2002 we focused our development efforts on the Cray X1 system; initial deliveries of the Cray X1 system began in late 2002. The Cray X1 system, designed for the high end of the supercomputer market, was the only new product we were selling in 2003 and the first three quarters of 2004. In 2004 we developed the Cray X1E system that significantly increased the system’s processor speed and capability; the first Cray X1E system customer shipment occurred at the end of 2004.

      In mid-2002 we began our Red Storm development project with Sandia National Laboratories to design and deliver a new high-bandwidth, massively parallel processing supercomputer system. The Red Storm hardware system was shipped in installments to Sandia, with the final hardware shipment in the first quarter of 2005, subject to subsequent installation of certain component upgrades when they become available. We are currently developing and installing system software designed to run applications programs successfully across the entire 10,000-processor system. The Red Storm project provides the basis for a commercial product, our Cray XT3 system, targeting the need for highly scalable, high-bandwidth, microprocessor-based supercomputers using a Linux-based operating system. The Cray XT3 system initial customer shipment occurred in the fourth quarter of 2004, and full production ramp is planned for 2005.

      In mid-2002 we also began work under a contract awarded by the Defense Advanced Research Projects Agency (“DARPA”) to develop a system capable of sustained performance in excess of one petaflops (1,000 trillion floating point operations per second), which we call our Cascade program. We are currently involved in phase 2 (the research phase) of this project, which ends in mid-2006.

      On April 1, 2004, we acquired OctigaBay Systems Corporation, a privately-held company located in Burnaby, B.C. OctigaBay was developing a balanced, high-bandwidth system, designed to be highly reliable and easy-to-use, targeted for the midrange market. We renamed OctigaBay Systems Corporation as Cray Canada Inc. and renamed the OctigaBay product as the Cray XD1 system. Initial commercial shipments of the Cray XD1 system began late in the third quarter of 2004, and full production ramp is planned for 2005.

      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.

The High Performance Computing Industry

      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 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. We believe there are three principal factors driving the growth in the high performance computing market: the continuing demand for advanced design and simulation capability, continuing concerns about national security issues and the recognized need to advance scientific research for domestic competitiveness of many major countries around the world.

      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 an 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 integration 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 one of the world’s most powerful installed computer system with a peak speed of approximately 40 teraflops (40 trillion floating point operations per second) and high sustained operating performance on real applications. The Japanese Earth Simulator validated our proposition that high-bandwidth and sustained performance are critical, and has provided Japan with the opportunity to lead in scientific research in fields such as weather and climate, geophysics, nanotechnology and metallurgy. The Japanese government recently declared that increased supercomputing technology was a high priority for the rest of this decade, and we believe that its stated intent to upgrade the Japanese Earth Simulator should reinforce the U.S. government’s desire to recapture and maintain supercomputing leadership.

      When we speak of “bandwidth,” we mean the ability of processors to communicate with other processors, with the system’s internal memory subsystem and with input/output (“I/ O”) connections.

      Today’s supercomputer market is replete with low-bandwidth systems and off-the-shelf commodity-based cluster systems that loosely link together multiple commodity servers or personal computers by means of commercially available interconnect products for several reasons. In recent years, the speed and capabilities of off-the-shelf interconnect systems and processors have continued to improve and independent software vendors have adapted their application codes to exploit the capabilities and partially mask the weaknesses of these systems. These systems offer significant performance and price/performance on small problems and larger problems lacking communications complexity. Secondly, the U.S. scientific, engineering and government users have had to turn to these systems for their more difficult problems primarily because they had no alternative. The imposition by the U.S. government in 1997 of anti-dumping duties on Japanese vector supercomputer vendors and the SGI cancellation of the development of successors to the Cray T90 and Cray T3E systems combined to eliminate the availability of high-bandwidth supercomputers to U.S. users. With no competitor planning to offer next-generation high-bandwidth systems in the United States, customer interest in these systems diminished substantially.

      We are dedicated solely to the high performance computer market. We differentiate ourselves from our competitors primarily by emphasizing the communication capabilities of our systems. We have concentrated our product roadmap on building purpose-built systems combining highly capable processors (whether developed by ourselves or others) with rapid interconnect and communication capabilities throughout the entire computing system. Our supercomputer systems are “balanced” in that our systems are fast not only processor-to-processor but also with memory subsystems and I/ O systems. Competitive systems may use processors with higher rated or theoretical speeds than some of ours — although at 18 gigaflops our Cray X1E processor is currently the world’s fastest — but even in those cases our systems typically outperform competing products by using their high-bandwidth communications to deliver more data to the Cray processors and keep them busier.

      As we design our supercomputer systems for the needs of the high performance computing market, we say they are “purpose-built” for this market. Vendors of low-bandwidth systems, such as IBM, design and build their processors and systems to meet the requirements of their larger commercial computer markets — for servers and personal computers — and then attempt to leverage these commercial server-based products into the supercomputer market.

      Low-bandwidth and cluster systems may offer higher theoretical peak performance than do our systems. Theoretical peak performance is the highest theoretical possible speed at which a computer system could, but never does, operate (obtained simply by multiplying the number of processors by the designed rated speed of each processor). Sustained performance, always lower than peak, is the actual speed at which a supercomputer system runs an application program. Due to their low internal bandwidth and distributed memory, however, the sustained performance of low-bandwidth and cluster systems on complex applications frequently is a small fraction, often less than 10%, of their theoretical peak performance, and as these systems become larger, their efficiency declines even further. Our systems, designed for balanced total system communications capability, provide high actual sustained performance on difficult computational problems, even though in some cases they may have a lower theoretical peak performance than competitors’ systems. While sustained performance may vary widely on different applications, our systems generally operate on a sustained basis from 1.5 to 10 times that of competitors’ systems. We expect our systems to provide price/performance advantages over low-bandwidth and cluster systems when performance on real applications is taken into account.

      The advent of the Cray X1 system in late 2002 provided the first new high-bandwidth alternative for the U.S. high-end high performance computer customers since the mid-1990’s. Our introduction in late 2004 of the Cray XT3 and XD1 systems extended the availability of high-bandwidth systems to all segments of the high performance computing market.

      Industry analyst firm, International Data Corporation (“IDC”), provides information regarding the high performance computing systems market, including historical data and projections. IDC estimates that the revenue for the entire high performance computing market totaled approximately $5.6 billion in 2003, and that the market added another $1.4 billion in 2004 for a total of $7.0 billion. IDC segments the high performance computing systems market based on prices and, at the higher end, intended use. IDC descriptions and estimates of revenue in recent years for each of these segments follow:

      Traditionally, we have focused on the capability segment of the high performance computing market where the features we are known for — high speed processors coupled with very fast communications — are widely recognized as necessary to solve the world’s most difficult computing problems. With the Cray XT3 system, our addressable market expanded into the enterprise segment. The Cray XD1 system further extends our reach into the divisional segment and parts of the departmental segment. We expect these two products will effectively quadruple our addressable market in 2005.

Our Target Market and Customers

      Our target markets for 2005 and beyond principally include the government/classified, scientific research, weather/environmental, and automotive and aerospace markets as well as exploratory opportunities into other markets such as life sciences and petroleum. 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 devote significant resources to porting widely used third-party application programs to all of our systems to expand their respective markets.

      Government agencies have represented a significant segment for Cray Research and ourselves for many years. Certain governmental departments continue to provide 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, primarily for our Cray X1E and XT3 systems, 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.

      The scientific research segment includes both unclassified governmental and academic research laboratories and centers. The success of the Japanese Earth Simulator has been important in spurring increased interest in balanced high-bandwidth supercomputers in basic research in areas such as climate and physics. The Department of Defense, through its Defense High Performance Computing 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, and the Army Center is acquiring an additional Cray X1E 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, X1E and XT3 systems and related services. The National Science Foundation, which funds the Pittsburgh Supercomputing Center, has acquired a Cray XT3 system with 10 teraflops peak performance. The Maui High Performance Computing Center, a U.S. Air Force Research Laboratory’s Directed Energy Directorate facility funded by the Defense High Performance Computing Modernization Program, has selected a Cray XD1 system with a peak performance of about 1.4 teraflops to increase the Center’s capabilities in space surveillance and image processing. Cray XD1 systems have been acquired by governmental and academic research laboratories and centers in Italy, Germany, India, United Kingdom and the United States.

      While short-term weather forecasting has largely moved to low-bandwidth and cluster systems, more challenging climate modeling applications require increasing speed and larger volumes of data and thus are targets for our high-bandwidth systems. Cray supercomputers are used in weather centers worldwide, from the United Kingdom to Korea. We have announced deliveries of Cray X1 systems with later upgrades to

Cray X1E systems to Warsaw University’s Interdisciplinary Center for Mathematical and Computational Modeling, the Spanish National Institute of Meteorology and the Korea Meteorological Administration. Using a Cray X1 system, the Army High Performance Computing Research Center ran a 5-kilometer resolution weather model for the entire continental United States and is currently validating the results of a 2.5-kilometer model. These models require eight and 64 times more computing power, respectively, than the 10-kilometer model that is the highest resolution typically used today. Scientists at the U.S. National Center for Atmospheric Research recently stated, based on their experience using Japan’s Earth Simulator and the Cray X1 system at Oak Ridge National Laboratory, that today’s vector systems deliver substantially greater performance on climate applications than other types of high performance computers.

      These industries 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 uses the system primarily to run structural analysis and computational fluid dynamics codes. The Cray XD1 system has demonstrated early impressive results on certain crash and computational fluid dynamics codes widely used in the automotive industry.

Product Offerings, Projects and Services

      Our high performance computer products provide high-bandwidth and other capabilities needed for exploiting new and existing market opportunities. Among supercomputer vendors, our intent is to offer the most comprehensive range of high-bandwidth products and related services to the high performance computing market. Our decisions to develop and market both the Cray XT3 system and the Cray XD1 system further this strategy. Our goal is to bring major enhancements and/or new products to market every 12 to 24 months. With the availability of the Cray X1E, XT3 and XD1 systems, we now offer the most comprehensive and capable lineup of systems for the high performance computing market.

Current Products

      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 prior generation Cray T3E system. Designed to provide efficient scalability and high-bandwidth to run complex applications at high sustained speeds, the Cray X1 system is an “extreme performance” supercomputer aimed at the high end of the vector processing and massively parallel systems markets. 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. The Cray X1E system, first shipped in December 2004, nearly triples the peak performance of the Cray X1 system and features the world’s most powerful processor, at 18 gigaflops, and the highest compute density. Our selling focus for the Cray X1E system covers a range of peak performance from 500 gigaflops to over 50 teraflops. Many of our Cray X1 customers are upgrading to Cray X1E systems.

      The Cray XT3 system uses Advanced Micro Devices Inc. (“AMD”) HyperTransport^tm and Opteron^tm processors connected via our low-latency, high-bandwidth interconnect network. It incorporates a massively parallel optimized Linux-based operating system and a standards-based programming environment designed to deliver unmatched sustained application performance in configurations from 200 to 30,000 processors. The Cray XT3 system features a tightly integrated management and operating system to provide high reliability and to run full-system applications to completion. The Cray XT3 system is based on the Red Storm architecture co-developed by Sandia National Laboratories and us. We began shipments of early versions of

the Cray XT3 system in the fourth quarter of 2004 and full production ramp is planned for 2005. Our selling focus for the Cray XT3 systems covers a range of peak performance from one to over 50 teraflops. List selling prices for a one cabinet system start at under $2 million.

      The Cray XD1 system, like the Cray XT3 system, is a purpose-built, balanced high-bandwidth system that employs standard microprocessors but is designed for the mid-range market. It provides superior sustained application performance employing the direct connected processor architecture to link processors directly to each other and memory, eliminating interconnect bottlenecks and providing greater bandwidth and lower latency than typical cluster systems currently available. The Cray XD1 system leverages high volume technologies such as the AMD HyperTransport and Opteron technology and a Linux-based operating system in connection with our automated management infrastructure and provides the opportunity to accelerate application performance through the use of field programmable gate arrays. Our selling focus for the Cray XD1 system ranges from 58 gigaflops to over 2.5 teraflops with processor counts from 12 to more than 512. List prices for one unit (chassis) systems start at under $100,000, with multiple units providing enhanced application scaling performance.

      Pursuant to our distribution agreement with NEC, we currently market on a non-exclusive basis the NEC SX series of vector supercomputers 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 classic vector systems offer high reliability in a balanced, commercial quality system. We have sold several SX systems to Canadian customers.

Current Projects

      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 processors from AMD connected via our low-latency, high-bandwidth, three-dimensional interconnect network based on HyperTransport technology. The Red Storm project involves critical network and Linux-based operating system development. We completed delivery and installation of the Red Storm hardware at Sandia in the first quarter of 2005, subject to installation of certain component upgrades when they become available. We are developing and installing system software designed to run applications programs across the entire system.

      In mid-2002 DARPA selected Cray and four other companies for phase 1 of an advanced research program leading to the development of a commercially available high productivity system capable of running real-world applications 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 research 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, with just under $50 million over three years to investigate advanced design concepts for the petaflops system. IBM and Sun Microsystems 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 are involved in several substantial research projects to develop vector-based, multithreaded and scalar-based offerings that will continue to advance performance and scalability. These activities include a successor to the Cray X1/ X1E line, code-named the Black Widow project; continued development of our

multithreaded system; and development of an integrated technology platform providing a single user interface and environment and improved performance by matching the appropriate processors to the needs of the user’s applications, code-named the Rainier project. These projects are expensive undertakings in terms of dollars, people and time. We seek government funding, such as funding provided for the Red Storm and Cascade projects and the Cray X1/ X1E systems, to help defray the costs of this advanced research.

Services

      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 professional services organization supports our emphasis on providing solutions rather than just computer systems to our customers. This organization provides consulting, integration of Cray products, 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.

Technology

      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 — vector processing, massively parallel processing and multithreading.

      Cray Research pioneered the use of vector systems, from the Cray-1 to the Cray T90 systems. These systems traditionally have used 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 that have been written to maximize the number of long vectors.

      Massively parallel processing architectures typically link tens, hundreds or thousands of standard or commodity processors to work either on multiple tasks at the same time or together in concert on a single computationally-intensive task. We build only massively parallel systems that have high-bandwidth and low-latency interconnect systems. As our systems employ very densely packaged connections and transfer data at very high speeds, they are best suited for computing problems that require many processors to communicate with each other, large memory systems and I/ O connections frequently. Cray XT3 and XD1 systems are purpose-built, balanced high-bandwidth systems that employ standard microprocessors.

      The Cray X1/ X1E system is the first supercomputer that combines the attributes of both vector and high-bandwidth massively parallel systems. With up to 64 processors per cabinet and a shared memory, the Cray X1/ X1E system can run small problems as a vector processor would or, by focusing many processors on a task, the Cray X1/ X1E system operates as a massively parallel system with a system-wide shared memory and a single-system image.

      Our multithreaded products are 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 system latency tolerant and, with flat shared memory, are 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:

      Our hardware engineers are located primarily in our Chippewa Falls, Wisconsin, Seattle, Washington, and Burnaby, B.C. offices.

      We design and maintain our system software internally. The Cray XT3 and XD1 systems exploit commercially available versions of the Linux operating system, as does the Red Storm system. In conjunction with the development of our integrated approach, we anticipate that we will merge our operating systems to one or more variants of the Linux operating systems. We currently provide and support separate UNIX-based operating systems for the Cray X1/ X1E system, our multithreaded system and the NEC SX products.

      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 our own resources where we add the highest value.

      Our software personnel are located principally in our Mendota Heights, Minnesota, Seattle, Washington, and Burnaby, B.C. offices.

Sales and Marketing

      We primarily sell our Cray X1E and XT3 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. We sell our Cray XD1 systems through our direct sales force and through channels we are developing in all of our geographical markets. About half of our sales force is located in the United States

and Canada, with the rest overseas. 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. Our marketing personnel are located in the United States and Canada.

      In 2004 one customer, Sandia National Laboratories, through our Red Storm project, accounted for 27% of our total revenue. In 2003 one customer, Oak Ridge National Laboratory, accounted for 11% of our total revenue and in 2002, no single end-user customer accounted for 10% or more of our revenue. Agencies of the United States government, both directly and indirectly through system integrators and other resellers, accounted for approximately 72% of our 2004 revenue, 74% of our 2003 revenue and 79% of our 2002 revenue. Information with respect to our international operations and export sales is set forth in Note 16 of the Notes to the Consolidated Financial Statements.

Manufacturing and Procurement

      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 personnel are located in Chippewa Falls, Wisconsin.

      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 X1E systems from IBM and for the Cray XT3 and XD1 systems from AMD, and field programmable gate array circuits for our Cray XD1 system from Xilinx, Inc. Texas Instruments will be acting as our foundry for future vector processors. IBM currently provides packaging for our Cray X1E and XT3 systems. We obtain custom cables and interconnect components for our Cray X1E from InterCon Systems, Inc. We obtain custom memory products for our Cray X1E systems from Samsung Semiconductor, Inc. Hitachi America Inc. is our sole supplier for Cray X1E printed circuit boards. We acquire power modules and spray cap cooling systems for the Cray X1E from SAE Power Incorporated and Parker Hannifin Corporation, respectively. We obtain power supplies for the Cray X1E system from Pioneer Magnetics, Inc., and for the Cray XT3 system from Valere Power, Inc. We use Benchmark Electronics to assemble our Cray X1E and XT3 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. See “Factors That Could Affect Future Results — Our reliance on third-party supplies poses significant risks to our business and prospects” below. Procurement personnel primarily are located in Chippewa Falls, Wisconsin.

Competition

      The high performance computing market is intensely competitive. There are significant barriers to entry into the capability and enterprise segments of the high performance computing market and the cost of remaining competitive is high. 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. These competitors have substantially greater research, engineering, manufacturing, marketing and financial resources than we do.

      We also compete with systems builders and resellers of systems that are constructed from commodity components using microprocessors manufactured by Intel, AMD, IBM and others. These competitors include

the previously named companies as well as smaller firms active primarily in the divisional and department markets that benefit from the low research and development costs needed to assemble systems from commercially available technology. These companies have capitalized on developments in parallel processing and increased computer performance in commodity-based networking and cluster systems. While these companies’ products are limited in applicability and scalability and can be difficult to program, they have achieved growing market acceptance. They offer significant 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 massively parallel systems, NEC offers 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 is difficult due to NEC’s aggressive pricing strategies and strong classic vector products.

      We compete primarily on the basis of product performance, breadth of features, availability of application software, price, quality, reliability, service and support, corporate reputation, brand image and account relationships. Our market approach is more focused than our competitors, as we concentrate solely on high-performance computing. Our products are designed for the needs of this specific market. We offer systems that provide greater performance on the largest, most difficult computational problems and superior price/performance on many important applications. Our systems offer total cost of ownership advantages as they typically use far less electric power for operations and cooling and occupy less space than low-bandwidth and cluster systems.

Intellectual Property

      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/X1E system. 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 or group of patents 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.

Employees

      As of December 31, 2004, we employed 889 employees. We have no collective bargaining agreement with our employees. We have never experienced a work stoppage and believe that our employee relations are excellent.

Factors That Could Affect Future Results

      The following factors should be considered in evaluating our business, operations and prospects, they may affect our future results and financial condition and they may affect an investment in our securities. Factors specific to our 3.0% Convertible Senior Subordinated Notes due 2024 (the “Notes”) and our common stock are set forth under the subheading “Factors Pertaining to Our Notes and Underlying Common Stock” below.

      Our quarterly operating results may fluctuate significantly. Our operating results are subject to significant fluctuations due to many factors, which make forecasting revenue and earnings for any period very difficult. First, 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 and within a quarter. This is due to the high average sales prices and limited number of sales of our larger systems per quarter, the timing of purchase orders and product delivery, and our general policy of not recognizing product revenue until customers accept our products and other contractual provisions have been fulfilled and the uncertain timing of payments for product sales, maintenance services, government research and development funding and inventory. A delay in an acceptance of a system at the end of a quarter or year or other factors affecting revenue recognition could move the associated revenue into a subsequent quarter or year and have a significant impact on revenue, earnings and cash receipts. For example, in 2003 we were successful in obtaining timely acceptances of major Cray X1 systems at the end of each quarter. In the fourth quarter of 2004, however, we were not able to record revenue for any of our late quarter shipments, which adversely affected fourth quarter and 2004 results. Delays in developing systems and enhancements could also result in cancellation or loss of orders. These factors will continue to apply to sales of our Cray X1E and Cray XT3 systems in 2005. We anticipate continued deferrals in recognition of revenue and associated costs for sales of products due to contractual provisions despite earlier installation and, in most cases, significant prepayment. At the end of 2004, we had approximately $38 million of deferred product revenue.

      Second, excluding revenue from our development projects, almost all of our product revenue in 2004 was due to sales of one product, our Cray X1 system, and was significantly less than anticipated. In 2005, our quarterly revenue and product margins will depend on the success in the marketplace of each of our newly introduced products — the Cray X1E, Cray XT3 and Cray XD1 systems — the timing of revenue recognition for several large transactions, and early product cycle sales at lower margins due to higher early manufacturing costs.

      Third, a number of our prospective customers receive funding from the U.S. or foreign governments. The timing of orders from these government customers is subject to the funding schedules for the relevant government agencies and delays that may be experienced in competitive procurements. Delays in the government appropriations process, including competitive procurements, could defer purchases and revenue recognition for transactions with government agencies.

      The timing of orders and shipments and quarterly results also could be affected by additional events outside our control, such as:

      If we were unable to complete system software development successfully for the Red Storm project and the Cray XT3 system, our 2005 results would be materially and adversely impacted. The acceptance of the Red Storm system at Sandia National Laboratories and the acceptance of several large system installations of our Cray XT3 system are dependent on our ability to complete the development of and to install stable system software that enables the scaling of application programs over a large number of processors. We are engaged in

a significant effort to complete this system software development project. A substantial delay in completing this work, or a failure to do so, could result in a delay in receiving acceptance or a default under our Red Storm project, delay or prevent revenue recognition on several large Cray XT3 installations, and adversely affect the possibility of additional orders for the Cray XT3 systems and our other products, particularly from the U.S. government.

      We were not successful in completing the Red Storm project on time and on budget, which adversely affected our 2004 earnings and could adversely affect our 2005 earnings and financial condition. Our 2005 revenue goals are dependent on the successful completion of the Red Storm project with Sandia National Laboratories. 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. We have experienced delays in receiving timely deliveries of acceptable components from third parties and development delays, which caused us to miss the contractual third quarter 2004 delivery date. Hardware shipments of the Red Storm system to Sandia commenced in the third quarter of 2004, and were completed in the first quarter of 2005, subject to the installation of certain component upgrades when they become available. We are developing and installing system software designed to run application programs successfully across the entire 10,000-processor system. Falling behind schedule and incurring cost overruns on the Red Storm project has adversely affected our cash flow and earnings, and we recognized the estimated loss in 2004. The Red Storm delays also prevented us from delivering Cray XT3 systems, the productized version of the Red Storm system, in time to recognize revenue in 2004. It is possible that we may have additional losses on the Red Storm contract in 2005. Failure to pass acceptance tests for the Red Storm system or to receive full payment for the Red Storm system would result in additional charges to earnings, and if severe enough could result in a contract default or termination. In the event of a contract default, we could be required to deliver all our knowledge and data that materially concerns the Red Storm systems, subject to the trade secret and intellectual property rights of third parties, and assign to Sandia all of our rights to our contractor developed technology, as such term is defined in the contract, subject to a paid-up non-exclusive and non-transferable license to practice such technology. In the event of a contract termination, we may be liable to pay Sandia for excess costs required to complete the contract. Such delays, default declaration and/or termination could adversely affect other transactions with other U.S. government agencies and our 2005 results and financial condition.

      Our product revenue and margins in 2005 depend on the success of three new products. Whether we achieve planned 2005 product revenue and margins will depend on whether we have sufficient internal engineering, marketing and sales resources to complete development and to market and sell successfully each of our newly introduced products — the Cray X1E, Cray XT3 and Cray XD1 systems — at sufficient margins in a highly competitive market. We must target each of these products at the appropriate markets so that there is minimal market confusion about our products. If we are not successful in these efforts, we may not achieve our planned product revenue and margins.

      We will use a significant amount of working capital in the first half of 2005, which could restrict our operations and could make it advisable for us to raise additional equity or debt which could be dilutive to our shareholders. At any particular time, our cash position is affected by the timing of payments for product sales, receipt of prepaid and regular maintenance payments, receipt of government funding for research and development activities and payment for inventory, resulting in significant quarter to quarter and within a quarter fluctuations in our cash balances. Our principal sources of liquidity are our cash and cash equivalents, short-term investments and our operations. We experienced lower than anticipated product sales and delays in the availability of new products in 2004, which adversely affects our current cash flow. We face increased inventory purchases and higher start-up manufacturing and selling costs with the introduction of three new products in late 2004 and early 2005. Our 2004 restructuring will lower our overall operating cash expenditures but not until severance and related obligations are satisfied. Until we are able to ship our new products, obtain product acceptances and receive payment, we expect to use significant working capital, particularly in the first half of 2005. Meanwhile, we are focused on expense controls and working capital efficiencies to maintain adequate levels of cash within each quarter.

      Depending on operating results, it could be advisable to enhance and strengthen our cash and working capital position by raising additional equity or debt capital. A financing may not be available to us when needed or, if available, may not be available on satisfactory terms, may contain restrictions on our operations, would reduce the percentage ownership of our shareholders, may cause additional dilution to our shareholders and the securities may have rights, preferences and privileges senior to the Notes and/or our common stock.

      If the U.S. government purchases fewer supercomputers, our revenue would be reduced and our earnings 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, including our products. From January 1, 2001, through December 31, 2002, approximately $101 million of our product revenue was derived from sales to various agencies of the U.S. government; in 2003 and 2004, approximately $145 million and $81 million of our product revenue was derived from such sales, respectively. Our sales of Cray X1 systems and contracts for Cray X1E systems have been largely to government agencies in the United States and other countries, and we expect that will continue throughout 2005. To date, however, we have entered into a limited number of significant new contracts for sales of Cray X1E systems to U.S. government customers, especially in the defense segment, and we do not expect sales of the Cray X1E systems in 2005 to match the level of Cray X1 system sales in 2003 to such customers. 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 agencies and departments of 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 losses in future periods.

      Failure to manufacture and sell Cray XD1 systems in planned quantities would adversely affect revenue and earnings in 2005. To be successful, the Cray XD1 system must be manufactured and sold in quantities much higher than our high-end products. We are redesigning our supply and manufacturing processes to accommodate significant daily production and shipment of Cray XD1 systems. The redesign of our supply processes includes finding and qualifying new suppliers. We experienced delays in receiving acceptable components from third parties, which delayed shipments of Cray XD1 systems in the fall of 2004. We are revamping our sales procedures to accommodate high volume sales through the retraining of our current sales personnel and adding sales channels — both distributors and agents — in various markets. We need to market these systems at sufficient margins in a highly competitive market and lower the cost of goods for the Cray XD1 system to achieve an acceptable rate of return. We are changing our service processes to accommodate the expected increased number of Cray XD1 systems in the field. Lack of success in so adapting our processes and sales channels and our manufacturing and marketing processes will adversely affect revenue and earnings from the Cray XD1 system in 2005.

      The decline in the vector processor market may make sales of the Cray X1 and Cray X1E systems more difficult, which would adversely affect our revenue and earnings. The market for vector-based systems has declined over the past several years, and is now served only by NEC and us. We expect that sales of Cray X1E systems primarily will be to domestic and foreign government agencies, including upgrades to existing Cray X1 customers. The Cray X1E system offers processor speed improvements and enhanced price-performance characteristics. We anticipate difficult competition with NEC for vector-based procurements in overseas markets and perhaps in the United States. If we are unable to market and sell the Cray X1E system successfully, our revenue and earnings will be adversely affected.

      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 2005 and beyond. Our success in 2005 and in the following years depends on completing the Red Storm project; completing initial development (particularly of system software) and successfully selling the Cray XT3 system, which involves adapting the Red Storm concept for the broader governmental, industrial and academic markets; successfully selling the Cray X1E system as a significant enhancement to the Cray X1 system; and completing enhancements to the Cray XD1 system, completing stable system software to scale application programs across multiple units and successfully selling the Cray XD1 system in the midrange market. In subsequent years we must develop further hardware and software enhancements to the Cray XT3 and the Cray XD1 systems, and develop our integrated technologies plan, which will allow customers to take advantage of

innovative scalar, vector and future processor technologies within a common high-bandwidth infrastructure. These hardware and software 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, and our 2004 restructuring has strained our engineering resources further. Given the breadth of our engineering challenges, we periodically review the anticipated contributions and expense of our product programs to determine their long-term viability. We may not be successful in meeting our development schedules for technical reasons and/or because of insufficient engineering resources. Delays in completing the design of the hardware components, developing requisite system software or in integrating the full systems would make it difficult for us to develop and market these systems successfully and could cause a lack of confidence in our capabilities among our key customers. At the beginning of 2004, we had planned on generating sizeable revenue shipments of the Cray X1E and XT3 systems in the second half of 2004. Due to development delays, however, we did not record any Cray X1E or Cray XT3 system revenue in 2004. We may incur similar delays in the future, which would adversely affect our revenue and earnings.

      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, printed circuit boards, connectors, cables, power supplies, software components and certain memory parts, on a sole or limited source basis to third-party suppliers. We use contract manufacturers to assemble our components for all of our systems. We are subject to substantial risks because of our reliance on these and other limited or sole source suppliers. For example:

      Our products must meet demanding specifications, such as integrated circuits that perform reliably at high frequencies in order to meet acceptance criteria. From time to time we have incurred delays in the development and production of key components for the Cray X1E, Red Storm, Cray XT3 and the Cray XD1 systems. The consequent delays in product shipments and acceptances adversely affected 2004 revenue and may affect adversely 2005 revenue and margins.

      We have used IBM as a key foundry supplier of our integrated circuits for many years. In 2004 IBM informed us that it would no longer act as our foundry supplier on a long-term basis, although it will continue production of our current products for a limited time. We have negotiated a termination of the relationship with IBM and we are completing contracts with Texas Instruments Incorporated to act as our foundry for certain key integrated circuits for new products planned for 2006 and later. Moving to a new foundry involves a costly redesign of components and processes that will adversely affect operating results in 2005, and may cause delays in the development of these future products.

      Our Cray XT3 and XD1 systems utilize AMD Opteron processors. If AMD suffers delays in the development of enhancements to its processors, such as in the delivery of its planned dual-core processors, our

Cray XT3 and XD1 system sales would be adversely affected. Changing our product designs to utilize another supplier’s microprocessors would be a costly and time-consuming process.