Stanford Case Study B

CASE: SM-126 BDATE: 10/04/05

MATRIX SEMICONDUCTOR INC. (B):TRANSITIONING FROM INNOVATION TO EXECUTION

When it comes to technology innovation, you don’t know what you’ve got until you’ve built a million of them.

—Dennis Segers, Chief Executive Officer of Matrix Semiconductor Inc.

THE PROMISE OF 3DM

In late 2004, the future looked bright for Matrix Semiconductor Inc. The company had pioneered the design and development of three-dimensional (3D) integrated circuits that enabled a new class of low-cost, high-density, non-volatile memory products intended to meet the unique requirements of the consumer electronic markets (see Exhibit 1 for a high-level representation of the product’s process architecture). Matrix’s second-generation product moved into high-volume production in July. Since then, Matrix had shipped more than 4 million units at a rate that climbed to 1.5 million units a month. With a growing number of well-known companies in its customer portfolio, an impressive list of strategic partners and investors, and a third-generation product waiting in the wings, the company appeared to be positioned for success.

However, fueled in part by the contagious enthusiasm of the high tech/Internet bubble, the company’s opportunities had looked nearly as promising five years earlier (see Exhibit 2 for a timeline of major company milestones). In late 1999, the Matrix team had just completed a prototypea three-layer memory array consisting of 27 bits (3x3x3)that demonstrated the technical feasibility of the world’s first 3D memories (3DMs). The company also had defined a business strategy to commercialize this new technology. The plan addressed many of the key strategic challenges facing the young company, including decisions about Matrix’s product direction, markets, partnerships, and business model.1 With its strategy in hand, the Matrix team

1 See Stanford GSB case SM-126 A for more information about Matrix Semiconductor, its technology, and the company’s early challenges.

Professor Robert A. Burgelman and Lyn Denend prepared this case as the basis for class discussion rather than to illustrate either effective or ineffective handling of an administrative situation.

Copyright © 2005 by the Board of Trustees of the Leland Stanford Junior University. All rights reserved. To order copies or request permission to reproduce materials, e-mail the Case Writing Office at: [email protected] or write: Case Writing Office, Stanford Graduate School of Business, 518 Memorial Way, Stanford University, Stanford, CA 94305-5015. No part of this publication may be reproduced, stored in a retrieval system, used in a spreadsheet, or transmitted in any form or by any means –– electronic, mechanical, photocopying, recording, or otherwise –– without the permission of the Stanford Graduate School of Business.

mailto:[email protected]

Matrix Semiconductor SM-126 B

began to generate market interest and investment in its technology. Leading venture capitalists and major consumer electronics companies appeared ready for Matrix 3DM and the price-performance value proposition it would potentially offer.

While Matrix’s investors, potential markets, and prospective customers began to fall in line, unfortunately, the company’s leading-edge technology did not follow suit. Originally, the team targeted late 2001 to begin shipping its first generation product. However, a series of unforeseen design and process technology challenges caused the launch date to slip. Just as 2001 came and went without a product that was ready for high-volume production, so did 2002, 2003, and the better part of 2004. As Dennis Segers (President and CEO), Siva Sivaram (COO), Dan Steere (VP of Sales and Marketing) and other members of the management team considered the company’s progress, pitfalls, and strategic direction between 1999 and 2004, they acknowledged that the road between innovation and execution had been complicated, difficult, and more time-consuming than any of them had anticipated. But, as Sivaram joked, “What doesn’t kill you makes you stronger.”2 Reflecting on the productization of the company’s technology, as well as Matrix’s future, the three men considered the importance of the company’s early strategic decisions, the challenges (and changes) it uncovered along the way, and its eventual breakthrough to high-volume production.

DEFINING A MULTIDIMENSIONAL STRATEGYLATE 1999 THROUGH 2000

As Matrix was completing the prototype that would demonstrate 3DM’s technical feasibility, the team recognized the need for a clear, focused, and compelling business strategy that would serve as the cornerstone of its message to potential customers, partners, and investors when it began to approach them in late 1999 and early 2000. Accordingly, the management team and its board of directors made a series of decisions that would set the company’s direction and define its first product offering.

Product: The Right Balance of Risk and Reward

First, Matrix decided to focus its time and resources on developing and productizing one-time programmable memory chips instead of read-write memories.3 Segers explained:

The driving consideration in the decision was to balance risk and reward. The prevailing thought at the time was if the company was going to succeed in building 3D integrated circuits at all, we were going to have to be exceptionally disciplined in the way we went about it. The activity of innovation is intrinsically defocusingthere's this tendency to want to go off and explore and put effort and resources behind every new possibility that comes to mind. But the founding team was a pretty seasoned group of veterans who understood, from day one, the danger in that process. As a result, we asked ourselves, what's the simplest thing we can do that has a significant enough reward to make it worthwhile? The

2 All quotations from Matrix representatives were gathered via interviews with the authors in spring 2005 unless otherwise cited.3 With one-time programmable (OTP) memories, data could be saved to a chip once but accessed many times. With read/write memories, users could save and access data without limitations.

p. 2


answer was OTP. We thought, let's take this one step at a time, demonstrate that we can do OTP first, and then move beyond that at a later date.

Time-to-market was a key consideration in this decision. The Matrix team felt that it could bring an OTP product to market by late 2001, whereas read/write memories were expected to take at least 12 to 18 months longer to develop (due to the technical complexity). Similarly, Matrix decided that the company would more quickly and successfully gain traction in the market going through leading OEMs4 and their well-established brands (instead of trying to create and credentialize a new Matrix brand with end-users). For this reason, the team shopped its OTP concept to major consumer electronics companies to collect their feedback. Given some initial concern that consumers might find OTP too limiting (when compared with rewritable memory alternatives), Matrix also began conducting primary and secondary market research in cooperation with some of its target OEM partners. “At first we thought customers might view OTP as largely restrictive. But that was an inside-out viewpoint,” said Segers. “When we looked at actual market behavior, particularly mass market consumer behavior, it wasn’t necessarily the case.” Even when consumers were given a read/write product, they frequently used is as a write-once medium (e.g., cassette tapes, video tapes). According to Dan Steere:

Some executives at consumer products companies told us they had spent the last couple of years thinking a lot about MP3 players, digital cameras, and digital video and where all this stuff could go. They talked to people in the semiconductor business and asked them for a less expensive OTP product. Nobody thought it was practical or possible to build. They were told to plan on using rewriteable Flash. But rewriteable Flash was still very expensive for mainstream consumer applications in 2000. And so we got a lot of encouragement from large potential customers that our OTP product would fill in a vacant niche in the overall usage model.

While the company perceived that rewriteable memories would ultimately provide bigger market opportunities, the team decided that OTP was a unique, less competitive product category in which Matrix could more quickly gain a presence in the market and achieve differentiation. Considering this, and given the responses it received from end-users, retailers, and target OEMs about the OTP concept, Matrix solidified its focus on developing an OTP product as its first generation offering. Generally, this decision served as an important focal point around which everyone in the company became engaged. However, Segers recalled, “There was a decision made that, in parallel with the commercialization of the OTP, we would maintain a small background research effort in read/write. I think there were probably a handful of the read/write resources who felt like OTP was still too much of a focal point that perhaps, from their point of view, starved the read/write effort from vital resources. I think that's true to some extent. But that's life in a small, resource-limited company. You have to make those choices.”

Markets: A Home Run . . . With Flexibility

With its high-level product direction defined, Matrix turned its attention to product issues at a more tactical level. The major consumer electronics companies that Matrix interacted with were

4 OEM stands for original equipment manufacturer.

p. 3


particularly interested in one-time programmable blank memory cards for use in digital audio applications and, in particular, digital cameras. As Segers put it, “Back in 2000, our strategy was to step up to the plate and swing for a home run on the first ball that got pitched. And a homerun was viewed as the digital photography market. We set out to produce a blank memory card that we would sell through strategic partners, and we would revolutionize digital photography with this one-time programmable, archival storage media.” At the time, Flash was the de facto storage standard in the digital photography market and Matrix intended to offer a product that could serve as a low-cost substitute for this technology. Accordingly, the team made a decision to ensure that its product was compatible with existing NAND Flash5 standards so that manufacturers could easily incorporate 3DM into their designs and consumers could easily substitute 3DM for Flash at the application level. “By making our product NAND Flash compatible,” said Steere, “we thought we’d be compatible with most of the momentum in the industry.”

The other significant market opportunity potentially facing Matrix was in storage for prerecorded content (e.g., gaming cartridges, embedded electronic content for other handheld devices such as maps and dictionaries). In this space, Matrix would compete with Mask ROM,6 one of the lowest-cost forms of available pre-recorded semiconductor memory. While the market was not as exciting as digital photography and audio, and Matrix’s cost advantage would not be as great (due to the lower average selling price of Mask ROM compared to Flash), the applications using Mask ROM were more stable, leading companies were more easily identifiable, and there were fewer unknowns with which to contend. Matrix 3DM also offered a compelling competitive advantage to Mask ROM in that it could be programmed in the field,7 and with shorter lead-time than Mask ROM. While the Matrix team was drawn to opportunities in blank media, prerecorded content appeared to be a viable alternative upon which the business could be built. Fortunately for Matrix, said Steere, “We recognized pretty early in that we didn’t have to choose. We believed that the same chip would be applicable to a reasonably large part of the overall Mask ROM market.” While Mask ROM memories had historically used random-access parallel interfaces, Matrix learned that major players were evolving to the serial NAND Flash interface standards, which further strengthened the flexibility of its product (see Exhibit 3 for sample products in Matrix’s target markets and Exhibit 4 for a high-level overview of the memory market landscape).

5 NAND Flash, developed by Samsung and Toshiba, was introduced in 1989. It has faster erase and write times, higher density, and lower cost per bit than NOR Flash, and 10 times the endurance. However its I/O interface allows only sequential access to data. This makes it suitable for mass-storage devices such as PC cards and various memory cards, and somewhat less useful for computer memory. The first NAND-based removable media format was SmartMedia, and numerous others have followed: MMC, Secure Digital, Memory Stick and xD-Picture Cards (Wikipedia).6 Mask ROM refers to a kind of ROM (read-only memory) whose contents are programmed by the IC manufacturer. In other words, it is not a user-programmable ROM. The terminology mask is used in IC fabrication. Since the process is costly, mask ROM is used when the needed volume is high and it is absolutely certain that the contents will not change (Wikipedia).7 Field programmability means that content does not have to be programmed by the IC manufacturer during the factory manufacturing process. Instead, Matrix 3DM can be programmed by a third party, on a more just-in-time basis, closer to the point of sale.

p. 4

http://en.wikipedia.org/wiki/Samsung

http://en.wikipedia.org/wiki/Read-only_memory

http://en.wikipedia.org/wiki/XD-Picture_Card

http://en.wikipedia.org/wiki/Memory_Stick

http://en.wikipedia.org/wiki/Secure_Digital

http://en.wikipedia.org/wiki/Multi_Media_Card

http://en.wikipedia.org/wiki/SmartMedia

http://en.wikipedia.org/wiki/Memory_card

http://en.wikipedia.org/wiki/Memory_card

http://en.wikipedia.org/wiki/PCMCIA

http://en.wikipedia.org/wiki/Sequential_access

http://en.wikipedia.org/wiki/1989

http://en.wikipedia.org/wiki/Toshiba


Strategic Partners: Matrix’s Dream Team

In conjunction with these decisions, the company set out to formalize its plans to go to market through an industry-leading set of strategic partners. By late 2000, Matrix had identified, negotiated, and confirmed some important strategic relationships. Steere described the approach:

We took a strategy to go to market based on strategic partnerships with leaders in the market categories that we thought were the best fit. We didn't want to have one single partner, because we felt we'd be overly dependent on one company. And we saw a lot of value in having a small number, perhaps three or four key strategic partners in imaging and audio. We surveyed the world and made a big list of the potential partners. A key question was how to fit together a group of three leaders who competed with each other. If we were going to have three partners, you'd want them to have complementary market perspectives, distribution channels, and market presence. And so it turned out there were a number of different groupings that we put together. The grouping that was on the top of the list was Sony, Kodak, and Thomson Multimedia. We started conversations with those three and explored others, as well. But it turned out those three were actually the ones that caught traction the earliest. All three agreed to invest in Matrix at the end of 2000, along with the venture capitalists who were prior investors.

In total, at the end of three rounds of funding, the case writer estimated that Matrix had raised approximately $80 million. However, this figure was not confirmed by Matrix management (see Exhibit 5 for a partial list of Matrix’s strategic partners and investors).

Business Model: Fabulous Fab Decisions

The other major strategic decision facing the Matrix team in late 1999 was what business model the company would pursue: licensing its intellectual property (IP) to other manufacturers, acquiring a fab and building its own chips, or entering into a fabless manufacturing agreement. After multiple discussions, the Matrix management team and board of directors agreed that, while licensing looked compelling from an up-front investment perspective, it was unrealistic until the company could prove that its product could be produced at high-volumes. The team also felt too much ownership over the technology to entrust its future to another organization by letting them bring it to market.

In terms of purchasing or building its own fab, as Segers said, “It was simply impractical for any company to get started, these days, by going off and trying to build a fab. Had we layered the capital investment risk associated with trying to build a factory on top of the technology development risk, we would have been doomed to failure.” The company determined that it wanted its core competencies to be in strategic technology development and strategic marketing. Everything else, including manufacturing, would be outsourced to top quality, established partners. This approach would enable Matrix to stay focused on what it did best. It also significantly reduced the time and startup capital required by the young company.

p. 5


However, there were two primary challenges that stemmed from this approach. First, the company would face a cost disadvantage to competitors that did not have to pay a 30 to 40 percent margin to a third-party manufacturer. Additionally, because Matrix would have to build its 3DM product using unique process technology (even though the company insisted on leveraging standard fab equipment and materials), it would not fully realize the efficiency gains that typically accompanied a fabless model. Segers explained:

There are certain aspects of the fabless model that our OTP technology doesn't match well. The ideal company profile for a fabless model is one where the company can use exactly the foundry's process and its most advanced process possible. The fab is putting together, on its nickel, a capability that's extraordinarily expensive and powerful. To the extent that a company can step in, without any investment of its own time or money, and leverage that capability, that's when it achieves the maximum benefit.

Matrix, on the other hand, would have to make a significant investment in designing a unique manufacturing process and completing a knowledge transfer to the foundry that would ultimately produce the 3DM. To accomplish this, Matrix needed to find the right manufacturing partner that would be able (and willing) to successfully execute a non-standard process for a young, unproven start-up. At this time, the fabless manufacturing model was well established, and there were numerous foundries that Matrix could pursue. In fact, members of the Matrix team had existing relationships with the two leading foundries in the industry. Segers had worked extensively with United Microelectronics Corporation (UMC) through his executive position at Xilinx. UMC was the second largest third-party integrated circuit manufacturer with 1999 revenues of nearly $1.1 billion. Sivaram, who joined Matrix from Intel, had connections to Taiwan Semiconductor Manufacturing Company (TSMC). TSMC held the industry’s top spot with more than $2.3 billion in 1999 revenues. “Siva had a strong relationship with the TSMC guys, and I had a strong relationship with the UMC guys,” recalled Segers, “So we carefully evaluated both options. Ultimately, at the end of the day, we concluded that TSMC looked like a potentially better partner for what we were trying to do.”

However, before Matrix would be ready to go to work with its new manufacturing partner, the company had to determine what generation process technology to use in manufacturing its product. As Steere explained, each generation of process technology conceptually represented “the font size used for printing circuits on a wafer.” The more transistors that could be fitted on a single chip, the greater its capability (and the lower its relative cost). At the time, 0.25 micron technology was in common use and was the highest volume process technology. The next generation, 0.18 micron technology, was in production in a few industry-leading integrated device manufacturers, and was expected to go into production in the foundries sometime in late 2000. “We had to choose where to go,” said Sivaram. “We were worried that if we went with 0.18 micron technology, we wouldn’t get enough attention, capacity, or people because they would all have other problems to solve. We also knew we needed a clean process because we would already have enough risks coming from our portion of the flow.” Steere added, “We favored 0.25 micron because we thought it would be a more mature production process and would give us a quicker time-to-market.”

p. 6


However, using the more advanced 0.18 process technology would increase Matrix’s cost advantage, which was a critical competitive differentiator for 3DM.8 Steere recalled, “The cost [of using 0.25 micron] looked fine, but product costs would be lower if we chose the more advanced process.” Ultimately, the team made a decision based on an assessment of the key drivers in the memory markets it was trying to serve. “As we began looking at how to best optimize the technology, it turned out that, for the markets we were attempting to serve, the technology did not need to benor should it beon the most advanced processes,” said Segers. “As a result,” added Steere, “we launched on .25 micron, initially aiming for an introduction timeline of late 2001.” The team felt this decision would enable it to take advantage of a clean, stable process and available capacity by the time the company was in high-volume production. The risk of potentially impacting Matrix’s 10x cost advantage was deemed an acceptable tradeoff for the first generation product, which had been given the internal code name Phoenix.

One additional decision related to Matrix’s fab issues in the late 1999 to early 2000 timeframe was how to develop the unique manufacturing process that would be used to produce Matrix’s non-standard chips using standard fab equipment and materials. “We needed to figure out the intermediate step between concept and manufacturing so that we could do a knowledge transfer [to TSMC],” said Sivaram. One option would have been to send process engineers to Taiwan to modify the manufacturing process at TSMC, but Matrix decided against this approach. According to Sivaram, “Manufacturing fabs abhor change. They are all about repeating the same thing day after day after day. But development is change. The whole reason for its existence is change, change, change. So, we decided we could not do development in a production fab.” Instead, the team needed access to a local fab environment that offered state-of-the-art equipment (unlike the out-of-date, overcrowded facilities they had been using at Stanford). Again leveraging the relationships of its team members, Matrix talked with a few local fab facilities before entering into an agreement with LSI Logic (located just down the street from the company’s Santa Clara headquarters). The fab at LSI would give Matrix the opportunity to do process development, run pilot production, and gain invaluable experience before brining its product (and process) overseas to TSMC for high-volume manufacturing.

PEELING THE ONION2001 INTO 2003

With a series of important decisions behind them, and a promising future ahead, the Matrix team was energized in 2000. In this timeframe, the team grew to roughly 30 people as Matrix began hiring specialized technologists who could meet the more diverse needs of the company as product development progressed. Sivaram recalled, “If there was one thing that we did extremely well, it was hiring. We made some spectacular hires in 2000 and added company-changing employees to the team.” Steere added, “Everything was very exciting in 2000. We had an innovative concept, we built a strategy around a technology path that we all felt very good about, we had a plan for the market, and we signed up the partners we were most excited about working with. Things were coming together very nicely.” Furthermore, in California’s Silicon Valley, the economy was booming and, at least for a time, small and large technology companies alike began to get a feeling of invincibility. “We felt like we had rediscovered capitalism,” said Sivaram, “With four or five PowerPoint slides people thought they could get funded and make a billion dollar company. But that optimism was misplaced, and it took an experience like

8 In 1999, Matrix believed its technology could offer a 10x cost advantage over competing memory products.

p. 7


Phoenix to get it straightened out.” Segers added, “I would characterize 2000 as the peak of our naivety.”

Between 2000 and 2004, conditions at Matrix changed dramatically. As product development began in earnest in 2001, the company was confronted with a series of interrelated and complex technology challenges that began to cause its schedule to slip. “We grossly underestimated the complexity of the undertaking,” said Sivaram. “Technically, the problem was not with the bulk of the bits. It was the tail that killed us. Every distribution has a tail, and the tail [for Phoenix] was big enough that we did not know how to handle it. So, we underestimated the scope of effort required in design, in process technology, in manufacturing, and in customer acceptance.” Steere concurred: “It was like peeling an onion. Every time we solved one problem, a new, more complicated issue emerged in its place.” As with many leading edge technologies, 3DM had new kinds of defects and new ways of affecting manufacturing yields that conventional memory designers and engineers did not have experience in addressing. “No one had ever done this before. The team was learning from its mistakes in real-time,” said Sivaram. “The fab process couldn't deliver, and the chip design couldn't cover for it. And, as a result, Phoenix didn’t work. This actually drove a schism in the company. Designers were pointing their fingers at the fab guys, and the fab guys were blaming the design team. It was rough.”

To make matters worse, with the Silicon Valley economy in a rapid state of decline in 2001, the company’s relationship with LSI fell apart. “Unfortunately, working with LSI ended up to be a fiasco,” said Sivaram. “The economy collapsed and LSI had to downsize. It basically shattered our whole agreement, and we were left holding the bag.” Matrix was forced to find another development fab in order to continue engineering the process technology that would ultimately enable the knowledge transfer to TSMC and Phoenix’s move into high-volume production. Fortunately, through its experience with LSI, the team had become much more knowledgeable and discriminating about its criteria for seeking a development fab partner. The company also took the opportunity to rethink the way it was approaching process development. At LSI, Matrix had been focused on engineering the end-to-end development process with the intent of then transferring it, in its entirety, to TSMC. In the interest of reducing complexity and increasing predictability, the company’s new approach was more modularMatrix would design only those aspects of the manufacturing process that were unique to 3DM and leverage as much of TSMC’s standard processes as possible for the remainder of the steps.

In the end, Matrix ended up entering into an agreement with Cypress Semiconductor Corporation. “We actually considered Cyprus earlier,” recalled Sivaram, “but we couldn't agree on a suitable IP model. They were more amenable the second time because the economy was different and they were more motivated to get someone into the fab. We got them to guarantee that the fab wouldn't shut down, and we developed a very clean IP ownership model.” However, despite the positive take-aways from this change, the failed relationship with LSI further exacerbated the slippage in Matrix’s schedule. “Overall,” estimated Sivaram, “it was probably a waste of six to nine months.”

With so many technical challenges to contend with, Matrix determined that some changes to the company’s strategy were in order. Segers, who signed on full-time as the company’s CEO in the fall of 2001, explained the shift:

p. 8


Over the course of time, it became apparent that it was going to take longer for us to be able to get the technology to work, and that the degree of functionality was not going take a quantum leap from ‘nothing works’ to ‘everything works at a pristine level.’ As a result, it became necessary to alter our go-to-market strategy. One thing that changed was our basic philosophywe said let's see if we can't step up to bat and get on first base with something before we swing for a homerun. So, we refocused some of our go-to market efforts away from digital photography and other blank media applications in favor of prerecorded content. We did that for a couple of reasons. First, it was technologically easier to build a part that could compete in that space. Second, it was an easier competitive field. In the prerecorded content space, we had a pretty clear, easy-to-define value proposition. And we could step into markets where we weren't trying to create a new usage model for memory. We also weren't taking on major marketing and market acceptance risk. Third, frankly, the scale of the market was smaller and we could learn to walk before we had to run. Finally, by first directing our attention to prerecorded content, we had the opportunity to program the content for many of our customers. In doing this, we would be able to see firsthand how many parts didn't work the way we expected them to. If there were programming errors or certain types of benign defects that were in the parts, we'd see them first ourselves. Then, we could go address them in the fab, alter the design, alter the test technology, or refine our processes before we put them in the hands of our customers.

At the time the company modified its go-to-market strategy, Matrix also “became a bit more conservative and modest in the scope of what we were trying to do,” said Segers. For example, rather than trying to build an eight-layer memory device, the team determined that a four-layer memory cell was a more reasonable goal that would have only a moderate impact on the product’s anticipated performance. According to Segers, the company also came to the realization that “some of the earliest estimates of price-performance for our technology were a bit misapplied.” Memory costs per megabyte had plummeted over the last few years (see Exhibit 6), making Matrix’s 10x cost advantage unattainable (at least for its first generation product). In 1999, the team anticipated that 1 MB of NAND Flash would cost approximately $1 in 2002, however, the actual cost turned out to be just $.20. Similarly, Mask ROM, which was expected to cost $.32 in 2002, was selling for $.28 per MB (see Exhibit 7). Further compounding Matrix’s price-performance issues was the fact that, by 2003, the 0.25 micron process technology that Matrix decided to adopt had already fallen two generations behind, with 0.13 micron technology now at the forefront.

Reflecting on the overall situation between 2001 and 2003, Sivaram said, “It was a morale-sapping, painful period for the company. We had promised the sky and the moon to customers. But we couldn’t make the product. We couldn't get yields up. We couldn’t get the process to be stable. And our partners were starting to doubt us.” As a result, “We were left playing defense for quite a while on the sales and marketing side, talking to customers about our strategy and also communicating product delays,” said Steere. “To some of our partners, like Nintendo, all the changes were fine because they just went up in the queue,” commented Segers. “However,

p. 9


others, who had chosen to partner with us for digital photography applications, were less than enamored.” To address the concerns of the company’s customers, Steere and team dedicated themselves to keeping these important stakeholders informed. “We communicated pretty intensely,” said Steere. “They understood that what we were doing was totally new and they had faith in Matrix's engineering capabilities and in the people they were working with. We kept them regularly updated on status, and in a lot of cases, if we didn't know what was going to happen next, we told them that upfront.” Sivaram added, “Many of these meetings were painful, but I think the customers and partners stayed with us because we were not hiding anything from them.”

The support of the company’s strategic partners also helped keep investors engaged. Steere explained further, “The strategic partnerships had a big impact on investor confidence. There were all sorts of technology uncertainties, but the fact that we had large potential customers who had invested significantly in the business and continued to reaffirm their interest in the products, once we could make them, made our investors feel more comfortable sticking with us. So, that was a big contributor in buying time for us to work through our issues.”

BREAKTHROUGH TO EXECUTIONMID-2003 THROUGH 2004

Despite these challenges, the company continued to press forward. As time progressed, Segers explained how the Matrix strategy continued to evolve:

There was a second part of our strategy that we began to create. To be competitive in a fast-moving market like memory, you can’t enter with a point product. I guess the best analogy might be, if you're in a NASCAR race and you're pulling out of the pits, by the time you've exited pit row you'd better be going 200 miles an hour. In our business, that means that you're not executing one product. You're executing on a roadmap that demonstrates ongoing advancement of your technology. So one of the things we had to do was begin to look beyond our first generation product. There's going to be another generation after this, and another one after that, and another one after that . . . so how do we build capability within the organization to capture the maximum learning from one generation of our technology and roll it into advancements in the next generation in as short a time as possible? We put into place a roadmap that said we want to be producing new generations of our technology with a period cycle of one to one-and-a-half years. In that context, as we looked at Phoenix, we realized it was bogging down the whole pipeline. But we didn’t want to completely abandon that first generation, because that could potentially doom the second and third generations if we didn’t capitalize on the learnings. Finally, we reached a conclusion to go to market with our first generation, on a selective basis, with just one or two applications within the prerecorded content space. We really wanted a demanding customer from a quality perspective. We needed to prove, once and for all, that this 3D technology could be really reliable. So we picked a top-quality Japanese supplier [Sharp], and their eDictionary product line. The demand was modest enough in scope that it would allow us to hone our skills in producing the technology at the same time that we did a few rudimentary things, like building up the capability in the organization to take a purchase order,

p. 10


schedule deliveries, mail out an invoice, collect our receivables, and all that kind of stuff. You have to go through that organizational learning. And, as modest as our production ramp was, we stubbed our toe every step of the way. But it positioned us in a much better way.

Steere added his thoughts on the company’s decision to move Phoenix into production in mid-2003:

Even though we recognized that Phoenix wasn't going to be competitive, it wasn't going to be a high-volume product for us, Dennis pushed the company to keep with it. He pushed us to pick a customer, pick an application, and run pilot production volumes of Phoenix, to make sure that we benefited from manufacturing and operational learning around setting up a production flow and running a certain volume. Then, we could throw all that learning into the execution of the next generation product that we would produce in high-volume.

In total, Matrix spent about nine months in production and shipped fewer than 50,000 units of its Phoenix product. “Those first 50,000 units were really painful,” recalled Segers. “But, they prepared us to ramp from 0 to 2 million units just in 14 weeks when the second generation product was ready. The difference was pretty dramatic.”

Matrix’s second-generation product, code-named Flagstaff, was ready for high-volume production in Q3 of 2004. “That was when we felt like, okay, now we're in business,” said Steere. Flagstaff incorporated significant changes in both the product design and the production process for manufacturing it and, as a result, took much less time to develop and transfer to TSMC (see Exhibit 8 for a product overview). That being said, the launch still had some issues. Segers explained, “As we were preparing to launch our second generation OTP technology into volume production, we stumbled in the transfer of the process from Cypress to TSMC. There were some process integration challenges that emerged that really threatened the viability of our OTP business in a pretty significant way. Given that set of circumstances, we had to make a decision to back-burner our read/write activities for a period of time” so that all resources within the company were focused on sustaining Matrix’s core business. Sivaram elaborated:

The initial transfer did not go well. When we put everything together at TSMC, we discovered some strange process interactions that we had not encountered at Cypress. So, in the spring of 2004, we started having these GOATGet Our Act Togethermeetings. Every morning we would sit there, pouring through the reports from TSMC, but we couldn't see past our noses. But Dennis could see it from a different perspective. Finally, he said, ‘You guys have the solution at hand. You don't see it, but it's there. So, let ‘er rip. Go today and start a thousand wafers.’ A thousand wafers costs about $2 million to produce, and it's not like we were putting any cash in our pockets at the time. I was a little bit skeptical, and I suggested we should try this on a smaller scale. But, Dennis put us all in a room and said, ‘It’s a thousand wafers and that's it. Go do it.’ So it was a big strategic decision. I didn't have the guts to make it at that time. I was afraid because we

p. 11


had just been beaten down for three years. But, it turned out to be a tremendous exercise.

At Segers’ insistence, Flagstaff rapidly moved into manufacturing and the company shipped 100,000 units in July 2004. By the end of the year, Matrix had shipped a total of 4 million units, with its chips designed into consumer products, like Mattel’s Juice Box, in time for the holiday season (see Exhibit 9). Reflecting on the experience, Steere recalled, “We actually didn't announce Flagstaff publicly until after we'd delivered a million parts to customers, so we could make the statementas much to ourselves as to the outside worldthat this isn't a technology curiosity, it's not something that's coming, that's got great potential. 3D memories and 3D integrated circuits are finally real.”

“The moral of the story,” added Sivaram, “is that manufacturing innovative, new technology takes at least three years, from proof of concept to production. Like Dennis says, until you've made a million units of something, you don't know what you have. For us to have thought that we could have broken that rule was being naively optimistic.”

INTO THE NEXT DIMENSION2005 AND BEYOND

With a successful 2004 behind it, the Matrix team felt confident that the company’s prospects remained promising. However, given the challenges of the past five years, the team’s optimism was now more conservative and tempered with more pragmatism than it was in the early years of the company. Emphasizing a continued focus on execution and increasing levels of discipline in its operations, Segers shared his immediate vision for the company’s future:

Our next challenge is to build a profitable base for the company in the existing prerecorded content market. We also need to expand our customer base and begin to move out of the mode of limited, fairly narrowly targeted sets of business development activities into more of a broad-based sales activitywhat I like to call the Johnny Appleseed sales model. We’ve got to get out there and scatter the seeds of our value proposition across as large a segment of the competitive landscape as possible and see what emerges. In some respects, I think we've mastered the technology. Now, we have to prove we can build a profitable business around it.

Matrix’s third generation product, called Trinity, was scheduled for a spring 2005 release (see Exhibit 10 for a high-level product comparison). Capitalizing on some major technology advancements, including a different fundamental architecture, Trinity was expected to achieve significantly higher performance levels at markedly lower price points. “We managed to get two times the number of bits on the same size die, using the same generation process technology and the same fab,” said Sivaram. While these results would not yet deliver on the 10x cost advantage that Matrix forecast in 1999, with each generation of its product, the company continued to improve its pricing differential. According to Steere, “With Trinity, the product is priced somewhere between 20 to 50 percent below Flash (see Exhibit 11), depending on the application. So the realities of getting to market and working through our challenges diluted our competitive advantage. But, for the price-sensitive market segments we’re currently targeting, our differential is still compelling.”

p. 12


While Matrix was committed to building a profitable base of business selling its OTP product line for prerecorded content applications, the company was also focused on expanding into new areas with its 3D technology. Mask ROM revenues, which accounted for the most significant percentage of the prerecorded content market, were forecast to decline by 14.5 percent in 2005 (to $297 million) and by 26.6 percent in 2006 (to $218 million). (See Exhibit 12.) While still a sizable market with strong near-term opportunities for Matrix, the team recognized the need to expand further into the industry’s growth segments.

One of the most immediate opportunities available to the company was to follow through on the launch of a line of blank OTP memory cards for applications such as digital photography (see Exhibits 13 and 14). While this market space was rich with opportunity (due to explosive growth of digital camera usage), it was also a high-risk proposition for the company. “There’s a very tough fundamental technological challenge associated with building one-time programmable memory cards for use by consumers,” said Segers, which revolved around the fact that OTP memory cells could not be tested like traditional memory solutions. With prerecorded content, the company could test the finished product after it was programmed to safeguard against excessive defects. But, for blank OTP memory cards that would be programmed directly by consumers (as they downloaded digital audio files or took digital photographs), no such testing mechanism was available. Segers explained, “As one of our founders says, ‘How do you test a stick of dynamite?’ That’s a problem that must fundamentally be 100 percent solved for the digital photography marketplace. We cannot afford to ship millions of Matrix memory cards only to have them fail in the hands of consumers.” Before launching a line of blank media cards, Matrix would have to be confident that the products would perform anywhere, anytime, under any conditions, with defect rates low enough to warrant consumers entrusting their photographic memories (and other digital assets) to them. “The product is very good,” said Sivaram. “But we’re not there yet.”

Another opportunity was for Matrix to again pursue read/write memory products based on 3DM. While the company’s background R&D effort in read/write had been temporarily placed on hold, the team of engineers had made significant progress in designing a viable product. Matrix still believed that sizable, lucrative opportunities existed for rewritable 3DM, but the questions was how the company should time its reinvestment in productizing this technology. If the company reinstated its read/write in early 2005, it had to consider whether or not this would potentially delay or derail its critical OTP initiatives (i.e., achieving profitability in prerecorded content and completing development of its blank media cards). Strategically, the company remained sensitive to the importance of not overextending itself.

In pursuing blank media and read/write memories, Matrix would compete directly with NAND Flash. Not only was this segment of the memory market significantly larger than Mask ROM, it was expected to grow at the rate of 22.3 percent in 2005 (to $7.5 billion) and 11.2 percent in 2006 (to $8.3 billion). (See Exhibit 12.) In total, when it considered the business it could gain as a substitute to both NAND Flash and Mask ROM, Matrix estimated its total available, near-term market to be at least $300 million. This forecast took into account the company’s existing relationships with its strategic partners, established applications that could potentially leverage

p. 13


its technology (e.g., digital cameras, MP3 players), and emerging portable applications (e.g., multimedia handsets) to which Matrix had gained visibility.

Looking forward, the future of 3D technology seemed secure. Yet, the team wondered what role Matrix would ultimately play in this emerging market. As Sivaram put it, “3DM is manufacturable. That's now been proven. But is it a one-trick pony? Is OTP all we can do? Ten years from now, I believe that 25 percent of all semiconductors will be made in 3D. Is Matrix going to make that happen? Or will we be just a footnote in history?” Steere elaborated further on this concern:

There is a well-recognized trend that says, in a new technology space or a new industry, you have pioneers and then you have the fast followers. And, often the fast follower strategy is the more successful strategy. One of our goals is obviously not to fall prey to that. Many times it seems that the pioneers have a certain view of the world based on the early problems that they had to solve, and the early work they did, and they don't always recognize when the game changes from innovation to execution. That’s why Dennis is so adamant that we understand that, at the end of the day, it’s all about execution.

Staying focused on execution was one way Matrix would defend itself against new entrants and potential competitors in the memory market. The company had also initiated and maintained an aggressive intellectual property protection strategy. So far, Matrix had 107 U.S. patents on its 3DM, a portfolio of international patents, more than 100 additional applications pending with the patent office, and another 100 or so technical disclosures that were in the process of being evaluated or turned into patent applications. Steere explained:

When we started looking at 3D in 1998 and 1999, we were surprised that it was completely unexplored territory. So we've been very aggressively patenting all the different aspects of 3D integrated circuits. That's one protection. But the other and more important thing we’ve done to protect our competitive position is to simply be the world's experts on 3Dto be able to do more with the technology, to move faster with new products, and to simply be better at 3D process and 3D design than anyone else. That's really our main focus. IP can help to a certain point, but your only real protection against competitors coming into your market space is to separate yourself and be better than they are.

New competition for Matrix could come from other young start-ups trying to break into 3D, or the company could face a threat from any of the established semiconductor or memory giants looking to expand their portfolios of traditional (two-dimensional) solutions. The memory segment was particularly ripe for intensifying competition due to concerns that Flash would be unable to keep pace with Moore’s Law beyond the end of the decade due to fundamental limitations of the technology.

As of late 2004, several major semiconductor and memory giants, as well as emerging start-ups seemed to be focused on memory research and development to address this longer-term challenge rather than on having a near-term impact in the market. For example, Intel invested in

p. 14


a small company called Ovonyx that was developing phase-change semiconductor memory devices, called Ovonic Unified Memory (OUM). According to the company, OUM memory technology promised to enable significantly faster write and erase speeds and higher cycling endurance than conventional memory types, like Flash and DRAM, by leveraging materials and processes used in rewritable CD and DVD discs. Ovonyx licensed its OUM memory technology to Intel, among other partners, and the two companies were working together to develop and demonstrate the feasibility of high-density, non-volatile memory based on the technology. Another company exploring new memory technologies was Unity Semiconductor. Unity Semiconductor was a venture-funded company founded in 2002 to commercialize a new nonvolatile memory technology based upon resistance switching in certain conductive metal oxides. Other exploratory efforts by emerging and established players were equally as experimental, involving the use of polymers, organic compounds, or nanotechnology. However, because they all involved the use of new materials and potentially ‘bleeding edge’ manufacturing processes, Matrix considered them not to be competitive threats until at least five years into the future. As the Matrix team knew only too well, it took time, money, and lots of patience to bring something so new to the market in high-volume.

In addition to facing new competing technologies, another scenario was that Matrix could become the target of an acquisition, or could consider partnering with a major integrated device manufacturer as a second source to its products. To date, no major competition (or offers) had been uncovered, despite Matrix’s best efforts to stay abreast of developments in the space. However, as its products spent more time on the market, the company fully expected to begin seeing more competitive activity in the months and years to come. In the interim, Matrix would continue to face significant competition from established, traditional memory products like NAND Flash and Mask ROM.

In the meantime, Matrix would continue to press forward, exploring new, untapped opportunities for its technology. “Much of what we're beginning to think about is how we can take 3D technology and apply it to the broadest swipe of the semiconductor industry possible,” said Segers. “Our core R&D team, our best and brightest, are coming up with new innovations in a lot of these other areas at a much more rapid clip than I ever anticipated. Long-term, our big challenge is how we can ultimately maximize the footprint that Matrix’s 3D technology will have in the semiconductor industry 10 years from now.”

p. 15


Exhibit 1Overall Process Architecture

Source: Matrix Semiconductor Inc.

p. 16

0.15um CMOS with HiV transistor

2 Levels of tungsten interconnect

4 layers of memoryPoly cells, tungsten wires

Al top metal


Exhibit 2Matrix Semiconductor –Timeline of Major Milestones

Date MilestoneNovember 1997 The founders (Tom Lee and Mike Farmwald) discuss 3-D chips for the first

timeMay 1998 Company receives its first round of fundingMarch 1999 Team grows to 7 employeesAugust 1999 Company defines its formal business strategy;

First release targeted for late 2001October 1999 Company completes its first 3D prototypeNovember 1999 Company raises its second round of fundingDecember 1999 Team grows to 15 employeesLate 1999 Company chooses .25 micron process technologyEarly 2000 Company’s emphasis shifts from technology to product developmentFebruary 2000 Matrix begins to pursue its first strategic partnersApril 2000 Company begins working with LSIOctober 2000 LSI closes Santa Clara fab and displaces Matrix effortDecember 2000 Company receives its third round of funding, including investments from its

first three strategic partners (Sony, Kodak, Thomson Multimedia);Team grows to 45 employees; First release date moved from late 2001 to 2002

October 2001 Company begins working with CypressSeptember 2001 Dennis Segers accepts position as company CEODecember 2001 Team grows to 75 employeesSummer 2002 Company modifies expectations for a 2003 product releaseDecember 2002 Team grows to 100 employeesDecember 2002 Company makes an explicit decision to make pre-recorded content its primary

focus (versus consumable blank media)February 2003 Nintendo invests in company as strategic partnerMarch 2003 Company raises fourth round of fundingApril 2003 First generation product ships in limited quantities (Phoenix)December 2003 Company has a 10 percent reduction in forceJanuary 2004 Read/write “skunk works” placed on holdJuly 2004 Second generation product begins to ship (Flagstaff)November 2004 Company announces second generation product (after more than 3 million

units have already been shipped)


p. 17


Exhibit 3Target Markets


p. 18


Exhibit 4Market Landscape


p. 19

First Field-Programmable Non-Volatile Memoryat Consumable Price Points

Matrix®3-D Memory

DECREASING COST PER BIT

INC

REA

SIN

G

FLEX

IBIL

ITY

NANDFlash

FACTORYPROGRAMMED

USERPROGRAMMED

REWRITEABLE

HIGH COST CONSUMABLE

Mask ROM

New Option in Memory


Exhibit 5Matrix Semiconductor Investors


Date Approximate $ Raised Key Investors1998 through 2000 $80 million Benchmark Capital, Skymoon Ventures, Microsoft,

Sony, Kodak, Thomson Multimedia, SeagateFebruary 2003 $15 million NintendoMarch 2003 $52 million Telesoft, Benchmark Capital Europe, Integral Partners,

Seagate

Note: Information compiled from publicly available sources; it was not verified by Matrix Semiconductor Inc.

Sources: “Delayed Expectations for 3D,” Electronic Business, May 15, 2003, p. 29; “Matrix Semiconductor Receives $15 Million Investment from Nintendo,” February 24, 2003, http://www.matrixsemi.com/media-resources/press-releases/2-24-2003.html, (August 28, 2005); “Matrix Semiconductor Raises $52 Million Round of Financing,” March 17, 2003, http://www.matrixsemi.com/media-resources/press-releases/3-17-2003.html (August 28, 2005).

p. 20

Investors

http://www.matrixsemi.com/media-resources/press-releases/3-17-2003.html




Exhibit 6Worldwide Memory Data (1999-2004)

Worldwide Memory Revenue (Millions of Dollars)

1999 2000 2001 2002 2003 2004

DRAM 23,149.0

31,551.0

11,626.0

15,481.0

17,521.0

26,141.2

SRAM 4,70

1.1 7,47

5.9 4,32

7.0 2,91

3.8 3,39

1.2 4,308

.9 NAND Flash

600.0

1,563.0

1,378.2

2,363.8

4,130.8

6,120.9

NOR Flash

4,133.7

10,141.1

6,873.2

5,820.1

6,764.1

8,771.3

EPROM 435.9

590.0

504.4

328.7

297.8

300.3

EEPROM 939.1

1,319.0

884.7

745.9

839.0

897.1

ROM 968.2

1,188.0

675.5

419.5

356.9

348.4

Others 723.0

900.0

451.0

349.0

374.0

468.8

Total 35,650.0

54,728.0

26,719.9

28,421.8

33,674.9

47,356.7

-

5,000.0

10,000.0

15,000.0

20,000.0

25,000.0

30,000.0

35,000.0

1999 2000 2001 2002 2003 2004

Rev

nue

in M

illio

ns

DRAMSRAM

NAND FlashNOR FlashEPROM

EEPROMROM

Others

p. 21


Worldwide Memory Average Selling Price (ASP) per Megabyte (Dollars)

1999 2000 2001 2002 2003 2004DRAM 1.20 0.99 0.22 0.21 0.16 0.16SRAM 21.74 21.33 12.18 5.11 3.68 2.94NAND Flash 1.61 1.13 0.45 0.30 0.20 0.13NOR Flash 3.84 4.26 2.35 0.96 0.68 0.53EPROM 8.65 4.86 2.87 0.91 0.50 0.39ROM 0.86 0.57 0.51 0.28 0.21 0.18Others NA NA NA NA NA NAAverage 1.60 1.43 0.44 0.32 0.24 0.20

0.00

5.00

10.00

15.00

20.00

25.00

1999 2000 2001 2002 2003 2004

DRAMSRAM

NAND FlashNOR Flash

EPROMROMOthers

Source: Gartner Dataquest Memory Research (May 2004).

p. 22


Exhibit 7Comparison Between Forecast and Actual Memory Prices (1999 to 2002)

$0.00$0.10$0.20$0.30$0.40$0.50$0.60$0.70$0.80$0.90$1.00

1999 2000 2001 2002

ROM ForecastROM Actual

1999 2000 2001 2002ROM Forecast $0.83 $0.59 $0.43 $0.32ROM Actual $0.86 $0.57 $0.51 $0.28

$0.00

$0.50

$1.00

$1.50

$2.00

$2.50

$3.00

1999 2000 2001 2002

Flash Forecast

Flash Actual

1999 2000 2001 2002Flash Forecast $2.58 $1.80 $1.39 $1.00Flash Actual $1.61 $1.13 $0.45 $0.30

Source: Complied from IDC Worldwide Memory Forecast (October 1998) and Gartner Dataquest Memory Research (May 2004).

p. 23


Exhibit 8Matrix’s Second Generation Product (Flagstaff)


p. 24

World’s smallest 512Mbit die:33 sq mm

>800 gross die/8” wafer0.15um Technology

World’s smallest 512Mbit die:33 sq mm

>800 gross die/8” wafer0.15um Technology

2nd Generation Product Family Announced in Nov. 2004 – after delivering over 1 million units to

early customers.

2nd Generation Product Family Announced in Nov. 2004 – after delivering over 1 million units to

early customers.

• High-Density Nonvolatile Memory (NVM)– Lowest cost per bit available– Currently shipping 128Mb – 512Mb– Production capacity: several million units per month

• Field-Programmable

• Archival Life > 100 years

• Compatible with Existing Standards– Standard NAND TSOP and MMC– Specs optimized for use in

standard digital devices

2nd Generation Product Family: Over 5 Million Die Produced


Exhibit 9Article on Matrix 3DM Applications in the Market

In the coming year, toy makers will begin to deliver more multimedia and interactivity at price points appropriate for mass-market toys. The key to making sure this trend continues: large amounts of inexpensive memory to store music, video, games and programming code.

Fragile hard disk drives and scratch-prone CDs have one thing in common - they are both completely incompatible with the usage they will receive from kids. So-called “solid-state” media - data storage that doesn’t use moving parts - offers durability and price points that toy manufacturers can accept.

Until recently, this solid-state memory for data storage came in two principal forms: flash memory and mask ROM. Flash memory offers extreme flexibility in terms of its ability to be programmed and reprogrammed, though it remains expensive. Mask ROM is far less expensive and a popular choice for toys, but decidedly inflexible - turning around a Mask-ROM-published content title can take months.

Owing to a breakthrough in chip design and manufacturing (building chips in three dimensions), Matrix Semiconductor of Santa Clara, Calif. has produced a memory that offers the low cost of Mask ROM with the added flexibility of one-time programmability and quick time-to-market for content publishing. The result is Matrix 3-D Memory.

High Tech Toys

In 2004, Matrix Semiconductor earned the opportunity to work with toy industry giant Mattel on a handheld personal media player designed for “tweens.” Dubbed the “Juice Box,” the lightweight player features a three-inch, full-color screen and a design that is both simple and durable. Video titles are pre-recorded and distributed on solid-state “Juiceware” cartridges.

Mattel had the choice of either using a drive-based system or durable solid-state media. Choosing the former would have meant that the Juice Box™ could have distributed music and video on inexpensive proprietary CD’s - a tempting marketing prospect mimicking conventional CD and DVD distribution.

Mattel instead chose solid-state media - Matrix® 3-D Memory from Matrix Semiconductor - and, in so doing, demonstrated an approach that combined the most amount of fun in the smallest and least-expensive package.

Here’s why:

CostDisk technology would have added either $8 - $10 for a CD drive or up to $60 - $100 for a

p. 25


hard drive to the bill-of-materials. By choosing solid-state memory to publish interactive content on a removable cartridge, Mattel was able to invest in a larger screen and faster processor, and still meet it’s targeted $69.99 MSRP, keeping the player in the critical price range.

Usage Model Mattel understood that if they used a hard disk drive tweens would have to dock their players to PCs in order to move content in and out of the device, a user experience that doesn’t fit much of the target market. CDs, while removable, are not as portable. Small solid-state cartridges provided the ideal usage model as proven by the popularity of the Nintendo Game Boy and other interactive toys.

MediaIn a hard disk drive, the media and the host device are linked - damage to the former means a catastrophic failure in the latter. For CDs, the data-retaining surface is exposed to scratching and staining – likely to happen in a middle schooler’s book bag. By using solid-state removable cartridges, the media exists in an enclosed, truly tween-friendly, form.

Size and WeightSolid-state media shaved the bulk by more than half of disk- or CD-based designs.

SecurityCD-based content can be counterfeited more easily by rogue replicators or PC “ripping” and allow much less parental control of content than cartridges.

Formerly a specialty technology, solid-state media is now an affordable, mainstream choice for toy designers. This holds the possibility of re-inventing older product types and creating entirely new categories of entertainment and educational products with a faster time to market and mass-market price points.

Source: Article written by Dan Steere (of Matrix Semiconductor) and published in TDMonthly (March 2005).

p. 26


Exhibit 10Matrix Semiconductor Product Overview


p. 27

51,500 117,000

2003 2004 2005

Product Flagstaff TrinityPhoenix

Architecture CheckerboardCheckerboard Segmented Wordline

32MB or 256Mbits

64MB or 512Mbits

Process0.15um0.25 0.15um / 0.13um

Mbytes per 8” wafer 17,000

16MB or 128Mbits 16MB or 128Mbits

32MB or 256Mbits

64MB or 512Mbits

128MB or 1 Gigabit

32MB or 256Mbits


Exhibit 11Worldwide NAND Flash Memory Forecast (2005 through 2008)

Revenue (Millions of Dollars)

Shipments (Millions of Units)

Average Selling Price (Dollars)

Note: 1 B = 8 b.


p. 28


Exhibit 12Worldwide Memory Forecast (2005 through 2008)

Revenue (Millions of Dollars)

Shipments (Millions of Units)

2005 2006 2007 2008DRAM 6,181.9 5,895.1 5,490.4 5,402.0SRAM 1,269.1 1,327.3 1,284.3 1,220.5NAND Flash 684.0 906.8 969.8 1,059.5NOR Flash 2,550.0 2,457.0 2,350.0 2,280.0EPROM 212.1 191.6 173.3 160.3EEPROM 3,565.0 3,417.0 3,116.0 2,906.0ROM 121.0 76.4 51.0 40.5Others NA NA NA NA Total 14,583.1 14,271.2 13,434.8 13,068.8Growth 1.5% -2.1% -5.9% -2.7%

Average Selling Price (Dollars)

2005 2006 2007 2008DRAM 4.75 3.42 4.46 5.63SRAM 4.11 3.78 4.04 4.57NAND Flash 10.94 9.17 9.81 10.49NOR Flash 4.14 4.54 5.30 6.37EPROM 1.31 1.16 1.21 1.31EEPROM 0.24 0.21 0.21 0.22ROM 2.46 2.86 3.75 4.63Others NA NA NA NATotal 3.75 3.24 3.96 4.84Growth 13.7% -13.4% 22.0% 22.3%


p. 29


Exhibit 13High Level Information about the Worldwide Flashcard Market

Flash card market revenue was expected to reach $4.32 billion in 2004 and $7.95 billion in 2008.

Flash card unit shipments were forecast to reach 169.3 million units in 2004 and 310.1 million units in 2008.

Total Flash card MB shipments were expected to reach 29.9 billion MB in 2004 and 397 billion MB by 2008.

The average Flash card capacity shipped in 2004 was expected to be 177 MB per card and 1.3 GB per card by 2008.

In 2004, the leading Flash card application was the digital still camera, which was estimated to account for 60 percent of the MB consumption in 2004.

By 2007, mobile phones are forecast to lead Flash card market share by application, with 43 percent of MB consumption.

Source: Joseph Upsworth, “Market Trends: Flash Cards, Worldwide, 2001-2008, Executive Summary,” Gartner, October 11, 2004, p. 3-4.

p. 30


Exhibit 14High Level Information about the U.S. Digital Camera Market

The number of digital camera shipment in the US has increased more than 50 percent each year from 2002 to 2004.

As reported by BusinessWeek, IDC expects shipments to rise to almost 30.5 million in 2005, up from 24.5 million in 2004 and 10.3 million in 2002.

Growth will eventually peak around 2007 and decline slightly in 2008 when penetration exceeds 60 percent. Sales will continue as consumers upgrade or buy additional cameras, but replacement cycles will lengthen as further improvement of features will have less of an impact on consumer satisfaction.

Digital camera shipments in the US, 2002-2008 (in millions of units and as a percentage increase/decrease versus prior year):

33.04 (-3.4%)

34.2 (1.3%)

33.76 (10.8%)

30.47 (24.6%)

24.45 (49.0%)

16.41 (59.9%)

10.26 (57.0%)

0 5 10 15 20 25 30 35 40

2008

2007

2006

2005

2004

2003

2002

With shrinking digital camera margins, memory cards are becoming an add-on accessory that retailers are promoting to increase profits.

In 2004, memory card-owning households owned an average of 1.4 memory cards.

The most popular memory card sizes were 128 MB and 256 MB, with each accounting for 28 percent of units purchased in 2004 (for a total of 56 percent). Higher capacity memory cards were also purchased; 512 MB memory cards had 12 share of volume and about 4 percent of units were 1 GB or larger. As prices come down, households will most likely purchase higher capacity cards.

Approximately 29 percent of households that owned digital cameras were worried about losing their photos stored in digital format. Another 20 percent admitted to never having thought about the issue.

p. 31


How digital camera users store digital photos, 2002-2004 (as a percentage of respondents who own digital cameras):

73%

40%

21%

24%

10%

6%

15%

8%

1%

0% 20% 40% 60% 80%

Folder on my hard drive

CD

Floppy disk

Printed phots

Online photo service

Zip disk

Flash media card

DVD

Other

20042002

The number of digital images kept in digital format in 2004 was 9.4 billion. This number will grow to 12.3 billion by the end of 2005.

Sources: “Shutterbug Standards” (May 5, 2005) and other information publicly available from eMarketer; “Photo Industry 2005: Review and Forecast” and other information publicly available from the Photo Marketing Association International.

p. 32

Stanford Case Study B

Documents