Dov Moran, President and CEO
Simon Litsyn, Chief Scientist, Future Technologies Group
Ronit Maor, CFO
Dov Moran, President and CEO
This first analyst day is much more than an analyst day. [inaudible] and go directly to the numbers. I know that analysts like numbers. This is what we have done in the last five years. $45M in revenues in 2001. $650M in 2005 with net profit of $53M. That’s what we have seen to date.
We made an announcement about a month ago that we expect to see revenues approaching $1B this year. If you compare it to the revenues of 2001, it is over 20 times the revenues in five years. I believe this is a unique achievement.
The question which we could ask or should ask is how did we do it? How did we get there? There are many potential answers to this question. Technology, that’s a very valid answer. Technology is very important. We are a technology company. We invest a lot in technology, but it is not what brought us here.
Strategic relations. We invest a lot in strategic relations. We have great partners, but it is not what brought us here.
Marketing. I think that we have great marketing. We are working in great markets, but that is not the reason.
Although we do have a very good OEM model. We are very much focussed on supporting our customers. We have great relations with customers, but that is not what brought us here.
Focus. We are a very focussed company. From our first day 17 years ago we are only doing flash data storage products. I think that timing (?) is critical. We brought mDOC to the market at the time that the market moved from regular standard NAND cellular phones to smart phones. We brought DOK to the market when floppies were not anymore the solution for transferring data from one computer to another. But again it is not the answer.
Inventory control. We do have excellent inventory control and very good logistics. Very good infrastructure, but logistics by itself doesn’t lead to success.
Excellent R&D. We have wonderful R&D people. I think that we have one of the best or the best R&D groups existing in Israel and in general overall, but it is not enough.
Our answer is innovation. We are a company which is driven by innovation. Actually msystems is structured to encourage innovation. This is a company organization chart. What you can see here is our four divisions. This is the enterprise devices. This is the mobile handset venders division. This is the embedded systems division. This is the mobile network operator’s division. And here is the corporate.
First of all, people. The first value of the company is people first. We do invest a lot in people. We do invest a lot in choosing the right people. We choose the people that have the abilities to innovate and think out of the box.
Actually, if you look at this structure, you can see here many things, many entities that deal with R&D and business development. Most of the divisions and the corporate. It is probably not the optimal organization structure for efficiency, but this is clearly the optimal organization for promoting innovation.
Innovation at msystems is a way of life. This began at our beginning in 1989 when we were founded in order to make flash data storage. And later on we brought to the the market the TrueFSS, the True flash file system, which was the first software to take standard flash components and make them something that emulates a hard disk. Something that today that is very obvious. Almost everything that you see today is a standard flash that functions one way or another as a hard disk.
The TrueFSS that we brought to the market in 1993, [pause] actually when we brought the TrueFSS to the market, the market perception was that standard flash cannot be used for disk emulation and we showed that we could do the impossible.
Later on the mFFD, or what we used to call the FFD, the fast flash drive. This was about 10 years ago. It was quite amazing to see a large company [Samsung] come to the market a year ago with a similar product claiming that this was the first flash-based solid state disk in the market, about 10 years after our introduction.
And then we brought the DOC, as a solution of a chip functioning in a computer as a hard disk. We took this DOC to the mobile market with our mDOC at that time that no one believed that it is possible to put on NAND flash both code and data. It used NAND flash as an embedded solution for cellular phones.
We brought to the market DOK or mDrive or the USB flash drive. We actually invented this concept back in 1999 and we brought the product to the market in 2000.
We converted the DOK to U3 together with Sandisk in order to initiate a huge change in the IP industry, a change that will influence each one of us in the way we are using computers and using data. We will see later on a demonstration of U3 and how U3 can change your life.
Then came MegaSIM. We brought MegaSIM to the market at a time when the SIM vendors were struggling to move from 64 KB to 128 KB and then to about 256 KB. We brought to the market at that time the MegaSIM, a SIM solution with 512MB. Not 2 times, not 20 times, but 2000 times the capacity that was there before.
Innovation is translated to IP, which is patents, technology and know-how. Patents become a very important factor in this market. Looking at the proof (?), in 2000 we were an immature(?) company, but we didn’t focus on filing patents. We did file patents. We did file very important patents. Each one of our patents that we filed before 2000 is critical in order to make flash data storage. In the last five or six years we began the process of filing many patents. Investing not just in quality, but as well in quantity.
Actually what you can see in this slide, is that we grew the number of patents from about 20 patents and patent applications in 2000, to 200 patents and patent applications at the end of Q1 2006. Each one of those patents is a very strong patent.
We need the IP in order to promote a business model that we created that is called vFab. vFab stands for virtual fab. Its another innovation of the company. I don’t think that there is another company that is working under such a model. In the vFab model, what we do actually, is trade our IP or technology or patents actually against a committed supply at favorable terms.
We had agreements in the last years with the majority of the flash players in this market. The names that you can see here. Those agreements are a variety of agreements that include JVs, jointly developed products, license fees, supply agreements, and so on.
Back to Arthur C Clark with his leading statement: “The only way to discover the limits of the possible is to venture a little way past them into the impossible.” Arthur C Clark was a SF author, a great one. What I am going to show you now is not science fiction. I am going to talk about reality.
Here it comes [talking about 2001 music and slides]. Ok, x4 technology. What is x4? Everybody who is familiar with the NAND flash market, knows the terms SLC and MLC. SLC stands for single level cell. MLC stands for multilevel cell which actually is two bits per cell.
If you look at the standard wafer, this is SLC, every juncture (?) represents a cell usually it is one bit, zero or one. In MLC a cell is actually 2 bits: 0-0, 0-1, 1-0, 1-1. Actually in the same wafer size, of SLC there are, lets say 100 components of 1GB, you will get in MLC, 100 components of 2GBs or 200 components of 1 GB. Now this is not exactly 200, there are additions to the silicon required to implement such things, but still theoretically the number is almost double.
This is x4. x4 technology enables 4 bits per cell. In the same wafer that has 100 components of 1GB, you can create, theoretically, 100 components of 4 GBs or 400 components of 1GB.
Its all very obvious, and clearly there are huge cost reductions involved in doing this. The question is why didn’t others do this before? Why is 4 bits per cell so difficult?
This is a very schematic drawing that shows SLC, MLC, and x4. You can see here the SLC which has 2 levels. Electronics generally is voltages, there are two voltages that differentiate between the zero and one.
You can see the MLC which actually is based around 4 levels of voltages. And here is the x4 which requires to differentiate between 16 levels of voltages. The distance between adjacent levels here in the x4 is significantly shorter.
Moreover, the electronics is not so simple. There is noise. There is statistical behaviors of electrons, of voltages. All of these statistical behaviors create overlap of both voltages that actually makes the life of reading the correct data, very difficult or almost impossible.
I can describe it in another very simple schematics. This is SLC here. Specific specifications at a specific price. When you go and implement MLC you basically get something which has better cost, but the specifications are much lower, which means lower performance and lower reliability.
When we try to implement 4 bits per cell, the phenomena that you will get is lower price and on the other side, much lower specifications. Practically those specifications will be below the market requirements.
What we are doing with the x4 technology is to combine the media with 4 bits per cell with x4 controller that together creates an x4 product that significantly reduces the cost while staying above a minimum requirement.
How do we do it. The solution that we bring to the market is very much based around a system approach. Usually when fabs or silicon vendors are working on solutions, they are very good at focusing on the silicon and on the difficulties that they are trying to resolve are silicon problems.
We actually look at the solution and problem as an issue that should be resolved not as a stand alone memory, but as a memory and a controller that works together. The x4 controller makes the 4 bits per cell NAND flash viable. It makes the immediate transition available from MLC to x4 with minimal cost addition.
Actually what we are doing is to push the x4 specifications up by advancing and focusing on unique algorithms for faster write performance and normal(?) data retention. When doing it we are actually providing [missed words] optimal algorithms that do much better and effective endurance.
What is the value of it? When looking at the value of it, we look at two scenarios. The first one is in a fab with 50% GMs and the second one is fabs at 20% GMs.
This is the scenario of fabs at 50% GMs. Lets look at the costs. We put the cost of MLC NAND at 1 GB at $10. It doesn’t matter in the calculation actually you will see later on, whether it is $10, $8 or $2. $10. What would be the price of an x4 solution? Theoretically we can use x4 technology in order to get 1 GB [2GB] at the same profit: $10, but in reality there is additional cost, additional overhead that is required in the solution. We took here a very conservative estimation of the additional overhead we made it at 40%. We believe that we can get, moving forward, to much lower overhead than 40% and actually make this overhead as close as possible to 0%. Taking into assumption 40%, we get 2 GBs at the cost of $14.
Now lets look at the price. We said 50% GMs. The meaning(?) is the component that its cost was $10 would be sold in the market for the price of $20. What would be the price of a 2GB x4 product? Practically or theoretically it is supposed to be double the price of the 1GB component. To multiply by $20 is $40. We know today that the price in the market between the 1GB SLC, the 1GB MLC is roughly 10%.
When we believe that we can get the specifications of the x4 technology, the performance of the x4 technology to very close or even sometimes better than MLC, we took a similar conservative assumption of 10% reduction and therefore the price of the 2GB x4 would be $36.
Now lets look at the profits. In the case of MLC, its a $10 profit. In the case of the x4 2GB, same silicon, same fab, same equipment, same wafers. No change at all at the fab, this is $22. Profit more than doubles.
Now lets look at the scenario of 20% GMs. Lets [do] again the calculation. The costs are the same costs, $10 cost for MLC. $14 cost assumption for 2GB component. The price, 20% GMs. This [MLC] would be a $12.5 selling price in the market.
Here again the calculation is x4 2GB will be sold at the price of the $22.5. Two times 12.5 x 10%. Here are the profits: 2.5 vs 8.5, more than 3 times the profits.
x4 means higher profitability. Actually much higher profitability. More than double in the case of 50% GMs. More than triple in the case of 20% GMs.
Someone asked me what about the case where fabs are selling at negative GMs. In this case it is profits instead of losses.
The next question is how do we fit it into our strategy and the answer is very simple. Innovation that leads to IP that leads to vFab that leads to profitability and success.
We have this innovation, the x4. We have many patents around it. Many technologies developed around it. Much stronger emphasis on the vFab model that works for us so well until now.
And now the future. According to the analysts, the flash market is going to grow. The growth expected is phenomenal. This is going to be, according to them, according to others, in many many new devices that are coming to the market.
Actually looking backward, what happened in this market is that we moved from some data, some files, some contact lists, and maybe applications into powerpoint presentations which were MBs already and then into music that was 100s of MBs or even a few GBs.
And we are just moving now from music to video, which is many GBs. And therefore all of us can dream, think about smart phones and GPS and smart pictures and many things around it, and gaming machines and personal video players, and many things that none of the analysts and not even us including me, know what is going to be there in the future.
Because the future is unexpectable [unknowable]. Exactly in the way that in 2000, when all of us at msystems and the analysts were talking about the great market of internet appliances. Today there are smart phones(?), but no(?) internet appliances. The two drivers were [inaudible] which no one expected in 2000.
So we truly don’t know the future. What I can tell you about the future and I am very much assured about it, is that there is going to be a lot of flash, there is going to be a lot of x4 flash and there is going to be a lot of msystems’ x4 flash.
**** end of Dov’s presentation ****
Simon Litsyn, Chief Scientist, Future Technologies Group
Good morning, my name is Simon Litsyn. Several words of introduction. Probably not needing the safe harbor [referring to not reading the safe harbor.] Simon Litsyn, Tel Aviv University, Electrical Engineering Systems.
For the last five years I have been working for msystems. Currently I am on a leave of absence from the University. Hold the position of the Chief Scientist in the Future Technologies Group of the company.
Here is the Safe Harbor statement. My goal here in the presentation is to elaborate on some research, scientific issues which are behind this development of the x4 controller.
I will start with theoretical basis of this invention and after it how we overcame the challenges of development and implementation of the x4 controller.
Our main goal, our main thrust is to accommodate as many bits per cell as we could. Clearly it is the main cost driver of the flash storage.
There are two options which can be used here. One is to double the number bits per cell, just to increase, to go from single level cell to multilevel cell and beyond it. Another one is to improve the technology of flash. For example by decreasing the transistor size used in the flash memory.
The question is which one of the two ways is better. The answer is clear, that both are good. Both dramatically cut the flash costs. While the first one, increasing the number of bits is much cheaper to implement. It is mostly algorithmic solution which doesn’t require changes in the flash itself.
But of course I want to emphasize that both approaches may be implemented at once because they are [inaudible] each other and go one thing another.
This is the picture of cell voltage distributions. Put some information into the flash, we do not have some specific voltages for each of the various values: 0-1, 0-0 etc. for the MLC technologies, but we have probability distribution of the position of the value of the voltage in the flash.
Clearly you see that the closer the distribution is the more difficult it becomes to fix the information if it is read in the wrong way. When the number of the levels increases, the probability of erroneous readings increases as well. Thus the number of levels cannot be increased infinitely and it’s constrained by the quality(?) of the information we can fix.
Now information about the distortion parameters. About the distributions of the voltages which are used for storing the data, we might get [missed word] the optimal number of the levels. This number should be good enough, this information should be good enough to keep the reliability of the flash under data retention and program and erase.
What is happening is that with the time after the information is stored, the bits of the distribution growth, intersections between the levels which causes errors. We want to correct these errors.
First of all let us look at the theoretical limits of the numbers of the quantities of the information we can store in the flash. The number of information bits per cell, IBPC, and this will be a normalized measure. This measure, to be fair, should be defined under the assumptions given the reliability is called for in this flash.
Here on the graph is the computation of the number of IBPCs, of the number of information bits per cell as a function of the reads (?) of the bit distribution. Here you see in the right hand of this graph, the green line corresponding to single level cell, to the two voltage levels prevails, and gives the best results. So it seems that even if the distribution is very wide, we have nothing to do but use the SLC scheme.
So then you see that the four voltage levels already become better. Eight and sixteen and then we have checked it, suddenly we have found out that actually we should be able to implement 16 voltage level cells.
Now what were the challenges. First of all the conventional error correction methods are not good for it. There are several effects that cannot be overcome from 4 levels to 16 levels. The theoretical models for the distortions of the voltage levels are not good. The standard Gaussian models which are used in most of the theoretical research are very far from the reality. The market requirements for read/ write performance impose severe restrictions as well as the restrictions for the silicon area and power consumption for the controller implemented base(?) for technology.
However we managed to develop the x4 controller by using a new set of advanced signal processor algorithms. By using advanced multilevel code instructions, specially designed for this project. All the algorithms were implemented and checked in hardware with optimization for multiple parameters such as silicon area, power consumption, and read/ write performance.
Here you see the graph comparing our designs with conventional design currently existing in the market solution. The upper curve, the blue one, is the theoretical limit calculated for the currently existing raw(?) flash.
The green line is what one could achieve using the existing currently in the market solution. The middle curve is the one which follows from our approaches. We are somewhere in the middle and you see that we double the number of bits per cell.
The natural question, what were the problems of the others? Why hasn’t it been done before? First of all, the use of conventional error correction methods doesn’t allow it to go beyond 4 levels. Then it was a general agreement that going beyond 4 levels is impossible. This was based on the traditional theories and implementations. We had to develop a new innovative set of flash-specific signal processor algorithms and we developed some specially tailored algorithms within that project.
By the way, all these new approaches are patented and we are protected [inaudible].
To summarize, we have a proven x4 technology. It is a viable technology. It is on the verge of commercialization. It is the fruit of conceptual change, followed by years of scientific and laboratory, lab, research.
Now that we have approached the theoretical limits of information capacity for, I emphasize, for current raw flash memory. I probably should reference the safe harbor, but it [x4] is not the end of the story. [this seems to imply that FLSH is working on a new design of flash memory] Thank you.
**** end of Litsyn’s presentation****
Ronit Maor, CFO
Good morning everyone. Let me introduce myself. My name is Ronit Maor and I’m the chief financial officer of the company. In my presentation I will discuss both the core competencies of msystems and how they have enabled us to show the strong financial performance that we have shown until today and I will discuss x4 and how we expect it to positively our future financial performance.
Before going into that, my presentation will contain forward looking statements. Let me refer you to the safe harbor statements done earlier by Ilana.
Again I will start with msystems core assets, our core competencies and the strong growth they have allowed us to show til today. As Dov mentioned our core assets, our core competencies, are our innovation, the IP, the know-how, the technology. These assets have allowed us to bring to the market innovative products, among them the mDOC, mDrive, MegaSIM, U3.
Both of these assets, both our core competencies and our products have allowed to both create new markets and grow within them: the mobile market, the USB market, the embedded systems market. And at the same time have allowed us to obtain strategic supply agreements. Agreements which gave us both guaranteed capacities and favorable terms.
The combination of all of the above, allows us to show tremendous growth in revenues and in increased profitability.
Just to remind everybody the numbers that msystems has shown until today. As a separate note, other than this specific slide, all the numbers are on a performa basis and do not include the effects of the consolidation of the venture.
Dov has shown you this graph, this tremendous growth in revenues that we have shown in the last five years, showing average annual growth rates of 93%.
On a quarterly basis also we have shown a very impressive growth rate that in average was 21% QoQ. That was a result of our sales to different markets. The ones that grew the most, the fastest in the last three years was sales to the mobile market, growing at an annual average of 185%. Sales to the embedded systems market growing by 34%. Sales to the USB flash drive market grew by over 100% on average over the last three years.
This tremendous growth has allowed us also to leverage our operating expenses, taking them down from 66% of revenues in ’01 to 16% of revenues in the last year, resulting in a very impressive growth in profits.
Let’s talk a little bit about the future and how we believe x4 can impact our financial performance. Let me take you back again to our core assets, we have added a very significant asset for msystems and that is x4.
With x4 now, we will have stronger core competencies allowing us to bring to the market smarter products. With that to both expand our existing markets and to create new markets and at the same time to allow us to improve our supply agreements. All of that expected to allow us to show continued growth in revenues and improve our profitability.
Let’s dig a bit into those things. As for the growing revenues, the expanded markets, the division managers which will speak in the second part of the day, they will give you more highlights about that and how x4 is expected to allow us to grow revenues.
I will focus a bit more on how x4 is expected to improve our profitability. With x4 msystems has broader and enhanced assets, as we spoke about our IP, our know-how, our technology. With these enhanced assets, msystems has a stronger negotiating position when we come to the flash venders to negotiate our supply agreements. Therefore allowing us to both increase our guaranteed supply and to get even better terms for that supply. What is the result of that? Improved cost structure for msystems.
What does our cost structure look like? Msystems’ blended cost structure, our GMs, are dependent on the blend of our supply. We buy from different suppliers ,as you all know. Clearly first of all msystems is utilizing the suppliers that give us more capacity at better terms, better prices, with the highest GMs. Once we utilize this primary supplier, we start buying from our secondary suppliers that give us lower cost structure, so in the blended GMs, the GMs are going down.
How will x4 impact that? x4 will impact that in two different ways. First of all, it will improve our cost structure, the prices, the terms in which we buy the flash components, allowing us to get better GMs and at the same time we believe it will allow us also to gain more supply on those improved GMs.
Looking at this graph, you can see how overall it is expected to significantly improve our cost structure.
Going back to msystems and showing how x4 again gave us more assets, better negotiating position, improved our cost structure, clearly allowing us to increase our profitability.
To summarize it, x4, our new breakthrough technology, acknowledges that we believe will change the NAND industry as we know it, will drive msystems’ financial growth. Thank you
**** Q&A Part I ****
Q, Dan Amir: How far along is the x4 right now? Have you talked to companies? Is this something that is in the very near future? Is it still very futuristic? Have you approached different foundries? Different relationships? Where does it stand from the commercialization perspective? It looks like you are making some statements, obviously that significantly impact your business model. I guess when shall we see it?
Dov: I would not come to speak in front of this audience if this was just an idea. We have been working on this effort for five years. Today we are in the position that we have high confidence in our ability to deliver. We are talking with fabs. We did some work with fabs. We are in the position today to have samples working and the ability to demonstrate x4 technology, read/write and all the activities required out of the product on 4 bits per cell.
We do believe that it will take time to get it into commercialization and mass production. We do believe that we will be in mass production beginning in 2007.
Q, Craig Ellis: With regards to commercializing x4, can you talk about where your priorities might be with respect to commercialization from licensing relative to commercialization through new or expanded sources of supply with current or prospective partners? Will you need to have it in production to generate, in you mind, licensing revenues from a third party?
Dov: Clearly our objective and our model is not collecting license fees and becoming like a company and I think that you us know well enough. What we plan or what we are looking for is to get fab capacity at the right prices. This is the direction. We will work with at few selected fabs in order to get these benefits and provide big benefits to those partners.
As I mentioned it is going to be a limited number of fabs. Its not free to anybody. We are already in the process of discussing this with a few partners.
Q: A couple of questions. You say you have manufacturing samples already. At what line widths are they? and are there any theoretical limits to shrink say at 90 nanometers or 45 nanometers? How difficult does it get as you start shrinking it? and does that set a limit?
Dov: One of the great things about this x4 technology is that you don’t have to go to any specific change in technology or fabs in manufacturing equipment compared to what exists today. So basically a fab can make an x4 media at the same technology you currently have, whether its 70 nanometers or 55 or in the future, 35 nanometers.
What we have today are samples at the most advanced technology, but clearly the idea is to move it with the technology and bring the x4 always with the most advanced technology available.
Q: Do you see yourselves having exclusive access to x4 or will your fab partners be able to sell x4 products to your competition?
Dov: We see the x4 concept as the concept where we are providing the technology to a partner and get a portion of the fab at a specific lower price. That’s the direction.
Q: I’m curious about the barriers to entering into the 4 bits per cell market for your competitors. Do you have a sense on when or if they are going to also emulate this approach? This is question number one. Question number two, you said that in 2007 you expect to be shipping in volume. I wonder whether you think that is way late 2007? Should we think about changing numbers in 2007? Or is this basically 2008 where we see the upside in numbers? and can you also talk about how additive this is to your GMs.
Dov: First of all regarding competition, without providing any names you can see several times in the history, some competitors claiming that x4 is impossible. We know there were some ideas about beginning to work on x3 at some competitors as well. We do believe that for x4, the advantage in time that we have is significant, is huge. In addition to the time advantage, we covered ourselves very well with patents around the concept that we are using to make this technology available.
Regarding how much GM can it add to us? Its too early to get into numbers, GMs and influence on revenues. Clearly, the impact on GMs and revenues, I believe, is going to be huge. I won’t go into details of numbers and exact time of when this influence is going to happen.
Q: I have a couple of technical questions. I just want to make sure I understand the fundamental technology here is software algorithms embedded in semiconductors that are able to differentiate the 16 levels [pause, while Dov laughs]. It’s not a circuit level design a la Saifun or something like that?
Dov [laughs]: It’s much, much, much more complicated. Its things that relate to the media itself. To the silicon of the flash itself. How construct the flash itself. We didn’t go into the details of what we have done there. Continuing, with a very complicated, a very difficult to implement controller with over a million gates. Very complicated and breakthrough technology of error correction and other things required to do there. And of course very heavy software that covers all of it. It is a total solution which involves silicon, flash, controller, software, systems’ product(?).
Q: A quick follow-up question to that. Is there any timing implication to the read/write action if you have to implement such complicated algorithms?
Dov: Clearly with such a device, there are aspects which relate to performance, but we do believe that with a lot of innovation that we have done in the controller we succeeded to bring the performance of the solution, the performance which is very close to the performance of MLC. In some occasions [respects] better than the performance of MLC. In some occasions less.
Q: A couple of things. One, you are familiar obviously with the NROM approach, the quad bit approach that is being tried by Saifun and their partners. How does x4 stack up on a lot of different metrics in terms of ease of implementation, read/write? Again I understand it is early, but I would like to get a sense of that. I have one more question too.
Dov: We appreciate very much the Saifun technology. As you know, we were one of the first investors in this company. We do believe and hope that this technology will be very successful. Today, the NROM technology, what relates to flash data storage is about 1 per million of the market. Its a market over $10 billion. We talk about a 5 million market for the NROM. Clearly NROM will become a player in this market of flash data storage. We believe that our technology can be utilized [with NROM]. NROM maybe having their x4, whatever, becoming x8 or similar theoretical x8. We are waiting for Saifun technologies to mature and succeed. We hope very much to work with Saifun.
Q: Ok and are there any particular applications where you think this will be commercialized first? Would you see it in the retail card space where you don’t play that much now? or do you see it in embedded handset applications? Where do you see it particularly commercializable first?
Dov: Let me tell you something about MLC. MLC was brought to the market several years ago by Toshiba and Sandisk. When MLC broke into the market, the approach was that MLC would be used only in very very specific applications where performance was not important and reliability was not that important. Actually what has happened today in the market, is that Toshiba and Sandisk are moving most or almost all of their production to MLC, more than 90% actually to MLC. Samsung and Hynix and everybody else in this market is heading as fast as they can into MLC and practically in the coming year already, most of the market of flash is going to be MLC. MLC is designed into any sort of application.
We do believe that a similar situation will happen with x4 with one change. When MLC came to the market, there were no MLC controllers. And actually, it took a couple of years for companies to develop the appropriate controllers that allowed the usage of MLC.
The difference between that and our solution, we are coming with the media and with the controller to supply a solution that can replace immediately, MLC. And therefore we believe that the entrance of the x4 into the market, the adoption of the market of x4 is going to be much faster than the adoption of MLC, while at the end of this game, you are going to see the majority of the market actually using x4 technology. Where? Anywhere: memory cards, usb flash drives, embedded flash data storage, mDOC, and clearly solid state disks. Everywhere.
Q: Are you going to be working with existing partners or new partners for x4? It might be hard to tell. Second question: How do you think about your ability to actually have higher GMs vs your ability to drive a lot more bit growth with similar GMs? Thanks.
Dov: First of all we are talking with, [pause] I wanted to say all, but I would say most to be more accurate [probably leaving Samsung out], of the flash vendors and it is clearly new ones [Intel?] and old ones [Toshiba and Hynix]. We are doing the x4 technology with the approach of improving the GMs and improving the ability to supply much more flash. We were in the last year to raise revenues very significantly. I think that quite a good job in fulfilling this mission.
In the coming years, in the years to come, we are expected, we should improve the GMs as well. We didn’t see the GMs as the most critical factor in our success. I think that we did right, focusing on growing the revenues and becoming a large company and a large player in a market that is expected to continue and grow significantly by a factor of many(?).
We want to be there. A very important player, one of the most important players. Clearly it should be done as a combination of revenues and profitability. One of the ways to reach profitability is clearly GMs.
Q: Just one quick question on the controller. You said it was around a million gates. Will the controller be separated from the media? Or will each piece of NAND have a controller built in?
Dov: Its a separate controller. It’s not part of the media. With such quantities, you don’t do it as part of the media. It’s different technologies, manufacturing and optimizations. So its separate, but one approach of selling such a component is the MCP combination of the controller and the flash as a component. It actually provides regular NAND MLC capabilities.
Q: For the fab, if you separate out the controllers, bringing in the x4 technology, on the present process, shouldn’t be too difficult for them?
Dov: Yes that’s what we said. The fab doesn’t have to be redone by new technologies or new equipment or anything else.
Q: Since you are coming to market with x4, which is for a specific controller, I was just wondering if you are going to be opening up the IP on the controller or is that something that you are going to license out? How will other potential customers, users of x4 actually get access to the controller.
Dov: They don’t. If we license it, we will license to specific partners. The ones that we won’t license the technology to, we believe, will not have access to the controller or to the technology.
Q: What I am trying to ask, if your partners in manufacturing the x4 are selling to parties other than you, how do those parties actually make use of x4? Where are they getting the controller?
Dov: That’s exactly my answer. They don’t.
Dov: The technology is worthless without the controller. Bringing to the market, NAND flash, 4 bits per cell, without the controller, it cannot be used. It is not something that anybody can use. The controller is part of the solution. One thing which is very unique about what we have done, looking at the whole issue of the controller and the media, as joint devices that always work together, never can be separated.
Q: So who are the other customers of your partners? Who are they going to sell this to?
Dov: We do believe that our partners will sell the controller as well.
Q: So you are giving them access to the controller?
Dov: Yes of course. We license our partners the controller [pause], no actually we will sell our partners the controllers. That’s the approach.
Q: Will you make IP on that? Or that will be included in the margins?
Dov: Through the process of negotiation, you always trade off between this and this. You can go into a situation where you sell them the controller with no profits or very low profits, but get more capacity at better prices, or maybe get lower capacity, but sell them the controller with high margins. Clearly there is a balance between those issues.
Q: I know you don’t have to retool fabs, but is there an actual change in the transistor structure of the NAND die? or is it more in how the controller and the die are integrated?
Dov: The transistor of the NAND flash is not changed at all. It is the same transistor, same technology. Clearly the circuitry around it [NAND transistor] has changed.
Q: Ok, that’s helpful and then secondly when you think of applications, obviously there is mass storage, but have you had any conversations with notebook makers for solid state drives and instant on PCs? Is there a particular sweet spot where you see this really being a killer application?
Dov: One of things we have done after several years of development, getting closer to a solution, is to go and talk with potential customers. We did talk with potential customers about their requirements and what they are looking for. The answer from everybody was cost. Usually for free, that’s the best, but clearly we have the requirements regarding reliability, regarding the number of write cycles, the performance. We did the solution to provide or comply with all the requirements.
We are yet not in the position to talk with customers about a total product of x4, let us go into production first, have products, then sell the products.
Q: In the lab, just one follow-up, in the lab, anyway, is the read/write and reliability similar to or better than a hard disk drive? Spinning media?
Dov: We didn’t go into testing solid state disk drive of x4 to compare it to a regular hard drive. We do believe that such solutions are possible, a lot of use(?) a lot of work. We still are in a way to create products that will replace hard drives in the market. This will eventually happen.
Q: A clarification on the patents. Are your patents actually granted? or are they still pending under application?
Dov: We do have several patents granted, but most of the patents that were listed there [referring to slide] , you can see the numbers, are patent applications.
Q: So the core patents are still pending under application then? Most of them.
Q: Can you talk about this technology in terms of the NAND, x4 die, and the controller? Which one of the pieces are you more comfortable with in terms of the development schedule?
Dov: It is very difficult to make such a division of where the efforts were, but I would say, I would put the emphasis, and point to where we put most of our power. This is clearly on the controller and the software. But the controller and the software would not work without ideas related to the flash itself. And clearly the opposite, the flash, 4 bits per cell, there is no way to work without the controller and system approach.
Q: Can you say on what part of the technology node you developed the x4 NAND die and also the controller technology. What source of technology was that?
Dov: I won’t go into it. First of all it doesn’t have a lot of importance. When we go into mass production, we will go at the most advanced technology. This is very, very clear, obvious. We did work at a very advanced geometry, but I think that the number is not relevant, and I won’t not expose it.
Q: Ok, the last question, in terms of Andrew’s question, in terms of reliability, and all that stuff. Do you have all the conviction on the reliability of this right now? How long does it take? In terms of production is this more in the first half of 2007 or the second half? Thank you.
Dov: Well, we do have, we are testing the solution. We are in the process of having working silicon, many tests around it. We do have a very high level of confidence about the reliability to go into mass production. As I said, mass production will be in 2007 and I won’t go into which month or which quarter of the year will it be.
Q: Just one follow up question regarding the patents, and I’m not sure you want to comment on it, but there has been some stuff in the press about some lawsuits around the patents. Is it relevant to this technology, or if it is, do you feel it is a problem or not?
Dov: I prefer not to refer to it. It is in the court. It was in the court, by the way, and rejected by the first circuit of the court. Went to another court now. I prefer not to refer to this issue at all. Practically we do have very very very high confidence that it is not an issue at all and that it’s totally unrelated even to the x4 technology.
Q: With regard to the partner side, the fab side to the x4. How confident are you with your partners at this point? And also what kind of yields do you need to [pause], obviously you are going to have 15 or 16 fractional charges here. That is a major step for people who are used to only just a few?
Dov: We are working with partners. We are in the process of getting into agreements and the process of development [this seems to imply an agreement has already been reached with at least one partner].
We do have a very high level of confidence in the issues that relate to the part of the design. Part of the concept is that actually by implementing very, very strong and unique error correction codes, might allow us even to bring in the market a product that will have better yields than MLC.
I know it sounds surprising, but as you will see, the concept and the way of what we are doing, there is the potential there, that the x4 technology will provide better yields than MLC. Now of course all it cannot be guaranteed. All of it is in the process of R&D. All of it is related to the fab’s requirements to ramp up.
Clearly in the process of ramp up initially the yields will be lower, will be low. And then clearly in time they will improve. We do believe that moving along with the technology, over the years, the fabs will have the ability to improve the yields overall compared to what they have today due to the fact that this is a system solution rather than manufacturing silicon that is measured for bit errors and if there are bit errors, its dumped away as garbage.
So the level of confidence here in the ability in providing high yields, exists [is high].
Q: Just a follow-up or switching gears here. Are you aware of anyone else in this industry that is looking at at 16 fractional charges. I know that one of your partner/competitors is working on 3D. What do you think about the viability of 3D? And who do you think is a potential competitor in trying to emulate your product?
Dov: Clearly we cannot go to the competitor’s labs and see what they are doing, but from what we see, we do have all the confidence that one can have that no one else in the market is actually working or was working on such a solution.
We heard several of our competitors, claiming one time after the other, that 4 bits per cell is impossible. We are always happy to do the impossible and we are doing the impossible. We did it already several times.
3D, 3D seems like a very interesting technology. We are not involved in it at all and I don’t have enough data regarding the potential success of it. We wish our competitors, partners, to be successful with whatever they do.
Q: Did I understand you to say that you believe the x4 technology can be applied to cell trapping technology also? To nitride-based storage? For example cell trapping that Saifun is using to double storage using cell trapping? Can you apply x4 to that?
Dov: Theoretically you look at such a solution [x4] which is a combination of controller, software, silicon systems. Clearly one can take such this solution [x4] to any sort of technology including NROM.
Q: Thanks for taking the follow-up Dov. Really two questions. One, at least in the initial implementation, is there anything about either the size of the controller, the voltage that it is operating at etc, that makes it more applicable to certain applications whether it be DOK vs others such as the mobile products? And then secondly, as we look at some of the standard products that are out there like memory cards, how does this implementation work with say an SD implementation or a standard MMC implementation?
Dov: The controller is basically a standard controller. It is a very complicated one, but you can manufacture it in any fab that makes controllers. There are plenty of those fabs.
The cost of the controller is tens of cents. That’s the level of cost existing for such as controller. As complicated as it is, if we would go to make such a controller several years ago, it would be many dollars or maybe tens of dollars. Today with the advance of technology we can make such a controller for tens of cents.
The solution can be used(?) easily for many sorts of memory cards. Actually one of the features of the controller is that it is implemented already, in addition to other things, specific card interfaces. So basically you can take immediately our controller and bring the card. Just like that.
Thanks a lot. Thank you and good bye.