Weebit Nano Limited (WBT) Earnings Call Transcript & Summary

We're all set. So welcome, everyone here and welcome, everyone on the Zoom. I see we have quite a few people on the Zoom there. Always good to see you guys. And very exciting times ahead as you can tell from the title of the presentation. I'll be talking about where we stand right now and some of the things that we're expecting during the next year or 2. So let's see -- okay, now it's working. So you guys don't need the intro to Weebit and I'll skip this very quickly. I think everyone here already knows Weebit and what we do in nonvolatile memory. And I like to tell people that my shareholders are probably the biggest experts in semiconductor in Australia. And these are slides that I presented last time. So this is more for other meetings. But I want to focus here on just what built up to where we are today, and I think that kind of sets the stage to what's going to happen in the next couple of years. And many of you have been following up on us, but starting even before 2010, I would say even 2005, the semiconductor industry already realized there's going to be an issue here with flash. It's going to be hitting the wall. There were all kinds of headlines like you can see here and people were talking about it. And so we actually had quite a few different technologies proposed. And I guess -- it's shared, so it's fine. So there were ferroelectric RAM, MRAM, CBRAM, PCM, all kinds of technologies that were proposed. And we actually started going through them. But one by one, they fell off because they were too expensive, too difficult to implement, to manufacture, et cetera, et cetera. So by the time we got to about 2020, there were really 2 technologies that were left in the game, which is MRAM and ReRAM. Now MRAM was already in mass production, okay? MRAM, there's a company in the U.S. called Everspin that their ticker is MRAM, and they were selling it, and others were selling it as well. So it was out there and people said, okay, we have a winner. We have something working. ReRAM was still not working and people set up MRAM manufacturing lines. Now they were very expensive to set up and everything, but we need a solution, right? So they were using it. But then you start having a situation where products go out to the market. And I guess you know M stands for magnetic. And we are surrounded by so many magnetic fields everywhere. Guess what happens? Some of those products ended up failing in the field. The memory was corrupted. So even as late as June of 2023, TSMC still had their road map, all MRAM. ReRAM was kind of a note on the bottom basically. But the big consumer companies realizing what's going on, they went and they made it very clear to TSMC and others, we can't work with MRAM. The risk is just way too high. Now meanwhile, over these past few years, on the ReRAM side, actually, quite a few of the players disappeared one after the other, but we got to the point where TSMC has its ReRAM now in mass production and selling. Weebit is qualified. There's UMC with a qualified ReRAM trying to make it work out to the market. So the bottom line was, June of last year, the TSMC road map was completely inverted. Everything is ReRAM, and MRAM is a comment on the bottom. So it's really the market now understands ReRAM is much simpler, cheaper to manufacture. It gives you all of the advantages that you want, and it's here and it's available. And so basically, this is the winner, okay? This is the technology that the market is going to focus on in the future. MRAM will still have its niche and whatever and people are already manufacturing. They invested in it. But this is really it. So I think that's the key point. And Weebit today, as you know, is the only independent supplier of ReRAM. And now in every industry, there's this great moment where suddenly everyone realizes, okay, this is the new technology, this is the new thing, and there's suddenly a big vacuum. There's this huge vacuum that is formed and it needs to be filled. Nature doesn't support vacuum. You have to fill it with something. And right now, we at Weebit are focused on just going in there and filling this vacuum as fast as we can, okay? And obviously, I'm always paranoid that somehow somewhere, I don't know how, someone will suddenly pop up and start trying to fill the vacuum as well. So we just don't have time and we're pushing as strong as we can, as fast as we can to just fill this vacuum. Now you know all of the advantages of our ReRAM. So again, I won't focus on this a lot, but everyone in the market knows the advantages of ReRAM. So the fact that we are much lower power consumption, much faster. And of course, we can manufacture below 28-nanometer, and we're lower cost to manufacture. So looking at Weebit's journey, and many of you actually have been with us since almost the beginning. So the company -- right now, we're celebrating our 10th anniversary. We started the company in 2015. At the time, it was that struggle of taking a new technology. And in Israel, the venture capitalists, they didn't want to look at semiconductor, they were too focused on cyber and AI and whatever. We ended up actually coming to ASX. In 2016, we already listed as part of the ASX trying to pull technology companies to Israel. But in those early days, we were just starting. We engaged with CEA-Leti. We needed a fab to do our R&D. And I joined the company 7.5 years ago, believe it or not, it's amazing just how time flies when you're having fun. We set up the French subsidiary. And then we moved to the R&D phase and there was really a lot of development and getting the memory bit working and then the memory array working and getting the chips and really building the company. And we got to the point where it was working. We actually went and we did qualification. We started pushing. We had SkyWater as the first company that we engaged with, and we qualified there. Unfortunately, as you all know, they decided not to really promote their foundry business. They're focused on R&D business. So we didn't get customers there, but we did get the qualification, the recognition, which led us to the deal with DB HiTek and we were moving nicely with DB HiTek. And now more recently, that already enabled us to make the next step forward with onsemi, which is already really one of the top companies in the industry. And it's really looking well. Now all of this is basically building up and the industry is seeing what's happening. The industry knows that it needs ReRAM. This is the future. And now with the recognition that we're getting, we're really scaling up. So there's a book that some of you already heard me quote from called "Crossing the Chasm", a must-read book for anyone who really wants to understand technology companies. And I won't go into all of the details. But basically, what this book talks about is how there's an initial idea. This guy, everyone thinks he's crazy, whatever, but starts building that idea, you go through an innovation phase. You take that idea, you play with it, you start forming something, you get to the point where you have an initial concept, and then you start implementing it. Now there are some early adopters, some geeks in different companies who like to play with it. So you have those visionaries and they tinker with it, they play with it. It's not a real -- it's not commercialization, but it's initial things. And then when you say, okay, I'm ready. I'm ready for the mass market, that's when you realize that there's this huge chasm and the customers are on the other side. And now you need to cross it. You guys got to see us struggling through the chasm, okay? Last year, everyone was all over me like, when are you going to announce something? When are you finally going to announce something? And we were just stuck in the chasm, and we were talking to people and all of these companies were telling us, we know ReRAM is the future. We know we're going to be using it. We know everything. We know how good it is. But you're a young company and this is a new technology and we haven't seen it in mass production and whatever. And just this natural human fear. It's just natural for humans to be afraid of something new that is "still somewhat unknown". And so everyone was saying, yes, we'll just wait. We'll do one more project with flash. We'll survive somehow until someone else takes the first step. And the book talks about the fact that what you really need is you need to find the companies that at a certain point, they realize the potential value that they can get from this technology is greater than the perceived risk. Now we were very lucky with onsemi, which is a great company. And onsemi announced at the end of last year, they announced their new Treo platform. And this is a big thing for them. They really are betting a lot on this Treo platform to really help the company grow. You don't come out with a new platform and all this good stuff and have old technology on it. And that's where they decided to, okay, we're going to bring Weebit in, we're going to bring the ReRAM, and we made that step forward with them. Now this is basically where we're finally past that chasm. And in the book, they talk about the bowling alley. You get to the point where you get that one customer. You have that SkyWater that helped you get DB HiTek and now you got onsemi. It steered up the market. I mean I have to tell you, everything now is different from what it was just half a year ago. Half a year ago, everyone wanted to talk to us, but they were sitting. Now there's movement. There's movement with all of these big accounts because, hey, onsemi already moved forward. We need to start reacting to this. This is the new standard. So now in the bowling alley, it's like one pin that helps you drop the 2 behind it, that helps you drop the 3 behind it and so on. And that's where we are now. We are now at that point where we're making this progress, and we have a few of these accounts that are already really moving forward and getting into negotiations and into deeper evaluations and it's moving now. We can sense it, and that's why I can say with confidence, we do plan to have several additional customers this year. You guys know that my compensation is directly tied to it. So I'll make sure that it happens. And at a certain point, when you have enough of these pins, the bowling pins dropping, people start looking to the right, looking to the left, and oh my God, my competitor here is using it, that guy is using it. And suddenly, everyone starts to say, we need to move forward before we'll be left behind. And that's where the tornado starts. And the tornado is really an amazing thing because suddenly, instead of you chasing customers, they're all over you. They start calling you up and everyone wants this, and it just starts going crazy. Now I've been lucky enough to be in the tornado twice in my life and it's just the most amazing feeling. And I can tell you, I can see the tornado building out there. I see it. I feel it. It's out there. I don't know if to tell you it's in 12 months or it's in 18 months or in 9 months or whatever. It's starting to build and I can feel it. And so Weebit right now is totally focused on just how do we prepare for this tornado, because if the tornado hits and you're not prepared, if you haven't fortified your house properly and whatever, you're gone. So everything that we're doing today, and I'll be talking about it in the next slide, is focusing on building that infrastructure, being ready to have multiple customers in parallel to support them and to make it work. And it's on the product side, it's on the sales side, and even on the governance side and building the company to be a bigger company. So that's really where we are today. You know our Board and our management team, and I'll talk about the new additions separately. So I won't stop here. But again, let's look at some of the big advancements that we had. So already old news is Naomi joining us, and we can definitely feel her impact already on the Board. She has so much experience, both on the strategy side and on the governance side and she's really helped us move forward. We got the first revenues last year. That's starting to build. There's research from MST and then from Unified. The company is already strong and solid. So one of the things, we don't need to have Yoav Nissan-Cohen as an Executive Director anymore. And again, getting the company ready to move to the next step on the governance side, people don't like to see executive directors on the Board. So he moved to a nonexecutive role. Of course, we did our raise. And the more recent news, we have Anne Templeman-Jones that joined us. She is already a Director that has been on the Boards of very big companies like Commonwealth Bank, Westpac, et cetera. She really has the governance experience that is required. I've been telling the team right now, just start thinking like an ASX 100 company. We need to already be operating like an ASX 100 company. We need to be prepared for that regulation-wise and governance-wise, et cetera, because the company will be moving forward now. And I don't know when and what and nothing is guaranteed, but we need to be prepared. And on the technical side, of course, you know that we demonstrated our ReRAM running on the 22 nanometers. We had a very good demo last year. We picked out on DB HiTek. We already have the wafers deck. There's a discussion going on now on when do we start the qualification. DB HiTek got feedback from customers and they were considering if we do some tweaks before we start qualification or not and what's happening. The wafers that we got are fully functional. They're looking good and everything. So we just need to decide if we're going to add that little step in the middle or not or start qualification. Onsemi, we've talked about it so much, but this is really an amazing thing. And the market reaction has been just overwhelming. It's really -- when you have such a big player, it's great. And of course, more recently, just in the last couple of weeks, we were at the Embedded World Show. For us, that's the biggest show of the year. We had two really big things happen there. So on the one hand, we had EMASS, which is an AI company. They were acquired by Nanoveu, which is traded here on the ASX. They have an image recognition product -- a gesture recognition, I'm sorry, that was using TSMC-MRAM, but like so many others, they really don't like the MRAM solution. So they decided to move to ReRAM. And they did a demo of their system running with our ReRAM instead of the MRAM. It was a very nice demo, very well received at the show and shows our capabilities when you're talking about AI, Weebit, and again, I have a slide to talk about it. Weebit is really a great enabler in the AI space. And of course, the big news that we had 2 weeks ago, the fact that we achieved the AEC-Q100 qualification. Again, we announced it on the first day of the show, and you should have seen the reaction. It was amazing. Basically, when you achieve AEC-Q100, that means that you can run for 10 years at 150 degrees Celsius, you can achieve 100,000 endurance cycles. I mean, you are robust. This is really at the point where once you have that, over 99% of the applications on this planet can use it. So it's really -- there's hardly any application I can think of that requires more than that. So this is really big. I mean it's not only big for automotive and suddenly the automotive companies are looking at it, but it's everyone. And just so you know, there are three qualified ReRAMs on this planet. UMC is still struggling to go above 105 degrees, okay? They haven't managed to qualify above 105. TSMC has not announced AEC-Q100 yet. I believe they probably can already achieve it. I don't know. But again, right now, we're the only ones that I'm aware of that have ReRAM qualified at this level. And this is a great message to the market. This is an amazing message. People are seeing now we're qualified. We have a great team. We're moving forward. We have now DB HiTek and onsemi and are really progressing. So there's a lot going on. I mean the level of energy now around Weebit is -- it's nothing compared to what we had half a year ago when we were just trying to push and push and push. Suddenly, there's this huge energy around us. So onsemi, again, we already talked about it a lot, an ASX -- I'm sorry, a NASDAQ 100 company, a major player in the automotive space, in the power management space. This is a very big, a very strong partner to have. Things are moving very nicely with them. Again, the Treo platform for them is a very important strategic move. They're putting everything they have into it and sometimes I think that they're trying to push us even more than we're pushing forward ourselves. It's just amazing the level of cooperation and everything is great. So really, really happy to have them and to push forward with this. The AEC-Q100 and automotive, you need to understand how automotive has become now a key market. It's a huge, huge market. The requirements are very extreme. I've been saying for quite some time now that there are already thousands of components in cars. We just saw the new numbers of 8,000 electronic components in some of the newer German cars and so on. It's amazing how much semiconductor goes into a new car today and all these sensors and all these things that are there, and AI is going into it and autonomous vehicles is a lot of computation. And even electric cars, by the way, they still have brakes and the brakes heat up and stuff like that. So they still need the 150 degrees and all of that. It's great. Now these guys, many of their chips either are analog and they want to have the smarts of a microcontroller with the analog, but then that doesn't work. I think I talked about it in the past. If you try to put flash next to an analog circuit, the flash is a front-end of line technology. It sits right next to the design and it impacts the design. You can't make an optimal analog design when you have flash next to it. So people have to do it in a 2-chip solution, and it's a big issue for them. With ReRAM, no problem. We're back end of line. We come on top, nobody cares. And we have a great solution there. So that's for some of the automotive products that are more on the analog side. Others just go below 28 nanometers, they need a solution. So we are a key element for this market and engaged with multiple automotive companies and again, great potential for us. I talked about EMASS and what they're doing with us in the gesture recognition. I presented this slide in the past, but I will go over it again because I think it's important. Today, AI, first of all, so that everyone understands, there are two key parts to AI. There's the learning part, where you teach it something; and then there's the inference part, where you leverage what you learned in order to do something in the field, okay, to implement what you learn. So the inference part is something that basically -- I mean, there are a lot of edge devices, what's called, the cameras all over the place or whatever other sensors and so on. And it's just not practical that they will send everything that they see to the cloud to be treated over there. So today, again, one of the things that now is very clear to the industry, you need to do inference at the edge device. Now that becomes a big challenge because you need to have the AI engine there. Edge devices operate very often on batteries. They need to operate at low power, et cetera. So in order to achieve the goals of what the AI engine needs, you need to go down to the smaller geometries. You need to go to the small geometries because that's where you can actually operate faster, lower power and more compact and so on. You go down below 28-nanometer and you don't have flash to put on the chip. So where will you put the coefficients? The weights or coefficients are different terms that are used for it. You have to have a separate chip. So basically, you have the chip with the AI engine. You need to have a separate flash chip, which will have the coefficients. And then when you turn the system on, you need to copy all these coefficients from the flash into a big SRAM that you will have on the chip, so that you can start running. Now let's start off with the fact that it's a 2-chip solution. It's more expensive than a 1-chip solution, obviously. A big issue here is the security. When you are actually copying from the flash to the chip, it's so easy to eavesdrop there and see what's going on. And then, of course, the AI chip, it's volatile. SRAM is volatile. So let's say, nothing is happening, the sensors aren't picking up anything. You can't really turn it off. So there's a power issue. Now just think of the solution, and this is what the market is really working on, is replacing SRAM with a ReRAM. Now you suddenly have a one-chip solution. You can have the coefficients in the ReRAM. You don't have a security issue. It's a nonvolatile system. By the way, the ReRAM bits are much smaller than SRAM. You can have a lot more coefficients on the same piece of silicon and you can have a higher level of accuracy. So this what's called near memory compute is what a lot of companies are looking at. That's what the EMASS demo was about. We have many commercial companies and research institutes that are now focused on this and we're working with them. And we actually started even a government project related to it. So there's a lot of activity there. Of course, the optimal solution moving forward is eventually to get to the neuromorphic and to have the ReRAM and the AI engine really combined together. That's further down the line. Right now, the big focus, and this is a huge opportunity in itself, is already the near memory compute. So this is the EMASS demo. I already talked about it. So it really shows the advantages of significant power savings, much larger bandwidth and so on. So as I said, and they're a great example of one of the many companies that are moving from MRAM to ReRAM. So where are we heading? So again, right now, with AEC-Q100, we can address any application that we want. So all of the applications that you see here are applications that we're talking to that want ReRAM, that need ReRAM, and there's an engagement with. You can see automotive, just how big of a business it is. It is a huge business. Power management, if you look at onsemi, the two reasons why they're working with Weebit is they are very big players on power management and on automotive. But even IoT and health care, Edge AI, as I mentioned before, which is growing rapidly, industrial, all over, there's need for this. By the way, I guess, being in Australia, I can tell you, we've had more than one mining company already come to us -- well, not the mining companies themselves, the companies that do the drills, and are telling us, the drills reach very high temperatures. We finally can see a memory that has a potential of addressing our needs. And now there's that talk about how do we enable them, because the drills -- you can imagine what goes on when they're spinning in there in the ground. So I mean, there's really all over the place. We have very interesting discussions. . We finally found someone who's analyzing the growth of the emerging memory revenues. And this is a market that is growing. This is a market that is growing. This is not just the ReRAM. There is MRAM in there. There is other stuff, and this is discrete and embedded. So I just want to make sure that I'm not misleading. But nevertheless, you can see just the direction that this thing is taking. Now that the technology is here, the move has started and more and more companies are looking at how do they get their hands on this stuff. . We talked already many times about the business model. So again, I will skip this. I will just mention that when we're dealing with foundries, we, on the one hand, give them a design license and they pay us a license fee and an NRE. Later on, as soon as we engage with the fabless companies, we'll be getting design license -- I'm sorry, the foundries are manufacturing license, the product companies are design license. Over the next few quarters -- you already could start seeing over the previous quarters a little bit of cash, a little bit of revenue coming in. As we move forward in the quarters, you'll start seeing how these numbers grow. We don't get right now all the license fee upfront. It's really tied to milestones. Companies are still very cautious with us. So the cash, the revenue, and everything will be coming in piecemeal as we progress, but the numbers will be growing and you'll start seeing them in every quarterly report as we move forward. Now of course, for us right now, the ideal is a company like onsemi, where we can actually accelerate a lot of things. When you have an IDM and they do both the manufacturing and the product, now you don't have the issue of the product company saying, "Oh, we'll wait until you're qualified." So I mean, with the foundries, if I go back to this slide, we're engaged with a lot of product companies, and we're pushing them to really commit. But obviously, they are concerned, they want to see more progress, et cetera. When you're dealing with an IDM, that issue doesn't exist. And so I can't go into what exactly is happening with onsemi. Obviously, that's confidential information of onsemi, what they're planning to do, which products they're planning to ramp up, and when and what, but it's obviously going to be faster than with the foundry. So what's coming next? Well, you know this slide and basically, so many of the names here are engaged with us now. Again -- but this time, the engagement is already moving forward faster than last year. Last year, we were engaged, we were pushing, we were talking, they were asking questions. We gave them data. But in reality, to a certain extent, it was kind of not really moving forward much. It was an engagement. They kept giving us positive signs, but we were not proceeding. Now there's real progress, much better progress happening. They are sensing now that they need to make that move. And in order to deal with that, again, we are now very focused on building the infrastructure to support many onsemis in parallel. That would mean we're going to be signing more agreements this year. There's going to be more than one technology transfer happening in parallel. We really need to ramp up everything. So one of the things that we've done is we hired Lilach Zinger. Lilach was the manager of Fab 1 at Tower before. If anyone knows what a fab manager feels, it's her. She actually grew from being a very junior engineer in the fab. She went through the ranks. She actually, with her hands, did all the work at the fab until she became the fab manager. So she is working now with the fab customers, and she understands exactly what they mean when they say something, and she knows what their concerns are, and she's the one representing them inside Weebit and telling us what we need to do. I think her title says it all. She's responsible for the customer success, and she's doing a great job. By the way, a key part of it is building this infrastructure. making sure that now when we have a new fab that we need to deal with, we don't start from zero. We have already a template. We just duplicate it. We modify whatever and we start running and we put as much automation as we can. When you have just one project working at a time, you can do it on an Excel sheet, you can do it in a much simpler way. When you start having many projects, you need to bring in management tools that follow up on everything that can generate automatic reports that you see, you have the big picture of things and so on. There's a lot of work that she's doing. We've brought in several people that are basically helping on that infrastructure level to set the stage, the tornado is coming, and we need to be prepared for it, and we're working very hard on that. And I guess you know we have what it takes. And I am very serious when I say I don't know of any other company in the world that has what Weebit has today in terms of the team, in terms of the device and the process and the analog engineers and the digital engineers and the tech and the characterization, and with all of this vast experience that we have in the company, with the management, with the Board and all, it is not easy to make a ReRAM. Again, even Intel and Samsung tried to do it. And believe me, they have all the resources, much more than Weebit has. They didn't have the management focus. And that's why they basically dropped those projects. We have this management focus, and we're working on it, and we're totally focused now on building that infrastructure and Weebit in a year, in 2 years is going to be nothing like what we are today. And we're working on getting the fabs and the product. So we have onsemi signed up as a fab. On the product side, actually, onsemi can qualify there, but we still have additional engagements to do, and the goal is to do more than these. So there's really a lot of work going on right now in parallel. DB HiTek, we need to now close this, how we start the qualification and do we add a little tweak or not or whatever, but that's running. So we have great targets for this year. By the end of this year, we should be at the point where we have enough of those pins in the bowling alley falling that the tornado will start kicking in. So that's really it. I think at this point, I'll stop sharing. And you know what, I'll keep it in case I need to refer to the slides, but I'll open the floor to questions from here, from, yes, Danny?

Before you proceed, as Coby said, we'll move to the Q&A session. For all those online if you want to ask a question, please type it in the Q&A box at the bottom of your screen. There are already a number of questions coming through, but I think it's probably best to start with the floor.

Talking about TSMC, in the past you said that ReRAM was a bit of a sideline for them as they were not serious about it, but they had it there and they weren't putting resources into it. It seems to me they've had a bit of the upside of the technology per se. Will that not indicate that they'll sharing a lot more resources into that to the extent that they possibly won't be looking at third parties because they will have all the boxes ticked for their own ReRAM.

So TSMC is a huge company, and they have a very good ReRAM, by the way. I'm not going to say anything bad about their ReRAM and it's out there, it's in mass production, it's moving. So I don't think they have a challenge with their ReRAM right now. However, having said that, this is a very secondary niche activity for them. TSMC is a foundry. We always need to remember, they sell wafers. They make money off of selling wafers. They don't make money off of ReRAM. This isn't a revenue generation generating activity for them or, I mean, it might generate a little bit or whatever, but that's not what they're building on, right? It's an enabler. So they are focused on this. Now just like I said in the past, and definitely nothing has changed, TSMC developed their own SRAM. TSMC developed whatever other IPs that they have. And yet when you look at their enablement library, they have other SRAMs from third parties in it and other IP. So I still believe that TSMC, it's a question of when they will become a customer of Weebit, as we grow, as they see that more and more companies are using our technology, as companies will come to them. And by the way, talking about -- if I go back to this slide, TSMC, as big as they are, their focus on the design side is minimal. They're not a design company. Yes, obviously, they have design engineers and they built a memory module and so on. But when companies come to them and say, we need to tailor that memory module, we want to have different versions of it. They basically say no. This isn't their core competence. This isn't what they want to focus on. This isn't mass revenue generating, let's call it that way. So these companies are coming to Weebit and they are telling us, we need a ReRAM that will be more tuned to what we need. And at the end of the day, if one of the big guys will come and say, I want a lot of wafers, they'll tell TSMC, we want to do that, we want it with Weebit ReRAM. I personally don't see an issue. It will have to be a large number of wafers, of course. That's going to be the key element. Yes?

As the refresh in leadership at Intel, Lip-Bu Tan, who was at Cadence Design, which I believe you had exposure to. Has that sort of reenergized any contact between yourselves and Intel? I know he has been very energetic and so [indiscernible].

Well, we have very good relationships. We had a great relationship with Pat Gelsinger. We have a great relationship with Lip-Bu Tan. Several people on the Board know him personally, as even I would say, Dadi and Atiq are personal friends of his. I know him, but I wouldn't dare say that he's a personal friend. He's an amazing person, by the way. I have unbelievable amount of respect for him. If anyone can save Intel, it's him. So he's really amazing. He has his hands full now with Intel. Hopefully, part of what he'll do will also -- I don't know what his plans are for the Intel Foundry Services. We'll need to see. Hopefully, either he will decide to really give it a push himself or he'll decide to spin it out and then there will be a focus there on it. So I can hope that things will move faster with Intel and so obviously, Intel, we've been in touch with, knowing all the people there, et cetera. But they have so many big issues right now. Let's see how he starts managing it and let's see what happens there.

I was going to ask you if you can [indiscernible].

Okay. So on the selector side, we are continuing to try to find a solution there. Obviously, it's a very low priority right now. I mean, I mentioned the fact that there's this huge vacuum that was formed in the embedded space that we need to fill as fast as we can. So right now, I ask the team to focus, let's just see how we fill this vacuum. Let's see how we get as many customers on board as fast as we can, ramp up revenues, ramp up the cash income, et cetera. So whatever is related to the discrete side is right now at a very low priority. We're still thinking and looking for this, and I mean even last week, I had a discussion related to it, but it's not at a priority level that it's really being pushed strongly. We're continuing to work in the background. With GF, I guess, I can't go into details about the discussion. They made it public that they were looking for some ReRAM of their own and trying to see if they can do all kinds of things. I won't go into -- I can't go more into what's happening right now between us and GF on that front. You had a third one? What was it?

You mentioned potential architecture.

Yes. Again, that's one of the negotiations that's ongoing, and I can't go into more details.

Negotiations at the moment, let's take that example. Is that more on the commercial sphere or that's still at a technical level?

That specific one is definitely at the commercial level. It's -- yes, with several -- we've passed already technical -- I mean, there are quite a few companies that already understand our technology. We're trying to push forward on the business side.

We've seen China recently absolutely disrupt the AI sector with DeepSeek needing to use fewer GPUs and not the top level ones. They are dominating the EV automotive sector. I saw before you had China written on one of the slides. Is that being a deliberative step that Weebit needs to be engaging more purposely with Chinese companies?

The situation with China is unclear, let's say it this way. There's this whole geopolitical war. Right now, it's not clear what Trump is going to do, where things are going to head. To a certain extent, anything related to China right now is a higher risk activity. So I'm very cautious about it. We are engaged with Chinese companies. We are talking and stuff. But I am very conscious to the fact that some of the companies that we might engage with might later on be on the black list suddenly. I mean, we are not engaging with anyone on the black list, okay? That's a statement that I want to make very clear. But there's a lot of issues here. So we have a lot of discussions strategically how to deal with China. So we're cautiously progressing there. Let's say it that way.

[indiscernible]

So the question was about we have here new companies that we plan to engage with and what will the impact be on the revenue and the cash and so on. So yes, first of all, the business model with either one of them, with the fabs and with the product companies, is to have a license agreement. That's the first step. As I mentioned, normally in the future, let's say, in a few years or whatever, when we sign a license agreement, we'll insist on getting the license fee upfront. Today, because of the situation, because of that concern and perceived risk and so on, the customers are insisting on paying the license fee in several payments based on milestones. And I understand them. And right now, we're still doing it that way. So you'll see payments happening over time. With each one of these deals, there's going to be payments happening over time coming in. And of course, the revenue is recognized based on your progress with the project, okay? So when we talk about the project of transferring the technology and manufacturing the prototypes and then doing the qualification and everything, the auditors are basically looking at it as one big project. And I know I'm a little bit simplifying things, but they look at what percent of the work was done until a certain point, and that's how much revenue they recognize, okay? So what's happening now is -- and you can see it also with SkyWater or with DB HiTek and you'll see it now with onsemi, revenue will be recognized every quarter -- or, I guess, in Australia, you recognize it half yearly. So every half year, we'll have a revenue report, which will include the part of the project that was done with that fab during that half year, and that's basically going to be the revenue recognized. The good news is that as we have more and more of these companies coming up, you're going to see more and more of these payments of license fee. And by the way, with the fabs, there's also a lot of NRE involved, because we're doing a lot of engineering work to transfer the technology to them. So it's the license fee and it's the NRE. Again, NRE is paid based on the work done. So we're not giving guidance yet because there's too much -- one deal that is pulled in 1 quarter earlier or 1 quarter late and all of your guidance goes completely off. So it doesn't make sense for me to give guidance yet, but we are definitely seeing every quarter we will be seeing growing cash and revenue as we move forward.

Sure. But I think if the company probably in the future had [indiscernible] sales target or something [indiscernible].

We will. It's a matter of time until we start having more of these customers and the impact. Right now, the impact of every customer is so huge that just having one deal slip to the next quarter or happen a little bit earlier and all my guidance is gone. So it's still not the time to give guidance. But as we grow, as we have more customers, the impact of each deal, and there are going to be many deals happening in parallel, at that point, I'll be able to give guidance with more accuracy. So we're getting there. The company is growing. That's part of scaling, right?

Yes. A quick question. The current endurance is quoted at 100,000 cycles. Is that minimum or is that...

No, that's what we managed to achieve. So to be clear, in nonvolatile memory, basically, the #1 limitation is how many times you can write into it before you wear it out. Now with flash, the numbers are normally between 1 and 10,000 cycles. So we are now already demonstrating that we can achieve 100,000. So it's basically up to 100,000. Obviously, we have a goal to continue to go to larger numbers. I mean, there's always R&D work going, and we're improving the technology all the time.

So this is based on how much writing you're seeing, and will that be sufficient, that 100,000 cycles, will that be sufficient for say the typical model vehicle that [indiscernible] will be using?

So you noticed that I gave the example, I was talking about the SRAM and the nonvolatile. So the SRAM is the scratch pad. That's where you keep writing and -- reading, writing and that's the scratch pad where you do all the calculations. The nonvolatile memory is basically to store, let's say, code. That doesn't really change very often. To store those AI coefficients, again, they don't really change. You basically read from a nonvolatile memory. You don't write very often. Now you do write sometimes. I mean, you have an update to the code. There was a bug that you fixed and you update or things like that, or on the AI side, they improve whatever AI calculation they have and they have a new set of coefficients to write in. So it doesn't happen often. And that's why with flash, even 10,000 was enough. Now we're showing 100,000, which is even much better. So...

Can you say something about the 2 government grants that you highlighted earlier?

Yes. One is a project in Israel and one is a European project. So basically, these are interesting -- the Israeli one is in partnership with several of the key players in Israel like Tower, like CEVA, like the Technion, et cetera. So it's a very interesting project there. The European one is still being defined. So we were already granted. I mean, we were told we have it. Now it's a matter of defining the details.

We have a lot of questions online.

Okay. We have a lot of questions online also. Yes, let's give those guys also an opportunity.

Yes. We've got about 20 questions in 25 minutes. So it's going to be a bit like sleep dating. So I'll try and pull them all together. I'll put them in the buckets like onsemi, DB HiTek, SkyWater, GF, and EMASS, because there's several questions on each of those. We'll start with the onsemi deal and more around clarification, Coby, around forecasted dates and time frames. So when will takeout commence? What is the forecast qualification date, to start off with?

So the work that we're doing with onsemi is such that -- also because they're a product company, they're more sensitive to talking about their plans and what they are doing and so on. And I think also on our side, right now, I'm feeling less comfortable to talk about it more also because of the NDA with them and in general. So you'll see it. I can tell you that it's moving really fast. I mean that's what I can say. We are working very closely with them, very intensive. On a daily basis, we have multiple calls with them and really moving forward well.

Okay. Further to onsemi. Because it's an IDM, will product design commence right away and will products be qualified at the same time as chip qualification?

So that's, again, one of the things that are under NDA. It's the confidential information of onsemi, which products and what they're doing. And until they don't announce anything, I don't want to comment on that.

Okay. We might move to a couple of questions on DB HiTek. As we're closer to the midyear now, is a more accurate qualification date available?

So as I mentioned, we have working silicon. I mean the silicon is working. It's in good shape. It's technically ready for qualification. The discussions that we're having now are around the inputs that they have from customers and the potential of a certain change that we might want to do to address some of these things. So it's under discussion between us and them, and we need to decide on what we do and when we actually start the qualification. So the goal is definitely to have the qualification this year. But depending on what we decide here, the qualification date actually might move around a bit.

And further to DB HiTek, more of a clarification, Coby. You've implied customer engagement in the past and imminent customer agreements. But what are the plans to qualify or is this not a priority at this stage?

When you're working with foundries, you always need to work with their customers in parallel. They need to have that confidence that once this is qualified, all of this work and investment will return itself in revenues, in customers coming in. So there's the parallel work happening. But again, the customers are always wanting to see that you're getting closer to the qualification before they actually engage and move forward. And so there's always that caution from all sides.

There's a couple of questions around SkyWater. I'll combine it into one question set. What does the SkyWater purchase of Infineon's fab mean for Weebit Nano? And is there more chance now of signing up a customer with SkyWater?

So unfortunately, the fab that they acquired has a very different technology to what we qualified on. So in that sense, I don't know yet. We're talking to SkyWater about what's going to happen there. But to a certain extent, that's a completely separate production line, and we'll need to do a technology transfer and everything from the beginning over there if we want to engage on that. So we're still trying to figure out what exactly SkyWater's plans are, and we've been burnt twice.

Okay. Just checking to see if there's any other questions on SkyWater. We'll then move on to GF or GlobalFoundries. And I mean, you said you can't say much at the moment. But it was noted it was on a 20-nanometer chip. Is this related to the GF 22-nanometer or another company?

You're talking about the EMASS demo?

Yes.

So yes, the EMASS demo actually used the 22-nanometer chip that we manufacture at GlobalFoundries. We are continuing to manufacture and test things. I just can't go into the details of the relationship with GF right now because of the situation with them.

And further to EMASS, can you maybe just give us a little bit of a background in terms of who approached who in that deal?

I think I can say they approached us.

Okay. Can you tell us the name of the foundry that manufactured the EMASS demo with ReRAM in it?

So actually, as I just said, EMASS, because we don't have right now a fab that's qualified at 22-nanometer, for the demo purposes, they took our chip that was manufactured at GF and basically had it as a 2-chip solution for the demo purposes. But the goal is to find the fab that will qualify, and we will be able to do it as a real 1-chip product.

That pretty much answers the next question. Is that embedded solution designed by the 2 parties or is this a 2-chip solution?

So it's a 2-chip solution that we work on together to make it work.

All right. Moving to some more generic questions. Will ReRAM replace annoying and expensive cloud storage for daily business computer applications?

So I think the answer depends on a lot of things. So let's start off with the fact and I talked a little bit about the differences between volatile and nonvolatile, et cetera. So the data centers need a lot of memory and a lot of memory that they read and write from all the time. They do a lot of computation. So for that, today, flash or ReRAM are not a good answer. MRAM, by the way, has much higher endurance, but still it is limited. So you need the SRAM and DRAM for the compute centers. There is, by the way, in the big compute centers, a lot of data that you don't access often, all the archives, for example, et cetera. And that's where nonvolatile memory comes in. Now of course, later on also -- right now, we don't have the discrete solution. I mean, they need terabytes, pentabytes, whatever, the huge numbers, which we can't even get anywhere close to at this point. But ReRAM will -- eventually, we will have it in a discrete chip. It will be growing. So it will have its place, but the cloud in general requires massive endurance abilities, which right now nonvolatile doesn't achieve.

The next question is around the addressable market, which I think is in the slide deck, Slides 20 and 21. So did you want to talk a little bit more about the addressable market?

Yes, the addressable market right now, as I said, once we achieved AEC-Q100, we can address any application basically on this planet that needs a system on a chip or that needs computing. So I mean, the semiconductor market is growing rapidly. In general, it's expected to cross USD 1 trillion by 2030. And all of that market, and I have the slide with -- let's see where was it, yes, this -- you can see here just the different market segments. This is part of the addressable market.

Okay. The next question is going back to discrete opportunities, Coby, you talked about through the deck. Is there a single selector work stream or various to tackle replacing flash? What state of success is each work stream in?

We've been trying several different types of selectors. As I said, it's a very low priority right now. We really are focused on the embedded.

Next question is around conversations with Chinese companies. Can you maybe shed some more light or progress on what's happening in that regard?

I think I already answered it.

I think you've already answered the next one. Is there still activity happening regarding the company that was interested in an architectural agreement?

Yes, I already answered that as well.

How do we think about TSMC's ReRAM as a competitor and not a customer? Do we have IP protection in place?

I don't think of them as a competitor. That's very simple. Yes, you're right. I mean, on the IP side, obviously, we want to make sure that we don't trip one over each other's IP. At this point, again, we don't know how they make their ReRAM or what they do. So each company is really working in parallel. I don't think they know what we do. But as I said, they're not a competitor on the IP side. Since we're both working completely in separation and in parallel, I don't think there's any issue there.

Okay. About the government-funded project that you mentioned, are you able to give any insights to the other?

I think I basically talked about it.

Yes. Will the R&D spend start to reduce now that Weebit has achieved extensive development already?

Actually, no. As the company moves forward, there's just more and more work to be done. So on the one hand, in this industry, you constantly need to research and move forward and improve your product and we will be continuing to try to achieve higher temperatures, higher endurance, higher -- a lot of other parameters, lower power, et cetera, et cetera. There's going to be -- we talked about the selector and stuff. There's a lot of work that's going to be done looking at that. There's on the AI front, there's a lot of work to be done there. So definitely a lot of R&D work. And by the way, when we're working with the different fabs, we keep being challenged by each one of them does things a little differently, and we need to somehow adapt to the specific machines that they use or other things. So it's not really R&D anymore. It's kind of between the R&D and the sales, but there's still a lot of work in that front.

We're at the end with the last 4 or 5 questions, 5 or 6 now, and we've got time, so we'll get through them. At the other end of the scale is BrainChip. Is there any synergy with their data requirements?

So we know BrainChip, and we talk to them. They have a different way of applying their AI type application. They have their own chip that they designed and so on. There is a discussion between the 2 companies on multiple fronts, but I can't go into more than that.

The next question back to SkyWater. Does the recent sale agreement between Infineon and SkyWater change the prospects for manufacturing...

I've already answered that one, yes.

How many of these top 6 companies, 6 of the magnificent 7, are we in serious discussions with?

We are engaged with -- I don't remember which ones are considered the magnificent 7, but we're engaged with really the majority of the big guys.

Okay. Going to LLMs. So the scale of AI LLMs is staggering. Does Weebit have a solution there?

Yes, again, LLMs are -- I would go back to the fact that LLMs, they need a lot of volatile memory. They need to do a lot of computations. So most of the work that they do is with volatile memory. Once some of that actually ends up getting to the inference AI, and that's where we come in.

We're down to the last 2 questions. So here's the last chance. We'll go back to the room at the end of this, but here's your last chance online, if you want to ask a question. We've still got about 5, 6 minutes. So online, the second last question is, how is FinFET development going?

We have several FinFET PDKs in the company. We've done work on it. I guess I don't think I can really comment beyond that at this point.

Okay. Another question has also come through. So we've got 2 more online and more about the financials and the markets. So they're actually a good way to finish online. How will our revenue compare with MRAM Everspin in the next 12 months.

Everspin actually is doing both embedded and discrete MRAM. So it's going to be difficult to compare the 2. But I think actually most of the revenue comes from discrete MRAM and not from embedded. So again, I don't want to give more predictions than that right now.

The last question online is around the recent cap raise. Have you seen any pressure negativity from institutional shareholders who invested in the raise given the retracing share price?

There's obviously some concern and people call me up, et cetera, but I don't see any major issue or something like that. I think the people that I talk to realize the potential and maybe this is a good way to wrap up things. But when you look at it and you realize there's this new standard right now, the market basically knows it needs this technology. Weebit is right now the only independent supplier of this technology. Onsemi signed up, we have AEC-Q100. People can see that the potential here is huge. I mean I'm not allowed to talk about what I think of the share price beyond the fact that I can tell you that I personally bought shares half a year ago. But I think people when they look at it and they really study it, they can see the potential.

Well, that concludes the Q&A online unless there's any final questions in the room. Back over to you Coby.

What's the biggest size of memory [indiscernible]?

The biggest one that we actually have right now in silicon is 8 megabit.

The recent talks you had about AI application?

8 megabit is actually already a big memory for these kinds of things.

For the AI application, in this field, but I was wondering maybe not enough, I think.

Well, first of all, what we have today is 8 megabit. Obviously, we're working on larger ones. And when we need, we can achieve larger memories. So it's a matter of priorities on what we want to focus on and what the demand is. But the vast majority of the applications that we talk to don't even need 1 megabit or maybe 1 megabit is enough. So there isn't really pressure to go right now to the bigger capacities. . Okay. Well, thank you, everyone, for coming, and thank you to the people online. I appreciate everyone's time. We are looking at a really exciting year ahead of us and looking forward to coming back in a few months with updates. I believe it will be interesting. So yes, thanks.