Stocks Down Under Videos
Get a 3-month FREE TRIAL to CONCIERGE now!
Concierge gives you timely BUY and SELL alerts on ASX-listed stocks
Weebit Nano at the Stocks Down Under Semiconductor Conference 15 November 2022
November 18, 2022
WBT, Weebit Nano
Weebit Nano (ASX:WBT) CEO Coby Hanoch presented at our 4th Annual Semiconductor Conference on 15 November 2022.
He spoke about the company’s rapid progress this year towards commercialisation, which is expected to commence in 2023.
He also talked about Weebit Nano’s discussions with tier-1 fabs, that accelerated following the announcement by TSMC that it is now offering ReRAM commercially.
See full transcription below.
Disclosure: Stocks Down Under/Pitt STreet Research directors own shares in Weebit Nano.
No time to do stock research, but you still want to invest?
Stocks Down Under Concierge gives you timely BUY and SELL alerts on ASX-listed stocks!
GET A 3-MONTH FREE TRIAL TO CONCIERGE TODAY
No credit card needed and the trial expires automatically.
Coby: Hi, everyone. Yeah, it’s great to be back in Australia after so long. Thanks for coming. So, I’ll be talking about Weebit. We are a company in the semiconductor space, more specifically in the memories for semiconductors, and even more specifically the non-volatile memory. So, we’re talking about the memory that doesn’t lose its content even when you unplug it from power. You know, let’s say the USB stick, for example, that can keep pictures and movies and so on. So, as Marc mentioned, we’re an Israeli company. Started in 2015, I’ve been developing this technology. It’s a very complex technology. We have a partner in France, where we’re developing the technology. In order to manufacture semiconductors, you need what’s called the fab, which is a manufacturing facility. These fabs are extremely expensive to set up. They’re billions of dollars to set up. So, we partnered with Leti, which has an R&D facility, a state-of-the-art fab. And that’s where we’ve been developing the technology. And the good news is that now, finally, we have the technology working. We have it proven, and we’re in the process of transferring it to the commercial market, and expecting to start sales in the near future. So, very big market potential.
Before I dive into the technology, I think a key part of Weebit and what’s making it successful is the team. We have a very strong board of directors, with Dadi Perlmutter as our chairman. He was number two at Intel. His team invented the Pentium processor. He was managing most of Intel’s advanced products for many years. Yoav Nissan-Cohen, who did his Ph.D. under the guy who invented non-volatile memories. He’s been in this industry from day one. He’s also set up one of the leading fabs in the world. Atiq Raza, another very well-known figure in the industry. And, of course, we have two Australian directors, Fred and Ashley. And the management team is also a very experienced team. Each one of the guys here has been doing memory development, design for 25, even 30-plus years. So, I’m very proud of this team, because without them, you couldn’t really achieve what we’re doing.
I know that many of you are not familiar with the semiconductor market, so just to give you a little feel, because today, practically anything you touch, anywhere you look, it’s semiconductors. I used to have a slide that showed 10 years ago, only 1 out of the top 10 market cap companies in the world was semiconductor-related. Today, 9 out of the top 10 market cap companies in the world are semiconductor companies, or very heavily in that business. The investments that are going into semiconductors now are unbelievable. You know, it’s just huge amounts of money, both from company, I’m sorry, countries. Everyone’s heard of the U.S. CHIPS Act, which is about $50 billion. But most of the leading companies today are investing heavily in this. In addition, you see the investment by the leading companies. I think when you total everything up, it’s about a trillion dollars. Not used to saying that number, but a trillion dollars going into semiconductors in the past year.
I think everyone here has seen in one way or another the lack of components that… You know, you wanna buy a car and you have to wait for almost a year. You wanna buy a refrigerator, etc. So, this is a crazy market. It’s a very, very… The demand is huge now [inaudible 00:04:30] are the largest segment in the semiconductor space. So, you know, that’s really, the pressure right now in the memory market is huge, and specifically when you’re talking about non-volatile memories.
So, the existing technology, you know, when I was showing the market is going to be $100 billion soon, that market today is dominated by a technology called flash. Flash has been around for several decades, but like anything in the technology sector, it’s starting to hit the walls in many different ways. So, the market is really looking for a more advanced technology. I’ll be describing our technology in a minute. So, it’s really growing very fast.
You can see, I mean, even in the Weebit logos, you know, we’ve kind of gone through the evolution of ReRAM and, you know, from the days that it started. Weebit started in 2015. And, you know, at the time, it was kind of really far out. We’ve progressed pretty fast in this market. Actually, most people don’t believe that in just seven years we’ve managed to get this technology to work. But we’ve gone through, and our logo said emerging memory and future memory. Today, now is the time. ReRAM is here. Customers, foundries, everyone is looking for ReRAM.
And I’ll change the order of the slides. What is it about ReRAM that people are looking for? So, our technology is significantly faster than flash, 100 times order of magnitude faster than flash. The technology is much, much better. Lower power consumption. Actually, we consume about 1% of the power relative to flash. So, really, a huge impact. Just think of the impact on any battery-operated device, your cell phones or whatever, when you suddenly consume so much less power, the impact on battery life. And, you know, we’re immune to radiation, and there’s a lot of radiation in different places, in aerospace applications, in military applications, in automotive. So, there are a lot of advantages here, but I think the two that are really the most important ones are the fact that we can go below 40 nanometers. Now, what does that mean? In the semiconductor space, for many, many years now, every two, three years, the industry manages to shrink everything so that you can have much more technology on the same piece of silicon. And, you know, it’s really basically doubling every two years, roughly.
We’ve been going down to…40 nanometers is already considered an old technology. The industry has gone down to 28 nanometers, 22 nanometers, 16 nanometers, 10 nanometers, 7 nanometers. The advanced designs today are already done at 5 nanometers and 3 nanometers. Flash is stuck. They cannot go below 40 nanometers. They have physical issues there. And so anyone who wants to do a more advanced design basically can’t have an embedded memory in their design. They can’t have a memory as an integral part of their design. So, what they do is they have their chip with their whole design, and then they have a separate chip for the flash, which they kind of glue on top and connect it somehow. Very, very non-efficient way of working. Our technology can shrink down to the smallest geometries, so that we can actually be embedded in the design, and be much more efficient for them.
But more than any of these advantages, the one that really counts is our technology is just much simpler and easier to implement, much cheaper to manufacture at the end of the day. And that’s what really counts. We add about 5% to 7% to the cost of a wafer, of a silicon wafer. Flash adds about 10% to 20% to the cost of a wafer. Other technologies go up to even 40%. So, I think this is one of the key elements that, of course, are impacting and making both foundries and customers interested in our technology.
And the technology, I already mentioned, any device requires semiconductors, requires memory. And the different advantages that we have are relevant to many of the major market segments. If you look at the mixed signal power management today, power management is a really big and very fast-growing segment. Again, you know, all of the battery-operated, not just battery-operated devices, IoT, what’s called internet of things, all of the wearables, the different gadgets that you have, etc., AI, automotive, aerospace, they all benefit from our technology.
So, in the past year, Weebit made huge progress, actually really fast progress. A year ago, we announced that we have gone down to already 28 nanometers, and we did a significant raise. After that, we announced that we’re starting to work on 22 nanometers, which we’re expecting to tape out before the end of the year. Tape out means sending the design out to manufacture in a fab. So, that’s already moving very quickly forward. And I guess, more important recent news, we announced that our ReRAM is qualified. I’ll go into the details of this, and also that we have our first mass production wafers in our hands.
So, really, that announcement, I think was a very important one. We just very recently received, for the first time, wafers that we manufactured in a mass-production facility. Up to now, we were working on R&D. A year ago, we announced that we have an agreement with a mass production fab called SkyWater in the U.S. We taped out, we developed a design that mimics what customers would normally use, and we sent it to manufacturing at SkyWater, and we just received the wafers. Actually, I have a wafer here. If after the meeting someone wants to see, I can show you guys the wafer. And we’re moving forward to qualify and to talk to customers.
Now, I think here is a point where maybe I need to stop and explain a little bit about how this whole industry works. Weebit is supplying a technology. At this point, we are focused on what’s called the embedded space. So, we’re basically talking about chips that are systems on a chip. Since everything shrinks so much, you can put a full system on a single chip, and a full system means a processor, and communication, and many other elements. And obviously, you need a memory there. So, you embed our technology into that chip. And that means that we don’t actually manufacture the end product. We license that technology to a product company that embeds it into their product, and then they need to go manufacture.
And since fabs are so expensive, even Apple, Facebook, Google, they don’t have their own fabs. It’s just way too expensive for them. So they go to general-purpose fabs and manufacture there. So, there’s this triangle here that needs to work together. There’s Weebit that’s supplying the technology, the product company that embeds it into their product, and then the fab that manufactures it. And we work with both sides to make that work. So, SkyWater is the fab side. And we signed that agreement, and we now have the first wafers from them. And we are now working with them to go with the customers, to work with the customers. So, they already have Weebit on their website. Their sales team is already trained on selling Weebit ReRAM, and we’re working together to bring the customers on board, so, we’re expecting to have customers signed up in the near future to get the first one up, and then the second, and start moving towards commercialization.
One of the key elements between us and that point is qualification. And actually, qualification is really a very important step. We announced also recently that we finished qualification on the wafers that we got from Leti. Now, what does qualification mean? Qualification is really that stamp of approval by the industry. There’s industry bodies that defined what technical parameters, how you test your device, so that you can claim that it’s robust, it’s ready for mass production. You know, if you think of IKEA and the chairs that they sell, and you sometimes see those robots that sit a million times on the chair and make sure that the chair manages to resist that, or in cars, you know, the robots that open and close the doors a million times and make sure that they function well. That’s basically what we do in qualification. We go through several production cycles.
From each one, we take a lot of chips, and we’re talking about hundreds of chips. And each one of those chips undergoes very intensive testing. So, we put them in ovens for, you know, long periods, for 1,000, 10,000 hours, and we see that they still function after that, that they withstand heat, that they withstand cold, that they withstand all kinds of extreme situations. And then we can tell the market, these chips, all of them, pass the testing. They’re robust, they’re ready for mass production. They all work the same. You know, one of the things that’s very important in mass production is uniformity, repeatability, and so on.
So, that’s what we did. We did that with Leti. We had very good results. And now we’re starting to do that with SkyWater. One of the results of that, one of the things that happened, and this is really the key thing that’s going on with Weebit right now, we’re getting ready for mass production. At the same time, TSMC, which is the biggest fab in the world, has announced that they’re starting to provide ReRAM. Which was great news for us. Actually, it was amazing news for us, because suddenly, there was an official approval that ReRAM is ready. ReRAM technology is no longer the future technology. ReRAM is here. ReRAM is available. And all the other foundries in the world were there, “Okay, we need to compete with that. We need to find a good technology, a good ReRAM technology.” And when they start looking around…you know, today… There used to be several other companies that were trying to develop ReRAM. For all kinds of reasons, they’re basically not here anymore, you know, each one of them, with the challenges that they met.
Weebit today is really, you know, practically the only other solution that the foundries have. So, suddenly, we have quite a few of the tier-one foundries in the world that are calling on us, and wanting to work with us, and they’re in different stages of evaluation. So, this qualification really triggered a lot of activity for Weebit, a huge potential. And again, going back to that slide from Yoav that showed that growth to a billion-dollar revenue, that’s really what’s happening now. You know, we’re talking to these guys. It’s a lot of work. Again, you know, you can see from the moment that you decide to work with a foundry until you actually transfer the technology to them, and then you tape out, and until it’s manufactured, you know, this takes time. But once you engage with these guys, you know, things start rolling, they bring in their customers.
By the way, today, the foundries are telling us, they’re telling us that the customers are coming to them and asking for ReRAM. They lost major deals because they didn’t have ReRAM. You know, that was a key statement to hear from these big foundries. And they really needed… We’re talking to the customers, and the customers are telling us, “We want ReRAM.” So, it’s really a very exciting period. You know, we just need to really get through this, get the agreements with some of these tier-one fabs, get the technology transferred. But that’s going to be done over the next year, year and a half, we’ll be getting more and more of these done, and then we’ll be able to go one after the other, qualify them, and eventually get into mass production.
There’s a completely parallel path that’s happening at Weebit. You know, I’ve been talking about embedded, and having the memory embedded in a system-on-a-chip. In parallel, many people want to have chips that are pure memory, all memory. And so, we’re working on developing that. The requirements in that market are much more extreme. One of the key elements that are required there is called a selector. It actually enables you to define which bit you actually want to read and write from. We’re making good progress on that front, but that’s a technology that’s gonna take, you know, that’s more medium-term. So, the big focus right now is, as I was saying, on the embedded market. But we haven’t forgotten this market.
And again, talking about the embedded market, I mentioned it. We’re constantly scaling to the smaller geometries. These are already geometries where flash doesn’t exist. So, this is a complete void that we’re going into. We’re going to be sending to manufacture chips at 22 nanometers. Again, to give you a feel, there are less than 10 fabs in the whole world that can manufacture at 22 nanometers. You know, it’s really a very, very advanced technology, a very challenging technology, you know, getting to this level. And now, by the way, some of the tier-one fabs are pushing us, and we’re already starting to work with them, even at lower geometries, in the teens. So, really advanced work, very challenging work. But 28 nanometers and 22 nanometers are very popular geometries today that people are using. And getting the technology working at these geometries is a huge milestone. So, you know, we’re gonna be sending that to manufacture. You know, when the chips will come back, we’ll start doing qualification on them as well. And again, moving forward, even closer to that commercialization.
So, that’s really, you know, the state we’re at. You know, I’ve already talked about the tier-one fabs and all of the work that we’re doing with them. So, in the next, I guess, towards the end of the fiscal year, in the middle of ’23, we want to finalize already the qualification with SkyWater, and start signing up customers for their fab. We’re talking to tier-one fabs, and I really wanna see an agreement already signed with one of the big guys, you know, bringing customers on board, and qualifying the technology for more and more extreme levels. So, you know, we’re continuing to do qualification. We qualified it at what’s called the industrial level. Now we’re working on automotive grade, and moving forward. So, there’s constant work here, really exciting times for Weebit. And by next year when I’ll be here, I hope to be talking about revenues. I think that’s basically it.
Marc: Thank you, Coby. If I understand correctly, the iPhone 14 is being shipped with embedded ReRAM [inaudible 00:22:49] right?
Coby: Yes, actually, that’s what we found out that… Again, one of those things that has been triggering the market to move forward is there is a chip with embedded ReRAM in the iPhone 14. So, again, now everyone knows it’s here, the technology is already working. That’s one of the things that’s been fueling the demand.
Marc: Right. So, Samsung is scrambling now, then, to get some [inaudible 00:23:13] ReRAM [inaudible 00:23:13]. All right. Well, before we go to the questions in the room, we had a question come in by email yesterday, and it sort touches on what you talked about a little bit, is…question’s from Stuart, not our Stuart, different Stuart. “I would like to better understand broadly how the income structure works. So, for instance, if a fab makes a wafer for us, what’s the percentage that we are likely to get in royalties? And also, what is the typical value of the wafers?”
Coby: So, first of all, the cost of wafers varies, you know, hundreds of percent, probably thousands of percent between small geometries and large geometries. So, it’s very hard to say what is the cost of a wafer. It’s really all over the board. What I can say, first of all, the business model, and that’s important for everyone to understand. We’re not selling a product. We’re selling a technology. We’re licensing a technology. So, when we engage, and by the way, I talked about this triangle. So, we license the technology to the foundry so that they can manufacture it, and we license the technology to the product company so that they can embed it in their product that they’re going to sell. Each one of them, when we do this licensing, they pay us a license fee up front, in order to start working. In addition, the customers, depending on the customer, but many of the customers will want to optimize the memory to their specific system-on-a-chip. So, they will pay us for engineering work, to change the size of the memory, change different parameters, etc., to make it fit into their SOC and operate in the optimal way. So, that’s engineering payment. The big money comes in once they go into mass production, because then we get royalties from their sales.
Now, we can either get the royalties directly from the customer, and basically we get a certain percentage from the sales of these chips, or, sometimes the foundry actually offers the technology to their customers. And when a customer, you know… Because, what is a foundry? A foundry is like a supermarket. You know, they manufacture chips, but they have all kinds of different technologies there, and their customers can take those technologies from the foundry. So, if the customer actually takes the technology from the foundry, the foundry uplifts the cost of the wafer, and then we get a percentage from the wafer revenues that the foundries have. Again, in terms of numbers, there are so many parameters, and it’s so across the board. Even, you know, just the size of the memory, the geometry that we work in, the number of instances of memory that they have in a single chip, because many times they have more than just one… It’s very hard to talk about specific numbers.
Marc: All right. Any questions in the room here?
Man: You talk about embedded chips, right? I’m a bit confused. Can you explain whether the embedding takes place at a fabrication level or assembly level?
Coby: Actually, much earlier than that. So, it takes place at the design level. So, the customer that is developing the product, when they are designing the product, they already embed our memory into the product, so that later on, when they send it to manufacturing, it is already inside their design, and the fab manufactures it with our memory inside.
Man: So, does this imply that if your customers need to practically redesign their chip or to accommodate…
Coby: So, actually, that’s where we will be providing several standard formats, if you wanna call it that way. And the customer can choose one of those, and then, you know, do their design around it. And by the way, customers today take a lot of IP. It’s not just going to be our memory. They normally take a microprocessor from another company, and some communication IP from someone else. Each one provides their blocks, and then the customer kind of puts their design around them, and connects everything, and makes the full system… Because it’s a full system on a chip. So you don’t do the whole system yourself. You take some blocks from some other people.
What I mentioned earlier is that there are cases, especially more advanced designs, where they will come to us and say, “We don’t like your standard one. You know, we want it to be in a different shape. We want it to be maybe bigger, you know, more capacity, not 8 megabit, but 16 megabit. We’ll want it to interface to a different kind of bus.” So, you know, they can pay us engineering, it’s called non-recurrent engineering, or NRE cost, that they can pay us, and we will do those modifications for them so that it’s optimal for their system-on-a-chip.
Woman: How will your chips be incorporated into the quantum computing technology that’s coming?
Coby: We are not looking at quantum computing at this point. Quantum computing has very specific needs. You know, it’s an evolving domain. You know, right now we’re focused on going into the mass market where we can make significant revenues, and not focusing on a future technology.
Marc: If there’s no question in the room, Stuart? Online question.
Stuart: Pretty interesting question from online. “When you get royalties, how much do you have to pay Leti and the original inventor of your technology, Professor James Tour?”
Coby: So, we have agreements with both of them on, you know, sharing that revenue. It’s a good question. I can’t go into the details of exactly the amount there. And again, there are some parameters to it. For Rice University, is really, you know, today it’s practically nothing. And to Leti, there will be a certain amount. It is capped, by the way. In both agreements, there were caps. So, at a certain point, you know, we reach a certain point and then all the revenue becomes ours.
Stuart: Next question. You talked about continued R&D. What are you able to tell us about what your geniuses are working on, in terms of the next steps?
Coby: Yeah. So, first of all, in this industry, you have to keep running forward or you fall behind, so there’s constant development. We’re developing, first of all, we’re going down to the smaller geometries. I already mentioned 22-nanometer. And some of the tier-one fabs are already pushing us into the teens. That’s a lot of work to get down to those levels. And that’s one major activity. You know, the basic technology, we’re constantly trying to improve it, so you constantly need to get to even lower power, even faster read times, even higher temperatures. I talked about moving from industrial level to automotive level. You know, there’s a lot of work to get the technology to work even at such high temperatures. And then, in parallel to all of that, I talked about the discreet memory, or the standalone memory, that we’re developing the selector technology. So, actually, we have so much work to do. It’s really crazy.
Stuart: Now, there’s a competitor technology to yours out there called MRAM. Tell us why you’re better or worse than MRAM.
Coby: Actually, that’s true. And I forgot to mention it. MRAM, the M stands for magnetic. MRAM has been out in the market earlier than ReRAM. So, when we talk about emerging technologies, it’s basically ReRAM and MRAM. And in the old report, I don’t know if you noticed, it basically said the whole embedded emerging market will be $3 billion, and they expect ReRAM to be one-third of that. I actually believe that ReRAM will take much more than that. But MRAM, the reason why I believe that is MRAM, first of all, is a much, much more complex technology. I talked about our technology being 5% to 7% added cost of the wafer, with really very minimal work to manufacture it. MRAM requires many, many, many layers, very thin layers. It’s very hard to manufacture it. The added cost to the wafer is 30%, 40%. So it’s much more expensive, it’s much more delicate.
Our technology can be manufactured in the same standard manufacture line that, you know, the chips are manufactured today. Same machines, same tools, same materials. So, everything is standard. MRAM is the most non-standard thing you can think of. You have to manufacture it in a separate facility. You know, it requires a lot of protection because of the magnetic material. And then the last thing is, and that’s what we’re hearing now from more and more fabs and customers, you know, there are a lot of issues with the MRAM because it’s magnetic. So, I mean, just put a magnet next to it and you erase the memory. Or, you know, all kinds of things can cause security issues there, and other issues. And that’s why we’re hearing now from the tier-one fabs, they’re telling us “We really want ReRAM because we know it’s robust, it can do the job, and it’s much cheaper.”