Good afternoon, my name is Blair Abernethy. I'm one of the software analysts here at Rosenblatt. Thanks for joining us. With us today is a newly minted public company, Silvaco. CEO is Babak Taheri, and as well we have the CFO, Ryan Benton, with us. Welcome, gentlemen.
Thank you very much.
Since you've only been public since May this year, there's probably a few people in the audience that aren't that familiar with your story. So I'm wondering, just to set some context here for our discuss-
Time on this call. Just a bit of background of myself. I've been in Silicon Valley over 35 years and held key role positions with companies ranging from small cap firms like InvenSense, to mid-size firms like Cypress Semiconductor, and Freescale, and large cap companies like Apple. I joined the company on October of 2018 as CTO, and became the CEO in 2019. I've been leading our strategic direction since then. I made sure all those directions are aligned with our vision, mission, and values, some of which include customer success, teamwork, and leading by examples, and strive for excellence. Silvaco enables semiconductor design and manufacturing, and the two key words here are design and manufacturing. For the past 40 years, all EDA companies have been focusing on the design aspect.
We are going to a new, white space as a software company, EDA company, that takes that space into manufacturing, and I'll get to that detail a bit more. But we do this through AI-driven digital twin modeling. Digital twin models are basically software models that enable simulation rather than prototyping to save time and cost. We also do software emulation for automation of a lot of these processes. Our digital twin models for simulation are provided using a combination of AI and our software platforms. So now a bit about our software platforms. They're considered to be the foundational technology behind the chip. As I said, there are two steps, the first two steps of a chip to be made and dispersed is design and manufacturing, so we are the technology behind those both.
And the markets that we focus on, the top three markets are display technologies, power semiconductors, as well as memory products. We also participate in automotive, high-performance computing, as well as products that pertain to IoT and 5G, 6G. That's a very quick overview. We do compete with larger competitors of ours. We were the last company, I would say, that went IPO in the EDA space was over 20 years ago, so it's a coveted space, and getting to where we are at, we are very thankful for that, and the team has worked hard to get us to this stage. Ryan?
Yeah, so I'll just briefly introduce myself, and again, I thank everyone for joining. Ryan Benton, the Chief Financial Officer. I like to joke a little bit and say that I'm a 33 year finance executive, which does make me the junior member of a very seasoned, experienced management team, so thrilled to be part of the team. Started my career as a CPA in public accounting, a few years there, and then I've spent over 30 years working at tech companies. Most of that experience has been in the public markets, so I've been a public company CFO multiple times, and most of the experience is in and around semiconductors, though split fairly evenly between chips and capital equipment.
In addition to my operating experience, I do think I bring a lot of transaction experience to bear, so over $2 billion worth of transaction tombstones to my name. So again, thanks for joining us.
Okay, great. Thanks, gentlemen. Maybe, Babak, we'll start on your side. You mentioned it a little bit in your comments, but maybe we can dig into it a bit, the Fab Technology Co-Optimization product. This is a fairly new area that you've released this year, although it's been something you've been working on for a few years. Maybe take people through a little bit about what that is and sort of how, you know, how big is that opportunity for Silvaco over the next few years?
Yeah, that's a great. So, given the fact that this is, as I mentioned, in getting a chip out to the customers, our customers tend to be the semiconductor companies that design and manufacture. The first step is designing it, and then followed by after the design is completed, the designers send this file to manufacturers to make the wafers. So think of Fab Technology Co-Optimization as a digital twin solution, right? I mention this because of the fact that I called our digital twins basically similar to what you would call a crystal ball. If you have a crystal ball that you can look at it and turn it around in order to see different aspects of future, this is what a digital twin model is.
It's a computer model that, now fabs, can actually look at this model. It's a very accurate model representation of a wafer, if you will, that they can analyze and see where they need to go in order to optimize their manufacturing process to get better yield, spend less money, on making extra wafers to fine-tune their fab, as well as getting, products to market sooner. So if you think of these digital twins, not only they serve the fab, but, and the way we create them is through AI and through our platform is by using customer data. And this customer data we never see.
The only components of our system that sees this data is our AI models that they get trained based on this massive amount of data in order to optimize and find out what's the corners in which the boundary conditions in which this fab line would be able to produce a wafer that deals in a good percentage, and that model itself is so accurate that rather than testing every aspect of equipment piece and tweaking it to optimize the fab, this can guide that process, hence you don't need to make wafers to test the capability of the fab. You don't need to make masks.
So ROI for this sort of thing is very easily justifiable, especially as you get down into the technology nodes that require you to look at a wafer that requires thousands of steps to process. So the wafers could cost in tens of thousands of dollars per wafer, and the masks could be millions of dollars. So imagine not having to run masks and test wafers of thousands annually in order to be able to do this through a digital twin by tweaking the fab process.
The other steps and aspects of this digital twin model that we haven't applied yet, but plan to do, is now that we have an accurate digital model that shows the boundary of what we call manufacturable design, this model can be provided now back to the designers and tell them, "These are the boundaries in which you can design the next generation of chips, that you can actually tweak these parameters within these boundaries that enable us to manufacture them, on the get-go." So, you know, the proof always is in the example. We have been working on this technology for over three and a half years with one of our customers in memory market.
That customer, we announced it to be Micron, and, Micron actually has been using this, providing us with feedback and commentary as an alpha customer. And we announced last quarter that, they are putting this in manufacturing, and as a result of that, we are able to actually go to the next level of our software with them and have signed up a five-year contract, as we announced already. And Micron actually initially in our IPO also invested $5 million in our company. But the point I'm trying to make with this is, we have a very solid, very high-level company that is that are the top of their game in terms of what they do with memory products.
And what we plan to do is now we are able to take this technology that is not memory-specific, but take it to foundries and fabs that are planning to come up with manufacturable technologies that are questionable at times. One example of it being silicon carbide technology or power technologies, technology for that matter. And as you know, silicon carbide has had this challenge of going from four-inch to six-inch wafer to eight-inch wafers to. And now they're at eight-inch wafer, and those are the challenges they're facing, and this technology can really help those companies to optimize. So we are going to go after the companies that are challenged by their technology in power.
Plus, in addition to that, as you know, as you get down to the curve in terms of technology, geometries, advanced CMOS technology nodes, FinFETs, and sub-7 nm technology nodes, those processes are very, very difficult to manufacture, and this would be a natural resource and way for helping them as well. So, initially, we have launched this in a memory product for Micron in their Boise, Idaho fab. But our plan is, and we've been discussing the possibility of launching this in both power market as well as what I call advanced CMOS technology now.
Okay, great. What is Micron sort of seeing in terms of returns on this? Is it dramatically speeding up their time from design to manufacturer, or just kind of... You know, I'm just trying to better understand the value proposition of what you've put together here.
... Yeah, so the value proposition has been initially not to run extra wafers to tune their fab, so saving wafers that used to be just utilized to tune the manufacturing but not go- I call that prototyping, not going to production for customers, and not running masks that would go making those wafers have been already justifiable to them in order for us to work with them, and in order for them to extend our relationship for five years.
Okay, okay, and then how do you- how big do you see the market opportunity for Silvaco for these FTCO digital twins?
That's a great question. So when we look at the fab technology and the way we can participate in that technology, our estimate, based on the three markets that I mentioned, that we are, memory market, power market, and advanced CMOS, I didn't mention anything that we will take it to display market, and that's another opportunity for us. But based on those markets, we've estimated that our SAM for this technology is about $500 million. Basically, this SAM expands the TAM for the whole industry. It's a white space, it's a green space and, you know, as we participate in our industry, we always strive in trying to expand the TAM for us and the rest of the industry, and that's how we all succeed and compete nicely, so...
Okay, excellent, excellent. So, Babak, just help us to better understand, so the core and part of the business and the main driver of your revenue today and in the last few years has been TCAD. And that's just maybe give us a sense of what you know what you guys deliver there to your customers, and what the competitive landscape looks like as well.
That's a great question. As you know by now, based on our press releases and our websites and, and as well as our calls, we have three business units. What we call IP business unit, EDA business unit, as well as what we call TCAD, which is technology CAD business unit. If you look at our reports in the, from the Q2 results, you'll find out that TCAD was 69% of our business, and typically, historically, TCAD has been between 60%-65% of our business. Getting TCAD into, this FTCO market, again, market expansion has helped us, and that's the proof in the pudding that our TCAD business is growing even more because of the FTCO. Our EDA business was 20%, and 11% was our IP.
Now, TCAD, if you think of it, is a very specialized software. There are only two companies in the world that have what we call TCAD platform. TCAD platform consists of, do you understand all these processes, steps, and brick and mortar of how to build these wafers? That's called process steps. The second component of the TCAD platform is what I call device. Now that you put these steps together, you make a transistor, can you model the process now? Can you model the transistors? And that's the second component of the platform. And the third component of the platform, our most important part of it, is modeling.
Now that you're able to simulate the process and device, can you generate a model such that now the designers can go and use those model to design their chips? And the answer to that is yes, we do that, and there's only, again, two companies in the world. We are number two in the world, but as we expand our SAM, we are striving to become number one in the world. And if you don't have any of those components, you don't have a TCAD platform. And you find out there's other companies that are trying to get into this space, but as you know, to get an EDA or TCAD tool accepted in the market, it takes a decade or so, and to a point that the customer trusts you with their multimillion-dollar chips.
It takes quite a bit of effort to develop the physics, chemistry behind it. The other limitation is the fact that there are only so many doctorates degrees that are given out, and that enable you to not only understand the physics or chemistry or optics or elements of these things, but also to be able to write software to implement what that physics model or chemistry model looks like. You know, we've been around for a while, and we have quite a few, like, paying customers that include universities. Those universities and research labs that we have worked with enable us to hire our next generation of, you know, staff that have the capabilities to do so.
Great, great. And then your TCAD is aimed at. Just maybe walk us through your core verticals that you sell the TCAD into.
Yeah. So TCAD mainly gets sold to companies that are developing. Before that, let me add a sentence to this. I want everyone to understand that we get our revenue. 95% of our revenue comes from advanced R&D projects. If you see a new display, if you see a display that's curved, if you see a display that's rollable, these are the projects that we have worked on many years ago with our customers. If you see a new device like silicon carbide or GaN, these are the projects we have worked on enabled with our TCAD with our customers. If you see a new memory technology, these are the products that our TCAD has helped and enabled customers to bring to market.
So TCAD is the fundamental, again, physics and chemistry aspects of any semiconductor design. And if you are providing new material, if you're in this place, for example, the new material historically has been material that are foldable, rollable materials, such as what are called quantum dots. These are QLED displays that you can go and pick one up from a store. Those are the type of fundamental physics and chemistry needs to use to simulate to see if these things are, first of all, designable, can you generate a model? And secondly, are they manufacturable? So if you think of the markets we are in, people who are into designing and manufacturing display power devices, as well as semiconductors, you find out that those are our main customers, and that's one level of customers we have.
You can think of us as a tier one, you know, and equivalent to what a tier one is in automotive industry. But also you can think of us as being tier two, providing solutions to other customers that use a fab. So in other words, the fabs are also our customers. If they have problems with using and understanding the process to steps or device steps, they use TCAD as well for those.
Okay, great. Great. Maybe shifting gears a little bit, we've talked in the past, you and I, about AI and how Silvaco was trying to apply AI within its product offering. Maybe just walk us through your view on what this technology can do for Silvaco.
No, absolutely. That's a great question, so if you think of application of AI worldwide is huge. We are not gonna discuss that. But if you talk about application of AI in the EDA, TCAD field in the world, I would split it into two portions. The portion that deals with design, the portion that deals with fab. In the design space, historically, us and our competitors have worked on three levels of AI for our products and to help designers, so if you think of that in terms of what we do in order to help designers is, can we make our tool user interface simpler?
Can we make it such that it can provide an AI assist such that not all menus when you're in a section of the design would pop up, for example, user interface, things of that nature that enable the designer to find things that they need to design with. That's the first level that we apply AI to. The second level is applying AI for optimization of problems that are complex. For example, in order to do simulations, sometimes you find out the manufacturer gives the designers this model with so many corners, you know, that it would take hundreds of millions of simulations to really cover all that corner in order to make sure this is manufacturable.
What we use AI for, for example, we do have tools that use AI in order to reduce the required simulations from hundreds of millions of simulations down by two orders of magnitude, a hundred times, smaller. So that's the second, portion of the assist for designers. And, as you know, the holy relic of AI for EDA is, for design portion is, can the tool just, you give it a data sheet that what you need, and it goes and spits out the design? So, I don't see that happen now because of the limitations of AI. And as you know, AI is in its infancy, even though I worked on AI in the computer science lab at SRI 30 years ago, still is in its infancy.
If you look at the history, in 1956, Alan Turing came up with the Turing test, right? And that test is a simple three-body test. And if you look at that test, basically, even the most sophisticated AI, ChatGPT, for example, the algorithms have not consistently been able to solve that problem. So saying that a problem was posed back in the 1950s, and the solution is not still there with the most sophisticated compute and algorithms that we have, it tells you we are in the infancy. So getting to that level of design such that a tool, an EDA tool, can go and complete the design, will take, in my opinion, at least a decade or so, if not longer. So now I talked about the three aspects of EDA that uses AI.
What we have done now, we are taking this EDA to fab, and that's the fourth dimension that we've taken it into. Now we are applying AI to look at terabytes and petabytes of data that's provided by fab. You know, I told you, you know, a memory product can take up to a thousand process steps. Every process steps on a wafer could have variations due to, you know, size of the wafer, how the process equipment works. Was it detuned? Was it calibrated or not? You find out that there are terabytes of data keeping track of this information. Now, how do you manage terabytes of data and try to make a decision out of it to find what parameters in the fab to tweak?
And that's where our digital twin model comes, that's where AI comes. AI reduces that you know, reduces the required understanding of that terabytes, petabyte data into much smaller space with a very simple graphical user interface to enable even the fab operators, not designers. Fab operators are very focused on yielding their products, and they don't even know anything about the design. So all they, all we're enabling them is to focus them on the only few parameters that they need to look at in this massive, I would say, jungle of data, and that's where we've actually proven that this is possible, so.
Excellent. Excellent. Thank you for that, explanation. Looks like there's still a long way to go for, for you guys in terms of, leveraging AI into your product set. Maybe Ryan, we'll bring you into the conversation for a minute, and just to... Maybe just walk us through how, Silvaco's looking at, you know, its margin model as it, as it kind of grows the business. When you were private, the company ran a fairly long time, at, a little above break even, if you will. You know, now that you've recapitalized and come public, just maybe give us some sense of how you see the margins, stretching out here.
Yeah, so I appreciate that, and it certainly is great to be public and have the balance sheet, and have the currency and the like. But really, I think it's a kind of coincidental timing that, you know, the great job that Babak and the team has done really focusing the company, and starting to hit its stride, and starting to get to the first kind of innings of scale, I think is making all the difference in the world. So what, folks, what's not necessarily intuitive is that the cost of our revenues largely is the cost of our staff that's supporting our customers, so by and large, they're fixed costs. So as we get scale, as we upsell the customers, add new customers, then it provides, you know, leverage to the model.
If you look at the gross margins that we have. I mean, I come from the chip space, and so I would dream about the margins that we have here in the software space. You know, getting into the mid-80s% in terms of gross margin, and certainly we've set the target model. We're on the doorstep of a 90% gross margin number. It's just really a great place to be. You can kind of just see that scale in practice. Then on an operating margin basis, I mean, it's a balance, right? Because we're of course starting to show that profit on the bottom line. You know, I like to kind of walk down the categories and think about it from an R&D perspective.
You know, no matter how much we add to the budget, it's never gonna be enough, right? We're always gonna wanna keep investing in the business and keep adding engineers, and we have to balance the short-term, long-term. But we definitely see these great opportunities with some of the business combinations, where we're able to pick up great talent, whether it be, you know, coming from other companies, or whether it be through acquisitions, and we're gonna wanna continue to invest in R&D. So that number, as a percentage of sales, we've always said, you know, we see it being there or even ticking up. Sales and marketing, you know, you know, we...
And when I say the royal we, you know, Babak had this great strategy of building out the geographic footprint to enable us to kind of compete with these big companies, and we've started to do that, and we've built out that infrastructure. And, you know, but we're gonna keep investing in sales and marketing. We'll get some levers, but we wanna just keep adding, especially whenever you look at adding, you know, adding FAEs and adding folks like that. It's just always, it always pays for itself in the long run. G&A is the area which, you know, of course, I personally am closely associated with, and you know, ashamed to talk about the percentage of, the percentage of sales we have in G&A.
We've added a lot of fixed cost, of course, to add the infrastructure, and add the team, and add the cost that it takes to become a public company, and we think it makes all the sense in the world. However, that number is not gonna scale with leverage. So we're gonna see leverage in our operating model from the G&A line principally. You know, and we've, you know, we talked about on the earnings call in terms of the target model that we set up, we set out for ourselves in terms of operating margin, and we talk it's a long-term target on an operating, you know, an organic operating basis. And one of the things that gets me excited, you know, this, we've all got a lot here, the team, a lot of experience with M&A.
When you do an M&A, part of the magic of it is enables you to essentially accelerate the timing towards those profitability metrics. You can take what I call the geography of the P&L. You're gonna continue to invest in R&D, but there may be some synergies as you continue to grow the combined company, where again, you're getting some leverage out of sales and marketing, but definitely getting leverage out of the combined G&A. So very excited to, as we grow and as we do an M&A, that we do it smartly and reward all the constituencies, all the shareholders.
Maybe on the M&A side of things, just help us understand what types of things is Silvaco kind of looking at? Or, you know, are you trying to just fill gaps? Are you trying to build adjacencies? Just what's your strategy on M&A?
Babak? You're on mute, sir.
Thank you, Ryan. That's a great question, Blair. Historically, as you know, and we announced it also, in our presentations, that we've done M&As in the past. We did actually over 10 M&As since 2015, and so we know how to do this strategically, and we know how to do this for four reasons. We do it for technology, we do it for talent, we do it for revenue. Historically, we've done it for those purposes, and one of the biggest metrics that Ryan and I are emphasizing going forward is we want any M&A activities that we do it to be accretive within six months of acquisitions and integration. So that's the additional metric we've added.
As you know, our bigger competitors have announced multi-billion-dollar acquisitions. Some of them are in the pipe, they're public. What benefits us with that kind of strategies, for us, we don't have to go and do large acquisitions to be effective and impact our, I would say our top line and bottom line. We are strategically looking, I would say, constantly on finding the nuggets that we know are there. We are actually aware that this, there's a big space for sub-$30 million, sub-$20 million revenue companies that frankly gets ignored or not paid attention to by larger competitors, due to the fact that they would not have any financial impact to them, but they're nuggets for us. We are trying to find those.
We are constantly working on this, and our targets are always in our sights. You know, there are multiple companies we are constantly looking at, and looking at our metrics to see if they make sense, and when we see that and the opportunity, Ryan and I have done this like routinely. It's not something that's new to us. We've done it in the past very comfortably, and we are ready to jump in and close the deals for us, so that not only our revenue and our, I would say, our top line and bottom line both grow, but also it would have a positive impact on margins, because we're very margin conscious.
We are not a semiconductor chipmaker that would be satisfied with, you know, with sub-70% margins. We really strive to have much higher margins and continue to improve that margin, so.
Okay, great. Great, just on the, sort of in the, the competitive environment, we all know that Synopsys is buying Ansys. What's your- what's Silvaco's view on that from your perspective? Does it help you, hurt you? You know, how does it change anything? Does it change anything for Silvaco?
The short answer is yes, but let me give you a bit of background. We have seen this, you know, acquisition play with larger competitors all over you know, for the past thirty years. If you look up the number of acquisitions these top three or four companies have done over the years, it's public domain, you see that essentially you find out that since EDA market is a very coveted market and it is high barrier to entry, even the larger competitors to get into a new area in which they need to get to, it would take them a decade if they do organic. So that's why there is a need for inorganic growth. The size of the companies they have to acquire to impact their P&L is much larger than what we need.
The understanding is that people need to have is where is the EDA market going from where it has been, to where it's now, to where it's going to go in the future? You'll see that Cadence announced an acquisition, right? Synopsys is also making acquisition in the same space like Ansys. And the reason that these acquisitions are there is where the market is planning to go. Where we are working on digital twins. Let me explain it in that terminology, since we talked about it, the better understanding is the physical world, you know, has dimensions, right? Our space in which we focus on is what I call from atoms to a wafer level.
We work from anything from atomistic level, all the way up to a wafer level digital twin. Then you find out companies like Ansys and Dassault. Dassault is also another giant in that space. They are not an EDA company. They are trying to get into EDA company, to become an EDA companies, and since there are not that many EDA companies around to purchase, then they get limited by what they can acquire. However, then the larger companies than them, Synopsys buying Ansys, they want to get into this new space of CAD, not necessarily EDA. CAD is computer-aided design, which is basically mainly mechanical simulations of engines, looking at thermal analysis of the engines, cars.
It's great to understand and acquire companies that do those kind of digital twins that are at that level, but that's a different physical space that where we play in. And by acquisitions of Ansys and what Cadence is, they are trying to get into system-level simulation and modeling. Systems such as a car system, for example. But you find out Dassault does this, but doesn't have EDA, it doesn't have TCAD, but Dassault does go simulation and digital twin models of airplanes, right? And then you get into Autodesk, and then you find out, oh, they do digital twin models of buildings and cities. And so the most important part of it is to see where this physical space, what portion of it is being addressed by each of these companies, and where their focus is.
Their focus is going to be system level, our focus is going to be wafer and atomistic level. And I think for us to really understand a system level model, design between model, unless you understand the fundamental physics and chemistry at atomistic level all the way to some physical parameter that you can hold in your, a ball size, if you will, it's hard to build on that. So that's where we are at, that's where the industry is going. And then now, how does that impact us? It impacts us positively in the sense that there is one less direct competitor for us.
And also, it enables us to actually not only go after the markets that we can address within that space that we are in, but also usually acquisitions result into a change of personnel, change of talent. And I think the talent pool for us will increase, and we'll take advantage of that.
Okay. Okay, great. You know, one thing we haven't touched on today is just the your IP business, which is a silicon IP business you've been in for quite a while. Just kinda walk us through sort of how you're positioned there competitively, and sort of where do you see that business going for you?
That's great. But as you know, we always tend to focus on the areas that we can, at some point in time, become a lead position. We're number two in the TCAD in the world, and we are striving to become number one. In terms of IP, we are far from that. However, we do focus on areas that we can lead, and our IP portfolio, you know, it's the youngest business unit we have. It's the youngest P&L that we have, as of five years ago. And we've gone from zero to 10%-11% of the revenue, based on what we focus on.
And what we've focused on is automotive IP, where we focus on is IP that we can develop without having to incur costs for tools, because we provide our own tools. There are IP companies that are pure IP players that have to pay a substantial amount of money for their tools. So our IPs tend to have lower number of R&D percentage associated with them, and cost in terms of tools. And then that has been historically the case. As you know, also, rather than developing and spending all the R&D to developing some IP, we've had agreements with one of the largest automotive companies that we commercialize their IP. We are actually doing that still.
We negotiated terms with them that in Q2 we started a new terms with them, so we could actually go and commercialize their IP. Half of our IP revenue has historically come from that. And then in the future, what we are focusing on, again, on value add IP, where we are not directly competing, but rather IP that customers need, but the bigger competitors of us cannot really or will not be able to, or have no interest in addressing. And those are... There's a huge market that's not addressed. And as you know, one of our strategic plans and work has been to be able to do what we call agile R&D. In other words, if there's an IP needed with- system features, we are willing to do it, and that's where we focus on, and that we think actually our IP business unit will be growing in the future, in the next one or two years, really fast, so.
Okay, great. We're, we've got about two minutes left here. I did want to drill into a couple more things if we can. Can you just talk a little bit about the power market, the power opportunity that you guys have been, you know, expanding into in recent times? And just maybe walk us through what you're offering there.
Yeah, so if you think of power, to be able. You know, you find out that power right now, besides the old technology of IGBT, that we provide solutions to simulation, but then there has been silicon carbide and GaN that has taken off quite a bit. And what we offer there are tools that enable companies that are developing their own product to be able to simulate those devices. And if you look at silicon carbide, just as an example, you find out if you go to the top 50 companies, there are that many or if you look at even top 10, you find out that their process steps, the variations that they do for their fab, even if they buy the substrate from a common source, the process steps in which they go, that helps them differentiate various changes.
You find out that, actually, we benefit greatly from that, because that variation is what makes them require more simulation tools. That variation and lack of understanding of what those parameters that they've added to this process will do, and that's where we actually help and come into picture.
Great. Great. Well, listen, we're at the end of our time here, Babak and Ryan. I really appreciate you guys attending today and sharing your insights with what you're doing at Silvaco. That sounds a really interesting story.
Thank you very much.
Thank you.
Bye-bye.