Silvaco Group, Inc. (SVCO) Earnings Call Transcript & Summary

Welcome to Silvaco's Tech Talk focused on the diffusion of innovation, investing in the ecosystem expansion. I'll be moderating this event. Before we begin, I'd like to cover a few housekeeping items. Today's event is being recorded and will be posted to Silvaco's YouTube channel following the event. I would also like to call your attention to our safe harbor statement and remind you that today's discussion may contain forward-looking statements, which are subject to the risks and uncertainties outlined in our filings, in particular, on Forms 10-K and 10-Q, and actual results may differ from those forward-looking statements. All statements are made as of today, and Silvaco is under no obligation to update these statements. Additionally, we will not be discussing Silvaco's financial performance or guidance. After the formal remarks, we will open the event to questions. [Operator Instructions] With that, I'd like to turn it over to Ian Chen, Silvaco's Chief Revenue Officer. Ian?

Good morning, everyone. Thanks for spending some time with us. So let me first just remind everybody with a quick overview about Silvaco. We are the first EDA company to have a successful IPO in the last 20 years. We are -- before and since the IPO, we have been growing in double digits percent. And we focus on applications with our growth such as power semiconductor, memory, photonics, et cetera. And our customers include the Who's Who of leading semiconductor companies, and we're attracting more talents since our IPO, our current employee number is over 300. Now while we are on the subject of numbers, let's talk about a few more. So we have -- 95% of our revenue each year comes from working with our customers' advanced R&D efforts. 68% of the annual revenue come from landing a new customer or expanding the commitment of existing customers. So -- that's to say that we are constantly working with our customers to deliver a new capability for their R&D efforts. In fact, on average, every quarter, we are actively working with at least 2 customers where we are trying to establish new R&D capability that they did not have prior to our engagement, and we are advising them on establishing new workflows. So getting customers to do something that they perceive as new is the classical definition of innovation. Today, we will explore a little bit about how Silvaco invests in our ecosystems to accelerate this diffusion of innovation. I will -- throughout today's talk, I will use primarily our FTCO or Fab Technology Co-Optimization solution as an example of what we do. First of all, I would say that enabling new technology, new R&D capability is core to what Silvaco do in almost all our products, not just FTCO. Now our FTCO solution is a way to deploy AI and ML for fab assistant. But first, let's talk a little bit about how the EDA industry is adopting AI and ML, right? There are generally 3 ways that EDA companies are working or have worked to incorporate AI. The first is making our design tools smarter. In other words, the goal is to get -- have the tool get to the optimal results that engineers want in less time by doing fewer simulations, for example. That's to say that with machine learning, AI tools can actually be more selective on what they do in order to satisfy a request. And this is something that is already showing itself in our tools as well as the industry in terms of building the next-generation more efficient tools. The second way that AI gets incorporated into EDA is as a digital assistant, right? So here, we're talking about an AI model can learn from the past experience of the designer as well as other in this community. And now they can understand new intents from the user and offer recommendations or decision strategies. Designers today are being asked to do more designs, more complex designs with the same number of headcount in the same period of time. So the second approach basically addresses the efficiency of the design. There is -- the third way is where everybody has been talking about using generative AI so that we hope in someday, we can actually provide just a design specification and the AI EDA tool can basically create a design based on that set of descriptions. So those are what EDA industry has been talking about in terms of AI. But I would also point out that the EDA industry for the first 30 years or so has exclusively focused on making designers more effective. After all, EDA, right, electronic design automation. So the industry has a myopic perspective that the complexity of semiconductor is in design. But in the last 10 years or so, that's no longer true. The complexity of making a semiconductor has gone beyond the design stage into fabrication, into packaging, into test throughout the entire value chain. So what Silvaco's FTCO's, I would say, innovation or the realization is that the same type of EDA capability and the type of tools could actually also be used by engineers in the manufacturing process, in the fab to actually assist in making decisions and improve productivity. Now to do that, of course, we have to do a few things different. Let's come back to the point that I said before. Innovation is getting customers to do something that they perceive as new. So here are the 3 main innovations for FTCO. First, it combines a physics-based simulation, quantitative analysis and AI/ML, all in the same unifying platform. Second, AI and ML requires a lot of data to actually become effective. So FTCO advocates starting an engineering creative process using simulations, in fact, using a large number of virtual design of experiment as a starting point. And then over time, engineers can judiciously pick data from physical experiments so that the model -- the learning model could be better calibrated with a goal that eventually we're building a digital twin. And then the third area is -- we are now talking about digital twin that are used by different people in the engineering process. So we have to make sure that the user interface of FTCO preserves how each engineering discipline prefers to interact with their data. So here, I'm just going to use the first out of the 3 points to illustrate how ecosystem can be leveraged to remove the friction of adoption. It will also apply to the second and third point, but for brevity, we're just going to focus on the first. If you look at the -- how a chip gets designed today, we can abstract it to say that there are roughly 4 stages. First, engineers have to select or invent a process by which the chip will be built. This means they would -- advanced R&D would have to consider materials, different process steps and how each material and processes to be applied to get the maximum yield for -- in the final product. Then given the design process, engineers have to create the device. The device could be just a small handful of transistors as in the power device or can be a handful of billions of transistors in terms of, let's say, an AI -- digital AI chip. So number three, then once there is a circuit design and then there is a process, engineers actually have to ramp the product to production. That actually means that the process engineers that help create the process, design engineers and the fab engineers have to all work together in this very crucial phase to figure out, a, is the product yielding the way that we had anticipated? If not, what's going wrong, we'll have to debug the process and prove it. And conventionally, this will mean that an upstream team like the process team will have to come in and help during product ramp. Now process engineers and fab engineers use different tools. Process engineers may use physics-based simulation like TCAD, whereas fab engineers want to look at statistical data. So what FTCO is presenting is let's put all these tools, all these ways of looking at the data into a unifying platform that we call the digital twin. So of course, we are talking about doing things a little different than what the current engineering -- from the current engineering practice. So Communication theory talks about the diffusion of innovation, meaning that a good idea doesn't automatically go out to everybody who has been considering it or considering to adopt it. It is a stepwise process that are controlled by 5 factors. The first 2 of these factors, relative advantage and viability is embedded into the product itself. So we can -- I certainly hope that we agree FTCO is a more efficient way of creating complex products. But it is something that existing engineers have to learn to use and the triability is delivered by our FAE teams working very closely with our customers to essentially show them how the FTCO would apply to their application and advise them on different workflows. So that's something that every company, every tech company has to do. But there are actually 3 other factors that control the rate of diffusion, compatibility, observability and the reduction of complexity. If you look at how ecosystem works for Silvaco and our EDA products, first, digital twins and AI are being used in many different ways in modern semiconductor manufacturing to help reach the next level of efficiency. So FTCO is a digital twin of a wafer going through the fab process. By working with other ecosystem partners, we can actually make sure that we keep track of all the other AI projects and digital twins that are springing up to help with the fab process and making sure that we're leveraging data from each other to make the result even better. So expanding utility is a part that is -- that addresses the compatibility part. We also look at research institutes and industry consortiums, which are part of our ecosystem. They are very important to actually not only develop the use case, but also the word of mouth of their experience applying FTCO. So our partners publish papers along with Silvaco presented conferences and that work with potential adopters. So here, the act of building evidence addresses the concern about observability. Also, if you look at what our customers and our ecosystem partners are doing is it is almost natural that when they use a new tool, working with different coworkers to address a problem, they are going to find new ways of doing something. In other words, innovation happens. So our FAE team working as senior consultants alongside with our customer will bring back these observations, which helps us make sure that our R&D team remove any frictions that our product may present. So the continuous feedback portion effort here works to reduce the friction of adoption or, in other words, reducing complexity. You've probably seen a number of -- on our website, a number of press releases regarding different investment and activities that Silvaco has made for its ecosystem. These are activities in the background that we've been engaging for years. And our collaboration sometimes go back -- the history of the continuous relationship sometimes go back a decade. And our -- some of our employees actually came from many of these collaborative efforts as well. Our work with ecosystem doesn't only benefit FTCO. It actually broadly benefits our EDA and TCAD products as well. We have announced our FTCO partnership with Micron to apply FTCO to memory manufacturing. But this is only one facet of how our products can be applied. So some of the other application of FTCO could include advanced display to look to improve efficiency, improve energy efficiency to reduce the weight of AR/VR headsets, for example, or reduce -- increase the pixel resolution so that the user have a better experience or in the case 5 years ago, we started working with display vendors to provide simulation capability so that today, we can enjoy cell phones that have a bendable screen. Similarly, in power semiconductors, we have continuously worked with our customers in terms of new process, new materials and how that actually translates into an efficient manufacturing process to help them get to production yield faster. With that said, I've almost focused exclusively about FTCO, which is only one facet of the products that Silvaco sells. Our ecosystem is, as I say, equally helpful in all the different markets and products we engage in. It is something that we have done over the last 20 years of our company to building different partnerships and keep expanding on it. And I expect we will talk about even more of them in the near future. That's all my prepared slides. Thank you very much.

Thank you, Ian. We will now take questions from the audience. [Operator Instructions] We actually have several questions already in the queue. Ian, how long do these partnerships take to negotiate?

Well, it's not -- I can't give a uniform answer. In some cases, we are talking about working with universities where there is negotiation between the larger -- the regions of the university on a broader contract that includes both a professor and Silvaco. There are times that we're dealing with small incubators that can actually move very fast. So I would say it varies from a few months to a few weeks.

Great. Next question, how many partners does Silvaco currently have and specifically in China and for AI research?

So, altogether, I would say we have about a couple of dozen of partners that are active with us. AI is really -- the way we think about AI is, AI is a design choice for our product, right? So it's an ingredient and many of these ecosystems are helping or applying AI in different ways. And in China, we have an incubation partner and a university partner.

Great. Next question. Can you expand on how these partnerships help Silvaco on the new product front or anticipate the next generation of R&D?

Yes, of course. So it's very important for -- in 2 ways. One, we are constantly working with advanced research institutes so that we are constantly at the forefront of new materials being applied. We're working with researchers that are consulting with leading semiconductor companies on manufacturing efficiency so we can understand what the current frictions are. And more importantly, as we work together with all the consortium like I said, when you bring people with different perspective together and put them on the same problem and have an intellectually in-depth conversation, you will find there are areas that requires innovation.

Next question. Can you discuss some other applications besides FTCO?

Sure. So for example, if you're talking about, let's say, a large digital chip with lots of memories, you can think of that as any data center or AI application you can think of. Well, the number of activities that involves different part of the memory and the size of the memory together could mean that a Six Sigma confidence in quality is no longer good enough. For example, we could find 1 billion transistors in a chip, they're all memory and certain part of the memory is going to be exercised more than others. So we are working with our ecosystem partners to look at how to -- what's the right way of going above Six Sigma, go to Seven Sigma, Eight Sigma. And at the same time, the number of simulation goes up exponentially with the larger standard deviation. So we have to figure out a more intelligent way using ML techniques to shrink the number of simulations so that, hey, it's still within the realm of engineering productivity to create those memory -- to test those memory before the chip goes out.

We've got a few more here. Next one, can you discuss how these partnerships are structured in terms of revenue sharing or new product development?

So again, almost every deal has its unique requirements. For the most part, we -- there is -- very rarely are there IP transactions in these systems. In many cases, the research institute actually become a favored customer for Silvaco. But -- and in some cases, it is just an agreement that we would provide inexpensive or even free access to our tools so that an industry consortium can work together. And so in almost -- in all those cases I've just described, the Silvaco retains all IPs and the revenue may not come directly from the ecosystem partner itself, but certainly, the participants in the consortium and can actually benefit from purchasing Silvaco tools.

Great. Next question. How many partnerships do you expect to announce this year and next year?

Well, we just announced one yesterday. So I figured that we -- so some of the -- some of our partners may or may not want to participate in a PR announcement, but I expect to probably announce a few more this year.

Clarification here. All the partnerships are with research institutions, correct?

Research institutes, which includes universities, but it also includes some industry consortium.

Perfect. We'll take one more here. Do you view the partnerships as channels to drive customer momentum?

Absolutely. They are -- they provide the observability that the earliest other adopter needs. So it basically shows other potential adopters in terms of what our tools can do.

And actually, Ian, we just got one more in the queue. Could you talk about the selection process? How Silvaco goes targeting or selecting a certain partner?

So the -- depending on the partnership, we want to make sure that a partner can bring in legitimate value to Silvaco, right? Because we are going to invest significant FAE and expertise to make the partnership fruitful. So if it were a research institute, we need to know that the researchers and the teams are aligned with where Silvaco is working on or will be working on. If it is an incubator, we want to make sure that the members of that incubation are indeed working on advanced technology and advanced products. So essentially, we want to make sure that the effort that we invest is well advised.

Great. That's all the questions we have, Ian. Therefore, that concludes today's session. As a reminder, a replay of today's event will be posted to our YouTube channel shortly. Thank you again for joining us today. Goodbye.

Thank you.