PDF Solutions, Inc. (PDFS) Earnings Call Transcript & Summary

Okay. Let's get rolling. Now we're on to the analyst portion today with Adnan Raza, our CFO and myself will be doing. It'd be -- of course, you all know our legal obligations to point out to you. But the first part of the talk, if we did our job correctly this morning is, I'm just going to bore the hell out of you, at least more than normal. But I want to touch on some of the high points we made earlier today, put it in the perspective of a little bit of what it means for PDF. Obviously, Adnan is going to come in and talk about not only our progress from the last Analyst Day, but how we see ourselves going forward, kind of our next long-term view. So I think it's compulsory in our space, if you're doing an Analyst Day, you have to start with, it's going to be a $1 trillion industry, data is growing to [ zetabytes ], et cetera. And this is all true. So I think in the context of PDF, why is this meaningful? Well, a lot of the growth going forward is being driven by AI, automotive, the collection of more data, meaning more sensors, more high-performance computing. Both high-performance computing and mixed signal are core markets for PDF. We have a demonstrated normally customer base success in bringing value to the customers. So the areas of the growth line up with places where we have experience and have had success in the past. Interesting, I didn't push the button. But this growth doesn't come without some headwinds we have to overcome. I think Sanjay spoke about it quite nicely this morning, as well as John. But the new architectures and materials mean that process flows are getting more complex. The maturity of the process is getting less, and that's not me doing that. The move to chiplets means the supply chain is extending, both in the importance and the complexity of the different stops along the line. But just the time from bare wafer into package chip outward has a lot of needs for more rapid understanding, while in line as to the causes of quality, reliability, excursion, more challenges for process control to have tight variation. And of course, we want to geographic disaggregation. Now this industry has always been geographically distributed with fabless companies in the U.S., with the OSATs and the primary foundries in Taiwan and the rest of Asia. But we're getting to the point where we're combining more chiplets, combining more of the supply chain and also standing up new capabilities in varied parts of the U.S., across Europe, et cetera. This is going to put a strain not only on the supply chain, but also on the engineering workspace and the quality of the experienced engineers and a number of experienced engineers that exist. So the industry is going to solve these. It always does. But these are challenges that are leading to inflection points that did not exist a few years ago. [Audio Gap] Of course, AI to the rescue. I don't know if, again, I've seen an analyst presentation in our space recently, didn't say they are an AI-based company, and AI was going to bring a lot of capability. And AI may be inevitable, but it's certainly not going to be immediate. Massive changes rarely are. But in our space, AI, although a requirement, will bring significantly more value than it even does today, requires some challenges again to be solved. There's requirements. One, again, we've been talking about quite a bit today is that need to combine that cross supply chain data integration. Not just from a semantic model and the ability to integrate the data, but also establishing the network and the security for sharing that data to bring the higher-level application. And it's not just about the amount of data. Our industry, my joke is semiconductors was big data before big data was cool. We always had a lot of data, but more data is not necessarily the answer. This is not an LLM type of space, where it's just the sheer multitude of data solve your problems. It's having data that's informative of the problem that the AI can drive the solution to good cause for you. And firstly, I'm impressed with the advances in models in machine learning and AI. Unlike -- some of you may recall, about several years ago, I'd say, look. And now, yes, it's coming, but it's more like the pixie-dust everyone's trying to sprinkle and cover the problems. I think today, given the cost of compute, the advances in the modeling that it is today already having significant impacts. But when we talk to our customers, it's less about needing help with deriving advanced models. The leading-edge companies, as John said, have groups of ML experts, a group of 20 to 30. But it's more about how do they go from the lab to the fab? How do they take these models and deploy them at scale? Even on a single model, you have to map from your ML expert through your production to all the different OSATs, if this is test-based ML, down to the specific testers. Now take that and multiply it by the number of products you run. Michael Sullivan talked about 75,000 SKUs. Now maybe you don't have to apply AI to 75,000 parts, but you certainly want to get beyond the tail of just your 2 or 3 biggest high runners where your experts can approach. And based on that feedback, we've really kind of focused and doubled down on the operations part of the machine learning model, to solve this challenge for the customers along with the big data, along with the DEX network for data exchange to really bring this to scale. Some of you have seen this. This is all aligned with our mission and vision. We really want to be the world's leading data and analytics platform for semiconductor and electronics ecosystem. We've seen this all integrating and being more interoperative as we go forward. Not just in automotive, but in other markets as well. And I think in many ways, it is consistent with the journey we've been on as a company for the last 30 years. When we started in the integrated yield ramp business, it was about combining design and layout information with fab and process information to understand the interaction. That has been expanded over time to test data, to more design information, combining chiplets to more assembly type of information. And we see it going upstream and in the third dimension to other organizations within a company. So we've talked a lot about the platform today. And again, this is the integration and how the development over the past 20 years has really grown together. I think there's 3 key things that are required. One is that AI drives the actionable insight. This is how do you solve the work force problem, but also how do you take the complexity of data to impact process control, capitalization, utilization, efficient use of your suppliers. And that's been a big addition on top of our Exensio and Cimetrix platform. No one owns all the data that they need to solve these problems. They don't even exist in any single company. So we felt for the last couple of years that partnerships are super important as well, both the ability to integrate enterprise level databases, but also to work with partners like the ones going next door, SAP and Advantest, to take not only their unique data, but their experience in their solutions to bring more value by combining them with the breadth of data and capability that PDF provides. An advantage to us is that does bring us new go-to-market opportunities for your partnerships. The vast majority of them are significantly bigger companies than PDF. They have reached into different markets or different customer segments. And of course, the transitional automotive electrification is a growth vector to us. We'll talk about not only silicon carbide, not only the mixed signal companies that are on pretty significant growth vectors, but adjacent markets. [Audio Gap] In PDF, we've really been applying AI and really machine learning across a number of our internal products. And this is really being driven to bring solutions, to bring actionable insights to our customers versus just analysis. But also to drive that total experience so that they can get their typical engineer to be as effective as their best engineers. That's been feedback to us from number of our customers as to why they believe that platforms such as PDF and applying things like ML can be most effective. Yes, it's the aggregation of data. Yes, it's solving the complexity, but it's also bringing this capability to a broader part of their customer base. We've seen this transition at a slower rate than the past, where more and more people consume the results coming out of Exensio in our platforms versus having to generate it at the expert level. In fact, our business model being mostly database or tool connectivity base is driven by the value being brought from the data and not the number of users. AI is just taking this to the next step. But it goes just beyond our products or our solutions, but our customer solutions across the equipment space, the design space, the ERP space, even to the cloud analytics that their solutions and capability combined with PDF bring even more advanced capability to the industry. And I think the work, again, that they're talking about next door with SAP and Advantest, I think, are highlighting that. That no one company can bring the cumulation of the data and the expertise necessary, but built on the PDF platform, it brings the scale and the ability to reach different parts of the customer base. Can somebody go forward? So in addition to the data or the applications that customers bring to us, I think the strategic partners bring specific benefits to PDF on the business side. One is it enables new products. We have new capabilities bring more value, not just from our core data, but bringing it to new users. SAP is an excellent example, where we can leverage our engineering data with the financial data to provide more detailed solutions such as product costing, RMA analysis, cost of yield to an entirely different group within our customer base. That makes PDF not only more sticky, but brings more value to the system. As I mentioned, customers for partners such as IBM, Advantest, SAP, Siemens are not only significantly bigger than PDF, but they also reach a different customer cohort within our customer base. And so this really is going to greatly help us in our go-to-market capability by leveraging their channels. And concurrently, we also, in several cases, have been able to introduce them to executives on the engineering side. So we're pretty happy of where we can leverage not only these enterprise platform partners, but also the partnerships on the equipment side in the engineering platform. In addition to machine learning and LLM models and growth in high-performance computing, automotive electrification is a big push. Whether the U.S. goes to completely electric cars by 2035 or 2030 remains to be seen. But the fact of the matter is the use of more compute within cars, the use of more sensors is an inexorable path that we've been on and will only continue. For us, in addition to this breadth of data, the traceability, the storage required across that supply chain, leading to more use of Exensio, more use of PDF as a platform. There's other specific growth areas, for example, silicon carbide driven by power and electronics, where the complexity may not exist on the product level, but certainly exists on the process level, from the bolts through the fabrication. We have business in the space. We have people reaching out to us with more interest. We see this as, again, one of the areas in our space that plays the PDF strength into understanding parametric data and complex analysis, combining both process information and the end characteristics of the product you care about. Lastly, we announced in the previous several months the acquisition of Lantern Machinery Analytics. And the reason we did that is we had a reach out from several customers or potential customers on how can they solve the manufacturing challenges in the battery space. This is a market where, as they told us, although fewer steps than a semiconductor flow has a great deal of complexity and variability, where the data analysis necessary not only to understand and diagnose the causes of quality or "yield failure", but new methodologies on how to control its mass production. Again, we believe this is a combination of having the right data and the right analysis. And through working through some of these pilots, we are pleased to see that not only does the Exensio platform and PDFS' capability very applicable, but the opportunity to bring this differentiated data. Again, one of the reasons we did the machinery analytics acquisitions. One of the reasons we partnered with Voltaiq is to bring to semiconductor the capabilities and to us, the business opportunities that we've seen in semiconductor. So a slide that you've seen before, we really view PDF today as an integrated platform. And is I mentioned in the previous slide that data access and integration, both through our metrics data connectivity, through our ability, through standards and to hook into literally thousands of different equipment models from our DEX network to combine data across the supply chain, is extremely valuable. But PDF is very unique in its ability also to augment industry data with that additional data required to have insights into root cause into process control to really have high profitable mass production in the future. I think Sanjay summarized in his talk this morning, both the challenges and ways that PDF brings value to our customers and helping to achieve these problems. When we did our last Analyst Day, we were just crossing -- it just crossed analytics being a dominant portion of our revenue. Today, analytics is the dominant focus today. And although integrated yield ramp may still exist, recall that it's mostly a business model standpoint and how we can help our customers achieve value and capture a fair portion of that, more than a technology play at these days. But the key is bridging that supply chain, providing solutions that drive value in mass production, not just technology development. So from PDF, the expansion is not just with the growth of the industry and via the partnerships, but from a customer base standpoint, we've continuously seen an expansion of the potential customers. Those of you who have known us for years, particularly when we were dominantly the integrated yield gram company. Our customer base was foundries, IDMs. And even within that space, it was when they are doing ramps of new technologies. So it's limited both in scope and time for interaction or gaining business with them. But since that time, we've grown not just from yield ramp but into mass production. Not just from the fast followers, but really across the fab and the fab and IDM business. Not just on leading-edge SSC, but quite a bit of business in mixed signal in other markets. We've grown into having a much more appreciable business with the fabless, if they own more of the supply chain responsibility, particularly the cost of test, the risk of OSAT assembly and time to market. With the Cimetrix acquisition in some of our partnerships, the growth on the equipment side has been significant and important to us. And we're seeing more, albeit at a lower level business on the electronic systems manufacturers, but also at the system level. Most of our business today on the system level are the companies that have chip arms. But even they are starting to talk about how chip capability, chip yields, managing the supply chain affect their overall product challenges, not just at the chip level. So we're pretty confident today or I should say, comfortable in the positioning of PDF going forward, right? I think John made the comment to me not that long in the past. I totally concur that we think PDF is more well positioned for the opportunity in front of us. And I think we have been at any point in the past of our 30-year journey. And with that, I will turn it over to Adnan. It just has a mind of its own.

How is this now? Can you hear me? Now you can hear me. Okay, perfect. So again, very nice to see a lot of familiar faces. For those on the phone or listening into the webcast, Adnan Raza, CFO at PDF Solutions. And to start off, I'll make an admission. I have not been here 30 years, unlike John and Kimon, who you met and obviously have heard about. Neither have I gone through Carnegie Mellon, although I did share the state of Pennsylvania, the other end, to go to school and met some fellow alums today. I will say sometimes I wonder if I said it really fast that my younger brother went to Carnegie Mellon, and maybe that's the piece of it that they found some affinity with and here I am. That said, also for those listening into the webcast, I wish you were here and able to hear some of the talks that we had from outside speakers. I think since you weren't able to hear them, I'll just mention it that we're very thankful for the gracious nice words that folks like EDI and Intel and GF had to say about the capabilities of PDF, and more so about the trust that we bring to the relationship. We sincerely believe we are the end-to-end market player in the semiconductor analytics space, and have been here for a while and have been developing those solutions, sometimes taking longer up to 10 years, like John mentioned in his presentation. The next section, market size and opportunities. Thank you for advancing the slides. I intentionally put it. Look, I mean, we were debating whether to put this slide or not. So let me give you a little bit of context. What we did was we worked with a Wall Street investment bank to look at 3 categories of companies as noted on the right side. The fabless logic, the IDM and the memory of the storage categories. And for each of them, we looked at what are their disclosed R&D spend and what are their disclosed cost of sales spend. And then the Investment Bank also had conversations with the IDC and partners of the world, some of the analysts that you heard throughout the day to say, in their estimation, there is no [indiscernible] to report out. So in their estimation, what is the percentage of those R&D numbers or cost of sales numbers that are spent by those companies within the analytics bucket. We did do a triangulation to compare it versus the fab, versus the EDA spend and got to a point where we felt comfortable enough to share that, look, a, in my mind, the takeaways are: a, the industry itself from 2019 to 2023, when you look at it, it's about a 40% growth. Why is that important? Just looking at the metrics this way is willed when we talk to the next phase keep that 40% in mind over the last 4 years. 2019 was our last Analyst Day, 2023, obviously, where we are. The second thing to think about is whether the numbers are big or how big, and whether you believe it or not or whether you find yourself more in Marco's camp this morning, who astounded me with a number of $7 trillion for the AI opportunity, which was great to hear. I think the key thing to note is as you look forward from now until the end of the decade, you're looking at just under 10% growth rate for the industry. So nothing short of saying, okay, great, good growth ahead of it that the industry has to serve. If you go to the next slide, this is a more PDF style on this bottoms up, SAM style assessment that you have seen us talk about in terms of what are the areas where we think we can provide our product? What was that opportunity? How big is the market where we can actually sell products into? If you go back and look at the 2019 deck, it was $1.3 billion. We looked at it today and did a full bottoms-up analysis, where we looked at all the individual players in the space, the different products that people serve, what we know about those industries, what are the estimated market sizes of those opportunities and we came up the $2.6 billion. There is a very small AI-driven [ loss ] factors that we incorporated into this, but nothing like the pixie dust extrapolated $7 trillion-ish kind of numbers, obviously. But it comes out to the $2.6 billion. So when we sat back and then started looking at this data and put it in the context of what we talked about on the prior slide for 2019 to 2023, today was a 40% growth for the industry. We felt pretty good that the market size today is about 2x what we felt it was in 2019. Second thing, if you go back and correlated versus the total growth rate that we have demonstrated over this time period, PDF is about double the size, just under double the size compared to where we were. I think it's also important to note that, look, PDF as a company that have invested in a lot of technologies. You've heard us talk about DFI. You've heard us talk a lot about AI and ML, but 2020 onwards is where we thought the momentum of our activity started to come through. 2020 and prior years, a few years before that, we were flat in revenue. Candidly, if you go back and look at the numbers and most of you know. However, 2020, our bookings were quite strong, well over 100%. We've talked about this. Some of those results of those deals then started to come into 2021, where we demonstrated and clocked in 26% growth. All of you remember the Cimetrix acquisition, which was done the year prior. So the conversation also was that in 2021, the growth rate partially came from acquisition. Admitted, absolutely happy with that acquisition. And by the way, it's performed better than we expected, particularly when the CapEx cycles were strong. 2022, therefore, was very important for us to clock in 34% growth, which we did, which was exceeding the prior year with the acquisition. So overall, we felt we were demonstrating good results. That's the kind of second bullet on the overall revenue growth rate. But then if you go and look just into the analytics revenue growth rates and think about, okay, how has analytics done over time? And you all know that we were transitioning from a historical gainshare model to an analytics model. This is where the remarkable number comes in, 2.8. For those of you who's good at math, 35% CAGR. I'll put it down on another slide later, we'll talk about that as well. Why has all this happened? In my mind, it's a perfect storm. It's where our cloud technologies have become mature. We are starting to make good inroads with multiple products, and I'll share with you some of that data in a later section of the slides, into the customers. And where some of our ML and advanced solutions are making headwind to the customers. And finally, what you all have noted as our spend in the DFI category, for many, many years that we've been doing and a few of you have rightly complained to us. But that has started to bear fruit. If you actually go back and look at our Q1 '21 earnings call, I think that's where -- kudos to Johnny, who came out and said that this is a year either will make some dollars out of this investment or will think otherwise. And pretty glad how things turned out after that, in terms of we were able to make the progress. And for a long, long time, we weren't able to say which customer we were working with. But thanks to the presentation from Intel today, and appreciate their gracious acknowledgment of our activity and help that we've been doing there. So it's a combination of those factors, which we're very pleased about the overall growth and the transition that we have done and also the analytics. Now I'll put this in another context. I talked about 2021 growth of 26% and 2022 growth of 34%. This year, obviously, just like we put out in the press release, no different than what we talked to you about at the end of Q2. This year, we're targeting lower double-digit annual revenue growth rate, which in of itself, compared to the industry, is still pretty good. And also, when you look at it over the last 3 years, it's still north of 20%. So a long term from a growth perspective, as a software company, creating new products and 20% delivering growth over a multiyear basis. Average, it's something we feel pretty good about. [indiscernible] for the next slide. All right, I'll control. All right. So this is a section where I wanted to talk to you a little bit about progress since the last Analyst Day. But as I thought about presenting this section, a lot of our conversations started coming back to me, in terms of how to explain PDF. A lot of you keep coming back and saying it's a complicated story, hard to understand, hard to explain. I can see some smiles in the audience. So for me, I had to make myself a little reference to you of, okay, what are the different pieces that go into PDF? What are those different product offerings? How do we structure the revenue for those different pieces? So let me just give you a walk through, and then that will help frame some of the thinking. All right. So at the highest level, if you go to our filings, everybody is familiar with this level of disclosure that we do in terms of revenue, we break it out to analytics and IYR. We've also talked about how analytics again, many times today, has become just around plus/minus 90% of our revenue, depending on how you look at it over which time period. Further description, within the analytics piece itself, let's focus on that one. There's the Exensio piece, which is the core software piece that we have developed, many ways to deliver that. I'll talk to you about in a second about those. And then there's the differentiated data. What is differentiated data? Differentiated data are the different ways in which like Said mentioned, when we cannot find the data available in the system, we go and create that data. What are some of the examples of that? This is the DFI machines that you hear us talk about. This is the CV infrastructure that you hear us talk about, wherein we use the knowledge of the fire software combined with those machines to create the differentiated data with the help of which we can use the unique analytics that we provide to our customers. The third piece that in the analytics bucket, of course, is the -- let's go back -- is the Cimetrix connectivity, which is the acquisition that we did. We feel -- I mean, if you think about the platform picture that Kimon drew earlier, it starts with having data generation, then it starts with having the connectivity for that generation of data, and then the analytics engine that you build on top of it, which is what makes for a good platform. If you go back and look at our platform slides, this back is posted. Obviously, all of that has to be encompassed with the trust that encompasses the entire span of that value chain on one side. And the other side, we talk about the partnership that you've heard throughout the day, including the SAP presentation that's going on. The bottom portion is obviously the integrated yield ramp, wherein we go in and we'll charge a fixed fee and then collect the gainshare, some of which gain share contracts even today that we have that we are reporting good until the end of the decade. A little bit more description of that I just described for you, but I can read it later. But then I think the important piece after that is to think about what are the different types of software? Or what are the different types of structures within each of these different categories? So if you stay with me on the analytics row, within that, the Exensio analytics, and just kind of follow it across to the last column, you'll see it under the type -- the kind of first 3 pieces. The way we sell our Exensio analytics software, like a lot of software companies do, is 3 ways: perpetual customers, the perpetual software. A lot of our customers in this industry, you heard the talk from EDA about the history of integration of acquisitions. You heard about it's been done a certain way. A lot of our customers will prefer still to use the software perpetual basis on-premise, which is we're happy to sell -- which is how we're happy to sell it to them. Some customers will get it from us on a term-based license basis, which is, again, installed under premise itself. It gets a little bit complicated. Some of the customers might come back and say, "Oh, we'll do it term-based license, but we're doing on cloud, but keep it simple. It's perpetual, and it's a terminate license. The third piece of that is, of course, the cloud piece. So when we're talking to you quarter-to-quarter on the earnings calls, when we talk about different deals that we have won or what kind of nature of those deals, this is the framework that at least I found handy for myself that I wanted to share with you. That how it fits in. The others start to get a little bit simpler in terms of differentiated data, how do we charge for that data. It's term-based license. You heard us talk about some of the large enterprise deals we've done. And the Cimetrix connectivity space, it's mostly perpetual software. We're starting to just dabble into looking into if we can make some of that recurring, but it's way too early with most of the all perpetual. And again, in the fixed fee and the gain share side of the things, we have talked about how it's service and time as well as royalty payments. This page is important. Also, when you think about the backlog number, the remaining purchase -- performance obligations that we talked about and disclosed in our 10-Q filings. What we disclosed is our committed, confirmed noncancelable orders. That is what is our backlog number. What it does not include is when customers will come in and buy some of the software on a perpetual basis. Because, again, we don't know when that order is going to come in, when it comes in, it gets fulfilled. So it's understanding that. It also doesn't include some of the -- and perpetual software, by the way, both in the extensive piece as well as in the Cimetrix connectivity piece. Then it also does not include some of the gain share pieces in the backlog portion, because gainshare is a function of the end customers' production capacity, and we don't know what that is going to be. We're obviously happy to collect that check when that production happens and comes through and it's very high margin for us. So we're pretty pleased with that. So hopefully, this served as a good kind of cheat sheet presentation page for the rest of you. Thank you for advancing. This is a high-level summary. So from here, I'll present to you a few data slides, but instead of just going slide by slide, let me just give you a little bit of a road map and kind of road signs to watch out for along the way, so you know where we are in the presentation. So we'll do 9 slides. Think of it in like 3 sections. The first 3 will just give you some data, some observations. The next 3 will be some insights that I will share that underscores some of that data, and the last 3 years are going to be just zooming back out at the corporate level for what that means for the entire company. So let's go. This is kind of, call it, Slide 1A, if you will. Let's stay back to that one. I think I've covered a lot of it, but just to highlight some of the things, 20% even on a CAGR basis for the total company, which is the point that I was making that even with the slower growth that we have talked about for this year and share with you all, it's on a CAGR basis of 20%, measured over the TTM period compared to when we last did the Analyst Day. And then again, the CAGR 35% that I alluded to earlier, a few slides ago as well. Analytics now, last bullet on the right, is also important to note, it's right around 87%. It will fluctuate depending on what kinds of deals, when we're doing them, what is the structure. Sometimes a gainshare payment may be high and expectedly, sometimes the gainshare payment may be lower than expectedly. But depending on that fluctuation is where that number will, plus or minus oscillate, but I think it's right around 90% is the good yards to think about. And that will be important when we talk in the last section about why we think of analytics as a driver for the whole company, and that's how we'll share some of the long-term targets that we will discuss. Next slide. Okay. Slide # call it, 1B, if you will, in that kind of 3-part structure. Look, we're pretty proud of how the business model has transitioned, and how our margins have expanded over this time. Of course, we're also proud at the same time of the balance sheet this we're not talking about here, continuing to maintain a healthy level of cash, not having any debt. But put that aside, as the business has transitioned to a recurring nature, as been able to better utilize our resources and people, as we've looked at how do we take benefit of the scale and put our people on projects where we can demand more from the customers, it has started to show itself in a higher gross margin. It has started to show itself in a higher EBIT margin. Gross margin, you can see 900 basis points improvement, 74%, as measured over the TTM period. Last couple of quarters, it was right around 75%. And then again on the EBIT margin side, right around 19%. Put this in the context of when we lost at the Analyst Day in fiscal 2019, our gross margin target at that time was right around 70%. So we've been in excess of that. EBIT margin, we're getting pretty close to approaching that, which is a 20% target that we set at that time. Again, we'll talk about new targets. Kind of my third and next slide. My third and last piece on the observation piece. This is new data. A lot of you have been hearing us talk about recurring revenue. And a lot of you have been asking us, okay, can you give us some sense of how much it is? What's the measure of it? How should we think about it in terms of scale within the analytics or within the company itself? And we went back and looked at what that number meant for Q4 of 2019. And then also on a trailing 12-month basis, which we think is the best measure. Because quarter-to-quarter, you could have some fluctuations depending on accounting and depending on rules around revrec. On a TTM basis, it's approaching right around 70%. Again, a number we're pretty pleased about. If you think back to the earlier slide, where I talked about there is a part of our business that is perpetual. There's a part of our business that is nonrecurring, if you will, that we also don't include in the backlog. There is always going to be a portion of our total analytics revenue that is going to stay either perpetual or it's going to be customers just buying the software so that it is not of a recurring nature. All of those will mean that there's a terminal limit, and also there's some services. So all of the which mean that there's a there's probably doesn't get to 100%, but it's a number of 70% is something that we're very comfortable with, and I think it's a pretty good metric as well. Okay. Now I'm going to start showing you kind of the second set of the 3 slides, which is some of the data underlying that. This particular data is about the analytics revenue for our customer. If you look at our quarterly investor deck -- actually, we put this as a bullet on the slide. Some of you have noticed it. Some of you -- actually 8 students in this room came to me just 10 minutes before this presentation and told me of the slides that were already posted, they have studied those and they had 1 or 2 questions. So kudos to you guys and call you out a little bit. So in terms of analytics, if you just look at the revenue per customer, pretty pleased with the fact that it is just under double, from $455,000 to $840,000. We are, as noted, not including Cimetrix in this metric. Cimetrix tends to be anywhere from a few hundred dollars to a couple of thousand, but it's very different. They're serving a very different end of the market. It's more connectivity software that's going on the tool when it's going to get shipped, a little bit different than what we do. Fits very tightly from a strategy standpoint and a platform standpoint. But when we think about analytics revenue per customer, we think XCG is the right metric. And like I said, pretty pleased with the fact that it's just under double. Why has this happened? It's all of those things that we've been talking about, which has been the theme throughout the day. It's the platform. It's the transition to the cloud. It's a recurring nature of the software. It's more stickiness. It's the ability to sell a larger platform -- a larger portion of our platform to the customers, and I'll explain that in one slide or two as well. Next slide. Okay. So I think we -- in the last Analyst Day that we did in fiscal '19, we talked around some of these numbers, and I think we may have had one slide or two at a higher level, but I just wanted to be a little bit more explicit here, this time sharing, since we candidly have a little bit more data and a history of having served the customers. So if you look within analytics and then think within the Exensio software itself, then within that, I think about the 4 modules that kind of mainly constitute the Exensio platform. It's those 4 modules listed on the bottom right, the manufacturing analytics, process control, test operations and assembly operations. Then we went ahead and we said, okay, let's look at the average revenue by the number of modules for these customers and then see how does that compare? And typically, when you add another module, the point of this slide is to show that it's not just the doubling of the revenue that we tend to see in our customers. The power of the platform is such, that we are able to see just under 5x the revenue that we are able to make with 1 module, when we go to a customer using 2 modules. Beyond 3, we're a little bit in the early stages, there are customers that are using those, but we are able to see a step-up of about 9x compared to the customers that are using the onetime. As customers adopt the platform, we think this is one of the areas where our opportunity lies to expand our presence and become more and more relevant and sticky with the customers and grow our revenues. Next slide. On the last slide, I alluded to, there are a few customers. So we put a bullet point on the top right. It's worth taking home. Over 20% of Exensio modules customers today are utilizing 2 or more. So it's just about 20%. That should give you an idea of how much opportunity there are in our customer base of those that could use 2 or more modules. And it also gives you a sense of the fact that yes, we have demonstrated this growth of CAGR of 20%, and CAGR well north of 30% for analytics customers over the last few years. But even now, we're just about 20% mark for the Exensio customers using multiple modules. Within that, it's a little bit plastic PDF tell title. You'll probably have to read it 3 times to understand the exact math behind it. But we looked at the average of the multimodule customer revenue, and if you add them up for any period, then we look at, okay, over that period, for example, the trailing 12 months of Q2 versus the entire calendar of 2019. So using the same methodology to compare. What is the growth in those [ fence ], in the pan amount for those customers. And again, it's a trend that we see. Average revenue per customer that we saw for analytics is increasing, average revenue when the customers are switching to multiple modules is increasing. And then this in of itself. What this is saying is, customers using multiple modules are spending a lot more today than they were a few years ago. But all the different pieces together is kind of where I think the story comes true for PDF, and why we are feeling like we are in a great spot given all the trends that Kimon also alluded to earlier in his presentation. Next slide. Okay. So this is the last kind of the 3-part sections, if you will. So kind of done the 6 data slides and we'll do the last 3 data slides. If you look at us compared to 2019 versus now. Again, moving back out to the corporate level. 6%, when you look at combined revenue growth plus the non-GAAP EBIT margins, about 6% the last time when we had the discussion with you guys for the full calendar year 2019. I guess, 2019 wasn't over then, but this is full year 2019 measure. But today, when you look at it for trailing 12 months of the Q2 that we've reported, it's sitting at right around 47%. How do we feel about it? A, we feel pretty good about it. But as we think about it, what is our North Star? How do we think about the opportunity? There will be periods of growth. There will be periods where we'll be able to charge higher margin. But for us, when you think about revenue growth plus in EBIT margins, we think of 40% is a good yardstick to think about. Again, this is our long-term target. There will be periods in which we may not meet that, right? Where we're investing for growth ahead of it coming in, and we are spending a little bit more on the operating expenses side than we might want to -- then it might be really apparent. I mean, if you go back and think about the time when we did the Advantest deal, back then our spend in the cloud was well ahead of the opportunity itself. Again, we were getting ready for a large opportunity, and that is a $50 million deal. If you think about some of the leading edge engagements that we did in 2021, we were again spending ahead of those opportunities, which is why our operating margin takes a little bit of time sometimes to come through. But this will remain sort of a good yardstick that we will continue to take by about ourselves. And the second of the last section of 3 slides. Another new data point. Some of you have asked us, yes, we see the disclosed number of the total backlog number in the filings. But how do I get comfort? Some of those people that asked the question to, how do I account for it to get comfort around how much of that is going to convert into revenue for the next year, right? So -- and a low number in this case is a good number, because the low number means that my total backlog in of itself is pretty large. However, the amount that I'm going to convert in next year, as a percentage of what I will eat out of that, what number is that? It's not a hard rule per se, but we think somewhere around 50%, does it oscillate around that? I don't know. Maybe we'll see 45%, 50%, somewhere around that range. We'll see if that number kind of oscillates around the 50% mark. Today, looking back at Q2, it sits at around 44%. And when you go back and translate that into the dollars, you'll see why we start to feel comfortable when we think about growth rates, why we think comfortable and how much of the revenue is spoken for next year. Candidly, since -- I've joined about 3.5 years ago now. And whenever we've done the AOP for the next year to our Board of Directors, we've included this metric about what percentage of next year's revenue is already in the bag. So this is another way that we track this internally that today we wanted to share with you. Lastly, turning a little bit to the balance sheet. I alluded to earlier, Kimon and John are very conservative folks. So that has not been on our books for a long, long time, and we intend to keep it that way. I'll let something really amazingly, in terms of opportunity presents itself, and we will revisit it. Not to say that it's a Cardinals never to look at that. Today, last quarter, a pretty healthy cash balance as well on our balance sheet. But then if you go back and look from 2020 through the last quarter, how much have you spent? It's really been in a lot of what we think good ways to return the capital to the shareholders or invest for the future stickiness and sustainability of the company. CapEx is around $25 million. So call it about $25 million to $27 million for both CapEx and share repurchase and the balance of about $30 million for M&A. The predominant portion of it, the older portion of it was our Cimetrix deal. Pretty pleased about the balance sheet that we've built. Pretty pleased about the margin progression that we have done. Pretty pleased about where we sit with our opportunities, and also the underlying metrics that we're today giving you a little bit look under the covers for. Next slide. Okay. This is the last section. Opportunities ahead and kind of how we think about them. If you give in the next slide. So long-term target financial model. We went back and we looked at what we talked about at the last Analyst Day in 2019. And you'll see that we put 2 rows there for the revenue growth. The last time we were all -- we were talking to you, we were talking about the analytics revenue growth. Now the business having transitioned over to most leading analytics, it makes sense to talk to you about the total revenue growth. Because essentially, that's a proxy for the analytics growth as well. We think, given the successes that we have seen, the stickiness of our platform, where we're seeing the business over the long term as we think about our revenue, as we think about our total growth, we feel that greater than 20% is the right target to set over the long-term basis for our total revenue. On the gross margin side, last Analyst Day, we had talked about 70%. We have started to clock in 75% quarter-to-quarter, again, there might be variations that are North Star of what we want the long-term gross margin target to be, is going to be 75% plus. And we think there's a path to get there, as we have demonstrated in the last few quarters. On the non-GAAP operating margin side, if you just go back to the last slide. We've stayed at the 20%. We're not changing that. The reason we're not changing that is because of the variety of opportunities that you're hearing us all excited about today. For us, it's always going to be a balance between what revenue growth do we deliver versus what operating margin. A few slides ago, we talked about roughly 40%, but some of the revenue growth and the EBIT margin, the way we think about it. However, on the operating margin side, we need to be able to continue to invest. And some quarters, it could be higher than that, if the opportunities are at the different types of investments, and some quarters may be plus minus that. But I'm pretty excited about the fact that we can talk to you about total company growth. And then the combination of those 2 as we look at it being 40% as our North Star, if you will. Some of the areas that we're looking to invest and candidly, these are not modeled into our internal projections right now, but we're pretty excited about. One investor called, as we were getting ready for this analyst day and said, one thing that would be exciting to hear about from you guys is DFI adoption is starting to happen, but it's still in the early stages. So what is the free option value? What is the option value essentially a shareholder gets in buying a PD stock? What are the opportunities that could become big over time? And candidly, as we sat back and thought about it, these are before that came to our mind. So DFI eProbe, like we discussed and like Intel talked about today, starting to see early signs of adoption. Pretty pleased with that. Obviously, as that scales, if there's more opportunities to that customer, more opportunities than some of the other leading-edge customers. Some of the analysts have talked about how big that opportunity could be some analysts were in this room, actually, I'd see. Premium, again, an opportunity that has started to have early proof points. We started with this in China. And it was -- we always say necessity is the mother of invention. But candidly, the inbounds that we're getting for the number of customers was a lot. And we couldn't serve all those customers. So we started offering a limited amount of database, limited amount of capability, things that people could do, and seeing who is using how much of the platform, who was going back in constantly deleting that space to reuse that. Well, there you go. That's my free marketing spend or free sales plan to figure out who is the customer that I market doing. We were surprised how many of those we were able to convert into actual paying customers. And some of the discussions we've talked about what those numbers are, but pretty surprised, higher than what we would have expected going into it. Partnerships, I think you heard us talk a lot about as well. When we think of partnerships, we think of them across the whole platform. So it is on the equipment side. It is on the test side of the seamless of the world. It is on the SAP side going vertical to the platform itself. And it's that ecosystem for us that makes for a sticky platform, and that's why we become relevant as well as we establish more of those. Battery, early stages. Obviously, some of the few touch points we've talked about here are the machinery analytics acquisition that we did. It's basically a camera system, not too dissimilar in some of the other data observation systems, if I can make it at that high level, like the machines that we have in DFI and CV. And it goes in and is able to analyze some of the battery characteristics during the manufacturing process to say, okay, we're running some machine algorithms on top of it to predict what the yields could be. Candidly, this is not an opportunity we were such forthright about and going out and seeking, as the battery industry was developing these batteries and there's 3 major players in the world in this market that kind of have the majority share close to 70%, I think, but check my numbers. One of those guys reached out to us and talked to us about process yield improvement, and which other areas they had seen software tools used for process improvement, semiconductor being one of them, and their PDF was the name that everyone talked about. So it was an inbound discussion that came in. And we did a pilot, along with machinery analytics and found that to be a pretty interesting solution, which is why we went ahead. But it's a combination of these opportunities. I mean if all of these come through in a meaningful way and start adding to revenue, there's sprinkles and starts, so some of these happening already, then I think we're very well positioned, which is why we're excited, which is why we think about expanding some of the dollars along the way into maybe sure those opportunities are invested in the right way before they start to bear fruit. And then I think this is the last slide. We kept this one. This is similar to the one that we had in the 2019 Analyst Day. We went back and looked at this. Pretty proud about all of the things here. These are the things probably been more promised. If you thought about it at all at Analyst Day in terms of improved profitability, increased visibility. Through the last few years, hopefully, we have done that by improving our margins and hopefully, by giving you some more data, starting to talk a little bit more about the backlog that we have been for the last few quarters. We have started to deliver on some of these metrics as well. And as Kimon is checking his watch, if you go back and go to the next slide. I think that is it over that. will. Thank you. And obviously, other sessions are going on. We're happy to take a few questions, given our booth here. But to the extent anybody wanted to hop back into the executive sessions, you always have that opportunity so I think they're playing on to 5 30 or so. But with that, we'll open up the floor. Any questions, happy to discuss [indiscernible] answer.

So Kimon, we started to talk earlier like if the DFI opportunity continues to expand, at what point do you have to decide the model? Is it a big CapEx commitment for you? Do you have to partner? What are the practical considerations in terms of? How you could scale that?

Yes, it's a good question. I think it's obviously a good problem to have. And I think from PDF, our perspective is we're going to do what's the most efficient from a go-to-market from a shareholder, and have the ability to go between scaling internally as we do today, to using third parties on the manufacturing, which is what most of the CapEx guide. If it makes sense to partner or outsource the CapEx portion of the DFI solution. That's a possibility. I think what's unique in PDF is in the direct scan and the DFI solution is there's a recognition in our customer base, the software wrap around the differentiated tools. [Audio Gap] So as a company that's analytics focused, we don't have to make all the money on the tool. So right now, the customers are more than happy to stay on a subscription model with the tool. If it per se works out that we're using our customer moving the tool to a cap [indiscernible] a PDF or someone else, you can kind of position in the sense that a lot of the value is still providing the software on the tool setup and the tool analysis, which I think our customers have understood, if you have a transistor with or a layout with 50 billion transistors and hundreds of billions of structures, understanding how to drive a recipe and what it means, to analyzing billions of measurements that come off of it is a big analytics problem regardless of tools. We've already solved that. And I think that is going to be a continuing point, regardless of where we move on the CapEx side, be it from a balance sheet management standpoint. [Audio Gap] Parts, et cetera, that we'll maintain, and I think we retain a lot of value.

When you think about the opportunity for DFI -- years ago, you talked about the number of tools that you could potentially place -- could -- is there something you could provide along those lines now for how you're thinking about the opportunity going forward, either in the intermediate or longer term?

Yes. I think there's -- one way to look at it is kind of a [indiscernible]. If the key for PDF and the reason we've gone into the business, we believe there was a need for mass production, not to technology. In mass production, by definition, no one is going to have one of the [indiscernible], because there's always maintenance, there's always downturns, et cetera. But by the way, our feedback stand up straight or talk. Okay. I guess everyone's in here. And by the way, the uptime on our tool, I think, is better than anyone in the industry right now. Minimum people [indiscernible] but purposely, that's assuming minimal sampling today from that standpoint. So it's tens of tools, dozens of tools out in the field over a very short period of time as you move into mass production. If you get to the point of you need high sampling, you need multiple measurement points. In fact, not only for the front end, but pretty 1x net to the complexity in the back end, where Sanjay pointed out today, maybe even on the backside power. Then you have to look and say what fraction of the entire section market moves from optical to it? And then you're going to look at KLA sales and market cap as a metric and take some hair cut of 30% of it going forward is an operate possible. So it's a significant market, which is one of the reasons we saw the need, but we also saw the potential opportunity, which is one of the reason. [Audio Gap]

For the presentation today, very helpful. I have a couple. First of all, can you just clarify on the DFI side. Is it strictly being used in the lab today? Or are there instances where it's being tested in production?

I think what -- if you refer back to Sanjay today is it's used not only on test chips and employing technology development. It's actually in mass production parts. Big news. So I think that we seen in our customers, the need and the ability to ramp and maintain complex leading products today with DFI. Now the part that I think is an open question to PDF in the industry as the mass production matures, as you get into that decade of mass production, to what extent and what sample rate is that virtual contrast inspection still necessary? As an industry we haven't got there, so we can't comment on how much it can scale back or how much is needed over the long term. But from a mass production need to date, it's been used on [indiscernible].

Great. Okay. Switching gears a bit. Of the 350 customers you mentioned, can you just clarify that 350 analytics customers?

That's right.

Okay. And can you give us a rough breakdown of those between fabs versus systems versus product customers or whatever it might be?

I think the last breakdown is probably the first level of equipment company versus.

Equipment companies when we did the metric acquisition, we mentioned it just over -- equipment companies. Thank you, [ Sonia. ] Equipment companies, when we did the acquisition for Cimetrix at that time, we had mentioned that there were around 220-ish customers. So the balance is kind of the analytics to metrics customers, if you will, and then within that comment.

And I don't have to break down that button. But even there is even some overlap.

Look, I mean, the other way to think about it is one of the slides that you presented, which had the system and the other players in the market and kind of arrow that goes up. You don't see back in the slide and you go back to it. But when we think about our presence, it's pretty strong in the equipment space. Then it's in the product is space, it's also pretty strong. But as we start to get into system, that's where we have some early wins, but there's opportunity there as well that we're pretty excited about. It's probably the best.

Okay. Great. And then finally, the recent round of restrictions -- trade restrictions, that is, any impact to your machinery or software that you're aware of at this point?

Look, we will always be compliant with all the laws and regulations, and our counsel is standing in the room. We always work together to make sure that, that is the case. Even China and of itself has tended to be a smaller portion of our overall revenue. It's right around the mid-teens. It's been that way and it stayed consistent. Sometimes when it will go high, it will be because we had the higher venture come in. . From where we sit today, we're not seeing restrictions that would guide us to think about any different metrics than what we have shared with you today in terms of revenue growth looking forward. So feeling pretty good about where we are and how we're handling lines in insuring it with all those.

You never know what [indiscernible]. But our exposure to China has [indiscernible], for example, and in their last call was somewhere north of 40%. I think most of our software products are not being restricted, based on the country of origin. We are fortunate that we have started our Shanghai office in order to have a consistent back-end and analysis and build people of in the capability of using the tool that we can serve the China market out of our China office. So the restriction on U.S. is spring call folds, et cetera, we're probably less sensitive to than some of the equipment manufacturers was publicized in one restriction. So yes, there's always a risk, because we don't know what the rules have changed tomorrow. But where we stand today, I would say we're probably more stable than many of the marketplaces in the industry.