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

Okay. So we're live. So I want to thank everyone for either attending on the webcast or coming in person. We do really appreciate that. And also for those of you that flew in from out of town or from -- drove down from the city or whatever, we do really appreciate any level of commute that you maybe know it's not easy to get around these days. And I think some of you came from the East Coast and maybe you -- where from, so that's a positive side of that. We're going to go through our Analyst Day briefing. We do this every 2 years. So the last one was 2023. And I will start it off. Adnan will pick it up, probably go over the parts of the slides that you really want to see anyway. And then we'll get back into Q&A, and Kimon, Adnan and I will answer. So basically, themes for today. Hopefully, you thought if you're attending here this morning, and I think it's relevant to both our investors and our customers. The semiconductor industry is evolving into this 3D era, I call it. I mean -- and we are looking to be a $1.1 trillion industry by 2030, if you look at the semi numbers or handheld numbers of IDS. And we think when you look at that, that means the industry does need a unified analytics platform. You probably saw with the discussions we had in the meetings today about what we're doing with secureWISE and generally the way that we're looking at the orchestration products. Our mental model many years ago, we started on this idea of putting Exensio in the cloud first around 2019, we first started attending -- bringing that out to customers. At that time, I thought was once you get the system in the cloud, you want to then make it usable by everybody else. And if you noticed, we hired our first partnership person, Bill Lee, I think, is in the audience, and to start our partnership program in 2020 because it's like, okay, once you've gone through the trouble of getting your manufacturing data in the cloud, then you want to start using it with other systems. And we then evolved that to then think about how do you then access the tools that you operate on. That was why we acquired Cimetrix, why we started thinking about the partnerships with folks like Advantest and Teradyne and how you actually can work with the tools first in test, but then more broadly across the entire test platform. And then with secureWISE, what we realized was you could have great stuff going on in the fab and at the floor and you got great stuff in the cloud, but if these 2 can't talk, then you can't get stuff done. And secureWISE really was another piece. And our mental model in all this is that eventually, PDF becomes a platform for the industry. And every customer has their own instance and their own capability, but it's connected to every other. And you saw that in Aziz' presentation, if you attended this morning, where their vision about why they went with Exensio and PDF was because they know a lot of the fabless customers they covet also use it. And they could provide a better level of connectivity between themselves and their customers if they can both communicate on Exensio. And that's really getting us to why we think the reaffirmation that the industry does need a unified analytics platform. And today, we are the largest independent analytics platform business out there, not attached to something else. And we think we're growing in industry relevance. So these are the themes for today. We'll talk about what that means for our stakeholders a little bit later. These numbers are from [ IBS and Handel. ] This shows about a 10% CAGR. This morning, I showed at 12% if you go back to a little different time period. But in the end, whatever it is, it does mean about a $1.1 trillion business, where 67% of all silicon really is related to AI, but not just AI in the cloud, greatly in the edge. And that's where most of the growth actually is. And when we see this with our customers, it is a tale of 2 worlds. Our customers that are not as exposed to what's going on in AI are growing less fast than our customers that are. Tonight, the fireside chat speaker is going to be Tom Caulfield, who is a great speaker and a great friend. And Tom pointed out -- like I've heard Tom point out a number of times, it took our industry about 60 years to achieve a $500 billion in scale. And in the next decade, we will achieve the next $500 billion of scale, right? And that's pretty staggering. And it gets, as I said this morning, through 3 different -- really big different things that we think have to come in. One is innovations in 3D; a much, much more complex supply chain; and leveraging AI to drive efficiency in manufacturing. 3D, again, as I said this morning, it's important to remember this, it is not just in the packaging world, it is also very much in the front-end factories. I know we talked a little bit about wafer-to-wafer stacking for backside power. But also if you look at 3D NAND, 3D DRAM, everything is looking at more applications for wafer stacking in the back of the room here is SW Sun, who's also a good friend of ours. And SW really what he did at XMC was really the first wafer-to-wafer stacking on vertical NAND. And I think that innovation is happening across the industry. That really does mean the way you think about manufacturing is changing greatly. These 2 drivers are a big part of the drivers of our innovation on what we do on characterization that includes the e-beam capability, what we do on data feed forward and the analytics capability because you have a lot more sharing across the production flow. I think our industry always felt like we were not appreciated kind of the roving danger field of the world, the chip industry was never appreciated that much, right? And now we are strategic. And what does that mean? That meant governments started investing in facilities and factories. It also means that governments have a lot to say about the supply chain. One of the things that we see is increasingly important for our customers is the way we manage and handle data and data governance. Most of our customers sell around the world, most of our customers have people around the world, and most of our customers have operations around the world. That means how we store and protect their data and what protections we have on individuals for the European engineers or employees is different than their Japanese ones or their American ones or their Asian ones in China and also data governance. When you look at secureWISE and providing a network, data coming off tools in China often has to stay resident in China on that network, even though it's coming -- it's being shared between an OEM and a factory. And this is, increasingly for us, we think, something where PDF is building a capability, and more and more important, a point of differentiation. Lastly, when you look at what's going on, because we're looking at building fabs around the world, it really is a huge step backwards from what the industry was doing. For the last 60 years, we concentrated production in a smaller number of sites, and that gave you the most economic leverage. And then for a variety of reasons, access to markets, as I said, access to human capital, access to energy, factories are being built around the world. And it's not cost effective, I think. When this first started happening, the founder of TSMC, Morris Chang, was quite vocal about the fact that it won't work. The economics will be terrible. And that's in part because capital costs, labor costs are higher, but it's also just because it goes against what we've all known about manufacturing, which is the learning curve, the person with the biggest volume goes down the learning curve fast, faster and first and with lower cost has the best ability to continue to compete. And our industry has lived well in that way. And now we're asking our industry to do something very different. There is AI, though, and there are ways now to get a lot more information out of the same amount of data than you had in the past. And we think this is super important for our customers because of the talent shortage, because of the ability -- the requirement to have smaller facilities located around the world and because of just the complexity, the systemic complexity of what they're trying to build. When I look at the room, I know some of you have been shareholders for quite a while. So really, for some of you and for certainly us, we start out really building on that first pillar, what you do for characterization. That started with our test vehicle for the characterization vehicle infrastructure. And we monetize that, if you remember on the gainshare business model, we would deploy the system and then charge a royalty based on the yield the customer ended up achieving. We've repurposed that capability and expanded it to include the electron beam in-line ability to measure product directly and put that as part of a subscription, much like the rest of the infrastructure. And we expanded it starting around the 2010s to include Exensio and the ability to start doing analytics for fabless customers and the product groups inside the IDMs. So if you look at our largest customer, part of the customer base, our 2 largest components of our customer base are the equipment vendors themselves and the fabless and system companies and the product groups inside the IDMs. So I think lots of folks always think about PDF for manufacturing, and we have a lot of fab customers and most of the fabs use our systems. But there are not that many fab owners in the world. The majority of our customer base is actually the vendors that supply equipment to them and the consumers of materials produced at their factories. Recently, starting around 2021, as we started getting the cloud systems up and the partnership started building, Tim and worked very closely with SAP in those early 2, 3 years to just to find out how could you start bringing a real-time view of manufacturing into the hands of the finance teams. How do you have a more accurate understanding of costs? When you have an RNA, how can you react more quickly and more thoughtfully? And this is really the Sapience manufacturing hub, which we then blossomed into this portfolio of supply chain orchestration. I think earlier today, they made reference to Sapience supply chain hub, manufacturing hub. And we will be demonstrating Supply Chain Hub, which is the newest member of that category. I think over in the room Nick to our left -- to my left, to your right. And there's some presentations about that tomorrow. But we recognize that orchestration goes on not just within the customer's organization, but across the industry. And the secureWISE network is enabling us to make that possible. Our customer base spans the ecosystem. I think what's very unusual about PDF, a lot of things that are unusual. One of them is just the breadth of customers. I go back to when we were a yield ramp-focused customer and supplier in 2009, I remember meeting with then Chief Procurement Officer and Chief Information Officer of TSMC, Dr. Stephen Tso. And he says to me, John, what's going to happen in your company? You have 3 customers left in the world. You have Samsung, us and Intel. That's it. 3 customers. Who else is going to buy from you guys? And I thought about that quite a bit. We thought about that a lot. And if you look actually from 2009 to 2014, we made almost no acquisitions. We really just thought about what did that mean. What we realized is you concentrated manufacturing capacity, you've had an explosion of the number of equipment vendors that were out there supplying them and the number of product companies that are using it. And I think when the slides were shown this morning, you saw the talk around application-specific silicon and more and more application-specific silicon. And we thought everyone is ultimately going to be responsible for yield and manufacturing. And we could really look at this as a different way. If we changed our products around, we could really think about the market more broadly. And today, you see the outcome from that. We really span the entire ecosystem because everyone is responsible for manufacturing these days. The equipment is so complex. Without the help of equipment vendors, the factories couldn't use them. The designs are so complex that just to understand the test programs and what they mean, the fabless seem to be intimately involved with what's going on in the front-end fabs and the test floors and up and down the supply chain. We really appreciate the feedback we've gotten from customers this morning. Mike Campbell talked about that a little bit. But over the years, at our different user conferences, we've had different customers say different things about us and kind of put these quotes here. But what we try to be as a trusted partner for our customers, one of the slides I showed in 2023 was the last 25 years of changes in compute environment from Linux systems and Windows and Linux systems and CPUs to GPUs. And then what do we do to help customers move their workflows forward while the underlying IT layer changes. And when you think about these manufacturing assets that are being built around the world, these fabs will last 30 years. For a long time, what we did in our industry is we built the fab, we installed the software and then you change nothing, change nothing on the software, run the same MES system, run the same software for as long as you can. And that's like saying your brain is never going to get smarter. What AI has done for our customer base has got them to start thinking, well, wait a minute, I could get smarter over time. I could use analytics to make -- so I can get more out of this factory to run this factory differently. So this trust is really important to us because to us, earning their trust means that we can begin the dialogue of how we can help them use software in more innovative and sophisticated ways to get more value out of the assets that they own. And that was kind of why, at the end of my talk this morning, the call to action was around, let's look at all the waste that's still in this industry and how do we help drive a lot more opportunity out of that waste. As I said earlier, we began partnering around 2020. If you look before that, we didn't really see partnerships as being all that important, again, because when the focus was just bringing up the yield ramp, there was really no time to work on a partnership. As we started moving customers to the cloud, we just thought more and more ways that partnerships were important. And this is a subset of partner with. We continually look for new ways to expand the partnerships. Kimon is a super big proponent of this inside the company that even if you say, well, don't we compete with this company on this or that, the most important thing is to deliver more value to our customers through partnerships, which is also why we're very focused on standards. And today, when you look at the company and the scale that we've achieved, while we're, let's say, $200 million business today, the scale of our -- and our reach in the industry is quite broad. We manage quite a lot of data. We transmit exabytes of data across the secureWISE network. And we start -- and we provide enterprise capability on what we do on the cloud and more and more what we're doing with standards bodies. I showed here SOX 2 and ISO 2701, but we also are supporting NIST2 and CRA and many of the European standards as well for our European customers. And this is a growing piece of what we're doing with this whole platform. I was meeting with one of the secureWISE customers that's a large equipment vendor. And they brought their CISO to the meeting. I thought it was going to be interesting. I wonder how -- what tough questions he's going to ask. And the person presented what we're doing on standards around the world for secureWISE and Exensio Cloud. And the guys stopped and said, wait a minute. So if we're using you guys for these things, I don't -- I can just refer to you when I go get my CRA certification in the EU. I don't need to go and worry about that. And the answer was yes. And this is increasingly an important point of differentiation for our customers. We are making sure our customers can trust our systems out to the edge when we talked about the fabless work and the encryption out to the edge, and they can understand that we're going to stay on top of the standards that are required so they're able to take advantage of the systems. And it greatly reduces their need to work on this stuff. So I want to provide a little bit of context when we go through TAM numbers for us and how we think about the business. So I talked about a $1.1 billion semiconductor industry. And if you would go back to 2019 or 2018, we really didn't serve the capital equipment industry. So our market opportunity was the manufacturing portion of the IC industry. And if you say the IC industry is roughly a 50% gross margin industry, then you say at a $500 billion size in 2020 or so, there was about a $250 million manufacturing business where software could add value. And then we add on top of that the capital equipment business because that's increasingly a customer base for us. Today, if you look at -- you kind of project forward to 2030, you take the portion of that $1.1 billion industry that is manufacturing oriented and add on to it the capital equipment piece. We think this is the customer base that we sell into. That is not our market opportunity. That's just the customer base we sell into. We think the addressable market for us, which has doubled over the last 4 years, will again more than double over the next 7 years. As software for manufacturing becomes increasingly important, the manufacturing volumes go up -- to the right-hand side of this chart -- and the need for more and more complexity and analytics continues to go up. When you look at how we've grown our business kind of -- that's why we put the left-hand side of this chart up. It doesn't mean that we bring the products to market the same way or the same capability. What's core and always been core in PDF was taking a system view to manufacturing to understand the design aspects, the process technology aspects, the manufacturing and the equipment aspects and how those integrate together. That was, as I just said this morning, PDF stood for portable document format. But what we show here is the way that we've layered on more and more capability reusing and then from a combination of reuse, development and acquisition to create new ways of helping customers capture the value in manufacturing. So today, still important is our characterization vehicle capability and the ability to characterize next-generation nodes. That continues to be important. We've enhanced that with what we've done on the DirectScan program. Layering on top of that, starting in the 2010, was really the way of having Exensio combine process control information, testing operations, offline manufacturing analytics to enable the customers to really look end-to-end at their data. That was really the next foray. Starting around 2019, customers started migrating that to the cloud, at least for the central and offline analysis capabilities. And that continues to be a very important piece of our business. Around 2020, we started adding the ability to give tools and capabilities to the equipment vendors. So the equipment vendors could be able to add more value to their customers, provide more AI-driven solutions, more analytics capabilities. And with that, we started recognizing the need on supply chain orchestration. For a couple of reasons. As I said, most of our customers can't deliver value without crossing organizational boundaries. They need to get from their organization to their supplier or their equipment supplier to their customer. So orchestration increasingly became important. We see that as a significant growth vector for us over the next 5 years. And then maybe it was lost on some folks when Said presented. But when you think about -- and we've been in the AI business for many years, in machine learning and analytics. When you think about how you look at large language models, agentic flows and how you bring that to manufacturing, guardrails are really super important. What -- as Said pointed out this morning, what is hallucination is also imagination. I think my wife is a super linear thinker, I'm kind of a scatterbrained kind of guy. I think I get more creativity of being kind of like all over the place sometimes because you see connections between things. And when you let things hallucinate, you get that kind of creativity stuff sometimes wrong. When you kind of go on a very linear flow, you kind of get repeatability, which is what manufacturing is all about. What we're doing with orchestrations and why orchestrations and workflows are so important in the next generation of our products is really around enabling customers to turn that dial between creativity and predictability. And that's why we see the next phase for what we're bringing to the customer base, really an AI-driven manufacturing flow and why we think it's so important for our customers and as a result, hopefully for our business. So that kind of gives you the update of how we're thinking about the business strategy and what's going on. Now I'm going to turn it over to Adnan, who's going to talk about the part that you really all sat here for anyway. So I was kind of like -- that was the preacher before the [ cookies ] at church. Thanks, Adnan.

Thanks, John. So when I was thinking about the transition about how to come in at this point into the presentation, yesterday, I had a different idea. But this morning, for those of you that heard John's talk, he talked about how parents always evaluate you on an integral function basis. So the point of this was to present a 10-year view to the investors. Yes, it's sort of a report card. I understand investors' past performance is not indicative of future results, but everybody, even after those statements, will ask what have you done for me lately. So this is a perspective on what we have done over the last 10 years. And really, if you think about it, kind of years 6 through 10 going backwards was where the transition was happening for us to the analytics space. If you recall back in 2019-2020 time frame is when we started to break out with our 2019 Analyst Day, the breakout of revenue into analytics and IYR. It made sense for them. Today, we'll present some other perspectives as well later in this discussion, so we'll go there. What has happened over the last 5 years is this growth. If you go back and think about the Analyst Day that we did in 2023, there were 3 long-term target in respect of the market target growth rates or target model metrics that we had set for us. One was on the revenue growth rate about 20%. This is 19.5%. Kimon and I had lots of discussion about it doesn't round to 20%, we left it at 19.5%. That's one metric here. But look, a lot of this is not just a function of, okay, industry happened. Yes, there are things happened along the way. For example, in 2020, we acquired Cimetrix. That was at the tail end of 2020. So initially, when we posted the 2021 numbers, of course, the curiosity was, well, you posted this 26% growth rate in 2021, but it was helped by Cimetrix. Acknowledged. But the following year, 2022, again, against the backdrop of a strong industry, we were able to post 34% purely organic growth rates. And along the way, why this has happened is because we've done some things. I mean John and Kimon and I have a relationship with who's who of the people in this industry. You've seen the names mentioned, you've seen people talk. I think one of the things that we did was we came together as a team, especially as our number of customers expanded. And in 2020, started to think how do we take a relook at our bookings and how we approach customers. So I remember we formed this exact booking meeting that we started to do on Monday looking at who are the key customers, where we start, who can unlock the ball and kind of move that deal forward. And between a few of us, legal, finance, product as well as John and Kimon relationships, we really made a difference in terms of the booking rates. If you go back, you'll see a big jump there. So that's our report card through 2024. 2025, as you know, we've guided kind of and reaffirmed in the press release today, the growth rates that we had talked about earlier in the earnings call as well of 21% to 23% growth rate with the acquisition, of course, of the secureWISE that we did in this part of the -- earlier part of this year as well. So within that, if you zoom in and say, okay, from 2020, and this one -- last slide was through 2024, and this one is through trailing 12 months of the Q3 of this year, it's been a 20% CAGR during this time frame. But again, within that, what was happening was we were going away from IYR and moving towards the analytics. 2019 is the time where we started to offer some of the cloud solutions. So typical 3-year deals, 2022 is when those came up for renewals. You've heard us talking, for example, Q3 call last year about how some of the customers' usage rates we were able to expand as those deals have renewed. So what's been driving within the total growth of the 20% has really been the analytics growth at a much faster CAGR at 29%. You can see the numbers on the bottom right as far as how analytics has contributed. And frankly, why we're thinking of another different cut back to [ 2020 -- 2019 ] when we presented the Analytics IYR, the business has transitioned to now 94-ish percent on the analytics. So that was some of our thinking why we present to you a couple of different cuts today as well. The other 2 metrics of the 3, one was revenue growth that I talked about earlier. But the other 2 metrics that we talked to you about in 2023 Analyst Day were gross margin kind of the horizontal lines of these charts. 2023 targets were 75% and then the operating margin target for -- operating margin target of 20%. And if you look at our progression over those years, it's pretty remarkable where we have come from. The non-GAAP gross margins on the left side from the low 60s to now in the middle of that -- slightly exceeding that range at 76%, which is why we're here at the Analyst Day. So to John's point, we'll see some of the cookies for later. And then, of course, on the operating margin side, we're just kind of inching up to it, but getting there where it makes sense to start talking about some of the new things that we will today. We're pretty pleased with this transition. And actually, if you think even more on the operating margin side, the fact that we've gone from a negative all the way to the plus 20% range. Okay. So this is one perspective we wanted to share with you on revenue. Historically, like I said, we used to report it as analytics versus IYR. To us, as you heard the word platform ubiquitously today. To us, it made sense to start looking at the revenue on a platform revenue breakout versus the volume. And I'll explain that in a second. The left top chart, we put it as reference. Really, to me, the interesting part is the annual and the kind of TTM patterns, if you will. And let me explain the characteristics of the right side in a second. But if you think about it, right, total growth rate from $149-ish million to the $207 million. But within that, what's interesting is the platform growth rate that has gone from $120 million to $170 million or so. So what is platform? Platform is everything that you know of and classically expect within PDF. It's a SaaS software that we sell. It's the term-based licenses, it's perpetual, as well as the DFI and CV systems, right? The reason -- when we separate out the volume-based revenue and that we include 3 things: the Cimetrix runtime licenses and secureWISE and gainshare. To us, the reason why we call it volume-based is it's directly tied to the customer shipments or production output. For example, Cimetrix runtime licenses. We go through a validation cycle, as John talked about earlier, of the software development kit, having the customer validate the software for their tool shipments. But really, the revenue on the Cimetrix runtime license is going to come when the customer is going to ship their equipment. Similarly, on the secureWISE, the data component of it, if the customer is going to use more data, that's what's going to enroll to our benefit. Lastly, similarly on the gainshare side, we do the fixed fee part of the work, and then we benefit when the customer will do gainshare. Now the interesting part about all these 3 is these 3 are high-margin contributions to our business. Really, if you think about it in all 3 cases, we have done the work in the platform side as well to get to enjoy that. If you think about the Cimetrix run time, we did the software development kit work, which we think of as formation as part of our platform revenue. secureWISE has a software piece that provides the connectivity that's also in our platform revenue piece. Same thing for gainshare. We do the fixed fee work with the customer, and that forms part of our platform revenue, which will then benefit us as the customer produces higher and higher volumes in the volume-based revenue side. This is another cut. A lot of you have been asking us about recurring versus upfront revenue, and I'll describe it in a second. Again, quarterly data on the left top side, annual data on the bottom left side. But really what's interesting to me is the first bar, which is a recurring portion of the revenue of $140 million and then the last bar, which is $180 million. When I explain on the right side, I'll explain why. So let's just get there. On the recurring revenue side, again, like investors would know and expect software companies to include, we obviously have the SaaS piece, the software piece, maintenance and services that we have as well as well as the DFI and CV systems and we sell them on a subscription basis. We also, of course, have the Cimetrix software pieces, the Cimetrix business itself as well as secureWISE and gainshare. Legal would be happy and wanted me to say this, I will say this. And within our recurring, those last 3 pieces, the Cimetrix, the secureWISE and the gainshare are -- you call them reoccurring businesses. Similar to the thinking that we read through the last slide, runtime licenses and the customers will ship more will get benefit more of that. And similarly for secureWISE as well as gainshare. But to us, these are businesses that will reoccur. And if you look at a little bit of a wider time horizon, they start to look like recurring businesses. We won't expect them to go to 0. Sure. In between some years, there might be some ups and downs, and that will cause the way we are measuring recurring revenue to have some different pattern. But over the long term, hence, I point out to the starting and the ending points on the left side, they will have a pattern and a characteristic that we like. If you go back and do the math, you will see that it's about 80% of our total revenue is of a recurring nature. So today, $207 million LTM basis on a revenue company doing about $180-plus million in recurring revenue. In the upfront revenue section, just to highlight, it's on the slide, but again, no surprise what you would expect. It's the 3 pieces of perpetual software. And when we do a CapEx sale with the AI systems, which we have done, and then also certain of the IP licenses where we'll use some of the technology and license it to the customers. So the other question that everybody has asked us, especially as we've taken out debt, especially as we've increased our CapEx rate is what is the return on the DFI business? Obviously, we've got 4 machines out there with 3 customers. And we've talked about how a subscription model is useful. So we thought it would be interesting to share some perspective as we look at our current deals and as we look to our future deals that we're engaging with customers on as to what sort of a return on the machine cost can we expect over the lifetime of the machine. And looking at it a few different ways, actually, we did even what our pricing would be versus what the customer might push us to. And we looked at those 2 different ways, and we're coming back with a number of saying, looking at our machine cost investment, we believe there is a 5-plus x return on that investment over the lifetime, which, to us, is pretty compelling and is exactly the reason why we have invested and continue to invest in that business. Of course, recently, we managed the CapEx, as you saw in the most recent quarter as well. And candidly, the machine subscriptions are a strategic shift. If you think about it, this is an industry, as you all know in this room, that is used to getting these machines on a CapEx basis. But to us, there was many benefits that made sense why subscription should happen. It's provides the customers the ability to upgrade the machine a few years in. We've already had some occurrences of that as we've talked about in the earnings calls as well as for us, as a software company, it makes our revenue be more of a recurring nature, which is exactly what we like as well. And people said, okay, this transition is going to be hard. I remember John making a comment that just kind of clicked to me at that time, was AWS. Of course, they're selling you hardware, but it is on a subscription basis. Behind that, they have hardware. So it's no different than that. Yet, it still requires a transition analogous to what the cloud industry went to where people used to have on-prem hardware or used to buy their own hardware and then eventually got comfortable today that we are seeing massive amounts of investments in the CapEx space. So what does all that mean? Most recent quarter, if you go and look, we talked about a backlog number of $290-plus million. And coming from the 2020 number of $100-plus million, it's a pretty good growth rate. Remember, during this time period, I also talked about how revenue had grown at about 20%. But what we like to see is, sure, there might be patterns that might -- maybe we don't like it from a quarter-to-quarter. But over the long term, we generally like our backlog to be growing faster than our revenue. That, to us, is what creates a long-term sustainable business. It's also a testament to the relationship that you're building with the customers. They're willing to sign up larger long-term deals with you. I'm pretty pleased about where that stands today. Global geographic revenue distribution, I think this is all numbers everybody can read. But if you look kind of the last 3 years, I think one pattern that's interesting to note is if you look at the bottom 2 colors, which is the U.S. and Japan, they were about 60%. And if you look at kind of what's happened over the last year, 2024, and even the trailing 12 months, we've managed to distribute that. So it's a little bit not as much a percentage as it used to be in the U.S., but it's more like 40-20 between U.S. and Japan. Obviously, people are -- people have seen our engagement with some of the key Japanese customers as well as our eProbe shipments that have been talked about. So it's been good to see that growth happening, not just in the U.S., but also in other parts of the world. This is a statistic slide you all know, but I can kind of require ourselves, it's important to put. Left part of slide, we talked about over the last 5-ish years or so, we spent about $272 million. If you break that up and think about what categories, largely speaking, 2/3 of it was spent on acquisitions. Obviously, we did the Cimetrix acquisition, which was a smaller one at about $35 million size. And then most recently, we did the secureWISE acquisition, which was $130 plus about 2/3 of that total on the M&A and investments. Of the remaining 1/3, it's also about a 2/3 split in CapEx and the remaining portion of about the 1/3 roughly, we spent on share repurchases. Balance sheet, on the right side, John has made the comment in the last earnings call that as we progress through the next year, we expect our cash position to improve. Doubling on that, as we pay down the debt because part of it has the amortization of -- amortized debt payment, that should create some expansion on a net cash level for us as well. All right. Last slide. This is probably the cookie slide. Let's walk through it a little bit. The first column is the prior targets that we had set in the 2023 Analyst Day. Today, we are revising those targets and putting in new long-term targets for our company. Obviously, much discussion, much debate, much planning that goes into it. But let's just walk through it. Look, I think John has talked about -- let's go step by step. John has talked about the industry growing at 10%. If you go back and even look at our history, we've grown ahead of the industry. That's where we aspire to be. And that's why the bottom comment is there that if the industry is projected to grow at about 10%, we would expect that we would be able to deliver a 20% revenue growth rate for our business as a long-term target model. That's where we felt comfortable. On the gross margin side, we're upping it to 77%. And the number there was chosen carefully in terms of greater than 77%. And some of you could go back and say, okay, you've come close to that number a few slides ago, we've talked about 75%, even 76-ish percent. We think there is opportunity there, but we wanted to be careful and commit to you a 77-plus percent number and hopefully kind of go from there against that metric. Where we're doing the bigger change is on the operating margin side, 20% going to greater than 27%. For us, as we're starting to see scale, we are starting to see some of that leverage benefit the bottom line. If you look at the last quarter, at 23% operating margin, it's something that felt good to us going from the negatives that we had 5 years ago. And as we look forward to the future, we believe that whether it's G&A or whether it's S&M, we can have the scale advantages come through while still allowing us the dollars to invest in the R&D line to manage that spend and grow it to serve the opportunities. You heard Said talk about the 100-plus people that he's had working on the new Exensio platform, so excited about that. So I think in terms of targets, those are the new ones to take and pretty excited about what that does. And of course, with that, what that benefits the rest of the financial statements as well over the years to come. With that, and the last cookie, I think we will open it up to questions.

Mics, correct? Mics in the back. Do we have a microphone for folks?

We have one here.

We got 2 mics.

Yes.

Are there any questions that folks want to ask in the room? Sure, go ahead.

So you talked about the subscription model and the lifetime revenue increases by 5x. Why can't I just purchase...

No. What we're looking at is if we look at what we spend to build a machine, what's the lifetime return on the machine? And what we -- when we did all this analysis, we did this analysis looking at both a customer buying the machine and depreciating over 7 years, a customer subscribing to the machine. And we set up the economics where actually the economics over a 7-year time period is cheaper to the customer than the customer purchasing the machine. And you look at the 7-year cost of ownership, and of course, they've got to put the capital upfront versus the machine cost over time. We are -- that's greater than 5x is in variant whether the customers purchase the machine or subscribe the machine. They're invariant whether the customers do. We make it a little bit more economically advantageous to them if they subscribe because we think we can deliver better value over the lifetime. But frankly, we have some customers who buy them. We have some customers that subscribe them. I think he was a little bit off on the number of machines in the field. I think the number is more like 6 or 7 actually, but over a few sites. But in the end, that 5x is just a return on the capital and you factor not just the machine, but the software drives, the services it drives, all the value that gets created off that machine spend.

I just wanted to ask you, this morning, there was quite a few product enhancements product releases announced for 2026, mostly coming GA in '26. What sort of has you most excited with these? And what do you think has the most near-term impact on the?

I love all my children. But the one that I'm very, very excited about is what we're doing on scalable analytics. I don't think anyone in the data analytics business, not just in semiconductors, really rethought what you need to do to make it so you can start operating on super wide data sets and super huge data sets. And was it back in 2019, Said, Kimon and I talked about this as we were doing the next plan for the next decade of Exensio. We said, okay, we're going to keep going with a classic BI tool on top of Cassandra through the remainder of this decade, the coming decade, but we think we need to start thinking about what's beyond that. And so we challenged Said, and around 2023, they kicked off this effort around what could you really do? If you just thought about -- we had put Spark, which is a parallel compute layer on top of Cassandra back in 2017. And it worked okay for batch jobs, but you couldn't make it interactive because there's no persistence of what you were doing interactively. And the way you still the data in the database to get the efficiency and speed that Aziz was talking about isn't the way you want to analyze it. So starting in 2023, they're like, wait a minute, what if we pull the data out, put it into another data frame a little bit different, Python-centric and operate in parallel with the data and give the user interactive experience. Like why bring 20 million data points to a customer to put up on a screen that's 4 million pixels. You're not going to see it anyway, right? So that's kind of crazy. So they started rethinking all of that. And by the end of '23, early '24, we realized that what would work. And they then kicked off this effort. And it's been quite enough as he said, over 100, I think he needs to understand the number because he doesn't want -- I have to get too upset on how much we put on that thing. I think that really has a way of just changing the way analytics is done. I think it is the first system out there in the world that really has this third layer. It's not just a BI tool attached to a database. It is really this scalable compute layer. I think it will do much like what moving programs to parallel computes like GPUs and other hardware environments, it gives that same ability to scale. I think it's super valuable. Question in the back.

All right. Following up on that, can you maybe put it in the context of the share of wallet opportunity that it gives you relative to what you've had before with some of your larger customers if you combine what they can do with scalable analytics, Sapience, consumption? I mean, how has that changed your kind of per-customer level revenue opportunity?

I think it's early just to figure that all out. But what I will say is the following. I mean you saw Aziz talk this morning. One of the super frustrations for us providing Exensio to the marketplace was always -- they always had internal storage anyway. He talked about the systems they built around Teradata and Exadata. And there's tremendous cost in those systems and not -- and in the end, they pull the data into Exensio for the analytics and capabilities. So we really want to be able to go back and say, we could build an enterprise-class system they could use broadly. And their commitment to Exensio that they signed earlier this year, as I said, it really is the largest -- he said it in the meeting today, the largest single deployment in Exensio out there. When it gets to full scale, it will be substantially larger. And it is also -- if you look at what they're -- they also will talk today or tomorrow about using secureWISE as their standard for communicating between their edge and their fabs and their equipment suppliers as well as their fabless customers. They talked about the way that bridges to -- between their data and customers, fabless customers that are using Exensio so they could communicate with them as well. And they also talked about what they're doing about connecting Exensio manufacturing data to their enterprise through Sapience capability. Overall, that is the single largest software deployment we have in the world by a substantial amount. And we think they are not -- I mean, they are unique in the scale that they have, but there's a lot of other customers that we see spend really significant dollars trying to build these central data stores that we think Exensio and scalable analytics would be a lot more valuable to that. A big factor in Intel making that choice was scalable analytics. They've been working with us on this over the last 9 months, and I think very excited about what that could do. So we think it's quite substantial for really taking our analytics to the next stage for our business. But we don't have enough data points that I could go and say it's going to give you an X.

But I think what you can say is we've expanded from being predominantly a company that supports engineering analysis to being more relevant to financial operations, but also engineering operations with the new tools and making that data available not within the -- just within the organization, but the ability to integrate across the organization. So we're seeing interest from IT groups, interest from manufacturing operations and product operations, the work with SAP getting more involved with the financial operations. So it's letting us grow vertically, if you will, within a company from a customer base over time, not just horizontally across the industry in our traditional engineering space. Sorry, I have to speak...

That's okay.

Another follow-up on the same subject. But probably the most impressive thing I heard today was Mike Campbell saying, this is our tool, this is the industry's tool. And like, yes, PDF sells it to us, but it's the industry's tool.

Absolutely.

So it's just a little bit of a disconnect. I feel like you guys are kind of underselling the opportunity here with the 20% number, that it's 20% plus the creativity of the industry to leverage this layer that you guys are supplying to do more stuff and for you guys to come up with a way to monetize more on that network. So why am I wrong about that?

So yes, thank you for the question, Jeff. Yes, we are honored that customers see us as a tool for our industry. We always -- Kimon and I have always seen ourselves as stewards of this business, right? We were the founders -- 2 of the founders. But this business is a business that our shareholders own and that our customers depend on and that our employees depend on. And we are stewards to make it more valuable for, first and foremost, our customers because they're the ones that drive everything. We thought really long and hard around the growth rates. I mean, to be candid with you, when Handel first started talking to me about a $1 billion industry by 2030, I was like, Handel, man, you happen to be around by saying that you're just putting that number out there. It seems farcical to me. And now as I've seen it actually evolve, it's like, oh, maybe this actually can really happen. So we've been -- we didn't want to lean too far on the growth rates because a lot has to happen, not just from our -- from our perspective, but for the industry overall. And we do want to be more important to our customers. We do want to make sure we're adding more value to our customers, but that's like giving the eye of the beholder. They have to tell us that by their commitment and what they do with us. That was why Aziz talked about what they were doing with standardizing and replacing internal systems. So it's a great responsibility that we need to take super seriously. We would like to win that across more and more of the customers and be more and more of the tool for the industry. We're happy if we can outgrow another. We're not looking to say that's the ceiling on where we're going.

I mean long term, the belief is there needs to be a standard in the industry for cross collaboration and spin. We're subscale today. Everyone is. You look at it, and it's sort of like the joke from the depression, I can't remember it was perhaps Mark or someone else, how do you go broke? Slowly then all at once. I think the growth in this industry is the same. The growth is going to be slow and then you're going to start seeing a flywheel effect as it goes. When did that point occur? Hard to determine. We think we're well positioned, but it's an opportunity. It's not a guarantee for sure.

And you got to appreciate how Kimon can do a joke they've done after the beginning.

My joke was going to say we're only 20% of because Mike is cheap, but Mike's in the room, so I can't use that.

Clark Wright with D.A. Davidson. You said in your prepared remarks, John, that the majority of growth in the semi industry is really going to come from that AI component. How does PDF look to increase exposure to some of these key players in the manufacturing of AI-related hardware?

Yes, that's a great question, Clark. [ Vicity ] is actually, I think that the next phase of AI growth is a broadening out phase. So we're seeing this -- over the last few years, most of the hyperscalers have become customers of PDF. I think not just relying on GPUs, but also building out special purpose systems like the TPUs and other capabilities. Our customers -- I think I saw earlier today, one of our users from Sensient, Peter [indiscernible] ,right. And they are building an edge processor for audio. That's when you talking the Google Alexa devices, it's the one that's processing. AI is going to start showing up in many -- our MCU customers are all looking at it. Edge devices are all looking at it. I think one of our customers was talking about their success in AR/VR as well as the Google guys, the Qualcomm guys on Snapdragon and AR/VR and what's going on there. So we expect it to be a much more broader part of the business. Part of our strategy on all of this is we serve the entire industry. We're not here to pick who's going to be successful or not. We sell to all of them. Some of them, they realize beyond what their expectations are in their business, and we grow with them and some don't, and we will participate broadly. But we don't -- we do look at things like, okay, what are we doing on advanced memories. This is an area that talked about, what are we doing on co-package optics. Each of these things we look and say, what do we need to make sure our capability is good. But then within that category, we just try to sell to everybody and whoever awareness that they win, it's going to be on their brilliance, and we're just going to be a participant of it.

Got it. And if we think about unlocking additional budget, can you help kind of describe that process? Because if I think about it in my conversations I have with customers today, some of the in-house solutions that are already existing that companies use to solve some of these analytics problems, where do you see the ability for you to displace some of those and continue to gainshare over time?

Yes. Aziz touched on that in his talk. And I think when Kimon and I were to talk about like our frustrations over the years, we would provide Exensio, primarily engineers. We've always sold to engineers and product and test engineering teams. And then we have customers, they would hire a Chief AI Officer, Chief Data Officer, and they start setting up a data lake system, right? And then they would spend $100 million between software suppliers and the system integrators and build the system, and it doesn't deal with the edge and all the action in our business is what goes on the edge. So we felt like we were communicating with the CIOs and the IT organizations, the finance organizations, the point that Kimon was making. And a big part of the reason why Kimon was so passionate about the partnership with SAP was it got us to start talking to different parts of the organization. I think a lot of what happened at Intel was the finance organization decided to use Sapience and the connection between manufacturing and their business process. So if that's there, why not use Exensio more broadly in manufacturing because they're going to be pulling the data out anyway, it takes out one of the -- now I have to worry about the MES system, right? And if I just use your stuff, heck goes away to support them anymore. It's the same data model. So the broadening out of the platform has helped us get to more touch points in the customer base inside the customer, which is allowing us to maybe be looked at differently. We think we have a lot to prove to the customers to be trusted in that way. So we're not here to say this is like this is a stamp and repeat. We think there's work to do. But that broadening out of the touch points from the customer is really important.

John, actually, for Adnan, just on the margins, 2 questions here. So where do you see the best opportunities over the next couple of years to move that op margin line 7 points. That's pretty significant. And how long is that going to take you? And also, just as in regards to the DFI, which you're still, in some cases, selling out right versus leasing, there's a different margin structure with that margin impact. Is there a bottom end on the margins that you are going to try and protect, I guess, is what I'm thinking about if you have that volatility in the top line?

Yes. I'd like to answer those in the sequence. So look, in terms of target margins, it's long term, and we chose those very carefully. Similar to the last time when we did this presentation in 2023. But if you go back and look at it, say, we have 2 years later and we're almost clicking in, I'm not saying that we're promising anything close to that. But usually, when people think about long-term targets, it's 3 to 5 years on average. So somewhere in that window, near that window is where we would like to be in terms of the time line. In terms of where do you add scale, look, there's naturally some synergies and scale efficiencies that you would expect in the G&A line. That's one. Second, on the sales and marketing spend, you -- many of you have heard us talk about how there is an opportunity that we are working to optimize, and we see opportunities there. I mean, candidly, we brought together the Exensio sales team, the Cimetrix sales team and the secureWISE. And with some of these initiatives that -- some of these discussions that you saw with the presentation from Aziz, you can start to see that there's a need to have one team for a client where you can present a portfolio of solutions rather than 3 different teams. So it's educating the sales team, educating the apps team and trying to create efficiencies there, that is another opportunity that we see. Thirdly, in the R&D side, I think there's opportunity there for us to provide the dollars that is needed for growth for the opportunities that we are seeing and then still be able to drop to the bottom line. So it's a combined effect of those 3 and again, over that time frame.

Yes, I think what's baked in there, if I could just [indiscernible] we anticipate revenue growth. We know we need to scale R&D, and we will continue to scale R&D. But we don't think we need to scale G&A and sales and marketing at the rate that we scale revenue or R&D. If you look at the last few years, we've been very much investing in those areas. We felt like when we switched from being a yield ramp company to a platform subscription company, we were very, very light on sales and marketing. So we needed to invest heavily because we really had 2 sales guys when it was an business PK&K. Yes, it wasn't needed. Well, we needed to change that. That meant a lot of invest. We don't think we need to continue to sustain that investment at the rate that it was growing as faster, in some cases, some years faster than revenue, right? So we think we can meter down that growth rate substantially and let more of that -- more of the growth in revenue go to the bottom line. On the gross margin side, we rest the numbers there in part because we don't know a lot on how the will impact gross margins. We do know when we look at Exensio Cloud Analytics, BI tools waste compute, massively waste compute because they're, by and large, semicon programs. They, by and large, are in-memory database systems. And so many cores get wasted. And what -- the reason why FY is able to show what he's able to show today is because you're using every core, right? You're using every core maximally. So we know for our compute customers, and part of Aziz's excitement when he looked at this was like, wow, you can be a lot more efficient on compute spend, right? So we do think for our cloud customers, there's leverage on the compute side because you just start using all the cores that we're basically wasting. And also by decoupling the parallel compute from the Cassandra node, it gives you a way to get efficiencies on the storage side, too. So there's some leverage there, too, in the way that we thought the product that we think would give us some upside there. But there's all this down headwinds that are out there in the business, too.

Just wanted to talk a little bit deeper on the R&D spend side. I mean you guys have put in some rails either by acquisition or by building that were not easy things to do that took real investment for whoever made that investment you guys are the company acquired. Now you're going to have AI at some point helping with coding quite a bit. I mean, is that rational with what you do? I'm not sure that you guys are tied to coding, but is it going to -- and so it implies that this scaling R&D budget is going to be more towards new products and new capabilities. [indiscernible]

Yes, that's a great point. So I think if you get a chance, go look at the product costing demo that the team is doing, I think it's probably in the room next door, Mike Norga that did that. Our team started working with generative AI techniques on coding. And what they started figuring out was very small teams, a mix of product folks, apps folks and developers and actually create virtual team members. So you have an agent that's writing requirements, an agent that's writing out test elements. You effectively build -- you scale out your team with virtual team members that are doing the same functions you do in a regular software development process, but they're virtual. And you've got a human that's working with them in each of those characteristics. And what he's been able to build, we put him in a small team on product costing because we knew we needed that for some of our Sapience customers by the middle of 2026. And they built stuff out. They got much further along within a quarter, right? And it was really around that innovation. And yes, when we think -- we still expect to grow our R&D spend to grow more than we're growing our G&A and sales and marketing expense. But we think we'll get a significant boost in productivity and the way we develop product because of what we're able to do with AI. And I wouldn't like arguably by coding, but it's really quite a bit different in the way that they're thinking about it. And the team, we did a lot of innovation, a lot of trial and error on that throughout later 2024 and 2025, and the rate of improvement in the tools and systems that we use is just staggering to the team. What they'll tell me is, John, what I was doing -- what I did last week, I couldn't have done 3 weeks ago or 3 months ago, right? But the capability is really improving, and it's really changing the way we're doing product development.

I can attest we're spending on those tools.

John, just one question. So with the subscription model, say, Exensio, you do plan to give regular updates like Tesla, they will upgrade the FLD every year or so. Netflix will give me new movies every week. So that's part of your model and that's included in your development so customers can expect regular updates from you?

Correct. Yes. And actually, that's one of the big differences between what we do on the sale of the eProbe and the subscriptions, the way the software gets smarter, too, on that as well, what Fire can do in looking at layouts and what Exensio can do on the other side of it in terms of identifying things. And the big shift on Exensio Cloud was the ability to make it more efficient to push updates. When we move customers to the cloud, the update got a lot more efficient, and we could get them -- it's super frustrating when you see customers running on the 2-year-old version of the code, you're thinking, well, I could have made the same money just flying everyone in R&D just running it out for the next 2 years. But you're only going to change it that rate. And really, there is a need for that. That's part of my motivation around AI for manufacturing. If we're going to run lots of factories all around the world, we've got to get smarter, continually smarter. And that means continual updates. That's not -- and that was a big piece of the reason why we acquired secureWISE because our equipment vendors want to deliver more value to their customers. And the only way they're going to do that and get a subscription business to be able to continually provide updates as well.

[indiscernible]

The more value given to the customer, the more you see. And they expect -- I mean, very early on, one of our first customers was Toshiba, and a great friend of mine says this guy Kukumo-sano, I know really well. And Kukumo-san was telling me, John, you have to understand, every year, I have to make the chip cheaper. You've got to give me more for the same money. If you want more money, it's got to do even more than that. And we understand that. That is part of the nature of the business. You've got to really be delivered. But it's -- as I showed earlier, it's an opportunity-rich environment. You look at the amount of money we waste as an industry.

Do you have plans to expand into silicon life cycle management and basically, you'll have more opportunity to engage with automotive and health care and data center.

Yes. So the question about silicon life cycle management. So for those of you that aren't aware, so the idea is, and I think it's really clever, putting agents on the chip, keep track of the data of the chip over the life, have the chip phone home, provide that data and enable ability to work with that. One of our partners was Proteintech. If you look on our slide, Proteintech is a leader in that space and continue to work with them. We believe our capability and enablement is very important for that overall capability. The difficult part is getting the information from the chips in the field back to the chip supplier. And we don't know how to solve that for our customers. And if you can't solve that piece for the customers, right, then it's really kind of -- the rest of it is hard for us to envision. So partnering with folks like Proteintech, we're happy to do this. Siemens also some really unique capability from a UNIX company they acquired out of the U.K. maybe 10 years ago. We partnered with them as well. We partnered with analog bits over the years. They have some capability as well. Around the analytics, once the data comes back, we are trying to enable that. But the real value piece is how do you get that data home from the chips in the field, and we haven't seen a way to solve that problem. It's a much -- the great thing about secureWISE is the ecosystem of semiconductor manufacturing, it's hundreds, hundreds of fabs and test facilities and OSATs, [ if not ] millions and billions and trillions. So the phone home capability is contained. We could tackle that for our customers in this space. That silicon life cycle management dream means you got to tackle that problem for the chip industry. It's above our pay grade right now.

I think the core for PDF and the customer base is being that central point of manufacturing data in that center of truth and bringing in if they get field data, as they talked about in the previous session, people running AI models, how do you attack the version of an AI model run on a chip back into the database and perhaps the output. So 2 years later, you can do analysis if you have a field repair. But the agents themselves on the chip, the phone home capability from the Tesla, et cetera, these aren't areas where I think we're going to be providing the expertise in the industry.

Yes. Understood, understood. It's interesting.

Yes. This may be a dumb question. In terms of analytics, are you just enabling your customers to provide their own analytics? Or are you able to offer some analytic solutions yourself? And I ask this question because this is a story issue of data ownership, even though the data is hosted on your platform, you don't necessarily own the data.

Correct.

I mean the customer owns the data. So by definition, customer has to provide the analytics. Do I have that correct?

It both happened. We've analyzed our test vehicle data over the last 30 years. And so we have a fair amount of expertise in analyzing data and building analytics to do different capabilities. And we do make that available to our customers. But you're 100% right, right? The real details, like there was a whole bunch of [indiscernible] Teradyne and us talked about data feed forward. The reality is in the audience for folks that know 10x more than us and our equipment suppliers about data feed forward and what they need to do. Mike kind of touched on that in his talk. And so a lot of the value gets delivered as here's a platform. Here are some starter algorithms that you can work with. And you can then build them, you can take them and use them in many different ways. And 90% of the customers take what we provide that's generic and then take it the next quantum leap ahead of that. And that's why Said talked about flexibility in the platform being so super important in what we do. We don't look to be the like single-stop shop. We're going to tell you how to do your analysis, but we are going to provide you a pretty rich toolkit from which you can build stuff.

I mean, simply put, we provide AI models, another simple -- more simple algorithm analytics. Customers either import or build models within our system, and we also support third parties, right? Our belief is, especially in this AI world, no company is going to have the cornerstone on the best algorithms, the best understanding for every situation. We are the trusted platform we would like to be for that repository of data, but we need to enable the customers to use it and integrate it with other data sources that serve their needs.

Can you guys talk about where you are from a supply chain perspective on the eProbe tools and just sort of the balance between supply and demand? And also maybe touch on the DRAM opportunity.

Sure.So we are ramping up our ability to produce. We expect to be able to double what we did last -- in 2025 and 2026. Most of what we can produce in the first half of the year, I think all of what we can produce in the first half of this year, we think -- we believe we know where they go between revenue-generating machines as well as machines that are [indiscernible] demo machines. With respect to DRAM, we've been working with a couple of vendors on that. We do expect to be able to ship to at least one of them in the first quarter of this year. That's why one of the machines will be an eval capability machine. And yes, we are -- I think we're pretty mindful of not getting over our skis too far on this. We do want to make sure we have -- everything that's out there in the field runs in very high utilization rates. We're very mindful about that because then we know that we can absorb more. And when we put -- when the customers had one machine, we start got a second machine, the second machine gets a 90-plus percent utilization very quickly. So we know that there is untapped demand. We have been providing under that demand, somewhat constrained by manufacturing and somewhat constrained by making sure we start opening up the aperture of customers and market opportunities. We'll get the back first and then [indiscernible] Are we supposed to open up the call -- questions for people on the call now?

I'm going to ask another DFI one, kind of a multipart one. The way you've -- you talked about it today, the direct scan, the combination with Tiber and Exensio, is that added capability? It was new to me. I didn't know how new that combination was. And just more importantly, as you talk about the added throughput that you've achieved, how do we think about what the SAM is at this point? And have we learned much last part on how you're going to sell these machines relative to the one you sold in the fourth quarter. Recently, they've been more the traditional lease model. What do you know today about the mix of how that will go?

Yes. Look, so I bought an electric car, and then I regretted it instantly because you should just lease electric cars because the rate of improvement and change is so incredibly fast that why would you want to own these things. And I think that's actually going on in e-beam. And we designed it with the ability to upgrade along the way. And a lot of that is because the gentleman in the phone asked the question around the software getting smarter. The integration of the layout software, actually even IP on the chip, original idea had that. We still work with customers on that. There's a lot of dead space inside chips, you could put interesting structures there. And the analytics to know not just the measurement result, but all the fields around the design attributes, that was kind of understood from the beginning, right? I always said to the team, if what we're trying to do is be a better capital equipment company in the capital equipment industry, we should just stop it now because we have to think about it from a system perspective and do it very differently for it to be a value. So that holistic view was always there. I don't think we ever -- I think the difference is Kimon and I presented it before to the investor community. And this time, we had internal present it. So probably that helps a fair amount because I think he could present it thoroughly and better than you did. Subscription versus purchase, as I said to them upfront, we try to make it economically slightly advantageous for the customer to subscribe because we think that there's value to them being able to get upgrades and support it that way. And frankly, at our scale, 1 this quarter, 1 next quarter. No, 2 this quarter, 1 next quarter, it's so lumpy, it drive our shareholders up the wall, right? So we'd like to make it as smooth as possible that does short -- it doesn't mean you get as much money upfront. But I think over the long term, I think it's more aligned with how we operate and more aligned with the system view and more aligned to the comment he made around about the Exensio time-based licenses. The things should get better over time, much like your Tesla gets better over time. The DirectScan application gets better over time, mostly with software, but in some cases, with hardware as well. So how will the market respond to it? And will it stay that way or not? I don't know. I think one of the things that we learned when we were super dogmatic about the IYR business model, I don't know that we get ourselves a service by being sold out of the box. So we thought to ourselves like our customers are orders of magnitude bigger than us. We show them multiple models. We let them decide. And I think so far, I think there is definitely parts of the organization that I think finds that hard to digest. But when you sit and think about it rationally and you show the economics, it actually kind of makes a lot of sense, especially in a part of the market where there's a huge rate of change, which I think they just go right now. But I know what it will be forever.

The SAM?

The SAM, if you look overall, I think ASML at the Analyst Day, ASML -- one of the best Analyst Day out there. I love what we do the Analyst Day stuff. We talked about additional wafer starts per year. When we back that out, we think there's somewhere around $800 million to $1 billion spent on e-beam inspection per year, roughly around now. That's assuming a certain attach rate per 10,000 wafer starts of wafers and an e-beam inspector. I think that's actually going up, there are more inspectors per wafer start. So if you see in the flat rate, it's in that range. And then we think about half of that is voltage contrast based and the other half is fine future inspection. So that would be $400 million to $500 million is your kind of voltage contrast-based inspection. I don't know if that's accurate or not. And then what fraction of that is direct scan useful for? I think we're still learning that inside that $500 million voltage contrast-based inspection add that you say. And if you flip it to a subscription, how would that look? Because that's looking at the -- if you sold that on a capital basis, what would it look like, which is what the market does today. It is -- I think Oliver Patterson from Intel gave a nice chat, right? I mean I don't think it's -- a chef wants multiple knives and not a Swiss Army knife, right? If you really want to be a Michelin star chef, you have multiple knives, you don't have a Swiss Army knife. I think early on, the e-beam tools were Swiss Army knives. I think they are evolving to be more like a chef's knife. And you talked about multi-beam machines, there's a value in a place for those. What we are doing with Direct, we think there's a value in a place for them. I can tell you that when we look -- I always tell our teams when we used to do yield ramps and we still do working with customers, any engineer that tells me the yield loss mechanisms are random. I know it's a bad engineer. There's nothing that's random that goes on in the factory. It's always systematic. You just haven't figured out what the -- there will be a part that fluctuates like the butterfly flapping its wings in Mexico, causing a storm in San Francisco. There's always some kind of a stochastic behavior to it, but there's a deterministic element. You need to figure that out. And so the whole idea behind direct scan was there's always a deterministic element to it, and you need to be able to go figure that out. And the power that it unleashes when you understand the connection to layout is very, very valuable. But when we talk to the engineers that use it, what they love, and Oliver hit on it a little bit, is after it finds all the spots of the defects, it goes back and takes an image and it goes back and tells you what the layout is. So you know what one scan on the wafer, this is what fails, this is what its failure rates were versus the other more genetic layout patterns, and this is what drove those layout patterns. Now by the way, here's a picture in case you don't believe me because it doesn't take a picture to figure out where the failures are. It is just a one pixel measurement. So that packaging up, we think, is valuable. It's a more system view. But what fraction of that $500 million would represent much more than what we're selling to now. So that's enough. Any other comments or questions? I think we're getting in between people and drinks. So I do -- first of all, I want to thank everyone for attending. Also, many of you are shareholders for a number of years, and we do really appreciate your support and your confidence in the business, and some customers are in the audience as well, and we appreciate them as well. As I said earlier, Kimon and I see this -- we see ourselves as stewards of this business. We are stewards of your shares of our customers' products and tools that they use and our employees' paychecks. And so we're really honored to have the opportunity to be in these roles as long as we have. We both grade and lost a lot of our here. But we continue to try to make it a better business. And we always appreciate the feedback. As Mike said earlier, every feedback is a gift. So thank you.