PDF Solutions, Inc. (NASDAQ:PDFS) Q3 2024 Earnings Call Transcript November 8, 2024
Operator: Good day, everyone, and welcome to the PDF Solutions, Inc Conference Call to discuss its Financial Results for the Third Quarter Conference Call ending Monday, September 30, 2024. [Operator Instructions] As a reminder, this conference is being recorded. If you have not yet received a copy of the corresponding press release, it has been posted to PDF’s website at www.pdf.com. Some of the statements that will be made in the course of this conference are forward-looking statements, including statements regarding PDF’s future financial results and performance, growth, rates and demand for its solutions. PDF’s actual results could differ materially. You should refer to the section entitled Risk Factors on Pages 16 through 36 of PDF’s Annual Report on Form 10-K for the fiscal year ended December 31, 2023, and similar disclosures in subsequent SEC filings.
The forward-looking statements and risks stated in this conference call are based on information available to PDF today. PDF assumes no obligation to update them. Now, I’d like to introduce John Kibarian, PDF’s President and Chief Executive Officer; and Adnan Raza, PDF’s Chief Financial Officer. Mr. Kibarian, please go ahead.
John Kibarian: Thank you for joining us on today’s call. If you’ve not already seen our earnings press release and management report for the third quarter, please go to the Investor Section of our website where each has been posted. Before Adnan discusses the financials in detail. I have some comments to make about our observations from the third quarter. Our view of the market and our business prospects for the remainder of the year. Bookings for the third quarter were driven by customers continuing to buy Exensio and Cimetrix connectivity software. Exensio sales include a large cloud customer renewing at a 50% increase in annual revenue due to the continued growth in usage as well as a number of customers deploying Exensio for process control.
Advanced logic, advanced packaging and high voltage semiconductor manufacturing fabs drove the increase in process control licenses. This is consistent with our perspective of where investments are being made in the industry. Improvements in equipment runtime licenses of Cimetrix connectivity was relatively broad based. Our integrated yield ramp business was weak in Q3 as wafer volumes were low and new contracts were slower to sign. We believe that both situations will likely improve over the next few quarters. Turning to design for inspection execution during the quarter. I am pleased to report continued great results. The eProbe manufacturing evaluation at our second customer is proceeding well, the customer and PDF team believe that the evaluation can be completed ahead of schedule.
This speaks the unique capabilities of the solution and the robustness of a hardware deployment. Utilization at our initial customer for eProbe DirectScan, where two machines in place remains high. While exact timing is always a challenge to predict, we anticipate both customers impacting our bookings over the next quarters. As our confidence in eProbe value and logic increased, we began exploring applications in memory R&D and production. In Q3, evaluations of the advantages of DirectScan on memory was very promising with the customer reporting that sensitivity and throughput advantages were over 10x superior than conventional methods. The impending completion of the manufacturing evaluation, continued application development at our lead customer, and new applications in memory increased our confidence in the DFI eProbe business.
As evidenced by our ramping capital spending this year, we anticipate the eProbe being a driver of revenue growth in Q4 and having a meaningful positive impact on our 2025 and beyond. Now, a few comments on our view of the environment and our perspective on the fourth quarter and beyond. As we talk with customers about their business, some are experiencing weakness, while others are growing. Consistent with our view last quarter, we believe our business will be driven by fabs developing advanced logic processes such as 2-nanometer, fabless customers deploying advanced test control software often with AI/ML to augment conventional test methodologies. And companies engage in digital transformation attempting to leverage data, whether that is IDMs, fabless, foundries, or large equipment companies So while we anticipate an industry where there will not be a rising tide lifting all boats, we believe we can extend the momentum we’ve begun in Q3 for continued growth in Q4, and while it’s too early to comment on specific numbers for 2025.
We expect robust growth then also. I do want to remind folks about our one day AI executive workshop in San Francisco on December 12, which is the day after the IEEE IEDM conference. We are bringing together a great collection of customers, industry experts, and PDF folks to talk about the advances in the application of AI for semiconductor manufacturing. Driven in part by our new model ops, guided analytics, and our Exensio analytics platform. I invite you all to attend. I want to thank all of the PDF employees and contractors for their efforts during the year. Let’s have a great Q4, so we can deliver another record year. Now, I’ll turn the call over to Adnan, who will review the financials and provide his perspective on our results. Adnan?
Adnan Raza: Thank you, John. Good afternoon, everyone. Good to speak with you all again today. We’re pleased to review the financial results of the third quarter and to bring you up-to-date on the progress of the business. We posted our earnings release and management report on the Investor Relations section of our website. Our Form 10-Q has also been filed with the SEC today. Please note that all of the financial results we discuss in today’s call will be on a non-GAAP basis, and a reconciliation to GAAP financials is provided on the materials on our website. Financial results for the third quarter of 2024 came in strong. Our bookings for the 9 months of 2024 have now exceeded our bookings for the full year of 2023. We ended the quarter with a backlog of approximately $240 million, essentially flat from last quarter, even with record revenues delivered this quarter.
Our total revenue for Q3 came in at $46.4 million, which is 11% higher versus the prior quarter of this year, and 10% higher versus the same quarter of last year. We are pleased with this strong performance in total revenues in spite of the decline in integrated yield ramp revenue. Our analytics revenue came in at $44.8 million which was 17% higher versus the prior quarter and 13% higher versus the same quarter of last year. Analytics comprised 96% of revenue for the quarter. The strength in analytics revenue. This quarter compared to last quarter was driven by all elements of our analytics platform. As John said, we’re pleased with the level of engagement with our analytics customers. One example of which was a multi-year eight figure renewal where we were able to increase the annual spend rate by approximately 50% primarily driven by increased usage and licenses the customer is deploying as more people inside their organization rely on Exensio for yield analytics and manufacturing improvements.
For our Cimetrix connectivity products. We saw a slight improvement in runtime licenses during Q3 compared to the prior quarter with strong year-over-year growth integrated yield wrap revenue was 4% of total revenues in Q3 and was lower by $1.9 million compared to the prior quarter and $1.2 million compared to the same quarter of the prior year. Overall, we are pleased with the growth rate we delivered for analytics, total revenues for the quarter and continued engagements with our customers across our analytics platform. On gross margins, we reported an unusually strong 77% gross margin for Q3 which benefited from one-time perpetual software license deals during the quarter. While we’re pleased with this result for the next quarter, we expect gross margin to revert towards what we have been seeing during first half of this year, driven by a shift in the mix of our product offerings.
As announced at our Analyst Day in October, we remain committed to our long-term gross margin target of 75% and making progress towards that over the coming quarters. Our operating expenses in Q3 grew compared to the prior quarter, primarily due to increased investments in sales and marketing and R&D to support our future growth. On EPS, we were able to deliver $0.25 per share for the quarter, our strongest quarter for the year. Turning to the balance sheet, we ended the quarter with cash, cash equivalents and short-term investments of approximately $120 million, incrementally higher compared to the prior quarter ending cash balance of approximately $118 million. This quarter, we used a portion of our positive operating cash flow for investments in the eProbe tool as well as investment in a private company where we see opportunities to partner for the benefit of our leading edge enterprise customers.
After achieving year-over-year revenue growth for the Q3 of 10% for total revenues, and 13% for analytics revenue, we expect year-over-year total revenues in Q4 to grow in line with our long-term revenue growth target of 20%. We are also thankful to our customers and partners for supporting the growth, uplift we delivered this quarter and look-forward to growing sequentially in Q4. With that, we’ll turn the call over to the operator to commence the Q&A session. Operator?
Operator: [Operator Instructions] Our first question is from Blair Abernethy with Rosenblatt Securities.
Q&A Session
Follow Pdf Solutions Inc (NASDAQ:PDFS)
Follow Pdf Solutions Inc (NASDAQ:PDFS)
Blair Abernethy: Hi guys. Nice quarter. Good to see the growth coming back there. I just wanted to ask a couple of questions. John just on the DFI, can you just — you’ve talked about customer number one and number two. What are you seeing out there in terms of pipeline of opportunity beyond the first couple of customers there?
John Kibarian: Sure. That’s a great question. Thank you, Blair. So as you know, we have also a third customer that is part of an overall integrated package. That subscription includes Exensio and vehicles and many other systems. I didn’t talk about that customer this quarter, but we do expect that customer to proceed. For the — my prepared comments, I also talked about memory application. And we have seen really great results there. We do see that as another growth vector for the systems. And when we look back at logic, as I said in my prepared remarks, with both the first two customers, we continue to see more and more applications, more and more types of problems as the customers develop more 3D processes. So gate all around is 3D, contacts and vias in 3D, backside power is 3D.
So more and more of these new features they’re bringing out on advanced process nodes depend on 3D integration and 3D yield problems. And so the machine is uniquely capable of looking at 3D on product and understanding exactly where it is in the product, and also being able to make sure it only excites the elements of the product that a person wants to excite to capture the correct failure mechanism. That’s really a lot of where the unique capability is. So, we expect more applications within logic, and now we’re starting to see applications within memory.
Blair Abernethy: Okay. Great. Thank you. And just add on CapEx year-to-date looks like around $12 million up a lot from last year, double the year before. So what should we be thinking about in terms of CapEx levels going forward?
Adnan Raza: Look, I mean, I think, Blair, nice to connect with you. I think we’re feeling pretty good about the progress the DFI platform and the eProbe engagements have had. Obviously, having a few machines with one customer and other one with another customer. And the evaluation that’s going on with the third one, and then John alluding to these memory applications being positive. All give us confidence to start to think about, okay, if we had to serve the needs of these customers, where should our order and the pipeline be? Obviously, we’re going to be careful in terms of looking at what are the longer lead times versus the shorter ones as we have talked to you before. But we see increased levels of CapEx and we have started to see some of that translate into revenue. And we are seeing this quarter in growth and we hope to continue that trend. So I think on the CapEx side, you should see some increased level from us even from where we are today.
Blair Abernethy: Okay. Great. And just one last one just for it to jump back in the queue. The SG&A step up in Q3 here from the first half of the year. Is that you should be thinking about that as kind of a new level or are there some one-time items in there?
Adnan Raza: Yes, this I mean, most of the increase, almost all of the increase is on the sales and marketing side. G&A, we’re careful about managing it, and of course we report as a combined SG&A level. But on S&M, I think you’ll see incremental expense if anything. We’re focused on kind of trying to maximize the S&M spend that we have and making the best out of that. Our investments in the future in the next quarter or so that we’re thinking about internally are probably more on the R&D side than on the SG&A side.
Operator: Our next question is from the line of Gus Richard with Northland Capital Markets.
Gus Richard: Congratulations on the good result. Could you talk a little bit about that perpetual license deal that you had in the quarter? I’m assuming that’s a onetime thing.it looks like, it grow incremental growth margin at like 98% and I’m just trying to reconcile those two things.
John Kibarian: Sure, I guess, is a set of my prepared remarks. Actually, the largest bookings in the quarter were related to a cloud renewals and new cloud total cloud deals, but they don’t drive very much incremental revenue in the quarter. We have some customers that have legacy contracts on process control typically because they will buy that on a perpetual basis with capital purchases typically, and that drove incrementally more perpetual license revenue in this quarter. That’s something that we’ve known about for quite a while. There’s a couple of customers out there that have legacy contracts on this stuff, and they tend to be they buy those in link with their capital or build out. This is related to advanced logic and advanced packaging, because you need more process control and advanced packaging, and it did drive the gross margin improvement in the quarter.
So, as you said, it’s, basically a very high percent. Legacy customers with no support, not a lot of legacy, handholding, it’s pretty much just license revenue. Most of the one-time licenses on Cimetrix, it’s basically all so.
Gus Richard: Worth it. Got it. And so, you’ve got, pretty decent sequential growth. Just given the 20% year-on-year you talk about, what’s — that perpetual license? I don’t expect to repeat and I’m just wondering if you give a little color on sort of what’s filling the GAAP in the fourth quarter in terms of the increase in revenue, is it IYR coming back? Is it, starting to recognize some?
John Kibarian: So I think there’s a three factor there for, as you look in the Q4. Number one, there is, as I said in my prepared remarks, we do expect improvement in the IYR primarily directly to customer contracts a little bit on the improvement on the way for fees from what we see reported from the customers. Number two, we do expect probably the bigger piece, DFI, as I said, in my prepared remarks, customer completing, earlier than we expected the — they anticipated the, evaluation as she has met all the criteria sooner than expected. And then, number three, as I said earlier, right, the largest bookings were really actually related to cloud and that tends to have a longer tail impact and over the next few quarters that will contribute, but it didn’t do very much in Q3 but nothing.
So those three drivers will, basically build out in Q4 and beyond. We do expect some modest improvement in run time licenses on the Cimetrix side as well. From what we see in customers, runtime licenses were, as Adnan said, a very significantly year-over-year and seem to be building as we go through the year. That’s less of a driver though, in terms of total dollars.
Gus Richard: Got it that’s helpful. And I, and this is going to, well, let me start with the easiest way to look at this, have you sized the voltage contrast market and given eProbe, how much do you think it expands that market as that comes into production or work where they could use some fab? And how much do you think memory, the potential of memory adding to the applications sort of expand the market opportunity for you? Any sizing would be helpful.
John Kibarian: Sure. Yes. I mean, I think you can go and look at reports, but I mean the e-beam inspection business has been, I don’t know, over north of a $0.5 billion. The number, I’m not clear in my head right now. But I think when you talk to folks, everyone expects it to grow. When I was meeting with a customer recently, they say with more 3D problems, even DRAM is going 3D if you look at it these days. Voltage contrast increasingly becomes important relative to see in line what’s going on with the products. So we expect that the e-beam and voltage contrast in particular business on the inspection side to outgrow the overall inspection market. And when I speak with even folks at other people that have other product in the market, the executive team, to see to say the same thing.
So I think everyone expects that number to grow pretty substantially faster than the rest of the market. And the part of e-beam that we think is most valuable is that voltage contrast because of the 3D nature of defects. There are also applications that are growing that are related to imaging. Historically, the biggest piece of e-beam business has actually been memory, not logic. And we started with logic, I think in part because of the complexity of being able to navigate around product and all of the software PDF had that made inspecting product possible at high throughputs. All the software we have around simulating, and simulated voltage contrast and actually knowing how to direct the machine where exactly to look, we call that point scan.
So we had started in logic, but ultimately, the market in memory will be probably very significant. Over time, which one is larger, Gus. I think that’s hard to say because looking backwards, that’s a joke, you can only collect the dots if you look backwards. I think you miss where the world is going, right. So the future of logic will be things like backside power and CFETs and all these things are three-dimensional issues. So the need for voltage contrast and logic, we think will go up. At the same time, the memory is also going 3D, and so the need there will also increase. So overall, we think e-beam will be well over $1 billion inspection market. The piece of voltage contrast, we expect to be the largest piece of it. And the fraction that what we’re saying today by saying that the eProbe has application both in memory and logic, we don’t really have to worry about which of those two is bigger.
We actually span basically both end market applications.
Operator: [Operator Instructions] We have a follow-up from the line of Blair Abernethy from Rosenblatt Securities.
Blair Abernethy: I just wanted to drill in a bit more John on the macro environment. Just sort of where you’re seeing things tracking and we’re covering in the last quarter. And I would say also, specifically, what are you guys seeing in the China market? It was down sort of fairly significantly for you guys from a year ago and prior and just kind of want to see how are you feeling with that end market?
John Kibarian: Yes, that’s so in general, as I said, in my prepared marks, we just see it’s a mixed bag. So if you look at our — what drove our revenue this quarter, it was clearly advanced packaging, advanced logic, some high voltage semiconductors and equipment going into those broad more broadly, but we did see an awful lot of back end equipment. I didn’t say that and then prepare myself. I thought it was kind of a model trend there, but for sure you can see some demand in those areas. Those tend to have a bigger impact, outside of China, right, advanced logic and advanced packaging, most of that activity is going on in Taiwan to a lesser extent, Korea and the U.S. So you see that shift in our geographic breakdown on revenue looking out, we did see we did report that IYR was weak and wafer IMEs weak that’s greatly influenced by China.
We did see weakness in China overall in terms of just the how the wafer fees coming out of fabs and the volume that said, we’ve seen continued heightened investment there and we do see new factories coming online, new notes coming up and new engagements on advanced development for them what would be considered maybe not advanced by non-Chinese standards. So we do expect that piece of business to recover. We don’t think it’s loss per, in any way, shape or form, and as we look into 2025 for that part of the market, we do think it also be a mixed bag there. We do expect some consolidation within the customer base. When I was chatting with folks in that country, I think there’s been kind of the Cambrian period of lots of new species and at some point, the total number of animals may increase but the diversity of species may decrease.
And I think you’re going to see that over there over the next year.
Blair Abernethy: Okay, great. And another question I had was just really around the partnership, particularly with SAP any progress to report on that one.
John Kibarian: So, we have, you have critical deployments with them ongoing. We were featured at their verticals meeting back in October, I believe it was in Seattle, where we met with a number of mutual customers and you’ll see them, they were on a panel with me at the GSA and you’ll see them at our AI conference as well. There’s a number of selling activities going on with them and us and a number of things we’re doing on the development side, because as we look at the future of deploying AI for our customers, when you want to operationalize it, tying what knowledge you have about the products in the ERP system with what’s going on the shop floor is very important. So, besides the impact it has for, we believe a number of these deployments, some, that are ongoing and moving quite well. We do expect it to drive or be relevant to a lot of our AI deployments as well.
Blair Abernethy: Okay. Great. Thanks for that. And one last quick one. I’m not sure if there’s anything — I haven’t seen anything in the press, but just in terms of the batteries manufacturing sector, your Lantern Technology acquisition last year, anything to note there?
John Kibarian: Yes. We’re in deployments at some battery manufacturers working with the car companies that are working with the battery producers. So really that linkage between the car company and the battery manufacturer, really leveraging our AI around being able to really, at high speeds, detect the variability in the manufacturing process at run time. And those pilots are ongoing. We’ll see as we finish this year what they result in terms of our expectations for 2025. But we’re quite pleased with the technology progress.
Operator: Our next question comes from the line of William Jellison with D. A. Davidson.
William Jellison: Good afternoon and thanks for taking my question. The first one that I wanted to ask was if you could provide an update on some of the pilot programs you’ve been running in MLOps, and how those have been going and what sorts of progress you’ve seen thus far as that product enters the market.
John Kibarian: Sure. Yes. And we’re going to talk about this at the AI workshop, Will. And we did talk earlier this year about first customers. We’ve now got a number of pilots ongoing with other customers, mostly around this kind of complex test flow. So using upstream information to make better prediction about downstream tests or actually our downstream make better predictions. So these are things like understanding virtual burn-in insertion test points. In other words, can I, in AI, predict the burn-in result and thereby skip the burn-in step or minimize it or change it, or reduce it. And similarly around matching, if I know the results at wafer sort, can I predict what this chiplet would look like inside the package. And therefore assign it to a collection of other chiplets that results in a better overall system performance at the package level.
So there’s a couple of them ongoing that are in that kind of category, those kind of categories, virtual insertion points, predictive insertion points. And they really leverage, really, the benefits of MLOps, which is that ability to span not a single test point but multiple test insertion points and take upstream models extracted from upstream data to make better access of downstream data and testing.
William Jellison: Great. Thank you, John. And then the follow-up question is, Adnan, if I heard correctly in the prepared remarks, it sounds like PDF made a small investment in a private company during the quarter. I was wondering if you could provide any more color on what that was.
Adnan Raza: Yes, absolutely. I’ll speak to the numbers and then I’ll let John talk about the qualitative aspects, which I alluded to that it’s related to the success of our leading-edge customers. But yes, it’s a $2 million investment in a convertible note. We felt the technology was quite differentiated. Sometimes there’s an opportunity to look at purchasing companies or sometimes there’s an opportunity to see if you can fund them to the next phase of growth. Particularly if they have engagements with customers that you deem important as well. So this was one of those situations where it made sense for us to support the next phase of growth to the next milestone. And it was a small enough investment for us that with the convert, we negotiated the terms that I think if the success happens, we all positioned, John, do you want to speak to the —
John Kibarian: On the technical side of it. Well, as I said in the Prepared Marks of Blair, right, the future leading edge businesses, they’re all 3D, they’re all the innovation goes on that the electrical behavior of the systems, we’ve had a test vehicle business that’s super excellent at that, super valuable customers of that. The eProbe brings that in line. And this is another way of exploiting our software stack to understand information about the design, the connection to our partnership in connection with Siemens, to understand the relationship between the design layout the test faults. This lets us get additional insight and information. It is more on the physical measurement side, but it links with our software stack and is also potentially very synergistic with the pro.
Operator: Our next question comes from the line of Christian Schwab with Craig-Hallum Capital Group.
Christian Schwab: Most of the questions I had were answered, but as far as ‘25 outlook. I think you guys said you felt positive or should we anticipate anything different than operating that your, at a minimal, your 20% long-term growth target, is that the way we should be thinking about it?
John Kibarian: Yes, we don’t give a specific number, but obviously, we said when we talked about this year, we said the second half of the year would set us up for annual 20% growth. You can see our Q4 number. So you can kind of see where we’re heading with this, right?
Christian Schwab: Right now, I’m talking about calendar ‘25. So now that we’re back…
John Kibarian: On our, our, we think our Q4 is more representative of how we’re thinking about 2025 than we think of 2024 on average, right? Because the first half year was flat. So obviously, the blended average for 2024 is less than our 20% growth target. But as we look at 2025 my words were robust because we feel like 2024 leaves us on that growth number and we expect to be in that zip code or better as we go into 2025.
Operator: [Operator Instructions] We have a question from the line of Andrew Wiener with Samjo Management.
Andrew Wiener: I wanted to follow back up on your comments around DFI it sounds like you’re making good progress with the lead customer and the customer doing the manufacturing evaluation, and we’re obviously further along in the year and add on referred to CapEx spending, and the long lead items. Just curious last quarter, you sort of framed either what you thought the potential demand could be from those customers and or at least what we were planning to have capacity to potentially ship, throughout, I guess the balance of this year and into ’25. So maybe you could sort of update that given what you’ve learned over the last sort of three or four months.
John Kibarian: Yes. I think by and large from a demand standpoint, we feel pretty good about where we are. We expect to be in the same range, so it’s going to be in that, I forget the exact numbers we gave out. But I think it was in the 4 to 8 range, if I remember correctly. In terms of shipments, we do believe pretty comfortable about that area, about that range. We will keep on monitoring our ability to build and ship as we go through this year, right? We don’t have all the material that we would need to make that number yet. We don’t have those things built and we don’t have the timing set up. So we would have to — we’ve got, unlike standing up the cloud. There is more orchestration required, but we think the demand is out there for that range.
Andrew Wiener: Okay. And that would be essentially from those two lead customers, and then if you got to the higher end perhaps an initial memory customer. Is that the way to think about it?
John Kibarian: Yes. That would be correct. Andrew, there’s additional eval that we probably should think about. We really don’t have — I think if we could put eval machines in other places like we do with the manufacturing eval, we would probably accelerate growth. So that’s something we do need to factor as we think about our supply chain. Because people have a lot of questions, I think the like the manufacturing eval, I think people — customer was able to see the results from our lab and wondered, okay, if you install a machine in a real facility and you run real lots through it in real time, does it actually really work and can we use it? And that’s why they were able to finish quickly because, we felt pretty confident that the physics were the same around the globe.
And so if we stuck the machine in there, it would result in sales more quickly. And it’s — let’s see, knock on wood, but we believe it’s on path to do that. So we would probably want to have some capacity to expand that program if we think about the year.
Andrew Wiener: Okay. And then you referenced, again the early completion of the manufacturing evaluation. Without putting you on the spot to get into too much of the details, assuming if you meet the criteria. Is the intention to convert that tool into a revenue-generating tool at the customer? Or does it need to be taken back and any upgrades or anything like that done in order to convert that to revenue?
John Kibarian: Yes. Without getting to the specifics, Andrew, we would expect it to convert to revenue sooner than a standard situation.
Andrew Wiener: Okay. And then I’m just curious separately, following up on the battery side question. Have you gotten far enough in understanding the capabilities of our systems and sort of customer needs to, if the evaluations go well to size what the opportunity could be? I mean, I’m not necessarily saying the first contracts, but sort of in general, if one thinks about a battery line doing X number of batteries and generating X number of revenue. What or is it more like DFI on a sort of tool and software basis? I’m just trying to think about what it could mean.
John Kibarian: Yes. I think we’re still trying to figure that out ourselves, Andrew. But what we realized going all the way back a couple of years ago to the first pilots we did, we were quite surprised at when you compare battery with semi — the customers found us because they said okay, what software is used for semiconductor manufacturing? That’s a sophisticated manufacturing line to improve yields. How would that apply to us? And we went and we started deploying, and what we found was the data collection rate on the equipment sensors is actually much more simplistic than what you would get off of a capital equipment tool in the front and drive even a back-end test assembly facility fab, so that we felt there was an opportunity there.
And then they collect a lot of images, but they do almost nothing with them. And part of that is just all around how fast the line moves and therefore how quickly any data you get off a sensor or off imager, how quickly you’d have to turn that into a control chart and operate. There, it’s much, much faster than semiconductors. Things are moving at meters per second. We talk about the eProbe moving the wafer at 10 millimeters per second. This is moving at many meters per second. So it’s a whole different scale of data generation rate. And what Lantern had was a very, very fast AI pipeline for being able to process and create alarms based on real information. And we’re trying to see how valuable that is. First of all, the customers want to believe that this will actually work in a real line.
So that’s what we’re doing right now, and then if it does really work, okay, how do they use that to control their line better? They have, there’s a lot of reasons why it should be very valuable, because a lot of your unlike semiconductors where most of your cost is the capital, the cost of the chip is, depreciation of the capital, most of your cost is consumable. So running a line even for a few minutes longer. Before you take action is actually quite expensive as a percentage of your capital build cost. So there’s reasons why it should be valuable. We are trying to work through all those things, over these next, months. The nice thing about it is the cycle times are so much faster. You have material in and days and weeks, not months and years.
So you learn your ability to see the benefit of the in line data and modeling and alarming is quicker. So that has helped us get learning cycles faster. But I suspect it’s 2025 initial customers and then we’ll get some understanding about what that means. In the out years, we made that investment, we didn’t think that was going to be a quick hitter. It was a small investment that was a little bit of a moonshot.
Andrew Wiener: Along those lines, John, but as you’re doing these pilots, are you seeing them deploying or, evaluating other, systems or is it an out of the box approach of looking at your solutions, it would be sort of a completely new way of sort of approaching, their manufacturing process.
John Kibarian: Great question, Andrew. We’re still trying to understand that too. When we did our own surveys, we didn’t think there was anything else out there that was quite like what they had and we thought, given our experience in Semiconductor manufacturing and how important it is to always, they say shift to the left, move upstream any ability to predict downstream problems. And we knew that was super important in our industry. We felt like it would be important in their industry. And we were quite surprised at how late they get information, relative to the production flow for this industry. So it seemed like there wasn’t, another alternative out there that was quite like what they had. There is in line data collection, but it’s quite simplistic.
And there isn’t, both the combination of in line and sophisticated that we saw. And that’s why we effectively started putting our toe in the water with this acquisition. But I personally, I always scratch my head on this. I think any manufacturing person would think it’s intuitive. You always want to get the best predictor you can as early as you can in the process. There’s — and, so we would think that this would have been something others would be doing, but we don’t know of another one out that is quite like what this team had built.
Andrew Wiener: Okay. And then maybe my last question. I know you pointed out advanced test as sort of an area of strength. Can you talk a little bit about sort of what you’re doing now with your Advantest and Teradyne? And how you’re working with them to capture that opportunity versus sort of going directly to customers?
John Kibarian: Yes. We’ve involved a certain amount of direct work with customers, right? A lot of what we’re doing with them. They have, particularly an admin test case, have brought a lot of compute. They call [ACS], to the — sitting with the tester that enables more sophisticated AI. And our — the standard way that models are run, they — when they’re running the tests, they’re very lightweight, often rules. Because you don’t want to overload the process of running the test program with a very, computationally intensive model. So what those edge boxes enable is to bring a much more sophisticated ML model or even some elements of design automation information to the edge. With a certain level of security because of the way they’re architected to work with the testers.
And so, you know, a lot of what we’ve been doing with them is coming up with ways that with their software and hardware stack, and MLOps and our other capabilities, you’re able to enable more sophisticated control at the test point. And the reason why that’s needed, when you look at advanced packaging, you’re testing many chiplets at wafer sort instead of just one chip. And you maybe have 3 test insertions for each. So 4 chiplets, 3 test insertions, that’s 12 tests at wafer sort. And then you’ve got test points in final test as well, multiple ones, after you’ve packaged it. So there’s a lot of value of being able to take information for upstream and predict downstream or make a more informed decision downstream. And a lot of what we’re doing with them is really looking at as you put more additional algorithms at the edge, how does that benefit the customer and their test?
What can you do beyond a simple or a thinner model? And this is true for both of them.
Operator: [Operator Instructions] At this time, there are no more questions. Ladies and gentlemen, this concludes the program. Thank you for joining us, today’s call.