PDF Solutions, Inc. (NASDAQ:PDFS) Q3 2025 Earnings Call Transcript

PDF Solutions, Inc. (NASDAQ:PDFS) Q3 2025 Earnings Call Transcript November 7, 2025

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 Tuesday, September 30. [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, including statements regarding PDF’s future financial results and performance, growth rates and demands for its solutions. PDF’s actual results could differ materially. You should refer to the section entitled Risk Factors on Pages 16 through 30 of PDF’s annual report on Form 10-K for the fiscal year ended December 31, 2024, 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 would 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 Investors section of our website, where each has been posted. Bookings in the third quarter were strong as we continue to realize the benefits from our investments in product development and customer support. As we announced in the September press release, we signed an extension contract with a large customer that involves bringing our characterization vehicle infrastructure, Exensio characterization software and eProbe machines to their manufacturing sites as well as expand usage at their R&D site. The eProbe machines under this contract are provided under a subscription.

Expanding beyond the machines they have in their R&D facility, we now have shipped 2 additional machines that are in the process of being installed at their first production site. Also in the quarter, we announced that we licensed Tiber AI Studio from Intel. With this license for source code, we are integrating the Tiber AI Studio’s award-winning data science operations platform directly into Exensio. Tiber AI Studio enables engineers to build and manage hundreds of thousands of AI models. Coupled with Exensio’s existing ModelOps, which enables model deployment in the fab and test floors, this integration is designed to enable engineers to use Exensio to both train models as well as deploy them. On a stand-alone basis, Tiber AI Studio already had hundreds of users.

As we talk with our customer base, we are hearing the same message. They have great proof of concepts using AI, but scaling and maintaining large deployments remains elusive. We believe the integration of Tiber AI Studio with Exensio, which we call Exensio Studio AI is an important capability required to close this gap. Among the Exensio contracts signed in the quarter, notably, we signed an 8-figure contract with a large IC manufacturer. We are honored this customer selected Exensio as their data analytics platform, the primary repository of manufacturing data and the platform on which to integrate their internal systems using Exensio’s big data APIs. As part of this contract, they will leverage Exensio Studio AI to manage their AI deployments in production.

We also closed an 8-figure contract for secureWISE with one of the largest equipment OEMs in the world, which extends and expands their existing licensing. Finally, contributions to revenue from our Cimetrix connectivity and control software were the strongest since the acquisition closed at the end of 2020. Historically, the large equipment OEMs make their own control and connectivity software, however, as our market share has expanded and more equipment is now — as our market share has expanded, more equipment is now shipped with our software installed on it than internally developed software of any single equipment vendor. This means, our software — our customers enjoy software that is proven across more applications in the fab, test floor and assembly facilities.

As we integrate Cimetrix, secureWISE and AI-enabled monitoring, we aim to enable equipment vendors to more easily deliver smart tools and value-added subscription services. Now, I’d like to turn to the environment. The industry is making significant investments in 3D manufacturing in front-end fabs and packaging facilities as well as product design. There is an increased geographic diversification of manufacturing locations. As those investments are ramping up, we see customers working to make the new processes, products and facilities economically viable. It is well understood that diversifying manufacturing comes with the risk of driving up production costs and slowing innovation. We see AI-driven collaboration as a critical capability to enable cost-effective and efficient manufacturing in many of these new locations.

Last month, we were invited to present our vision of AI-driven collaboration at SEMICON CEO Summit in Arizona. My comments, which appear to resonate with the audience, outlined how the industry can leverage our secureWISE network, Sapience orchestration products, and Exensio AI to collaborate with their customers and suppliers. We have also noticed that our equipment, fab and fabless customers are looking for ways to move from human-driven collaboration to AI-driven collaboration, in part to enable more efficient production around the world. As we look to Q4, we are preparing for our users conference and Analyst Day. Since 2020, when we acquired Cimetrix, and began the journey to be a comprehensive analytics platform for the industry, we have driven results equal, or exceeding the long-term revenue growth, non-GAAP gross margins, and non-GAAP operating margin goals we set in 2019, and also have exceeded the revised goals we established in 2023.

Before 2020, we had approximately 150 customers that were primarily fabs and fabless. We had few equipment companies and almost no cloud suppliers as customers. Today, we have over 370 customers, including most of the equipment industry and multiple cloud providers. It is a unique customer base as we bring analytics capabilities to every aspect of the semiconductor supply chain. Today, our cloud systems manage petabytes of data and secureWISE network transmitted exabytes. Our systems are used to control tens of thousands of tools. We believe the success we’ve achieved to date is due in large part to the relentless investments, including the acquisition of secureWISE and the build-out of our eProbe machines, both of which required us to use our balance sheet this year as investments were ahead of the growth they enabled.

We expect the profits generated from these investments in 2025 will enrich our balance sheet in 2026 and beyond. Finally, I encourage you all to attend our Analyst Day and Users Conference. There you will see our customers, partners and PDF folks talk about the needs and opportunities for AI and analytics and manufacturing. We are honored to have Mike Campbell, SVP of Qualcomm; Aziz Safa, Corporate VP of Intel; Tom Caulfield, Exec Chairman GlobalFoundries; and Jean-Marc Chery, CEO of STMicro, among others, share their perspectives. Now, I’ll turn the call over to Adnan. Adnan?

Adnan Raza: Thank you, John. Good afternoon, everyone. Good to speak with you all again today. We are 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 in the materials on our website. As you saw from our press release, with our Q3 results, we achieved another record for quarterly revenue. Our bookings for this quarter totaled over $100 million, as a result of multiple large deals signed across our product portfolio of leading-edge Exensio and secureWISE.

During the third quarter, our bookings were greater than the prior 2 quarters combined. On a year-to-date basis, for the 3-quarter period, our bookings were 49% higher than the comparable period of last year. With the contracts John mentioned as well as additional business closed in the quarter, we ended Q3 with backlog of $292 million, which is 25% higher than last quarter and 22% higher than the same period a year ago. We are pleased that we were able to grow our backlog while delivering record quarterly revenue. Our total revenue for the Q3 period came in at $57.1 million or 10% higher than last quarter and 23% higher on a year-over-year basis. Our Analytics revenue came in at $54.7 million or 12% higher versus the prior quarter, and 22% higher on a year-over-year basis.

The growth in Analytics compared to the prior quarter was driven by business from leading-edge customers and equipment software. Integrated Yield Ramp revenue was 4% of total revenue in Q3 and was lower by $0.5 million compared to the prior quarter and up on a year-over-year basis by $0.8 million. On gross margins, we reported 76%, or slightly ahead of last quarter, and down 1% versus last year’s comparable quarter, which had meaningful perpetual software revenue in that quarter. As you will recall, our long-term target for gross margin is 75%. We’re pleased that we were able to be ahead of that target for this quarter. Our operating expenses in Q3 grew 3% compared to the prior quarter, primarily due to spend related to development improvements for our platform and increased variable compensation accruals due to strong results.

On EPS, we were able to deliver $0.25 per share for the quarter, our strongest quarter for the year. For the first 3 quarters of 2025, our EPS of $0.64 is now $0.06 ahead of the comparable period of last year. We generated positive operating cash flow of $3.3 million this quarter and $6.7 million for the first 9 months of this year. We ended the quarter with cash, cash equivalents and short-term investments of approximately $35.9 million compared to the prior quarter’s ending cash balance of approximately $40.4 million. We repurchased $0.2 million of our stock this quarter at a per share price of $19.55 per share. During the quarter, we invested $6.3 million in CapEx, which is lower than the $8.5 million in Q2 and the $8.2 million in Q1 of this year.

2025 has been an important investment for us, like John said, as we use significant cash on the acquisition of secureWise and related integration expenses while only benefiting from a partial year of ownership. During the year, we also invested in building eProbe machines to meet customer demand in 2025 and 2026, without the benefit of full subscription run rate return on the investment within the year. Now with 2 additional machines shipped and going through qualification on a subscription model as well as the integration cost of the secureWISE acquisition largely behind us, we anticipate cash to grow over the next year. Given the strong business activity, the growth in our backlog and the customer opportunities in front of us, we reaffirm our prior guidance of 21% to 23% annual revenue growth range for this year.

As we get ready for our Analyst Day and user conference on December 3, we look forward to sharing more details about our long-term targets for the next phase of PDF growth with you at that time. We are also thankful to our customers and partners for supporting the growth we delivered this quarter and look forward to growing sequentially again in Q4. With that, I’ll turn the call over to the operator to commence the Q&A session. Operator?

Q&A Session

Operator: [Operator Instructions] Our first question comes from the line of Blair Abernethy from Rosenblatt.

Blair Abernethy: Nice quarter. I just — I wanted to just ask you a little bit about the BFI. I see that [indiscernible] more machines. The machines that are under the lease model, when does that start to generate revenue? Is that some point next year? Or is that first half of this year?

Adnan Raza: Yes. We are going through the deployment and qualifications of those machines. Blair, as you know, those can take anywhere from 1 quarter or a little bit plus or minus on that time frame. So given that we have shipped, we expect within the next quarter or the quarter after, depending on the timing of those qualifications and the customer acceptances to start converting and generating the revenue.

Blair Abernethy: Okay. Great. And how does the pipeline of opportunities look for the DFI right now?

John Kibarian: Quite strong, actually. This is John, Blair. We do have other places where we would like to be able to ship machines. That’s why if you did notice, we did spend some on CapEx this quarter in part to continue building machines, which we expect shipping in the first quarter of this coming year. We are hoping to squeeze in one more shipment this year, but it may be tight, just given the timing and what we’re doing to bring up the machines already. So quite strong across a handful of customers. It’s not a huge market for eProbes, but there’s probably between 5 and 10 customers in the world, and we do have probably closer to 5 where we are actively engaged in discussion.

Blair Abernethy: Okay. Great. And then just on the secureWISE, you won a large contract there this quarter. How is that — how is the go-to-market there now that you’ve had it for a couple of quarters? Just kind of a sense of how that is building.

John Kibarian: Sure. Yes. We actually had — at SEMICON, right after SEMICON last in Phoenix this year, we had a little, what we called Connected Summit because in the past, the secureWISE team had had our users conference in conjunction. What was different about that was we had not just equipment companies come, but also fab companies attend. And in that conference, Intel presented how they are using secureWISE as their standard connectivity platform for both internally as well as to support the equipment vendors. We had felt that the way secureWISE had been run, it really only focused on the needs of the equipment vendors. And when we — because of our DEX network, because of our work with the fabs and our — just our general footprint, which, as I said in my prepared remarks, goes from everything from wafer makers through to system companies, we thought connectivity to the fabs was actually desired by lots of folks.

In fact, when we announced that acquisition, the first congrats I got was from one of our largest fabless customers, who was very intrigued with the ability to get remote access at the OSATs and fabs. And so the Intel contract was a way of us, saying, okay, we’re going to provide base capability on every machine at Intel. They announced that they’re going to put this on every machine front end, back end and test facilities for their use as well as to make it available for some modest level of usage at every equipment vendor. When we talk to the equipment vendors that use secureWISE, one of their elements was not secureWISE itself, but the fact that they were not able to get it on every factory in the world, even if it was actually already installed at some factories, that factory may not give a specific equipment vendor access, maybe only the largest equipment vendors in the world typically got access at every factory in the world, and the smaller equipment vendors didn’t feel like they were getting access, but they would much — very much like it.

They knew it makes them a lot more efficient, both for human-level collaboration as well as for AI-driven collaboration. So now that we’ve had this company in our hands for the product in our hands, for maybe 7 months now, if I think about it, what we started doing is selling it much more broadly into the fabs as well as into the equipment vendors, creating collaboration across them, which is what I talked about at SEMICON West. And we started piloting it at the OSATs and really merging it with our DEX network because the security and some of the features that secureWISE enables are very desirous, more broadly. So it becomes an integral part really within the first quarter, we were selling combined contracts, if you look at that first contract that we announced in Q2, which is really now that it’s been announced, was effectively the Intel contract.

And we see a lot more of that coming down the pipe.

Operator: [Operator Instructions] Our next question comes from the line of Clark Wright from D.A. Davidson.

Clark Wright: Quick question just around the customer concentration mix. Your Customer A that you guys referenced year-over-year went from 19% to 38%. I’d love to understand kind of how you’re winning bigger and as well as how you’re looking at using secureWISE as a potential point of the spear to expand the overall customer base?

John Kibarian: Yes, that’s a great question. If you look at our business, really, we kind of think about it in 3 categories. There is the fabs. They tend to buy almost everything from us. If you look at just the discussion I had around secureWISE as well as the test vehicles, the eProbe, Exensio, et cetera. And those are very large contracts, typically multiple contracts with the same account. And hence, you see the customer concentration. It’s really not often of one contract, but of many contracts that gets those customers. They represent typically between 40% and 50% of our business in any given quarter, just looking at how things work. The fabless and system companies are around 35% to 45% of our business, and that’s about 100-something, 150-ish companies.

We are seeing more interest in AI on Exensio, the Exensio Cloud, the secureWISE connectivity and some of the what we call orchestration products, Sapience products. There, we’ve had a number of wins over this past year and continue to drive business there. And then lastly, about 15%, and we think in the long term, closer to 20% are the equipment vendors. We have about 200 of them with the acquisition of secureWISE that certainly grows our business with those customers. And they have access — they have desire for the same access points that the fabless did. So the way we think about it over the long-term clock, you can think about fab customers as a nexus point. And they run factories and they control the data, but they need collaboration with their customers to get qualified, and their equipment suppliers to be able to reach effective use of the machines that they purchase and put into use in their factories.

And so secureWISE, as you pointed out, is the point of that spear because it allows us the collaboration across a number of those customers. And the Exensio contract that we did this quarter, it has an element with regard to secureWISE because they want to be able to reach their customers through secureWISE on the Exensio platform from a collaboration standpoint. So this is how we see PDF becoming a platform for the industry rather than a platform for each individual company. And secureWISE is a very important element to that.

Clark Wright: Awesome. Appreciate that color. And then just as it relates to the announcement made in the end of September around the landmark contract, you reaffirmed your guide for this year. Is there anything we should be considering as it relates to kind of the 2026 picture and what you can say so far around how that deal potentially sets up the company for kind of the next leg of growth?

John Kibarian: Yes. We haven’t given guidance for 2026 yet, and we’ll do that as we get through Q4 and we do our Q4 call. But obviously, as we grow backlog, you asked a question about other, we do see a number of other opportunities on the horizon for the company over the next couple of quarters. We hope to have a strong 2026 on top of a very good bookings 2025.

Operator: Our next question comes from the line of Gus Richard from Northland Capital Markets.

Auguste Richard: I’m just curious on the systems you’re sending to the production site. How many tools per fab do you think the customer is going to need?

John Kibarian: Yes. I think it’s early to say, Gus. Obviously, we’ve put two in the first site. Two is a good number because at least as these things are used in mission-critical manufacturing, if one were to go down, you would want to be able to at least route critical material to the other. So I think very rarely would it be one? I think the minimum number is two. And then the question is, with any inspection capability, how much of the dance card can you fill up? What we’ve noticed as we’ve installed machines around the world is it gets very quick to get these things filled at very high utilizations in part because they can see things that are very hard to see or maybe nearly impossible with other systems. So we’d like to go beyond the two, but right now, we think two.

Auguste Richard: Okay. So these are near production, but not necessarily in line, not — there’s not…

John Kibarian: Not in product, that’s in your words. No. I said the reason why they want to is if one goes down, they want to be able to continue production, right?

Auguste Richard: Okay. I just want to get it clear. And then of the — you have several systems going out end of this year, beginning of next. I just want to understand, are these evaluation systems or are they for revenue?

John Kibarian: These are mix. It will be a mix and then the next set of machines. There’ll be a couple on eval and the remainder will be revenue machines.

Operator: [Operator Instructions] We have a follow-up question from Blair Abernethy from Rosblat.

Blair Abernethy: John, I just wanted to follow up on the Hybrid AI studio. Can you just — you said there’s 100-or-so customers for that. I’m just wondering if you can give us some sense of the time line of when that can go to market with the Exensio platform, i.e., when is the integration sort of ready for customers?

John Kibarian: Sure. And I said actually hundreds of users. It’s actually a very, very small number of customers, Blair. And — but thanks for asking that clarification. We will have an integration at the end of this quarter. We actually signed this contract quite a long time ago, but due to timing of other contracts, we needed to wait before we could announce it. And we had had discussions with them going all the way back to 2024 around this opportunity, we had evaluated it in early Q1. As you can imagine, we were pretty busy in early Q1 because we’re also closing secureWISE and then signed contract towards the end of Q1, then executed some activities in Q2 and announced in Q3, but we do — so as a result, we’ve been working out with this code base for — since all of Q3 and into Q4, and we expect to release some first level of integration with Exensio at the end of this quarter for some early access customers.

Blair Abernethy: Okay. Great. Great. And then just on the Exensio Analytics business, what is the — what’s the renewal book look like as we kind of head into the end of 2025 here versus last year? And are you — and maybe in recent contract signings, what sort of any changes in the term length of Exensio contract?

John Kibarian: Yes. Typically, term lengths are 3 years. There are some that go as long as 5 and some that are short as 1 or 2. Our book is quite robust. The large contract we signed, the 8-figure contract we signed this last quarter was probably one of, if not the largest, stand-alone Exensio MA contract in the history of the company, as other larger contracts tended to have test operations or other components included in there. What we are seeing, we’re going to talk about this at our user conference. Customers really want to have a scalable AI-first analytics capability. We’re going to show our road map and what we’re doing to have kind of analytics with AI-first, and what that means in terms of parallelization, the advances we’re making around how to get to very large data sets interactively.

They want the human interactivity with data still, but they want to be able to operate on data where you’ve got 1 million parameters and 10 million data points. And you really can’t do that with conventional business intelligence tools, right? If you look at Exensio or any of the tools out there, you’re limited by the compute. And — we’re going to show what we’re doing to break through that problem. Studio AI is a very — is an element to that, a very critical element to that because even when you’re interacting with 1 million parameters, you need to use AI methods to screen and to tell you what part of that data set you should look at. And so what we’ll demonstrate in December is ways of being able to operate interactively, leverage AI first and move — work with data sets that you really couldn’t do anything, but a batch mode in the past by leveraging basically a native AI approach and a natively parallel approach.

You can think of a lot like moving algorithms from CPUs to GPUs, you get to scale with compute. What we will show is how you can leverage GPUs and other computing elements in an interactive analytics capability. And we are hearing from our customers that this is what is desired. We have a number of renewals that are coming up this year, and primarily next year that I think will benefit from this capability.

Blair Abernethy: Okay. Perfect. Great. Yes, excellent. The other question I just had was around Sapience. Anything to report there or any progress with the partnership with SAP?

John Kibarian: Yes. We’ve got a number of activities going on, on Sapience contracts. We do expect to announce something related to Sapience in Q4 in terms of customer — additional customer business. And you’ll see us announce something in the Sapience family at our user conference that’s really building on top of Sapience some capability targeted to the fabless and system company that we expect to announce at our user conference.

Operator: Our next question comes from the line of Clark Wright, D.A. Davidson.

Clark Wright: Awesome. Appreciate the time. Look, any findings from SEMICON West in terms of how the end markets and the health of those sound relative to the beginning of the year or your expectations?

John Kibarian: Yes. Clark, it’s a great point. It was an interesting SEMICON, I think, because it’s the first time SEMICON West was not in San Francisco, people couldn’t go in and out. You’re kind of stranded in Phoenix. So there was a lot more informal conversations. And we did meet with also a lot of our — not just our equipment customers, but I would say our fabless customers and a lot of our fab customers. And what we heard were a few things. Yes, the build-out on AI is continuing to go on. All the equipment customers participating in advanced packaging, everything around advanced nodes, on logic side do see and on the DRAM side, do see a pretty rosy outlook for 2026. I do think they’re in a pretty strong position. Our customers selling into automotive, industrials and communications, the ones with very differentiated products do seem to be talking about a robust 2026, I think that’s still a mixed bag.

There are customers in that sector that have some challenges to work through, but I would say for the first time, the kind of ones that have very differentiated products that are much more bullish. So I would say you start seeing kind of a more broad base of enthusiasm within the customer base, than, I would say, 3 months ago or 6 months ago, where it was really limited to just the people on the very advanced nodes and advanced packaging. So I do think it was more broad, not fully, I would say, not fully broad to everybody, but definitely more broad than where it was in terms of positiveness than where it was 3 or 6 months ago. And then lastly, I would say, our fabless customers as they are becoming more and more embracing advanced packaging are recognizing they’re becoming a manufacturer.

And that means ability to communicate with the OSAT, more complex test data feed forward and other test flows. We had a lot of dialogues with customers on that topic and how can they effectively become more aware of what’s going on in manufacturing. In the past, they would order a wafer from the foundry. And once they hit wafer sort, there wasn’t a lot to worry about. But now they’ve got everything from operationally make sure they have organic substrates available and manage the supply chain of that — of the production post the wafer sort as well as having many more test insertion points and needing to be efficient in how they leverage. And we’ve heard a lot of dialogue from customers in that regard. And I think that will be a growing area, as we move from just the very few high-valued, not terribly high-volume chips driving advanced packaging today to a much broader set of customers trying to leverage these advanced packaging test flows.

Clark Wright: Got it. And then in terms of Cimetrix and kind of the shadow backlog, last quarter, you kind of referenced the fact that tens of millions there, has that upticked as well this sequentially?

John Kibarian: Yes. As I said in my prepared remarks, we had a very strong quarter. We referred to it as revenue because the booking and the revenue happened in the same quarter on the runtime licenses with secureWise. In other words, we get designed in on the SDK and then they ship. What we noticed is, as I said in my prepared remarks, at the end of 2024, and we see it again this year, is more equipment is shipping with our software than with any of the proprietary software systems that the companies build. And I think that’s really giving our software the reputation of being very robust and very applicable. A lot of our equipment companies that used to be on the front end are now trying to bring in tools to the back end. Our software is already proven in back-end assembly facilities.

Our tester companies are doing much more sophisticated system-level tests with more robots, and our software is very proven with that capability, too. So we do see a fair amount of activity, design activity in some of the customers evaluating our SDK. Net, when you look at our runtime licenses, Q3 was very significant. It was a good quarter for us, a very good quarter for us. We don’t get a lot of visibility. So we hope to sustain that in Q4. We don’t have as much visibility because we only see it when they ship. But overall, while quarter-by-quarter may be difficult to predict on an annual basis, the trend is quite positive as more and more equipment ships with our software, and they use our software for more functionality, which means they buy more of the Cimetrix modules.

Operator: Our next question comes from the line of Andrew Wiener, Samjo Capital.

Andrew Wiener: I wanted to maybe follow up on, I think you touched on it a little bit in the last answer, but I noticed that 2 of your partners in the test space, Advantest and Teradyne both posted very strong results and outlooks. And you’ve talked in the past about Advantest and the complexity around test being a strong secular driver for the company. Can you maybe elaborate a little bit on what you’re seeing and how the opportunity for us? Is it coincident with — as they see strong shippings, does our bookings or opportunities lag theirs and sort of any other color you could provide?

John Kibarian: Sure, Andrew. Thank you for the question. Yes, you’re right, it tends to lag. We see that, by the way, with our yield ramp business over the years, too. And kind of my prepared remarks saying we’ve seen lots of people building out 3D production and test and assembly with — and now they’re trying to figure out how to get a good return on a lot of these investments. We see that with — that’s part of what’s driven our characterization vehicle and eProbe bookings over this year and a lot of our evals that are ongoing are folks recognizing they’ve got to get a return on what they’ve just put on the ground. So we would expect on the advanced test for advanced packaging, we will also lag our partners in that regard.

The biggest application we see, and we’ve got a number of pilots going on with customers is data feed forward. And what this means is they have — as there’s many more packaging steps, they have a lot of test insertion points. So they test more than once at wafer sort, more than once at final test and once at system-level test. There’s multiple wafer sort tests, multiple test points package, even sometimes as part of the package flow. And then finally, system-level test. And they do these at different temperatures and different conditions for these very large data center chips. The nature of that is they want to see forward data. Typically, they take the raw data, run an AI model, extract features and send it downstream to another test insertion point.

And as the data is coming off that test, they will use that as a basis to decide to test more or test less. Our original strategy on this, Andrew, was to not be in the modeling business at all, but provide them the infrastructure for kind of orchestrating the data up and down the supply chain. We’ve got pilots ongoing with that. We actually have a customer deployed using that on tens of tools now over a year in production across multiple OSATs. And even that customer has come back to us and said, “Hey, we need capability to be able to build and maintain the models, not just move the data around.” And that was really the Tiber studio — Tiber AI Studio, I forget the name, right. We thought our customers would be able to do that on their own, or there were systems to do that on their own, but the reality is there’s a lot of friction to getting that work done, too.

And so the integration with Exensio Studio AI with Exensio ModelOps is to help them not only run the model in production, but manage the model through the build and through the life cycle in their central servers. So I don’t think this is a 1-quarter bang and everything is going great. Andrew, we’re going to see win by win by win with customers, but I would say we have a handful of pilots going on in data feed forward at this point. We’ve got some in production already, and we do anticipate that to becoming an increasingly important part of our Exensio test business.

Andrew Wiener: Okay. And would you think that would be like a 2026 sort of time line?

John Kibarian: Yes. I mean, the majority of that business impact will be in 2026, to be honest. This year, there may be some additional contracts won, but the revenue impact will be de minimis.

Andrew Wiener: And then following just — I want to clarify something. So the 2 tools — 2 DFIs that were shipped and are in the process of qualification, it sounded like the way you described it, they could get qualified this quarter, or maybe not. Is it then fair to say that, Adnan, I think in the past, you’ve talked about multiple ways to get to guidance. You guys feel comfortable with the guidance, even if there — these tool qualifications slip into Q1?

Adnan Raza: Yes, Andrew, exactly. I mean, any time we’re looking at a quarter, particularly when we’re speaking to comments such as the ones that we put in my prepared remarks about, sequential growth for Q4, and also both in John’s and my remarks about the reaffirmation of the 21% to 22% guidance, you’re absolutely right. We’re thinking about this and then other ways to get there. So look, if the timing happens this quarter, great, but like John said, there’s many other opportunities we’re working on across the product platform portfolio that we have.

Andrew Wiener: And then maybe…

John Kibarian: this is a little bit our timing. The timing on qualification would be towards the end of the quarter in any case. So the in or out is not a tremendous amount.

Andrew Wiener: Okay. So again, the qualification of those tools regardless would be more of a tailwind to 2026 versus the earlier question about, correct. And then maybe just a little more color on — I mean, obviously, you talked about rough engaged with 5 customers or potential customers on DFI. How — is there memory customers in that bucket? Or are we still primarily focused on logic? And to the extent it’s logic, is it — you have 2 sort of other customers that have accepted tools? Are they looking at what your lead customer has done and the conversations around sort of a similar broader deployment along those lines?

John Kibarian: Yes. It’s actually all of the above. So with our existing customers, we see interest in more machines as they see the value, including the value of putting these in manufacturing and using these to monitor lines. They — with new customers that are in the same areas our earlier customers, maybe at different feature sizes, but logic manufacturers. We do — we started getting ongoing pilots with customers where they’re sending us wafers and showing them what you can do. And it is a very unique capability. So we’re able to show usually within a wafer or a few what you could see that’s hard to see elsewhere. And then lastly, as I’ve said throughout the year, we have interested in doing pilots on DRAM for just about 12 months now, maybe 13 months.

And you got by sending wafers to our facility here in California. And we’re getting ready to be able to ship. I think the limiter on shipping is really us not at least one of those customers being in a position to ship given what we’ve been doing to bring up these machines and manufacturing this year. And at the engineering level, there’s a lot of similarities to what we’re doing in the logic side, but it’s very, very different process technology, if it’s memory versus a logic technology. But also exploiting the unique capability of the machine and the software.

Andrew Wiener: Okay. And then just lastly on the Tiber. So just to be clear, like — so it didn’t — we didn’t come over with existing revenue. But in the press release, it said something about like supporting customers through this transition. Anything you — sort of that contributes will be going out and selling once the integration is complete?

John Kibarian: Yes, that’s correct. The majority of the customers were not in semiconductors. Some have interest for us to support. And we may — we are talking to only a very small handful about having them be supported on the stand-alone version of the product. Obviously, we licensed this from Intel. Intel itself was an internal user of the product. And for them, we will support it — offer to support it both ways, both stand-alone and with Exensio. There’s a lot, a lot of value getting it integrated in Exensio for a semiconductor customer, because it didn’t really support any visualization of the data, the model, the results, just really visualization on running the models and algorithms. So Exensio and you didn’t have a database for storing the data set you want to use.

And then once you’re done, you want to be able to store all of that information, so you could always go back and recreate it, and Exensio has capability for that model registration and things like that. So I think for that — the user base, which is primarily at Intel, that was in semiconductors and using it, I think it’s quite advantageous for them to use the integrated version. For the small number of non-semiconductor customers, they may use a stand-alone version.

Operator: [Operator Instructions] There are no further questions. Ladies and gentlemen, this concludes the program. Thank you for joining us on today’s call.