Good day, ladies and gentlemen, and welcome to the PDF Solutions, Inc. Conference Call to discuss its financial results for the third quarter ended Sunday, September 30, 2018. At this time all participants are in listen-only mode. [Operator Instructions] As a reminder, this conference is being recorded. If you have not yet received a copy of the corresponding press release and management report, both have been posted to the 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, anticipated cost savings, growth rates, collections and demands for its solutions. PDF's actual results could differ materially. You should refer to the section entitled Risk Factors on Pages 13 through 19 on PDF's annual report on Form 10-K for the fiscal year-ended December 31, 2017, and similar disclosures and 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 Christine Russell, PDF's Chief Financial Officer. Mr. Kibarian, please go ahead.

Thank you, and welcome, everyone. If you've not already seen our earnings press release or written management report, please go to the investors section of our website where both are posted. Today, we will discuss events that occurred during the third quarter 2018 and their impact on our business and its evolution, as well as our view in the general semiconductor economy environment. By way of background, I will start by saying that, in 2014, IBM selling its microelectronics business and its alliance on foundries shifting their working model, was a catalyst for us to transform our business. We believe that moment was a harbinger of a number of change in the industry: one, there will be much less competition on the leading-edge; two, the rate of new nodes would decelerate; three, fabless and system companies would need to learn how to squeeze more out of their manufacturing technology and supply chain; four, solutions for the leading-edge customers would only be relevant if we have proprietary unique data sources that are critical for ongoing manufacturing; and five, when one competitor disappears, new ones would emerge and China would be the new source of investment. Over the last four years, as part of that transformation, we made significant investments in developing DFI and expanding our Exensio business, both organically and through modest acquisitions to make Exensio more valuable to OSATs, fabless and system houses. The primary objective of the transformation was to change our business from bringing up new nodes to controlling ongoing manufacturing for yield, reliability and overall operations efficiency. Secondarily, we drove to diversify the customer base by making our analytics and data valuable across the supply chain from foundries and OSATs to fabless and system companies. GF's recent decision to abandon 7-nanometer has accelerated how fast we need…

Thank you, John. Most of you would have seen our financials in our earnings release. In addition, we posted in the Investor Relations section of our website a management report with financials and comments regarding the results of PDF for the quarter. So I'll focus my comments on a few key areas. Revenue for the third quarter was $20.2 million, a 4% decrease from the prior quarter. Solutions revenue decreased quarter-over-quarter by $1.3 million and gainshare revenue increased by $400,000. The decrease in solutions revenue was primarily a result of lower hours spent on IYR projects, only partially offset by increased sales of Exensio solutions. Exensio revenues benefited from the commencement of a time-based usage license for the largest China-based deployment of Exensio at a major IDM. The increase in gainshare revenue was primarily at the 14-nanometer node. Cost of sales were basically flat at $10.7 million, down 2% quarter-over-quarter. Our cost of sales includes many fixed costs such as engineers not only charging time to revenue-bearing projects, but presales activity as well. Total operating expenses at $12.4 million decreased by around 6% compared to the prior quarter. Compensation was lower as we managed down our employee base through voluntary attrition and slower hiring. Travel expense, cost related to subcontractors and financial services were also lower. We are reviewing our services contracts, including financial services, and have already made changes, which will result in $700,000 of annualized savings. We moved our headquarters to a lower rent location in Santa Clara in August, which will result in an annualized savings of $252,000 per year. We're examining all the costs of the business negotiating with suppliers and targeting that monies are wisely spent. As a machine development for DFI was largely completed during 2017 and we shifted the lower cost or Exensio-based…

Thank you. [Operator Instructions] Your first question comes from the line of Jon Tanwanteng from CJS Securities. Your line is open.

Hi guys, thank you for taking my questions. Can you repeat the impact of the GLOBALFOUNDRIES movements in Q3, if any? Did you quantify that?

Well, what we said was that we expected the fourth quarter to be impacted.

I think we also said we'd stop work partway through the – reduced work partway through the third quarter as we're notified by the customer, and it had an impact on the third quarter. It was a modest impact on the third quarter compared to the impact that we'll have on it.

Yes. And I will repeat for you that we expect that quarter-over-quarter, $2 million of revenue primarily will be a – reduction is a result of stopping work on the GLOBALFOUNDRIES project. But I also qualified it by saying we are in discussions with the customer right now. And if we reach an agreement, it could be different than that.

Okay. Got it. What kind of discussions are you having with them? What agreements are you counting for? What kind of services can you still supply to them as they rotate away to seven and to derivative nodes?

Yes. So obviously, we're not going to talk about specifics of any discussions we have with a specific customer. But if you listen to their call, like many of the semiconductor industry, they are trying to bring up derivative versions of the technology, derivatives like high-voltage or RF or embedded non-volatile memories. Most of the – we have a number of Exensio-based contracts where customers get a combination of vehicles and Exensio. For things like phase change memory and other nonvolatile memories, high-voltage technologies, et cetera, all those things would be just as valid for them on any other number of nodes that are valuable for the customers that we have today.

Okay. Got it. And then, Christine, you talked about the streamlining you're doing a reduction in force. Any ballpark that we can look to in terms of an annualized cost reduction? I know you're not quite done with that yet, but I think the majority of this cost you're doing a restructuring. Is there anything else in there that we can use to kind of guess what it would be like?

Not really. One of the reasons why we wanted to wait and discuss this on next quarter's call is there are still some items to be determined. What offices we might not need to use anymore, which locations we are addressing. And also, some of the severances are negotiated severances. And so until those are completed, we just really can't come up with a good number. But it will be several millions of dollars.

Okay. Thanks, and then just on the gainshare side. You mentioned that the majority of sequential increases was in 14-nanometer. Was that due in part because GLOBAL stopped investing in seven or was there something else going on there? And kind of what can we expect going forward in the fourth quarter beyond? Do they ramp-up more or is this kind of the steady state we should be seeing?

Yes. This is John, Jon. Yes, so generally speaking, volumes were up in 14-nanometers that drove that amount. I think when you stop doing a development program, one development wafer, typically, is replaced by more than one production wafer. So we would anticipate customers that stopped development in a line would drive up there existing node volumes and they have – customers that we've discussed have expressed that to us. How quickly they can do that and what the end demand is, is of course hard for us to forecast. And as you get out past a year or two, their ability to keep factories full is always something that we watch very closely. That ability to keep a mature factory, following the foundry business is clearly an art, not a science and some do it better than others, so we'll be watching that. In the short term, it's should be generally better.

Okay. Great. And any movement at 28 at all or any signs of life in that market?

The foundry leader made a call this quarter, basically. So they thought it was going to take a couple of years to swallow up the glut of capacity on 28-nanometer. I think they are, by far, the most sophisticated overall of the supply chain in understanding the best visibility, by far. If you look at the words in their script, and we study that pretty closely, they talked about the need to bring up derivatives. In other words, it's not going to backfill with just vanilla logic technologies, but high-voltage applications, RF applications, embedded non-volatile applications. All those are wonderful opportunities for all of the second-tier foundries, who will need vehicles to characterize them, because what is different about these flavors is the electrical characteristics. And we would argue the best way to characterize electrical behaviors, silicon is our vehicles and systems. So I don't think – they were very clear that they didn't think it was a short-term rebound and they think it's going to be off derivatives. We have been pointing that out to the customer base overall because we think all of them need to listen to that.

Okay. Great. And one last question. The change at GLOBAL, does that mean that they would not be a customer for your DFI systems in the future or would they still be able to use the 12 and 14? Are they interested in that?

Yes. So I think for our lead customers, particularly on the fabless side, what they really like is the ability to understand, at layer, what the intrinsic quality or liability of a wafer of silicon is. Particularly as people try to develop derivatives for things like ADAS, high-performance computing, automotive applications, this information is quite useful. And the customers that are – besides those two, there are others that are putting on their test chips or are looking at putting them in their products. They're all folks that are trying to do large chips for AI, ADAS, automotive applications, et cetera. If you listen to GLOBALFOUNDRIES as well as other customers, what's going to go into 14-nanometer and 16-nanometer of this next wave is a lot of those kinds of products. So, there remains the technical reason why an application like DFI would be valuable for a company like that.

Okay. Great. I have one more last one from an investor, if you don't mind. You had an Investor Day about two years ago. You were very excited for the part future of Exensio and DFI. Maybe that sentiment around gainshare and the solutions business was probably a little bit better than its now, obviously. Just, can you give us a high-level overview of what happened? What's taking so long and kind of what investors should expect from that time line going forward, and especially as you get into this new Investor Day next year?

Yes. Yes. It's a great question, Jon. So yes, when we looked at that, we've heard two things. We knew China was going to be late at bringing up their volumes and we handicapped it about four quarters off what they were thinking. The reality is the volume ramp has been much slower than that for a variety of reasons. Some of which is, I think if you go look at the leaders comments about the need to do derivatives in order to fill 28-nanometer, I think they could heave that set of words quite well. In other words, developing a vanilla version of 28 at this stage and bringing it out, probably isn't going to fill out the factory these days. You're going to have to have it come up with a non-volatile or high-voltage application. I think we didn't forecast how much behind they would be in terms of generating customer demand. And secondarily, I would say that Chinese consumers of silicon, while I think they always want to support their country's policy towards in-sourcing, they also want to run a practical business. And a practical business that is going to be qualified suppliers in the world. So the demand is not shifted into China as much as I think we would have anticipated. Number two, while we did believe at some point folks like GLOBAL would stop driving the leading-edge, we knowing that particularly, AMD was a very important customer to them, and what the future we saw AMD have on server and high-performance computing, particularly if you look at the press right now, we had anticipated them staying in this round because of the opportunities that sat in front of them, AMD in particular, but in general, high-performance computing. That, I think, we really missed. We…

Great, thank you. That was very helpful.

Your next question comes from the line of Tom Diffely from D.A. Davidson. Your line is open.

So, yes good afternoon. So when you look at the guidance for the fourth quarter, it sounds like the solutions drop-off only partially offset by perhaps a little more strength in the gainshare. So, would you expect to kind of, I guess, drop into the negative pro forma earnings due to lost node for the first time in quite some time?

We're not forecasting that, Tom.

Yes. I think, we were not – yes, it's hard to forecast that, Tom. But I will tell you is the following: if you go to look at our history of the business, there are very, very few times where we lose money, and we don't lose money on extended time period. Even if we were to lose money in one quarter, it will not be a sustained thing. That's not how we run our business.

Yes. Okay. I mean, do you have any sense of what the cash burn would be in the quarter, I mean, taking into account some of the restructuring you're doing?

Well. Yes, we know we right away that the restructuring we gave a range of those numbers, which was in the range of $1.3 million to $2.3 million, and so we will have that cash charge. We will not have additional charges, at least not substantial, related to the building and the tenant improvement. However, we will continue to use cash for the build of the 250s.

Okay. And speaking of the 250, I mean, do you have the order in hand at this point? Or what gives you the confidence that it'll finally get signed this quarter?

Yes. We are exchanging contracting papers back and forth with the customer now.

Okay. Good. Then based on that and the shipment in the first quarter, what is the timing of either revenue or maybe a build-out of a more production suite of these tools in the field?

Yes. So we will see revenue relatively quickly. There is continued use of the 150 machine before the 250 is up and running. So there would be value customers we'll be achieving even before the machine is up and running and then we'll – slot machines for them. In terms of building out additional ones, we've built out a small number of them, a couple from a development standpoint. And as Christine said, a little bit of additional money spent to build out additional machines. What we're really doing right now, in my prepared remarks I think I said we are working with lead fabless – couple of lead fabless customers and the lead manufacturer who is supporting these companies on what the use cases and applications will be. We are communicating with them what our lead times are for these things. But we will look at how they respond with the applications that the fabless are asking for, and then we will push the button on what we do about additional builds out over – after that first handful of machines.

Okay. Will this be a situation where the fabless customers tell the foundries to buy or would you actually sell it to fabless to install with the foundry?

Yes. Tom, our initial business has been to deploy the IP at the fabless and then provide that capability to the fab to collect that data. Over the long term, how it evolves, we're going back and really looking at that. In my prepared remarks I said, there would be much less competition on the leading-edge. I think in the second and third quarter, one foundry was clear that they had 100% market share at 7-nanometer. What that does to the industry overall, over the long term, is that it would sustain itself for multiple years. I think it really changes the way where the technology risk is taken. With the multiple manufacturers, the risk is on the manufacturer and the fabless has a choice. When you get down to a single manufacturer over on extended period of time, not just over six months or a year, but over an extended period time, you're going to have to go and ask yourself the question, who's taking the technology risk at that point? I would argue over that time period, it ultimately be the fabless companies who will take that technology risk. And that may change where we turn out to monetize it, but that's kind of my speculating and thinking about the long-term future. But we're not doing anything in the way we're setting up our contracts that preclude us from shifting or adjusting should it be appropriate to do that.

Yes. Okay. That makes sense. And then shifting over to kind of the really bright spot, the Exensio software. Looks like it's a $8 million a quarter run rate at this point. You talked to a lot of penetration into different fabs and OSATs, system houses of different kind. What is the kind of the future growth potential look like from here? I mean, how saturated is the market that you're serving and what do you think you can do on the future years?

Yes. So I think just a couple of things, Tom, on that. So I think in my prepared remarks I said Exensio-based revenue was over 60% of solutions revenue. So you should go back and check your math on your estimate. And then, secondarily, where do we see the future and growth? What's been really super about this, we've turned this more and more – PDF has always been a data business, and what we're really doing with the Exensio is making it a data business. That doesn't mean that you consume the data through a PowerPoint that was created by a set of consultants or project team members, which is how we ran the yield ramp business. Christine's remarks she said, hey, more cost-effective based Exensio deployment because, by and large, the customer consumes the data via the analytics provided by Exensio, not the PowerPoint provided by a team. The neat thing about that is we have a tremendous amount of algorithms and systems that we built over the year that we're able to bring out to the customers. What we see is customers, particularly at the fabless and system basis, wanting that control of their manufacturing from – and operations efficiencies, but also a yield and quality. More and more growing of quality and reliability factor. And we think that's really where the communications – and by the way, DFI is just another data source for that. When we look at the customers that are driving DFI, they're interested and then using the DFI data and Exensio combining out with downstream package and downstream system information to understand quality up and down the supply chain. And you'll see us continue to enhance what we can to in that regard.

Okay. That make sense. And then final question on the model itself. Where is most of the cost reduction is coming from? Is it SG&A or R&D?

Yes. I can answer that. So I think there's some natural decrease in R&D because, over time, we've spent less with third-party suppliers on DFI hardware development. Majority of the cost reduction, as we look at yield ramps in the west, let's say, U.S. and Europe, we don't believe there'll be very many people bringing up advanced nodes in U.S. and Europe. And so we've taken on a fair amount of costs associated with being able to serve customers in that region of the world, and that's where you'll see a relatively meaningful reduction in cost. As Christine also said, on G&A, we've gone back and looked at our cost there as well, and we do see ways that we can improve cost there.

Yes. Tom, we are looking at everything. And one thing that naturally falls out if you have a lower number of employees is your travel expense goes down. Travel expenses probably our third or fourth largest expense that we have in the company after compensation and subcontractors. And so while you're going to have a natural lowering of travel, just simply because of a lower number of employees traveling, we also have undertaken a project to actually focus on travel cost, make sure that we are traveling wisely and not traveling too frequently, and that the number of employees who are going to any one location is a number that is appropriate for the opportunity.

Okay, makes sense. Thanks for your time.

Your next question comes from the line of Gus Richard from Northland. Your line is open.

Yes, thanks for taking the question. I hear you're in the midst of negotiation. Hi, John is the business model or arrangement with your lead customer going to be similar to prior arrangements or does the business model shift at all?

Yes. I won't speak specifically about that customer, but I will talk in general about what you've been doing. So, we started noticing in that 2014 time frame, customers that had relatively matured nodes, 90-nanometers, 65-nanometer, et cetera, who wanted vehicles and Exensio for improving their manufacturing control. There is no time to volume. It's already at some volume. And there isn't an thing obvious thing to gain share against. So we moved to a model where they paid us for a number of tool connects. In other words, how many tools are in that factory that need to connect to Exensio, how much information they process. And then the unique data sources that PDF provided, how to pay for those. And in some cases, when they were using the systems to bring up derivative nodes, let's say, in an embedded non-volatile memory or something, they volume scale on the license, depending on how many of those wafers got processed. It may very well be the case when a company pivots to focus on trailing it's nodes. That is a better business model for us to move to. And in those cases, in those customers that we deployed them like that, and its – of the top 10 Exensio customers, a couple of them are this way. The customer uses the vehicles themselves, but there's no time component to it so we don't provide this team to help them squeeze the information out in time to make a product release. Going forward, it may be a large portion of the fabs that we service will be on contract service as well.

Okay. And then any thoughts to the TAM for DFI at this point? Any method along those lines that you can provide will be helpful?

Yes. It's a great question. And we're still figuring out – roughly, in comparison to what we showed in the past, it's still in that same range. When we started showing the results to the fabless customers, they ask if we could inspect every wafer, right? And typically, inspection is done in two wafers per lot. And the way fab is set up, you inspect two wafers per lot, the wafers have – the lot has about four hours, so you have about two hours on the wafer and you've got to show whatever information you can show in two hours. What the customers are asking for is if you could be able to see a lot more of that from a control standpoint and of course, the way you look at metrology data. We're obviously not there yet that with the 250. All of our original 10 estimates, we're assuming that it was used more like yield inspection tools not like a control tool, like a per wafer control tool, like an overlay tool, for example. So as we get through these next few months to 6 months, I would say, with the lead foundry and the lead fabless customers, we're going to have a better understanding of how we would adjust the TAM for the product.

Okay. And then just out of curiosity, what do you believe your cycle time will be for manufacturing?

That's a great question. If we close our eyes and behave like software guys, it will be quite slow on the order of over nine months. If we're thoughtful about how we work with our supplier industry and what we look at in terms of critical elements, we can bring that down meaningfully.

Okay. And your existing FASE is at this point?

We will have a handful of 250s produced by the end of the first quarter next year.

Okay. So, I can count them on one hand.

Yes.

Okay, I think that is it from me. Thanks so much.

Thank you Gus.

Your next question comes from the line of Andrew Wiener from Samjo Capital. Your line is open.

Hi, good afternoon John. I wanted to clarify a few things. First, with respect to Exensio, you talked about it being greater than 60% of the solutions revenue. I assume that was for the quarter on a run-rate basis and not year-to-date?

It's actually both.

Okay. Second, can you then tell me what percentage of Exensio or ballpark what percentage of Exensio revenue is of a subscription or time-based nature versus sold on a perpetual license basis?

Yes. There are still some perpetual license contracts, but by far, the minority – the majority of it is time-based licenses or maintenance and support of existing older perpetual licenses. The majority of it is ratable. The vast majority of it.

Okay. So it would be fair to say that 80% to 90% is ratable and the remainder, perpetual?

Probably, yes. Probably 80% or greater.

Okay. Second, is there any – whether it's revenue year-to-date or bookings, can you give us a rough sense of what sort of the current growth rate is?

Yes. So as I said, over the last four years, it more than doubled. When we got on to the analyst call in Q1, we'll probably provide more granularity across that. But you can go back and put a growth rate to that. That was relatively uniform across those four years.

Okay. So something north of 20%?

Sure.

Okay. Can we maybe then also talk about what types of customers and applications you're seeing the strongest demand for Exensio today? And when you sort of look at over the next 18 to 24 months, where you see the biggest opportunities?

Yes. I don't think – know if I can answer that. When we look at 2014 or 2015, 2014, the pipe that I thought we totally missed was the fabless community. So we bought some assets, I believe in Q1 of 2015, if you go back and look at our SEC filings, of a company going bankrupt and then rebuilt that stuff on the Exensio platform. What we found was when we did that, the first wave of customers that drove that growth was the asset-light IBMs, primarily in the mixed-signal market, and they had their own test and assembly facilities. We subsequently bought a technology called ALPS from a company called Kinesis, I think in the summer of 2017, that incrementally added capacity that was really useful for those companies from the mixed-signal world because they make sensors. And often in those packages, there is a sensor that has no chip ID and there's also other elements. And the end customers, particularly in the phone supply chain, we're demanding transparency across that assembly flow. So I would say, that has been a very important part of the growth of Exensio over this time period. Subsequently, in 2017, we announced that we had signed a contract with Exensio for a large fabless flash system company in China to deploy Exensio's, call it, OSAT community. And that's why in my prepared remarks today, I talked about how, now, Exensio is available at – this year, we hooked up an additional seven OSATs, and I think almost 1,000 tools as a result of that. What we're finding now is that capability we were bringing to those asset-light companies who control their supply chain, the fabless communities also are interested in getting, too, and they don't control their supply chain. So what drove the business in the past will continue to drive a broader collection of customers in the future. Moreover, in China, I would say the majority of the business growth has been at fabs because they're still building a very big capacity there. So what we call Exensio-Control and Exensio-Yield. There's been a lot of uptake in that part of the market. So the IBM that Christine and I talked about as well as most of the foundries in China have some amount of Exensio in them.

Okay. I wanted to clarify something with respect to the comments about the six, seven – the lead customer who discontinued the 7-nanometer. I believe in your prepared remarks you suggested that even though we might stop recognizing revenue against the contract, there's cash obligations and that we've continued to bill or receive cash against them. Was that correct?

That is correct, Andrew.

So even if there is – so is the assumption there even if we don't come to an agreement in Q4 and there's this revenue drop off, there may be cash collection without associated revenue component?

Yes. The cash payments and the revenue recognition are different from one another. John remarked that the revenue was recognized on a percentage of completion basis, so that was dependent on the number of hours that were worked. However, we have billing milestones within our contract with the customer that permit us to continue to bill.

Okay. Okay. And I think you touched on it in response to, I think it was maybe Tom's or Gus' question. But to the extent that we would expand work with that customer around derivatives, is the nature or the arrangement likely to take the form that deemphasizes gainshare or unit-based royalty arrangements and focus more on this some sort of either data provided or information provided nature?

Yes. I mean, like I said when I answered the question for Tom, how it turns out on that specific customer, I'm not going to comments on. But in general, Andrew, we have customers where they pay on number of tools connected, the amount of data processed. Metrics that are around the information process, not on future volumes because these are nodes that exists in production. These aren't nodes that we're doing development on, let's say, something like 7-nanometer and it's not going to go to production and a couple of years. So charging you for the data processed doesn't recoup the value. These are things like 40-nanometer or 90-nanometer that are in volume now. So we can charge you for tools connected, which is the surrogate for data processed.

Okay. Can you talk a little bit about marketing efforts for DFI outside of the lead customer that you're in negotiations with on the 250? I think in your recent call, you had suggested that you would expand, you sort of – after taking some time off with regards to that, you'd sort of reach back out. Have – are we still trying to sell 150s or is the focus entirely 250s? And has that shifted the nature of the potential customer base that we're marketing to and sort of where do you stand in some of those efforts?

Yes. That's a great question, Andrew. Yes, so we have begun – in the third quarter, we started talking to more customers. One has recently shipped us a wafer or couple of wafers because they foundry for some of the fab. As we've talked about in the past, we have over 100 chips out there in the world, many – more than 100 chips that had DFI content on them. So many fabs have actually DFI in them. Some fabs have DFI wafers in them. We can get one of those wafers and show them what you can see. So we're beginning that activity in this fourth quarter. Other leading-edge fabs have now – we've been looking at what its – what we're capable of doing and we've begun dialing for them to send us material as well. So if that activity picks up, we expect that we will have more demo activity in this fourth quarter and ongoing. That's why when we have a handful of machines, we will keep at least 1 or 2 of them in our lab for future customer selling capability and incremental software development. With respect to the 150, we find it to be very useful for early R&D of the node and even some production control. And we anticipate for our customers, particularly in China, there's an ongoing need for that product. We have machines – a machine in China already, and we believe we could see additional machines going into that geography.

And is the emphasis still entirely logic or at one point we thought there might be some opportunities within memory. Has the recent sort of weakness in that area sort of created – on those marketing efforts?

One of the demos that we started to undertake in the third quarter was for memory itself, and there are other – we expect in the fourth quarter other applications where will begin some exploratory demos. On other areas we're measuring in line electrical data would be very useful like sensors, for example.

Alright, thank you.

[Operator Instructions] At this time, there are no more questions. Ladies and gentlemen, this concludes the program. Thank you for joining us today.