PDF Solutions, Inc. (PDFS) Q2 2016 Earnings Report, Transcript and Summary

Good day, ladies and gentlemen, and welcome to the PDF Solutions Inc. conference call to discuss its financial results for the second fiscal quarter ending Thursday, June 30, 2016. [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 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, growth rates and demand in solutions. PDF's actual results could differ materially. You should refer to the section entitled Risk Factors on Pages 12 through 19 of the PDF's annual report on Form 10-K for the fiscal year ended December 31, 2015 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 Mr. John Kibarian, PDF's President and Chief Executive Officer; and Greg Walker, PDF's Chief Financial Officer. Mr. Kibarian, you may begin.

Thank you, and welcome, everyone. Today, I will start our discussion with a brief summary of our second quarter results, then I'll provide more detail on our business activity in the quarter. Next, I will turn the call over to Greg, who will walk you through the financial results in detail. We will then take your questions. Consistent with the past several quarters, we drove strong bookings in the second quarter. We engaged with new customers for both our Integrated Yield Ramp and our Exensio Big Data solutions. We also broadened and extended many existing agreements. However, our Q2 gainshare revenue of $6.1 million declined by 6% when compared with Q1 2016, primarily driven by lower 14-nanometer production volumes. We believe this volume weakness was primarily the result of product transition at one of our major customers as well as softness in the smartphone market. At the 28-nanometer node, we did see mild recovery in the production volumes, and we believe that Q1 may have been the trough for 28-nanometer gainshare. As a result, we expect total gainshare for the second half of the year to be materially higher than the first half of the year. The overall weakness in gainshare revenue was offset in the quarter by strong solutions revenue, which increased by 10% over the prior quarter to $20.6 million. I will provide more details on the business activity that drove this in a minute. From an overall spending standpoint, during the quarter, we continued to step up R&D expenses related to our DFI initiative. This increased level of investment reflects acceleration in the activities related to the production of our eProbe 150 series systems and the continuing development of our second DFI system, the eProbe 250 series. Now turning to our solutions business in detail. We experienced another strong bookings quarter. In addition to broadening our business with our traditional foundry and fabless customers, we also continued to expand into new markets and customers. As an example, we booked a new deal with a major semicap equipment supplier to embed certain modules of Exensio in their hardware products, making our Exensio Big Data solution more readily accessible to customers in the future. The pipeline for Exensio Big Data analytics business continues to expand on a worldwide basis. Many of our customer engagements and deals in our pipeline now include or, in many cases, are being driven by demand for our new solutions around Exensio-Test. For example, in the quarter, we also booked our first major deal with an outsourced assembly and test house or OSAT, which we believe firmly establishes our presence in the test and assembly market. This expansion in both the breadth and depth of our Exensio engagements is aiding in our efforts to rapidly diversify our customer base. For example, during the quarter, we booked new or extended deals with a total of 27 customers. As examples, we closed on an extension of an existing Exensio agreement with a Japanese logic manufacturer out of mature nodes. We closed -- and we closed an Exensio agreement with a new Chinese foundry. In our Integrated Yield Ramp solutions, we continue to see strong interest at 14-nanometer, 10-nanometer and 7-nanometer nodes, and we're heaving early discussions with our customers on the 5-nanometer node. Overall, booking activity for IYR, including DFM, was very strong. In the quarter, we closed an agreement for a new 10-nanometer DFM engagement with a leading Chinese fabless semiconductor company. We closed an agreement for a 10-nanometer extension of an existing DFM engagement with a major fabless customer. We closed an agreement for a 14-nanometer extension of a DFM engagement with a major system company. We were able to extend existing multi-node Integrated Yield Ramp agreement because we met performance targets a quarter early. We closed an agreement for an extension of an existing 14-nanometer Integrated Yield Ramp engagement with a foundry. We closed an extension of existing 14-nanometer Integrated Yield Ramp with another major foundry. And we continue to see significant opportunity in China as the country develops an indigenous semiconductor manufacturing capability. As we've stated before, the large-scale investment by the Chinese government and private enterprises are fueling rapid expansion of both local foundry and fabless businesses. We had over 60 executives and key technical leaders from the Chinese foundry and fabless industry in attendance at our 10th anniversary celebration in China. We are thankful and honored that our partners and customers in China were able to celebrate PDF's long-term presence and unique place in the Chinese IC community. Finally, each quarter, I provide an update about our Design for Inspection initiative, referred to as our DFI solution. As a reminder, Design for Inspection solves the ever-increasing challenge of inspecting production chips for electrical defects. Conventional inspection allows semiconductor companies to see visual differences in patterns on a chip. However, many electrical faults are not visually detectable. PDF Solutions' DFI technology is designed to change the paradigm for inspection by placing small proprietary characterization vehicle test strips on chip in a product design. PDF Solutions' first eProbe 150 measurement system is up and running at the first customer and is already generating information for them. As you may have seen in our recent press release, in the quarter, we received an order for a second major logic foundry for another eProbe 150 system. This DFI system is scheduled to be complete -- scheduled to complete deployment at this customer in Q3. As we stated in our initial guidance, we were not expecting revenue from DFI until late 2016. However, our customers are finding the eProbe 150 system to be valuable for their research, development and ramp efforts and are eager to leverage the capabilities of the system. Thus, we now believe that there is a market opportunity for DFI in process R&D in early ramp. As a result, we are putting in place the capacity to deploy more eProbe 150 systems. Given the fabless design activity that I'll talk about later, we continue to see value in new product introduction and process control applications. And as a result, we are also working to pull forward the development of our second-generation DFI system, the eProbe 250. These activities continue to drive our R&D spending. As I said above, we also see strong interest in our DFI technology from our fabless customers. For example, a second large fabless company is placing DFI structures in its 10-nanometer product designs. Beyond the 6 DFI tape-outs that our fabless and fab customers did for their test chips in Q2, fabless companies reported to us that 6 full product tape-outs also occurred in the last quarter. These wafers have up to 8 billion on-chip instruments. These customers have DFI instrumentation worked into their standard design flow. This means many tape-outs will occur without our direct involvement. To date, this gives us a total of 46 tape-outs that we know of. In summary, we are excited about our DFI achievements and the support we are finding in the foundry and fabless customer base. We believe DFI will significantly expand our market opportunity. This initial traction supports our belief that non-contact electrical process control will become a critical solution for the increasing manufacturing challenges driven by advanced semiconductor technologies. We will continue to provide updates on our progress on our DFI solution. Now I'll turn the call over to Greg to discuss in detail our financial results for the second quarter. Greg?

Thanks, John. As a reminder, in addition to using GAAP results when evaluating PDF's business, we believe it is also useful to consider our results using other non-GAAP measures. For internal purposes, the company focuses on non-GAAP net income and EBITDAR. Non-GAAP net income excludes from nonrecurring items -- sorry, excludes nonrecurring items, stock-based compensation expenses and amortization of expenses related to acquired technology and other intangible assets, their related tax effects, as applicable. Additionally, the income tax provision has been adjusted in our non-GAAP net income to reflect cash tax expenses only. EBITDAR is equal to earnings before income tax adjusted to exclude nonrecurring items, depreciation, amortization and stock-based compensation. You can access the earnings press release that contains a reconciliation of EBITDAR and non-GAAP net income to GAAP results in the Investors section of our website located at pdf.com. Now let's turn over to the financial results. Total GAAP revenues for the quarter were $26.7 million, resulting in GAAP net income of $2.2 million and GAAP EPS of $0.07 per fully diluted share. Revenues on a non-GAAP basis also totaled $26.7 million, with non-GAAP net income of $5.3 million and $0.17 per fully diluted share. Cost of sales and operating expenses together were $22.9 million on a GAAP basis and $20.4 million on a non-GAAP basis. Moving on to the revenue details. As stated earlier, total GAAP revenues for the quarter were $26.7 million. This was higher than the previous quarter by approximately $1.6 million. Total non-GAAP revenues were also $26.7 million and were approximately $1.5 million higher than in Q1. Total non-GAAP revenues were comprised of design-to-silicon-yield solutions or solutions revenue of $20.6 million and gainshare performance incentive or gainshare revenues of $6.1 million. Our top 10 customers represented 86% of total revenues in the quarter. Two of these customers contributed revenues of 10% or greater for a total of 52% as compared to 2 customers and 64% in the prior quarter. Looking at solutions revenue in more detail. 15 project-based engagements contributed at least $100,000 of solutions revenue in the quarter, 2 more than in the previous quarter. This increase in the number of product engagements was driven by the addition of a major Taiwanese foundry and a leading Chinese fabless customer. Q2 solutions revenue at $20.6 million was $1.9 million higher than in Q1. Gainshare revenue for the quarter was $6.1 million, a decrease of approximately $400,000 from the prior quarter. The total number of node sites, which we define as an individual fab and process node combination, contributing to gainshare revenue was 13, down 2 from the previous quarter, as some older nodes continued to shrink down below our cutoff levels. On a geographic basis, North America accounted for 36% of total revenues, which is down 13% from the previous quarter. Europe accounted for 21% of total revenues, an increase of 4% from the prior quarter, and Asia accounted for the remaining 43% of total revenues, an increase of 9% over the prior quarter. As you can see from both the geographic revenue data and the top 10 data, our efforts to both expand our customer base and focus on the Asia markets are being reflected in our actual revenue results. Moving to expenses. Cost of sales for the quarter was $10.7 million on a GAAP basis, which was approximately $500,000 higher than in the previous quarter. This increase in GAAP cost of sales was primarily driven by higher salaries and benefits due to our annual worldwide merit increases, hiring in Asia and higher travel expenses during the quarter, partially being offset by lower stock compensation expense. GAAP gross margin was 60% in the quarter as compared to 59% in the previous quarter. On a non-GAAP basis, cost of sales was $9.6 million, which was approximately $600,000 higher than in the previous quarter. This increase in non-GAAP cost of sales was driven by the same factors as in GAAP cost of sales. However, there was no offset for the lower stock compensation expense. Non-GAAP gross margin was 64% in the quarter, the same as in the previous quarter. Total GAAP operating expenses at $12.3 million were approximately $700,000 higher than the last quarter and 46% of total revenues, the same as last quarter. R&D expenses totaled $7 million, approximately $700,000 higher than the prior quarter. R&D expense as a percent of revenue was 26% in the quarter, up 1% from the prior quarter. SG&A expenses totaled $5.2 million or 20% of total revenues compared to $5.2 million, the same as in last quarter, and 21% of total revenues in the prior quarter. The overall GAAP operating expense increase was primarily driven by a ramp-up in the development activity related to our DFI solution, which includes R&D hiring and increases in the use of third-party contractors. Also contributing to the increase was the previously mentioned worldwide merit increases and higher travel expense. Lower stock compensation expenses once again partially offset these increases. On a non-GAAP basis, looking at operating expenses and cost of sales together, total spending was $20.4 million, which was $1.5 million higher when compared to the prior quarter. As previously mentioned, this increase was primarily driven by the development activity related to our DFI program, higher compensation expenses related to merit increases and finally, additional hiring in Asia. The GAAP income tax provision for the quarter was $1.5 million, which represents an effective tax rate of 40% compared to 33% in the prior quarter. This rate increase is primarily due to having fewer discrete reserve reversals in Q2 as opposed to Q1. Cash tax liabilities for the quarter were approximately $1 million. This represents an effective cash tax rate for the quarter of 26% of pretax GAAP income. As we have stated before, our cash taxes are primarily comprised of foreign withholding taxes. GAAP net income of $2.2 million for the quarter resulted in GAAP EPS of $0.07 per fully diluted share compared to $2.1 million and also $0.07 per fully diluted share in the prior quarter. EBITDAR, which I defined earlier and is also defined in our press release, was $7.1 million in the quarter as compared to $6.8 million for the prior quarter. On a non-GAAP basis, net income was $5.3 million and non-GAAP EPS was $0.17 per share for the quarter compared to $5.4 million and $0.17 in the prior quarter. Total cash at the end of the quarter was $122.2 million, a decrease of $7.2 million when compared to cash on March 31. Cash used in operations during the quarter was $2.7 million. The cash impact of fixed asset purchases during the quarter was $3.3 million, which were used primarily related to our DFI program. The company also repurchased $1.8 million or 129,500 shares of stock during the quarter related to our board-approved stock repurchase program. Additionally, the company also repurchased $1.1 million worth of shares related to employee tax liabilities on RSU grants. Trade accounts receivable DSO was 66 days for the quarter compared to 72 days in the previous quarter. The trade accounts receivable balance at the end of the quarter was $19.4 million, a decrease of approximately $600,000 from the previous quarter. The unbilled accounts receivable balance, including long-term, was $23.3 million, an increase of approximately $7.1 million over the prior quarter. This increase in unbilled receivables was primarily driven by the company meeting previously established yield targets and other milestones on multiple solutions projects ahead of scheduled invoicing dates. This also resulted, by the way, in a positive impact on solutions gross margins in the quarter. Of the $42.7 million of total current and long-term receivables, approximately $2.6 million or 6% was aged greater than 30 days. Since the end of the quarter, an additional $3.4 million has been collected. Total DSO for the quarter, including unbilled receivables, was 146 days compared to 129 days in the prior quarter. Headcount at the end of the quarter was 412 compared to 395 at the end of last quarter. And as I stated earlier, it was primarily increased in Asia. Now let's discuss the remainder of 2016. As we previously stated, we've been very cautious with regards to 28-nanometer volumes for this year. Given what we saw in Q2 and the expansion of capacity at new 28-nanometer customers, we are a bit more optimistic than in that we may have seen the trough on 28-nanometer volumes in Q1. Offsetting this good news on 28-nanometer, we saw the impact of industry inventory corrections and product transition at the 14-nanometer node. Looking at the combined effect of these 2 variances, we expect total gainshare for the year to be flat to slightly down as compared to 2015. In respect to solutions revenue for 2016, due to the incremental strength in our solutions bookings as well as some upside in our DFI program, we expect solutions revenue to grow in the mid- to upper-teens range versus our prior estimates of low teens. Therefore, our overall revenue outlook is to outpace the logic semiconductor market and grow at a rate in the high single digits for the year. In regards to spending, as John stated earlier, growing demand for our eProbe 150 system and accelerating development on our eProbe 250 system are requiring increased investment levels in R&D and cost of sales from both an expense and cash standpoint. We would expect R&D expense to grow by approximately 3% per quarter for the remainder of the year. Finally, we expect our capital spending to increase by $10 million as compared to 2015, once again, driven by the DFI program. Finally, as you may or may not be aware, the SEC has recently provided updated interpretations related to disclosure of non-GAAP financial information. In order to comply with this guidance, we will be changing our methodology for calculating non-GAAP financial results. The most significant impact of this change will be that non-GAAP net income will be calculated on a full tax basis as opposed to our current methodology, in which it is calculated on a cash tax basis only. As we have stated previously, we expect our full effective tax rate for the year to be 38% to 40% of pretax income. We will fully implement this change in the reconciliation of GAAP to non-GAAP table included in our Q3 earnings release, and I will discuss the impact of this change in detail on the methodology on our next call. Now I will turn the call back over to the operator for Q&A. Operator?

[Operator Instructions] And your first question comes from Jon Tanwanteng.

You guys mentioned pulling forward the 250 series DFI product. When do you expect that to launch now? And what do you think the potential market size is or annual run rate might be for that product after launches?

Jon, this is John. So we're pulling forward some R&D. We still expect it to be available in -- scanning wafers in our facility in early 2017 and shipping in the second part of 2017, which is on our original plan. In terms of the market opportunity, we do believe, and we -- as we talk with our fabless customers, they see the opportunity for new product introduction, when you bring in a new product into a fab, to more quickly bring that product up. As well as we believe there'll be process control applications. We don't know what fraction of the process control market this represents. We believe it would be a meaningful part of it, but we don't know exactly what that is now until we understand where customers can put this in their flow. It's not really replacing -- it's a very different type of inspection, right? So it's really allowing for new information. So at this point, we don't fully know how people will use it, just like we didn't expect people to really use the eProbe 150 in the way that they're using it. They found applications for it that were in process R&D that we had not anticipated. And that's why we're meaningfully spending on the eProbe 150 this year and expect to deploy more of them than we have so far.

Okay, great. Greg, you mentioned an engagement with a "major" Taiwanese foundry when you talked about the solutions business this quarter. Is that a new customer? Or is that an existing one?

That would be a new customer.

Okay, great. And there's only a couple of major Taiwanese foundries, right?

That is correct.

Okay, great. Okay. Another question just -- when you talk about the solutions business, are you ever going to break out the DFI or software portion of that? And I guess the following question is, what is the trend in the traditional yield ramp portion of that that leads to gainshare after you pack those out?

Yes, so our current plan is to not break that out. The reason for that is in many of the engagements, these will be sold as either a direct part of an ongoing yield ramp deal or be merged with several of our other components of our business like Exensio Big Data. So it will be hard to kind of break out the elements. So you won't see that, although we will give you color on the scale of the business and the growth rates as we proceed through. And what was the second part of your question?

I was just wondering what the trend is in the traditional yield ramp portion of the business.

We actually saw quite a number of engagements get booked this quarter, particularly in both the DFM side of the traditional business, but also across all the nodes, 14, 10 and 7. As we said, we're seeing some early interest in 5-nanometer also.

And our second question comes from Tom Diffely.

So Greg, what is -- I missed your comments on the expense side, the operating expense side for this year.

Yes. So we're expecting R&D expense, driven by the DFI program, to increase from Q2 to Q3 and Q4 by 3% or so per quarter.

Okay, great. All right. So when you look at -- I think, John, you mentioned the 46 tape-outs. Is that with multiple customers? And what are they doing? How are they testing that if you only sent out one system so far?

Tom, it's a great question. So yes, it is on multiple fabless and fabs. I've lost track of the number now, but it's over 5 or so fabless companies that have a similar number of fabs at this point. And now there's 2 machines out there that I can measure. We also have capability in our own clean room here in Silicon Valley. And at this point, wafers do come back to us, and other fabs are scheduling to ship -- fabs and fabless are scheduling to ship wafers to our facility, where we can demonstrate the kinds of information that you can generate from these instruments. Remember, that's really what drove these first 2 systems. In the summertime in 2015, we had wafers come from those first 2 fabs to our facility here in California, and we showed them what kinds of information you could get. So some of those tape-outs will go through -- most of the tape-outs will go through one of those few machines, the ones that are in our facility as well as the ones that are in the existing customers. Also, for our fabless customers, once they put this into their design flow, then it becomes a standard operating procedure, and every tape-out has these instruments in there. And they recognize, when we talk with our fabless customers, and we recognize that in order to get fabs to adopt, if you tell a fab, "I have a new inspector, and I can just use regular wafers," then they can make a decision about that inspector and not worry about what the fabless do. But our approach is you put something on the wafer to make the inspection problem much easier. So when we go back to the fabs, we need to be able to point to a community of designs and wafers that have these instruments. And now what we've got with a couple of customers now doing this on all of their tape-outs built into their flow, foundries can get comfortable with the fact that there will be tape-outs that are coming into their factories that are fully instrumented. And therefore, this methodology is possible. So this is a chicken and egg problem, right? Tom, we are actively getting fabless to adopt and then turning back to fabs and showing them what benefits they could get. And the fabless are helping us with the fabs with this.

Okay. So if a fabless potential customer decides not to do this, is it because of cost or complexity or just time?

If somebody says, well, does my fab have a machine to -- does my foundry partner have a machine today? And in many cases, the answer is no, right? "So why am I going to do it?" would be a response you get from a fabless company. What we see with the leading-edge fabless -- and they're very enlightened about this. Hey, we know that it is very hard to see the problems that limit our new product introduction. We think we know what product layout patterns are tricky to build. We want to give instruments for our fabless -- our foundry companies, who can -- to make this problem easier for them. So that's the dialogue that starts. And they put those instruments down, and then literally, they go to their foundry partners and say, "We've done this work. We'd like you to work with PDF to inspect these." And we've had that happen with numerous fabless now across multiple foundries. And the first 2 foundries are a result of those fabless companies bringing this up to their foundry partners.

And I assume your hit rate's pretty good for, say, 14-nanometers and below.

Yes. Because we run electrical test vehicles on every flavor of 10 and 14 that's out there, we have a lot of insight about what instruments you would like to put on a wafer to make the process control problem easier. And so we're -- and we also -- our layout software runs through most, for our fabless customers, runs through most of the major tape-outs on application processors, graphics chips, et cetera. So we have a good handle on what patterns are tricky in their layouts. So when you marry those 2 datasets together, you can build an on-chip instrument that has good sensitivity to customer-specific problems.

Yes, that sounds very compelling. So you've also talked a lot about how China is just a really fast-growing market for you. When you do your engagements there, is most of that at the 28 and above level? Or is it really moving quickly to 14 and below?

Yes. What we've seen on the fabless side, and we talked about one of the engagements signed in the last quarter, was 10-nanometer for a Chinese fabless, DFM for a Chinese fabless. So on the system and the design company side, we see a very accelerated use of advanced nodes. Now, of course, when they're designing a 10-nanometer or 7, they're not using foundries inside China that are Chinese-owned foundries in China. They are working primarily with our customers in the U.S. and the rest of Asia. And again, part of the value that PDF can bring them is we run vehicles in those facilities all the time, so we have a lot of expertise about helping them characterize their overseas suppliers. Now what's happened in the last couple of quarters is those fabless companies are also trying to use that 28-nanometer and, soon, I think, on future nodes, internal capacity inside China. And again, because PDF has business with many of those companies, we're able to provide them an infrastructure to make that a more smooth on-boarding process as well.

Okay, great. And then finally, when you think about gainshare kind of recovery, I guess, from trough levels right now, is the bigger driver the 28 -- the new 28-nanometer customers? Or is it a ramp of a second 14-nanometer customer?

We expect the ramp of 14-nanometer in '17 -- in '16 and early '17 to be the big driver, and as you get out to '17, the second wave 28-nanometer customers also become an important factor.

And your next question comes from Brian Freckmann.

I think my questions have mostly been asked. I just wanted to follow up with what Tom was saying on what was going to drive to the back half gainshare improvement. So you guys have sort of touched on that. Is there anything more you'd like to discuss?

No. I think we do believe -- and we've seen in our foundry partners an uptick in volumes, and that would drive the third and fourth quarter gainshare.

Okay. And then forgive me for not having my EE from Carnegie Mellon and PhD as well. It was my impression in the past that sort of the e-250 was much more sort of the fab level and the 150 was kind of the lab to make it simple. Is that, in fact, correct? And so is this -- are you guys thinking that some of your customers are going to bring the product down to the fab floor versus just keeping it in the lab?

Yes. So generally, your description is right, Brian. I don't think you need an EE degree. You're doing just fine. But what happened was our original thought was the 150, we were going to put a couple of them in the world for collecting data to verify the on-chip instruments worked and to give -- because Tom brought up a very good question for fabless, right? People will ask, "Hey, are there machines out there that can measure these things?" We knew it would take a while to develop the 250 platform, and so we though while we're doing that, we get 1 or 2 of these out there to create some data sets for our fab and fabless customers to gain confidence in the instruments and learn how to use the data. And that's why we said, if you remember, on our previous calls, we think there'll be a small amount of revenue at the end of 2016. And what's happened is, as the data sets tended to be more valuable for our foundry partners than we thought, fabless customers are also putting more of them on their test vehicles than we initially anticipated. And they're seeing good application for the data. And it's still R&D, so you're moving a little bit from the lab to kind of the R&D process inside the fabs. And we are starting to find some applications that could leverage the 150 even for production, although it's still very early in proving those out. And we started to realize that the 150 has a place separate than the 250 for a different set of applications. And it's a very -- we were able to strike relationships with customers that were very compelling for them. And for us, a profitable business that's something that we could afford to bring out. And now we see other customers wanting that same capability, so now we're making plans. And this is part of the reason why the expenses have gone up, to continue to develop applications and capacity for the 150s while we continue to develop and accelerate the development on the 250. And we think they're going to complement each other in the field rather than our original intention, which was to replace the 150s with 250s.

I guess there's indications for orders on the 250s. Are those from customers that have the 150s? Or is that a new , different set of customers?

Yes. So we've not gone out and asked for orders and solicited orders on the 250s, but we have gotten folks to express their interests and if it could do what we said it would do. And primarily, there are different applications. As I kind of alluded to in my prepared remarks, for our fabless customers that expressed interest in the 250, they are quite interested in the new product ramp-up because being able to scan billions of on-chip instruments in an hour is kind of an out-of-the-imagination capability, at least 100x faster than what you could do alternatively. And they think that, that would greatly shorten their learning curve. And so that is a little bit different than the 150 application, which really is getting around process window verification and more IP pattern development as opposed to full product bring up.

And your next question comes from Christian Schwab.

So after this quarter, can you guys give us a quick update on what you think the pipeline is for additional DFI orders? Do you have any insight there you can share with us?

Yes. Christian, the pipeline is really those tape-outs that are out in, I think, at this point, 5 different foundries. We anticipate generating -- getting wafers in our facility in Q3 and early Q4 from other foundries as well as the existing 2 kind of digesting what the initial capability can be. We would -- we are not looking to push it so hard that we trip over ourselves. So we would be excited if we had an additional system in the second half of 2016, with kind of an on-ramp or an off-ramp in early 2017. But we don't need to get to a big number this year. And also remember that the model that we use for this is a ratable revenue model; it's almost like a time-based license on software. So the machines that we ship are still generating revenue. When you're in a capital equipment model, once you ship and recognize the revenue, you need to -- you start again from 0 on your revenue for machines for the next quarter. Whereas for us, we build off the base of what's already been established. So we don't need to ramp them up very quickly in order to sustain ourselves, and we will ramp them up at a rate that we feel comfortable we can meet customers' expectations and exceed them. So if we could get one more in this year, that would be super, and then a ramp into early '17 would be great. And that would meaningfully impact our revenue.

Yes. Are you still thinking about incremental revenue in DFI year-over-year in the $5 million to $10 million range in '17 versus '16? Is that still fair?

That's reasonable.

Okay. And given a lot of excitement that's going on for you guys in bookings, in particular, on the Exensio side, as well as a lot of activity in China, is it fair to assume that we could see a similar type of improvement from those 2 areas in '17 versus '16 as well?

Yes.

And your next question comes from Gary Schnierow.

I wanted to follow up on your CapEx. I think you said it was $10 million incremental to last year. Can you give a little color on that and maybe what that implies for 2017? I mean, I know you said it was obviously good news that you're spending money. But what does that look like for next year?

Yes. So if we think about '16, maybe being in the mid- to high-teens level is probably rational. When you look at '17, it's hard to call because, as John pointed out, we may be expanding and ramping additional machines on the 150 series overlapping with going to late development, early production on the 2 series, which is a significantly more expensive capability. So we've got a variety of potential projections, depending on timing of those things, but I would expect that a similar to slightly more spending in '17 than in '16, at least.

Okay. So simply, if business is good, CapEx is going to stay high in 2017?

Correct. Yes, and that's driven by the fact that our plan is to not actually sell any of the machines associated with this business. They'll be treated as capital goods and then depreciated over probably 5 years.

Got it. And I just want to make sure I heard it correctly. At the beginning of the call, I think you said you either added or extended contracts on like 20-something customers, which sounds like an awfully big number. What's the number? And what would it have been, say, last quarter to put it in context?

Yes. I believe the number was 23, although we're going backwards and looking. I'm sorry, it was 27 that we booked new or extended deals in the quarter, 27 customers. That number can vary quite a bit quarter to quarter because of the software business, but it has been growing is a good way to think about it. I can go offline and get you last quarter's number.

Year-over-year, the number's up quite significantly.

Quite significantly, yes.

And your final question comes from Jon Tanwanteng.

Just 2 quick follow-ups. What do you expect the margin trend in the solutions business to be through the second half?

Yes. I think Jon, you and I have talked about this before. I think you would continue to see the solutions margin improve slightly quarter to quarter. Now there's some timing that will fall in there, but on a half year basis, if you're thinking up 1 point or 2, driven -- on the good news side, being driven up by increasing mix of the software business and now even some of the 150 series business. Offsetting that will be -- as we ramp up business in China, it takes a while for those margins to get back to full levels.

Okay. But in general, you expect the margins to go up?

Overall, if you think -- yes, margins should go up but only by 1 point or 2, I think.

Okay, got it. And the second question, just on the DFI side. Do these fabless customers who put the instruments on their designs -- how do they pay you for that? Is it part of the software package that you offer as an additional module? Just wondering what the model is there.

Yes, it's a great question, Jon. We, right now, charge nothing to fabless who want to put instruments on their wafers. They then generate a key, in effect, which is where the instruments are located within the design, and PDF has the right to distribute the key to the foundries. So really, the strategy here is to -- the fabless put these on as a way of making their designs easier to bring up and control, and we, right now, monetize the foundries. And that will be the model for the near term.

And we do have one more question from Eugene Fox.

John, I apologize. I got on the call very late. Did you -- I know you commented on the gainshare for the second half. Any thoughts on the global foundries agreement with AMD and sort of where they stood as of the end of Q2?

Thank you so much for the question. Unfortunately, we really can't comment on our customers' contracts with their customers and tend to not have much visibility into that. We'll probably see those things when the customers are required to file their SEC documents. It would be about the soonest we'd see it other than hearsay.

At this time, there are no more questions.

Thank you, everyone. We look forward to talking with you at the end of Q3. Have a good day.

Ladies and gentlemen, this concludes the program. Thank you for joining us today.