PDF Solutions Inc (PDFS) 2012 Q1 法說會逐字稿

Good day, ladies and gentlemen, and welcome to the PDF Solutions, Incorporated, conference call to discuss its financial results for the first fiscal quarter ended Thursday, March 31, 2012. (Operator instructions) As a reminder, this conference is being recorded. If you have not received a copy of the corresponding press releases, they have 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 demand for its solutions. PDF's actual results could differ materially. You should refer to the section entitled ``risk factors,'' on pages 10 through 16, of PDF's annual report on form 10K for the fiscal year ended December 31st, 2011, and similar disclosures in subsequent SEC filings.

The forward-looking statements and risks stated in this conference call are based on information available to PDF today. PDF assumes no obligation to update them.

Now I'd like to introduce John Kibarian, PDF's president and chief executive officer, and Michael Shahbazian, PDF's vice president and principal financial officer. Mr. Kibarian, please go ahead.

Thank you and welcome, everyone. PDF Solutions started off 2012 making strong progress toward our goal of being pervasively applied to leading-edge logic manufacturing and generating gain share from our clients' successful yields.

In the first quarter of 2012, we achieved revenue of $20.6 million, non-GAAP profit of $0.16 per share, and GAAP profit of $0.12 per share.

On the new business side, we successfully closed contracts for the following, all with existing clients. A 40-nanometer yield ramp with a foundry client, an extension to a 28-nanometer DFM engagement with a fabless client, and a 28-nanometer DFM engagement for a Japanese integrated device manufacturer, which is starting to use foundries for production.

These contracts continue the trend of fabless and fabs, using PDF Solution's Characterization Vehicle technology across the IC process lifecycle to improve the yields and manufacturability of their designs and processes.

Considering the recent activity in the chip industry, [one-sies] met challenges in achieving strong yields and production volume, meaning that logic fabs need to invest more in characterizing their processes. While inspection provides a visual of the defect, circuits work on electrical signals. Our Characterization Vehicle infrastructure provides a comprehensive electrical characterization of the defects, both the hard defects, which are often seen at inspection, as well as the parametric defects, which are undetected by visual inspection. Ultimately, what works or fails electrically is what matters.

This explains why both fabs and designers continue to expand the use of our Characterization Vehicle technology across the lifecycle of their processes. PDF Solutions CV infrastructure is now applied to yield ramps at a larger number of processes than ever, ranging from 65 nanometer to 14 nanometer.

As the industry moves to thin sets and other complex structures to improve their parametric performance, the logical characterization will become essential to achieving competitive results.

In the fourth quarter of 2011, we signed a commericial license at a beta customer for our next generation YieldAware process control software system, which we call [Ascensia]. It's a big data system that leverages the latest in software architectures, making it scalable on the [clock]. This quarter we successfully demonstrated the desired performance and receive customer acceptance of the system. This contract is for a manufacturer's facility for image senses, so beyond proving out our new YieldAware process control solution, it also validates the applicability of our technology to an adjacent market.

Gain share results in Q1 are improved, as we expected and stated in our call about the fourth quarter, 2011, results. The primary driver for this improvement was increased 45 and 32-nanometer volumes and yields.

Although we continue to expect gain share to grow year-over-year as leading-edge volumes increase, we also expect volatility quarter over quarter. As characterization- as logical characterization becomes essential to ramping new products and processes, we continue to see strong interest from the fabs and logic designers for our Characterization Vehicle technology.

However, we are mindful of the cyclical nature of the chip industry. So, while we see opportunities to continue to grow both our solutions and gain share revenue, we will be cautious, as we make investments and disproportionately grow earnings with increased revenue.

We will continue to refrain from providing specific quarterly guidance and focus, as we have, on long-term profitable growth.

Thank you for your time and attention. Now I'll turn the call over to Mike, who is filling in for Greg Walker while he's on leave. Mike will discuss in detail our financial results for the first quarter. Mike?

Thank, John. As a reminder, in addition using GAAP results when evaluating PDF's business, we believe it is also useful to consider our results using non-GAAP measures. In this case, non-GAAP measures exclude stock-based compensation expenses, amortization of expenses related to acquired technology, and other intangible assets, restructuring charges, and the tax-related effects, as applicable.

You can access the earnings press release that contains a reconciliation of non-GAAP to GAAP results in the Investors section of our website, located at PDF.com.

Now let's turn to a review of the financial results. Let me cover some of the highlights of the quarter. Total revenues were $20.6 million, with a GAAP net income of $3.5 million. This resulted in an EPS of $0.12 for both basic and diluted shares. Net income on a non-GAAP basis totaled $4.6 million, or $0.16 per diluted share. Cash decreased by approximately $500,000 during the quarter. Cost of sales and operating expenses were $16.8 million, an increase of $1.4 million from last quarter. We are pleased with the strong improvement in revenues and earnings for the quarter.

Now moving on to revenues - total revenues of $20.6 million for the first quarter were up 17% as compared to $17.6 million in the prior quarter and up 37% compared to $15 million for the same period in the prior year. Total revenues were comprised of design to silicon yield solutions of $13.4 million and gain share revenues of $7.2 million. Gain share revenues for the quarter increased by $3.8 million, or 111% sequentially, and yield solutions revenue for the quarter decreased by approximately $770,000, or 5% sequentially. As John has indicated, the significant improvement in our gain share revenues was due to increases in 45 and 32-nanometer volumes and yields. Our gain share revenues for the quarter as a percentage of total revenues increased to 35% from 20% in the prior quarter. The decrease in yield solutions revenue is due to the completion of engagements at the end of the fourth quarter, while newer engagements begin to ramp in the first quarter and was also the result of lower software revenues.

The total number of customers contributing to gain share revenues in the first quarter remained at five, same as it was in the prior quarter, which is consistent with the increasing trend of the concentration of the leading foundry manufacturers.

As John mentioned, we added two new contracts during the quarter, along with one extension to an existing contract. In the quarter, 13 engagements with a total of six different customers each contributed at least $150,000 of revenue. Our top ten customers represented 92% of total revenues in the current quarter. There of these customers contributed revenues greater than 10% each, the same number as last quarter and one less than the first quarter last year.

On a geographic basis, Europe accounted for 41% of total revenues, North America represented 34%, and Asia accounted for the remaining 25% of total revenues.

Moving on to cost of sales, the total cost our yield solutions was $8.7 million, an increase of approximately $500,000 as compared to the prior quarter. Due to the higher gain share contribution, first quarter gross margin increased to approximately 58%, compared to 53% last quarter. Non-GAAP gross margin was 60% for the quarter, compared to 57% last quarter.

Our total GAAP operating expenses were $8.1 million, or approximately 39% of total revenues, compared to $7.2 million, or 41% of total revenues in the prior quarter. R&D expenses totaled $3.2 million, or 15% of total revenues for the quarter, compared to $3.1 million in the prior quarter. SG&A expenses totaled $4.9 million, or 24% of total revenues, compared to $4.1 million in the prior quarter.

Due to the effect of foreign currency fluctuations, other income and expenses's $142,000 expense compared to $282,000 income in the previous quarter.

The income tax provision decreased quarter to quarter by approximately $100,000 and slightly increased as compared to the same quarter last year. In general, our tax provision is primarily comprised of form withholding tax on sales.

Cash used in operations was approximately $1.4 million. This was primarily the result of a delay in payments from a single customer. The majority of this balance was collected after the end of the quarter. Employee stock purchases and option exercises contributed $1 million in the quarter, to partially offset this delayed payment. Accounts receivable increased by approximately $7 million, primarily due to an increase in billings during the first quarter, further strengthening the balance sheet.

Headcount at the end of the first quarter was 323 worldwide, compared to 319 at the end of the fourth quarter. That concludes the review of the financials for the quarter. Now I'll turn the call over to the operator for questions.

(Operator instructions) Tom Diffely.

First, a couple of questions on just the model itself. You know, you talked about some of the comps going up sequentially. I'm just kind of curious with, you know, with the design-to-silicon yield revenue going down, what drove the increase in the direct costs in that segment?

Well, overall, we have continued to add headcount in our cost-of-sales organization, so if you look at period over period, so that- certainly we did see some increase, primarily due to increases in terms of our headcount increases. Also, we did increase our variable compensation because of the higher profitability. We did make a larger accrual for our bonus accrual for our employee bonus programs. So, I think those were two of the most significant factors regarding our cost of sales.

And then, also when we're talking about SG&A, we did end up with some favorable impacts in the fourth quarter regarding lower legal expenses that were not repeated in the first quarter, and so the reversal of that, or the non-recognition of that, in the first quarter, plus we had higher year-end audit fees, caused our expenses to go up in that category.

OK. So kind of on a go-forward basis, would you expect this to be a good starting point, and then, you know, just some incremental adds for headcount going forward?

Well, we hesitate in giving any guidance, but I think the expenses for the quarter do reflect a pretty good reflection of our current expense trends, and they do not reflect any offsets or one-time impacts, like we had in the fourth quarter, so I think the first quarter is a very clear reflection of our current expense base, related to our employed population and the level of activity the company is experiencing today.

And then John, when you look at the gain share during the quarter, does that represent just kind of ongoing business, or is there any one-timers that could possibly be in that?

No, it's pretty much ongoing business at this point.

OK. And you mentioned obviously it could be volatile, on a quarterly basis, but if you assume that just the demand and the growth of these advanced nodes continues to be strong, then you would expect this number to be strong and growing as well?

Yeah, you always, on a- we always like looking on it on kind of a four-quarter rolling average, because there's always quarter over quarter stuff, you know, factory and [convert] stuff, and products go in and out. So there's always- we're [inaudible] but generally speaking, yes.

And then I guess question two, you mentioned that the customer accepted your beta software. Does that mean that it was revenue during the quarter, or is it revenue later?

There was some revenue in the quarter. It's mostly ratable, so I think there was a modest revenue this quarter. I think there were also some expenses that might have driven some of the cost of goods sold expenses on hardware that we shipped with that beta. That was also recognized in the quarter, which probably drove up the cost on the COGS in Q1 a little bit, too.

And then finally, when you look at just the market, or the industry right now, and the 20 nanometer ramp, does it look to you as if it's just progressing as the normal ramp is, or is there something about 20 nanometers that makes it perhaps a lot more difficult than the ramps you've seen in the past, say, 32 and 40?

Yeah, we looked at that pretty closely, both at our customers as well as- we have a lot of [inaudible] customers, so by and large, we're running vehicles at all the logic facilities, all the logic foundries in the world, either directly or indirectly.

You know, what we hear are few things. We see, you know, the start of the 40 ramp - if you go back and look at that, everyone says the 40 ramp was bad. In the first one or two quarters, the leaders basically said, ``Yeah, it's bad, but it's behind us and it's going to be fixed in the next quarter.'' They did that for two or three quarters, actually, almost four quarters, if you really listen to their conference call 'scripts. And we see similar things in the 28 node, where the first quarter [bring up] and the crossover when, in the foundry business, 28 nanometer represents 1% or 2% of foundry revenue, happened in Q4. Q1 was a, you know, improving from a revenue standpoint, but if you look at the amount of capital that's gone into the node versus the- and what that should represent in terms of volume output versus what volume actually came out, there's still an ineffectiveness of that capacity, due to the difficulties in controlling the process. And I think a reflection of that would be the increase you're seeing on the metrology and inspections, as people try to control these technologies. And I think you'll see that over the next few quarters, as folks try to control the 28 node, so the capital intensity will be relatively high and relatively inefficient until the control improves. And that's, of course, good for us, too, because that means characterization, not just with inspection metrology but also electrically, will be more important.

OK, so you expect to see top line growth continuing, just because of a combination of both the software plus this increased intensity of, you know, yield issues, on a near-term basis.

[Um-hmm.]

OK. And then finally, when you look at the potential to go into the IDMs that are outsourcing the foundries, do you see that as a growing part of your business, where even though a bulk of what you earn ultimately is going to be gain share from the foundries, just this upfront business with the IDMs being more and more important.

Yeah, it's a good question, Tom. So you know, if you go back to 2007, or 2008, the Japanese IDMs were a very significant part of our business, and as, you know, the recession started in the middle and end of 2008, we recognized that the Japanese as integrated device manufacturers on the leading edge were really going to move away from that model.

I think early on, when we met with them in 2009, 2010 timeframe and said, ``Hey, look, our leading fabless customers all see the importance of running our vehicles, and we can be very- our systems can be very instrumental to you as you try to bring up your products in the foundries, I think they had very idealistic, almost maybe immature expectations of what it would take to actually bring up complex SOCs at the foundries. In this past couple of quarters, we started talking with them again, and we do see a lot of activity in Japan, this last quarter being the first- why I think the most vanguard of the Japanese IDMs. But we expect more of that, and they've all reflected upon this, no matter who they selected as their foundry supplier, if they don't do some come characterization of their IP end designs ahead of production, achieving the performance and yield targets they expect is quite difficult, no matter who they select.

Right, OK. And I guess finally here, anything new on the adjacent market front?

Well, we did mention that the [sign off] on the [image sensor], which we were quite pleased with, and we continue to kind of make our forays- if you look at the broader inspection and use KLA as a good proxy for the investment in metrology and inspection, the bulk of their business was logic last quarter, and I think they've projected that to be the case for a good part of year, as I listened to their call. And I think that's kind of is a good proxy for the overall interest level in process control and inspection and metrology and our electrical characterization. So, we, in general, expect logic to be still be the driver of the business and we're still dabbling in trying to understand the adjacent markets, to make sure we make good, profitable investments in them.

[Adam Fisher], [Sanchez].

There's been a lot of industry discussion this quarter about kind of TSMC's capacity shortages, you know, some of their major clients having to move away, you know, to other fabs. Can you talk about the implications of that on our business as well as TSMC's decision to accelerate 20 nanometer spending, how that might affect not only our relationship with TSM, but investments that some of your other partners might be making?

Sure. So let's see if I can answer that question. So the first part is, of course, we work with most heavily with the [ISDA] partners, and they're, by far, the largest alternative to TSM in terms of leading-edge logic foundry manufacturing. So another large, fabless company expresses desire to move their products to- apply to TSM, then of course, the partners are probably the net beneficiary [of that] should they execute. And of course, it's our role to help them execute. So that's a great thing for us, is another driver of potential incremental volume.

In terms of the overall 28 nanometer challenge, I kind of answered a little bit on the question with Tom, but the efficiency, the capital efficiency, or in other words, the amount of output wafers for the amount going in is quite difficult in the 28 node. I think that's why the ramp-up is challenging for everybody, even the most competent foundries. And the process control or process windows are quite difficult. That's why we've been putting so much investment in our YieldAware solution we call Ascensia. We believe that's very critical to future control.

I'm sorry, I thought there had also been some discussion by TSMC on increasing their 20 nanometer.

Yeah, I was going to come to that. The third part of your question is, of course, the 20 nanometer. That's a very interesting one, right? Because the industry overall has been looking at the effectiveness of the 20 nanometer solution in terms of staying on Moore's Law, with respect to, you know- what the industry always expects is, I can't get that same [gate] for roughly, you know, almost half the cost of the previous node, and incrementally, about 25% of the performance. Due to the incremental cost and multi-patterning that is required at 20 nanometer, the less-than-usual amount of increased performance versus power, in terms of the transistor on 20 nanometer, a lot of the fabless companies have been looking at the 20 nanometer [planar] solution and saying, ``OK, is that going to be incrementally beneficial enough over 28?'' It's clearly beneficial. The question is, how much beneficial it is relative to the previous node. Or should I wait for some kind of thin technology?

TSMC's announcement was interesting because what that basically says is, you know, much like prisoner's dilemma, none of the fabless guys want to be left behind at 28 if their competitors are moving on to 20. So while they may all kind of wring their hands and say, ``Gee, the 20 node doesn't look that effective, maybe I should wait,'' I think TSMC is at least communicating with this, that maybe enough players are moving on to 20 that probably that's going to drive a lot of the industry to move on 20 anyway, because you don't want to be left behind.

So, I suspect that's going to increase the number of incremental nodes. TSMC also said they're going to decrease the number of versions of nodes available at 20 nanometer. Frankly, I've seen every supplier that we've ever worked with always say that at the beginning of a node, and by the end of the node, we're providing vehicles for more [derivatives] than we did the previous node. That clearly happened at 28 nanometer. They cancelled the 32-nanometer node to decrease the number of versions, but in the end, there ended up being more. I suspect the same will happen on 20. There will be more flavors than the suppliers will want to admit.

OK. And kind of the corollary to that, how are other fabless customers reacting to the TSMC shortages, and is that creating some opportunity for PDF to work with the fabless- you know, to work with some new fabless customers who may not have been available to us prior-

Yeah, so last quarter, the one kind of newly minted fabless, as we referred to them, you know, a Japanese IDM, I think they're starting to use our technology for their 28-nanometer products, in part, it's because they recognize they need to be agile, the availability of capacity is a question. The challenges of just making their products work with whoever they select is difficult, so I think that drives us. They need to be as efficient and effective as possible with the silicon they are going to get.

On the new YieldAware product, I guess Ascencio, now that it's out of beta, can we- is your attention to kind of start selling it more broadly, within the next couple of quarters?

You know, these are not fast things, so in other words, the sales cycle on this is not quick. We do- we are much more aggressive in terms of bringing it out. We really wanted to do a bang-up job on the first installation. I mean, obviously, people are trusting a very expensive fab to this system, so you really want- you want to make sure the results are quite stable and the performance is quite strong, and you know, we really did fly through the success criteria, well ahead of our original plan. So, we feel quite good about the performance of the system and that will probably make us more bold, as we go out and bring it to other potential customers. But you know, like going to the dance, you may want to dance and she may not want to, so we have to work through the challenges of selling a very complex system to a very sophisticated customer. That will take time.

And on the Japanese customers, how much additional capacity would potentially- would they add to kind of the foundries, if the majority of those kind of move from- to a fabless model?

Oh, I think that's a tough one for me to handicap. You know, what we had been seeing, you know, we've done a lot of work with them over the years, is as they went through the shrinks, their capacity was going down. In other words, they couldn't find good uses for the incremental transistors that Moore's Law would provide them, and so dye sizes were coming down in Japan, from about the 90 nanometer node on.

So, I don't know, in general, how much volume they represent for the foundries. It is probably the last bastion of integrated device manufacturers that are going to the foundries, and they will probably go there over a long period of time, so we're not forecasting this is going to be a super-huge opportunity in terms of incremental wafers, but it's not insignificant, either. Maybe it's, you know, collectively, less than a single factory.

OK, I think that covers it. Thank you. Thank you very much.

[Gary Genaro], RiverPark Funds.

When you mention that you expect volatility in your gain share, is that a market comment or a customer-specific comment?

Well, you know, we have a small number of customers, so the market and customer are somewhat synonymous, but in general, we don't, you know, quarter over quarter, volumes can go up and down - some yield sensitivities, as we get towards the end of, you know, if we're still in the deployment phase and we're collecting gain share. So, you know, it's specific to our customers, but our customers kind of basically go up and down in terms of the overall volumes in the industry. It's not that we have any specific knowledge that we think the gain share is going to up or down from here, it's just, you know, we want to caution folks from taking Q4 and Q1 and putting a line through it and extrapolating out, you know?

Right, right. So obviously Q1 was a great gain share number. About when did you know how good that would be? You know, what kind of visibility do you have on that, timing-wise?

Well, you know, it was- as we went into Q1, we had a pretty good idea that it was going to be good, and we had some idea about a range of where it was going to be. As we got towards the middle of the quarter, we saw that it was probably going to be above the higher end of the range that we expected, and it turned out to be around, by, you know, March, where- by maybe March 1, what we expected, March 5 or something like that, around there. It ended up being what we expected it to be.

Got it. OK. And when you look at the gain share for the rest of the year, is that simply volume from existing contracts, or would you expect another contract or two to come into gain share?

Well, the 32/28 contracts have been fine and have been executed for a while, but those are all starting to hit their measurement points and starting to contribute gain share, so throughout the remainder of 2012, and into 2013, we expect incremental nodes and facilities to come online on 28 and 32. The timing of when they come on and you know, and how significant their volumes are, that's always very difficult for us to predict, so you know, we're- you know, that's why we're trying to caution everyone. We're not out here, trying to predict timing, but there's more facilities, you know, in this generation that are not contributing to gain share now that we anticipate to contribute to gain share in the future, over the next year or two, and whenever the customers basically complete their ramps and go into volume.

Right, got it. And the last thing is, at the beginning of the call, you mentioned two new contracts and the one extension. Can you just- I missed that. Can you just review that quickly, with any color you had?

Yeah, we just said one was a 40 nanometer yield ramp up of a foundry, one was an extension to a 28-nanometer fabless company's DFM engagement, and one was a DFM engagement for a Japanese, what we refer to as ``newly minted fabless company.'' It's really an IDM.

OK, great. So the foundry contract, that- that's a new foundry?

That's an existing client.

Oh, existing client.

A new facility, or actually, a new node.

Got it.

You have a follow-up question from Tom Diffely of D.A. Davidson.

John, just a quick follow-up, then, on the nodes themselves - so how does the number of flavors per node and perhaps number of intermediate nodes impact your view of the length of the gain share tail? I mean, in the past, we've talked about kind of a four to six-year tail for some of these foundries. Is that impacted by these different flavors?

Yeah, to some extent. You know, I think first of all, the 32/28 node, there's two real nodes there, or there's two half-nodes, depending on how you look at it. So, you know, obviously, in our commentary, we said 45 and 32 are contributing to gain share now. We didn't make mention of 28. That wasn't an omission by mistake; that was an omission because 28 is not contributing. 28 will then start up. So, that collective 28/32 node is obviously going to, you know, in total, be longer than what the 32 period would be, because you're going to get the 32 period plus the 28 period that comes on. Right? So overall, yeah, but if you looked at kind of that node, that's kind of a major node, plus its half-sister. The total length of number of years that will contribute gain share to PDF would probably be longer than you would expect, because it's kind of two half-nodes, and there's a lot of flavors in there that will start off with different periods.

OK, so kind of the way to look at it is you get more of the, I guess, the solutions work along the way if there's more flavors, and, you know, at least as much on the gain share side?

We would hope to see that they represent gain share tails, of course, but we don't control the market.

Yeah. OK, thank you.

At this time, there are no more questions. Ladies and gentlemen, this concludes the program. Thank you.