Welcome to the 26th Annual Needham Growth Conference in-person day, and glad to see everybody here. My name is Charles Shi. I'm the semi-cap analyst here at Needham. I cover FormFactor, and on the stage here, we have CEO Mike Slessor, CFO Shai Shahar, and the management team gonna present a few slides, and then we're going to fireside chat, and then we open up for Q and A. Let's get started.
All right.
Hey Mike.
Thanks, Charles, and thanks to Needham for hosting us again this year. We'll run through a few slides just to orient everyone on FormFactor, and then, as Charles said, get into a Q&A session. Key facts about FormFactor, revenue over the last 12 months, between $600 million and $700 million, down a little less than 20% from the peak before the industry entered the downturn. But I think as you look at our results through 2023, we've weathered the downturn pretty well. We did a significant adjustment in capacity and downsizing and cost reduction when we saw the downturn coming in the fourth quarter of 2022, earlier than most people. That's a result of some of the characteristics of our products, very short lead times.
But got our cost structure in shape and have really been operating in a fairly stable mode from throughout 2023 and now into the first part of 2024. We have two major products that we operate in two separate segments and report for you in all of our SEC filings in two separate segments. The first, and most people think of FormFactor; when they think of FormFactor, they think of probe cards. These are device-specific consumables that our customers use to connect their chips, usually at the wafer level, at the end of the production process, to automated test equipment built by people like Teradyne and Advantest. There's some really unique characteristics about this product. As I said, it's consumable. It's unique to each customer chip design.
So when a customer, like an Intel, like an Apple, like an NVIDIA, changes their mask set and moves to a new product, that requires all new probe cards, and you can think of us as being this interface layer that then allows the test equipment to be utilized for each of these new chip designs. So that drives a much higher frequency of replacement than most capital equipment and drives some different demand cycles. You can see our demand cycles decouple from capital equipment in some pretty significant ways over the recent history. The other segment, about 25% of our business, but has been growing pretty significantly, is engineering systems. These are sets of tools, services, instruments used by our customers, also for electrical test and measurement, but in their early development, in their labs as they're developing their next-generation devices.
We've seen some nice examples recently of growth in this business from applications like silicon photonics for co-packaged optics, a really interesting application that's set to be adopted in data centers around the world as we move through 2024 into 2025. Another area where we've invested in the systems segment is in cryogenic test and measurement for quantum computing. This is also a very early stage in the development of an industry of applications, but you can think of our systems segment as really being engaged in these next-generation developments that are gonna reach volume production a year, two years, maybe even three years from now. Whereas the probe card business is very much leveraged to here and now semiconductor production and design releases.
As most major suppliers in the industry, that serve the semiconductor industry, we are global, with a footprint all around the world. More than 2,000 people strong around the world, and you can see that, from our recent 10% customers, we do have strong relationships with all of the key driver customers in the industry, and that's something that's fundamental to our strategy. We believe that, making sure our engineering teams and our customers' technology teams are linked up very closely, so we're driving our roadmap against solving their next-generation requirements and taking the 15% of revenue we invest in R&D, driving it to solve the right problem, so that we're driving, innovation, market share, and gross margin enhancement with these leading customers.
Overall, the company, we talked about, you know, a picture of a probe card on the left here, but we are a market leader, and that's one of the central tenets of our strategy, is to lead the market in all of our served segments. We're there in some of our served segments. We have initiatives to either gain leadership or regain leadership in other parts of the segment. But we believe that being the market leader drives a virtuous cycle of, gross margin enhancement, being effective with your R&D, again, making sure that you're really partnered with these key customers in capturing, in capturing key share there. Trends that drive our business, very similar to what drives the overall industry, a variety of different pieces, but one of them I wanna highlight today is advanced packaging. So-...
When the industry really is adopting these techniques of taking what used to be a single chip and breaking them into chiplets, and then reassembling them in the back end after the wafer fab, this has a profound impact on test, and I'll detail that in a minute. But you're seeing FormFactor's recent results really be driven by a couple of interesting, emerging advanced packaging applications. One, high bandwidth memory in our DRAM probe card business. The other being the adoption of advanced packaging for the first major client microprocessor in industry history. And those are two key themes that have driven pretty significant growth for us in the second part of 2023. Overall, a focus on profitability, as many companies in this space do.
We wanna make sure that we're being good stewards of shareholder capital, and making sure that even through a cyclical downturn, we remain profitable, and we're proud to have been non-GAAP profitable in every quarter in this downturn. Finally, you know, mergers and acquisitions are a big part of the company's history and will continue to be a big part of the company's future. We wanna lead in our served segments, but we wanna take the cash we generate from that leadership position and expand our served markets, continuing to diversify and grow the overall business. I talked about... Can we get the next slide, please? All right. Maybe we should have tested the chips in this TV a little more exhaustively.
Let me talk about it conceptually, 'cause I do wanna make the point about advanced packaging and breaking die apart, before reassembling them using these advanced packaging techniques. You can imagine if you're gonna, stack a set of die or chiplets or tiles together, if you're a customer, you wanna make sure that each of those individual tiles or chiplets is good before they go into that stack. It turns out there's relatively limited repair or redundancy in these stacks of chiplets, and so what that does is it drives up both test intensity, you can imagine customers are testing each of these chiplets more comprehensively than they otherwise would, but it also raises the requirements or the complexity of test. Customers are testing these chips at higher speeds, at stressing them with higher power, at higher temperatures.
And so as a probe card supplier, as a supplier of these interfaces, that raises the technical requirements for us, and when we're able to do that properly, creates competitive advantage. These are hard technical problems to solve, and when we solve them, we create significant value for our customers and significant competitive differentiation for ourselves. With that, I think we'll open it up to Charles' questions, if we're still stuck. Yeah, there's another slide in here on recent results and that sort of thing, but we don't need to go through that. Let's get to questions.
All right, then. Hello?
Do I still have the mic? Okay, good. All right, thanks, Mike. Yeah, and Mike, Shai, let's get this started with our fireside. Look forward to the next few minutes to discuss a little bit more color about what's going on about the industry, about the probe card business, and FormFactor specifically. I wanna start with the DRAM. So one thing that you have, but your closest peer, let's say Technoprobe, doesn't have in your product portfolio, is the DRAM probe card business.
So last year in 2023, right, I remember both you and we thought the DRAM probe card business might, may take another leg down, I mean, in the middle of the year, after the pretty meaningful dip in Q1 last year, right?
Mm-hmm.
But it turned out the business has recovered a little bit earlier than we had expected. Obviously, HBM is one of the drivers, but as we look forward to 2024, any thoughts on the continued improvement of the DRAM probe card business? And any color would be great. Thanks.
Yeah. So we do, when we report quarterly, break out probe card revenues into three segments for you. One is foundry and logic, our largest business. These are probe cards to test microprocessors, application processors, RF components, things like that. Our second biggest probe card market is DRAM. And DRAM has really been the tale of two sub-applications or two applications as we went through 2023. As Charles noted, and I think for those of you who follow the industry, DRAM and memory in general, right, DRAM and flash, were in a pretty tough shape, right? We had overcapacity, spot pricing going down, our customers not really investing in capacity, given the situation like that. And so as Charles said, we saw some pretty significant weakness in our DRAM probe card business in the beginning of 2023.
What happened through the middle of 2023 was we started to see significant adoption of high-bandwidth memory for generative AI applications, and that, for us, represents a really interesting probe card opportunity. Again, I'm gonna tie it back for you to advanced packaging. HBM is essentially a stack of individual DRAM die with a, a logic control chip at the base. But as our customers, people like SK Hynix, who's the leader in market share for HBM3, the third-generation memory, which is really driving significant growth in generative AI, or supporting significant growth in generative AI.
You know, that's an area that really took off as we went through the middle part of 2023, and we saw our DRAM business, which, you know, had been operating at kind of cyclical lows of $20 million in revenue a quarter, all of a sudden step up to around $30 million a quarter. That incremental $10 million was all HBM. And so the strength in our DRAM probe card business as we went through 2023, almost entirely driven by HBM. We expect that to continue through 2024, right? The momentum, that our customers are seeing associated with growth of high-bandwidth memory.
Micron presented an interesting slide at the SEMI ISS conference last week that showed that in one of these GPU plus memory HBM packages, the memory silicon area, the HBM silicon area, is at 7.5x that of the GPU area. So you can imagine from a production capacity, from a supply chain, you know, whether it's process tool suppliers, probe card suppliers, tester suppliers, HBM is a much more compelling opportunity than the GPU, even though the GPU gets all the headlines. You know, it's an area where, you know, if we're able to enjoy a bit of a overall DRAM, regular DRAM recovery together with the growth of HBM, I think you see some very strong results from our DRAM business as we go through 2024.
Great. So, Mike, I think you guys have been pretty good at, I mean, at least to put some quantitative color on the incremental opportunities, like things like a chiplet-based client CPU.
Yeah
But, any thoughts right now you think, for, what's—how much incremental revenue opportunity is there gonna be for your DRAM probe card business coming from the HBM?
Yeah. We've said as we went through the back half of the year, back half of 2023, began to understand, the scope of the opportunity, at least in the near to medium term, mostly driven by our key customers, their forecasts, and the purchase orders they're placing with us. You know, we expect at some point in 2024 to be able to double the run rate of that HBM probe card business. So again, at least the way I think about it, our DRAM probe card business kind of split, bifurcated into two pretty distinct applications right now. Regular DRAM, which is mostly DDR5, it's that right now, about $20 million a quarter in probe card revenue. HBM, which is 10, we expect can double to 20 sometime here in 2024.
If you superimpose some sort of general DRAM recovery on top of that, that's where we get to the logic of having some very strong DRAM quarters in 2024.
Got it. Got it. Appreciate the color. Yeah, and but how should we think about the competitive advantage FormFactor has in HBM probe card? How easy or difficult is it for your competitors-
Yeah
... to really develop a competitive alternative?
Yeah. So it's an interesting question. It goes back to the strategy, the philosophy I talked about of being very closely partnered with some of your key customers. You know, we've been building HBM probe cards, co-developing with key customers in the HBM application for about a decade, going back to first generation HBM, which I think was four die stacked together. We're now at eight, 12, even 16. And there's some pretty specialized requirements as you start to stack these die. One that I referred to, generally associated with advanced packaging, is higher speeds, right? Customers really wanna make sure that each of these chiplets, or in HBM's case, DRAM die in the stack, are operated at speed, and they understand they're gonna be good before they commit them to that stack.
So that drives a set of probe card requirements, that really push the speed requirements, the speed capability of the probe card, and that's not an easy thing to do. RF in general and high-speed digital are areas where we have significant know-how and competitive advantage going back. It's part of what motivated the 2016 acquisition of Cascade Microtech, a leader in RF, and we've integrated a lot of that know-how and technology into our overall probe card roadmap, including HBM. The other interesting thing about HBM, this was another area I referred to as being a general characteristic of advanced packaging, is temperature. Customers wanna test these chips at different temperatures, from down to, think about it as minus 30 degrees Celsius, so only a little colder than it is outside today.
Up to, you know, close to 100 degrees Celsius, essentially boiling water temperature. And you can imagine that if your a DRAM probe card makes contact with the entire wafer at once, and you can imagine that as I heat up and cool down this wafer, it contracts and expands, and because it's a stack of die, it does so in pretty unpredictable ways. The FormFactor technology, developed over the years, allows the probe card to stay in contact with all of the contacts on that wafer as it expands and contracts through this temperature range. So those are two key pieces of technical differentiation, both this temperature flexibility, temperature expandability, but also high speed.
Got it. Got it. So moving a little bit beyond the DRAM and HBM, now Intel has finally launched the Meteor Lake, the first, probably arguably the first high volume-
Yeah
chiplet-based consumer, processor-
Yeah
microprocessor product. As you said, I mean, this is one of the major opportunities for FormFactor. I know, you, in the past, you did provide some initial quantification about how much incremental probe card demand is going to drive, for this chiplet based microprocessor. I mean, conceptually, right, for us to look at that is like, instead of, having one monolithic chip, you probably probe once, you now have to probe four or five different chiplets [cross talk] that, that came from different wafers. But definitely the TAM, we probably cannot put a rosy number, like it's gonna expand by 4x or 5x. But, how should we think about how much incremental demand [cross talk] is gonna be there?
Yeah. As we look at all the different factors, it's definitely positive, right? When customers move to chiplets or advanced packaging, for the reasons we talked about, the need to screen out bad chiplets before they're committed to the stack are driving up the overall test spending by something like 20%-30%, is a good rule of thumb. And that's coming from two places from a probe card perspective. One, you know, goes back to the very nature of probe cards as a device-specific consumable. So if I break a chip up into four chiplets, I'm gonna need four different families of probe cards to test that. And so there's some increase there. Now, it's not four times, because each of the chiplets are simpler than the composite chip would have been. But there is an increase associated with that.
The other increase is associated with... The other component of increase is what I talked about with high-bandwidth memory, right? You're seeing higher speeds, higher test intensities, more comprehensive tests, which means each chip is, or each wafer is, essentially in contact with the probe card for longer. Test times are longer. And so that means for a given number of products out at our customers, they're gonna need more probe cards, more test cells, more tests. And so you roll all those things together, it's a big rule of thumb, right? But you roll all those things together, and on a like-for-like basis, we see something like a 20%-30% increase in overall test spending, test intensity. You see that reflected in what the probe card market's done over the last decade or so.
You know, 10 years ago, probe card spending, advanced probe card spending, was about 0.35% of semiconductor revenue, and it's climbed up in recent years to 0.45% or so. So a pretty significant increase in overall probe card spending.
Hmm, interesting stats. Maybe I wanna ask a bit more about the competitive landscape. You have a European competitor, Technoprobe. They apparently made some noticeable share gains in microprocessors, I would say probably in 2022, judging their results versus yours. It does appear the business may have retraced a little bit more than FormFactor did, I mean, through this downturn. The question really is, do you think the market share shift between FormFactor and Technoprobe has reached some sort of a stabilization right now? And any chance you guys can regain the lost ground?
Yeah. Well, I think if you look at... We've still got a quarter to go in terms of reporting 2023, but I think if you look at 2023, our results versus Technoprobe results, as you say, they're our primary competitor in the foundry and logic probe card space. I think it's pretty clear we've gained some share in 2023. The industry, like many segments, you know, in each of these major segments, there's really two leading competitors. And I think it's good for the overall industry to have two leading suppliers supporting our customers. Our customers want that for a variety of reasons, obvious competition, but business continuity and risk reduction. I think as suppliers, we all understand that we're operating in these essentially two competitor, two supplier segments.
Inside those segments, we're both gonna compete vigorously for whatever share we can get. Customers are gonna hold you within a band, right? They're not gonna let anybody run away with 80% or 90% market share. But we're each partnering with those customers, finding places where we have competitive advantage, really driving innovation around that competitive advantage and partnering with the key customers, and making sure we're capturing the most market share, but also profit share, that we can out of those segments. It's a similar story in DRAM probe cards, where we compete with Micronics Japan, a different competitor, and in the engineering system segment, where we compete with MPI, a company out of Taiwan.
Our strategy's really been about trying to bring together these different elements of related served markets, create a little bit of diversification and scale out of that, to make sure that we do weather volatility, like the current downturn, better than some of our competitors. I think you've seen that as we've traced through this downturn.
Got it. So maybe a little bit more, a topic related to Technoprobe. Last quarter, they announced that they entered into agreement with the Teradyne. I believe it's an important company that provides ATE.
Yep.
You provide the complements, complementary product to ATE, right?
Yep.
The probe cards. Technoprobe, they bought Teradyne's PCB operation.
Yep.
But more interestingly, they sold a 10% stake of themselves to Teradyne. I mean, for a long time, we thought that vertical integration in semiconductor test, I mean, kind of between ATE and probe cards, did not really make much sense. But yet, they seem to be taking that step in that direction. So what's FormFactor's view there?
Yeah.
Do you foresee any impact, positive or negative, to FormFactor business?
Well, I think one of the other important things about that announcement was both Technoprobe, but also more importantly, Teradyne, reaffirmed their commitment to an open ecosystem, where all probe cards work equally well on all testers. Whether it's a Teradyne tester or an Advantest tester, Technoprobe and FormFactor probe cards will still work on all of them. I think that's, you know, in discussions with customers, in discussions with the management teams of both Teradyne and Advantest, that remains an important principle that we're all committed to. We've with key customers in these discussions really strongly reaffirmed our commitment to an open ecosystem. We think it's important for the industry and important for us as well. I mean, think about our opportunity space with an installed base of tens of thousands of testers worldwide, split pretty evenly between Teradyne and Advantest.
We wanna make sure we work on both of them, otherwise, we've effectively cut our served market in half. So we remain committed to an open ecosystem. At least in the messaging and action since this, I think it was an October announcement, both Teradyne and Advantest continue to be committed to an open ecosystem. Time will tell whether this is the first step in vertical integration, but for now, FormFactor, and supported by and encouraged by all of our key customers, remain committed to working with both Teradyne and Advantest in preserving and driving this open ecosystem forward.
Got it. Thanks.
You're Shai in the mix.
Yes.
Yeah.
I'm starting to think, I'm starting to think I'm only here for my looks.
Okay, gross margin. I mean, we know you guys have been making incremental improvements in on gross margin line since, let's say, the dip, the 12 was fourth quarter 2022, right?
Right.
The gross margin, well, it has improved, but it's still a little bit below where it used to be. I mean, is it still a volume problem, which by would resolve by itself, I mean, as your volume goes back up? Or is there anything you have been doing or you can do, to so you can recover the gross margin a little bit faster, ahead of the time you think, the volume revenue can get back to that $200 million per quarter run rate? I mean, can you-
Get the margin back a little bit ahead of that?
Sure. So just to draw the line between where we were and where we plan to be.
Yeah.
Yes, we were at 38% or high 30s at Q4 of last year. We are now at the low 40s, at revenue, at similar revenue levels, so you already see improvement, even though revenue stayed about the same. And going forward, when revenue recovers to the $180 million a quarter level, we expect the gross margin to grow to about the mid-40s. And when we get back to the $200 million a quarter, which is close to our target model run rate, we expect gross margin to be at the high 40s. So 47%, that's the target model.
Mm-hmm.
A lot of the headwinds we see now is because of the capacity we've been adding in the last couple of years, and with a lower volume and the added capacity, this probably had somewhere between 5%-6% of gross margin headwinds.
Mm-hmm.
So with volume going up, we should be able to recover to these levels. But also, we need the volume to recover from our stronger market, which is the foundry and logic market.
Mm.
We talked about foundry logic, DRAM, and flash. The foundry logic gross margins are, on a relative basis, relative basis, higher. The recovery should come from there, which gonna help us improve the margin. So we talked about volume, we talked about mix. The third component is two main areas the company is focusing these days is increasing market share, and Mike talked about that, and the other one is improving gross margin.
Mm.
Things like, manufacturing efficiency, automation.
Mm.
These are things everybody in the company is focusing on. I expect to see this helping us, providing us more tailwind, and improving the gross margin to the target model and beyond.
Mm-hmm. Got it. Got it. So maybe you, can you talk a little bit more about the capital allocation, Shai? So over the past few years, obviously, I mean, investing yourself, investing, put money in CapEx, the money and capacity had been the main focus, right? And now, that you have sold the FRT business to Camtek, we'll likely see a little bit of bump in your cash balance. And any thoughts on how you plan to deploy the cash there? And any thoughts on going forward, what are the priorities for capital allocation?
Sure. There isn't a big change in our strategy when it comes to capital allocation. The priorities, as you said, in the last few years, were first CapEx, increasing capacity. We did say in the last earnings call, we're gonna trim down our CapEx in 2024 'cause we achieved most of our goals. We still need to make some investments, but a lot of the capacity expansions we wanted to do are behind us or almost behind us. So CapEx will go down from $60 million-$70 million a year to probably around $40 million-$50 million a year. M&A has been, and will continue to be, a big part of our strategy, and Mike talked about that, diversifying the business.
When we announced the sale of FRT, we also announced a buyback, a $75 million buyback to offset dilution from stock-based compensation. So these are the main, you know, capital allocation priorities. No significant change, with the exception of, you know, little less CapEx and the increase in buyback.
Got it. Got it. Okay, I think I run through all the questions I have. I think, we still have time. Maybe let's open up for questions from the audience. Any questions? Yes, please.
Can you explain why test and measurement can't be vertically integrated for either memory or microprocessors? You know, why wouldn't that be something to expect in the future?
And by vertically integrated, you mean inside our customers, or?
No, I meant like you, instead of having a Teradyne machine [cross talk] and a probe, you just have one.
Yeah. Well, well, there's no reason why it can't be, right? In many cases, these industry structure, dividing lines, borders, are historical in nature, and, and this one is probably historical in nature. If you were starting the industry today, you would probably integrate automated test equipment with the interfaces. But over the years, I think we've developed a relatively efficient set of interfaces and processes to allow this open ecosystem. And, you know, customers do take advantage of that open ecosystem, as we talked about, for a variety of reasons. And, you know, where business continuity wasn't a really big deal before the pandemic, it's a much more significant part of our customers' supplier management strategies than it used to be.
So if you vertically integrate, you know, as with most segments in the industry, you've kind of got two foundry and logic probe card suppliers, two ATE suppliers. If you pair them up together, all of a sudden, you've lost some of the optionality for our customers, and they don't like that. So they would certainly not be helpful in such a vertical integration move, and we've seen some of that in the discussions around the Teradyne Technoprobe transaction that Charles referred to. If you look at the benefits of vertical integration, they're not all that compelling, right? Yes, you could... As I said, we have a relatively efficient way of interfacing to these two different test platforms and other tertiary test platforms as well.
You know, there's just not that much to be gained by true vertical integration, either from a cost perspective or a performance perspective. There are certain applications where you can get rid of some connectors and do some things that might lower the cost a little bit, but I think from our customer perspective and the overall industry perspective, those benefits are significantly outweighed by things like the business continuity benefits of having multiple suppliers and an open ecosystem where the different subcomponents work equally well with the other subcomponents.
Great.
For the high-bandwidth memory, you said that you expect high-bandwidth memory will maybe double this year. Considering that generative AI activities are more than double, maybe you can give some much more insight why, where that double estimate comes from or why it can be more than double?
Yeah, I mean, you're looking at some fairly steep adoption trajectories with a variety of assumptions associated with them, right? So, you know, we wanna make sure that we're being balanced in our view of setting expectations, in sizing our factories, in making customer commitments. So a lot of our, a lot of the support for the data that comes behind the doubling of the HBM DRAM probe card business in 2024 really comes around direct customer interaction, us understanding what their forecasts are, what their wafer start ramps are, and it roughly triangulates with where most estimates of the generative AI end market and spending are likely to go.
You can imagine there's a lot of error bars on each of those estimates, and we're trying to be mindful of making sure that we do a balanced job of estimating all those different error bars as we think about planning our capacity, as we think about hiring, as we think about spending, making customer commits, and setting expectations for where this business could go.
Okay.
We have time probably for one more question. Okay. All right.
All right.
I think that's a wrap, and thanks, Mike.
Thank you for having us.
and Shai, for your insights, and appreciate your support of Needham Growth Conference, and hope everyone enjoyed the rest of the conference.