Good afternoon, everyone. This is the 28th Annual Needham Growth Conference. My name is Charles Shi. I'm the Semicap analyst here. Joining me here is FormFactor. I'm thrilled to have Mike Slessor, CEO, Eric McInnis, CFO, here on the stage. And Mike, you've been coming here every year.
For a few years, yeah.
For a few years. Eric, first time and welcome.
Thank you.
Management would like to start with a few PowerPoint. Then we turn into fireside chat. Mike, stage is yours.
Great. Well, thanks, Charles, and thank you to Needham for hosting us yet again this year. It's a great way to kick off the year, and we really appreciate the continued support and invite. Great set of meetings, and hopefully, we'll be meeting with some of you as we go through tomorrow as well. We thought we'd start with just some level setting, making sure everybody understands where FormFactor's at this point. Trailing 12-month revenues for us about $750 million, although, as you'll see, we guided the fourth quarter up closer to an $800-plus million annual runway, so we're seeing growth in the business. Business composed of two segments: one, the largest business, and the business most people associate with FormFactor, a device called a probe card, and this is the interface.
Think of it as a connector between the test equipment built by people like Advantest and Teradyne and the customer wafer. Each of the chips on the wafer needs to be tested before it goes to the downstream assembly processes to make sure those chips are going to work. And with advanced packaging, things like chiplets and die stacking, a great example driving our business today is high bandwidth memory. That wafer sort step becomes much, much more important. Because you can imagine, in a 16-high stack of HBM DRAM, if one of those 16 die is bad, it's going to kill all the other 15 die in the stack. And so we've seen a substantial increase in what customers are expecting from us, from a technical capabilities standpoint, to make sure we are able to screen bad die out in these new die stacking paradigms.
Our other segment, engineering systems, a very important one, and has been responsible for our exposure in another exciting growth area called co-packaged optics. This is a set of tools that we've developed over the years to essentially help customers in early R&D as they develop brand new processes and technologies. And co-packaged optics, or CPO, is a great example of this, where we've been, for about the last decade, working with key customers in their labs, now transitioning this new technology into production. Of course, one of the big problems today in data centers is power consumption. And both scale-up and scale-out co-packaged optics is poised to help at least offset some of the energy consumption requirements associated with data centers. Light, a lot more energy-efficient way of transmitting data than electrons. We are a global company. Recent 10% customers you see down the bottom here.
But fundamentally, our strategy is to be closely partnered with the leaders in the industry to make sure our R&D investments are aligned with their needs. HBM is a great example of this, where we've been working with our top customer, SK Hynix, for well over a decade on HBM and the test strategy for HBM. And that's produced a strong leadership market share position for us in this new application for memory. About 2,250 people, again, arrayed worldwide. Eric's going to talk in a minute about some of the recent changes we've made. But one of our goals is to, and I'm going to skip over this slide. We touched most of that. I want to focus you on some of the recent results, both the top line, but probably most importantly, the gross margin line.
We've had some of these strong market share positions, great technology, good customer relationships over the years, but admittedly, we've come short in turning that into gross profit and earnings per share, and about a year ago, we undertook a pretty significant organizational change, brought in an experienced leader, Missy Stigall, who spent most of her career at TI, essentially as our global operations lead, and she's been working to build her team, a set of experienced people who really know how to run high-volume manufacturing, and you've seen that result in gross margin improvements as we've gone through 2026. We'll report the fourth quarter sometime first few days of February, but we've guided a sequential improvement from the 41% of the third quarter up to 42%, and as Eric will talk about, we think we can continue to make progress there as we go through 2026.
So taking strong market share positions, making sure that we're turning those into gross profit and real leverage through fundamentally taking advantage of these market share positions and driving leverage and scale. To that end, last week, we announced factory realignment, factory consolidation, and a reduction in force. I'll turn it over to Eric to provide some context on that before we get to the question.
Yeah, so before I dive in, maybe a little bit more context around our targets for the company. We've published a target model, which would have profitability at 47% for gross margins on $150 million in revenue. That target model was publicized back in 2020, and we've had a winding, curvy path since then. We went through the pandemic, and then we went through a cyclical downturn, and as Mike mentioned, we got to a point where it looked like things weren't really going in the right direction, so we made this correction in our global operations team that Mike alluded to, bringing in a team that's a lot more effective and efficient, and really, our next step on that path is what you saw us publicize this last week in 8-K.
That is really us taking the next steps that we need to do to consolidate our footprint, become more efficient, more effective, creating a more fungible and flexible manufacturing footprint. And it's also the first step in our transition of certain activities to our new site that we purchased a few months ago, back in May of last year, in Farmers Branch, Texas, which is in the Dallas area. So a couple of things of note. We're closing two sites in California. One is Baldwin Park, a relatively small site. That action is happening immediately. The Carlsbad site is actually not closing until the end of the year. So that'll be a very controlled transition as we're bringing that capacity up in our new Farmers Branch site.
Despite the fact that we're closing sites, this is actually the first step in us increasing capacity in the strong demand environment that we see, and doing that with the backdrop of becoming more efficient from a cost perspective.
All right. Let's start the fireside chat. Thanks for, I mean, giving us a little bit of color on demand, on technology, on gross margin. Let's get to it one by one. Maybe I'll start basically with HBM first. The recent developments, HBM, we think it probably will remain the, well, the central role in the AI memory, especially. But that's where your involvement is pretty deep over the last few years. You basically cover all three top three HBM players in the world. I think a question that often comes to us is, how are you positioned across all three HBM memory customers, especially in the context of the laggard out of the three? Now it looks like there's a chance for them to catch up.
Yeah.
And we know you do have competition. And how would that change, let's say, market share, competitive landscape in HBM, given the changing customer landscape?
Yeah. Well, as you can see on this slide, we are a key supplier to all three major DRAM manufacturers and have strong relationships with them. Having said that, our top customer, SK Hynix, who you can see a 10% customer in our filings, has really been a historical strong relationship for FormFactor, even predating me arriving at the company back in 2012. And so we've collaborated with that customer very closely. We have a strong market share position there. And candidly, it's a sweet spot for us right now with their very strong market share in the HBM market. Having said that, again, as you can see, we do have strong supply relationships with the other two DRAM manufacturers as well.
With HBM, one of the interesting things that's happened with DRAM probe cards is the requirement for testing speed has gone up quite substantially, not too surprisingly, as the chip speed has gone up. I think lots of people who pay close attention to the industry have seen this 11 gigabit per second specification that's been put out there that really stretches our customer's ability to build DRAM HBM stacks that operate at this speed. FormFactor's legacy. We really have a strong amount of know-how in high speed and RF test. I think it's fair to say that we're the only ones who can build a high parallelism, high productivity probe card that operates at these 11 gigabit per second operating frequency.
So if our customers want to validate that they're meeting this specification and do it in a cost-effective way where they're testing hundreds of devices at once, that means a FormFactor probe card. So that's a competitive advantage for us in this part of the HBM application. And it's driving market share growth at some of these other DRAM manufacturers as well because of that differentiated capability.
Great. So higher data rate has helped.
Yep. And will continue to help.
Will continue to help.
That's right.
This HBM is a race. It feels like anyone who has a little advantage can change the direction for the whole industry, where things should be, what specs everybody should achieve. So maybe I want to ask you a little bit more into the future. So let's say, at some point, HBM does transition from microbump to hybrid bonding.
Yep.
Will that have any impact on FormFactor at all?
It'd be an interesting technology change because it probably then requires us to probe on copper. Again, remember, when I described a probe card, think of it as a connector. The connection you make to the chip is typically through either solder bumps, aluminum pads, or copper pads. As the leading market share provider across all applications, foundry and logic, DRAM, all the sub-applications, we have years of experience probing on copper. So that's something that is on our roadmap to fuse together our copper probing technology with our DRAM test technology. I think that's an interesting inflection point for us that may offer some further competitive advantage.
Interesting. Let's talk about the other side of AI, GPU, ASIC. I know maybe going back two years, you were here. You talk about HBM, basically seven times the silicon area versus GPU. And HBM is where the money is at, I mean, from your perspective. But I think the AI volume gotten greater and greater. I think at that, you probably need to tweak a little bit.
Yeah.
But tell us about your GPU AI ASIC position as of today, and what do you think going into this year, next year, how things could evolve?
Yeah. No, that's fair. While we were very focused on developing our HBM roadmap, I think we were late to the game for the GPU test of this, again, because of the relative unit volumes. The interesting thing is that GPU test times are extremely long. This is a very test-intensive piece of silicon. The tens of billions of transistors on a GPU drive long test times, but they also drive some very interesting different, for example, temperatures that the chips require to be tested at, and so the volume there is significant, and the opportunity is significant. We have a very good relationship with the GPU leader. We have strong market share in their networking business. I talked about CPO being a growth driver for us. Obviously, there's some connectivity there.
Over the past six months to a year, we've worked hard on adapting our technology to make sure we can build differentiated probe cards for the GPU, the actual GPU test application. We've packed all the technical qualifications at the world-leading foundry. We have cards there that are now being exercised in pilot production. We've been very forthright with people that we expect to compete for this year's GPU business, kind of rounding out that overall piece with the large AI silicon customer.
Yeah. So one of the things more recently being talked about among your test peers is GPU AI ASICs. The power requirement has gone up.
Yep.
And I wonder if it's only about a package test, or is it something you are seeing on the wafer test side as well? And how's that changed, the probe card requirement? How does that help you? Or, I mean, I do expect that probably help you, but how does it work?
Yeah. Yeah. Power is a pretty big deal for a probe card provider because each of these connectors, these hundreds of thousands of individual probes on a probe card, you can think of them as tiny little structures, about the diameter of a human hair, and they have to carry a tremendous amount of current. Well, if you drive too much current through them, they turn into a fuse, and they're no longer a very good connector, right? That causes the test cell to go down, causes a productivity problem for our customer. We've spent a lot of R&D resources on making sure our probes can tolerate the maximum amount of current that's going to drive through that device. That then helps uptime.
And so as power and current continue to advance in the GPU and the ASIC application, we think there's an available competitive advantage for us in driving more power, more current at higher temperatures in the wafer test application. Now, that's important for the same reason I talked about advanced packaging and chiplets driving up the complexity of wafer test. Here's a great example. That GPU, or that set of GPUs, need to be verified as good before they go into the CoWoS package or the CoWoS package to make sure that each of them is not going to kill all the other HBM and GPUs in that package. And you can imagine the composite yield math is pretty daunting. That's what's driving some of these higher power requirements, higher temperature requirements that used to only exist at final tests are now moving upstream into wafer test.
So I want to ask you a little bit of double-clicking on that. Higher power helps you. I think a higher speed in HBM, you kind of tie your advantage, your technical advantage, back to your experience in RF.
Yep.
Right? But in terms of being able to handle higher power, where is your advantage? Where does your advantage reside?
Yeah. It's got to do with some of the investments we've made historically in our MEMS technology. We build each of these individual probes, again, 100,000 or so of them on a probe card, using a very advanced in-house MEMS technology. Think of these as, again, they're about the diameter of a human hair. But microstructures that carry a lot of current have to be very compliant to absorb the topography on the wafer, have to act essentially like RF transmission lines for high speed, and be super robust in some pretty unfriendly production environments. That's caused us, or driven us, to develop these microstructures that have six or seven different materials, right? Think of them as a Boeing Dreamliner airplane wing squeezed into something about the size of a human hair. It has to do multiple jobs. It's made of multiple materials.
And they're kind of, when I take a step back, astounding little pieces of engineering. Fundamentally, that's where the competitive advantage derives from. And there's only a few companies in the world that can do it. I know we're going to talk about the competitive landscape a little bit. But our competitors have, I'll call them comparable technologies. Part of that's what's caused the consolidation. About two-thirds of the advanced probe card market is owned by three suppliers. And that MEMS technology and those investments are one of the core reasons why.
I want to ask you more about CPO, co-packaged optics. This is a 10-plus-year continuous investment for you guys into this area, and finally, it seems like it's about to take off in some big way in AI. Can you walk us through what positions, what kind of products you sell into co-packaged optics today, maybe later this year, maybe next year? How are you participating in CPO RAM?
Today and at present, it's really a business and a set of tools inside our systems business, our system segment. And you can think of it as a probing system that doesn't just probe electrically, but also probes optically, right? Co-Packaged Optics, not surprisingly, you need to interact with the chip both electrically and optically. And one of the things we've done is taken the core pieces of enabling technology from that and ported them onto TEG probers. And together with Advantest, the 93K tester, is the first step in production silicon photonics, production Co-Packaged Optics. One of the other steps we've taken, we announced an acquisition right before the end of the year of a small German company called Keystone Photonics, who builds a version of these optical probes. So I talked about MEMS probes being an enabler for our electrical probe card business.
Keystone brings the same piece to the optical probing business, and so integrating electrical probing and optical probing technologies together down the road a little bit, we would anticipate turning this into an electro-optical probe card and making a much simpler test insertion for our customers, driving the kind of productivity we do on the electrical probe card front, but for today and for through 2026 in the initial ramp, these will primarily be very flexible, complex systems as our customers figure out exactly how to optimize their test flows for Co-Packaged Optics.
Got it. So far, more systems.
Yep.
Maybe in the future, probe cards, not just the electrical probe cards, but maybe optical as well.
That's right.
You can build on to some of the recent acquisitions, some of the technology you got from there.
Yeah.
Okay. I want to ask you next, I mean, on the more traditional side of the business. I'll start with conventional memory, DRAM and NAND. You have a sizable DRAM, conventional DRAM business.
Yeah.
It looks like it's coming back a little bit. It was pretty much run rate at a pretty depressed level for a long time. And until the most recent quarter, it seems to have ticked up a little bit. What's your view on DRAM growth? Do you still see that as cyclical? Can we break to a higher high this time? And the other area I want to ask you is NAND. Your NAND probe card has been a relatively smaller piece in your overall probe card business. And I believe it was associated with the Baldwin Park facility. And now you're closing it down. Are you consolidating that business over to your other sites? And what's your view on the NAND probe card business going forward?
Yeah. Let's start with NAND. NAND's never really been a big focus area for the company, right? Tends to be a much simpler wafer test step. You've heard me talk a lot about complexity and the value we deliver and things being challenging. NAND's not that challenging. And so for a company focused on expanding gross margins, we don't want to operate in areas where we can't differentiate. And I think NAND's probably best characterized still as a commodity probe card application where there's not a lot of available differentiation. So not a focus area for it. DRAM, on the other hand, is. DRAM test is much, much more complex. We've talked a lot about HBM. But one of the interesting things we're seeing, as Charles notes, is an upswing in our DDR5 probe card business.
When we guided Q4, we said that we expected to set another DRAM probe card record. But the strength was not going to come from HBM. It was going to come from DDR5. And that, I think at the time, surprised people a little bit. You now look at what the spot markets are doing with DDR5, and it's pretty obvious what customers are doing, right? They're shifting some of their mix back to commodity DRAM. It's an interesting time, right? Because there's a ton of demand for DRAM across a whole set of applications from high-performance compute through mobile, through general-purpose server. And our customers don't really have the ability to add a lot of capacity. They're clean room constrained. And so unlike any past cycle, I think we're going to be stuck in shortages for a while as long as the demand persists. And so an interesting time.
We'll be able to grow our DRAM business as we continue to make optimization in our existing footprint, improve yields, but also as we move, and I want to make sure we leave some time for Eric to talk about factory footprint and expanding in Texas as we view this as kind of a long-term expansion of our opportunity in DRAM probe cards.
Thanks. Maybe the next two questions I'm going to ask, one, back to the logic side of the probe card, then we'll move on to gross margin.
Okay.
So, Mike, in terms of logic, I'm not going to ask more about the smartphone, the mobile SoC side. But I think I'm going to more asking you about the microprocessor. This has been an area a lot of folks, including myself, were more concerned about the downside. But I think things seem to have shifted a little bit in terms of market sentiment, et cetera, on that part. Maybe there is more of an upside for you to explore. And then, speaking of that, the recent reports, which says that there are some shortages at Intel on their older nodes, 10 and 7. Any of those may impact your revenue in some surprisingly positive way? And by the way, the recent launch of some of the 18A-based products, like Panther Lake, does that help for your microprocessor business this year and beyond?
Yeah. Well, one of the subtleties of the probe card business is what drives demand is new customer chip designs, right? Being this connector or adapter between the test equipment and the chip means anytime the chip changes, that's a probe card refresh. And so something like these new designs in the PC space are pretty exciting opportunities. They're demand drivers for us. I will say a lot of people associate us with the PC and CPU business. As Charles said, that was a key risk for us. A few years ago, the world's largest CPU manufacturer was 25% of revenue. Recent quarters, they've been less than 10% of revenue. Yet we're guiding Q4 to all-time record revenue. So we've undertaken kind of this shift in focus towards high-performance compute. Still a very important customer and very important set of applications for us.
But as Charles said, probably better characterized as upside from where we are here.
Thanks, Mike. Maybe the last question on gross margin. So we understand from an operation perspective, you're doing a lot of work internally, right? And including some of the key hires and some of the manufacturing footprint optimization expansion, right? So 47% gross margin target, that was set about to be six years ago. But then back then, the assumption was very different from today. There was no such thing as a huge HBM revenue for you guys. And back then, it was, if you remember, going back five years, it was a chip shortage, right? A lot of things are happening on the logic side. Smartphone, PC was going very strong. So your product mix today is fundamentally very different from, let's say, five years ago. Your cost structure, I believe, is also very different.
Do you think 47% is still achievable given your much higher HBM, much higher DRAM in your revenue, which have been carrying slightly lower margin compared with the logic? And do you see the pathway to get back there given your product mix even today? Are you comfortable to say you have the line of sight to get there with all the pieces you know that will get you there? Or we're still waiting for one or two important pieces for you to bridge the gap all the way to 47%?
Yeah, so yeah, what we've said, to reiterate, is that we believe we're going to get to target model gross margins in the course of 2026, and you mentioned mix, and I think historically, maybe we've been a little bit victim to mix and not just mix, volume, right, so if you go back to when we put this strategic model into place, I mentioned that some things happened after that, we went into the pandemic, we actually were pretty healthy in terms of demand during that time, and it wasn't until we saw demand drop off and a cyclical downturn that we started to see some of the operational issues that we had, we had built up some capacity, we had loaded some costs into our cost structure.
It wasn't until we saw demand kind of decline after that that we started to become pretty apparent that we had built in some inefficiencies from a cost perspective. And not only that, we also started to see some shift in terms of our product mix, foundry and logic, less of that, and more memory products, which traditionally have a little bit different gross margin profiles. And so that became a problem. And we saw that come to fruition, if you will, in Q2 of last year, where we hit a low point of 38.5% gross margins. Now, what you've seen since that point is we came up from 38.5% in Q2 of last year to 41% Q3. And then we guided Q4 to 42%. So why is that occurring? It's not because of mix, right?
It's because we are focused on really, as one of the top priorities for the company, driving fundamental improvements in our cost structure in terms of things like cycle time, yield in our manufacturing process. Some of those things that we saw were underlying fundamental issues. As demand dropped, we're solving some of those problems, and probably the most recent thing that you've seen is what I mentioned earlier, a closure of a couple of sites that allows us to become more efficient in our deployment of resources, so we're able to make better use of our workforce when demand shifts, say, from DRAM to Foundry and Logic or vice versa. It allows us to reallocate those resources. We have some ability to use similar equipment across those product groups as long as they're co-located in a site.
And so you can imagine how us consolidating our sites together helps lead to some of those improvements. So in Q3 and Q4 guidance, what you see is some of those actions starting to take hold. And they are fundamentally changing our cost structure. I view them as durable changes. So trying not to be in a position where we're reliant on mix and even volume to hit our target model, right? We want to make sure that we have the fundamental cost structure that's supportive of our business model at any reasonable mix level. The other thing about focusing on yield and cycle times and just eliminating waste in our manufacturing process is that you get an additional effect. Not only do you decrease the unit cost of your products, you actually decrease your cycle time.
You get more good product out for each product or material in, which effectively increases your capacity. So in an environment like what we're seeing right now with DRAM, where we see capacity that probably exceeds supply, the ability for us to squeeze a little bit more out of our existing sites and our existing tools is pretty important and allows us to capture some incremental demand. So I think those are probably the fundamental drivers, and again, trying not to be victim to mix as we move forward, so in short, yes, I believe that we are on a path to get back to target model gross margins. Maybe the one point to clarify is tariffs. Those continue to be a headwind for us, so when we put our target model in place, tariffs were not something that we foresaw.
Those continue to be about a 200 basis points headwind to our model. But targeting that 45% after the 200 basis points for this next year. Now, maybe just a couple words more about Farmers Branch. So while we are consolidating some of our footprint, becoming more efficient there, we're simultaneously bringing up a new site in the Dallas area in Texas. As you might imagine, Texas has a different operating cost than, say, California. And so as we expand our capacity, what we're trying to do is make sure that we're doing that at a more effective, more competitive cost structure. And our Farmers Branch is a key part of that strategy. We're going to be ramping that site up over the course of this next year with first production out at the end of 2026 and then continue to ramp that site into 2027.
It gives us good optionality going forward as it relates to, again, meeting what we believe is the demand in growing markets, and we are going to be very focused on making sure that we bring that capacity online in alignment with the demand that we see.