My name is Amanda Scarnati. I am one of the semiconductor analysts here at Citi. I'm joined today by Shai Shahar, the CFO of FormFactor, as well as Stan Finkelstein, the Director of Investor Relations at FormFactor. The format of today's call is going to be a fireside chat. I have a couple of prepared questions that I'm going to dive into, please feel free to click that submit question button or send me an email directly at amanda.scarnati@citi.com, and I'll be happy to ask some of those questions a little bit later on in the session. Let's just dive right on into the logic and foundry business, the largest piece of the portfolio. You saw about a $20 million pullback in your largest customer Intel in the June quarter.
Was this more of sort of seasonality, or was it a shift in the business in terms of how they're looking at production capacity and manufacturing? Was this somewhat more of an expected quarterly decline?
Well, first of all, thank you for hosting us. Hopefully last time we do it virtually, and face-to-face next year, or hopefully even sooner. As for your question, prior to the second quarter here, we had several quarters where that customer was around and above actually $50 million or a quarter or $200 million annual run rate, which we said we expected was kind of an elevated run rate due to a variety of factors. In Q2, the pendulum probably swang a little bit hard in the other direction. Embedded in a Q3 outlook, some sequential strengthening with that customer, and we believe a more normalized annual run rate for that customer is probably around $150 million, $160 million a year. Higher than the Q2 run rate, but lower than the $200 million historical run rate or the last year run rate.
As you know, longer term, there are even some really exciting things going on with that customer, right? Recent announcements are very encouraging. We have a long-term relationship, decades long, where we have been a key supplier to them, with strong market share and both their strategic initiatives in entering the foundry business, as well as their adoption of advanced packaging technologies that they highlighted earlier this quarter, and they are really exciting opportunities for us, and I do think allow us a path to go back to kind of this $200 million run rate in the future. Now, it's going to take a while until we get there. They are not entering this foundry business in a material way tomorrow.
We've got such a solid foundation with this customer and a nice competitive position there with the technologies that we have and their strategic initiatives, they do offer us longer term path back to this $200 million run rate.
Perfect. The other large customer in the foundry and logic space, obviously is TSMC. Historically, that was not necessarily a larger customer for you, but it did cross the 10% threshold for the first time this year on a quarterly basis. Can you talk about the growth that you're expecting at TSMC, and why TSMC is just now only becoming an important customer?
Yeah. Actually, TSMC, I think, made it to the 10% customer list in the past. It's a great success story for us, growing our share with them from almost zero, if you look back four years ago, to around 50%-60% today, basically sharing the share with our competitor. We are the main or the only suppliers for everything, which is 10 nm and below with TSMC. Now, there is some seasonality with the orders from us, as we saw in the last two, three years, associated mainly with the mobile application processor business. They are going to be in and out of the 10% customers list in any given quarter. We should expect that to soften in Q3, as we said in our earnings release, in our last earnings call.
At the same time, we are making good progress in expanding that business with the customer to more fabless customers, more designs, moving into high-performance compute. Mobile is such a big part of that business now. We could see that customer not be a 10% customer in Q3, but continue to come in and out of that business of that list in the future.
TSMC is still sort of in-sourcing a lot. You mentioned that 10 nm and below is somewhat outsourcing. How much do you think TSMC will continue to outsource, and do you think that your share could remain at this 50%, or do you think there's going to be some puts and takes in that 50% range?
Yeah. As I mentioned in the previous question, TSMC started outsourcing their wafer testing with the 10 nm and below processors. This is when they realized that their in-house technologies just cannot support the process, and they needed to use our technology, or to be fair, the technology of our competitors. Today, we estimate, as I said, about 50%-- well, not as I said. About 50% of their business is being outsourced, and that number is growing as they move more and more processes into 10 nm and below. As you can imagine, there are some designs that will stay on legacy nodes of 16 nm, 18 nm and above for a long period of time. This just doesn't make sense economically to move them to newer nodes.
Now, as TSMC move more designs to the newer and more advanced nodes, it means the outsourced part is getting bigger, and with the significant market share we have, it means more business for us. Now, as I said, we have about 50% market share at TSMC, and since like many of our customers, they do like to have a second source, this market share will likely stay around this number going forward as they allocate each design between us and our main competitor. It can be on any design. It can be 50/50, 60/40, 40/60, within these ranges. The encouraging part is that more and more things are moving into 10 nm and below, which means the pie is getting bigger, and as long as we keep our market share there, that's a business that we think can grow into even the levels of our largest customer today.
Are they splitting out this market share based on purely technology? Maybe the apps processors go to you, some other technology might go to your competitor, or is it more sort of dual sourced on each technology process?
It's really dual source on everything, and it can be 50/50 or 40/60 based on how they allocated the previous design, what's the on-time delivery commitment, things like that. Our technology is better in some aspects. Our competitor's technology is really different in others, but no significant changes there or differences.
At your Analyst update, you highlighted that you expect to add about $200 million in incremental annual revenue, driven primarily by logic and foundry. What's driving this incremental growth outside of what's happening at TSMC?
Right. As you said, last year in our August Analyst Day, we increased our target model by $200 million, from $650 million to $850 million. The growth will come from all the markets we serve, but mainly from foundry and logic, as you mentioned. Advanced packaging and 5G are really the main drivers for this growth. The growing business with TSMC is a key component of that, as I mentioned. We also have some opportunities to grow share at customers we currently don't see at our 10% customer list. Our RF business, which is also included in part of our foundry and logic market, is also growing nicely. All of these together, these are the drivers that will get us to the $200 million additional revenue.
Switching gears over to DRAM, moving away from logic and foundry. In the past, you've mentioned that you expect DRAM to be closer to about $30 million per quarter across the cycle. Can you talk a little bit about the dynamics in this business and what drove revenue to about $40 million in the June quarter?
Sure. We have seen and continue to believe that the DRAM business is cyclical and will continue to be cyclical. It reached a high revenue peak of low $40 million in a couple of quarters in the last two years, including in previous quarter, in Q2. We also saw it going down to $20 million, $25 million a quarter after such a peak when our customers went through a digestion period of the I-consumption from us. We now think an average run rate of mid-$30 million a quarter is what we should expect going forward. From DRAM revenue, will continue to fluctuate from these low $40s to low $20s. In the past, we talked about maybe $30 million a quarter, where now we think about $30, $35, mid-$30s probably more makes sense for that run rate for the business on average.
Since our products are really design-specific, the design activity is what drives these fluctuations. Every change in the design by our customers require a new set of probe cards, and the peak quarters we saw were related to ramping of new designs, things like a move to DDR4 and DDR5 that we see now. What we saw specifically in Q2, when revenue reached $42 million, almost the second highest year on revenue in the last 12 years, we saw a significant ramp of one specific design with one of our largest customers, which drove DRAM revenue to be above $40 million. That design is still significantly contributing in Q3, although not as strong as in Q2, we expect DRAM revenue in Q3 to be probably in the mid to high $30s.
On the March quarter earnings call, you highlighted different margin mix within the DRAM business, and that had a major impact in what we saw in the margins in the March quarter, and then a little bit of a rebound then in the June quarter. Can you talk about what the different product types are within this business and, what will drive that significant swing in margins?
Sure. Maybe I'll start with a high-level review of the different gross margins in the markets that we serve. We operate in two segments, the systems segment and the probe cards segment. The systems business has the highest gross margin, ranging from the low 50s to high 40s, followed by foundry and logic, DRAM, and then flash. With the variety of products that we have, there is of course some overlap in margins between these markets, right? A high-level DRAM design could have a higher gross margin than a low-level foundry and logic design. What drives the margins are the complexity of each design. For example, a one touchdown design, a card that touches and tests the entire wafer with one touch, will likely have a higher margin than a probe card that uses two touchdown designs.
What we saw in Q2, and to a certain extent it will still be with us in Q3, is a significant ramp of one specific two touchdown design that has lower margin. That was the biggest impact on the mix. In future quarters, we expect such fluctuations to continue as we make our way to our 47% target gross margin in the long-term target model. Mix certainly has the biggest impact, as I described above.
Let's shift gears over to the systems business. The business has been quite a bit lumpy over the last four quarters, bouncing between $27 million a quarter-$35 million a quarter. What causes this lumpiness, and how should we think about this going forward?
Maybe I'll start with an overview of what is our systems business, because some of the audience might not know. This business accounts for about 20% of our business. These are tools, essentially wafer handling and environmental control pieces of hardware, that are primarily used early in the development and pathfinding by our customers for electrical and increasingly optical measurement of their devices to make sure that they are operating correctly. They are used to improve the yields, to improve the performance, even before we get into the production. This is in the R&D part of the business. This is a business that we break out separately because it has some pretty interesting financials, as I said, higher gross margin, for example, but also equally strategic importance for us.
It allows us to engage with the leading semiconductor device manufacturers very early in the development cycles, and then help them debug and improve the yield and performance for all kinds of new devices, from regular CMOS semiconductors all the way to innovative things like CMOS image sensors, VCSELs. The main customers there, in addition to the R&D organizations of our customers, will be universities and research institutions as well. This business, since it's R&D driven, it's less cyclical than what we see in the probe cards. We do see some seasonality there, mainly stronger second half of the year, and particularly Q4, as our customers are looking into using R&D budgets before year-end and before they lose this budget.
It has been more volatile recently also because of the impact of the pandemic, because sometimes the pandemic caused delays in installation, mainly due to travel restrictions and the quarantine requirements and things like that. Now we kind of learn to live with that and together with the seasonality, that business was strong in Q2, and we expect it to grow even more in Q4, as we said in the last earnings call. I think, yeah, I answered your question, right?
Yeah. Obviously, a hot topic in the sector today is shortages and capacity constraints, and what's happening in the market in that regard. You've added quite a bit of CapEx in 2020 and are planning for even more CapEx in 2021. Can you talk about sort of the underlying demand that's driving this? Is it more sort of secular longer-term demand, or is it a function of some of the semiconductor tightness that we're seeing in near-term demand?
We did talk about $70 million-$90 million of CapEx investment in 2021. This decision to add capacity was made in late 2019, first by leasing an additional building in our Livermore campus, and then deciding to purchase that building in early 2020 when the landlord presented us with the opportunity to buy it, and the terms were just great for us, and so we decided to do that. That decision was made late 2019, and this can tell you that we saw the long-term secular growth drivers even before what we currently see as the semi market tightness that you mentioned. For us, the main growth drivers, as I mentioned before, are advanced packaging and 5G, more than specifically the shortage that we see today. That capacity is supposed to be online during the second half of 2021, so very short.
We expect it to contribute to revenue just a little bit in Q4, nothing significant, but more significantly in 2022 and going forward. This capacity addition will help us to get to our $850 million of revenue target model and beyond. We need that to get to the $850 million, but that additional capacity can, of course, give us more than that if we need it. We plan to be very measured on how we turn on this capacity. We want to make sure we align it with the demand that's coming in from our customers. We're going to do it very slowly in a very measured way.
With some of the tightness that you've been seeing, and you've been vocal about having tightness and under-shipping to demand, is there any concern that you've lost out on potential designs due to some of these capacity constraints? Or are you really just seeing maybe some lead time stretching out in order to meet some of that demand?
I think we were capacity constrained mainly at the end of last year when revenue reached $196 million, peak revenue for us. To a lesser extent, we still have some pockets of constraint even today, here and there, because we have few factories around the world. Currently, it's not significant. We, again, learn to live with the constraints and all the COVID restrictions and social distancing, et cetera. The additional capacity that we put in place also helps with that. In terms of losing designs, from what we saw, we did not lose any designs, mainly since our competitors were in a very similar situation and were also capacity constrained at the same time. For example, our largest competitor, Foundry and Logic, is a company located in North Italy. As you can imagine, they had issues with capacity constraints as well.
How are you being impacted by any of this supply tightness that we're seeing in the market? We're seeing supply tightness leaking into other areas of the market, materials spending, equipment spending, or securing supplies and materials and equipment. Are you seeing any impacts in that regard in securing supply?
Well, we continue to see strong demand for our products, as evidenced by the last four quarters, including the current quarter revenue levels. Since our product is design-specific, it will not make sense for our customers to pre-order or order products that will just sit on the shelves. We are a six to eight weeks lead time business, so our customers will order us the last possible minute they can when the design is final. On our side of the supply chain, we don't see double booking or double ordering and things like that. On our side of the supply chain, so far, we haven't seen any significant impact.
We saw small things here and there, maybe a probe card, a $300,000 probe card was waiting for a small transistor to arrive to finish the production of it, but it was really small here and there with no significant impact on our revenue. It's something that we continue to monitor very closely as we hear more and more of our peers struggling. It's kind of a part that we are helping to manufacture semiconductors, but we have a shortage of semiconductors for some of the things that we need, and it will take some time to resolve in the industry. For us, we don't see a significant impact so far.
Let's talk about some of the recent acquisitions that you've done. Last year, you added two new companies, mostly looking to sort of extend some technologies. You have Advantest probe card business and then also High Precision Devices. How have these acquisitions been tracking towards your expectations in terms of ramping up that technology know-how?
Yeah. We are very happy with the progress made with these acquisitions and how they're integrated into the FormFactor family. The business that we acquired from Advantest was mainly technology-driven, and we are in the process of qualifying these products with additional customers, and also using it in markets that were not used before. That business was only serving the flash market. Now we are looking into how we can use these technologies in the DRAM market as well. Good progress was made to date, but these things take time, as you can expect from qualifying a new product with new customers. I expect this to start contributing in a more meaningful way probably next year, not this year. HPD is a little bit different story. They brought us the cryogenics and quantum computing technologies.
It is very interesting, very exciting opportunity for us to participate in this market. This is more of an emerging growth market for us. Until we see it significantly contributing to revenue, it will probably be 2024, 2025. It's not even included in our long-term target model of $850 million, but will be on top of that, more longer term, because this technology is still in early stages when it comes to mass production and adoption.
Are there potential for additional acquisitions? What could some of those potential areas of interest be to expand the portfolio, either organically or inorganically?
We still have over $250 million of cash on our balance sheet with very limited debt. M&A continues to be very important part of our strategy. As the market leader in almost all the markets we serve, and with our customers' intentions to have second sources, we have small opportunity to increase market share in the current markets that we serve. Nothing significant. We talked about from low 30% to probably mid-30% market share. With acquisitions, we plan to increase the share, which is exactly what we did with our recent acquisitions of FRT and HPD, that you mentioned. We plan to continue to do that. We're looking into additional, smaller tuck-in acquisitions, probably in our system segment, just like we did with FRT and HPD. These were like $15 million, $20 million acquisitions, in that range. Can be a little higher than that.
We also look into larger acquisitions, similar to what we did with Cascade Microtech back in 2016. This will be in adjacent markets because there is no obvious Cascade out there anymore that will be a good target. There is also no point in us buying one of our competitors, since our customers will just qualify another vendor as part of their second source strategy, right? These larger acquisitions naturally take more time to close, especially these days when most conferences like this one is virtual. As you know, it's kind of hard to bump into someone in Zoom. It continues to be part of our strategy, for sure.
With that $250 million in cash, if you're looking more towards some of those smaller tuck-in acquisitions, how do you think about returning cash to shareholders, either through a buyback or through a dividend or other means as well?
We do have, in terms of capital allocation priorities, obviously the expansion that we have in place is the highest priority at this point, as long as continue investing in R&D. We talked about M&A, and we do have a buyback plan in place. We announced it in October. It's a $50 million two-year plan, mainly to offset dilution from stock-based compensation. So far, through the end of Q2, we executed about half of it, $24 million of this $50 million. It's a two-year plan. It's been one year. We executed about a half, everything is going as planned. I would say this is what we have in mind for now is the priorities for capital allocation.
What would you say are sort of the four key takeaways that investors should really focus on for FormFactor? Is it really focused in on that opportunity at TSMC and the growth there, Intel? Are there other opportunities that you really think investors should focus in on?
I think it's important to understand that the whole concept of us being a design specific, almost a consumable. We are not really driven by volume, but the more design activity there is, it means our customers need more probe cards. In the past, it is not only about node transitions from 14 nm to 10 nm to 7 nm to 5 nm, but also in the same node, if there is a change in the design, a little tweak, it means they need a new set of probe cards. That help us diversify the business in terms of with DRAM and flash and foundry and logic and systems. That's something that really helped us show our revenue being stable and growing versus our peers in the past. The advanced packaging and 5G, these are mega trends. They are certainly helping us. Moore's law is slowing down.
Some say they are dead, it's dead. We see more and more customers innovating on the back end because on the front end, it's harder to innovate. That's important. We are serving all the big five manufacturers in the world. They're our top five customers, even if they didn't make it to the 10% mark at least, they are our top customers. We are the market leader. We are the only one that serves all the markets. We have a different competitor in every market that we serve. We're the only one who serve both or all, foundry and logic, DRAM, flash, and the systems business. That gives us a view of everything that R&D that we deploy in one area, it's something that we can learn from and apply in different areas as well. I would say these are the main things.
We are serving all these customers, and we have opportunities to gain share with other customers that currently are not up there on the list. Very exciting for us.
Yeah, I think that's something of a misconception, something you highlighted before was that FormFactor is really a consumable business. A lot of investors still focus in on some of that DRAM story and then maybe some softness coming out of the market, pricing coming down, and there really has been some of this decoupling between pricing dynamics and what Form is able to do. Can you just talk a little bit about what your expectations are if this market starts to soften a little bit on the DRAM side? Do you see that having an impact or is it really still driven by technology node changes which could happen even in a softening environment?
I think that a great example is if you go back to 2019, where our peers on the CapEx area went down 25%, we grew. That's mainly because most of our business today is design driven, and these fluctuations in the DRAM prices really don't impact us significantly. We are really about the design changes. DRAM also is not a huge part of our business as it used to be. It's running around, as I said earlier, on average of mid-$30 million a quarter, while the foundry and logic is around the $100 million a quarter, and the systems is $35 million-$40 million a quarter.
DRAM is still cyclical, but with the diversification that we have in place in customers and in the markets that we serve, and the fact that we are design specific and design changes driven, it's not the same company that it used to be 10 years ago when it was really a DRAM cyclical company.
I would also add that over the course of the last decade, obviously DRAM industry has gone through pretty major consolidation, and you're down to three players that are fairly rational. Whatever fluctuations in prices we absorb, overall, I think it'll have minimum impact on new design cadence, which fundamentally drives our business.
Just got a question in from a client in terms of volumes versus ASP on the probe card. Is there any sort of major difference between ASPs across the board, logic and foundry versus DRAM versus flash? How do volumes work? It's a consumable product, so are you shipping high volumes per design or is it one probe card that gets continuously used over and over and refreshed when there's a new design?
It really changes from customer to customer and from the different markets. In general, the reason we don't serve the flash so significantly and revenue is only mid-single digits, maybe low teens, is because this market is just the probe card is relatively simple. It's a simple die to test. We only participate at the high end, skim the top, high margin products. You could have a high-level flash probe card that is more expensive than a low-end DRAM probe card. Even within the DRAM, you can have a one touchdown probe card that costs $300,000 and a two touchdown probe card that costs less than half than that. It really varies in the different markets and between the customers, and it can be kind of all over the place. In terms of volume shipping, we usually don't ship one.
It's serious. We recently celebrated sending the 1,000th probe card to one of our customers, not in one year, but over a few years. It also depends if the customer have how many tester they have available and how many testers they want to use to test and how fast they plan to ramp it. It really can be all over the place, but it's usually not one unit. They want to duplicate the testing program is over few testers using the same type of probe cards.
Perfect. All right. I think we are just about out of time here. Shai, Stan, thank you so much for your time, and enjoy the rest of your day. Thanks.
Thank you, Amanda. Thanks for inviting us for this conference. Always a pleasure.
Thank you.
Bye-bye.