Okay, we're going to get started. I'm Tim Arcuri. I'm the Semi and Semi-Equipment Analyst here at UBS. Very pleased to have SiTime next. With SiTime, we have Rajesh Vashist, who is the CEO, and we have Beth Howe, who is the CFO. So thank you to both of you.
Thank you.
Very happy to be here.
So before we start, I have to read a statement. Before we get started, the company asked me to remind you that today's discussion includes forward-looking statements that involve risks, uncertainties, and assumptions, which are further described in SiTime's SEC filings, including SiTime's most recent Form 10-K, and that actual results could differ materially and adversely from those anticipated or implied. SiTime assumes no obligation and does not intend to update any such forward-looking statements. For more information, please visit SiTime's Investor Relations webpage at sitime.com. Okay, great. So let's just start out with the fact that you've had a very impressive growth story, particularly in data center over the past year. I think it's gone a little bit underappreciated until very recently by many investors, and can you just first start in data center and remind people where you play?
Yeah, so thank you for having us. What I've done is three businesses, business units. One of them is CED.
Hello.
Yeah. So one of them is CED, which stands for Communications, Enterprise, and Data Center. So data center is part of that. The short answer is we are pretty much everywhere because precision timing, the category we invented, is everywhere. We are in the XPU, the GPU, CPU, APU, TPU part of the business. We are in the switches. We're in the accelerator cards. We are in the networking piece with the optical modules, the smart cables. So sort of everywhere.
And within, can you remind investors why MEMS timing is so important in data center? And just generally, can you just speak generally to why timing is so important?
So timing, as we famously say, is the heartbeat of all electronics. So for people for whom it's a little hard to understand what it exactly does, think of it as a metronome. Think of it as every semiconductor in a system, a processor, communications, anything, power management, they need to know what is the reference point at which they need to fire, at which they need not send a signal and not send a signal. And so we send out famously a square wave. It literally is a square wave, looks like that. It's a timing square wave. And based on the edges of that square wave, they have to decide how to fire.
So now in a simple thing like a TV remote control, it doesn't really matter if you miss it or click it a few times, but anywhere where timing is critical, think a satellite, think a weapon system, think an ADAS, think data center, communications, it's very, very important. So we sell products that are in the 10 cents, let's call it 30 cents, all the way to $3, to $30, to $300, depending upon the timing. So timing, precision timing in particular, you mentioned MEMS. We invented this technology called MEMS back from 2007 to 2015, 2012. And we competed with a lot of companies, some large companies, some small companies. But in the end, we're the only ones who have shipped, let's call it 4 billion units or so to date in cumulative. The next level companies have shipped maybe 100 million, maybe 120 million units.
So the better way of thinking about SiTime is, yes, we have MEMS technology, but we have analog technology, we have systems technology. What we see ourselves is, Tim, as the timing company. So timing, wherever it's needed, that's where we want to play. Highly differentiated timing.
Can you talk about just the two parts of the market? Talk about oscillators and clocks.
Yeah, actually, there are three parts of it, getting to $10 billion , excuse me, as a TAM. So where we started was oscillators. Oscillators are made up of a resonator, at least one resonator, and a clock integrated into one. So we built that, we did it because it was the surest way of getting to the market. It's the highest order bit. It's a system play. Then we split out and we got into clocks, both natively and through acquisition. And recently, we've gotten into standalone resonators. So when we talk about the $11 billion market, it's about $4 billion-$4.5 billion resonators, dollars, $4.5 billion oscillators, and the remaining is clocks.
And can you just talk how bundling your MEMS oscillators with your clocking portfolio that you've acquired, how that helps?
Yeah. So it turns out that the clocks that are sold independently are typically very early in the system thinking. So a system engineer picks a processor, picks an architecture, and very quickly picks a clocking architecture. So that means that clocks get decided on by the customer before oscillators get decided. So by acquiring the Aura product and by building our own clocks, we get further up the food chain in making the decision-making, and then we can show up with our oscillators. And that's a unique thing because as of now, there are no companies that do clocks at scale and do oscillators and resonators at scale. The market is divided in two out of the $11 billion. There are clocks, there's clocks, that's mostly semiconductor companies, and there's oscillators and resonators, which is mostly quartz companies, and we're the only one that does it all.
Yeah. So let's actually talk about the competitive landscape. So from 2019 until now, you're still the only timing company. But back a long time ago, several tried to get into the market, all the analog companies, Maxim and IDT back at the time and SLAB. And the MEMS piece really is the key to all of this, figuring out how to do these moving parts in silicon. So can you speak to sort of what your moat is and how difficult it would be and how long it would take for someone to replicate what you do?
Yeah. So while it's true that MEMS is the defining technology, at the level at which we're playing, at the level of performance that we are playing, the analog mixed-signal becomes super, super important. So SiTime is an analog semiconductor company. And then putting the two together, which is, call it, the systems piece, makes it an even more challenging thing. So SiTime does have all three now, but rewinding the clock back to when I showed up in 2007 to a 30-person company pre-revenue, at that time, the MEMS technology getting that right was very critical, very important. And there are three aspects to getting MEMS right. One, MEMS doesn't have a process other than the one you build yourself. In other words, SiTime has built its own MEMS process. We're on the fifth, call it the sixth generation of that process.
We make it at Bosch, but we tell them how to make it. It's our process. It's not their process. We tell them, do this, do this, do this. That's important. The design is important because it's material science goes into it. The design of the resonator is critical. And finally, there are no tools. So there is no Synopsys. There's no Cadence tools. So SiTime uniquely rolls our own tools. So we have several engineers that are material science, software, physicists, mathematicians that build our own tools with which to design. So that's all only in the MEMS. Then, of course, as I mentioned, the analog, we are experts in PLLs. We're experts in oscillators, in mixed-signal circuits, and then we're experts in putting it all together. So I think the moat is around IP, actually patents, but it's also on know-how and the years of doing that.
One interesting thing back to the MEMS, when I look at other MEMS companies in other non-timing areas, Tim, once they're established, the way SiTime has been established, they tend to be always top of the heap. Think Broadcom, ex- Avago. Think Avago before that from HP. Think of Bosch. Think of all these other companies that have established themselves.
And are there pieces that you don't have from an M&A perspective? Are there areas that you could acquire? I mean, you did this recent deal with Aura, but are there other areas where you could consolidate?
Clearly, clocks, we're pretty small. We're sub $20 million in revenue right now. As I said, the market's about $1.5 billion. We're clearly relatively small. Our Aura deal was very successful in getting us design wins, but the design wins have taken their time as we expected. Clearly, in the clocking area, that's an opportunity. There's an opportunity in software. There are various synchronization softwares. We got some of it through Aura. Some we've rolled ourselves. There are opportunities in modules. There may be opportunities in IP. There's quite a few opportunities for M&A.
And so I just want to go back to data center. Is there sort of a rule of thumb? We've talked about this, but can you sort of go through what your content is within the rack? If we hear an NVIDIA rack, for example, is there a way to sort of aggregate what the content per rack is? People want these handy numbers to say, "Oh, if they're shipping an NVIDIA rack, how much SiTime content is there in that rack?
Yeah, I think Beth has a pretty good handle on it.
So as you think about the rack, and it really varies across, right? So NVIDIA is unique in that they're building an integrated rack themselves. And as you think about this timing content, really along two different vectors, compute vector, so CPU, GPU, XPU, as well as the networking fabric. And timing is critical in both, but if you have to kind of pick, networking is really where you see the benefits of our precision timing, whether it's in NIC cards, whether it's in optical modules, whether it's in the switches as you look. And so there's content across all of those different elements of a data rack there. And it really would vary from product to product or rack to rack, but you can think of certain racks where we've got high penetration and really well penetrated across the topology.
That can be multiple hundreds of dollars in a fully integrated rack.
And can you just talk, Rajesh, about just the MEMS piece of it and how really you're the only one doing it in silicon and how everyone else is using quartz? And can you just talk about just how there's the capacity benefit because you get 120,000 die per wafer? They have fixed volume, so it's a lot more flexible for you than them. I mean, there's a lot of factors there, but can you talk about those?
Sure. So back in 2010, as you said, there were those three large companies, but there were also five startups, and SiTime was the sixth startup that competed. So we had a lot of competition, and collectively, all these other companies spent $300 million-$400 million in about seven, eight years, and they kind of gave up. They gave up because they weren't focused enough on the three elements of MEMS that we talked about, the process, the design, and the tools. SiTime was, and we came through. So that we established. Going forward, what happens is the MEMS part, and I wish we could show it, is about one tenth the size of a quartz crystal. And that's a great benefit on its own. Okay, it's smaller, so that's good.
But what's really good about it is it's also small in mass, which means it's got low thermal mass. It's got low mass for shock and vibration. In other words, it's very, very much more resilient to shock, to vibration, to temperature movement, to temperature changes. And so what SiTime does is it delivers a uniquely stable reference through the MEMS product. And of course, it's much smaller and it's lower power. Also because it's semiconductors, it's built in a very, very clean environment. And so the quality and the reliability of that product is exceptionally bigger. It's 10x or 20 x better than in some cases, as much as 100 x better than quartz. Quartz is just not semiconductors. Semiconductors is the cleanest product known to humankind so far. So we get all the benefits of semiconductors, plus we are an analog company.
We combine these and we have a very unique product.
Can we pivot to the consumer opportunities? You're being carried along at a large consumer company with their internal modem. So I just wonder if you can talk about that.
Sure. Beth do you want to go?
Sure. So we have a large customer, and it's been reported that we do have two chips in their internally developed modem. And we've worked with them for a long time, really excited about that. The innovations that really went into that, as Rajesh was talking about, oscillators and clocks, we've basically married those two into a single system for them, along with another oscillator that we sell as part of the modem. The teardowns have shown that they're in two products right now with the internal modem, the 16e and the 17 Air. And I think we see a lot of opportunities as we head into 2026 for our consumer business as that proliferates into additional products that we've got great growth opportunities from that customer as well as other consumer customers that buy our oscillators.
When we think about the complexity and your timing gets more complex with new PCIe. So PCI shouldn't affect timing complexity much, I guess, but can you kind of talk about how the frequency changes is helping you?
First of all, as frequency goes higher, it's generally harder for quartz crystal to get to that same level of frequency, although they've done a phenomenal job of coming up from the last 10 years. So they do a pretty good job, but not at the levels that we're talking, say, 625 MHz. That's a hard frequency for them to hit. But beyond that, I think the throughput that we're talking about and the latencies we're talking about are very differentiated in our product. And finally, quartz crystal doesn't hold its stability as well as a MEMS product does because quartz crystal has, they call it hops and pops, where it starts vibrating at different frequencies, which throws the system off. And the SiTime ability to stick to the frequency is exceptionally great.
That's why it's a chosen product, particularly at 800-Gb and 1.6-Tb optical modules, for example.
And can you go back to what you talked about within data center? Is the opportunity bigger in the compute tray or more within scale-up and scale-out networking as the racks get more dense and the clusters become larger? And you did mention some ASP numbers, but maybe along the lines, are you engaged with all the different rack vendors now?
Yeah. So as we talked about, the opportunity is probably bigger in the network topology or the network fabric as we look across the opportunities. There are timing with CPUs or XPUs on the trays, but if we think about NIC cards or optical modules or switches, there's really timing content. Precision timing is really required across all the different elements of the network topology. And one of the great things about SiTime is we're addressing this AI opportunity is that we're working with most of the different players, right? So 10, 15 different cable providers, the different hyperscalers. So we've really got relationships kind of across that ecosystem, if you will. And so we think about optical module makers and the various ones there. We think about some of the switch players.
As you said, the content can vary from $1-$2 ASPs for some of the optical modules. As you go into some of the switches, you can have $7-$10 parts. So as you go up the stack, you can have meaningful content across the networking topologies.
Just going back to this most recently reported quarter, it was a watershed quarter, I would say. Is there any concern that there might be a little advanced procurement of your solutions and there could be any digestion on the other side?
Not from our side. We're very clear that we did see that in the 2022, 2023 shortages timeframe, and we got very sensitized to it painfully, I would say. So we've been very diligent about checking not just the disks and the OEMs, but also the contract manufacturers all the way that we can. And we think that this is real demand, and it's bolstered by our customers when they put out their numbers. We can see that, right? When we see a Credo publishing banner numbers yesterday, we can see as a customer of ours, we can see that that is very much in consonance with the actual numbers put out.
Can you talk about the potential for more oscillator opportunities? You're still a relatively small player in the market. I think Titan is probably not going to be until late 2026, I think.
That's right.
So can you just talk about sort of the ramp of the oscillator opportunity and how that rounds out your portfolio?
Yeah. I think, as you rightly point out, we're still a small company. Analysts have a set around $300-odd million this year, which is a significant growth year from last year, but it's still a relatively small number. The best way for investors, I urge them to think about SiTime, is to think about SiTime as a player across markets. It turns out that data centers are big right now and a big lift for us, but we're also in the C part of the CED. We're in the communication systems. We're in enterprise systems. Moving from there to what we call IAD, Industrial, Automotive, and Defense, we have a pretty decent automotive business. We have a decent-sized industrial business, and we have a small but very growing, very, very profitable defense business. Think drones, smart fuses, missiles. In automotive, we are very big on ADAS.
The more it gets towards level three, level four, level 3.5 to four, our content increases in radar, lidar, cameras, and in industrial, if you want to put robotics and humanoid robots, we think that's a great opportunity. We have several design wins in the big names known there. And finally, as Beth talked about, one very popular phone, but consumer products like watches, smart watches, a variety of home use case like doorbells and so on. So highly diversified and growing across all of it. So that's why we think, given our SAM of around $3 billion-$4 billion, given our TAM of around $11 billion, which will probably double in the next decade, I think we have a great opportunity ahead of us.
Just in terms of a large modem company, a large merchant modem supplier, all their timing is in quartz. However, they're also talking about building out their own racks for data center. Now they're pushing into data centers. So is that an opportunity for you to actually penetrate that customer and sort of turn the tide and say, "Hey?
We think so. We think so. We think that they're very open to using not homemade technology in many areas. You know that they have a big automotive push as well. So we think there are many opportunities all around. Most people don't have religion around this because we think we're not for everybody. We're for the highly differentiated products. So we're still only 10% of our SAM and 3% of our TAM. So we have a long way to go, but SiTime has the highly differentiated premium priced. We are always higher priced. So we have high gross margins. We guided towards 60% for this quarter. And 60% gross margins for a high growth company, I think that's pretty good.
Very good. Beth, so I want to talk about how to think about seasonality and particularly as you go into March because normal seasonal March seems to be down 10%-15%, but you're also ramping this modem opportunity. You're ramping in data centers. So how should we think about how much normal seasonal really is a factor for your business?
I think it's a great question that folks maybe don't always understand. Actually, our historical seasonality, if you look kind of three, five years, is more closer to 20%, 19%, 20%, 21% sequential declines from Q4 to Q1. As we've changed the complexion of the business and we've got higher CED, the data centers, I think it is moving more to the range you said around the kind of 10% to 15% sequentials as we see. There's a number of factors that go into that of not just our consumer business, but also early in the year, you see that across different customer bases. And so I think that's a reasonable assumption for the March quarter.
Relative to the consumer modem opportunity that we talked about, how much of a headwind would it be if they end up relying more on merchant modem for longer for your business? I mean, obviously, you want to see them adopt their own modem as quickly as possible, but how much of a headwind do you think that would be for your business?
Yeah, it's probably not much. It's a meaningful part of our business, but there's so much growth in the other parts of our business that it's okay.
And maybe also onto auto and industrial, it sounds like much of the secular growth is more in 2027. And just can you talk about why MEMS is important in those markets and just as important in those markets versus in data center?
Yeah, because as you rightly point out, most of the high level of level three, level 3.5, four, that's when they're emerging. They're emerging in 2027, which is where we play in ADAS. We also have another initiative around safety. SiTime has a product called FailSafe, which is the first of its kind. It's a safety-oriented, think of run-flat tires, which fail, get punctured, but still run for 50 mi or so. SiTime is coming up with this new category of product, which would make sure that timing in an ADAS system is fail-safe. That's an axis of development that is pretty innovative and hard to do, and we're one of the ones doing it. So we have high hopes for that.
Great. I also wanted to talk about cash, and I wanted to talk about CapEx. So over the past few quarters, CapEx was a lot higher than the mid-20% range and before coming down to like 6% in the most recently reported quarter. Why would CapEx have been that high for a fabless company? And maybe is it fair to assume that CapEx is going to stay below 10% going forward?
As we think about CapEx, we really think about the investments we need to make in the business. And so as you rightfully said, over the last couple of quarters, we did have elevated CapEx as we were ramping our new products and basically preparing our supply chain for those productions as we ramp into mass production for several of these new products. And so when it's important to make the investments, we will do that. In the second half of this year, we said it would step down meaningfully, and it has. And so as we think about it longer term, we're going to continue to invest in R&D for tools and our labs as we bring new products to market.
And also on the supply chain, where it makes sense for some of our unique tooling that we talked about, or where we want to make investments for some of these new products in order to support the production ramps at those OSATs, we'll make those investments. But it's really based on those new products and what we think we need to do in terms of the right ROI to basically drive the revenue that we want to drive.
And I would imagine that the consumer gross margins are lower. So as this modem opportunity becomes a greater piece of your mix, I mean, it depends on how fast that grows versus the data center stuff, obviously. But does that become a headwind to you being able to get margins above 61%? Can you kind of talk about the relative margins between the different segments?
Sure. So as we talked about earlier, we guided for 60% in Q4. As we think about margins over the long term, really that's why the mix is really the biggest driver of our margins as we think about our data center, our AI business being accretive margins, and why as we continue to grow that very significantly, it does provide us the opportunity to continue to expand margins. As you rightly pointed out, the consumer business does tend to be more dilutive, so in certain quarters where that may be a bigger component of the business, then you may see some impact from that. But over time, we do believe we can continue to expand gross margins, both in terms of our product mix and also as we grow the revenue, the leverage on the manufacturing overhead and being able to see some benefit there as well.
Really, our long-term plan is to continue to expand the gross margins over time.
Great. And I guess maybe we can. I just wanted to end on, you're not guiding 2026, but this year, I mean, your growth is going to exceed 50%. So I would think you can grow at least in a similar range next year, but given how strong data center is. But can you just talk about the puts and takes? I'm not asking you to guide next year, but just talk about the puts and takes on your growth rate next year versus this year. Do you think you can grow 50% next year again?
We haven't guided for 2026 quite yet. Trying to get that in there. We do see a lot of opportunity and strong growth across our segments we've been talking about over the last 25, 30 minutes in terms of AI and data center, in terms of consumer and industrial, automotive, defense. All of those create opportunities for us across our different markets. We see lots of vectors of growth that are going to support really strong growth for 2026. If I look at the strong momentum in the marketplace, I look at our design wins, our product funnel, all of those things I think give us a lot of confidence for a really strong 2026.
Great. Well, this is a very interesting story. So I really appreciate that.
Thank you very much.
Thank you.
Thanks.
Thank you.
Thanks so much.