All right, we'll go ahead and get started. Welcome, everybody. Thank you for joining us at the 28th Annual Needham Growth Conference. My name is Quinn Bolton. I'm the Semiconductor Analyst for Needham. My pleasure to host this fireside chat with SiTime. SiTime is a precision timing company whose semiconductor MEMS programmable solutions offer a rich feature set that enables customers to differentiate their products with higher performance, smaller size, lower power, and better reliability. The company has cumulatively shipped over 3.5 billion timing devices to date. Joining me from the company is CEO Rajesh Vashist. Rajesh, thank you for joining us.
I'm very happy to be here.
Before we start with my questions, the company has asked me to read the safe harbor language. So these discussions include 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-Q, 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 forward-looking statements. For more information, please visit SiTime's investor relations website at investor.sitime.com. With that, Rajesh, SiTime is a leader in precision timing. Can you start off and just sort of describe what is precision timing?
Yeah, so first of all, what is timing? Timing is the heartbeat of any electronic system. Simply put, a signal from a CPU, a signal from wireless, a signal from anything needs to know when to go, when not to go. And the precision part of it comes in that there's a lot of performance issues, which Quartz Crystal, which is the incumbent technology, has managed for the last 70 years. But what SiTime brings with precision timing is all that performance, but under tough environmental conditions: temperature, shock, vibration, power, airflow, size. When you factor that in, our approach, which is semiconductor-based new technology vs non-semiconductor, quartz, we think wins.
How large is the precision market, sorry, precision timing TAM? And how much of that opportunity do you address?
Yeah, so the full timing TAM, which is made up of three different kinds of components, is around $10-11 billion. It grows at 5-7% a year. It's been doing that for the last 60-70 years. And so in a decade, it's probably going to be twice as much, so $20 billion in a decade. The SiTime SAM, which is a share of the market, is about $3 billion. And of course, our revenue is significantly less than that.
What is your sort of strategy to increase your SAM within that total addressable market?
Yeah, so quartz does a fantastic job of taking care of most of the existing use cases by definition. SiTime's job is to deliver a highly differentiated solution, which they can't do. So differentiated in terms of speeds and feeds, we can do better phase noise, we can do better jitter, we can do better stability. But also, we see ourselves as a solution set. We see ourselves as solving a tough timing problem, which includes having a resonator solution, having an oscillator solution, having a clock solution. There's no other company in the world that has said that, that is capable of that, and that is delivering that. So we are delivering solutions that are premium priced, high gross margin to us, but premium priced, which gives us a single source, 70%-80% of our business single source. So what's not to like about that?
Right, right. Just went through some of the advantages of silicon MEMS timing devices vs quartz. Is there anywhere any metric where quartz is still competitive or where you are trying to catch up with quartz, or do you think you now have sort of performance leadership pretty much across the board relative to the incumbent technology?
I think when we went public, we were still deficient in some areas. That was six years ago, 2019. But since then, I think it's true what you just said, which is by and large, SiTime, there may be still some isolated pockets where we are not as good. For example, in space, we haven't done the work for deep space and rad- hard, radiation hard. But you know that's a relatively small, maybe $100-$200 million market.
Okay, but in terms of the metrics itself, jitter, temperature, stability, you feel like you have equal, or if not better, performance?
That's right.
Okay, perfect. If you look back at 2025, it was a record year for the company. Maybe talk about some of the biggest growth drivers that you saw last year.
Yeah, for the first three quarters, which is what we have talked about, obviously, including our November call earnings, just reiterating that. Obviously, AI, no surprise, has been a huge growth. In November last year, we said that it was our sixth consecutive quarter of growing our what we call CED, which is Communications Enterprise Data Center, by triple digits, so 120%, 130%, 140% for six quarters in a row. So that was clearly the banner, the headline. But the other was a slow creep up on improving gross margins. Another was consolidating our Aura acquisition from 2023, December, and getting those products out, getting those design wins, showcasing that our viewpoint, which is clocking and oscillators selling together for certain markets, is very valuable. So those were three of the highlights that I can think of.
Does anything change as you look into 2026 in terms of your priorities? I mean, you've shown strong growth. You've seen the margin expansion. Any other priority you would have as you look into this year?
I think, as we've said before, we look to get to 60% gross margins and keep that. 60%-65% is our sweet spot for gross margins. I think we keep on getting into that part, that side. AI continues to grow for everybody, and one can't say where things go, but in general, we see AI growing. We kind of gave a sort of soft guidance in November, and for 2026, our base growth rate, base, meaning our is 25%-30%. And we have done that consistently, and then in years like 2024, we did 40, and 2025, we'll do whatever we think. As of the last quarter earnings, there's more than 40. So as long as we have a tailwind of some kind, we have 400 applications, about 5%-10% of that, they grow more than about 25%-30%.
And we have just kind of begun, right? Our revenue is really still pretty small compared to our SAM and certainly compared to our TAM.
Wanted to move now to the CED segment, which, as you said, is certainly led growth last year. Needham sort of estimates you're probably on track to grow about 130% year on year in 2025. Can you talk about applications within the data center and use cases that are sort of driving that growth? How broad-based are your timing devices within some of these data center applications?
Yeah, as we have maintained in the past, we are in almost all aspects of the data center. But the part that we really seem to shine in a lot is in the networking and the communications piece. Because yes, we are next to the GPU and the CPU, but the switches, the accelerator cards, the optical piece, both scale up, scale out, all of those are great opportunities for us. And we continue to see that networking, we see the GPUs are underutilized many times. And part of it is because they're not able to get the data around the way they should. So clearly, networking is the bottleneck.
If we want a better utilization of the GPUs, we need better networks.
That's right.
Better networks probably require higher precision time.
That's right.
Got it. Can you discuss sort of, you know, we're kind of moving in the data center from unit of compute from the server to sort of the unit of compute is now the rack?
Right.
Roughly, how much dollar content would SiTime have in a, say, NVIDIA rack of equipment? Is that $100s ? And how do you see that trending over time?
Yeah, so it depends, of course, on how the rack is configured and so on. But in general, we have said in the $100s for a Blackwell, for example, picking a number, $500-$600 per rack, fully configured. The previous generations would probably have been more close to $400. The future generations are probably going to go up. So yeah, it's an increasing SAM for SiTime.
Okay, and you look at NVIDIA, obviously, a number of the hyperscalers have their own ASIC platforms. Is it generally similar dollar content opportunity in some of the ASIC platform racks?
Yeah, I think NVIDIA is sort of the purest example of them. Others are not quite everywhere in the same way that NVIDIA is. So it necessarily is a little bit lower. But eventually, at scale, everyone will need that kind of timing, right? The most important thing to understand about timing is that in any use case, but let's talk about data center, is that the data throughput, the data, in other words, bits per second is critical. And so shoving more data around is critical. Low latency is critical. But the new thing that's coming up is the synchronization of data. Synchronization between GPUs, between CPUs, between GPUs and CPUs and TPUs, between racks, between stacks, between eventually data centers, right? The data must be coherent at all times.
This notion of synchronization just doesn't stay in the data center, but goes into any kind of compute and AI, for example, in robotics, in humanoid robots, for example, in cars, where, again, everything must be synchronized. And you can achieve synchronization through some software, but that software sits on top of hardware. And that hardware is the oscillator, the clock, a combination of the oscillator and the clock. And the more instantiations of that, the more synchronization needed.
Knowing that there's more synchronization needed, what's sort of the company's strategy to continue to increase its position in the data center market? You've talked about having resonators, oscillators, clocks. Is it a bundling strategy? Is it getting closer to your biggest customers? Just sort of thoughts on how do you continue to enhance your position in the data center?
Yeah, so the beauty of semiconductors is you have to deliver the functionality, and then the rest comes with it. So the functionality that we're looking for is reducing latency. One of the ways you reduce latency is through the, you know, we've gone from picoseconds to femtoseconds, reducing the jitter, increasing the stability, making the phase noise better, making the size smaller at the same time while you're doing that so that there can be more of these populated in any use case. And then the strategy as far as the go-to-market customers is clearly connect with the semiconductor companies, clearly connect with GPU, TPU, the switch companies, people who do like Broadcom or Marvell. But also and follow that because they make reference designs, so they may use it and then work with the people in Taiwan who are putting that together or the CMs. So follow that.
The second is talk directly to the hyperscalers because some of them are doing the same thing as some of the semiconductor guys, but they're also doing their own take on it. They have their own architectures. So work with them. We do that. And finally, the people who supply the networking, the Credos and optical modules and people who do the retiming, like an Astera, those people are kind of important to round it all out, a Molex, a Coherent, you know, delivering coherency. So all aspects of that.
You talked about trying to reduce latency, picoseconds going to femtoseconds. To the extent SiTime can supply the entire timing tree, does that enable you at a system level to help customers, you know, deliver better value, or is it really, you know, you don't necessarily need to have everything under one roof to achieve some of those goals?
Yeah, I think actually our strategy is to deliver the full in the data center, in the communications, enterprise business in particular, what we call CED, to deliver the full suite of solutions. Because previous to us, a customer had to go to the quartz crystal oscillator companies and get the oscillators. And then they had to go to the clocking companies such as a TI, a Skyworks, a Renesas, Microchip, and get the clocks. Because SiTime has the clocks as well. We've just started doing it, but we do have it. We can bring synergistic use case, sometimes virtually. In other words, the independent chips, they just work together better with each other. Sometimes physically, we put oscillators inside clocks to make a single solution possible. So yeah, it's very much based on that.
Are there any areas within the data center where you feel the company is underpenetrated, whether it's major customers or major applications within timing?
Yeah, I mean, you know, the good news is we have done very well. The bad news is we need to do more, or maybe that's good news. The point is that there are many, many semiconductor players in this space, and we haven't penetrated them to the same level that we should. We're doing that. We've created a semiconductor sales group inside the company. It's relatively small. It'll grow through the year. So that's going to be important. But the other is we need to call more exclusively on the hyperscalers as well. And that's very important. And again, we've done that a little bit, but we need to do more of it. We also need to do a little bit more outside of the U.S., which so far we have sort of stayed away.
And also there are companies that are being formed in the AI space all the time. And you know, we need to pay attention to that. So it's an embarrassment of riches, as it were, but there's a lot of work to do.
Your company has 800G optical modules. It has been a pretty strong area of growth for SiTime. Can you talk about your position in the 800G market? And I don't think you were as penetrated at 400G . And so maybe just talk about what changed as you went from 400- 800 that enabled you to enhance your share position in those modules?
Yeah, so we typically don't give out what % of the market we are, but suffice to say at the 400G , we were fairly low, and that was at the height of COVID, and there were so many other opportunities that we sort of didn't quite spot it the same way. We did get some design wins. At the 800G , we clearly came into our own, and we had the performance that was ready to match it, so that was a good spot. At 1.6 terabit in particular, we are in a very good shape. We have said that before, primarily because the trajectory of adoption of 1.6 has been a surprise to everybody in the industry. The volume, the speed at which it's been adopted, and the use case in terms of units being used.
SiTime is one of the very, very few, perhaps the only one that can fully deliver all the capacity that's needed, can deliver the performance that's needed. That puts us in a very good spot in that. It's very early stages. The interesting thing about this market, I saw a third-party report which showed even out in 2028, 2029 that there were still 200G and 400G shipping. Everything is growing. As you know, data centers are being built outside the United States as well. There's this notion of sovereign data centers. People, countries, you know, middle-sized economies are also adding their own data centers for their own sovereignty, I suppose.
Wanted to sort of shift from wire to wireless infrastructure. It seems like wireless infrastructure from the base stations to the radio nodes, you need a lot of synchronization. Seems like it's a content-rich opportunity for SiTime. Maybe start with where SiTime plays and just sort of thoughts about the opportunity to grow that business as we get to 5G Advanced and possibly 6G in the coming years.
Yeah, so it elicits a chuckle whenever I talk about it because people are so focused on the AI that talking about communications and Ericsson and Nokia and such, like people go, "Really?" But it's happening, as you said. Not everywhere is 5G extant. Not everybody has 5G. And 6G is definitely on the horizon. We specifically play in the radio access network or node or RAN, where the cellular signal emits and the digital unit, the DU. In both of them, performance, size, environmental robustness, resilience is very critical. And we have done well with the customers. There's also this notion of fixed wireless access, FWA, which is happening in countries like India, where there's a large portion left of the population that doesn't have anything.
There are some government mandates that require that they get access so that they can be connected to the economy in a good way. I think there are some great opportunities there. We see that. We never ignored that because part of the beauty of SiTime is out of the 400 applications, maybe five, seven, 10% are always growing more than 25, 30%. And the rest are growing at 10, 12, 15, 20%. By definition, we are always paying attention to the bright shining stars, but we are also feeding the rest of them because we never know when they're going to explode. That's what happened with AI. That's what happened with ADAS. That's what happened with, well, way back when and with crypto mining when that was a thing in 2016. Yeah.
Shifting over to the auto industrial A&D part of the business, can you talk about just some of the biggest or fastest growing opportunities in that end market?
Yeah, I mean, people are surprised when we talk about automotive as a growing opportunity, but it is. We expect it to be a $100 million business in fairly near term. The reason we grow is because, again, we're focused on the highly differentiated, the high-growth businesses like the automated driving ADAS business, or in fact, GPS-aware, or in fact, some kind of high-end infotainment inside the system. So ADAS, of course, you have Level 2, Level 3, Level 4. Level 4 is the pinnacle where we probably have north of $30 per car. Level 2 is the starter kit at pick a number, $5, $6, and everything else in between. So we have customers, particularly Chinese customers, other than the well-known ones in the United States. We're starting as well with the European guys as they become more and more clear on ADAS.
The GPS one is interesting because it spans between military, aerospace, defense, industrial, and automotive, and I guess I suppose in humanoid robots. So it turns out that we used to accept GPS as a highly reliable, invariable way to tell where you were, but also as a source of atomic time, free atomic time. Well, it turns out with the war the Russians started that there is no way to depend upon GPS anymore. There's GPS being spoofed, GPS being corrupted, GPS being not available. So our chips have the timing set up so that when GPS is corrupted in some way or denied in some way, they can hold the time for eight hours, 12 hours, 24 hours to keep on going with that atomic time-based level. That's a huge thing.
And we see ourselves in everything from drones to undersea exploration to, of course, military aerospace, the cars we talked about. And you know, you can think of, you know, a 135-pound humanoid robot that you don't want going berserk in the movies like we've seen. So yeah, timing, very important for all that.
Your timing devices in a GPS-hindered or denied environment can keep local time until the system reverts back or reestablishes connection with the GPS satellite.
And one day when we have atomic time, true atomic time, which we don't, we're far from it. But if we have atomic time and, you know, a 2.5-millimeter by 3-millimeter chip, we would not need any of that. We just have true atomic time.
Anything else you want to highlight in the broad-based or the industrial A&D segment before moving?
Yeah, I mean, that's our biggest segment by use case. Out of the 400, probably 300 are in that. It's the most exciting. It's high margin, actually. It's slow growth, except when it isn't. When something, you know, spikes, it really spikes. And it's sort of a mini SiTime in a sense that SiTime is a broad-based company. And we don't have one, you know, you'd ask somebody, you know, quick, tell me what is TI's number one use case. People go, we don't know. Because TI is such a broad player or an ADI is such a broad player. SiTime is in that same category, very small, with no TI. But we are playing on that same model.
Great. Moving to mobile IoT consumer. Last March, you announced that you had secured two sockets on the internal modem for your largest customer. Do you expect to remain sole-sourced on these timing solutions going forward? Do you see competitors, you know, trying to develop solutions for that modem?
Surely. I think that it's a big forcing function among competitors when they see a very clear, you know, revenue stream. It would certainly sort of focus their minds to coming into this market. So we certainly expect it. We know there are certain, you know, large Japanese companies that have been for years talking about doing this. There's at least one startup in Canada that recently secured funding that's quite capable. There may be others. At the end of the day, we have taken kind of an approach of predictive revenue along with high-growth revenue, along with high-margin revenue. And the predictive strategy tells us that we're fairly, we think we're pretty good for the coming couple of years. And just out of an abundance of caution, we don't know what happens after that. But let's not get away from the fact that timing is hard to do.
MEMS-based timing is hard to do, and the package of performance and performance under tough environmental solutions is hard to do. Particularly if you have to ship 500 million units or 1 billion units, it's hard to do. It doesn't mean we won't have competition. I always take competition very seriously, but I think like any winning team, the winning team decides its strategy and always checks to see if its strategy is the right strategy, but then executes to that strategy, so that's what we do. We have an executing to a strategy we have while keeping an eye open for competition.
I guess it kind of raises the question, you know, to the extent you have a socket, whether it's mobile, whether it's CED, and you lose that socket, do you think you would lose it back to quartz? Or do you think it really takes a MEMS-based or a silicon-based timing solution from a competitor? I mean, do the competitors to displace you really need to develop MEMS or silicon-based timing?
It could be both. I'm not too caught up on the technology. The Japanese have been doing quartz-based solutions for 70 years. Every time I turn around, somebody tells me, "Oh, they're tapped out." Every time they deliver another surprise. That's because they're very good at it. They're very good. They understand what they're doing very well. Similarly, in MEMS, there are people who can do it. It's a question of focus. It's a question of drive. The question is, SiTime has 500 people with 350 engineers that are focused on one thing and one thing only, unique in this timing. That's all we do. Does anybody else have that capacity, that financial ability, or indeed the drive to declare that? It's certainly hard to do, but you know, it's all doable.
Okay. Within the consumer IoT, you highlighted the Symphonic clock generator for some mobile applications. Maybe spend a minute. What does that product do? Where are you seeing deployments of Symphonic?
Yeah, it's basically a clock. It's our own clock and it's a clock with a difference. It's a clock which has a resonator, or actually an oscillator built into it. That's what makes it unique and of course, its performance characteristics are amazingly good. We have a lot of headroom built into it in terms of phase noise and size and power. Right now, a modem use case is the main one, but we see a bigger use case coming in Internet of Things, in robotics, which need to be connected, sometimes to cellular, sometimes to something else, so we kind of think it's got a 10-15-year life going forward. We put a lot of money into it, but I think it's going to do very well for us.
Excellent. And maybe a similar question. You've also introduced your Titan resonator platform. I think this is across multiple markets. You know, talk about the resonator opportunity. Where do you play within the resonator portion of the market?
Right. So the resonator opportunity is one more product that exists in functionality, but doesn't exist in the form factor and qualitative ways that we bring to it. So it's a $4 billion market. As always, we're going for a small slice of it. $400 million is all we ask. We bring a dramatically smaller size, like one-fifth, one-seventh, one-tenth. We bring dramatically lower power, dramatically greater robustness and supply chain. And we can sell it directly to people who make PMIC or make Wi-Fi or make Bluetooth Low Energy and sit right next to it. That's one use case. The other, a higher bar, is to go to chip companies that do this and get it packaged inside their chip.
It's a higher bar because now you're asking somebody who does an MCU to depend upon your product the same way they depend on theirs for quality, reliability, cost, availability. But once you get over those hurdles, and they're not small, but once you get over that, the benefits of integrating timing into an MCU, into a Wi-Fi, into a BLE are immense. So we see it's not immediate revenue. It won't show up much in 2026. It'll show up a little bit in 2027. But four years out, this could be a $100-$150-$200 million business at greater than average, corporate average gross margins. And the functionality of this thing has taken us 18 years to develop. So.
What's the response been from some of the silicon partners you've approached?
Very good. Very good. They're almost like, "Too good to be true. Let me check this out.
Okay.
Right? How can it be all these great things you say, and yeah, too good to be true, but let's check it out. We have design wins. We have launched some. More coming.
Excellent. I wanted to ask just on the M&A front, you've been vocal about your interest in potential M&A. I know you're limited in what you can say, but can you, at a high level, talk about what types of assets would be of interest to the company? And then any sort of criteria? It's got to be accretive. It's got to be additive to growth. Probably has to be strategic. But just sort of thoughts, how you might approach a future M&A?
Yeah, it's very simple. As you point out, we have a rather large amount of cash, $800 million. No debt. That's a good place to be. We are looking to scale up in timing. We're doing one thing, right? Just timing. We've got to scale up. I like a billion in revenue. I think that's a good number to get to be at the right scale to play in a lot of areas. So we want to scale. We want to scale in timing. We want to scale while still keeping growth at 25%, 30%. We want to scale while keeping our 60%+ gross margins. And we want to scale while keeping 25%-35% net profit. So it's kind of a tall order, but that's how we see it. I don't want to compromise our financial. I don't want to compromise our strategic direction.
We just want to be a great timing company.
Great. Well, I think we're getting close to the end of the session. I guess I just wanted to, you know, ask you a last question. Do you think, you know, we missed anything? Is there anything you'd like to share with the audience?
No, I think just for scale, I've been in the Valley since 1984, so it's been 40+ years. This is the second time I've taken a company public. The first time when I started in 1999, in 2005, in 2006. What I learned was that one needs to go after existing large markets. So timing is an existing, when I come to it in 2007, is a $5 billion market. Today, it's an $11 billion market. A decade from now, it's going to be a $20 billion market. There isn't any other company that is solely dedicated to the task of delivering exceptional timing solutions on all timing solutions. That's what SiTime does. We've been doing this with me and the companies for about 18 years.
It's in an amazing position to be at that, having formed such a company and done it while maintaining the financial template that we have. So I'm very happy, but I think the best years are still to come.
Excellent. All right. Well, with that, Rajesh, we'll wrap. Thank you for joining us at the Needham Conference.