Synaptics Incorporated ($SYNA)

All right. Good morning, and welcome to JPMorgan's 54th Annual Technology, Media and Communications Conference. My name is Harlan Sur. I'm the U.S. semiconductor and semiconductor capital equipment analyst. Also with me on stage is our SMID-cap Semiconductor Analyst, Mayur Ramdami. Very pleased to have the team from Synaptics here with us today, Rahul Patel, President and Chief Executive Officer; Munjal Shah, Vice President of Investor Relations. Rahul is going to kick us off with a brief overview of Synaptics. It's been a busy earnings season. So I've also asked the team to just give us a quick summary of the March quarter, June quarter outlook, and then we'll go ahead and kick off the Q&A. Gentlemen, thank you for joining us today. Rahul, let me turn it over to you.

Thank you very much. Good morning, everybody. So a little bit about Synaptics, and then we'll touch on the March quarter. Synaptics is a company that's been -- is a semiconductor company that's been in existence for 40 years, and it's a very strong, formidable legacy of the company. I think when I came into the company, I -- first thing I realized is the VLSI textbook from where I learned VLSI design was -- and which is the textbook that's used across the world is authored by one of the founders of Synaptics. And so that's how the legacy of the company is. The company came into existence first with the ClickPad in the PCs. And even now, we do a lot in the PCs with our TouchPad products. Then came the revolution about the iPhone and the touchscreen, and Synaptics was at the forefront in that arena. And then came the time where Synaptics pivoted to IoT, which is where we are right now. And the company is now focused in areas of Edge AI and Physical AI, along with everything that they've been doing. And that means the company is largely built its product -- on products on 3 pillars. And the 3 pillars are Processing, Sensing, and Connecting. And that's what Synaptics is all about. We build semiconductor solutions targeting processors, wireless connectivity, Wi-Fi, Bluetooth, Thread capabilities, processors ranging from microprocessors to microcontrollers. Our touch-sensing capabilities, our sensing capabilities also get supplemented with fingerprint as well as video interface capabilities. And so these are the core IP capabilities in the company, along with many other things. And so this becomes the basis for us becoming a lot more focused in Physical AI and Edge AI on a forward-looking basis. In the March quarter, we had phenomenal results. That was, I believe, a sixth quarter or seventh quarter, if I get this right? That demonstrated year-over-year growth. And we grew substantially, 30-some percent in our IoT business. Our EPS grew more than 20% year-over-year, while our revenue was in double-digits growth as well. And we also guided that this year -- our fiscal year ends in June would yield greater than 40% growth in our IoT revenues year-over-year. And so that was our quarter. And in this quarter, we also talked about a big unveiling basically in our Physical AI play. We have been talking about a lot of the design wins and capabilities that we have demonstrated in the marketplace in Edge AI. But in Physical AI, we came out and said in the prior quarter, we had indicated we had a design in humanoids. And 90 days later, we were at 35 various OEMs engaged on Physical AI, especially in the field of robotics that range from a small robot with no display to a full humanoid. And so that was what we have reported in the March quarter.

No, that's great. And as you mentioned, it's been great to see the transformation of the business, right? Over the many years from sensing to connectivity to compute and then now being able to go after the new opportunities within Edge AI and Physical AI. And to your point, we can start off with the IoT business, which you just articulated the team expects to grow 40% year-over-year in fiscal '26. You've also outlined a path to sustained above-market growth across your core IoT and Edge AI portfolio, supported by the multiple growth drivers and product ramps. Could you just break out those drivers by time frame? What's near term versus what's intermediary term, like specifically, how should we think about the expected contribution from some of your newer product categories such as your Astra processor, Wi-Fi 7 connectivity family of products, integrated MCUs, wireless connectivity, some of your semi-custom products and so on?

Yes. On a go-forward basis, our focus has been on Edge AI and Physical AI. And so the core elements are our Astra processor family of products and wireless connectivity, and our touch-sensing and video-bridge capabilities in this area. Astra is coming to life about now, and it's going to be the growth driver for our IoT segment in fiscal '27 and, more likely in calendar '27. And so if you look at what we have done over there, Astra is a category that is in the class of microprocessors and microcontrollers. However, with a huge difference versus what you see from the peer sets that have been in the marketplace for a very long time. Astra is AI-Native processors. They are -- we are not building any processor without having the ability to do inference at the very far end of the edge. So it's a microprocessor or a microcontroller from Synaptics under the Astra family. It comes with in-built neural processing capability, along with other processing engines that we have. And so within the Astra processor, last year, we in the calendar quarter 4, we started sampling our first microprocessor that we had also integrated Google's Coral NPU. NPU that we co-developed with Google Research, along with our processing engines. And so that's gone into production just about now, and it's going to see its production ramp towards the end of this calendar year and throughout 2027. We also sampled recently a Wi-Fi 7, Bluetooth 6 MCU with NPU in a monolithic die. If you look at our peer set in the field of microprocessors and MCUs, likes of NXP, STMicro, Infineon, Microchip, Renesas, they do not have the wireless capabilities that we have. And our implementation in a monolithic die on the microcontroller that not only has a microcontroller, but also has a Native NPU for inference at the far, far end of the edge, along with wireless connectivity in a single monolithic die, is very exciting for our customers because now they have access to a power envelope that otherwise would not be easily attained, doing multi-chip implementation, a form-factor envelope that otherwise would not be attained if they have to again go multi-chip implementation. The BOM envelope would have been also a lot more efficient with what we have implemented. And so that is -- that product is in sample stage as of now, and we anticipate that going into production in calendar '27 as well. And then the semi-custom microcontroller with Coral NPU and ISP capabilities that is targeted in the field of wearables and has been co-developed with a big customer that believes in building out software stacks and scaling software stacks on top of silicon. Their business is not to sell silicon, but their business is ultimately to drive their software stack into billions of platforms. And that is the opportunity that is going to ramp in '27, second half of calendar '27 with the semi-custom partner of ours. And so that -- the word is semi-custom. We've developed the architecture in collaboration with this big customer so that what we compile, we implement transformers, all is done in context of what their software stack needs may be. In a power envelope, in a BOM envelope that makes sense. And so again, the word is semi-custom and what that means is although it's co-developed in many situations, we have the ability to take it to the larger marketplace. And so that also presents us another vector of revenue realization. So all these aspects of Astra come into play in '27 -- calendar '27, and that's very exciting for us. And then towards the end of this year, we'll be having our first implementation of Wi-Fi 8 in silicon. And now you can think of likes of Broadcom and Qualcomm having Wi-Fi 8 for things like access points and phones, but when our peer set are barely having Wi-Fi 6. I don't know how they get to Wi-Fi 8 this year. And so when I say I peer-set the likes of MCU and microprocessor players. And so that presents a nice tailwind to our business in calendar '27 and '28 as well. And so very excited about these aspects. And something that I have not included in our revenue plans, but I've talked about it as well, and I've also suggested not included in our revenue plans is our activity that's going on in Physical AI with -- especially in robotics and humanoids. That has definitely surprised us. And I think it's also validating the technologies that we have in our sensing portfolio that is lending extremely well in the field of Physical AI. And so all of these, again, build that confidence that Edge AI and Physical AI are going to be our forward-looking growth vectors for the company.

Yes. And we'll talk a little bit more about some of the humanoid-robot programs in your pipeline. But you actually gave us several good examples of how the team goes to market, especially with your IoT business and Edge AI business, right? And you've got this really great portfolio of connectivity, touch, analog, mixed-signal, compute, as you pointed out, you -- some of these products are integrated, right? So you're selling both compute and connectivity at the same time. Some of them are not. The team has discussed increasing content per engagement by delivering more complete solutions, leveraging things like reference designs, the broader attach. But can you just elaborate on how this strategy is progressing? And where do you see the biggest near- to mid-term opportunities?

Yes. I think -- so before I came into the company, we were 3 groups that were largely siloed, and they're operating as a processor team, as a wireless team, and as a sensing team. And so earlier this year, we consolidated processors and connectivity is one team from an engineering execution point of view as well as a go-to-market point of view. That team is going to sell processors and wireless together. And so what that means is we're going to sell solutions. We're going to build solutions. We're going to build software platforms that ultimately help reduce the cost of engineering at the customers that help scale a lot of software capabilities that would be at a higher level of abstractions just beyond our SDKs, and also leverage our open-source, open developer platform strategy across the marketplace. And so this is where we kind of are very differentiating versus our peer set. Again, the same big names that we talked about or I have mentioned on the process side. They have a very walled-garden approach on the software side. It's their environment, it's their SDK. And if you want something different, if a developer wants to come in and play, you have to sign license agreements, you have to get through the scrutiny of are you going to be able to consume a lot of resources or not and what is going to be meaning for support dimension for the company. All of that for us, we are open developer platform, and we let the developer community build on our platforms and the developer community in turn, supports each other basically, so that's our strategy that's very differentiating and it's going to help us go-to-market. The traditional way to go-to-market for the microprocessors and microcontrollers is go through distribution. We believe the opportunity that we have with AI-Native capabilities in our end-products, and the developer strategy gives us a jump start very inexpensively related to developing a distribution strategy. Ultimately, when we are a lot more broader in our SKU map and capabilities and AI-Native becomes a lot more prominent in the marketplace, we will definitely go on the distribution vector, but at this point, I think the strategy that we have is yielding us really good design wins and design pipeline is building up very nicely as well as a result.

Let's focus on the humanoid-robot program at one of your major customers that you articulated in your prepared commentary. It is a great example of Physical AI, where you're supplying Touch Controller/Interface Solutions with content in the range of, call it, a few tens of dollars per unit. That customer is now on its third-generation platform and has articulated ambitions to scale to meaningful volumes. Against that backdrop, how should we think about the opportunity to expand content in future platforms as you pursue additional sockets and additional customers, right? Such as processing, wireless connectivity, where your total content could potentially exceed like over $100 per unit?

Yes. I think it's not outside the realm of possibility, I would say that. Having said that, I think -- the one that we have publicly talked about and mentioned, and this big customer in North America has also publicly announced, they will be doing pilots at the end of this year. It's a humanoid and not the first generation of humanoids, basically that they're doing it. And so it's very well demonstrated, talked about. Our content in that platform is largely a few tens of touch controllers as well as a video-bridge implementation. And I'll talk a little bit about both. The touch controllers are in the palm of the humanoid, and it's in the order of 10 to 20 in each palm, there's 2. So you can see where it goes. And the video bridge is a high-bandwidth bus interface from the main SoC to various subsystems, including the display. And so that itself also is fairly rich in silicon content. And so you add all these things together, and you get to a few tens of dollars. And I think we're not putting a number, largely because I personally believe this market is in the early phase. And the customer -- this big customer said they pilot at the end of '26, and they anticipate going into production end of '27. Some reports have said the first year will be 1 million humanoids, and we'll see what that does. But coming back to the numbers, right? If you do a few tens of dollars into this million units in '28 -- calendar '28, right? I think you get a sense of where this is going with one design. And having 35-plus designs basically now in combination of sensing capabilities, our interface capabilities, our Astra product capabilities now and wireless connectivity. I'll talk about Astra and wireless connectivity. Every subsystem in a robotic platform has its own MCU and microprocessor -- and/or microprocessor. And every subsystem has the need for machine learning as well as inference locally versus having to send it to the main CPU or GPU. And its reason for not loading the main CPU, GPU and also latency of inference basically at the edges of the humanoid of the platform. And so that itself is another $10 to $20 portion in the extra processor. Every time you use an extra processor, it's $10 to $20 of content. And then robots or cobots or industrial platforms will need to communicate as they mobilize across platforms or even in your homes, if you have a cobot, they will need to remain connected. It would be peer-to-peer, a humanoid-to-humanoid, or humanoid-to-the-Internet, or to the data center. And that requires a certain level of wireless connectivity, a latency equation that does not deprive of the experience that the end humanoid application has to deliver. And so long story short, there's a wireless capability over there as well. Various industry reports come out and talk about this whole market in context of trillions of dollars. And so there's a lot of forecast out there, it's not easy to say this is where the plane is going to land. And so at this point, I think we are not adding a whole lot in our financial models, just keeping our heads down and remaining engaged with these customers. Our touch controllers go from the palm to other locations of the humanoid, including the feet, because that's where some of the sensing capabilities need to reside. Our Astra processors can go from the hand to multiple other subfunctional sections of the platform, and so is our wireless connectivity, I think, largely for data communication.

So I just wanted to clarify one thing. I mean, Rahul mentioned 35 -- we have 35 engagements, and we have -- we talked about 3 additional designs this past quarter.

Yes, we started -- thank you, Munjal. We started shipping silicon to three additional other than the one that I mentioned is a large customer.

I see. So let's focus on your compute family of products. This is the newer -- I would say, as we followed Synaptics over the years, this is a new facet of technology, addition, and product addition to the portfolio. You gave us an example of the potential opportunities with the Astra processor. But the Astra product line, correct me if I'm wrong, includes processors and MCUs, but I think your SR80, SR100 are also MCU-focused SKUs that are within the Astra family of products. One of the key differentiators, as you mentioned, is all of these, whether it's processor, like the 2600 series or the microcontrollers like SR80, SR100, they all come with the ability to process AI and machine learning workloads, right? You've got what we call neural processing engines, NPUs, right? And for your flagship 2600, 2160 platform, your NPU is called Torq, I believe that's correct. And which leverages open-source technology from Google Research and integrates a lot of hardware and software innovations and accelerations. Outside of some of the humanoid opportunities, if I just think about all of the Edge AI, Physical AI opportunities in front of you, like where has the team with Astra been able to see the strongest market adoption for which applications, which products? And is it skewed more towards processors? Or is it skewed more towards MCUs?

It's an excellent question. Also a loaded question. I'll try to kind of operate at a higher level of abstraction in my response. So I don't take up too much time on this topic. But I think you should think of our processing engines as engines that are very differentiated versus what's available in the marketplace. Now I'll try to use examples. SL is our processor class-products. SR is our microcontroller-class products, all are AI-Native. Coral NPU is Google's NPU that's open source. However, Torq is our architecture that encompasses multi-processing engines along with Coral NPU. These engines are general-purpose ARM CPUs. These engines are application-specific audio-processing capabilities, video-processing capabilities, ISPs, all in a single fabric that we call this Torq architecture. And the compilers that sit on top of these platforms are co-developed with Google Research. It's an MLIR compiler. And what that does is that it is intelligence in the compiler that says, if this is what the workload looks like at compilation, this is how the workload needs to be distributed in terms of where it goes for processing, given the intelligence of the design and the pipeline is available in the compiler. So that is how this whole architecture works. Having said that, I'll share with you a few examples and then respond to the larger question of where we're seeing traction. So the first, I'll share with you 3 examples. The first example is one of us being in our family rooms basically with our television screen. Our television screen before we turn on, gets to know that I'm a Boston Celtics fan, right? And is aware that there's a Celtics game on a particular channel, ESPN, it may be NBC or whatnot. And you don't have to worry about what channel you need to scroll to. If it is going to be aware, contextually aware of your presence, human presence and effectively aware of your preferences based on your prior viewing habits, it will take you -- the first set of eyeballs will land basically on that particular channel of choice. Now you can obviously mobilize from that position, but that is inference and being human-aware and contextually aware. If in the family room, before you turn on the television, if the television knows that here is a family with kids, it auto turns on parental control. That's being contextually aware. That's been very much of a use case that we would care for, right? I'm just giving you a couple of examples. But you can imagine where this goes. And this is through the AI-Native microcontroller in the SR-Series of products that you mentioned. This is how AI-Native, contextually aware, human presence works in consumer applications. And I just use television as one place, but you can think of a lot of things, your doorbell, to your thermostat, to how you mobilize in the home, your security, all of that is going to see this level of inference capability that will be supported by our Astra class of products. All of these platforms would need to be consistently communicating with various places, and that requires a certain level of wireless connectivity that's integrated in this platform. And so that's where we go. And this is, again, a consumer-class application. We recently in the last quarterly earnings call, just to highlight, Astra is also getting into medical devices. And so we highlighted a medical device that does a scan for the well-being of the mother and the child during the phase of pregnancy. And this scan can be done at home versus being done in a hospital or a clinic. And so this ultimately has certain AI-Native capabilities that are leveraged for the experience that comes, the level of accuracy that comes to the forefront. So again, this is medical. The third area that I would highlight is industrial. We also touched on an application, and I publicly talked about -- I talked about it at the conference call, is fleet management. Through our embedded ISP capabilities and being AI-Native, the ability to manage a fleet of vehicles basically in an industrial application is also at the forefront of what we are engaged in basically in design wins and stuff like that. And so in all of these applications, making decisions natively, after being contextually aware of what all can happen if what we are sensing is happening is what our products are capable of bringing to light. And so going back to the larger question, this is obviously the first few innings of our engagement in the Astra class of products in the marketplace. And so like with every other market that you see in the processor world, the first set of designs that are going to turn into revenue are going to be consumer-class products. Industrials are slow to ramp, but longer to hold on the revenue front. Consumers are fast to ramp and fast to turn basically. And so that's how we see our business profiling on a forward-looking basis.

That's perfect. You talked about in your prepared commentary semi-custom projects. You secured one with a fairly large customer. You've indicated that program remains on track to begin sampling in the fall, with initial production ramping in first half '27, larger volume in second half '27. So I guess first question is how significant could this program be for the company? But as you expand into potentially more semi-custom solutions, what criteria or metrics does the team use to figure out do we pursue, do we not pursue, right? And is the team currently engaged with other customers on similar semi-custom opportunities?

It is very important for us and given my experience, very important, when you are engaging in a new market, you want to be -- especially in the semiconductor marketplace, the investment equations are fairly substantial, right? In this age, it's even more than what it was maybe 10 years back, right? And so having some customer skin in the game is -- upfront is very valuable. And that is what semi-custom presents. What semi-custom also presents is a design win that you know for sure is going to ramp into production. And so not only that skin in the game to support the project, but also that "bit of a shorter bet" on going to production with that silicon a lot sooner than you would have to otherwise if you don't have a semi-custom partner to go with. And then the word semi in semi-custom also is that we profile what we are going to do with this partner to be able to take that design to the broader marketplace. And so those are the core principles of engaging in semi-custom, skin in the game, time-to-market, a large market opportunity with that one customer as well as the ability to scale the product into the broader marketplace. And if there is anything like you cannot go wrong in semiconductor business, then it is a semi-custom portion, right? I think that's why it makes a lot of sense. Your second part to the question was others, basically, yes, there's -- on the core competencies of sensing, processing and wireless connectivity in all leading capacities, we draw a tremendous amount of attention from large players wanting us to do things that are very differentiating, especially when you are building contextually aware AI-Native platforms and where standard products are not available, right? We see that in data center. You've seen players who are doing ASICs do extremely well with semi-custom ASICs in data centers. And I think the same is going to play out at the far end of the edge, and we are at the forefront of that dimension growing to being substantial. Now in going down that path, we do have opportunities that we are engaged "evaluating" not in context as much of customers having skin in the game and the ramps and all that, but also given the finite amount of resources, can we really make this into a semi-custom versus a custom -- and something that will ultimately also further our road map across broader marketplace and so those are key criteria in how we kind of go about engaging in semi-custom opportunities.

We have just under a minute here, but if you could just briefly touch on gross margins. You have several new product ramps expected to contribute to revenue over the next 12 to 18 months. So how should we think about the trajectory of gross margins as these products ramp?

Well, I think gross margin, is going to be very important dimension that is going to come to play in all semiconductor business, not just Synaptics. And we have been maintaining 53.5% gross margin, plus or minus percent. That's what we guide despite the input costs having gone up in the last few quarters. Based on my anticipation, I don't see relief on input costs in the coming quarters. And it's widely broadcasted and talked about. And so while Astra is going to be gross margin accretive and it's going to ramp in calendar '27, I remain watchful on how this will play out on a forward-looking basis given the input cost equations and how they are trending at this point, right? And so this is a story for, I think, the entire semiconductor industry. Astra by design is going to be our growth engine for the future -- the near-term future of the company. It is going to be expected to be gross margin accretive, but there's going to be headwinds with the input cost that I can't ignore.

Perfect. Rahul, Munjal, I appreciate your participation today. Look forward to monitoring the execution of the team as the year unfolds. Thank you very much.

Thank you very much.