Ambarella, Inc. (AMBA) Earnings Call Transcript & Summary

Great. Good afternoon, everyone. Thank you for joining us. I'm Vivek Arya. I cover semiconductor semicap equipment at BofA Securities. I'm really delighted to have the team from Ambarella join us this afternoon, Casey Eichler, the CFO; and Louis Gerhardy, the Head of IR and Corporate Development. And what we will do is perhaps go through a quick word of introduction from Casey on Ambarella to folks on the call who might not be familiar with the company, and then we will get into a fireside format. So with that, a warm welcome to Casey and Louis. And maybe let me turn it over to you, Casey.

All right. I think several people are probably familiar with the company, but I'll just give a few comments. I'll let Louis [Audio Gap] about anything I missed and we can kind of go from there. So the company was founded in 2004. It was a video-based processing company. You might know it from GoPro or DJI or a lot of other markets, which is when the company went public on in 2012. Following that, we realized that we needed to create a much bigger, broader company that could really access a lot of other markets. And so Les Kohn and Fermi, who are the 2 founders, went off and started working on CV technology to be able to go off into not some of the markets that we were in, but also in some of the new markets, most notably to most people in the automotive space. And today, the company is focused on both professional and home security. It's focused on automotive and a series of other markets where computer vision technology can really be applied in a very new and distinct way. The company has been cash flow positive, has a strong cash position, has invested about $0.5 billion in this technology today and has also returned about -- or close to $180 million to shareholders through the process of this. We've had several different announcements. We just had our announcements last week, but have several different announcements that we can cover and talk about, and we're happy and pleased to be here and have a great partnership with Bank of America. Louis, if there's anything else you'd like to add on an intro, that would be terrific, and then we could just go to Q&A.

Yes. Thanks, Vivek and Casey. I'd just add from a 40,000-foot level, what's happening in Ambarella is we're moving from a market driven by the human population, in other words, collecting information from the environment and disclaim it for human eyes to one where we're collecting the same information, but now we're processing and analyzing it for machines to sense the world and then either partially or fully autonomously make decisions, that cuts across all of our markets, not just auto. And to do this, we leveraged video processor, where we have an expertise for collecting information as efficiently and accurately as possible, and we add a deep neural net AI processor to it, and it's all integrated together into a single chip, either a 10-nanometer or 5-nanometer for our newest products. And we're able to do this with -- in a very differentiated and proprietary way. Because of our people, we have our unfair share of engineers' experience with video hardware and software, our technology and our IP and it's our approach as well, which we'll probably touch on later. And what -- why are we winning in this market, why is our CV business beginning to take off, it's because of 3 reasons: one, efficiency, which we measure in performance per dollar or performance per watt of power consumption; second reason would be, we're an open platform, we provide these computer vision, AI hardware resources in the chip and customers will develop the software that runs on them; and then the third reason is, is our platform, which scales from CD28 to CD5 at the high end, at least for now. That platform offers tremendous flexibility and delivers to investors a strong return on investment because one chip can be used in a wide variety of different verticals. So with that, I'll pause and Vivek pass it off to you.

Excellent. Maybe just kind of part B of that. When investors think about low power computing, they always think of smartphone companies, right, Qualcomms of the world. When they think of high-power computing, right, or kind of mid-power to high-power computing and AI, they think about the NVIDIAs of the world or Intel, right, of the world. What kind of niche has Ambarella established for itself? And how defendable is that, right? Because for -- what does it take to be successful in the niche that you have? And what kind of competitive landscape do you see coming from both kind of companies who are very established and large on the low-power side of the market and the high-power side of the market?

Yes. I would say, when you think about the smartphone market, the video processing technology has been developed with completely different use cases in mind. For example, when you use your phone, you might take 10 or 20 second video clips. If you were to leave your phone on for 30 minutes, you probably run out of power. It would get very hot. It might even freeze up on you. The markets that we've served with our video processing technology that, again, we're leveraging this core that I'm talking about by adding a deep neural net processor and integrating it together. The markets we've traditionally served have been applications like enterprise-class or public infrastructure security [Audio Gap] that operate 24/7 non-stop outdoor environment, sun, rain, environment beating on them without missing a beat, all in an externally low power envelope. And so it's that type of camera market, that type of operation that we've targeted. And when you think about some of the applications that will need computer vision technology, whether it's a machine vision application that operates 6 days a week, 24 hours a day or an automobile where safety is at mind, this type of video performance that we focused on since day 1 is absolutely critical. And how do we do it, to answer your question, whether we need to do it, well, first of all, it's the people. And we have our unshared fair engineers that have been passionate and focused on video technology. In fact, most of the members of executive team have worked together since day 1 at Ambarella even at predecessor companies like CQ Microsystems. It's also our approach to the market. And so for example, when we developed the first CV chips, rather than just begin to lay out transistors with computer vision in mind, Ambarella acquired a company in Italy called VisLab that was excelled at computer vision software stacks. They were not and are not a chip company. And so Ambarella took 3 years to work with this lab and understand what is the most efficient way to execute computer vision algorithms. And it's this very thoughtful approach. But by the way, it took longer to get to market. But now that it's here, we have this extremely efficient silicon relative to the more general purpose approaches used in the market.

Got it. We'll come to the technology and the demand drivers. I just wanted to address the supply side quickly. On the supply side, right, just widespread constraints in the industry. How is Ambarella seeing the supply situation? I mean, when I look at the last few quarters, your results have been very strong, right? They have outperformed expectations. So where has the supply situation impacted you? When do you think you get back to trend? What steps are you taking to get back to trend?

Sure. So our operations team has done an exceptional job, and we're very quick to react to the situation, which a lot of times, you have people in denial, as you've seen in the past. And these guys did a terrific job. But we're not impervious to what's going on. So we have some video-based products that were in -- or are in Texas. And Texas actually had some impact and Samsung was one of those impacts. And so we've had to deal with that, but we've been able to do that in a reasonable way. It's not over. But over the balance of the year, I think you'll see things continue to get stronger and stronger. And I think we're in reasonable shape there, though. It's not like there's no impact at all. From a broader perspective, we are not with TSMC, where there's been challenges that people have talked about. We've been with Samsung. Since we were the second foundry partner behind Apple, they were a little larger than us, but we've been there a long time. We have a long relationship with them, and they've been good for us. And so again, we have worked with them very closely since it started. And while we have not been -- there's been no impact, we've been able to manage ourselves through it. Finally, it's substrates and packaging test, et cetera. We've got a majority of our employees in Taiwan and China. And so to have people on the ground, on the floor, working with these partners is, again, helped us in a difficult situation where a lot of people are struggling. It's not like, again, it has an impact to us. But we've been able to, to your point, get in front of it, guide and understand what the situation is and I think deliver pretty effectively. I think you're going to see that. I think Fermi mentioned that he thought this quarter would be the kind of the bottoming of the move, although there's still going to be impact through the balance of the year. And so that's what we're doing. We're focusing on it every day. We have phone calls and try to make sure that we're in the right place and feel pretty fortunate that we've been able to deliver in a way that I think has been pretty consistent, pretty effective for our customers. Now the things we can't control are the other components they need and some of the other things that are going on. And so that we have to work with those partners to make sure that we're doing that in a balanced way because we don't have any control over that. But right now, I have to say hats off to Fermi for reacting quickly, our operations team for doing the right thing, and I think it's shown in our numbers last quarter and our guidance for this quarter.

Got it. Casey, how are you making sure that the demand signals that you're seeing a real demand signals and it is not a sign of customers perhaps getting nervous or holding inventory? How are you ensuring the quality of the demand and the funnel that you have?

Well, I always joke and say as the CFO, I sleep like a baby. I wake up every 2 hours and I'm crying because you're always worried. You're always concerned about those type of elements. What we do is we have a very deep relationship with most of our larger customers and have pretty candid conversations. In other instances, if somebody feels they absolutely need something, we might have to go out and say, "Hey, look, if you want to pay for a super hot lot, we can get that to you. If you don't want to pay for that, well then, we're going to have to delay that." That tends to sort people out a little bit because if they're just trying to build inventory, they're not going to pay for a super hot lot. If they really do need it and they have the rest of the components to build out their products, they might do that. And so it's partnership, it's communication. And in some instances, it's making sure that we test people's approach to whether or not it's inventory or whether it's really demand. But we don't have any -- other than, again, relationships, et cetera, we don't have any super secret sauce to do that, but we try to use all of those to understand where we are. When we build inventory like we did this quarter, the other part of it is, we're buying wafers. As soon as we know that we've got demand, we'll put those into work in progress. As soon as they get to finished goods, they're out. We don't carry a big finished good. And so that gives us the flexibility, again, to move across customers and across situations to where we can accommodate what we actually for demand. But when you see our inventory go up, I think sometimes people get concerned about building finished goods. And that just isn't really the case in our situation. What we're doing is making sure that we have wafers so that we can quickly go to work in progress and delivery versus building up a bunch of inventory that sometimes people get concerned about.

Got it. What are your lead times now in your CV products, right, versus what they are, usually? I know you're early in the ramp. So I know there isn't a ton of history behind it. But what are the lead times now versus what you would feel comfortable?

Yes. To your point, it can be 26 weeks, it can be 40. It depends on the product and what we're doing. Some things are much more difficult to manufacture than others. Some need different treatment after that. So I would say that certainly, lead times have gone out. And I think partners are looking to have better visibility into what we're not only going to do this year, but next year. And so again, you sit down, you work from an operational basis, and you work through that. But lead times have definitely extended, in particular, in certain products. For example, if you had a vision-based product that was in Texas, obviously, when they had to shut down or restart, that increased that. Over time, they can tighten that up and get back there. Other places, the lead times, maybe it was 40 weeks or whatever the case might be, and it really wasn't impacted. And so it really is kind of a case-by-case basis that we continue to try to monitor across the products, whether it be vision-based products or CV-based products. All of our -- just as a reminder, all of our CV-based products are done in Korea. The only thing that's done in Texas is the -- some of the older vision-based products, which we -- they are still great products for us. But all of the CV is being done and, as I think Louis mentioned earlier, right now, we're working on 5-nanometer.

Got it. Excellent. So now on to the demand side, and you've clearly identified automotive as a key area of growth. Talk to us about what are the applications in automotive. And how many of those applications is Ambarella kind of the sole source? And then how many of those applications, perhaps you're sharing that opportunity with other suppliers?

So I'll let Louis jump in on that because I want to make sure that we both have an opportunity to chat a little bit. But it really does depend on the product. And as you know, we've been in automotive in certain ways, like data loggers and the like for a long time. Other things, we're just entering the market. So it can be different depending on the market and the project and what we're using today. But Louis, do you want to add some color to that?

Sure. Yes. I mean, there's 6 different camera systems we primarily target. The one where we're established is in data loggers or sometimes called drive recorders. And that market for us, the opportunity going forward, is really having them installed in new vehicles, and we have strong market share there. And the penetration of this type of camera system into new vehicles is less than 10% but rising. And so as that penetration rate rises, we think that will be -- continue to be a good growth market for us. The rest of the markets I'm going to talk about now, 5 of them principally, are all incremental to Ambarella, meaning these markets need, for the most part, computer vision, and we don't have revenue in these markets. And so what's happening is we're beginning to realize revenue in many of these areas, and that's what is really a key factor causing our automotive revenue to double in fiscal '22 versus fiscal '21 that was just completed in January. And these markets include Front ADAS, where Mobileye has strong market share. Xilinx is also in that market, as is TI. And this is a market where, just in this last earnings call, 1 quarter, we announced 4 design wins where the front-facing camera is included in terms of the functions being run on our chip. The next market to notice would be in-cabin camera systems. This can be more than 1 camera, by the way, such as a driver monitoring camera or a occupant monitoring camera. Electronic mirrors is a market that isn't quite as active yet as the in-camera -- camera systems but we are very optimistic about the longer-term potential for e-mirrors, and we think we'll have a very strong share in that market. And then L2+ and L4, L5 are expected to already begin to contribute revenue this year, although they become much more significant in calendar year '25 and beyond. We just announced arrival as they represented a design win for L2+. And for L4 vehicles, we announced Motional where we're providing our CV2FS chip to do all of the exterior cameras in their next-generation vehicle. So that's where we are today. Over time, you'll see us introduce new products where our content in the vehicles will expand beyond just camera perception with the next step being a domain controller and what happens with the domain controllers, we'll continue to do all the camera functions in these different applications, but we will also do the sensor fusion where we'll take the data we're already collecting for the cameras, and we'll fuse the data coming out of adjacent sensors, whether it's radar, LIDAR, ultrasonic or some other sensor, we'll fuse it all together into this real-time point cloud that's updated 20, 30 times every second. And to do so, you need a tremendous amount of incremental processing power. And as we add more processing power to our chips, that causes our average selling price to rise.

Got it. No, very clear. Can you give us a sense for how much of your, let's call it, traditional vision business was exposed to automotive and how much of your CV business is exposed to automotive?

So historically, automotive, if you take last year as an example, was 15% to 20% of our revenue. And most of that -- almost all of it was video processors primarily for drive recorders. And again, this will be the first year where we're beginning to show traction in multiple computer vision market opportunities such as in-cabin or front-facing ADAS cameras or even L2+. And for CV, we -- as of May 31 this year, so a little more than a week ago, we announced that on a cumulative basis, we've shipped 450,000 units into the automotive market for CV. And most of those 450,000 units have been shipped in the last year.

Got it. And roughly, what does that imply for content or sales? Because I know in the prior quarter, I mean, you reported a very steep ramp, right? I believe it was 300,000 units in the prior quarter cumulative, and then it went to 450,000. What does that represent in terms of just the revenue growth in automotive sequentially on the CV side?

So our ASP and what we're shipping today ranges from -- CV ranges from $15 to $20 to $50 to $70. So there's a range of different chips in our portfolio. We've currently got 6 chips in production and the seventh one is sampling this quarter.

So that is content for the car or that is content per CV chip, right?

That is for each CV chip. And in some instances, for example, take Motional or we announced there's multiple CV2 chips in that vehicle, same thing with Arrival, multiple CV chips. And relating this back to our earlier description, today, what we're -- the problem that we're solving in a car is at the perception level specifically for cameras. The next step for us will be sensor fusion and adding that incremental processing horsepower to our CV family in order to do sensor fusion.

Got it. That was the other topic that I wanted to ask about because that seems like a very interesting development, but also jump into more horsepower, right, and more processing performance. Does that increase the competitive aspect? Because, again, back to the all the competitors that I mentioned before, they are also very well engaged in the sensor fusion side, right? But obviously, they are coming at it from various perspectives, right, whether it's NVIDIA or Intel or others. How do you -- how successful do you think you will be able to be in kind of making that jump into sensor fusion versus the camera applications that you have right now?

Well, we're scaling into the sensor fusion domain, the domain controller, if you will, from the perception and working our way in towards the middle of the car some of our competitors, you mentioned, start in the middle of the car and may have more challenge working towards the edge given the efficiency metrics, whether it's performance per watt or performance per dollar that we've spoken about. So again, our sensor fusion chip will integrate the camera perception that we're already doing and then the incremental horsepower, if you will, whether DMIPS or TOPS for the CV flow architecture will be providing the sensor fusion processing, and we'll be able to charge more for that. So we're leveraging our way in from a position of strength and towards the middle of the car.

Got it. Talk to us about the opportunity for upgrading the professional security business with the CV product. What is the use case there? Where are you in terms of upgrading your -- is it upgrading an installed base? Or is it actually increasing your addressable market with the CV product?

Yes. The answer is both. And to back up and give some perspective. There's about 900 million security cameras in the world today installed base and the replacement ranges from 4, 5, 6 years. So what is that 15% to 25% replacement rate. And then to your point, the installed base, that 900 million, is continuing to grow the way we see it for several reasons. One is the deployment and use case of security cameras remains compelling, whether it's in a public infrastructure or it's an enterprise or it's in the home security market. And the other reason is [Audio Gap] allocations that we'll be using these security -- or these CV cameras is expanding quite rapidly beyond just the security function of catching bad guys. So for example, different types of automation products, retail store, analytics, intelligent transportation, these are all examples of markets where they're not driven by security and catching bad guys. They're not even human viewing applications often. But what we're doing is our CV chips are enabling a machine to perceive the environment and semi or fully autonomously make decisions based on what it's seen and what that's causing is this installed base of let's just call them IoT cameras to expand well beyond security market that represents the majority of that 900 million installed base today.

Got it. So if we kind of bring it all together, what is the right way to think about a business model for Ambarella in terms of sales growth, right, where margins can grow? Because we have seen a very impressive recovery, right? You have been able to hold costs. You have been able to manage them extremely well. We are seeing this really strong ramp on the operating side. So how are you able to kind of grow top line while keeping expenses right in that very tight band? So talk to us about what the business model looks like for the next 2, 3 years.

Well, as we've talked about on the top line, CV is about twice the revenue as the traditional products, right? On a margin basis, they're roughly the same. But right now, we've been performing at the top end of our guidance. And so that's good. I think from an OpEx standpoint, obviously, we see an opportunity. And we saw an opportunity 5 years ago, but now we're actually being able to execute and demonstrate the opportunities. So we're not going to hold back to make sure sure that we take -- after all the hard work and suffering, take advantage of it. So we're hiring, not only in engineering, well we've been hiring and engineering for years, but we're building out the sales, the marketing, the field application engineers in Europe, Germany in specific, and in U.S. and around the world. And so we're trying to make sure we invest in that. I think the interesting thing that you probably saw in the last couple of quarters, and we talked a little bit about is the leverage. The leverage in the model, I mean we were able early on in the best days to 20% to 25% margin. And we feel that we've demonstrated pretty good growth over the last few quarters. And I think that's really the thing to keep an eye on is we're growing revenue because of the leverage I talked about there. We're adding to the expense line because we're not going to miss this opportunity. But in light of all of that, we're still able to grow the op margins to a way that people have been looking for. So we're excited about that. We continue to have a lot of cash, but in this business, you need it. Because in particular, in automotive, they want to know that you've got cash for 10 or 15 years and 20 years. And so we're going to do that. But we're going to invest in the business, we're going to -- Louis has been a terrific partner to the company and looking at opportunities for us. And we've been selective in what we've done, but we will do what it takes to make sure that we're able to execute right now because we feel that we really are in a position to have a very unique presentation of our opportunity that just doesn't exist out there today. It's a combination of the 16 years of work done in the video processor, but it's also a combination of buying VisLab over in Italy, leveraging that capability and starting to generate the CV technology that it's just now coming out. So we're pretty excited about it. We feel we have an opportunity to demonstrate, and we feel we are starting to do that. So I would keep an eye on those things in particular.

Got it. Final question in the 30 seconds or so that we have left, is there a certain mix of end markets that you want to be in, say, 3 to 5 years from now, right, the consumer versus enterprise versus automotive?

I will be flipping, giving Louis a minute to think and say anything that makes money, because I'm a finance guy, and so we'll focus on anything that makes money. But no, I think we do have a focus and we've demonstrated from the traditional consumer markets that we were in to the markets we're in now really where our heart is. But Louis, why don't you add just a little bit if you have anything to that because I do think that, that's an interesting part of the story?

I mean it would be surprising by 2025 that auto doesn't become a majority of revenue. And that's not to say security won't be posting solid growth. We've been more cautious on the other revenue as it's mostly video processor business. But we have seen that some of our new products, like CV5, there's also some interest in that other business as well. So we'll see how that impacts the outlook there. But clearly, auto should be the key revenue driver. And we'd all be disappointed if it's not around 50% of revenue in calendar year '25.

Excellent on that positive note, and I hope we come back and look back in calendar '25, Louis, and I'll mark it down, more than half the business. So on that positive note, let's wrap up the session. Thank you so much, Casey and Louis, for joining us this afternoon and giving us your insights about the industry and Ambarella. Really appreciate it.

Great. We appreciate it.

Thanks very much. Thanks, everyone.

You bet. Bye.

Bye.