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

Hi, welcome back, everybody. Before we start with Ambarella, I do need to read a quick safe harbor. For important disclosures, please see Morgan Stanley research disclosure website at www.morganstanley.com/researchdisclosures. If you have any questions, please reach out to your Morgan Stanley sales rep. I know you've heard that a lot of time, but anyway, we're here with management team of Ambarella, Casey Eichler, CFO; Louis Gerhardy, Head of IR, after reporting a really strong quarter last night. So welcome, guys. I don't know if you need read safe harbor. We'll just go straight into Q&A.

Great. I appreciate it. Appreciate everybody joining us here today. Louis and I would be happy to answer questions as they come up.

Great. So maybe if I could, to start kind of big picture, I mean, you guys have come a long way. And I feel like the stock has sort of reflected it in the last 3 months, but over the last time we did one of these in November, there's still a lot of negative sentiment. You guys have been going through a consumer transition and kind of building up this computer vision business and, like I said, the stock seems to be reflecting a little bit more now. But just talk about this bigger picture, computer vision, what are you guys trying to achieve? And where is it heading?

Yes. So as you know, this has been a long journey. It started off around 5 or so years ago, bought a company VisLab, combined with Les Kohn, who's one of premier computer architects, and Fermi and the team. They went through a process of starting to develop our CV technology that really has just coming in -- to your point, is really just coming into to focus from a revenue standpoint and from a market standpoint in the last year. We talked about 3 waves, and that means the -- firstly, professional and then the home security market. Second being the -- I'm sorry, the second being the home security market. And then the third being the automotive and some of the other markets that we're in today. And what I think we're seeing now is, obviously, we've got several products out today, and we have several engagements in all of these markets as well as new markets like industrial and others that we've talked about. And in the first few years, we were waving our hands talking about how great this is going to be and how good the products are going to be. And today, we actually have the ability to demonstrate that and our customers' capability to demonstrate that. So for example, our first wave in professional security, Dahua has just brought out products in the last year and now into this year in CV, several different products that really demonstrate the capabilities. That's just one example, of what I think, to your point, is generating some of the proof to what we've been talking about.

Great. And I think, as you said, the leadership of the products has been kind of clear for a while. It takes a while to generate revenue. We talk to customers, whether it's customers in surveillance, customers in autos. It's been pretty clear that you guys have really leadership capabilities in computer vision at the edge. Competing with like programmable solutions, these are dedicated solutions. Competing largely with other programmable chips. Can you talk about the approach that's led to that? And why you guys have leadership capability?

Sure. What I'll do is I'll talk at a high level and then let Louis chime in as well. But we started off with our compression and our video solution that I think was well received. Today, we've put a neural net processor on the same chip to provide not only human vision, but computer vision in a single chip solution. In a lot of places -- in most places, what we're doing is we're providing that platform. They're able to then generate their own software capability that demonstrates or gets their product to demonstrate the capabilities that they want to address. But it's really -- it starts with our chip and then builds off of the software capability of our partners. Louis, do you want to add anything to that?

Sure. Yes. AI, as you know, has been around a long time. And historically, it's been executed in the cloud or in the data center servers that would do the processing. We're serving in the third tier of the market, the actual endpoint device, some people call it, the IoT device, and we're doing as much AI as possible. Sometimes all of it in that endpoint device. And the advantages of doing that is you don't have all the costs and the latency of moving the information back and forth to the cloud or the server edge. And it's a very difficult thing to do because frequently, these IoT endpoint devices are operating off of battery or they have a form factor with significant thermal constraints. And so the architecture required to do this inference AI processing in an IoT endpoint device is much different than the silicon architecture used in servers and data centers in the cloud. And that's fundamentally how our leadership is different than some of the other companies that are known for doing AI and servers.

Great. And so maybe I'll do this a little bit in reverse order. In the sense of Wave 1 is professional surveillance; Wave 2 is home surveillance; 3 is autos. But a lot of the developments from last night's call were in autos. So maybe I'll kind of jump into some of those, and then we can go back. Maybe just starting, there's a question from the webcast about that 300,000 units that you've shipped computer vision into cars. If you could talk about what applications drove that 300,000? That was a higher number than I would expect, frankly, because you've always talked about CV in cars as being more next year. Where -- how much of that is coming from in car monitoring, other types of features as a question alludes to pre-CV products as well, but you can just give us sometimes about 300,000?

Sure. So just the big picture with auto is this last year that ended January 31, auto was 15% to 20% of our revenue. Think of that as $35 million to $40 million. Yet we talk about an automotive SAM for us of about $2 billion roughly. And the reason the SAM is so much larger is that it's increased a lot because we now have the CV products out in the market. And you're correct, we did announce 300,000 units. We expect to have shipped 300,000 CV units by the end of our April quarter. And those products are not going in, for the most part, into markets that we've traditionally served like driver quarters, but they're going into markets, for example, commercial fleet, aftermarket, ADAS boxes that might do some combination of driver monitoring or front ADAS, even simple L0 type applications. And then other units of the 300,000 might be going into pre-installed front ADAS solutions. For example, we've announced one customer called MAXIEYE that's in mass production for those front ADAS products today. Guarantee, yes, we talked about reference designs to help us become more successful in the passenger vehicle market. And so I think calendar year '21 or fiscal '22, we're going to show progress on multiple fronts in the automotive market, not just driver quarters, but front ADAS, the in-cabin cameras, including driver and occupant monitoring electronic mirrors and then, of course, with the announcement yesterday, for L4/L5 vehicles with the relationship with Motional.

Great. So maybe we could talk about that. Motional is sort of not a household name for people, but it's actually -- because it's a relatively recently formed JV between Hyundai and Aptiv. But it seems like a really important player in sort of the Balfour economy. I sort of describe that as a breakthrough win. Can you talk a little bit about that? Like what exactly are you doing for them? And what's the -- I mean, I assume L4/L5 is a longer path to revenue, but can you talk about what that path may look like?

Sure. They've made some announcements, I think, in December about their commercial status and different programs they have in other areas. And I would refer you to the rollout to what they've said, and there's a lot of information out there now about it. In fact, you might be able to get rides in some of these vehicles in the cities that they're talking about. Our role here is that we have multiple SoCs in these vehicles, and we're part of the central processing module. And we're handling the image and computer vision processing for all of the cameras and this includes the front-facing camera as well as the other cameras around the vehicle. And our chips not only do the image capture and doing so with very high resolution and in difficult lighting conditions, but we're doing the AI processing, we're doing the encoding, we're providing the ASIL B functional safety capability, which is critical and all of those -- in the encoding capability. And we're doing this all simultaneously with a very attractive thermal budget and, of course, efficiency in terms of price. So it's a very exciting opportunity, a very complex system. Motional is running their software on our chips, and that's very consistent with the open platform approach that Ambarella is promoting in this market versus the black box approach that some of our competitors have historically taken.

Great. And the fact that this is a joint venture between Hyundai and Aptiv, obviously, both of those are potentially interesting customers down the road. I know they're separate organization, so it doesn't necessarily imply traction. But does it open the door a little bit with those other customers?

I think this is a very strong testimonial for the capability of our chips in HMC, or Hyundai Motor Corporation, is one of the leading automotive OEMs, as you said. And Aptiv is a leading Tier 1. And those are very important prospective customers for us. We don't have anything to announce on that front. But obviously, they'd be very important prospective customers, and we're going to have to earn our business with them.

Great. And I think the other observation I have from these announcements and really from what we're seeing across the automotive supply chain is that the automotive customers really do want to develop their own systems around this, and it's been a tremendous source of value actually, right? If you just look at GM crews or the way that these autonomous organizations are being valued, it's a large component of the value at the OEM level. It seems like that plays to your strengths because you guys -- you compete with mobilize the entrenched level 2, tends to be kind of a black box solution, and you're competing with companies that have other programmable companies. But as you move into electronic vehicles, it seems like the battery power is going to be an important feature. So am I getting too excited about all that? I mean it just seems like a world in which people need a dedicated chip solution that allows them to develop their own platform. And you guys are really the only people that are providing that in the way that you do.

Yes. I think you're dead on. As you know, we have a slide that kind of divides it into 6 different markets. And the first markets are the ADAS markets being recorders, obviously, the forward-facing and the mirrors and the in-cabin area. That's really generally been the Tier 1's world. To your point, there's a lot of investment and a lot of activity going on in L2+ and L4 and L5 that's in the OEMs. And the OEMs are taking a much broader leadership role because when you get out there, it's really going to be fundamental to the car itself. And so that's not to say that there's not going to be activity with Tier 1s as well. But you're absolutely right. When you get out there, the OEMs are going to be much more active. Today, we are the leader in the recorder market. It's a smaller market. It's not emerged as much as the forward-facing ADAS market, which is a market the Mobileye is leader in today. And then the other markets are really kind of open to everyone, and everyone is going about it differently. We have an approach, to your point, where we have an open solution. We're providing a platform and you can deploy your software onto that platform. Mobileye has been more of a closed box solution. In fact, they don't even develop the chip itself. They're really doing the software. They combine those, and that's a single solution that you can buy. We're finding that OEMs and even Tier 1s want to be able to differentiate and have an open solution where they can differentiate based on what the customer wants. And I think over time, that's going to be, at least in our opinion, a more successful approach to the market.

Great. And a lot of the strategic value of the company is going to come from Level 4 or Level 5 because it's such an important market, and it's driving so much activity across the industry. Obviously, a very long path to revenue for that type of stuff. And so maybe we talked about some of the other drivers. You just mentioned recorders, and we sort of give secondary consideration to announcing Ford, NBW sort of China JV. And it seem like pretty big potential wins. What -- and that's what's going to drive the revenue presumably in the short-term recorders, digital rearview, stuff like that. The penetration there is relatively low. What are the prospects there for those parts of the business to continue to grow where, as you said, you already have the benefit of incumbency?

Yes. I mean from the recorder business, as you know, that started out primarily in Asia as a market. It's now becoming more popular globally. Again, we have a slide that goes through that market and says that it's a 5% to 6% penetration. We are the dominant player there with the Mobileye, if you will, in that market. And I think you're going to see that grow quite a bit over the next few years. And also, it was interesting, and Louis will have to correct me if I'm wrong about this, but Tesla came out with, I think, it was called Sentry, which was basically just a recorder that when it was brought out in the U.S., everybody thought that was just kind of this wild new thing that Asia has been doing for several years. But it really did wake people up to the importance of the data logger, as they say, or the Ford of the recorder business. And I think you're going to see that continue to grow pretty nicely over the next few years as the penetration rate gets deeper and deeper.

Yes. So seems like it goes hand-in-hand with more autopilot type functionality. You're going to need to have more accident resolution to figure out who the faulter was and things like that. And then the last automotive question, in cabin monitoring, seems like a category that can -- CV category that can drive revenue sooner than something that manages the driving of the car. Because it's a safety feature that makes you more safe, even if it doesn't work 100% perfectly. And I know it's part of the European luxury car standards to get certain safety ratings. And last night, you announced several Chinese customers going to production for in-car monitoring. And we talked to our your customers there, there's a lot of enthusiasm for the Ambarella product from the Tier 1s and Tier 2s to supply that market. So can you just talk generally about that market? And is that going to be relevant this year?

I'll let Louis go ahead on that.

Yes. Driver monitoring, the penetration rate of that technology is low for passenger vehicles, let's call it, 1%. But you're right, the new car assessment programs are increasingly recommending different types of in-cabin camera systems, whether it's driver monitoring, which is most frequently discussed, but also occupant monitoring and eventually other sort of in-cabin monetization schemes. But for now, the market is evolving from commercial vehicles. In-cabin technology, driver monitoring, particularly, has been around 10 years. But it's been used in like mining vehicles and low-volume opportunities, but where the ROI investment is very high. And the software has been executed with traditional machine learning type algorithms on general purpose processors. And here we are at the point of time where we're about to put this in passenger vehicles, and the algorithms initially are not as complex as they will become. And what Ambarella offers that's very attractive to our customers is a scalable CV family where they can start out with a product like CV28, which doesn't have a lot of AI capacity in it, but it's got great video resolution. And they can use that as an entry way -- the entry-level product into an in-cabin system. And then over time, as they continue to develop their software and exceed the capability of CV28, then they can use that same software in CV25 or CV22 or even CV2, our high-end product. So it's a very attractive market, very early stages and we offer that scalable family of chips that will help them have staying power in this market and gracefully evolve as the algorithms become more complex.

Great. So maybe we could shift to Wave 1 of the CV story, which you've already sort of started to see last year, professional surveillance. That market, if you talk to your big customers there, which we have, you hear a lot of positive things about Ambarella. You also hear about the potential for putting AI into the camera to really expand the TAM. But it's part of the reason that companies like Vision or so enthused about the directions that they're going. Can you talk a little bit about that? And what's the penetration today? I assume it's still relatively small at 10% of your revenues exiting last year of those professional cameras. And where is that penetration going to go over time? And understanding that's 2x plus the average selling price when you convert a win from video processing to CV?

Yes. I'll talk a little -- I'll let Louis talk about some of the new uses that you referenced. But you're right, the majority of the revenue last year was Wave 1, which was the professional side of the business. The home side started to come in the second part of last year, and we'll start to mature into this year. That's going to continue to drive. The other thing that's going to be a driver in that market is the fact that we've kind of talked about last quarter kind of getting through the burn of this inventory that was built up. And so we're starting to see both with Hik and Dahua. We're starting to see some of that come back. And then, as I mentioned earlier on, with the CV, the Dahua is not bringing out several different products, that really sets up the year pretty nicely. Barring a change like we've seen over the last couple of years that you can't predict COVID or whatever you might want to pick, I'll let Louis talk a little bit about that because some of those products that are coming out now do apply the CV technology in a different way in the market. And I think that's what does expand our opportunity in that market.

So this market, security cameras, the installed base globally is about 900 million units as we speak, with less than 5% of that were inference AIs enabled in the camera itself. And so we're just beginning this new product cycle. Meanwhile, that installed base of 900 million units is also increasing. So you think about this market fundamentally, as you know, Joe, would both a replacement rate and new subscriber growth. And to emphasize the point Casey was making, most of those 900 million installed base cameras are being used for security applications and catching a bad guy, but what Casey suggested is now there's applications, now that machines can independently proceed their environment through cameras and make decisions partially or fully autonomously, that opens up all kinds of new markets, for example, access control. And in access control, there's a number of subcategories. For example, electronic locks that use a camera as well as some other sensor like time-of-flight or structured light to improve the accuracy of the results. So that it gets to that 99-point whatever percent that's required for that product to take off in the market. Enterprise access control panels is another example. A number of different robotic applications, those fixed robotic applications like factory automation, where cameras are used in the production line. That's another category. Again, has nothing to do with security function. It's a new market for us. Mobile robots, both aerial robots like drones or mobile terrestrial robots like AGVs or AMRs are also going to need cameras very much like autonomous vehicles will. These transportation or delivery vehicles are going to need cameras in those markets. And then there's another example called, what we call, sensing cameras. And these are cameras where the image and your private information or your image is not shared. All that's extracted is data about what's happening. And for example, this last quarter, we announced a design win with a company called GeniusPro where they have a sensing camera used to count people, how many people have entered and exited a room. These cameras can be used for aiding in ventilation applications. There's -- they can be used in public transportation applications. So there's a wide variety of new nonsecurity applications emerging that will help drive this installed base growth higher. And then we also have the replacement rate in the traditional security camera market.

Great. So we're running up against the end of our time. I'll go a little bit -- a few minutes late because we started late. But maybe you could talk just briefly on Hikvision. You mentioned that Dahua has sort of come in with orders for its CV portfolio. It seems like from the call that you're not banking on Hikvision moving in that direction. On the other hand, once they run out of inventory of high silicon product and Ambarella product, there's a lot of enthusiasm at that customer to do AI cameras. I've heard it from them first hand, like it seems like there's still an opportunity there. So just -- I don't know if I can -- you want to get into any individual customer, but it seems like you fairly derisked this China surveillance exposure and maybe there's upside, if that works out better than you're expecting. Is that a fair way to look at it?

Yes. We've talked about how we don't have that built in to our numbers today. To your point, it's not like they haven't licensed SDKs in development opportunities and they haven't looked at our products, understand what we have, want to understand our capability and so that is an opportunity. Right now, in the high end, as we've talked about in the past, HiSilicon and ourselves are kind of the 2 players, in particular, when it comes to CV. And so if they were to get into a situation that they were trying to push, and I think you're right, they see the importance of pushing CV as the products come out going forward. I think there's an opportunity there, but we'll take that as upside if it happens. And it's really a difficult situation, as you know, with the global tension, et cetera. They're -- maybe you might know better than I, but they're probably 30% owned by the Chinese government. And so there's obviously different issues and pressures there and different probably opportunities for them there off of that. So we'll just have to see what happens. But clearly, I think they understand the capability of our products and they understand the importance of using CV going forward.

Great. So I have a couple more questions. One for me and one from the webcast. Consumer surveillance seems like as a category, even before computer vision has started to penetrate. It's been a pretty healthy category. Things like Amazon ring doorbells have been some of the bigger Christmas promotional items that we've seen. And there's just a lot of growth there. One of the -- you guys have sort of gotten out of consumer businesses, and yet this really plays to your strength. So how are you thinking about consumer surveillance? And then I think you announced last night that you have kind of a -- you didn't name the customer, but kind of a flagship customer that will adopt CV in the consumer part of the business. So maybe just talk to that opportunity a little bit.

Sure. Again, we talk about both the professional and the home as being much broader than just kind of the security to catch a bad guy at the door kind of thing that you see today. And ultimately, as you drive up into your driveway, the camera can recognize that it's you open up your garage, turn your heating to the level of heating you want, turn the music on that you like and turn the lighting scheme on that you like. All of that can be done today. And I think as time goes forward, it will be about security, but it will be about kind of smart home automation and smart delivery and several other markets. And so I think that's the real opportunity that's much broader than just the security side of it that there really hasn't started to evolve in a way that I think is possible in the future. So that would be my spend. I don't know, Louis, if you have anything to add to that.

Let me just briefly say that the consumer we talked about 5 years ago was more of, what you call, in economic terms of consumer leisure good, where maybe its life cycle was 2 years and you throw it away and maybe get a new one. Maybe you want. Whereas the consumer business we have today is more consumer durable where you have the system in your home. It's there for 5 or 6 years, and then there's this replacement rate built behind it. So really a totally different type of consumer good, much more durable good than this discretionary or consumer leisure type product we sold in the past.

Yes. And it does strike me just listening to you guys. When you think about -- if you teach a car to drive, there are so many things that you can do to replace human functions that are much simpler than driving a car, whether it's on a factory floor or a robot or the home automation type stuff. So exciting stuff. So I'll finish with one question from the webcast. It seems to be a proliferation of AI silicon start-ups across all verticals, including AVs. How does Ambarella plan to stay ahead? And what provides you guys with a durable advantage?

Well, I think, to a large degree, it's the technology. You've known less and for me for a long period of time. Less is considered one of the premier computer architects in this space. The add of VisLab and Dr. Broggi and his team in Italy really kind of cemented that for us. In like any technology market staying ahead is by having a better product and further advances ahead of everybody else. Yes, I mean it's difficult when you've got stacks and other companies they're going public or doing a stack and they have no revenue, and they're talking about valuations off of 27 and 28. And here, we -- getting higher valuations than we have as we continue to kind of muddle along actually selling products and generating revenue. But we'll stick with it. I'll put my money on Fermi and Les and Dr. Broggi. And I think at the end of the day, people are going to see the value of CV, the value of the products that really, it seems like for all of us, and I'm sure for you, Joe, you've been involved in a long time. It's been a long road, but we now finally are getting enough products out and enough markets address. But I think people are really appreciating the value of what we've been talking about over the last 3 or 4 years.

Yes. Well, I think there's kind of an illuminating comment last night when Fermi said, Motional is a project you've been working on for 4 years, right, which predates Delphi changing its name. Obviously, these are things that where the CHF to be planted well in advance. And it's one of the frustrating elements is that it takes a long time to generate revenue. It's like once you get revenue, it's pretty solid revenue. All right. With that, I'm afraid we do have to wrap up. But good talking to you guys, and good luck with everything.

Thanks, Joe.

Thanks, Joe. Appreciate it.

Thank you.

Bye.