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

Good morning, everybody. I'm Joe Moore. I'm very happy to have with me today, Casey Eichler, the CFO of Ambarella; and Louis Gerhardy, the Head of Corporate Development. I have a quick disclaimer to read. Please note that this webcast is for Morgan Stanley clients and appropriate Morgan Stanley employees only. This webcast is not for members of the press. If you are a member of the press, please disconnect and reach out separately. For important disclosures, please see the Morgan Stanley research disclosure website, www.morganstanley.com/researchdisclosures. If you have any questions, please reach out to your Morgan Stanley sales representative. So welcome, guys. I do want to point out that Ambarella is in the quiet period, so we're going to keep the company focused on kind of bigger picture things. I won't be discussing any different things. But if you do have any questions, please feel free to enter into the portal, and I'll read those as we get towards the end. So Casey, Louis, thanks for joining me. Maybe we just start at a really high level.

You guys have been in transition. You were primarily kind of video processing before you went through a commoditization cycle there. And now you're sort of driving towards this computer vision. Can you just maybe talk about that transition and big picture where we are before we get into the details?

Sure. It was probably about 5 year ago that the team got together and thought about their markets and how to scale this company in a significant way. And I think what they realized was that some of the markets that they were pursuing at that time were just not big enough markets to really drive the revenue profile that they wanted. And 2 of the markets, both sports cams and drones were kind of dominated by a single kind of 80% customer, if you will, for those markets. And so they made a decision to take this kind of groundbreaking video technology they have and combine it -- that vision system with a computer or machine vision technology on a single chip, so you could provide both human vision and computer vision or AI technology on a single chip. They made a small acquisition of a company called VisLab, which had been doing AI stacks for autonomous driving since the early '90s, and they embarked on going off and designing a new CV, what we call a CV or computer vision architecture that would allow us to pursue much bigger, broader markets as we move forward. So the last 5 years has really been a transition of moving from vision-only products to computer vision products. And that's still, to your point, a transition that's going on as well as we are changing markets and expanding not only a market that we were currently in, which is both professional and consumer security as well as level 0 to level 5 autonomous, and we can talk more about that, robotics, smart manufacturing, smart warehouse, a whole set of larger markets that I think are really going to give us the opportunity or starting to give us the opportunity to drive a whole different profile in the company. And so we rolled our first products out a couple of years ago. We just started generating revenue in CV at the end of last year and along with COVID and everything else, matured out through this year. And we really feel that we're at a point where we can now exploit the $300 million -- I'm sorry, $400 million, $450 million of investment in CV that's now starting to generate the revenue profile that we've been focused on for the last 4 or 5 years.

That's great. And I think 1 of the things I find kind of interesting about this, when you guys started this journey, there was a certain way of looking at computer vision analytics where you were using heuristic algorithms to identify images and things like that. And then as it's evolved, machine learning, deep learning has become sort of the core approach around computer vision. Maybe talk about that. And you guys pivoted along with that. You have a software ecosystem that really is designed to take advantage of AI. And it's clear from talking to your customers. And I always tell investors, they should really be talking to your customers, not to me or you probably, because there's so much enthusiasm for your customers about these products. Maybe just talk a little bit about that evolution and how you've pivoted to take advantage of the new way of looking at computer vision?

Joe, It's Louis. I'll answer that. You're right. We do deep learning. There's many types of AI processing. The most sophisticated of which is deep learning. And before we even laid out any transistors in our first computer vision chip, we understood what is the best-in-class computer vision software stack that would run on our chips toward a part to figure out how can we create an optimized architecture that incorporates both the video processor we've always done plus this entirely new AI, deep learning processor. And 4 years later, this led to our first deep learning CV SoC. And we've just started to generate revenue in the last year from this. So what's important about this architecture, though, it's fundamentally different from what's used for AI today, for visual AI today. Whether it's -- today, a lot of AI is training and inference. It's done on servers and a network edge and the cloud or data center. Let's just call -- say the AI today is heavily done in the network. And what's different about our CV SoC, the way we go from the get-go was to be very optimized to do this deep learning in the edge endpoint device, or let's call it the IoT endpoint or node. And this is true, whether it's a security camera, a vehicle or some other IoT device that we're going to target. And when you could do all of this in the endpoint or camera in our case, you save all of the expense of the data transmission that has to be sent into the network. And in some applications, much more importantly, you eliminate the latency and the delay of all this data going into the cloud and then coming back to you. So in summary, we really have a fundamentally different approach. We target inference, deep learning in that ultimate endpoint device.

That's great. I think it's a great overview. So you're going to be roughly 10% of revenues this year for computer vision products. I think it's an impressive number given you were mid-single digits in the first half of the year. And you're really only seeing 1 major market, which is the professional surveillance market kind of adopted. But can you talk a little bit about that 10% and kind of maybe some sense of how quickly that can grow over the next few years from here?

Yes. So as you referenced, we really are just starting. We've talked about 3 waves in our revenue, the first wave being professional security that matured the second half of last year and into this year. The second wave is consumer security, which is maturing or starting in the second half of this year and maturing into next year. And then the third wave being some of the other markets we talked about, like robotics and industrial and the like, but also some of the early stages of automotive and some of the automotive that we can have more things like aftermarket and fleet, in other areas where we can start to see some revenue there. Obviously, some headwinds around -- over the last couple of years around trade and then more recently, around COVID, but as we talked about in our last call, we continue to have pretty good adoption across CV and we remained -- at the beginning of the year, we suggested that we thought we could do 10% in CV revenue, as you referenced, and we reinforced that again in our last call. I think that with some of the discussion that we've talked about, we're going to continue to see that mature and grow. And there's 2 exciting things. One is that it gives us a whole another revenue stream for growth. The other thing is CV, as a reminder, is twice the ASP of the vision products. And so as we get more and more adoption from vision to CV, that gives us leverage on the top line. The margin profile is about the same, but obviously, the op margins benefit from that, and that's exciting as we continue to build that revenue stream. So like I say, we've been talking about it for a couple of years. It's now just starting to happen, and we're very, very excited about the progress there, even though there has been some geopolitical and obviously pandemic type of headwinds that we've all faced, I guess.

Yes. Yes. I think it's pretty exciting. I mean, this is -- these are really important markets when you look at the importance of deep learning and you look at the drivers in the stock market, having specialized product to deal with vision at the edge, I think is really important. So maybe -- I mean, you mentioned this, but I think the company would already be in a strong growth mode this year, if not for the trade issues in particular with the Chinese surveillance customers. That was over 1/4 of your revenue this time last year, you're down to mid-single digits now. Can you talk about the longer-term drivers of that business? And how important are your China customers? And do you see that -- long term, how do you see that evolving?

Well, China is, without question, the largest market today and Dahua and Hikvision are the largest providers in that market today. So it's important. We talked about at the end of last year because of some of the trade concerns and considerations, some of what we thought would be in this fiscal year was pulled into last fiscal year, kind of an inventory build, if you will. We estimated that to be around 10% or so. We don't have perfect visibility into that, but we kind of tried to put a number on that. And we've talked over the last couple of quarters that our focus or our concentration of revenue, as you referenced, had been down. And I think part of that is just the environment and pandemic on top of that. But I think part of it is the use of some of that inventory burning off. And we had estimated and we thought that was going to happen in the second half of this year when we first talked about it. And it feels like that is the case. As we look forward, I think CV adoption across that market, whether it be in China or more broadly, is going to be key. And the differentiation of what Louis talked about in our CV approach, it's not all created equally. And when we look at our CV, we really are able to do things at a price point and a performance in a power metric that really is very attractive for this market. So we continue to feel we have the right solution for the market. We don't control the geopolitical or pandemic elements of it. But I think that there'll still be opportunities there for us. I think it will continue to be an important market, not only with those 2 customers, but there's another set of customers in that market as well, kind of being out there, the next wave down that we'll be interested in looking at this new technology, and we continue to feel confident that we have the right solution, not only in China, but just broadly for the market globally.

Yes. And in this professional security market, like I said, one of the things I think investors ought to do is talk to your customers because there's so much enthusiasm from Motorola, Avigilon and from Hikvision and some of the customers about what computer vision can do for the camera by having this capability like in the camera itself. So maybe you talk a little bit about that. What does it do to the market? Do you think it -- this should be a shot in the arm for the surveillance camera market? And then what does it do for Ambarella specifically from a dollar sort of -- you mentioned 3x the ASP, but does it -- what does it do for your market share? Or what does it do for your economic opportunity within this market?

Yes, Joe, your question was on professional surveillance, but just more broadly, I think everyone is familiar with the SAE definitions of levels of autonomy in the automotive market, L0, L1, L2, L5 and L5 being fully automated, robotaxi. But if you think about every single market we're in that's moving to CV, you can think of them with the same sort of framework of no automation or some automation or L1. And almost every single market is in that very early stage of the L0 to L1. And so for professional surveillance, for example, L0 -- no autonomy today would be a security guard looking at 8 monitors and study show that after 20 minutes, looking at 8 monitors, the security guard starts to get fatigued and make mistakes. But as you enter CV into that, maybe that becomes L0, all of a sudden, detection of whatever you want to look out for, crossing a perimeter or a grouping of individuals that's too large or reading license plate, all becomes automated and removes the mistakes that can get introduced by human. So simplistically, CV is now allowing machines to see. And for the entire value chain, this is very profound, because you now have new business models being introduced. For example, reoccurring revenue potentially for our customers, eliminating the bandwidth expense, entirely new machine-based applications. Just as an example in the recurring revenue, with CV in a camera, our customers can do over-the-air software upgrades and completely change the purpose of the camera from, say it was doing license plate reading, upgrade it, then it can do line counting or upgrade it again and then it can do contact tracing or upgrade it again and it will understand -- checking in and out to see if you're taking too much time at lunch. That's something Casey is looking into for me. But there's an infinite number of applications being developed and what is happening to the software community for all these different markets is that there's just an explosion of interesting applications and things that none of us have ever thought of that are being added. It reminds me of when the cell phone industry added broadband data, it went from the 2G to 3G era and all kinds of interesting applications emerged.

And you guys announced this last quarter a thermal camera with Motorola, like that seems like it could be -- you mentioned contact tracing. It seems like that could be kind of important given what we're dealing with COVID?

Yes, that's a real nice market for us, measure the market in terms of digit -- millions of units versus the global security camera market, both professional and smart home is measured. It's more than 200 million. But it's a nice market. We do have multiple customers there. What I would highlight about the thermal cameras is that to do the thermal part of it, you need a different sensor than a traditional CMOS image sensor used for visible light. And so in these applications where you've got the visible light camera and then a thermal sensor sitting next to it, what we're doing is the fusion of 2 sets of data coming from the environment, the thermal information and then the visible light information and then the CVflow AI processor in addition to doing the sensor specific signal processing, it's also fusing the 2 sets of data together. And so that's like a new layer of processing for Ambarella in addition to the sensor-specific signal processing that we're so well known for.

Great. And then on the consumer surveillance side, the evolution of computer vision is kind of just getting started there, but it seems like some of the hot products for Christmas continue to be things like video doorbells and stuff like that. So you guys have a pretty nice presence in consumer surveillance. When do you think we might see computer vision becoming more of a factor in that part of the business?

Yes. I mean, there's already some computer vision. As Louis mentioned earlier, all computer vision is not created equally. The AI technology, what you can do with a very basic processor is very different than what you can do with our chip. But having said that, we are starting to see -- as you mentioned, we've had some announced, some have not been announced. We're engaged with vision and CV with most of the major players, whether it be Nest or Ring or Vivint or alarm.com or Comcast or whoever it might be, where we have a pretty broad base there with our vision products. And I think you can expect to see that start to move to CV as that market starts to come on. And we, as I mentioned earlier, predicted that would be second half of this year maturing into next year. And even with COVID and some of the other things going on, I think you're going to continue to see that grow. And certainly, those 2 markets will be the primary driver for next year.