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

Good afternoon, everyone, and welcome to day 2 at the Stifel 2021 Virtual Cross Sector Insights Conference and the Ambarella session. My name is Tore Svanberg, and I'm a senior analyst, covering analog, connectivity and processor semiconductors at Stifel. It is my great pleasure to introduce Ambarella. With us from the management team, we have Casey Eichler, who is the company's Chief Financial Officer; and also Louis Gerhardy, who is Head of Corporate Development and Investor Relations. The format for this particular session is a brief presentation by management and then a fireside chat. But we do invite you to ask questions as we go along, and you can type those into the chat function, and I will certainly address them as they come in. With that, let's just jump straight into it, and I want to thank Casey and Louis again for participating, and welcome.

Thank you very much, Tore. We appreciate being included. It's been a great day so far. And I'll ask Louis to kind of briefly walk through a high level so that give some background to those who might not be as familiar to the story, and then we'll quickly get into Q&A and other questions. And again, welcome and appreciate it.

Yes. Thank you, Tore. I'll just give some 40,000-foot perspective and save a vast majority of time for questions. Ambarella was founded in 2004, so 16 years of experience, and that time has been spent primarily focused on video processors. In other words, capturing an image for human eyes to digest. And in the last 5 years, the company has invested more than $500 million now into the development of computer vision technology. And what that means is this is a representation, an illustrative diagram of our chip. The green would be the video processor where we have an established reputation for state-of-the-art video processing technology. What's new is this deep neural net AI processor that when that's integrated into the same chip as the video processor, it's our computer vision chip that specifically targets the edge-endpoints of a network. And this is differentiated from other chips in the market that primarily target computer vision and AI in the enterprise edge or in the data center. Our chips, to operate effectively in the IoT endpoint, require a fundamentally different chip architecture. As we move to computer vision and incorporate this deep neural net AI processor into our silicon, it not only creates a new product cycle in our existing markets, which today are 60% surveillance cameras, 15% to 20% auto and the balance other markets, and in those new product cycles, our ASP is about 2x on a like-for-like basis of what it used to be. But in addition to new product cycle in our existing markets, these computer vision chips enable us to reach into many new markets that are not human viewing in nature. In other words, this video processor role in the chip moves from collecting an image to display for human eyes, but in this case, the video processor is collecting data as accurately as possible, and it's feeding the AI processor with this data, which allows machines to either partially or fully autonomously make decisions based on the software stack that sits on top of our chip. And so this transition is well underway, but still in the early innings. Last year, this computer vision revenue was a little more than 10% of our revenue or maybe 4% to 5% of units. And this year, we've said that CV revenue will be at least 25% of our revenue for fiscal year '22. And just to give you some evidence of how this is playing out, we see -- every quarter, we provide some indication of design activity for our chips. And you can see that of the 11 -- sorry, 12 representative designs this last quarter, 11 of them were these new CV products that carry the higher ASP for us, and they're also enabling us to reach into new markets. I'll just wrap up with a discussion of our SAM and how it's changing. Five years ago, Tore, we would have talked about our SAM being less than $1 billion, but where we are now is with the inclusion of the computer vision chips in the numbers, you can see that our SAM is growing and it's significantly higher than our revenue today. Also, I would point out here, auto becomes our largest opportunity because of all the processing that's needed in vehicles and some of the other non-auto IoT markets also have very attractive opportunities for us today. Tore, I'll stop there, and we can go into questions.

Right. Thank you. Yes. And as a reminder, if the audience has questions, you can type it into the chat function. So let me start with the differentiation for both Casey and Louis. I still think there's a lot of confusion out there in the investment community exactly what's so different about CV versus some of the competing solutions out there. And I think the one really notable aspect, of course, is this ability to program and configurability and using other software. So can you just reiterate again how different Ambarella solution is vis-à-vis some of the main competitors out there? And if you want to name them, that's fine, too.

Yes. So -- yes, go ahead, Casey.

No, no, that's fine. Go ahead.

So first, I'd point out that oftentimes when we're asked about competitors, competitors in the top 2 tiers are compared with us when the reality is they may be serving different parts of the market. Again, our focus is on the IoT edge-endpoints and, second of all, we're specifically doing deep learning. And when you apply those 2 filters, I find it takes 90% to 95% of the competitor folks want to talk about out of the equation. When you do narrow down the competitive landscape to the companies we do feel we compete with, how are we differentiated was your question. And the answer is, we've got our unfair share of experienced video engineers. There's simply a shortage of them out there, and that's point one. For example, most of the executive team work together over the last 20-plus years, including going all the way back to CQ Microsystems when the video industry was moving from the analog format to the digital format, and much of that team is still together today in Ambarella. Another major way that we're differentiated is how do we apply those engineering resources, and this is really important. Our approach to the market has been significantly different. And what that means is that rather than, okay, we're going to do a CV chip, let's start laying out transistors with general purpose blocks of IP, Ambarella took a different approach where they acquired state-of-the-art computer vision software stack VisLab in 2015 and spent several years understanding the algorithms and how to most efficiently execute the computer vision algorithms in the VisLab stack. And so what that yielded was a very optimized computer vision chip for IoT edge-endpoint applications, leveraging the video processor that the company has a very strong reputation for and adding this deep neural net processor that is highly optimized for treating video as a unique type of data. And so that is how we're differentiated in our approach. And the reason we're winning today and we expect to continue to win would be really threefold. One, efficiency, and we measure our efficiency in terms of performance per watt or performance per dollar. The second reason would be, for our CV SoC you see on this slide, we don't provide the application software that sits on top of it. We provide an open platform, which allows our customers and third parties to develop differentiated software that then takes advantage of these resources on our chip. And then the third reason that's becoming very apparent right now and we highlighted on our last earnings call is the flexibility of our chip platform, which is allowing customers to develop differentiated solutions. And flexibility is represented here by the word combination. And what we're able to do is, one, say, CV25, in this case, is supporting 3 camera systems: driver monitoring, occupant monitoring and recorders. So in other words, there's 3 cameras, there's 3 CMOS image sensors sitting in front of our chip. They're all feeding into CV25, and CV25 is simultaneously running multiple AI applications for each of these different cameras. And so that type of flexibility and being able to run so many different things simultaneously really requires the efficient architecture, the low -- the attractive performance per watt characteristics of our chip. And so we're finding this is the third reason beyond efficiency, beyond open platform flexibility and the ability of our customers to do these combo-type products.

I think -- one other thing, Louis, that is also interesting is kind of the algorithm-first approach, which the VisLab team was really able to help leverage not only in automotive but also, more broadly, in just developing the chip that you referenced earlier.

Yes. That was very well formulated. Second more general question, and this is something I've been asking a lot of the companies at the conference this week, which is, what are 2 or the 3 things that are different for Ambarella now versus before the pandemic? And you could pick -- I'm sure there's a lot of them, so you could pick whether it's customer behavior, inventory management, your own culture? Or are you seeing some accelerated or reaccelerated demand because of the pandemic? So yes, you pick the 2 or 3.

Gosh. Well, I'd say, first and foremost, we've always been very focused on our employees, our investors and our customers. But in the pandemic, we've spent a large amount of time making sure that our employees were safe. And as we talked about a minute ago, we have VisLab, which is in a beautiful place, Parma, Italy, but in a place that was very challenged early on during the pandemic. We also have people, as you can see here, all around the world. And Fermi and the team have been very, very focused on making sure that we continue to be able to move things forward, but we do it in a way that's safe for employees and, thankfully, we've had pretty good success with that. And I would say, that certainly is one of the things that I would walk back from. The second thing, I guess, would be the natural things you don't think about much, like, the fact that we're doing this online versus at a conference or that our employees -- a lot of costs came out of the business from travel, T&E and all of that, and that will come back in, we all hope. It could come in a little differently, but it will come all back in. And just making those adjustments, not exactly spending adjustments, but what you're not spending, I think that was probably a second thing of focus. The third thing, and we've all experienced this in a different way, but while there were challenges, without question, it was somewhere between breathtaking and surprising at how well our teams participated globally and just continuing to keep the trains running. I've been closing the books now with nobody in the office, including the auditors, for a while. And that really has gone much better than, I think, one would expect it. We've taped out chips. We're about ready to start sampling a chip, and that's moved on much more seamlessly than people would have thought. And so just the ability for us to continue to keep the business moving forward in a very important time, I think, was probably the third thing that would jump to mind to me. Those are things that as we go back to the normal or new normal or whatever you want to describe it as, I think we'll have learned some things and maybe make some adjustments. And then I think we'll go back to business as usual in a lot of other ways. So while we've enjoyed all of the Zoom conferences in that, we -- I know I can speak for both Louis and I, we look forward to being able to do this in person again because so much of it is the connectivity of really being able to sit down with people, both in the meetings and then kind of in the casual settings, and being able to have broader discussions. And that's something, I think, we're going to look forward to as things move forward.

Yes. No, thanks for that. And I think if you talk to investors in general, I think the answer is they don't want to do another Zoom conference. I think they're pretty sick and tired of them. Maybe a little bit related to sort of the environment from the pandemic. And obviously, capacity is really a big topic right now. I'm just curious, are your customers giving you any input at all on where they prefer for you to partner with? I mean, I know Samsung is obviously your main foundry partner. But do they really care who your foundry partner is and where you do your manufacturing? Has that rhetoric changed at all or no?

It really hasn't. We've been with Samsung. We're their second foundry partner behind Apple. They've been very good partners of ours. And we're taping out in the leading edge geometries, and they've been very, very supportive to us. I think it might have been a bit of a different story. First of all, TSMC, as an example, is struggling, I think, more than Samsung is for a variety of reasons. And the company of our size I'm not sure they get the same attention as we get at Samsung. So that's been critical for us. We do all of our CV chips in Korea today. We do some chips in -- the video chips in Texas, and we have had a deal through that, like everybody has, not just Samsung, with everybody. When I look at our ability to meet our customers' needs, our operations team really were very effective in moving very quickly. We don't carry a lot of finished goods, but we did make sure that we had the wafer supply that we could quickly get into work -- to WIP. And then as soon as we got them done, we send them out. So it's given us the ability without much exposure to really be able to meet our customers' needs. Now they have a lot of other parts they're trying to get as well. And so sometimes, thankfully, we haven't been the long pole in the tent, as they say. Typically, we've been able to be supportive. Everybody is, I'm sure, building a little inventory so you have to figure out how much is demand for current product and how much they're trying to build inventory. In certain situations, if it's something they feel they need, we'll test that by asking if they want to pay for a hot lot, which is more expensive. And if they don't, that means they were probably building for inventory. And if they do, they have the other parts they need to get to market. And so we've done pretty well there. And then the final note I would make, which is up here on this slide, is a majority of our employees are in Taiwan and in China. So we have all of our feet on the street right where the action is happening, and that helps you with the relationships that you have with those people, and it helps you with the reaction time. And so like I say, hats off to our operations team, Zemo and everybody else involved. They've done a terrific job in helping manage through a difficult system. We -- or I should say, Fermi mentioned that he felt this was probably the low quarter as far as the challenge to getting things through. Not that there isn't going to continue to be challenges for second half of the year, but I think we're all optimistic that we'll start to see some improvement and then we'll see what we get when we get into next year. But for right now, I think we've been in pretty good shape, been in front of it, and I've been very pleased with the way the team has reacted.

Got it. Got it. So moving to CV and, as you mentioned, you're expecting it to be 25% of revenues this year. And sometimes, you give some reference for which markets are contributing. But when we think about Wave 1, Wave 2 and Wave 3 and maybe you'll have to clarify that to the audience, how much of that CV revenue do you expect to come from those 3 waves this year?

Yes. I'll let Louis get involved and answer that because we did give some color on that, I think, in our last call. And so we can give you some thoughts around that.

Yes. So we've articulated 3 waves for our CV revenue. Wave 1 in the professional security camera market, we said, would be material in calendar year '20. And that has occurred, and momentum in that market continues to be strong with a customer base that's growing broader and more and more new programs ramping. So the trends in professional and the overall security camera market are represented on this slide. And you can see the green sliver, both in terms of the installed base as well as the annual unit shipments. We feel that we're early in this, which again, it's not only a new product cycle with 2x ASP, but we think CV is also going to enable the installed base to continue to grow in areas that are not security camera related and that are not human viewing in nature. The second wave we talked about was in the home security camera market. And we said that would become material in calendar year '21. And we've announced so far this year 2 design wins, both with Amazon. And those programs are in production now. And we expect to have more customers in the home security camera market, adopt our CV products. And so needless to say, we're very comfortable that Wave 2 is going to happen as expected this year. Wave 3, which is the largest wave, is in the automotive market. And you can see from our serviceable market opportunity, or SAM, on this slide, this year, which is fiscal '22 for us, it's almost $2.2 billion of SAM. And our revenue in auto in the prior year was $35 million, $40 million range. So a very significant opportunity in auto. We've characterized auto CV revenue wave to become material in calendar year '22 or '23. And given the amount of wins and -- that are going into production now, I think, in general, most people see us in a very strong position to achieve the Wave 3 guidance.

And what do you expect the mix to be over time, Louis? Maybe over the next 3 to 5 years, will auto maybe be like half with security being the other half? I mean I know consumer is going to come into the mix, industrial is going to come into the mix. But yes, just some reference for the next few years?

Well, I would -- if you use calendar '25 as kind of a pivot point that around that period of time, auto, I think we'd all be disappointed if it's not 50% of revenue and that compares again with 15% to 20% last year. And of course, the expectation is the overall revenue number will be growing at the same time.

So we should use this SAM slide as a good proxy for how your CV revenue mix will evolve over time?

Very roughly speaking, yes. We have a lot of share to -- share gain opportunity in auto. And so we're very early in terms of proving that we can do it.

Right. Right. And maybe on the topic of auto, some of the investor questions that I get is on sort of Tier 1 OEM wins. And it's interesting because I got this question the other day, and I actually went back and looked at all your design wins that you've had in auto because I know you disclose those every quarter. And it was a pretty impressive list of customers. So what are some of the things that you could say to sort of ease those concerns that Ambarella is absolutely being designed into Tier 1s?

So to answer the question in 2 parts, one would be human viewing where you'll recognize some of the Tier 1 names, very well-established companies like Denso Ten and JVC and so on and so forth. And then there's some OEMs that are known to be using our chips here also. So in the human viewing market, which is mostly recorders, data loggers, many different kind of sub-applications here, including auto black boxes that's emerging, that's a very healthy market where we're established. And some of those Tier 1s and OEMs that are already using our parts are also doing these other chips. And so it's not like we're completely unfamiliar -- they're unfamiliar with us. In fact, you can see from this list, we're strong in China in these recorders, in Japan, in Korea. And so having that established presence for the human viewing products is helpful. And the question is transitioning them as well as a lot of other new prospective targets, both Tier 1s and OEMs, over to our computer vision chips. And I think based on the evidence we can give you in this auto industry, we're extremely limited in terms of announcing wins with the name brand, OEMs and Tier 1s. And so we've put this funnel together as a way to articulate what's happening in our automotive business. And we announced this funnel on November 23, and the date is as of Halloween last year, and we'll update this again later this year. But at that time, we announced a $600 million funnel covering 6 years, and 2/3 of this was business that was won and discounted, and the other $200 million of it was pipeline that was heavily discounted for a variety of factors. And so we'll be providing this information. So even if we're unable to announce a major OEM or Tier 1, you'll begin to see what's happening as an investor in the funnel.

Got it. And what are you seeing among some of the more sort of emerging auto companies as far as doing their own semiconductors, maybe even their own image processors? I mean I know Tesla is obviously not a new company per se, but when you think about that new generation of companies, are they more inclined to do their own solutions, including maybe their own ASICs? Or would they be willing to work with Ambarella?

As you know, the auto industry is quite fragmented, and that makes it difficult to generate an ROI if some of the largest manufacturers are doing just 10 million vehicles a year. I think of the few, very few that have decided to do this, it makes sense if it will give them a time-to-market advantage and they can charge more for being the first to market. But there's -- majority of the players, as you can imagine, are not first to market and are going to need to compete, not only with technology and price and differentiation, and that's where we think we're in a very strong position to help them by providing this open platform with all the flexibility we talked about earlier, to come out with differentiated solutions that are much different than what was used in, say, to the first generation of ADAS vehicles.

Great. And I know we're running out of time, and this is not one of my favorite topics, but I feel like I'm going to ask it anyway because investors are so focused on it. But Dahua and Hikvision are now combined, what, roughly 10% of your business. On your recent call, you talked about Hikvision probably not going to be a growth customer anytime soon just because HiSilicon has tons of inventory, so on and so forth. So maybe update us a little bit on the dynamics between those 2 companies. And my understanding is that, Dahua, for sure, is going to be a growth customer this year, especially with them migrating some products to CV.

Yes. No, that's -- you've answered a lot of it in the correct way in the sense that Dahua has brought out several products in CV. And when you look at Hike, Hike is over 30% owned by the Chinese government. They're very incentivized to use domestic product. They've been using a lot of HiSilicon in the past, which is part of Huawei. And as that became increasingly challenged, they built up a large amount of inventory, I think. Now HiSilicon can't get chips from TSMC. That could change. But for the meantime, I don't think there's any pressure on them to switch from that because of the inventory that they built up over time. We still provide vision-based solutions to them, less than we used to, but we still provide vision-based solutions to them. I would say a lot of the business that they were doing has been moved out across the customers you're seeing here. And you're also seeing some new names, and that's because the next 5 to 10 people that were in China going after this market, they didn't have the ability to build up inventory. And so they've come to us, and we've partnered with them to give both vision and CV products. And collectively, they can be meaningful as they mature. And some of them are coming to market today. And I think that's something that has been beneficial to us. But we still are a partner with everybody. We're trying to do the right thing. We're trying to work in the current environment. But that, I think, is probably a consistent look across those 2 customers, which are down from where they were, to your point; Hike, in particular. But yet, all of our other customers, the Axis and Avigilons and others of the world are probably picking up share because of concerns. And then we have the new customers in China. So net-net, that's a good business, a strong business and, I think, will continue to grow for us.

Excellent. Well, with that, we've run out of time. So Casey and Louis, thank you so much for attending the Stifel CSI conference this year. Hopefully, next year, we will all be in person. Really appreciate it. And for everyone online, thank you so much for attending the Ambarella session. Have a good rest of the afternoon. Thank you.

Thank you, Tore. Bye.

Bye.