Ambarella, Inc.
Q1 2019 Earnings Call Transcript

Published: 3 Dec 2019

[Joe Moore:]

All right, good morning everybody. So, I'm Joe More, very happy to have with us today from Ambarella, CEO, Fermi Wang; and CFO, Casey Eichler. Ambarella is actually our favorite small-cap stock. And company has been going through very interesting transition. Terrible, I guess, way to put it, but very positive on some of the developments in computer vision. And we think computer vision is sort of one of the most important kind of new developments in semiconductors and Ambarella has emerged with leadership position in silicon.So, maybe we could just start with that a little bit, if you could just kind of walk us through this wasn't a reaction to shortfalls in the other businesses it was an investment that you guys have made that predating the VisLab acquisitions. So can you just talk about the importance of computer vision when you started investing in it and what brought us to this point?

[Fermi Wang:]

In fact [indiscernible] had the IPO in 2012 and at that time we start thinking about what's the technology related to video products and one thing -- one common thing from the customer feedback is always that they would like to analyze video while they capture the video on the camera side. And so we look at it -- that’s called traditional computer vision algorithm. We just not convinced that’s the right way to approach to solve the problem, but also fortunately in 2012 to 2013 there's many new paper published around the new network competition for computer vision. And at that time we convinced ourselves that that's the right approach that's a direction we should spending money and resources on that.And the way we did it is just like what we did for our image processing and video compression in the past, we took the approach we called algorithm first. We tried to understand the algorithm will be implemented on the application like security camera, automotive or robotic. And we tried to understand the algorithm that required for those applications, and tried to afterward understand tried to find the best way to define that silicon architecture that optimize with die size performance and power consumption.And it took a long time for us. In fact we spent that we talked about $350 million in the past few years. And we spend several years to understand algorithm in the process we acquired VisLab to understand the software and algorithm on automotive side. And we start doing chip implementation, probably I'd say 2017 and we introduced chip 2018 and we're just at the face of start ramping up mass production in silicon with our customers in the professional security camera side.So although, it's a long development cycle, but I think this is really a cycle not only we convince ourselves we are in the right track on the technology side, but also we get enough feedback from the market side that there is a multiple high volume market that will require this computer vision technology. And that's why we're so excited about it.

[Joe Moore:]

And I think if you look at -- I mean people have been excited by the major investments that have helped like NVIDIA on the training side. Certainly, the largest workload that's been trained on GPUs has been these kind of [indiscernible] nets deep learning approach to solving vision problems. And if you look at the most important development so that your surveillance camera customers are dealing with or AI cities, it's all vision problems autonomous driving is all vision problems. So I think when people think about these markets they think about the programmable solutions that are the first stage of solving those problems whether it's graphics chips or FPGAs or microprocessors. What's the argument for dedicated silicon versus those types of applications?

[Fermi Wang:]

Well, in fact I would like to take case for video application particularly. When we started this company, let’s say 15 years ago when we start approaching the security camera market that market was dominated 100% by DSP approaches and quickly after we understand the requirements, algorithm requirements and the function requirements of those applications we can design a chip which is not only programmable, but also very efficient on the performance of power. And that's where we really show up with better performance at low costs compared to all those programmable architecture. And video such a business because just if you look at the video inputs of gigabit per second you have continued to serve it 24/7 for the security camera, for automotives when you are driving it you have continue serve that real-time. So if you don't have efficient architecture than how to scale performance, how to reduce the power consumption they all become an issue with programmable architecture.So in the last 15 years in many markets that we entered we always enter to compete with the programmable architecture first than when we show up with a solution that can compete not only in the performance power efficiency, but also at a lower costs which really helps us to address the need of market and the goal with -- or the growth of the market.

[Joe Moore:]

So where are we now from a product roadmap perspective you had first silicon in 2018, you have multiple product families out, it's pretty small portion of your revenues computer vision at this point to be clear, but a pretty significant part of your design win pipeline. Can you talk a little bit to just where we are productizing this?

[Fermi Wang:]

Yes, we look at three wave of CV designing. We already talked about the first wave of designing which is on professional security camera. We talk about 9 out of 10 major customers are with design with our chip and two of them already in production with us. And we believe all of that nine customers will be in production was sometime next year before next sometime next year and will start ramping the revenue, in fact, we already give the guidance that we believe our CV revenue on the security camera will be material, which is more than 10% next year and hopefully we’re going to continue ramp up from there.The second wave of the CV revenue will be on the consumer IP-cam side. Basically we believe that the consumer security camera will follow the trend of professional and the designing cycles like a year behind and hopefully that we can start talking about design wins sometime later next year.And the third wave is automotive, which is definitely the biggest market opportunity is ADAS or plus -- level 2 plus cars. And that's something we are talking about is third wave of CV opportunity for us.

[Joe Moore:]

Great. So maybe I want to dig into those markets a little bit more, but maybe before we do I know there's a lot of attention being focused on shorter-term trends, you've had two consecutive beaten raise quarters now after a series of challenges in the consumer side, your consumer exposure is pretty small and not really a headwind anymore. So the main headwind now I guess is trade. So just focus on that a little bit?So maybe start with, Casey, you made some comments on the call about pull-ins, you quantified the number I think so maybe just talk a little bit to that. Your two Chinese customers are pulling in why and how much of a headwind is that going forward?

[Casey Eichler:]

Yes, to your point for the last year or actually almost a year-and-a-half now there has been a lot of attention and a lot of disruption in the supply chain related to U.S. based products going into China. And two of our larger customers Dahua and Hikvision are customers that have been impacted by this as well. They are the largest providers of security cameras in the world and certainly China is the largest market.So as that's progressed over the course of this year we've seen an increase of orders that really are building their inventory. And so we tried to over the last quarter getting a sense from our fulfillment house as well as our customers directly to get a sense of what level that was at and how that impacted this fiscal year that ends in January and next fiscal year. And what we came out with is saying that we felt there was about $10 million of revenue that was pulled probably from next year into this year at this point. And that's because they want to have a carry an extra set of inventory into next year because it's an uncertain market today.So we wanted to make sure we got that out there some Huawei and others are holding much, much more inventory. And I think people are concerned about what the number was. So we just wanted to go out and that's an order of magnitude. It's not an exact number, we don't have an exact number from our customers, but I think give or take a small amount that's a pretty good sense today.

[Joe Moore:]

And what's the behavior now around that inventory, I mean, you -- both of those customers are obviously on the U.S. entity less you’ve also said that the majority of products you can still ship to them. So is there a comfort level now that they could start bringing those inventories down or does that happen next year. How do you think about that?

[Casey Eichler:]

That's a little unclear, what we said was we thought that probably that inventory would come down sometime over next year. But it's a little unclear and depending on the day some people feel more comfortable or less comfortable with where we are in the discussions today. So I don't anticipate that coming down anytime soon, but I think at some point, if we do get to a point where there's some sort of agreement to provide some comfort you might see that come down over the next year. But really that's going to be our customer's decision and we don't have much insight today, if we get more visibility of course we're going to communicate that.

[Joe Moore:]

Okay. So daily tweet changes…

[Casey Eichler:]

Just one tweet away.

[Fermi Wang:]

In fact, over the weekend probably just say they might change the rules to qualify. So I have to imagine that our customers probably have different concerns still today.

[Joe Moore:]

I guess from that perspective, I mean, what about the design pipeline with those customers as well, I mean you alluded to some risks around the call to the extent that you've seen design activity that's been put on hold because of all this, do you see any signs of resumption and is there should I think differently about their domestic business within China versus their exports?

[Casey Eichler:]

Before Fermi answers that I think I missed one part of your question, which was around CV. We did state that none of the CV products that we have are impacted by this for those two customers. And so for the majority of the revenue that we're going to be seeing going forward over the long-term, I think that is not impacted today. Now, again, as Fermi indicated that could change. But the CV products can continue to be shipped.

[Joe Moore:]

So it seems like I mean we talk to both of those customers and I mean the implementation of these computer vision AI features into their cameras is really central to everything that they're doing. So is there less of an opportunity because of these anxieties and what's the alternative for them, can they use HiSilicon?

[Fermi Wang:]

Right. So, obviously, we still continue to engage with our customers on this CV chip -- CV design wins. And of course HiSilicon continue to be our competitors with regard to this trade war situation. What’s different -- difficult for us is really that after this entity these things I think all of Chinese customers or all Chinese system companies trying to reduce their dependency on the U.S. components, which I think is probably the biggest threat to not only Ambarella but all the U.S. semiconductor companies, because China tried to reduce the dependence.And for Hikvision and Dahua I can say that continue to engage with our CV discussion with Dahua come out to say they are going to production our chip just said in their latest conference call. So I think that we still continue to engage but long-term uncertainty, these things that we worry about that if the pressure continues to pile up our customer going to consider replacing us with other non-Chinese -- non-U.S. component which is possibility I think.But I rethink that our biggest risk is on the low-end side, there is a lot of companies that can do chips like ours, but on the middle and high end side I think our main competition is HiSilicon at this point.

[Joe Moore:]

Okay. And your sense of HiSilicon’s competitiveness outside of China for export. And also obviously Hikvision, Dahua compete with Huawei. So is there a roadmap for them?

[Fermi Wang:]

Yes, first of all I really think that with this trade war also the HiSilicon which is part of Huawei, the chance that can sell their security chip outside China is become less. And also we do see more opportunities that using our silicon to replace HiSilicon’s design wins outside China, which is happening. And also inside China Huawei has become very verbal that they want to become the IP security system company competing with Hikvision Dahua so that put Hikvision Dahua in a very difficult situation because they are buying a lot of high silicon solutions, which is from their competitors. So it's really a very touchy issues with our customer and they have to make a decision whether U.S. is bigger threat or Huawei. So I don't think that's easy decision for them.

[Joe Moore:]

Okay, great. Well, that's helpful. I think maybe we could shift a little bit to thinking about some of the markets and customers. Surveillance is about 60% of your revenue on a run rate basis. And the majority that’s professional you talked about traction there. Can you talk about what that does to your business, what it does to average selling prices, what it does to market share and what it does to the growth of the markets?

[Fermi Wang:]

So, outside China I think we definitely see good growth in those professional security camera, many on the momentum of the market as well as that we get design win, extra design win from the -- to replace HiSilicon design wins. So, both of that we have -- for example we talked about Hanwha Techwin, which use HiSilicon solution in the past and they have transitioned to Ambarella solutions just in the last several months. We start seeing they're ramping up. So, we do see that the growth on that.On top of that I rethinking that CV transition is important transition for us. My feeling is at the right price range. Although the existing installed base of security camera will be replaced by CV security camera. We're talking about roughly 700 million units of security camera install base out there. So that could be a huge wave for CV, but have to assume that the price is right. And that's why we have -- there is a family of products from the high-end to low-end so that we can enable many different product lines for our customer at different price performance point.So, that's -- I really think that's a main growth area for us not only -- I think gross margin going to be similar, but the thing is not only that replacement cycle, but also our ASPs a lot higher than our video products. So that definitely accounted for a major growth area for us.

[Joe Moore:]

Yes. You said over two times.

[Fermi Wang:]

Two times ASP. Yes.

[Joe Moore:]

And, as you look at 2020, I think you’ve talked about ramping most of your customers over the course of next year any sense for the penetration over the course of the year that you'll get?

[Fermi Wang:]

Well, it's hard for us to predict because we have two customers just start ramping up. And ramping up at the pace that we think is very -- is reach to our expectation very good. But whether that apply to older customers that really depends on the market situation. But I really think at the end it's depends on the price point and how aggressive our customer want to deploy those CV based camera.So, we are still trying to be conservative on that, but we do give a guidance that we believe next year our CV revenue will be material total revenues.

[Casey Eichler:]

And second wave of the CV revenue next year is also in the security market on the consumer side. And so we'll start to see that in the second half as well which will start to amplify that CV revenue, as well.

[Joe Moore:]

Okay. Maybe talk a little bit about the consumer business. First, how do you see it this year? It seems like as we look at a lot of heavy Black Friday promotions we've seen a lot of -- your customers’ products on shelves and seems to be doing pretty well. What's the expectation around that there is still that consumer element of your business that is there a risk of an inventory overhang? And then what is the pace of CV adoption in those markets?

[Fermi Wang:]

First of all, I really think that we see a strong growth this year on the consumer side. There are a lot of design win opportunities for the video only solution and we start seeing that go into production and in fact that for example really shipping a lot of products not only just [indiscernible], but their flood light, their indoor outdoor camera. So I do see that we see strong growth in the U.S. market for this type of products.And moving forward we start opportunities -- CV opportunities, design win opportunities for our customers. And we do believe that we are in a position to compete with those design win and hopefully we can start talking about the design win and also revenue potentials starting next year.

[Joe Moore:]

Yes. I mean, I think everybody -- every investment that I talk to that owns these products sees the benefits of better triggering mechanisms for the warnings that you got. So yes, I don't get squirrels and deer setting them out.

[Fermi Wang:]

In fact, I think all of those kind of products has one compliant is really that although the older product has given you some warnings about events, but very soon you turn that all, because most of them are wrong messages. So I really think that we're putting a CV on the edge will solve majority of that progress and we think that with right algorithm especially on the detection on new network algorithms will dramatically reduce the false alarms on those systems will become a lot more appealing to the customers.

[Joe Moore:]

And longer term when you think about these types of cameras. Obviously, you think about traditional uses of surveillance, but we're seeing in AI cities there's a much wider use case. And then you see in areas like in the Amazon Go talked about their concept stores that are completely computer vision based. It seems like there's an awful lot of opportunity once you sort of put these capabilities into cameras to really expand the TAM and replace human capital on a broad scale with computer vision?

[Fermi Wang:]

It is a great point, I really think, the one thing we notice in the last several months, is a lot of our traditional security camera costs. They start seeing new opportunity outside surveillance application. For example, Amazon Go you mentioned that people putting security camera in there for totally different applications and by using different software and different network application. And I can give you probably many different examples, that you use a surveillance camera as a hardware by loading different software suddenly you enable different application, smart cities, smart home or for the centers or for the medical application, for smart traffic modeling. So suddenly because you can add different smart feature in there, the application that you can address was that hardware and software combination very, very high.So in fact internally we have been talking about how we address this to provide better guidance instead of just calling all of the surveillance cameras, how we have a different way to category our market segment, so we can give a clear explanation to our investors and customers how we are really doing.

[Joe Moore:]

Yes, I think that would be helpful. And then shifting to auto I mean, it's going to take a long time, particularly when you think about the highest profile sort of full autonomous applications, it's going to take a very long-term. But you’ve had a really impressive 2019 in terms of the pipeline of customer wins that you’ve announced and I know it can be hard to get customers to make those announcements and I know there are some that you haven't talked about as well. So you talk just maybe before we go into the specifics, just generally about your progress in autos, who are you competing with, what are the most important applications to be thinking about?

[Fermi Wang:]

Right. So in fact we have a several important applications, but consider limited time let’s focus on two things, one is ADAS, the other one is L2 plus opportunity. I think ADAS is a very well defined front camera facing forward single camera to help the car to do emergency brake or some limited functions. But so it’s a wealthy frank market and is very high volume already. So from that point of view we already announced chip that available to do that and also we announced partnership with HELLA Aglaia, our approach is different than mobile app which is our biggest competitor in this space. Mobileye is designing a black box solution which means they deliver everything, hardware, software and the -- they don’t even allow customer to do differentiation.Our approach is totally the opposite way, we have software, but we use software to prove that we have performance hardware side, but we open up the platform to help our customer to put an algorithm on to our platform. So one example which is HELLA Aglaia we work with them to put their ADAS algorithm on to our chip in less than two months' time. They can functional and start that more into the OEM Tier 1s and that’s just show you how powerful this platform can be if you have a running algorithm running software that you want to have a platform to do it. So that’s -- we continue to work with our algorithm partners to demo to the OEM Tier 1s on ADAS side.We also start seeing a huge opportunity on L2 Plus, there are many different definition let me put my definition out there which is in a car you are putting multiple camera, anywhere putting 5 to 12 cameras. And you probably have RADAR, and you probably have LIDAR but you do level 2 plus functions that means majority of time you are -- in fact all the time driver need to continue to pay attention to the road condition, but the system can probably help you to drive on the highway, avoid most of the problems.With that you need just considering that how many camera you need to deal with and the performance you need that’s really usually require much higher performance chip, you need to have SO level of chip and that’s definitely another area we think that we will start addressing and approaching customers to talk to them by their requirement and how we can deliver to the market.So this is definitely two major opportunity we're working on and by demoing our product right now. But in addition to that, we talk about e-mirrors and surround view and recorder business that’s still meaning business. For example we talk about fleet management design wins just in this conference call. They use ADAS, they use e-mirrors and use recorders in those product line, which will have much shorter revenue cycle we're talking about maybe 12 to 18 months, we can start talking about revenue for most of our fleet management design wins. But I think the most important thing for us is to win a design with some Tier 1s and OEMs.

[Joe Moore:]

I mean it seems that you’ll have pretty clear growth in your automotive business even though there's a long path to revenue in the CV oriented businesses because you've the OEM businesses growing in the near-term.

[Fermi Wang:]

Exactly. So right now 20% of our total revenue come from auto. We have to believe that any meaningful design win will increase our revenue potential on auto side. So I hope that in the long run, and I do believe that our auto revenue growth will give us auto at one day will become our largest revenue generator from the market point of view.

[Casey Eichler:]

And, Joe, as you've noted in the past, some of this can get to market a little sooner, like aftermarket. Obviously doesn't have to go through the OEM cycle. So that can get to market a little quicker. Things that don't do critical safety, like driver monitoring, cabin monitoring, those type of applications using our CV technology probably can get to market a little bit sooner. Obviously, the OEM announcements are the ones that are of most interest and have the biggest revenue stream. But, you do have that.

[Joe Moore:]

Yeah, I wanted to ask about that. I mean, you have several wins in cabin, in car monitoring. We talked to some of your customers in that space who have been very enthusiastic about Ambarella silicon in that and it's coming into the standards relatively quickly. So you are like as you said you'll be in revenue quicker. And then also, logistics as a new category you talked about on this call with Mercedes Benz and some of the work that I mean, you sort of think about monitoring packages coming in and out of a Federal Express truck. It's sort of less sexy, but potentially a pretty high revenue opportunity over time. Maybe talk to those things a little.

[Fermi Wang:]

Yes, I really think the design win we talk about is really helping people doing delivery, product deliveries. And in the Mercedes Benz, they basically putting two of our chips at the door. And so that every package going there automatically scan the barcode as well their address. So you don't need to do many scanning anytime.So as soon as it goes through that package goes there, they understand what the package is where they go for. And also they decide where to put the packages on inside the van. So when you reach your destination, the system will tell the driver where to pick up the box without them scrambling to search for the package and they can grab the box and when they remove from the van automatic check it out. You don't need to -- so and you can deliver.So this become a total automation system for the delivery van. And we think that's a huge opportunity. We talked about some customer told us about 3 million to 5 million delivery van in the worldwide. And our total we sell two to four chips per van in the system, which is our high end chip. So we talked about roughly anywhere between $100 to $200 revenue from per cars.

[Joe Moore:]

Pretty substantial. Yes. So I want to leave time for an audience question maybe. But maybe last one for me, just bigger picture, how are you thinking about 2020 earnings? Because I think people are -- and I am not asking for a number but how are you thinking about planning processes you deal with these various uncertainties?People have talked about how expensive the stock is on near term earnings. And my view has been the pipeline is sort of the thing to watch not necessarily the earnings levels. But how do you think if there are further setbacks with trade. How do you think about you're monitoring OpEx? And, whether you adjust your priorities as you think about planning for next year?

[Casey Eichler:]

Yes. When you say 2020, I assume you're talking about our fiscal year 2021. And when we look at it to your point, we are in a transition. And from an earnings perspective, we've invested and I think you’ve pointed this out. You can kind of separate the businesses and say, we've got the core technology in business. And that's a P&L. And then you've got a startup company, where we’ve put $350 million into CV technology today, without real any revenue of substance really coming. And that's all kind of in front of us. So if you strip those two apart, I think you see that the underlying business has actually performed pretty well. But we have made this investment. And that needs to pay off over the over the next couple of years.When we look at OpEx over the next year, I think you're going to see it grow most people are modeling at roughly 10%. That's less headcount for us today, but as we go to 7 nanometer and 5 nanometer, the cost of masks and tape outs, the cost of CAD tools, as you know all tend to increase. And so we'll continue to add headcount. But we're also going to have those costs and that's really the main driver in OpEx, SG&A and those functions really aren't going to grow that much in cost.But once we get out and get twice the ASP on some of these CB products, they start to become more substantial, while the gross margins are the same. You'll see the operating leverage come back into the model. And I know everyone is as anxious as we are to see that. In the meantime the thing that gives you confidence that that's going to happen to your point is the announcements that we've made. And they've been significant. We've done 15 plus announcements each quarter all across all markets. A lot of them are CB based. And I think that's what's giving investors the confidence that we really are in a path to be very differentiated in these markets.

[Joe Moore:]

Yeah. Make sense. See if we have an audience question.

[Unidentified Analyst:]

Thanks very much. It was pretty clear that’s a good question. You basically announce also that you’re working now with Cayerns, Panuck, Cofneck [ph] and so on. Now these guys, obviously the way they work they have huge engagement, lot of system integration what do you do with them actually, they actually white labeled the products you're getting involved in the selection of the system. Can you elaborate a bit more on that?

[Fermi Wang:]

So I think because with assistance supplier sorry -- semiconductor supplier, so what we do is try to engage to them and try to show them that our chip can help them to provide higher performance solutions to what they have today. So the engagement is really that we go there, demo our system to show that what kind of new network performance our other video processing technology we have and convincing that to kick off project with us. And then we support them to put their software onto our chip. And that's a progress on the business development and engineering side.So it's really about supporting system customer to using our chip which is a very traditional business model that we do for any other customer for example, the Daimler design win we talk about is a very similar approach.

[Unidentified Analyst:]

Are you shipping to these guys now, are you shipping to Cayerns are you shipping to Panuck?

[Casey Eichler:]

One of the things that they have to do is rewrite their software. And so as you know they bought a company in Switzerland, they bought a company in Korea. These are people who are taking and writing the software that we get deployed on our chip. So one of the things they have to transition is that then they need a platform to run those -- that software those neural nets. And that's really what the progression is going to be once they're able to port over to our chip they are not able to move forward.

[Fermi Wang:]

The answer to your, we haven't announced them as a customer yet.

[Joe Moore:]

So it’s right time, we'll wrap it up there. Thanks, guys.

[Fermi Wang:]

Thank you.

[Casey Eichler:]

Thank you.