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

My name is Matt Ramsay, the semiconductor analyst at Cowen. I see a lot of folks in the room are still in the same chairs as the last time I said that. So I'll reintroduce myself. But thank you for spending a little bit of time this morning, and obviously, thanks for coming to Cowen's TMT Conference. I'm really pleased to be joined by my good friend, longtime friend, Fermi Wang, who's the Founder and -- Co-Founder and CEO of Ambarella, one of the leading -- and they are the leading video processing company in the industry. And making their transition from a lot of markets into the Level 2 autonomous driving space over time. So thank you so much, Fermi, for spending some time. I know you guys just reported earnings the other day. So it's -- we appreciate you making the track east and spending some time with us.

Always. Thank you, Matt.

Yes. Just because earnings were just a couple of days ago, maybe you could spend a couple of minutes just kind of recapping for everyone? I know the quarter was a fairly clean one. The guidance had some challenges to it, with the operational side coming out of Shanghai. We've finished the conversation in the previous session here about, hopefully, things opening up in China in the supply chain. So if you could just give us a little bit of context around where we were with earnings the other day, and then we'll go into the longer-term conversation.

Right. So like you said, Q1 is really -- no surprises there. And Q2 guidance was down because we were talking about kitting issue for 2 quarters. And the China shutdown, particularly Shenzhen and Shanghai shutdown, I think, caused a major impact to our Q2 forecast. And in fact, our customers start pushing out their orders early April, which is really just the timing that Shanghai start shutting down. And we're continuing to see our customers want to push out their orders for Q2 starting April all the way to the end of May. So that's really the reason that we have to adjust our forecast in Q2, and that's why we report $78 million to $82 million of forecast for the July quarter. I think, like previous company report, that we also see similar things. One, our Shanghai GM show me a picture in here enjoying coffee in the Starbucks, which is the first time he can do for 70 days. That was a major breakthrough, and we're happy for him, and hopefully, that we can move forward with better news from there.

No, that's great to hear, especially, I know you guys have lots of employees over there that have been affected and so many of the companies in the semiconductor industry do. So it's great to start hearing about those stories proved that things were going in the wrong direction for quite a bit of time and things. So I guess just to get this whole supply team conversation out of the way. There's been 2 aspects of it. One is the amount of your customers' product that sourced from the port of Shanghai and the area around Shanghai and the impact there. And then the second part of it has been some of the supply challenges that have happened with your fab partner, Samsung, in Austin, on 14-nanometer. If you could maybe just walk us through both of those situations. You touched on China a bit, but just what's going on with Samsung as well would be helpful.

Well, last quarter, we reported that Samsung told us that we have -- they have a problem with a 14-nanometer process node, and they will be short supply to us. And we -- the situation didn't change. In fact, we still expect $5 million revenue impact to us, which is included in our forecast. I think, Q3, we expect a better supply situation from Samsung 14 nanometers. We believe that, in Q4, that situation will be resolved. But I really think that the kitting issue that we continue to see, in addition to just Shanghai and Shenzhen shutdown, the short supply of microcontroller chip, Wi-Fi chip and PMIC chip, the situation didn't change. In fact, we probably had more question around that because of Shanghai shutdown. So I think that we continue to monitor that progress. And hopefully, the recovery in Shanghai will take one variable away from this whole equation. But until we see better supply on those components, we continue going to be tight on the kitting issues.

No, that makes sense. One thing I wanted to ask about really quickly, so yourself and Brian and Louis and I did a number of investor meetings about what, 5 weeks ago in Chicago and a couple of other places. And there was a lot of conversation at that time around Samsung. And the message that I took away from those meetings was that the third quarter, the October quarter, you might be hopeful that things would get a little bit better on the supply in Austin. And it seems like your language is a little bit different about that now. Maybe you could talk what visibility has improved in the last, I don't know, 1.5 months around improved supply in Austin?

Right. Well, the -- last quarter, when we talked about this, we were expecting a similar amount, $5 million impact in Q3, too. But with the current forecast, I think we see some improvement now. I don't think the whole problem is being fixed. But I think that definitely, from a forecast point of view, we see some delivery improvement. And also, I believe the foundry relationship is always about contract and the relationship. So I think continue to -- managing the leverage of the strong relationship we had with Samsung Foundry from top to bottom is continue to be a priority for me.

One of the questions that I've gotten in the last, I don't know, 4 or 5 months is just -- and we've talked about it in-person a number of times, just the consideration to bring on another fab partner, just given this disruption issue. And while it's temporary, it still has to make you guys think a little bit. So I know in the process nodes that you're going forward on 7 and 5 and other nodes, like mass sets are crazy expensive. There's a lot of additional people you might have to bring on to support another fab partner. Like what are the puts and takes there? Or is it just not in consideration right now?

Well, it is in consideration, however, it's difficult consideration. We are setting 5-nanometer with Samsung. It's hard for me to believe that we can enable another 5-nanometer suppliers even we want to because it's so expensive to have a true fab partner for the same process node, particularly for a company like us. I think that moving forward, when we go to 4-nanometer or 3-nanometers, I think that it's our job to make sure that we cover our bases to understand all of the process technology out there. I think TSMC and Intel are the only possibility for us to consider. It's just going to be a difficult transition. But however, like you said, it's important for us to understand what the other foundries are offering and making sure we make the right decision.

No, that makes sense. So I just wanted to say real quick before we kind of make a transition and start talking about the product road map in the future. But I don't want to monopolize the conversation. So if there's folks in the audience that have questions, please, in the back.

Can you just review briefly percent of revenue coming from auto? And just in your guidance, what are your assumptions is that the auto market for the period in guidance? And then we see a weakness in consumer and all of market deteriorates, what's the sensitivity to the guidance in here?

Right. So automotive is roughly 25% of total revenue. And in our script, we talked about in this call, Q2, the impact is almost equal between AIoT and the automotive business from the reduction on our revenue forecast. So I think that's really reflecting that the impact of the Shanghai lockdown has equal impact to us on both AIoT market as well as the auto market. And also, in fact, this morning, I saw that U.S. automotive production dropped like 30% in May. That just -- I think it's probably in a similar impact that we are seeing in the market. So the issue is really, in my mind, what's the -- how fast they can recover from that. It's really about how fast the supply chain can recover itself on this Shanghai lockdown. And also, still, at the end, even with the results of Shanghai lockdown, how fast we can go back to normal supply of microcontroller. I think that's continued to be a very big question in automotive business.

My question is more around auto demand. [indiscernible] everyone understands that. So what I'm trying to understand is the guidance, the assumption around other demand. And while demand falls off, would that impact your guidance?

If that demand drop, absolutely. Today, in fact, in our script, we also said that outside China, we believe -- even today, I still believe that our customer can sell as many as they build, right? So the demand is not an issue, it's really that they cannot find enough components right now. So I don't see that automotive demand is an issue, it's really supply is the issue.

So the guidance assumes all the demand remains constant?

Remain. Yes.

We're going to make a little bit of a transition, I think, from supply and foundry to new products and -- it sounds like you've gotten first silicon back from CV3 within what a week or so ago?

7 days ago.

And I mean, I've been following the company with yourself in the last for a long, long time. And this is the, by far and away, the most capable product you guys have ever had. And maybe you could tell the audience and sort of remind folks about the product. What the big jump is to CV3, the new capabilities and then what's gone on in the last 7 days? How is the bring up gone? And what's the first impression of the silicon coming back?

Right. So first of all, CV3 is probably the largest and most complicated chip we ever built inside Ambarella. Although, in our history, we built some other major big chips in different companies. But this is a $10 billion transistor chip, a 400 millimeter square die. It's definitely any time when you deal with this kind of chip, design was a problem, bring up was a problem. But in the last 7 days, we brought up all the major functions successfully. We are now bringing up the -- try to integrate all the sulfur layers onto the silicon. And you can imagine that there's a kind of bottleneck you've gone through this part before that it will take probably a few weeks to go through all of the trouble. And then I hope that in a few weeks, we're going to be able to demo. The most important thing is the new network performance and the power efficiency that we promised to our customer. And that's 2 things we have to demo quickly. And also in the next few months, we want to be able to show VisLab software and Oculii software running on CV3. And then eventually, we show EVA car driven by CV3 and all software stack, complete, Ambarella software stack in the near future. So I think this is a progression we designed for ourselves. And hopefully, we can quickly show demo to our key customers on that.

So the other night -- I mean, so much has happened in the last 2 days. I guess, it was 2 nights ago on your earnings call, I asked you about sort of sampling CV3 and the time line and the variables to get there. You had mentioned the intention in the summer to sample to, I think, more than 10 customers. The follow-up question that I've got a few times in the last couple of days is just like the breakdown of those customers. Are they all auto OEMs? Geograph -- are they geographically diverse? Is it Tier 1s or just OEMs? Just any kind of color around the folks that are queued up waiting for these samples.

It's OEMs and Tier 1s, obviously, because we still believe that -- although OEM is making a lot of decisions for Level 2, Level 3 cars, but, at the end, it's still a new Tier 1 partner. And in fact, working with Tier 1s continue to be very important for us. And we have a people in Europe, U.S., China, Japan, Korea waiting for this, right? So we are definitely trying to sample to multiple customers that we continue to talk to. And -- but I think they all see our CV3 presentation for several months now, right? So it's not -- talking today is not important, it's really bring this system to improve the performance number and the power number that we promised, that's probably the most important milestone for us.

That makes sense. So if you just think about where CV3 is and when you're giving your CV3 presentation to all these customers or potential customers, what are you guys pitching and what are you pitching versus the competition? I know there's capabilities of sensor fusion around radar and LiDAR and camera and other data, there's the VisLab software and the ability to drive the car, there's power, there's differentiation on quality of video input. I'm just wondering if you sort of rank order the differentiation points of CV3 in that system that you're trying to sell to customers that they're going to look to validate. What's the -- what's your bullet list of picture products?

Yes. First of all, I think most importantly is our computer vision performance need to be multiplex of our competitors. And with that performance improvement, we also promised much lower power consumption. We're not talking about 10, 20x -- 10%, 20% lower, we're talking about 50%, 70% lower. So that's the number. In fact, anybody you saw that number, they say -- their probably first reaction is prove it to me. So I think that's the -- we took in our credibility for our customers when we show up with that. The second thing we are showing is with all the integration we did, we believe there's a major bank cost saving. Because of power efficiency, you can save enough battery cost or cooling systems or cost. And if you look at the Oculii software, in fact, that's about the third thing, in fact that we also promised a much better sensor fusion solution because we can do camera and the radar sensor fusion at a low data level. I think a lot of people are talking about they can do camera and radar and the LiDAR sensor fusion today because they are doing it up to this level. But if you -- any time you want to try to go to raw data level, that means you have to bring raw data, 10 camera video raw data and 5 to 6 radar raw data at high resolution into that, and the performance you required to process it and the power consumption, you can sustain with that kind of processing is a major challenge. And we are promising that not only we can take in all the data to do this low level, but also we will be able to show a much better sensor fusion result because of the low level of sensor fusion. And that's probably, I would say, the top 3 differentiation we talk about.

Maybe you could just share -- you and I talked about it last month when we saw each other about the antidote around battery cost savings and power savings from the chip.

Yes. So in fact, that we compare -- it's not only just chip by itself, usually, when you give to your customer, you build a sum, basically a chip with memory and other PMIC chip so that you can be selling small module. If you look at the module power versus our one major competitor, we think we are at a higher performance of our power number is 140 watts lower. And if you translate to 2 things: one is what's the extra battery cost saving for that 140 watts? I think that's anywhere between $250 to $300. And on top of that, we don't need a water cooling system. So that is extra weight and extra cost saving. So I think that's just true on the [ bond ] savings side, I think that's important for our customers.

That's -- I guess the question then on the flip side, there's a lot of exciting stuff coming with CV3. What has been so far the hindrance of getting more business in auto, right? Like what are the things that the customer -- when you go into these meetings, and your team starts to pit stuff what other than give me the silicon improve what you say, that's one angle. But like as a relatively small company that's breaking into the automotive industry, what are the challenges? And what are you guys doing to address some of them?

I think seeing to believing is probably the most difficult problem, but I think we're addressing that. I think the second issue that we need to continue to develop is that how to support key customers, right? Support means we -- in fact, we promised that we're going to help them to put their software to our platform. That engineering support, not only just helping them put the computer vision software but all the other software from a single processing point of view to the high-level C software that converting the whole software platform to our platform. That's a major support issue, and that's something we need to build up. And that's the second issue. The third issue is really all of OEM and Tier 1 is focusing on quality. We have been shipping automotive-grade products to a lot of the customers, but some of the customers who haven't been using our chip, we need to prove to them that our quality level can meet their standards. I think those are the things we're talking to our customer.

Got it. That makes sense. Please.

Starting just on the [indiscernible] opportunity in each size. How big that revenue...?

Well, we talk about 10. There are 2 numbers we discussed in the past. I think that's probably a $4 billion TAM every year for us. And also, we talked about a different number we call sales funnel, right? In November last year, we disclosed that our automotive sales funnel for the next 6 years is $1.8 billion. A portion of that is CV3 potential design wins. So I think that is another matrix we showed the potential. Of course, that we hope that by winning more CV3 design wins the sales funnel number will continue to grow.

Sorry, just on the sales funnel number, how is the right way to think about how like actually convert to revenue?

Yes. So there are 2 portions of it in the sales funnel. One is [ guarantee ] design, we already confirmed the design win, that's $700 million. And there's $1.1 billion of potential design win that we're working on RFQs. And in that business, we have a multiplier in terms of our confidence with win and our confidence about the size of the business. So there's multiple levels of multiplying there to charge down the potential size of the business. So that's 2 levels of controls.

I guess just following on with that, Fermi, a portion of the funnel is CV3-based. And as you say, if you start to win a couple of designs and announce them, then hopefully the pipeline will continue to expand. But the pipeline it is CV3-based. Like if you sample the product in a month or 2, like what's the time line we should think about starting to convert some of the CV3 pipeline into one business? I'm just trying to -- and it's a different cycle than some other end markets. Ambarella has been in, in the past with sports cameras and security cameras and other things where there may be a shorter design cycle. What's realistic in automotive?

Right. So for CV3 is -- CV3 is not cheap for me. CV3 is a family of chips for me, meaning that for all our OEM/Tier 1 customer, they need multiple chip, too, because they have a high-end -- middle and the low-end cars from Level 2 to Level 4. And you have to have a family of chips to serve that. So that we are -- the first chip we did -- a sample very soon is to help them to start doing software development. And then we're going to build -- start building up a chip. In fact, if you study the script we talked about just 2 days ago, we did mention that we are working on our next-generation auto-grade chip, which is an ASO chip, designed for the CV3 family. And so we believe that the design win activity was -- for any design win that's secured in the next months will be production was talking about probably '25, '26 time frame.

So that's the time frame to revenue. I think I'm trying to understand the time frame to just converting business that you're chasing to one business within the funnel? Can that happen in 6 months? Is it a year? Is it...

I hope that is in that range.

Okay. Anything else on the automotive business from the audience? I know that it's now 25% of revenue. So there's a big part of the rest of the business that we need to talk about. But if anybody has anything else on automotive. If not, maybe we could talk about the security end markets a little bit. That's been a period of transition where you had a concentration of business in China that's now gone way down, and you're growing your CV-based camera in security and -- enterprise and home security internationally. Maybe you could give us an update there about how the markets are trending in those businesses? And have you felt -- I guess you felt the same level Shanghai and China disruptions in those businesses that you felt in auto?

Yes. So first of all, we call the AIoT market, what -- explain and why. But big in the AIoT market, security camera is still a big portion of our business. Enterprise security cameras, we are doing quite well outside China. In fact, like you said, if you look at 3 years ago, our enterprise security camera probably more than 40% come from China. This year, it could probably go down to 2% just that's because the geopolitical situation that happened in the last 3 years. But outside China, we're doing quite well. In fact, I would say almost all the major enterprise security camera system company, they're using partially all over their product using our solution today. For example, Motorola, [ XS ], Bosch [indiscernible] and Ipro in Japan, they are all using our solution for the enterprise. And in fact, our growth last year was because of the heavy growth in that area. CV continued to play a major role in the market because last year, our total CV revenue is 25% last year. This year 45%. A big portion of that growth will come from enterprise security camera also. On the consumer, security camera continue to do well. In fact, we just announced our second major customer, [ Vivint ], who is -- they are going to use our CV25 for the next-generation CV-based home security camera. And they will only ramp up the production, and they just made an announcement. So we believe that market will continue to grow that way. The third way for the CV3 is really on the -- will talk about that. But I think outside China, I think the enterprise and the consumer security camera continue drive the business for us. But we also mentioned, with the same hardware of security camera by loading different software, different new network will start enabling new market, new AIoT market, access control, the automation market. In fact, the slow motion robot for the warehouse, all those applications, if you look at the requirement for those applications, they need sensor fusion, camera, radar sensor fusion -- some they need LiDAR sensor fusion. They need to have a powerful CPU to do past controlling and the decision-making. They need to control the arms, they control the motors to moving around. It sounds for me is just another type of automotive products. So you can imagine that from the product spec point of view, those new application requirements is very similar to the requirement of automotive. And we believe that CV2 family moving to the low-end CV3 family will be a great solution for the market, and that's why we continue to spend a lot of business development resources to develop these new opportunities.

I guess there's a lot of opportunity there and also no secret that your, I'd say, primary competitor that was pushing forward with a CV road map on the chip side, HiSilicon and Huawei is -- has been taken out of the market. I just wonder what -- how far are we in that transition, right? They -- a lot of these security camera companies built up a ton of HiSilicon inventory, way more than I would have. I knew it was a lot, but it's maybe more than a lot. I just wonder how far are you along in the transition of converting the business that HiSilicon have to leave behind both in China and externally? And then second, are you seeing new competitors pop up in China? Like how -- are they small? Are they capable? Is it just R&D groups out of HiSilicon that have renamed themselves into something else? I'm just trying to -- and how are they -- if they've done that, how are they getting supply? Is it from [ Snap ]? Is it some from somewhere else? I'm just getting an idea of the -- that was a pretty big disruption with HiSilicon getting taken out. Like what are the ramifications of that for you?

We definitely believe that some of the success we're seeing outside of China is because of HiSilicon is not part of the equation anymore. So that shows on our -- that reflects on our success in the last 2 years. However, inside China, is a totally different story. I believe the HiSilicon team went to different companies. And I also believe some of the technology probably went to a different Chinese company, too. We're starting seeing pin-compatible HiSilicon solution popping up in the market. I just don't see how that possible if it's not, there are some business have been happened. So although it's all speculation, but that also shows me that in China, I think HiSilicon still, one way or another, a supplier to the Chinese market. And there are also new market -- new competitors inside China because I think the Chinese government want to make sure that there's -- they want to make sure that the Chinese components can supply to this market. So I think, in the future, we probably need to really separate this to market, to look at the possible opportunity for us. For security camera, like I said, China business for us is less than 2% total revenue, which I don't believe that will grow in the future.

So it's going to be all about an international growth story.

International growth story. Yes.

Got it. I think we just have a couple of minutes left, and I always like to pick your brain on things. Just given all the supply chain things that are going on, and then all of the excitement around CV3. Next 6 months, where are you spending your time? Are you dealing with the supply chain things and on customer escalation calls and all that? Or are you -- and lessen the team actively part of going to customers trying to get CV3 samples, getting through the process development? I mean, where are you spending your time?

Everything you said is part of my job. But in fact -- but however, what I define my job is I know what we need to do for CV3. I know that we need to take care of the supply chain problem. But the most important job for me is to make sure the company continue to have -- know what's our long-term goal is. I think there are a lot of time -- I told myself not to be distracted by the short-term problem. And the most important thing for me is, now we have CV3, how we going to -- who's the customer? How are we going to improve the CV3 from there? What's the product line we need to develop for CV3? I think this long-term question is probably will take more time than ever because I always believe when you are in a crisis, that's the best time to make sure you have developed your road map because when everybody is trying to cover their short-term problems, and if you can continue to invest and do the right thing for the company, when this important crisis finished, you probably come better than competitors. I hope that's where I can spend most of my time. And I hope that won't be distracted by too much by those short-term problems.

Any other things that you wanted to get across here? We covered a lot of ground in different directions from supply chain to China to road map to security. So anything you want to leave the audience with? I think we just got a couple of seconds here.

Yes. So I really think that -- one thing we didn't spend a lot of time on is Oculii integration. And we are extremely happy about the technology we require and the people we work with. And I think that's a very successful transition. And more importantly, we -- now we are convinced that we can fully integrate and take advantage of Oculii algorithm into our CV3 family. And like I said, we can do a level sensor fusion, but more importantly, I think that the levels of quality that we can do and also because of the fundamental technology that really give us another extra layers of differentiation and also another layer of cost [ sunk ] cost savings, I think that will significant for us. And hopefully, in the near future, we'll be able to demo it.

No, I think that's interesting. One of the things that I appreciated when you bought Oculii is that you were going to take their software for radar and run it on the ARM cores on your current products, and I thought Okay, well, CV3 is kind of fully baked. It might take till the next product where you can really do it. But if I'm not mistaken, you're able on CV3 to actually do a lot of things in hardware that accelerate their stack?

You know less, right? He won't wait. So -- and we -- we found a way to make it happen. And of course, that the next chip -- next CV3 derivative will be even better, but we won't be able to demo CV3 with Oculii software.

Awesome. Well, thank you so much for the time for the audience here. Fermi, it's always great to see you.

Thank you.

Thank you for spending a little time with Cowen, and all the best.

Thank you. Thank you, Matt.