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

Let's get started on this afternoon session. I'm Vivek Arya. I cover semiconductor, semi-cap equipment at BofA, and really delighted to have Fermi Wang from Ambarella, CEO of Ambarella join us this afternoon to share his thoughts about the company. I'll go through my Q&A, but please feel free to raise your hand if you want to bring up some of your questions in between. But -- welcome to you, Fermi. I appreciate you joining us. And maybe for people who might not be familiar, right, with Ambarella, if you could give us a quick overview of your kind of growth strategy and product pipeline that will be very useful.

Right. So today, if you look at our technology is really applying AI to video content and particularly focus on edge devices because our differentiation is performance per watt, which is critically important for any edge device. It doesn't matter whether it's a security camera or automotive for those applications that not only need a high-performance but need very limited power consumption. So that's where we are focused on in the last few -- several years, and our CV revenue -- our AI revenue go from 45% total revenue last year to close to 60%. So that's really our core business these days. And we're going to continue to focus on these 2 markets with particularly IoT, which is dominated by security camera, and the other side is our automotive business. The current revenue ratio before this inventory correction was roughly 3:1, 75% on the IoT side, 25% on the auto side. And like I said, the technology differentiation is the only -- the reason that we can continue to compete. In the [ post ] phase we're competing with companies like NVIDIA and Qualcomm and some other smaller companies on the low-end side.

Got it. Maybe we can start with a few kind of near term, get those out of the way and then really focus on the longer term, Fermi. So just in terms of the near term, what are you seeing in terms of demand for IoT and automotive products in the current quarter and as you look into the back half versus the assumptions you had at the start of the year?

Right. So first of all, the demand is still strong, especially on the IoT side. For example, I heard that Motorola was here just this morning. So their business based on the video product continue to grow, I think, 15% a year. And that's consistent with our expectation at the beginning of the year, so that didn't change. What's the difference is really that our revenue from definitely is we're seeing downturn compared to last year. That's basically because of inventory correction. And that gap between their demand and our shipment, that's where we see our revenue weakness. And we see the same -- similar effect among almost all of our IoT customers for auto, that kind of impacts much less but still there.

I see. Is there a way to quantify for me how much IoT inventory is out there in terms of the number of weeks or months or...

It's hard because every customer, for example, certain customer, they are willing to share with us their data, we know that. But we have a lot of customers out there. We have 20 meaningful major IoT customers, not all of them willing to share with us their inventory situation. So it's very hard to generalize it. So the only comment we gave before was we continue to watch our booking number. If -- when the booking number start growing again to a healthy level, then I would call it the sign of that -- the end of inventory correction, but we are not there yet. And other than that, we haven't really given other prediction, except that we believe our Q1 revenue is at the bottom based on the current inventory correction situation.

I see. And then in terms of automotive, how do you see the trends there? Do you have inventory in that end market also? Or how would you look at that specific market?

I think for that market, the inventory level is much lower than IoT. We do see some push-out demand requests from our customers. And -- but the size and the scale of the push-out is much lower. So I would say that autos haven't been impacted severely as IoT market with -- for the inventory correction.

I see. And then can you maybe remind us what are your key focus areas within the automotive today and where are they going in the future?

Well, there are a few areas we definitely focus, the biggest focus is how to monetize our CV3 investment. We -- right now, the 60% of AI revenue all from CV2. And we invest heavily on CV3 for several years already. So how to monetize that technology most critical time. So for automotive market, we made 2 major announcements. One is we announced with Conti that one of the Level 4 customers using our silicone end-to-end software that Conti and us codeveloped should go into get a design win. That's our first design we announced was our silicon software. So it's a really major milestone for us. The other one that we mentioned in our script this time is we believe our CV72 which is a CV3 derivative chip will become -- well, we are using CV72AQ to bid on several Chinese Level 2+ design wins and we believe that we have a good chance to win some of them. And most of those projects are on the schedule to be in production the end of next year. So the design cycle is 18 months versus 4 years in any other area. So that's where I really think that we will monetize our CV3 technology in China automotive market first. So that's 2 major area we're focusing on. In addition to monetize CV3, we also continue to make sure 2 things; one, is we want to make sure the software, including our VisLab software, [ AD ] software plus our Oculii radar solution will run on our CV3 chip and that we want to demo this year. And that will be the first time we demo a single-chip solution with our own silicon or software, and I think that's a meaningful milestone for us. And that's something we continue to develop. So that's another thing we should talk about for the automotive market.

Got it. And then we'll come back to automotive. On the IoT side, I think mentioned that you're undershipping demand by 25%, 30%, right or so. When do you see that kind of coming back, right? When do you see kind of sell and sell-through start to match? Is it possible this year? Or do you think that's more likely happen in '24?

I think it's possible this year, but we're not calling for it. We're not saying that when is that going to happen? Because I think every customer of us will have different timing. Some customers start the process to clean up their inventory earlier, they'll probably finish earlier. But there are some customers just started probably a quarter ago that will take little longer. So overall, it will be average out. So I think that -- I'm hoping that for this year, but we never give a guidance about what exactly has happened.

Of course. Now in that IoT, how much would you kind of call industrial versus consumer IoT? And the reason for me I asked the question is that sometimes you get like a seasonal rebound, right, in the second half or anything that is consumer related. So what is that industrial versus consumer mix? And is there a seasonal benefit that we might be able to see?

Right. So for the IoT space, the biggest portion is security camera, and our professional security camera versus consumer is roughly 2:1. So that's the ratio that we're looking at right now. And historically, that Q4 for consumer IP cam, there is definitely seasonality. But we are not sure whether that's happening because when you have inventory, the seasonality maybe people -- our customer will use the inventory too for the season if they have plenty of that. So we are waiting to see, and we should see that very soon because if they want to really have a season for Q4, they probably need to stop sending in PO in June, July time frame so that they should have -- they can have a chip for that shipment. So I think that's where we're waiting for to see whether there's an upside or normal seasonality this year.

I see. And then on the automotive side. How would you describe the competitive landscape, right? Because one aspect of this is that you have very innovative products coming out. The other aspect of this is that, look, you have these very large, right, competitors, whether it is Qualcomm or Mobileye or NVIDIA. I think it will help to understand -- I'm sure the competitive dynamic is different versus each of them. So it will help you understand what it is.

Right. So I think -- the biggest problem we run into in this market just like because of our scale. So that's the reason we really early on, we focus on to bundle, to partner with the big Tier 1, the Conti and Bosch and that's a strategy because we know that if we go to bid on a project with one of those big Tier 1s, that itself kind of address that issue. So then the question is really -- your question is really about how we compete with each one of them. For NVIDIA, I think the -- it's clear that the graphic engine is just too power hungry and that the performance per watt is really their biggest problem. The product they are shipping right now need a water cooling. So I think from that point of view to get performance per watt, right, for the automotive industry, it could be -- that's our biggest differentiation. And for Qualcomm it's similar. In fact I would say that we have been competing with Qualcomm with IoT space, too, for the last 9 -- 10 years now. And the reason can win, same thing. We have better video quality, better compression and also better power efficiency. And that's where although that Qualcomm tried -- many times tried to use -- sell [ the AI ] processor into our IoT space, they didn't get much of a momentum. I think the similar things what you are seeing in the auto space, too, is really that while you try to refurbish a processor, which is not designed for this video AI application, the inefficiency is very noticeable. For Mobileye, definitely they have a strong momentum on the ADAS market, and they really dominate market by a black box approach, right? So customer doesn't even to understand engineering. They can -- they sell the whole solution. It's really just manufacturing. And that's why for people who want to guarantee the production time, they choose Mobileye because it's guaranteed it will go to production. So -- however, for most of the Level 2+ OEMs, they want to have their software. They want to add this control portion of software, and that's not viable. At least Mobileye is not offering some viable solution on that. So I think from that point of view, when we compete to Mobileye, it's really about whether if a customer really want to have their own software stack or they don't want to be -- they don't want to use a black box solution, that -- those kind of projects usually be done by Qualcomm and NVIDIA and us.

I see. But if I'm NVIDIA, I'll say, look, I have somebody like Mercedes partnering with me, right, so they understand the specs and they are comfortable with the power consumption. So why isn't that a good push back to be competitive?

Yes. But however, it's water cooling system, of course is high, it's complexity is high. So Mercedes might decide that they can take that into production, but that's for high-end chip -- high-end product, however for low-end if you -- low-end cars still need to have a water cooling system that obviously become a problem. So I think for high-end car, okay, I can tolerate that, but how about for your [ host back ]? So power consumption -- power efficiency is important. You just cannot tell people continue to -- for all the car, you put another water cooling system for that, right? So I think, yes, I agree with you that on high-end car ,you can get away with that, but you may not get away with that in general.

Got it. And then who is the decision-maker usually in the process? Because when I talk with NXP and other, they will say, well, we are also designed with Continental or other Tier 1s. What really matters is has an OEM actually put purchase orders, right? Because the Tier 1 will have relationships with many different suppliers that's not exclusive with one because that's their job to have as many options as possible. So how do you take that relationship and get the comfort and visibility that it's going to translate into x dollars of revenue in some predictable time frame?

I agree with you on that, particularly for Level 2+ design wins, OEM want to make this decision. It's very clear, and that's why we continue to try to make a relationship with that. Even in China, we talk about this CV72 design wins. We have to go to OEM to convince them that this is right chip, but, however, you need to have a Tier 1 to provide a solution to OEMs and we are ready to go into production. So you have to get both, but decision is done by the OEM level. But for other business, it's a little bit different. For example, for the ADAS today, most OEMs say, I'll just buy from any Tier 1 who can give me a production worthy solution. For the [ EMEO ] or for the recorders or even for the DMS or radar, right? Tier 1 is driving the decision for the solution that they're going to propose to OEM. Obviously, Tier 1 have multiple silicon supplier. They can provide all of that, but you have to build a relationship with Tier 1 to make sure you have a solution that they can propose for you.

Right. So when do you think these relationships with these Tier 1s will become tangible? Is it '24 or is it '25?

Well, we just announced the first design win with Conti. You can see that, the momentum is there. We definitely want to continue to leverage on that. And also, we were trying to make sure that we continue to build our momentum to convince like Chinese Tier 1s. That's also because Conti and Bosch also in China, but Chinese -- there are also different Chinese Tier 1 that we need to work with. So that's definitely an effort that we continue to put in to make sure those Tier 1 can propose our engineered solution to OEMs when there's RFQs.

Got it. Do you think the business in China will prove sticky over time? Or do you think that business will face a lot more competition?

You have competition everywhere. I think sticky is -- in China -- well, that's a good thing and bad thing about China. China is if you have technology, they will use you because you have a lot of scale, right. It's also price performance right, they will give you a chance. That's building Chinese market. And that's why we think that the CV72 will penetrate the Chinese market. But of course, at the same time, China market has a problem, which is that they want to use their Chinese component as much as they can. So that's why like Horizon Robotics has [ holdco ] advantage in China. So our strategy has to be we have to provide better price performance than our Chinese competitor and making sure that we give the -- give our customer a reason to use us. Technology has to be there, right? So we think that we have a chance to win China because our price performance is there. And also, we have a road map that we -- our high-end chip already is multiple [ Orin-X ] performance, while the low-end chip is selling at -- we're already quoting a $50 in Chinese market. So you can see that we have a road map that cover really from high end to low end for automotive. And I think that's another advantage we have over our Chinese competitors.

Got it. Has some of the weakness in the business been because of the weakness among China EV makers? Like is there a correlation there? Or was that not business you would have gotten anyway for this year, it was a longer-term business? So it's not the trends this year that matter, right?

Right. So we do sell our other solution like recorders and also [ EMEO ] type of solutions into EV market. We do see that is weak. But however, it's not really a big portion of our business. So it's not very -- it didn't have a significant impact on us. We talked about our China total exposure is 15% being consumed in China. So there's only a portion of that is in automotive. There are other businesses there. So overall, I think that because our penetration in Chinese automotive market is not big at this point, so that impact to us over today's economy is not severe at this point.

Got it. Now I have to ask you about the AI market, right? That seems to be very exciting growth area. What does that mean for Ambarella? Because when I think about it, especially large language model, right, there is the well-known data center market and then there is also people like Qualcomm, who are saying I can put a 10 billion or 20 billion parameter large language model on the phone itself, right? So what kind of opportunity do you see for Ambarella? When do you think it starts getting more tangible?

On the edge side, we are doing it already. For example, if I'm driving a car, one of the most popular network is called transformer. Really, that's a network, if you have a Tesla on the display, you see this bird's-eye view of your surrounding, that surrounding view is generated by transformer. So everybody, all of the car makers want to copy Tesla need to have that transformer. And if you ask around about who has the best implementation -- transformer implementation of silicon, I think you'll hear all of them. And also, we demoed that transformer at CES who is one of our largest Tier 1s. So we know that [indiscernible] type of network, we have probably one of the most efficient silicon architecture. So from that point of view, in fact, we have less -- our CTO and I have been looking at the server market for many times, and we decided not, that's not our market in the past because it was dominant by the training workload, and we don't have training data. So we don't have a training solution. And also even for the server inference engine, the market is segmented and our differentiation is not really highlighted there. But I think that argument changed -- was changed when the [ ARM ] was introduced. We believe -- everybody believed that ARM inference will be the dominant server workload in the near future. And if you ask people who are really generating all of the workload, what's the most painful thing with the current solution is, one is, to -- the price is too high and the other one is power consumption is too high. So then we start asking ourselves with that, do we have a solution? Do we have opportunity in that space? So in the last 5 months, we spent time to put a large ARM model on to our current silicon, silicon ensemble to our automotive OEM in August last year. We put a real silicon and kind of confirm that. We believe our -- ARM performance on the silicon is similar to A100 and 1/5 of total power number. With that, we believe we can address both the cost as well as the power consumption problem, right? Why power consumption is so important, right? Each data center, the power supply is a fixed number. So with 1/5 of performance that means you can stop in 5x more AI performance into that data center. That's meaningful. So I really think, especially when the workload becomes so dominating there. So I believe that there is a demand for an ASIC-type inference engine for server, and that's where we are coming in. So while we are currently at there, we need to really quite quickly finish our demo to our customer and start doing a business environment to talk to customer to make sure they understand our performance and also convince them that to port their ARM model onto our chip. We don't need a lot, we just need a few examples that prove our performance and the power consumption. And with that momentum, that will probably really trigger us to do more investment if we can convince ourselves there's a real business behind it.

But is comparing to the A100 a fair comparison because that's part of a lot more of their training setup. So if you were to compare it to the T4 or L4 or L40, right, which are more inference PCIe SKUs, then do you think you would still have a favorable comparison?

Yes, I think so. In fact, definitely, that's A100 is really GPU, right? So if you look at using A100 to inference, their performance is still pretty high. So I think we can definitely talk about how to do a fair set by set comparison, it's really about to prove that. But we have done this with automotive trying to compare our other new network performance compared to NVIDIA that we get similar performance numbers. So we definitely think, from our point of view, that's a fair comparison. If the people -- customers think that we need to prove further, we'll definitely do that.

I see. Is that a worthwhile market for me as we were chatting before because you mentioned that there's a number of players, right? I mean the incumbents, right, it's not just NVIDIA, it's Intel Habana, right, it is AMD with the MI300 planning to enter the market. So you -- and of course, your Broadcom TPU, et cetera. So is that market worth investing in when you have 3 or 4 incumbents and when you have several others who are also trying to -- right, like how much of your R&D resources is going after that market?

First of all, we're using our current silicon.

I see. That's incremental.

And also the software that everybody is trying to build. We already have that software for our auto business, right? Today, people talk about competitors try to convert neural network to our chip. And we have 100 customers come from 500 even new network on our chips which is based on our tools already. So we have tools, we have an environment, we know how to convert, helping our customer to convert neural network to our machine. And we need to optimize the software for [ LOM ]. At the same time, through working with IoT customer and automotive customer, guess what, we are helping them to put software running on CUDA to our platform twice already, helping them to do one more time to find the server in this engine to our chip. It's just another access that we have done. So from that point of view, the resources required to enable a handful customer to prove the concept, get design win, we don't need a lot of people because we already have that investment on CV3 based on leverage, the heavy investment that we put on CV3 in the last 4 years. So from the additional incremental OpEx, I don't think that's a lot. But when we convince ourselves there's a real business behind that and we can win that market, you bet we're going to ramp up that engineering investment to try to get an even bigger market share. But today, our approach is let's convince ourself and the customer that there is a business model that we can win.

I see. On -- back to the automotive side. I think in the past, you have also made important investments, right, on the radar, right, side and then in domain controllers. Could you give us an update on where they are, those investments?

So domain control is really CV3, and we talk about that the design win already. So CV3 as a domain control, we're doing very well in terms of continuing to deliver software on our chip and then with customer we are talking about that. We already have first revenue in China when we get a more design win there. So that's happening. On the radar side, I think it's a great important thing because Tesla just announced that they're going to put a radar back. So all the copycats who want to copy Tesla design were starting how we're going to use 4D imaging radar to do a ground truth. That definitely will help us to do business development. And we -- and also, we have momentum on the design win side with Tier 1 to really look at our centralized architecture and putting their software over. So I think from the radar point of view, I think we also start continuing to make progress on the current situation. So our target is really -- at the end of the year, we have to demo that our VisLab software, Oculii software running our single-chip CV3. That's the important milestone because at that time, we basically have a total solution offered to OEMs with everything they need.

Got it. How would you contrast that with, say, somebody like NXP or Infineon, right, who are the incumbents in some of those markets?

I think that -- I think there are still -- for 4D imaging radar there is still focus on edge device, basically building a module on the edge and using very big antenna array to do that. That's not our approach. The approach is how to use smallest possible antenna array, 4 x 3 ideally and we use a single processing at the center to generate similar point cloud and the performance in terms of range, resolution and accuracy. And that's 2 different approach. So from that point of view, we think our costs will be lower, our power will be lower, but we didn't trade off the win. We can provide similar [ resolution ] performance like what they do. So that's a different approach, just like a video already being centralized, domain controller centralized all the video processing. We believe that will happen to radar also.

Got it. Then maybe something on the foundry relationships. Do you think you have the right foundry partners, you're comfortable with that, right, given some of the issues we had last year? Or do you think you are now well set and you have all the support that you need to execute on CV3?

I believe so. 5 nanometer is what we need and we already secured the supply for the CV3. And then we have -- we definitely built a very strong relationship recently with Samsung again because with a change management team, we didn't follow u. So we -- now we build a relationship with a new management team. And I think that our relationship with Samsung strong And also we get a commitment to continue to support our supply 5-nano supply chain. So from the foundry current generation point of view, I think we are set. I think the biggest issue is really, the way I look at the process node is gate overrun is coming maybe not this year, maybe next year but it's coming. Who is going to be the first one to give us efficient get around implementation? I think that's the biggest decision we need to make in the next 2 years. I think that will change everything because that's probably the most important technology on the process. Basically, think that continue to close down like Samsung 524 or TSMC 523 is all FinFET. And from that point of view, it won't give you a lot of advantage. The biggest advantage is gate-all-around and we haven't made a decision yet.

So you think after 5, you would just directly make this step to gate-all-around, you're not going to stop at 3 or FinFETs or anything else?

Well, for Samsung 3 nanometer gate-all-around. So the only milestone is 4, should we do that? It really depends when gate-all-around is going to be much short. If this takes too long, we may have to do another step. But I will definitely look at. However, I want to point out even if gate-all-around is available, it's not going to be also great. It's very difficult to get all the way. So for auto, I think 5 and 4 is probably the key. But for the non-auto business, straight -- gate-all-around become very meaningful discussion.

Got it. And last question then for me is as a company, you think you have all the resources, all the technology, all the IP to execute on your growth strategy? Or you think that some additional M&A might be required to execute and expand like this AI area. Do you think it's -- you can go stand-alone? Or you think going through M&A will, no pun intended, accelerate your plans?

Well, there are 50 stellar companies planning to compete with [ 50 ] people. I'm not saying that's the right approach, but I think we calculate the resources we need, we think we can fund it ourselves with our own current resource. Hopefully, if our revenue recover to the normal level, we can continue to invest. I'm not going to assume that somebody is going to pick us up. . Our business, for us, we need to make sure that we can stay independent. And if that's what's required, we need to make sure that we can independently continue to fund the project that we need to do. So CV3 and now AI, I think, we estimate the performance, I think we still have -- we do have a resource to do that.

Right. Makes sense. Perfect. Thank you so much. We really appreciate your time.

Thank you very much.