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

Good morning, everyone. Thanks for attending the Berenberg U.S. CEO Conference. My name is Andrew Buscaglia, I'm the U.S. Industrial Tech Analyst here at Berenberg. So today, we have one of my personal favorite companies, Ambarella. This is an artificial intelligence company with roots in video processing and compression chips. However, in the last 6 years, Ambarella started developing computer vision, which is finding really tremendous growth opportunities in security, automotive and other industrial applications. So with us today from Ambarella, we have the company's CEO, Fermi Wang; Director of Corporate Development and Investor Relations, Louis Gerhardy. And the format of this call is going to be a fireside chat. I'm going to start out -- Fermi is going to start out making a few quick comments, and then we'll hop right into Q&A. After about 20 minutes, I'll open it up to questions. If you want to ask a question, you can put it in the Zoom message box or e-mail me. So with that, Fermi, if you want to take it away.

Thank you, Andrew. Good morning. First of all, Ambarella was founded in 2004. In the last several years, we have gone through several generations of technology change. In the first 10 years, we focused on video processing and compression, helping our customers to capture video. But since the IPO, we started working on AI technology we call computer vision basically to analyze video real-time in our chip. And we spent several years and hundreds of million dollars of R&D investment into this technology. And we started seeing the results of that starting 2 years ago. We talk about -- we have 3 phases of revenue from computer vision. We'll talk about that in detail. And also, not only we build a foundational technology, now we are seeing the revenue result. But more, we think that recently, we announced -- we acquired Oculii for the radar technology. The reason behind that is really, we believe that the vision processing become or computer vision become a critical technology for many different industry. When we look at those industries all of them require sensor fusion, which means they need a different system of modality, and the radar usually become the most popular one next to video. So -- and with that acquisition that enable us to really start offering the sensor fusion solution to our key customers. And we believe that with that we'll continue to increase our serviceable market as well as reaching out to more new customer bases. This is just a really quick summary about the company. I will be happy to answer more questions.

All right. Well, thanks for that, Fermi. So maybe to start out, we got a lot to talk about with the Oculii acquisition and all the growth drivers that are kicking in. But maybe you can start off taking a step back and talking a little bit about Ambarella's uniqueness and that they are a platform technology. I think that's something that people don't quite understand. So could you explain what that means and what the benefit of that platform is to customers? And I'm talking specifically for your computer vision chips.

Right. The platform that we talk about is really a technology we use to not only just showing up our technology but also how to support our customer. We call this an open platform, which is very important for most of our customers because most of the customers who come to us, they want to leverage what we are good at. But in addition to that, they need to add to their own differentiation technology into this joint product. And this open platform allow them to do that. Not only allow them to do that, we help our customers in the last 20 years, continue to help them to put their software onto our platform with our software in there. So this is an open platform, is powerful to enable customers to provide differentiated technology. That's one. The second thing is important is this platform is a scale, meaning that, for example, we offer the CV2 families a CVflow chips. You can see that we have roughly 7 different CV chips here, but they are all so very compatible within this platform. What that means? If our customers develop software on one platform, one chip, the same software can easily scaled, in fact, we allow -- just with the compilation, they can easily taking their software to a different chip. What that means is because different chip has different performance and price point, it really make our customers easy to adopt this whole family of chip, without reinvest all of the software development. The third thing is this platform is based on some really critical technology. We differentiate our technology based on video quality, we believe, we have the best video quality among our competitors. We have the best video compression. And more importantly, our computer vision technology has shown up and show people that we really know how to drive the performance per watt. And our power consumption continue to be the shining point for our platform. So those strong support here really helping us to differentiate us against our competitor. Now we add Oculii's radar technology into this platform that makes this even more powerful offering to our customers.

Okay. Yes. That that's helpful. I think before we get -- I think I want to talk about Oculii in a second. Right before we get to that though, can we talk about your computer vision and this platform has been -- the growth has been driven by a few waves. I know you guys talk about professional security, then home security and then automotives are starting to pick up. Could you talk about some of the applications, computer vision is finding in professional and home security? And then we'll talk about automotive after that, but maybe start with the security side and why that's driving growth right now?

Right. In fact, I think from a professional security camera point of view, is really all of the existing professional security camera application is transitioned from the video to the computer vision. In fact, the smart city is probably the most obvious one that we're seeing a lot of different projects. If you look at our customer like Motorola, when they come out to talk about their system, how they help to develop a smart city system with the camera that basically is driving force behind this, converting from the video-centric professional security camera to a computer vision. So I will say that although this is really beginning phase of the transition from video to computer vision. But I expect that at the end, all of the existing professional security camera will go to a computer vision solutions. On the consumer side, I think we are just at the beginning, we announced Ring as our customer. There are a few other people are working with our solution and that we will probably talk about later this year, early next year. But if you look at the driving application, it's really about how to provide more safety and real-time alarms to your home, for example, the most popular application today is [indiscernible] light type of camera also outdoor or indoor battery-based cameras. So all Those new applications is really continue to drive the safety and the security at home and with computer vision that enable real-time alarm and also tremendously reduced the false alarm rate. That's really the driving force behind it.

Okay. Helpful. And in automotive, you guys relatively recently talked about a $600 million funnel. So I think understanding where you guys -- where computer vision will play in automotive could be helpful if you could explain what are -- maybe talk about some of the design wins you most recently won and then kind of where you expect growth to come from?

Yes. So first of all, this page shows you all of the automotive applications that we are working on. And of course, our traditionally work on the recorders and data loggers, which are popular in the Asia side. But if you look at our computer vision chip, the most popular application is on the right-hand side, is really starting with ADAS, then moving to DMS, in cabin monitoring, and now move to an L2+ eventually go to L4, L5. If you look at our design win on the DMS, OMS side, we announced Dongfeng [indiscernible] our customer using our solution today. And with fleet management, we announced multiple solutions already like the Hyundai's KeepTruckin, Momenta. There are quite a few that fleet management companies are using our solution today. In terms of the L2+. We announced Arrival, Motional, all of those guys as our key customers moving forward. But I think that this is really just the beginning of the trend because, for example, I really think that Level 2 or Level 2+ is going to be the mainstream revenue generation for us moving forward. And this is where we focus on. And we have L2 -- we have a multiple chip solution that can address this market. But again, Oculii's radar software directly helping us to secure more design wins in this category.

Okay. Yes, that leads me to my next question. So Oculii, you guys just announced an acquisition more recently. So there's some confusion around what exactly -- how this technology fits in, I guess, with the spectrum, radar has never been really considered maybe long-standing technology that will be that influential, but this seems to be taking radar to another level. And then combined with computer vision, you guys are going to come up with some interesting solutions. Can you talk a little bit about the process behind acquiring that and why you think it's going to be helpful?

Right. So first of all, if you look at all of the sensor modality for the autonomous driving or Level 2+ or even security camera, you know, lidar, radar and camera has been mentioned many times. But if you -- from an Ambarella point of view, we try to pick the most complementary technology to us, we think radar is natural because both lidar and the camera using optical sensor, all the problem that impacted the performance of camera also impact the performance of lidar. And radar is really, for example, at a low light condition rain heavily or deep fog, in all those weather conditions that the radar can continue to perform well and really help the sensor fusion. So you talked about in the past, radar has a lot of issues. I think the resolution is one of them and also the distance is also another one. But however, when we evaluated Oculii, we understand that Oculii is taking a totally different approach to address radar solution. Instead of trying to using a lot of hardware like multiple transmitter receiver, what we did was -- what Oculii did was really using a signal processing on the software side to address this issue, to solve this issue and to provide much higher resolution and much longer distance that which was a problem for radar, and with Oculii solution, we think they can solve that problem. So with the combination of our computer vision processing and the Oculii radar, I think this will address most of the sensor fusion problem moving forward. I truly believe that from low end of L1 to L3, that's going to be vision plus radar only. Level 4, Level 5 probably continue to want to use lidar, but considering the cost, considering the complementary technology, I think that the video plus radar makes sense for all of the application that we're talking about.

So lower -- so yes, the lower cost and combined with computer vision makes it a as good enough technology to compete resolution wise. And I guess what -- can you size that opportunity? What you think -- I don't know if you have units in the mind or growth rates or however you guys were thinking about what this could -- like how big could this get?

Well, in fact, in this slide, you can see that we think that in calendar year '21, we have 90 million units of market and grow to 240 million units in calendar year '26. I think this is just by itself a very big market. Of course, there's from high end to low end, there's huge range just like a video. But I think that all of the key rationale behind us is that all of our applications that we are selling to will require radar or different type of a sensor modality to help the accuracy and the redundancy of the video systems.

And you guys had mentioned this influencing your ASP a bit. It could be positive and your market share -- or recurring revenue rather. Can you talk about those dynamics to how that -- how you see that playing out?

That's really about the Oculii's business model. Oculii is selling 2 things today. One is they are selling a complete module, which is being used to evaluating their technology by Tier 1 OEMs. And they have to have this reference design so that people can easily just connect those modules to the cars or other systems and can start evaluating the performance. But the main business model moving forward is licensing software. So their original business model is giving software to the Tier 1s and charging software royalties. So in the future, our plan is -- we'll keep this Oculii business model as this will continue to do this without interrupting the current business. But moving forward, we also believe we are going to integrate Oculii's radar software into our CVflow engine. And then that will enable diffusion type of sensor fusion and then therefore, we can charge or increase ASP because we are really providing more function and more redundancy to our customers.

And another interesting angle, we talked about the opportunities in automotive with Oculii, but what are some other opportunities outside that? Maybe you guys mentioned industrial robotics and even security. What are those applications, that you're referencing?

Well, in fact, after we announced this deal and we went out to talk to all of our existing customers, automotive and the security camera. And we got overwhelmed response by a security camera company because they are all looking at radar. Maybe they look at radar starting with the high-end system, but that's because the cost and also the complexity. However, if we can -- I think when we integrate that software into our chip and try to minimize the bond, that will help the adoption on the security camera side. And for example, I knew that some of our key customers already have a high-end product with radar in there. And for example, even some of the consumer security camera already integrated radar in there. So I think this trend will continue. And I truly believe that the security camera can be a huge market that people ignore for the radar system moving forward. Of course, that all of the other robotic application we're talking in the past like AGV or other smart manufacturing, as long as you need to work with -- you need a redundancy or you need a more accuracy, a radar become very easy way moving forward. Because now in the future, we're going to integrate this to -- we're going to offer software solution for both the video and radar on top of our CVflow engine. That will make it easier to adopt and reduce the cost.

So we're just about at the 20-minute mark, I wanted to -- I got a question from an investor here. So it's very topical, asking if you guys will -- what is your view on the Metaverse? And can Ambarella be a play on that? Or will you guys participate in that?

We are not. But however, the virtual reality is a big space. I think that as far as I know that Meta's plan is to do all the silicon development in-house. However, I really think that in the past, we have done many AR and VR solutions. Because although the market is not very big at this point, but I think there's a potential moving forward. But more importantly, I think all the VR systems is going to require computer vision, different sensor to help them to navigate and also understand the environment better. So I think that definitely, there's an opportunity for us. But at the same time, I understand that the Meta's plan is to doing most of the silicon development inside their house.

Okay. Okay. Well, maybe we'll come back to Oculii in a minute. I was trying to get through a few others that are more topical maybe right now. So obviously, the topic is towards supply chain challenges. However, you guys have been doing fairly well navigating that and a lot of that has to do with your kind of supplier relationships. Can you talk about, as of last couple of quarters, how you guys have managed to handle this so well?

Right. So in fact, that we have a strong relationship with our supplier like Samsung Foundry, the ASE testing and the packaging house. And we have a strong relationship there. And we -- from a foundry point of view, I think Samsung although is tight, but it's not really creating a severe problem for us. So in fact, in the last several quarters, although that -- the biggest problem for us was the Texas, Samsung's Texas foundry got shutdown because of weather. But after that problem is resolved, although we are very tight, but we don't -- we are not short to our customer. The biggest supply problem that we are facing right now is really that our customers cannot get enough parts from other vendors. For example, MCU, WiFi chip, power supply. Those components are continued to be in high demand and short supply. And while talking to our customer, everybody says they don't see anything, they don't see any improvement probably until early next year.

Yes. Okay. So it's not your problem securing parts. It's more, I guess, on your customers being ready to move forward with, I guess, where they are with their development?

Well, but that also entitle us because it becomes hard for us to forecast what the customer needs. Because our forecast becomes dependent on other suppliers components. So it's a tricky situation for everybody. It's just nobody can avoid this problem at this point.

Right. Okay. Another question comes up often, just kind of form of the landscape of this computer vision industry. I think a lot of -- I think it was a misconception that it can be commoditized, and there's a lot of players around doing this. Can you talk more about competition, how you perceive? Whether it's start-up players or some industry giants like Qualcomm that have showed interest in this area. How do you guys fit in? And where -- I guess, relative to that, where are you unique?

Well, first of all, even video space, when we start the people kind of say, "Well, this is easy technology people can easily compete with you." But we haven't seen even very strong competitors globally. Computer vision is something. I think what people misunderstood is, computer vision doesn't mean it's just a bunch of multiplier accumulation. That's a too simplified way to look at it. Computer vision by itself is because it's continued to adapt because new algorithm continue to come out. And our approach is different than our competitors. We look at algorithms that we want to implement first, then we decide our silicon architecture. With this approach, we really can drive not only the performance as well as the die size. But more importantly, we can drive the power efficiency, which is critically important for all applications that we're working on. Not to mention that the small form factor is security camera but their power consumption is important. Even for automotive, right, telling the people to have a power consumption to handle 300 watt chip is difficult and that we can continue to show multiple X better power efficiency than our competitor is important for us. And our competitors today on the automotive side is, you mentioned Qualcomm and I think Mobileye and Nvidia there are very, very well known. On the security side, outside China, we don't see many of high-end competitors and a lot of our competitors staying on the low end side in China and Taiwan, there many of them. But on the mainstream and the high-end side, we don't see competitor. In China, in the past, our biggest competitor is HiSilicon. And the situation there is very unclear, right? In fact, that although HiSilicon cannot get a supplier on their foundries, but we continue to hear rumors about a different way to showcase their limitation. And we have no proof of that, but it's just a complicated situation we need to continue watch. But even in China, I think today, HiSilicon still is our biggest competitor, but there are other, many other smaller competitors there.

So you're saying exactly, so HiSilicon having -- being owned by Huawei and effectively being blacklisted, is created this dynamic where you -- maybe companies that were using you now are more so using them and vice versa, that companies outside of China. You're starting to see some -- or you have been seeing incremental demand based on that? Is that part of the dynamic?

That's absolutely true. In fact, outside of China become very obvious that because of the short supply of HiSilicon as well as that our new CV platform continued to show up with multiple different levels of performance level, and we continue to win a majority of design wins outside of China. . Inside China, I think there are people still holding up a lot of HiSilicon inventory. And I was surprised to see how many inventory they built before they got the blacklist. But there HiSilicon still quite widely available inside China and then we continue to consider them as our competitor.

And I don't see any questions coming through yet, so I'll just keep going. On this competition topic, Mobileye is one of your biggest competitors that will be in Forward-Facing ADAS, but they have such a dominant market share right now. So I think they have over 50% market share. So how someone like Ambarella expects to compete there? Or maybe explain 1 or 2 differences between the businesses where you think you have an advantage. And then again, like -- how do you expect to gain any share in this kind of like market controlled by Mobileye at the moment?

Right. So 2 things. In fact, 1 is really business model, that Mobileye has been shipping a black box solution to everybody. And everybody using the same technology, same solution, which is a problem for most of Tier 1 OEMs, and then we talk about our open platform, which help our customers to differentiate. I think that's one biggest reason people looking at us. The second reason is on the technology side, we are there competing side-by-side with Mobileye. For example, our video processing capabilities, definitely better than theirs, on the RGB camera and also that our computer vision performance is definitely the performance per watt, I think is definitely better than theirs. But however, what they have is they have a lot of data. They have a first mover advantage that's where we want to compete. And so I think what we need to do is continue to build up our competitive advantages, Oculii again is one of the reasons.

Yes, so back to Oculii. They give you -- in some of these more futuristic automotive and ADAS markets, they give you a real competitive edge. Again, against one of the bigger competitors in that area.

And also, I think that just one more thing. I think that the approach that Oculii is taking and the way we're going to integrate our chip will make the sensor fusion between video and radar, a lot better than just using a [ blue force ] way to sell the radar data technology that people are using today. We'll give you more details at the Analyst Day in January.

Okay. In that Forward-Facing ADAS market, can you size that opportunity? What could that mean long term? I don't know, I just think big numbers, I guess, units or market share-wise, how you think about it that growing?

Well, first of all, I really think that the adoption rate will continue to go up. So I think that very soon, when the ADAS solution getting cheaper and widely available, I think the adoption rate will go up. And I think that I won't be surprised that almost all the car has a ADAS system in a few years. So for us, really, the penetration need to focus on where we can differentiate usually, like I said, is technology-driven as well as business model. So just penetrate one big design win can give us a few percent of this huge market, and that's where we're counting on that. We need to continue to find a way to secure design win and helping our customers move into a production soon.

So I don't see any questions, so I'm just going to keep going here. So I think we talked about security quite a bit. We talked about automotive quite a bit. But there are some applications that I think you guys can be influential and with robotics, we touched on with Oculii. But I guess, what are some realistic applications you could see Ambarella being specked into, when it comes to kind of more of industrial, maybe warehouse setting or anything that comes to mind?

Right. So there are a few things that we are working on. First of all, that's called robotic such a segmented market. There are many different small locations, small volume application today, but all of that has potential to be big. For example, there's a lot of people using existing security camera and loading different software to enable totally different application. For example, people using a security camera become doing people counting, doing a face recognition sort of, to become access control solution. So this is kind of a new application with existing hardware across new software. That's definitely one area where I believe the security standard -- typical security car market will transition into multiple different application, which has nothing to do with the security and safety. That's first. The second thing is really about other robotic application like AGV, right? Just those AGV basically is a small moving vehicle and the technology that we develop for the automotive can then easily apply to those things. And we believe those market can be huge in the future while we are still talking about small volume production at this point. But I think those markets that can potentially to have a high volume in the future is going to continue to be the area that we need to put our business development in. And also I think there are a lot of traditional machine vision type application using traditional algorithm, using a huge PC, industrial PC with industrial camera to do, but there are many things can be converted to with the computer vision, with a neural network-based computer vision and using small form factor camera to replace that. So all of those can be a new market for us. We are watching those markets. In fact, we're doing more than just watching. We are putting business development resource in there to understand the market and try to spec our solution moving forward that can address the needs of those new markets.

And the last comment, you're referencing maybe machine vision, some of the machine vision players, it's how big spot defects in factories and things like that, you think Ambarella could be useful in that sort of application. Is that what you're referencing there?

Yes. In fact, again, back to platform, we are not providing total solution. We are providing the platform so that helping those machine vision customers can move their software onto our platform and leverage our CVflow engine to using neural network-based solution to solve the same problem they are solving today. I think a lot of the machine vision company are moving to that direction. We are -- we need to prove to them that our platform can be useful and deliver much better performance than other platforms they are evaluating so that, that we can move into that direction. But definitely, we believe that we will be start seeing some of the companies are using our solution moving toward that direction.

Okay. So Fermi, you bought -- you guys just closed on Oculii. What's your use of cash going forward from here? I mean, clearly, M&As of interest. But is that maybe is there some technology you can -- but you could still sound useful that could still help further round out your offering? Or is it going to be technology that can get you into new markets that you're not currently in? I guess how are you thinking about cash allocation? And just if it's M&A, what else can you talk about?

Well, first of all, our cash utilization is definitely focused on M&A and Oculii is a great example, but we do believe, we're evaluating a lot of different companies and see how they match to our offering in the market. What we are looking for is not a silicon company, we're looking for is really algorithm type of approach that we don't have today. Just like I said, computer vision and other sensor fusion modality, they continue to adapt. They continue to have new research coming out. And I actually believe that 10 years from now, the computer vision will be totally different than what we have today, and we just need to continue to evolve. And I also believe that, of course, that the growth inside the Ambarella is important, but when we look at, when we found the unique technology in the sensor fusion areas in terms of algorithm development, that's the area we're going to focus most moving forward.

Okay. Maybe we could take a high level, some -- maybe some high-level thoughts on just the model going forward with, you got this great top line growth coming and you're seeing that already. Can you talk about how the dynamics with the margins playing out? Your gross margins are 60%. Can you talk about -- is there a chance that those expand from here? And then secondly, where is the leverage in the model going forward? It sounds like if the top line is there, what are sort of the levers you can pull that you'll see that earnings really start to ramp?

Right. So from a gross margin point of view, we still think that our current guidance is right. We have -- our current corporate gross margin guidance will be our long-term gross margin. And the reason for that is our computer vision definitely gives a boost on the gross margin. But I think just like you said, on the low end side, there's a lot of competitors. And also, we need to continue to -- the ASP continue to -- basically every year, it need to go down, although with a short supply, we haven't seen that for a while, but we should assume that any silicon or probably ASP will go down or go up. So the only way we can maintain our gross margin is continue to introduce new products, and that's what we're doing so. I think we are very comfortable with our current gross margin guidance in the long term. From a leveraging point of view, I think the operating margin, we see a significant improvement in the last several quarters. And we believe that we're going to continue to show leverage. However, I think that we need to continue to invest at this time, like we're competing with Mobileye, Qualcomm and NVIDIA. And I truly believe that continue to invest on the technology is important for us. For example, we are going to continue to invest on 5 nanometers. We are evaluating 4-nanometer today. That's a huge investment. And also from a hiring point of view, we need to continue to boost up our technology on the computer vision as well as now with radar technology, too. So while we continue to increase our R&D expense, but we do believe that we can continue to show up leverage on operation margin, not as fast as last year, but we definitely think that 20% to 25% is a target that we want to hit in the future.

And so you're saying with R&D, though, you'll still be spending quite a bit, obviously, to maintain competitive advantages. But that's going to kind of level out like that R&D spend relative to the growth there relative to your top line growth is, you'll start to see some leverage on that level. Is that kind of how you're saying?

Yes. We definitely believe that, it's our job to continue to show up strong leveraging on the operating margin side and which is our goal, right? So I think that our R&D expenses will be [ limited ] about how fast we can grow our top line, bottom line. And that will be the variable play and try to make sure that while we continue to grow our technology, we continue to show up with a better performance -- financial performance in the future.

Yes. Okay. And what about -- a question I get too, is if there are -- is all this growth ahead? What about investing in capacity and -- you are fabless. But how do you think about invest -- having to invest in the company to keep up with all that growth? I guess where, I don't know where do you see any additional expenses down the line that you don't currently have?

Well, there are 2 things. First of all, in the short-term, I think that we need to -- instead of trying to invest around the capacity, which we need to continue to build out our inventory. Just be prudent that if something happens for any different reason, we have -- at least give us a buffer to react to. But however, in the long-term, I think that our expense will continue to focus on R&D and business development. . And that's where I think that would give us best return for our investors, particularly on the -- in the past, we continue to move forward with almost the leading edge of process node. And that definitely help us, and that's one way to continue to keep competitive. And but it becomes so expensive, not only just on a wafer taping out point of view, but also the R&D investment point of view, we need to build a strong [ VI siting ], strong software team that we can leverage this 5-nanometer or 4-nanometer technology. That's going to continue to be costly. But I think that's definitely worthwhile because we have proved that we know how to build chip, we know how to specify chip. As long as we can continue to do that, the high investment on the R&D side will turn out to be the ROI for the investors.

And I got one -- just got a question in right now. So I'm just going to read it to you, the R&D footprint is spread largely between 2 geopolitical opponents and their allies, U.S., China and Taiwan. How do you manage risk of trade wars and conflicts between these "frenemies"?

Right. So I think -- in fact, it's kind of a hedge on all of the places, right? So definitely, we believe that this geopolitical situation will not get better anytime soon, and we need to continue to make sure that we have -- we don't have disruption on our business. And at the same time, I think China is an important market, right? There's no -- I don't think any company can afford to say they work away. And for Ambarella, we need to continue to stay in China and win the automotive business in there for example. So I think that we continue to heavily invest in all the spaces is important. But also like in terms of risk, we do believe that we need to continue to ramp up some R&D place in addition to the current Santa Clara and China and Taiwan. I think you now -- Oculii is in Ohio, and we have Italy Research Center, which is continue growing, we have 70 people there. So we do understand there is a risk we are taking -- having a resource sitting on this unstable geopolitical situation. But however, at the same time, we are building up resource somewhere else, try to provide a balance on that.

And I don't see any other questions and I'm running short on questions now. So maybe I'll just finish off with one last one. And it's Fermi, you've done a great job. This company is -- things are really starting to perk up here. So at this juncture, what keeps you up at night? What do you think is the bigger risk here? Is it competition, geopolitical, I don't know, just what -- what do you think about? Where are you most on guard on related to risk?

Well, competition is always in my mind, right? Because for a company like Ambarella size to compete big company, is always a challenge. But however, we know how to do it. We have done that for 17 years, so that didn't scare me. What really keeping awake in night is whether we are moving to the right direction from a technology and market point of view because things changed so quickly, right? 5 years ago, nobody can see a computer vision come to this point. And luckily, and strategically, we invest heavily on that and it helped us to come to this point. And now for any semiconductor company, when you build a product, you need to forecast what happen in the next 3 to 5 years because every silicon take that long to develop and really is whether I got all information, like I made the right decision from which market to go after, what's the technology going to -- that we need to build, where to put our R&D resource. That's where I think that I spend most of my time.

I see. And actually, I think I got one last question, I'll throw it out there, and then we can wrap it up. But I got the question is, why is the comp incorporated and came in?

Yes. When we started the company 17 years ago, and the only reason for that is to try to minimize the tax consequences and that's how we did it. And we started the company of out Cayman. We didn't do the reverse. We started company in Cayman and then build all of the other countries as a subsidiary of the Cayman company. And you can see that the impact to our tax rate and that's the only reason we do that.

Okay. Okay. Well, we're just right out of time here. So thank you, guys. Thank you, Fermi. Thank you, Louis, for doing this. We really appreciate your time. I hope we can have you back again next year. Hopefully, in person.

Thank you, Andrew.

Thank you guys. Bye.

Bye-bye.