China EVs & More

MAX Episode #7 (ME) - Omer Keilaf, Co-founder & CEO of Innoviz

May 27, 2022 Tu Le & Lei Xing
MAX Episode #7 (ME) - Omer Keilaf, Co-founder & CEO of Innoviz
China EVs & More
More Info
China EVs & More
MAX Episode #7 (ME) - Omer Keilaf, Co-founder & CEO of Innoviz
May 27, 2022
Tu Le & Lei Xing

In this latest MAX episode, Tu & Lei welcome Omer Keilaf, Co-founder and CEO of Innoviz, which designs and develops. LiDAR and perception software for ADAS systems & autonomous vehicles. 

Omer talks about the early days in the Israeli military, starting Innoviz six years ago, why he is confident that this is the right time to enter the Chinese market, his views on LiDAR competition, technology, the difficulties in achieving mass production, and how Innoviz is tackling these issues with unique approaches. 

Climate Confident
With a new episode every Wed morning, the Climate Confident podcast is weekly podcast...

Listen on: Apple Podcasts   Spotify

Show Notes Transcript Chapter Markers

In this latest MAX episode, Tu & Lei welcome Omer Keilaf, Co-founder and CEO of Innoviz, which designs and develops. LiDAR and perception software for ADAS systems & autonomous vehicles. 

Omer talks about the early days in the Israeli military, starting Innoviz six years ago, why he is confident that this is the right time to enter the Chinese market, his views on LiDAR competition, technology, the difficulties in achieving mass production, and how Innoviz is tackling these issues with unique approaches. 

Climate Confident
With a new episode every Wed morning, the Climate Confident podcast is weekly podcast...

Listen on: Apple Podcasts   Spotify

CEM MAX#7 Omer Keilaf, Co-Founder & CEO, Innoviz
Recorded on May 3, 2022

Tu Le:
Hi everyone, Tu Le here, one-half of the China EVs & More duo. Lei and I are always thinking about different ways to bring you, our audience, relevant and compelling content about the China EV, AV and mobility sectors. Especially now that several companies that we’ve tracked over the last 60 or so episodes have become global phenomenon. 

China EVs & More MAX is where we bring you that special content, in the form of conversations we have with special guests from those sectors.  

In this episode, Lei and I speak with Omer Keilaf, co-founder & CEO of Israeli LiDAR company Innoviz, which officially announced earlier this month a major design win for its InnovizTwo LiDAR with one of the largest vehicle manufacturers in the world. The selection represents the third major design win for Innoviz, which will deliver the InnovizTwo LiDAR and perception software as a direct supplier to the multi-brand group to support a series of new products and features for the OEM. This milestone comes less than a month after Innoviz entered into a strategic partnership with Chinese LiDAR system supplier LiangDao Intelligence to accelerate the introduction of series production solutions of high-resolution LiDAR for the Chinese market based on the InnovizTwo.

Omer talks about the early days in the Israeli military starting Innoviz six years ago, why he is confident that this is the right time to enter the Chinese market, his views on LiDAR competition, technology, and difficulties in achieving mass production, and how Innoviz is tackling these issues with unique approaches. 

Tu Le:
Thanks for joining us today. We're excited, Lei and I excited to have you on the China EVs & More MAX episode, looking forward to learning more about your company and your journey as an entrepreneur. So the first question, can you tell us a little bit about yourself? Your background, how you moved into the LiDAR space and started Innoviz?

Omer Keilaf:
Yeah, sure. Thank you very much for inviting me. So I am Omer Keilaf, co-founder and CEO of Innoviz. I started my career as an engineer in the army of Israel. And as you might know, in Israel, everyone goes to the army, and I served for 7 years, in a unit, which is very technical. So I was acting a as an electrical engineer in the intelligence force, developing quite interesting technologies. It was a very, I would say it's a very small kind of, I would say, unit, but developing quite unique colleges. I think that through the course of those 7 years, I’ve seen how impossible problems are solved by enough commitment, knowledge, and will and dedication, and I think that bug bit me. And it left me with an infinite desire to continue to do things that are very challenging. And I was always looking to do things that are impactful. After 7 years in that unit, I was in three startups, one of them was acquired by Apple, and the other one was acquired by ST Micro Electronics, and the third one is still running. 

And I started Innoviz 6 years ago to develop LiDARs because someone told me that it's impossible. Someone argued with me that it never will work. That's a problem that too many have tried to solve. And it really is something that is very difficult. And I was very intrigued to do that. And I also come from, my personal background, I come from a family where my older sister went into a very serious car accident when I was a kid. In a way, even though I was not physically hurt in that accident, it obviously had a huge effect on the way that I lived my life later. And through the change of the career of my parents and dedication to help my sister, I definitely know the importance of solving possibly the most dangerous activity that we are all still facing on a day to day basis. I mean we still need to travel from one place to another. And that's the most dangerous activity that we are still carrying with us. And there is no good reason behind it. There is no good reason that with better technologies, we can avoid that. And again, not only being hurt, but also the damage, the collateral damage to many families and societies. And it's really in our grasp, this is really achievable. I'm happy to push in that direction. It is very difficult. The guy was right. It is very difficult, but it is possible. I'm happy that we decided to be committed to make it happen. That's what we've been doing in the last 6 years and actually seeing it coming to reality now. 

Lei Xing:
Great. Any special meaning to the Innoviz name?

Omer Keilaf:
Yeah, so Innoviz, it comes from “innovative vision.” And so in a way, we are developing a vision system, but a very innovative one and with a good a feel of the vision that we have for autonomous safe mobility. And we also, eventually people mostly know Innoviz as a LiDAR company, but I actually see ourselves as something slightly different. I want to see that what we are doing is really trying to solve difficult problems that are important for mobility. And we will do anything that we will see as a roadblock technology wise that would become a roadblock to enabling scalable solutions for autonomous vehicle. We will go in that direction and try to solve it. And one of the, following that, I would say methodology or vision is explaining some of the activities that we've done in parallel.

For example, we are not only developing LiDARs, we are developing also all of the software, the perception software that is needed in order to take the 3D data and translate it to meaningful information: objects and classifiers. And you might argue, at the beginning, many have argued why are we wasting our time to develop software while the biggest challenge is still developing a LiDAR. And maybe we should let our customers deal with the software and only focus on developing the LiDAR. And after talking with the industry and trying to understand how they see the challenge in front of them, it became very clear to me that most of the carmakers, if not all, do not really want to deal with the road data of the LiDAR. They, eventually they do want to get a secondary sensor coming from a LiDAR that provides them redundancy from cameras and radars, but they don't have internally, the desire or the capacity to really build all of the data collected and annotated and develop the algorithms that are needed that eventually would allow them to use it. Carmakers want a product, they don't want the sensor. They want a product that the interface between the LiDAR and the compute model is basically identical to what they get from a camera. Imagine that there is just another interface that if you only look at it, you could actually think that it comes from a camera based solution, but it's not. And they want to do the sensor fusion in a manner which each interface backs to the other, gives the redundancy. And they really care less whether it's a LiDAR, a telepathic sensor, or a bird that tells them what happened. They just need something that is robust and redundant to each other. And that's how they want to develop the software stack. Understanding that and understanding that it is difficult to develop perception software. It's not similar to cameras. You need to deal with a new modality. And it requires a different treatment and preprocessing and really a new, I would say, a world of algorithms. We started to learn what's available, what are the tools that are available to analyze 3D data? And we actually saw that it's quite new, at the time that we started, there was almost no work done on developing perception software on LiDARs. We knew that eventually it might become another bottleneck. I saw it always as a blind spot in autonomous in the mobility space that people are underestimate the complexity and the dependency on such an automotive grade software stack, and that carmakers will eventually need.

And that's why we decided to do it from the first day. It's completely, you can say it's not a LiDAR solution. It's a complete stack that we developed. And I think that today we have three design wins so far. We have a design win with BMW, that design win was completely dependent on our ability to provide them also the perception software. We just announced a few days ago, another design win, a very major one with one of the largest car groups in the whole world. And that also required us to serve not only our LiDAR, but also our perception software. So definitely, our strategy and understanding the needs of the market was, I would say, was right. And I can say that as we go further and we see the hurdles of developing a platform that is based on 3D data, there are more activities that we see that are causing difficulties and delays.

And those are things that we are now incorporating into our, I would say, activities. Those are related to the data validation, production validation. All of the things that when dealing with a very unique new kind of sensor, which no one ever have done before and they're struggling with it. We don't want every carmaker to go through those hurdles and we understand that we need to take more and more roles in having a LiDAR going to automotive programs. It doesn't make sense that each carmaker would do it.

And that's why in some way, the new program we announced a couple of days ago, we're not acting as we've done in the BMW, we're not a tier two, we are actually acting as a tier one. And that's a huge step and for sure being a direct supplier for such a car company requires quite a lot of responsibility, because when I know many tier ones that would have loved to be a direct supplier for that carmaker. But it comes with much responsibility. And obviously we had to go through a very long process with the customer to make them feel comfortable that Innoviz could become a direct supplier for such high volume, in series production, serving, by the way, multiple brands and multiple cars, car lines. It's not a very simple thing. We were able to reflect to them all of the benefits that come from a program that is not now split between a tier one, tier two, and our ability to scale and our efforts between different brands and different carmakers in the sense that eventually they get a more mature and less expensive, I would say also from development terms, made a lot of sense.

And so we did, this new design win is the first for our second generation product InnovizTwo. We have two design wins within InnovizOne. Now it's the first design win we have with our second generation, which is for us very important. It lends a lot of confidence to many other carmakers, but also the first design win for us as a direct supplier. And that's even bigger because that's a huge, I would say advantage to any carmaker that is now doesn't need to be dependent on a very specific tier one globally, right? Once you win a program with a tier one, you're basically somehow dependent on it to be able to serve another carmaker, which could be in a geographic that is not relevant for it, and then you need to split your work. And it's very difficult. So now being our own, I would say, tier one, allows us to amortize a lot of effort between different carmakers, allowing us to provide them a very attractive commercial, I would say, offer and shorten the time for everyone. Sorry, for this long answer.

Tu Le:
No worries. No worries.

Lei Xing:
Great. That was a lot of information. We have a pretty good idea of who it is, but I guess the announcement will be forthcoming. Let's go back to the future a little bit. Let's go back to 6 years ago when you started the company. If we talk about 2015, 2016, 2017 timeline, if you remember the big news during that time, was one was Intel's acquisition of Mobileye, $15 billion, right? The other one was the BMW Intel Mobileye project that was aiming to put AVs on the roads by 2021. And that hasn't happened. Remember, right? So and going back then, and now comparing now, what's different? What do you think is different in the global, we’ll go into China in a bit, but in the global LiDAR competitive landscape, what's different than versus now? If you can.

Omer Keilaf:
I think that, look, 2017, you would hear about a new LiDAR company every day. I remember, it was such a horrible time where every day I get a new text message with, oh see, here's another LiDAR company, oh see, here's another LiDAR company.

Lei Xing:
And from China.

Omer Keilaf:
Yeah I mean, it's like a faster than mushrooms after the rain. And obviously as an entrepreneur and as a person that wants to lead this team, it's obviously very important to do a reality check, right? That you're doing the right thing. And you don't want to be surprised that eventually maybe there's a better way to do what you're working on and of course, I always told my team, the people that we brought into the company, I want them to be busy. We're doing things that eventually matter, right? And if eventually we that's why it was so important for me to understand that I’m targeting them in the right direction that we are really doing, the technology is the right one. Every time we hear about another LiDAR company, we would study, try to find the background and try to understand whether we see any advantages or disadvantages. And I have to say I’m very proud that 6 years later, very confident with the architectural decision that we made. I think the first point of time where we felt that we made a good decision was our design win with BMW in the beginning of 2018. And at that time, we had basically almost nothing. We didn't have yet our LiDAR available. BMW shows us for two main reasons. One, they agreed with our strategy, they believed that the way that we are thinking about how to solve this problem is the right path forward. And the second was that after I think 8 months of the diligence, they felt that we are capable that eventually we will be able to perform to all of what we've shown them. And we went in very deep details. And since the kickoff they've been super helpful, the help that we got from BMW was crazy. Innoviz today is really just because of all of the support and help that we've been given by BMW on a daily basis even today. And I owe them a lot of gratitude, and we announced a couple of days ago the new design win. And I was important for me to tell the person that leads the activity there. He's a good friend of mine. It was very important for me to give him my personal commitment, this that doesn't change our focus in supporting their launch next year. We are determined to have our LiDAR on the road and make them very successful. I think that's part of, for us it's very important our customers succeed. And I know that I told them also that possibly that it's not even needed because how committed we are anyway, but I didn't want anyone to think that now a new design and we are diverting resources to somewhere else. That's for sure, not the case.

Now, over the course of years, of course we continue to develop our technologies and we did our first design freeze in the beginning of 2018 to start ramp up the production and industrialization of InnovizOne to meet with BMW plans. But obviously, in parallel, we continue to develop the technology and bring in the second generation of our chipset. InnovizOne is based on a chip set that we developed through the process of the program, but we continued to improve each of those components. And that chip set is now used in InnovizTwo. That's the second generation. There actually we are relying on the same architecture, but there are many improvements, very significant improvements that allowed us to improve the performance by a factor of 30 and reduced the price by 70%. Now, obviously, it was a very major, I would say, improvement. And I think that the creativity of the team, but also learnings that we have on the first program and understanding better how to do things in a better way resulted in the new product. I have to say that the new customer have really pushed our envelope very strongly. It was finally because we did a celebration two days ago and I was showing it was a two-year process. We started the discussion about two years ago. I remember coming to the customer with, first we came with a new design and they said no, it's not good enough. And then we came think we like we redesigned our products probably 15 times until they said it's like, that's good, like that's what we're looking for. I think that in many places, external forces, customers really drive the most out of you, understanding that you need to achieve better from an important customer is probably the best external, the best force that you can get to push your team to try to make the most. In many ways, the result of the InnovizTwo is just because they pushed us strong enough to do better. Who knows if they have continued to say, no, maybe, I don't know where we could have gone. But I’m happy that at some point, they said it's good enough, but it is really a good product. I'm very proud of it. And after we met all of the requirements and timeline plans, we had to go through a very long process to become a tier one, because it was important for us to go in that direction. Becoming a public company allowed us to raise quite a meaningful amount of money, which we saw is an opportunity to show the customers that we are financially robust and they can count us as a tier one. We didn't need to buy any company here or buy any technology. It was mostly to fund our activity. We started the process with the customer now coming into this process, it was very obvious, at first they were, I would say, very, not very positive that this could actually lead to us becoming a direct supplier because this is a very conservative group that they don't have too many direct supplies, but we actually show them that Innoviz, although we were a tier two in the first program, we are much more than that, because even in the BMW program, we were responsible for the development of the production tools, we were responsible for all of the quality testing, the design validation. There were many things that over time we learned that because we know better the technology and we know all of kind of the corner cases makes more sense that we will take ownership for it. And we were able to show it to the customer, showing them that we are actually a tier one and ahalf, more than like a tier two. And very quickly, and after several discussions over things that they had to test us for, I saw that they became very intrigued by the potential of this, I would say, engagement. They found a team that is very knowledgeable in automotive. Those are all things that are unrelated to the LiDAR. Those discussions are for things that make sure that you meet their group standards for supplier management, cybersecurity, AutoSAR, quality assurance, etc. These are things that you like, when I started Innoviz, I never thought that these things will, that will be where my team would be so strong, right? You start by trying to develop a technology, a LiDAR. Being such a strong company in automotive was something that I actually learned through this process. I learned how capable and knowledgeable my company is in automotive after the experience we gained. It concluded around the November last year when they gave us the official tier one kind of position. And we were waiting for a decision. Decision was supposed to be done in November, then it was pushed to December, and then it was pushed to January. And then it was pushed to February, and then it was pushed to March and lucky for us, they also started to feel that it takes too long and decided to give us the formal nomination. I got the letter last Friday. I can't express enough how stressful my last 6 months were, being a public company is so fun. No I know, it's not, it's a complete, it's not fun. We came from being a startup company, which obviously mostly focuses on its growth, right? You need to invest in your future. You develop technologies to win big deals, and then you become a company that basically need to give a report every quarter. You need to show how one activity affects the next quarter and the next quarter, and the sale cycle in automotive takes time, right? This is the deal that we've been working for 2 years. And although we were expecting this to be decided already last year, some investors might not understand what things are happening in the background. And you can't even say anything because this the nature of the industry that we are working at is also very discrete. You can't, even today, I haven't told you yet who is the customer. It's hard like how messed up this is, right? You can't. It's like this is a very discrete industry. It's not very common that you hear, when was the last time you heard that I don't know Conti won the camera business for one company, they don't issue press releases for any decision made in this industry because right, it will create a lot of noise. And at the time that we announced on our design win with BMW they told us it was a mess because now every tier one asked to do a press release about the gear, about the wheel, everyone wants to say that they are in the next generation product of such a premium vehicle. It creates precedent that they don't want to be dragged into. But that's going to happen anyway. We expressed the importance of doing a joint press release with this customer, the new one. And we do expect to do a joint press release later this year, hopefully not too late from today. I think it will be very exciting, although you know who it is, right?

Lei Xing:
We pretty much know who it is.

Tu Le:
First of all, congratulations, originally on your BMW tier two partnership. Now, with this new customer, as someone who worked at a legacy OEM in the sourcing department, I do know how difficult it is to become a tier one supplier, all the qualifications, all the credit checks, all the checks that need to happen in order to become an OEM supplier. But let me assure you having spoken with, and Lei can also assure you, having spoken with a lot of the legacy OEMs over the last several years, you're teaching them as much as you are learning from them, because this is brand new to them as well. So I think you're probably injecting a certain amount of new blood into their systems. And they're getting excited because they're going to use Innofiz as a validation that their technology is going to be leading edge, right? And so it's a two way street. You're going to be using them because they almost automatically validate in a way as a real technology, right, or solid u player in the LiDAR space. So one question and we'll move it over to China right now. You recently established your China subsidiary and announced a strategic cooperation with LiangDao Intelligence. Why do you think it's the right time to enter the Chinese market?

Omer Keilaf:
So there are two, I would say key strategies around the relationship with LiangDao, the companies that we work with that we have a design win with are also very focused on China, so they plan to deploy these vehicles in China and as such, obviously our perception software needs to be trained also in the Chinese environment, right? Now, we can’t collect data in China. That's not, we can’t do that and we need to rely on a partner that can do it. And of course this is a perfect partner, which also has partnership with our customers. It's a really good fit. Beyond that, they are also a partner to deploy our technology in in the Chinese market. As you said, the reason that I think it's a good timing is because InnovizTwo, the new generation, solve some of the friction that we had before in the Chinese market, because InnovizOne was a really good product, but it was too expensive for the Chinese market. In the deployment of some car lines in China, there was a very strong, I would say, tendency to cost versus the performance of InnovizOne. I think that with InnovizTwo, we are not only solving that friction because of the price, which is becoming even competitive in China. I think it's also going to be a I would say, a very strong, a solution that can take them further and further along. The solutions that today you see being used in some car vehicles in China, are mostly for ADAS or I would say first launches that I don't see, I don’t expect them to go very long because of the limitations and of those solutions in terms of resolution and range, InnovizTwo brings the ability to support, not only Level 2 because of the pricing, but also Level 3 and Level 4 because of the performance. So from that perspective, I feel more comfortable and confident this solution has a long term opportunity in China and not too short. We had some thoughts of possibly doing kind of like a foot in the door kind of project: let's bring InnovizOne with one of the customers, even if it's even if we lose some money, even if we like just like it's not a large volume. But I felt that it will be a waste of time. I actually prefer to wait for InnovizTwo, come in with a product that I see has a long lasting, I would say, kind of potential. And I think that's why it's a good time.

Tu Le:
Now, there's a couple of your competitors, your contemporaries have also recently opened offices in China and Shenzhen. So it's the right time, I think there are a lot of mass market companies that are putting LiDAR into their ADAS systems, so I think the volume is going to be there. And as you said, at a 70 price point, a 70% price point down from InnovizOne, it sounds like it would be competitive in the China market.

Omer Keilaf:
Definitely it will be. Innoviz’s strategy is to continue to secure more design wins in the automotive space with this new design win, which is backed by one of the largest carmakers in the world, gives us a huge commitment for very high volume. And that allows us to support and to commit to even carmakers that don't come with that high volume, right? Not all carmakers in the world can give you commitment to millions of LiDARs. And you don't want to lose that opportunity. On the other hand, you don't want to do a design freeze and to commit to a certain timeline and product, I would say that doesn't have a good fit to the rest of the market. Having a design freeze with such a strong customer, allows us to reach high volume which definitely will allow us to offer our LiDAR even for car companies that are playing a more soft launch, and we will be able to support them, in very good, I would say, pricing. And also, we don't need to offload R&D cost on them. And so I think that's what the market needs. That's why I think that this design win will also have a big impact on other carmakers. Everybody's been waiting, right? You've seen many carmakers sitting on the fence trying to kind of figure out where this technology is going, which LiDAR is going to be the winner, right? I think that everybody knows that the level of the diligence that was conducted by this carmaker for the last two years and eventually decided to use Innoviz is a very strong vote of confidence. And they rely on them to be able to bring the product to the right level of quality and maturity. And now we know this because we are a direct supplier. We can actually serve everyone. We don't need to replicate a tier one relationship in different geographics, and we can do it internationally.

Lei Xing:
So a quick question, can you confirm this design win is for, it's global? So it includes, it includes different markets, including China?

Omer Keilaf:
Yes, including China.

Tu Le:
So Omer, one of the big LiDAR companies in China is Hesai, right? And so one of the Velodyne employee that started this company in China and quite a few of the Chinese EV companies are using Hesai LiDAR. So what are the advantages that Innoviz, let's say InnovizTwo has over the Chinese competitors. It sounds like you're going to be able to compete on price. So how do, how do you differentiate your product from some of the Chinese competitors?

Omer Keilaf:
So I won’t pretend to know everything that they are doing, but I would say that InnovizTwo architecture is very lean. It relies on in high level, we're using 905nm, which is basically, in a way, some of the reason of why we're able to take our costs dramatically lower than some of our peers that are using 1550nm. And we're using only one laser. Most of the LiDAR companies are using tens or I don't know some kind of above 100 laser diodes to cover the entire scene. And we are using only a single diode. So obviously it allows us to bring, I don’t think there is a way to make it significantly lower than that. I'm kidding. So it's really is the leanest possible architecture which I can think about. Obviously, it's one laser, one detector. So we only use one collimating lens and one collimating detector. I mean it breaks down to very few components. But we don't come like, we really don't how to say, lose performance. We actually have improved the performance by 30 times between the first product. So the one aspect is of course the performance, the other aspect is the price. But the other aspects that I think less discussed in this market is related to the automotive experience that we bring in. Innoviz now spent now 4 years working with one of the best car companies in the market BMW, now signing agreement as a direct supplier to another very big car company. This adds a lot of value to the customer. As you said, we do bring a lot of knowledge and knowhow and lessons learned. Part of what we tend to do is when we meet with customers is to share our experience and in bringing LiDARs to automotive and in Level 3. And they definitely appreciate it. We're trying to save kind of the lost time and not repeating mistakes that we see in project planning and like trying to give our experience of what worked well, what didn't work well, how we do should do things. I think that risk is a key in those programs because look, take into account this program that we just won. The customer have spent 2 years in making a decision. And then now it's going to 3 years of development and testing now to make the car available and the platform available and then deploy it for 8 or 10 years. You're talking about a decision that is going to be very critical for total of almost 12 years, if not more. It's a very dramatic decision because of that timeline. So you need to make sure that your decision is the right one and because LiDARs are so different than each other, making a bad mistake could be very costly. And I have to say that I’ve seen carmakers making such decisions in the past. You can reach a dead end with a platform that you already invested hundreds of millions of dollars that eventually don't have any future, because it doesn't meet the requirements. It became too expensive all the time and doesn't meet the features that you need. Being able to chosen by two major carmakers allows carmakers to feel more comfortable, more confident. We have already have millions of kilometers driven in testing the architecture. That's a very important thing because with LiDARs, there is a very big difference between a LiDAR and an automotive LiDAR. And the reason is that a LiDAR is an optical sensor. It's also very susceptible for different optical artifact that could happen on the road, because of different things that you can’t see on the road that can shine or blind you or create noise etc. It takes one artifact that happens too much that can kill the entire concept. And those could be things that you only learn when you drive, I don't know, a few hundred of thousands of kilometers, and that's not, it sounds not a lot. But you know to get to a point where the car is already driving that mileage, sometimes take a long time, right? Because at first, you're doing test in the lab, and then you do in a few kilometers, by the time that you get to drive millions of kilometers, it could be too late, right? To learn about something that could actually kill the entire concept. And I think that this is something that carmakers understand that there is a risk in using new technologies in the car, which they might not, unpredicted, all of the issues that might come up. And without technology, this, I think that we carry enough to understand everything that we need in order to not have those issues in the future.

Lei Xing:
We talked about some of the technical details, now moving onto production. A famous line would be prototype is easy, production is hard. And we've heard that many times. What makes it difficult to go into mass production. So one reference point, we can mention is Valeo just produce their 100 million camera in order to go into that massive scale, what has to happen?

Omer Keilaf:
You're right. Production is production hell, right? I have to say that when we first opened, ramped up our first production line, it was in Germany, was managed by Jbill, which is our contract manufacturer. It was on March 2020, now reminding you what happened in March 2020. We couldn't send any engineer to our new production line for over a year and a half. And while you're trying to ramp up your production line, none of your engineers can actually enter it. You can't imagine how horrible that is, okay? And it did take a huge effort from all of us, including myself. I was in a room for over, I think 6 months we were in a what we call the war room, which all of us was in the room, trying to talk with the engineers in Germany, trying to understand now what's going on. And it was very difficult. That's part of the reason, by the way, that today we are producing our NPI line, which is the production line of InnovizTwo, the new introduction line is done. It's Innoviz in this building, everything is done here. I'm not willing to have it anywhere in the world, just here. And that's a huge advantage today that we have everyone that we need to come to the clean room and see the process and understand what it is.

Now. To your question of why it's, what's it takes to get from a few samples to 100 million? It doesn't happen in one step. There are different stages on automotive world. They are called A Sample, B Sample,  C Sample, D Sample. And this is a very known method in which carmakers are using together. So there is a certain list of requirements that you need to meet in each stage. B Sample basically means that you need to go through design validation. It is very important that you do not move to the volume production before you make sure the design feature wise is met, you want to make sure that nothing is doesn't meet the design. When I’m talking about design, I’m referring also to the temperatures, vibrations, that nothing in the design of the product is allowed to change once you pass to the C Sample. And it requires sometimes many iterations. In the InnovizOne, we've gone through five or six iterations of the B Sample to get to where we need to do a design freeze and move to the next stage of going into volume production of the automotive, you finish the B Sample when you have an automotive-grade product, you go into the C Sample where you basically need to meet all of the production validation test. And there it means that you need to take into account all of the tolerances of all of the components that are part of the design.

Now, a lidar is using multiple components, lenses and optics, right? And you need to take into account in that all of those components are produced in different places in the world. And you need to make sure that no specific batch of a certain supplier is somehow suddenly taking you out of the spec. You need to make sure that all LiDARs act the same. It is a repetitive effort that you need to do in very high volume. Now, specifically because this is also a functional safety product, meaning that this product acts for safety, you can't build it manually. Everything has to be done automatically. So most of the effort that you need to do, meanwhile, designing the system is also designing the production, because when you talk about automatic production, it means that you need to design machines that develop that produce LiDARs. For InnovizOne and now for InnovizTwo, we have designed a clean environment machine that basically has robotic arms, that places all of the optics accurately. And yeah, eventually you need to have a station where you press a button and you get a LiDAR at the output. You need to make sure this button, even though you press it 1 million times, it does this like the actual result would be the same.

And the process has to be accurate. You need to make sure that the gluing process, the tolerances of the machine are all repetitive. Once you get to a station that does that, then you can basically ramp up volume, because you don't want to do that while it is producing bad units. The next step is related to the testing. So before you ship a product to the customer, you need to test everything and calibrate it. Of course for a sensor should be able to detect small objects at 200 meters. You don't want to have in your production line, a test line of 200 meters or 300 meters, and that obviously requires some creativity on how to test a LiDAR that needs to be able to see things at 200 or 300 meters. But making sure that no product that you ship to a customer not meeting everything. So that's also automatic. So we have designed what we called an “end of flying” tester, which is, again, it's a chamber, which is a few meters in length where you place the LiDAR, it uses several targets that move in and move out and adds different ways to calibrate all of the components. And once you go through the script, you ship it to the customer. It's a very long and exhausting process. It is production hell but seriously, but once you have tailored it, then you have a machine that produces LiDARs. And that's fun. Then you just produce them. You can duplicate it. You can do it in many regions in the world. We have produced several machines that produce InnovizOne. We keep one in the U.S., we keep one in Germany, we keep one in Israel. And you don't only produce LiDARs, you produce capacity of LiDARs, and it makes it very scalable. And in being much less dependent on a very specific manufacturing line. Because even if I want to produce now LiDARs in Thailand, I don't need to train the team in a very long process, because they are getting a machine that produces LiDARs and they need to feed it with components.

Tu Le:
Good.

Lei Xing:
Produce where you sell. Is that a strategy for you?

Omer Keilaf:
We sell LiDARs to automotive and non-automotive, automotive is mostly to support customers in their pre-development stages, but non-automotive because we do need to aggregate revenues across multiple industries. There's going to be still sometime before that all carmakers would use a LiDAR. In the meantime, we are generating revenues from many markets that are now built by many other LiDAR companies. Next year, we're also coming with another product, which is the 360, Innoviz has also developed a very interesting category LiDAR for the 360, which is going to be also significantly cheaper and better than the existing solutions, mostly serving automotive applications. But the market there is quite large. And it's quite funny to see that even though most of the LiDARs used today in automotive that are 360, the LiDARs haven't made any progress. It's still very expensive, still very big. Performance is not very high. And we were able to take the InnovizTwo architecture and basically use the same optical back bones but spread the light differently. InnovizTwo is like a projector that looks to the front of the vehicle. With the Innoviz360, we are using the same architecture only that we are using a different scanner that scans around the scene. Since the optical budget is so high, the performance is so high, we get very, very good performance, even that we spread it in a 360 manner.

Tu Le:
Can you give us a quick tutorial on like MEMS versus OPA and why Innoviz chose MEMS versus OPA?

Omer Keilaf:
It's because OPA doesn't work.

Lei Xing:
Let's go the next question. 

Tu Le:
Let’s ask you this…
 
 Lei Xing:
You're welcome to expand on that.

Omer Keilaf:
When we started, we tried, we try to explore the different ways to develop a LiDAR. And it was important for us not to just rely on what we see available today, but also what will be available in the next few years. Because you need to set a certain vector that is also aligned. And what's the time that the market would actually make a decision? Our analysis showed that OPA is a technology that is far from being mature, but also from potential point of view, it's very limited. And I don't see it very relevant for this, I would say, industry. We've seen that MEMS is a relatively mature technology at the time that we started. And we understood that the there still is a lot of room for it to improve. And that's what we decided to do. So we took a relatively, very unique technology in MEMS. And that was very, very at its early stages. We brought the right talent. I myself have quite a lot of background in developing a solution by MEMS. We decided that would be the right fit in terms of time and performance and price for the market. And I feel that it was the right decision and not others, other technologies that we believe, No.

Lei Xing: 
So Tu, I'm really interested to hear Omer’s opinion on the debate on the number and location. Maybe you want to ask him.

Tu Le:
So Chinese EV companies right now are really leaning into, we have over 30 sensors, we have over 40 sensors, we have three LiDARs, we have four LiDARs. And is, in your opinion, is more just more or is more better? And if more is just more.

Omer Keilaf:
My CFO likes that there is more. I won't fight anyone who wants to buy more, but eventually it's a business case, right? I believe that probably you're referring to companies that are trying to develop Level 4, Level 5 kind of applications. Our key focus right now is mostly Level 2 and Level 3. And even at Level 3, you don't really need more than one LiDAR. I'm happy to sell anyone more, but one is enough. That's what we see also in the market. And the reason that we are focusing on those levels is because those are the levels that we expect to go into volume faster than the rest. Those are the passenger vehicles. You won't see a passenger vehicle with Level 4 anytime soon in the next decade. And that would be a waste of our time and focus. And I think that was also part of our strategy from the first day. We were very certain that the volume would come from passenger vehicles. They cannot absorb more than one LiDAR per vehicle because of the cost, because it's not I in level three or level two. The burden is on the customer they need to pay for a cow. It's very different than or robotaxis were. The owner of the platform is not a single individual. You can justify the cost by the business that you are trying to improve by efficiency. When it comes to a passenger vehicle, eventually cost is dramatically more important than and that cannot be done by using multiple LiDARs.

So it's just a matter of the business case and the use case and the end owner. I can justify ten LiDARs for a truck that saves two drivers that each one cost $100,000 a year, right? It makes a lot of sense when it comes to a private vehicle, it doesn't.

Lei Xing:
Yeah, this is exactly the same response. I got from value and and mercedes when I attended the there a tech day in california when they showed the pro pilot level three with a scholar. That's the same. It's all about business case. And I think you've just answered some of our other questions.

Tu Le:
Yeah, so let me ask you this Omer, because obviously, you have reduced the price of InnovizOne to InnovizTwo by 70%. And in your estimation, once you get to this volume for passenger vehicles, right? That you're talking about, what price point do you see that being at? Is it $300 a LiDAR? Is it $1,000 a LiDAR, wo where it becomes pretty standard, even on mass market vehicles? Where's that price point do you see that the tipping point for adoption gets reached? 

Omer Keilaf:
Sure. So obviously, even with a LiDAR that costs $1, you won’t see it being used on all vehicles in the next probably 5-10 years, right? Because even using a LiDAR is complicated and you need processing power and a lot of testing, not all carmakers are going to be using Level 3 technologies as soon as the others. There are carmakers that are usually adopting technologies faster than the others and they want to use technology faster than the others. And then there is a price point that unlocks, different brands and different categories of cars. So LiDARs don't need to cost $1 today. I think that the next entry point for at least 10% of the market is $500, and that's what we are targeting with the InnovizTwo. Beyond that, the LiDAR would need to go lower, $300 and eventually $100, when you will see a LiDAR on every vehicle, it will be around $100. And then it makes sense. And I think InnovizTwo definitely can get to those levels. It's just a matter of industrialization. I would say hardening the design. The situation right now is that you have different carmakers, each one trying to ask for this or that and some changes, they have some reference to the different things. And it means that when you're doing a design, it's like one of the benefits of a certain design, it’s flexibility without changing too much and having that allow you to tweak and do things. That means that a lot of the design needs to keep a lot of flexibility, flexibility usually goes back like invert, invert to the cost, right? Because if you look at a DVD player 15 years ago, it will be costing, I don't know, a few thousand dollars, you would open up. You'll see many boards and many electronics and flexes and optics. Once the industry has kind of like determined, a certain, I would say, architecture and determine a certain functionality, you start to harden it, you start to make things more turn, like FPGAs turned into ASIICs, and then different ASSICs become one chip, dye over dye, and everything becomes like a chip set that you build in terms of millions or more. And it's a stage in which the old industry is not at, right? But once you get the consolidation of architectures and you will see, let give you an example. Let's say that I get an order now of 15 million LiDARs, okay? I close the deal at a certain price point of, I don't know, $500 or so. Hopefully. Now my target right now is to make more profit of it. So if you think about the ROI let's say I can spend, I don’t know, $20 million in taking this board into something much more lean. I would harden it the most I can, right? I can take all of those components, make a new like a chip set, like the ROI of investment is obviously very positive. It's all about a volume that justifies an investment that allows it a further cost reduction. Moving from InnovizOne to InnovizTwo, we dropped the price by about 70%.

It was mostly around an investment of R&D down around very few components. The next step in which we want to do another cost reduction of products. There are other things that we haven't changed at all. This is InnovizOne. InnovizTwo, if I would show you, you would actually see the same electronics. We didn't change it. It's like, it's actually somewhere else. Of course this doesn't look like a product that costs $100. It's still very flexible. It still allows us. It's about, it's really about making a decision of hardening a certain design, because you want to make sure that the design is solid for many years. And you don't want to harden it as long as you you want to keep some flexibility. But it's achievable. It really is achievable.

Tu Le:
And really, let me quickly follow up Lei. And then you can take the last question. Are you seeing a squeeze on pricing in the short-term like increase because of the supply chain issues and the supply issues? Are you seeing an increase in pricing in the short-term and some of your commodities?

Omer Keilaf:
Yeah, of course right now the market is difficult with components, but the volume of the market, the LiDAR space is not that big to make it any effect. We sell a few thousand LiDARs then it's not that meaningful. And I think that by the time that the market would scale to millions, I don't think that by that time that would be continue to be an issue, right? InnovizTwo also has significantly less components than InnovizOne, really small portion of it, it's one quarter of the components. That's how we managed to get to the 70% cost reduction. So no, I believe that by the time of the inflection point of the car market that would be behind us.

Lei Xing:
One question to confirm, are you not looking at the robotaxi, serving the robotaxi market? 

Omer Keilaf:
We are. Obviously, I'll sell LiDAR to anyone who's willing to pay. We need to spread, we need to put our chips on different players and expect that some of them would succeed and grow and etc. But in terms of focus, we are mostly focused on signing deals like with carmakers that plan to go to high volume. There is no robotaxi right now, company that I can get to an nomination of the volume that I get from a passenger vehicle. That's just too many years ahead of us. I have to keep my focus where I can sign a deal, not just R&D and selling a few samples, ect. Many LiDAR companies, by the way, that's what they do. You see a lot of announcements. The market is, there is so much noise in the industry. People don't see the difference between a series production announcement than here's a partnership with a company, I just sell them like 10 units and we did the press release together. Everything looks the same. That's something that's part of the noise that we're dealing with. But what can I do?

Lei Xing:
One more question from me, and then Tu, you can end with the Chinese smart EV startup question? So what is the end game for LiDAR?

Omer Keilaf:
LiDAR would be on every car, but much beyond that. A LiDAR is eventually a 3D camera, right? People tend to connect LiDAR equals a car. It's not. A LiDAR is a 3D sensor, so if I would tell you I developed a camera, I don't think that the only thing that comes to your mind would be a car. Even though cars are using many cameras, there are many applications for 3D sensing. I actually think that long term, any 2D camera would be replaced with a 3D camera, because there are many benefits of getting another layer. And interactions would be different. And we all talk about metaphors and augmented reality. Those are all relying on 3D data.

Tu Le:
Well, iPhones already have LiDAR in them, right?

Omer Keilaf:
Yeah, definitely. So LiDAR is a strange name for a 3d camera. Really, if you think about it, the VR goggles, the Oculus, right? And I know if you've used it before, it's quite amazing. Eventually it's a 3D display, right? You actually consume 3D data. And the thing is all of the data that is available in those goggles is completely synthetic, because there is no way to film 3D today in a good way. So everything that you are able to consume today in those VR are mostly synthetic. A LiDAR is a 3D live camera, which you could actually use. So I could have actually placed a LiDAR here in my room. You would be wearing VR goggles and actually see the room and travel in it because it makes it accessible. So LiDAR is just the first market which will actually consume a lot of LiDARs in automotive. But there are many other markets that the adoption of LiDARs would translate to large adoption, surveillance and construction sites, farming. We live in a 3D world, so using a 3D sensor somehow makes more sense than a 2D, and it will grow fast.

Tu Le:
Omer, final question here. I don't know how closely you've been following the Chinese EV sector, but who do you think has a chance of becoming the first Chinese EV maker to go global to be successful in Europe and in the U.S.?

Omer Keilaf:
To be honest, I hope I'm not saying, I've been following obviously all of the big names and we hear a lot about Baidu and Pony recently several news and of course I'm sure there are many other players that are very serious. And I don't want to make a bet, because I'm probably kind of not for sure, I don’t want to insult anyone for not mentioning him. So I'll pass on that question.

Tu Le:
So if you want to have an offline conversation about the Chinese EV landscape, we can tell you about it. Just give us a holler, but, so thank you for your time. This discussion was great. I learned a ton and we're glad you were able to join us.

Omer Keilaf:
Ok guys. Thank you very much.

Tu Le:
Thanks again, Omer.
 
 Lei Xing:
Thank you for your time.

Lei Xing:
Hi, your co-host Lei Xing here. There is no doubt that LiDAR has become a prime-time product or technology in the smart EV era, driven in large part by Chinese companies like RoboSense, and Hesai which are finding their products increasingly equipped onto production vehicles and robotaxis in China and across the world in large quantities. “Faster than mushrooms after the rain,” as Omer puts it. For Western LiDAR suppliers like Innoviz, this is almost like David going up against Goliath because the Chinese companies right now have the advantage on cost, go to market speed and local knowledge, yet many of them have only been around for not much longer than Innoviz. Listening to Omer, you can feel the drive, determination and confidence coming from his military background, and obviously, by becoming a direct LiDAR supplier to a multi-branded, multi-regional global automaker, that gives Innoviz a huge leverage as far as China is concerned. It is also interesting that Innoviz is working with Chinese partner LiangDao Intelligence, which recently launched its own solid-state flash LiDAR. Will these partnerships translate to success in China? I guess time will tell and we wish Omer and his team the best of luck.  

Tu Le:
Lei and I will be sharing more of conversations with the men & women around the world moving the EV/AV mobility sectors forward as part of this China EVs & More MAX series. Some folks will be instantly recognizable, but some will just be people that are doing amazing in the space that we think deserve to be highlighted.

Don’t worry though, Lei and I will continue to host our live weekly China EVs & More Twitter Spaces room that summarizes that week’s most important news coming out of the China EV, AV and mobility space. For those that can’t catch the live show, you can find the China EVs & More

 

 

(Cont.) MAX Episode #7 (ME) - Omer Keilaf, Co-founder & CEO of Innoviz