Oculii Corp. (AMBA) Earnings Call Transcript & Summary

Hello. Thank you for standing by, and welcome to the Ambarella's acquisition of Oculii conference call. [Operator Instructions] Please be advised that today's conference is being recorded. [Operator Instructions] I would now like to hand the conference over to your speaker today, Louis Gerhardy. Please go ahead.

Yes. Thank you, Josh, and good afternoon, everyone. Thank you for joining our call to announce Ambarella's acquisition of Oculii. With me on the call today is Fermi Wang, President and CEO; Les Kohn, CTO; and Casey Eichler, CFO; and Oculii's Co-Founder and CEO, Steven Hong, who will become Ambarella's Vice President and General Manager Radar Technology when the acquisition closes. The primary purpose of today's call is to provide you with information regarding the acquisition and its expected impact on Ambarella's business. The discussion today and the responses to your questions will contain forward-looking statements regarding expected timing for completion of the transaction, anticipated benefits of the combination of Oculii's Radar Technology with Ambarella's visual AI technology, projected financial results, financial prospects, market size and growth and demand for our solutions, among other things. These statements are subject to risks, uncertainties and assumptions. Should any of these risks or uncertainties materialize or should our assumptions prove to be incorrect, our actual results could differ materially from these forward-looking statements. We're under no obligation to update these statements. These risks, uncertainties and assumptions as well as other information on potential risk factors that could affect our business operations and financial results are more fully described in the documents that we file with the SEC, including the annual report on Form 10-K filed on March 31, 2021, for fiscal year 2021 ending January 31, 2021, and the Form 10-Qs filed on June 8, 2021, and September 8, 2021, for the first and second quarters, respectively, of fiscal year 2022. Access to our quarterly results, press releases historical results, SEC filings can be found on the Investor Relations portion of our website. Fermi will now provide an overview of the acquisition, key terms of the transaction. And then, Fermi, Les, Casey, Steven and I will be available for your questions. With that, I'll turn it over to Fermi.

Thank you, Louis, and good afternoon. Thank you for joining us on such short notice. We are very excited to announce an agreement to acquire Oculii, a privately held company based in Ohio for $307.5 million. Oculii has developed a proprietary AI software platform that enables radar to have significantly higher levels of resolution, longer range and greater accuracy than existing high-performance radars, providing higher levels of safety and reliability at a very economical price points. In other words, this is a software solution to a hardware problem. And this software platform can run in the performance envelope of our current and future CV SoCs as well as the existing radar SoCs in the market today. The acquisition has 2 key strategic implications. First, Ambarella's addressable market will expand as we can now provide a software layer into radar using automotive ADAS and autonomous driving applications, but also other IoT market like mobile robotics and surveillance. Based on ABI data, the vehicle radar market is around 90 million units in calendar year '21 and growing to 240 million units in calendar year '26. And as Oculii technology penetrates this market, we expect to generate software licensing revenue. The second implication, the most important one, long term, is that by combining Oculii's radar software stack with Ambarella's video AI technology, we expect to be able to unlock levels of perception that could not be achieved with the traditional discrete camera and the discrete radar approaches coming in the market today. So in the long run, we expect Oculii to not only generate software licensing revenue, but also increase our SOC market share and strengthen our SOCs ASPs. As you know, Ambarella has always pushed an algorithm-first approach in the camera market, and that is exactly what the Oculii team has accomplished in the radar market, where they are high regarded in the industry. With more than a dozen commercial engagements with leading OEMs and Tier 1s, some of whom have invested in Oculii. There are already a number of strong endorsement and more to come. The 2 companies have very similar cultures, and we are very excited to commence the collaborative camera and the radar technology development. Now I will provide some comments on the transaction details. We have signed a definitive agreement and the transaction is expected to close during our fourth quarter of fiscal year 2022. The acquisition will be 100% funded from cash on Ambarella's balance sheet. As a reminder, our cash, cash equivalents and marketable debt securities totaled to $449 million as of the end of our second quarter of fiscal year '22. Oculii is expected to earn $3 million to $4 million in revenue during calendar year '21 from preproduction programs. With the expectation this level of preproduction business will increase in calendar year '22. The initial production program for Alkali are expected to commence in calendar year '23. Oculii's operating expense are approximately $1.6 million to $1.9 million per quarter, and we plan to continue to grow our investment into this technology. During our Q3 earnings call, at our January 4 Capital Market Day, we expect to provide additional information on acquisition. We do not expect to break out Oculii data after this point in time. At this time, I would like to turn the call back to the operator to poll the questions. Thank you.

[Operator Instructions] Our first question comes from Gary Mobley with Wells Fargo Securities.

Congratulations on the acquisition. I know Oculii received a lot of good trade press and seems to be a good target. But my question really centers on the go-to-market strategy for Oculii and part of the question is, do you intend to license this technology, these algorithms on a merchant basis to perhaps even some of your competitors? Or -- and as well, does Oculii productize in the form of a radar transceiver chipset today, like a 4D or radar transceiver? And then moving forward, will just, again, strictly just be software licensing?

Yes. So -- at the beginning, we'll continue the current business model with Oculii, which is software licensing to OEM, the Tier 1s as well as selling the modules that they already build. And this will continue to do this business model. Moving forward, we expect that we will continue to with existing incumbent silicon provider, which is working on the RF technology. And we are expecting to start putting Oculii software into our current CVflow chip -- current chip and the future chip so that we will continue -- so we can provide a combined video and radar perception moving forward. But we do expect to continue to work with the partners like TI, [ ASP ], Infineon in the RF chip side.

Okay. And so just to be clear, this allows for 2D, 3D radar transceivers to work as if they function as a 4D radar transceiver developed by the likes of, say, [ Uhnder or RPU ] or something like that.

Yes. And Steven, can you help me on that?

Yes, certainly. Well, great to meet, and thanks for the question. So this is Steven, the CEO of Oculii. And as Fermi mentioned, our core technology is software that actually can run on any hardware platform to enable significant expansion and enhancement of performance, including angular resolution. Traditional radars need new RF chips because historically, resolution has been tied directly to the number of antennas and active transceivers that are in the system. Our software effectively breaks this trade-off by using a dynamic adaptive waveform to use software to virtually create the same effect. And so as a result, this software can run on any radar hardware with any RF front end and effectively multiply its effective resolution by 100x or more. The beautiful part of this software is that as it runs on better and better hardware, of which Ambarella is leading and developing the cutting-edge silicon, the software can enhance the resolution and performance even further. And so this allows us to have both a business model to address the existing radar solutions on the market with the existing RF and digital solutions, but also work closely with Ambarella to deliver an optimized solution for the future when performance, safety, sensitivity and resolution are of utmost importance in these systems.

Got you. Appreciate the color. And as a follow-up question for the Ambarella team. The last sensor to fuse into a fusion process or, of course, will be LiDAR. Do you feel that as a necessary element to have a full fusion processor to complement radar and computer vision?

Les, do you want to answer that question?

Sure. So yes, we will support people that want to use LiDAR and to fuse that with radar and camera. But we are excited about the potential of combining our high-definition stereo with the high-definition Oculii radar to see how far we can push that to potentially minimize the need for LiDAR.

Our next question comes from Jeremy Kwan with Stifel.

Yes. And let me add my congratulations on the transaction. Just a question first on the -- you mentioned addressing, in addition to automotive, the robotics and security markets. Is this something that you plan to do in parallel with automotive? Or is it something that you're going to stagger? And when can we expect to see some engagement on those fronts?

Well, we definitely plan to do in parallel. Obviously, automotive represents the biggest opportunity. But with our momentum on the security camera. You can imagine after this meeting, we're going to kick off process to asking our marketing sales to engage all our current existing customer in robotic and as well as the security camera business and see whether we can identify customers who are looking for the combination of video and radar perceptions.

And how big can these markets be? Is automotive is still going to be the largest by far, but can you give us a rough ballpark of where you see these -- the other 2 growing to be?

Yes. This is Louis. So we mentioned the automotive market in calendar year '21, the various radar systems there, whether it's corner, front or even in cabin is around 90 million units, growing to roughly 240 million. And just conservatively looking at the other markets. And if we just kind of narrow it down to mobile robotics and the surveillance market, that could easily add another 5 million unit up to those numbers.

Got it. And in terms of the cost savings that your solution will -- the [ IQ ] solution will enable for your Tier 1 OEMs, can you help us quantify that a little bit? What kind of cost savings that you can enable for them? And also, what kind of potential ASP bump that you can see on your solutions once you integrate the full stack?

Well, if you compare -- if you look at the current traditional setup for the camera and the radar system is really -- on the camera side, you have SoC to do all the video processing, then you have other different SoC to do a radar processing. But then in the future with Oculii and Ambarella combined solution, you can imagine that on the digital side, all the processing will happen in 1 unified chip. But we still need to interface with RF chips that are required for any radar system. But you can imagine on the digital side, that we can save some significant [ BOM ] and also associate DRAM power supply and so on. So I think we didn't come out a number because it really depends on how many radar you need, what application. So the range can be quite wide. But just think about doing ADAS, we can easily remove probably, I would say, more than $10 of [ BOM ], just for a 1 channel of radar systems.

Your next question comes from Kevin Cassidy with Rosenblatt Securities.

Congratulations. Just to understand that as you move toward the fusion on 1 chip, would that require your CV5 -- or 5-nanometer, do you need your next generation to have the processing power to do this?

Les, do you want to answer that question?

Yes. So one of the great things about the Oculii technology is that it's very scalable. So we can actually run a version of the processing on our current CV2 family chips, and we plan to do that. But at the same time, as we scale up performance with our next-generation chips, we'll be able to scale up the capability significantly.

Okay. Great. And just for competitive reasons, did Oculii -- you have a discussion with your customers that may be competitors to Ambarella and they're going to remain customers. I just want to know how those conversations went.

Well, we haven't talked to any people outside the company for -- just for the confidential reasons, but we definitely will engage in discussions with them as quickly as possible.

Okay. Great. And just the -- you said $1.6 million to $1.9 million per quarter, which really isn't very significant compared to what you're spending currently. But is that mostly software engineers? Or how does that spending?

So the whole company has roughly 50 people, most of them are engineers, just like Ambarella, that's why we said the culture is so similar. And most of the engineers sitting in Ohio and some engineers in China. So overall, that's the current situation. But we do -- like I said, when we acquired this -- acquired Oculii, we do see the potential and we definitely will continue to invest on that.

Our next question comes from Matt Ramsay with Cowen.

Congratulations all around. I guess my first question is for Les. Les, correct me if I screw anything up here, but my sense of this is you'll be able to run the Oculii software on the ARM cores or something on your current product? And are you able to benefit from that? And kind of as you guys did back in the day, dissecting the video algorithms and then building specific accelerators on your chip to get much, much better performance than you did with human video back in the day, that the idea would be in future versions of the chip to do the same type of hardware acceleration for the Oculii algorithms as you move forward. Do I have that right?

Yes, that's exactly right. And in addition, there's a lot of opportunity on the software side on the perception processing to very tightly integrate the radar processing algorithms with the vision processing algorithms to provide further boost in performance.

Got it. Perfect. Just a follow-up question for the team. I know that I guess I would imagine that the majority of the focus here will be on automotive. But can you talk about some of the opportunities that might exist in robotics, security, other verticals outside of automotive? And I guess what percentage of the focus might be in those verticals versus the primary ones in automotive? Congrats.

Thank you. So there are two things. First of all, even at the security camera, you can see that the integration of radar already started. In fact, some of our customers already announced their products as a security -- whole security camera -- professional security camera, but they already bundle discrete radar solution in there and just to improve the accuracy of the perception. So the trend already started, in fact, that if you go through some customers' website, you see they are promoting that. So we do see that trend will continue because the -- especially for the outside security camera that -- because of the weather condition, because darkness and sometimes radar definitely is very complementary to optical sensors. So that's definitely one area we think we can immediately help and identify the proper customer to engage. On the robotics side, in fact, there are a lot of slow-moving robots for different applications like moving objects or doing inventory control. And you can just view that as an automotive application because the requirement is basically similar at a much lower speed. So that also is a place which we will require a very high-quality and high-accuracy perception, which usually comes with sensor fusion type of technology.

Our next question comes from Andrew Buscaglia with Berenberg.

I was interested in the -- you're talking about the acquisition helping generate software licensing revenue, and you made the comment about strengthening ASPs. Have you kind of run some scenarios of what this means for -- broadly for your ASPs and how -- how and when do you expect that to hit your revenue stream?

Yes. So Andrew, this is Louis. We've got -- two ways this can impact. There's the software licensing revenue. And then the other way is when the algorithms, the software layer is incorporated into our chipsets, incremental processing power may be required to support that, and that's what would drive the ASP. As you know, we've got a very rich new product road map ahead of us, things we can't really talk about yet. But this software layer will very gracefully use up much of the processing power in those -- in that road map.

Okay. Interesting. Another question I had is very simplistically, why now. So you -- the company -- I think you said generating $3 million or $4 million in revenue. What's going to be excited to make this deal now or I guess, more so to wait if you could eliminate some timing risk if, I don't know, if things don't transpire the way you expect?

All right. Thank you. I think that's a very important question. In fact, that I would say, tracing back to last year where we rationalize our thinking, we realize that we're going to do well with our computer vision silicons and we start seeing the first sign of that revenue growth. We ask ourselves what's -- if we have a chance, what's our next technology we need to integrate into our platform. And it becomes very clear that moving forward for all our target markets, radar becomes the most meaningful as well as most important sensor modality for us. So that's when we start looking at the candidate. And throughout the multiple studies, and we identify Oculii being the best in the market that we can find. And for us, just like what we -- because we talk about this algorithm-first approach, very quickly, we realized that [ state ] as a partner doesn't help each other because we need to understand their algorithm, so we can modify our hardware and silicon, accordingly, to make the combined solution better. So it's the algorithm-first approach that both sides appreciate basically reinforce the need that we need to work together as a company, instead of 2 separate companies as partners. And that's why we want to move as quickly as possible so we can start pushing the combined road map out to the market, which will help us -- help both of us.

Our next question comes from Martin Yang with Oppenheimer & Co.

And my first question is regarding the ideal silicon partner for Oculii. So before the deal, what would be considered the most ideal silicon architecture for the software?

Well, Steven, you want to answer that?

Yes. I mean to be honest, if you could have painted the perfect silicon partner for us, it would have been a company like Ambarella. It would have been a company that has leading edge, latest technology capabilities, one that's executed at every advanced technology node to deliver the most advanced silicon. It would have been one that had tremendous vision expertise so that we could leverage our radar technology on that same platform to effectively come up with not just a discrete solution for radar and camera separately, but one that integrated them in a very, very low raw level to enable higher performance, more efficient, more compressed networks. And then I think you look at the third piece, you wanted a company that's growing and going into this market very aggressively, and that's Ambarella as well. Ambarella obviously, has been making tremendous strides in the automotive industry as well as the security and the IoT robotics applications. They're on the rise and just with tremendous traction. And so the way that we see this is this is a rocket fuel for the rocket that they've already built. And we're very, very excited. And so we really couldn't have dreamed of a better partner than Ambarella.

Got it. I have one more question regarding the applications, especially for robotics. When I think about the combination of vision and radar, it seems like drones already have that, some of the drones on the market. So do you think that's, say, like a near-term opportunity for you besides automotive that where you can go commercial in 2023? Or is that somewhat maybe too small market for you?

Well, yes, thank you. In fact, we will into markets. In fact, we are still into a market. DJI is still using our chip in some of their drone models. For us, the size of the drone market is limited, particularly on the consumer side. And I think now you're seeing a lot of drones having a first vision and then radar, most of them are in the commercial side. I definitely think that when the market size grows in the commercial drones that become more interesting. But however, the current size, we were in that market. We know how big that is. And I just don't think that's big enough compared to the opportunity we have today.

Understood. Congrats on the deal.

Thank you.

Our next question comes from Suji Desilva with ROTH Capital.

Fermi, Les, Steven, congratulations on the deal. A couple of questions. The architecture of CVflow, Fermi, or Les. Is it really already designed to incorporate radar data conceptually? Or is it -- as Matt maybe asked earlier, does it need to have some modifications over time to optimize for radar data flow versus the imaging, multiple imaging cameras that you've used together already?

Les?

Yes. So as I mentioned, it is possible to run the Oculii software on our current CV2 family using even just ARM processing. But going forward, we do see more and more of the processing or the bulk of the processing moving to the CVflow side of things, and we will make further enhancements to CVflow to optimize that.

Okay. Great. And then a question maybe for Fermi. I mean the conversations you've been having with automotive customers as you try to penetrate there. What are the elements of those conversations that motivated you to incorporate radar into your product portfolio?

Well, in fact, everybody we talk to, they all have different type of radar program in-house. And in fact, some of them already asked us not recently, but in the last several quarters, that have as -- our future road map will support radar. And I think that definitely is a driver for us to do this deal, but it's not the only driver for us to do this deal. But definitely, that we will see most of the Tier 1 OEMs, if not all of that will use radar as a sensor modality in the future.

Maybe one quick last one. As you get all these different sensors you're fusing together now camera radar, perhaps LiDAR, can you talk about the edge function of perception and object detection versus it being centralized in video? Does that dynamic change at all with this kind of acquisition for your part?

Les, do you want to give a try?

Yes. So yes, we -- in talking to these automotive customers, we definitely see a trend to want to move radar processing from the edge to a central domain controller and take advantage of the fusion opportunities there.

Our next question comes from Richard Shannon with Craig-Hallum.

I jumped on late. So I apologize if I'm repeating what you talked about before here, but in terms of the support of other hardware radar chips, is there -- is everyone included, addressed your TI, Infineon and some of the new start-ups here, they all address are there ones that you can't support now or you don't intend to support in the future?

No, we do. Yes, we answered that. We do -- we will continue to support the current business model of Oculii, which is using TI, [ ASP ] or Infineon chip for their current software licensing as well as for the module, which will continue. And even in the future, I think we still need to have a partner on the [ RF mixed single size ], which we hope that we can continue to work with this company as a partner on that -- on that front.

Okay. That's helpful for me. Second question is in the automotive space here in terms of the kind of the levels of driving here L1 to L5, is there any overlaps that you do or do not have with that in general? It sounded like with the performance you're expecting here is certainly going to cover the high end of the application set here, but how about the low end as well if that's something you expect to intersect with?

I think Oculii software is scalable. It can go to Level 1 plus Level 2 all the way to Level 5 is really about how many -- what's the algorithm and also the performance require. So the current CV2 family chip can probably address ADAS Level 2, Level 2 plus. But moving forward, you need to probably next generation of a CV chip to address Level 2 to Level 3 and 4. So I think that's the scenario we're looking at. And we definitely believe that the Oculii software not can only scale inside the automotive business but also can scale to the security camera business.

Okay. That is helpful. Last question for me. If I missed it, apologize, but do you have any commentary on the commercial engagement pipeline? Any wins out there -- any way you can address where you're seeing wins, how broad they are, or anything like that, that would be great.

We haven't addressed that yet, but we did talk about that we have engaged with more than 15 Tier 1 OEM at this point, and we have some products will go into production in calendar year '23. And we will provide more details later in our Analyst Day on [ January 4 ].

Thank you. I would now like to turn the call back over to Fermi Wang for any further remarks.

Thank you. And in conclusion, this is Ambarella's second acquisition, and we expect it to be just as impactful as the first one. And we look forward to sharing more information with you including our Q3 fiscal year '22 earnings call and our January 4 Capital Market Day at our CES Hotel in Las Vegas. Until then, thank you very much. Looking forward to see you next time.

Thank you. Ladies and gentlemen, this concludes today's conference call. Thank you for participating. You may now disconnect.