Ambarella Inc (AMBA) 2021 Q3 法說會逐字稿

Ladies and gentlemen, thank you for standing by, and welcome to the Ambarella's Third Quarter Fiscal Year 2021 Earnings 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 host, Louis Gerhardy, Corporate Development. Please go ahead.

Thank you, Sarah, and good afternoon, everyone. Thank you for joining our Third Quarter Fiscal Year 2021 Financial Results Conference Call. Calling in today from different locations will be Dr. Fermi Wang, President and CEO; and Casey Eichler, CFO.

The primary purpose of today's call is to provide you with information regarding the results for our third quarter of our fiscal 2021. The discussion today and the responses to your questions will contain forward-looking statements regarding our projected financial results, financial prospects, market growth and demand for our solutions, among other things. These statements are based upon information available today, and are subject to risks, uncertainties and assumptions. Should any of these risks and 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 financial results, are more fully described in the documents we filed with the SEC, including the annual report on Form 10-K filed on March 27, 2020 for fiscal year 2020 ending January 31, 2020, and the Form 10-Q filed on September 9, 2020 for the second quarter ending July 31, 2020, and Form 10-Q filed on June 8, 2020 for the first quarter of the fiscal year ending April 30, 2020.

Access to our third quarter fiscal 2021 results press release, historical results SEC filings and a replay and prepared transcripts of today's call can be found on the Investor Relations portion of our website. With that, I'll turn the call over to Dr. Fermi Wang.

Thank you, Louis. Good afternoon, everyone, and thank you for joining us this holiday-shortened week.

We are pleased to see our multi-year visual AI investment as the major force driving our accelerated business momentum which is apparent in our Q3 results and outlook.

Q3 revenue of $56.1 million was above the high-end of our original guidance range. CV led the way but we also faced stronger than expected demand for video processors, despite the persistent geopolitical and public health risks. Gross margin of 62.7% was above the high end of our long-term model for the second consecutive quarter, reflecting a favorable customer mix and continued operational execution.

CV revenue was slightly more than 10% of total revenue in Q3 and we continue to expect it to be 10% for the full year. With an ASP about 2x our non-CV ASP, CV units were less than 4% of our total units shipped, and an even smaller portion of the installed base, highlighting the very early stage of this opportunity and the significant multi-year headroom for growth. We anticipate CV revenue will present -- represent about 25% of our total revenue in fiscal year 2022 and we have a flattish outlook for video processors.

The megatrends for security, safety and automation are very favorable, and the pandemic appears to be accelerating this digital transformation. To support this anticipated growth, we continue to build our team globally and we intend to further expand our presence around the world to support the rising interest in our CV SoCs from both existing and new markets.

I will now talk about our markets and the customers. As we have previously mentioned, Chinese IP security camera makers have become concerned about the continuity of supply from existing Chinese suppliers and have been evaluating alternative camera designs on other solutions. Ambarella's CVflow AI SoCs have won a number of design wins based on their AI performance, high-quality imaging and low power. We have now received orders for shipment beginning in the fourth quarter of this year, including orders for our CVflow SoCs from Dahua, China's second-largest security camera company.

Outside of China, we are continuing to see customers introduce new IP security cameras based on our solutions including our CVflow AI SoCs. During the quarter, Johnson Controls launched its new Tyco Multisensor Camera. Based on Ambarella's S2E SoC, the camera is equipped with four separate image sensors to provide up to 360 degree coverage via four separate nonstitched video streams. Each of the four sensors are able to capture 4MP images, enabling the camera to cover an area which might normally require four separate high definition video surveillance cameras.

Honeywell introduced its 60 Series IP cameras including 5MP indoor and outdoor dome, bullet and outdoor speed dome models and a 2MP outdoor speed dome model. Based on Ambarella's S5L SoC, the 60 Series offers increased resolution and provide fast notification and verification of potential threats and responses.

And FLIR introduced its Elara thermal camera for accurately measuring skin temperature at medium- to high-throughput entry control points. Based on Ambarella's CV22 AI SoC, it is equipped with one -- with on-edge, intelligent face detection and issues on-screen prompts to individuals when they need to remove glasses while also guiding them to the correct position for best measurement.

And also Panasonic iPro continued to expand its portfolio of Ambarella-based models with the introduction of its Vehicle Capture Camera, the WV-X5550. Based on Ambarella's CV22, the camera can capture clear images of vehicle occupants on one channel and license plates on another, even in challenging lighting environments, including night time.

In the body-worn camera market, Panasonic iPro introduced its BWC4000 camera. Based on Ambarella's CV25 SoC, the BWC4000 records 1080P video and has 12 hour battery life.

Also during the quarter, Motorola introduced its VB400 body-worn camera aimed at security professionals and based on Ambarella's low power SoC. Integrated with Avigilons Control Center the VB400 can augment fixed video deployments by displaying first-person perspective video from the body-worn camera alongside fixed cameras for superior situation -- situational awareness.

During the quarter, Comcast announced its new Xfinity Home offering called Self Protection designed for consumers who want high-quality security cameras, but who may not need a professionally installed and monitored whole-home security system. The new Xfinity's HD home security camera is powered by Ambarella's S3LM SoC and is built for both indoor and outdoor use.

And this month, we introduced our CV28M camera SoC, the latest in our CVflow family, supporting the introduction of advanced AI features targeting a variety of cameras in new high volume markets. Its highly efficient AI processing and low power will enable a new class of smart IoT devices for applications including smart home security, retail monitoring, home robotics, and occupancy monitoring.

For new AI sensing applications, like retail monitoring or occupancy monitoring, the SoC provides AI performance to make all decisions in the camera preserving privacy and avoiding heavy video processing running on back-end servers. In home robotics applications, the CV28M can be connected through a wide range of sensors such as visible, structured light, and time-of-flight, to capture and then process, the data required for navigation.

And we are currently sampling CV28M to our customers. The CV28M software compatibility with existing CVflow families is allowing customers to enter production very quickly, and we have already received mass production orders for shipment beginning in Q1 next year.

In the automotive market we are continuing to make great progress as we promote our AI solutions for a variety of automotive applications including ADAS and AD for customers worldwide. Today, we are announcing our total automotive revenue funnel for the first time, and we intend to update it at least once every year.

Based on our best judgement, we currently estimate our automotive revenue funnel model is about $600 million with a majority represented by CV for a variety of ADAS and AD applications. The auto funnel revenue runs through fiscal year '27 with some of the more significant programs commencing production in calendar year '22 or '23. And this funnel covers business we have won as well as a pipeline of business we are currently competing for.

We are seeing shipments of our CVflow-based ADAS solutions for commercial vehicles in China continuing to ramp with customers including Yutong buses Shanqi trucks and Dongfeng trucks. We are currently the market leader in the OEM car recorders category with customers including Nissan, Honda and Toyota in Japan, and Great Wall, Geely and BYD in China. We are now seeing a recovery in orders for OEM car recorders in both regions, as car manufacturers begin to ramp production back up following the impact of the pandemic.

And also, we are pleased to announce that we have our first design wins for our CV2FS automotive functional safety SoCs, which we introduced during CES at the beginning of this year. These SoCs include the AI processing and ASIL features required for safety critical ADAS applications.

During the quarter we signed a development agreement to supply our CV2FS SoC for level 2+ ADAS systems with an OEM manufacturer of electrical -- electric vehicles. In this application, the CV2FS provides the AI processing performance and flexibility to support a wide array of driver-assistance capability to improve safety and convenience.

Also during the quarter, we signed a new development agreement with a tier 1 supplier of electronic mirrors for the European commercial vehicle market based on our CV2FS functional safety AI SoCs. The CV2FS was chosen for its ability to process video from multiple cameras and perform advanced AI processing to support features such as vulnerable road user detection and prediction.

And we are continuing to win new designs for fleet management solutions for commercial vehicles as well as ride-sharing and taxi services. With 140 million commercial vehicles in the US and hundreds of millions worldwide, the ability to retrofit existing commercial vehicles with cutting-edge driver monitoring capabilities represents a major opportunity for Ambarella. Our CVflow AI SoCs have the required performance to support front ADAS applications, active driver monitoring and video recording.

And in November, Ambarella announced a joint reference design for automotive camera applications that simultaneously monitors drivers while capturing vehicle occupants for one-way videoconferencing. The reference design uses software from Smart Eye, a world leader in developing AI-powered eye, mouth and head tracking technology and the image sensors from OmniVision Technologies. Ambarella's CV22AQ SoC simultaneously processes both RGB color image and IR and runs Smart Eye's algorithms to analyze the driver's state and alert the vehicle to any unsafe indicators, such as drowsiness.

With the holiday season approaching, there have been a number of new consumer product introductions featuring Ambarella SoCs.

During the quarter, DJI introduced its Pocket 2 hand-held camera, equipped with a 3-axis motorized gymbal to stabilize movement. The Pocket 2 is based on Ambarella's H22 SoC and shoots 4Kp60 video, takes 64 megapixel photos and includes 8x zoom capability.

And On November 4, DJI introduced its Mini2 drone, the replacement for the popular Mavic Mini model. Featuring Ambarella's H22 SoC, the Mini 2 increase its maximum video resolution up to 4K at 30 frames per second while also shooting 1080p video at up to 60 frames per second.

Also during the quarter, Insta360 introduced its OneX2 pocket camera. Based on Ambarella's H22 SoC, the OneX2 includes dual lens 5.7K capture for high resolution 360-degree images, H.265 encoding, AI-based editing and advanced image stabilization.

In summary, we have heard -- you have heard today more evidence on the broad and expanding adoption of our SoCs in automotive and security camera -- security camera markets.

In the professional security camera market our CV momentum, already strong, should build as we have added new CV customers in Asia, and we expect to see CV become material in the home security camera market over the next year. In auto, our CV traction is strong and a significant part of our expanding funnel. In addition, we continue to develop other IoT markets including enterprise access control, smart locks and counting/occupancy sensors, and we look forward to providing you with updates on our progress.

We are making significant progress in our multi-year transformation to a Visual AI company, and the return on the investment is ramping. I'm proud and would like to thank and acknowledge all of our employees worldwide for their contribution to our leadership position in the market, and for their execution in the face of the turbulent environment. I am also thankful for the support of our customers, vendors and shareholders during these times.

I will now turn the call over to Casey who will give you more details about what we are seeing and expect for the business.

Thank you, Fermi, and good afternoon, everyone.

Today, I will review the financial highlights for the third quarter of fiscal year 2021 ended October 31st and provide a financial outlook for our fourth quarter.

During the call, Ill discuss non-GAAP results and ask you to refer to today's press release for a detailed reconciliation of GAAP to non-GAAP. For non-GAAP reporting, we have eliminated stock-based compensation expense adjusted for the impact of taxes.

Our revenue of $56.1 million was above the high-end of our original guidance. This represents an increase of 12% from Q2 and a decrease of 17% when compared to the same quarter of the prior year. In Q3, on a sequential basis, Automotive revenue and Other increased while Security decreased.

Non-GAAP gross margin for Q3 was 62.7% compared to 62.4% in the preceding quarter and remained above the high-end of our guidance, primarily due to customer mix as revenue from the two largest professional security companies in China remained at low levels. Margins were also supported by continued operational execution in a challenging environment.

Non-GAAP operating expense for the third quarter was $32.4 million, compared to $30.2 million in the previous quarter. OpEx was slightly above the mid-point of our guidance range and the sequential increase was primarily due to additional chip development costs and higher employee-related costs outside the US as normal activity resumed.

Other income of $673,000 was due primarily to lower interest income as rates continued to remain low.

Non-GAAP net income for Q3 was $3.3 million, or $0.09 per share compared to net income of $2.1 million, or $0.06 per share in the second quarter. In the third quarter, the non-GAAP earnings per share was based on 35.8 million diluted shares as compared to 35.4 million in the prior quarter.

Total headcount at the end of the third quarter was 766 with about 81% of employees in engineering. Approximately 69% of our total headcount is located in Asia.

Cash and marketable securities were $423.6 million, up from $410.7 million at the end of the second quarter. In Q3 we had positive operating cash flow of $13.1 million.

Total accounts receivable at the end of Q3 were $24.1 million or 39 days sales outstanding. This compares to $23.3 million or 42 days sales outstanding in the prior quarter. Net inventory at the end of the quarter was $23.7 million compared to $23.9 million at the end of the previous quarter. Days of revenue decreased to 102 days in Q3 from 109 days in Q2.

We did not purchase any shares in the third quarter. In May, Ambarella's Board of Directors approved an extension of the current $50.0 million repurchase program for an additional 12 months ending June 30, 2021. As of today, there remained $49.0 million available for repurchase.

We had 2 customers over 10% of revenue in Q3. WT Microelectronics, a fulfillment partner serving multiple customers, came in at 62% of revenue and Chicony, a Taiwanese ODM who manufacturers for multiple customers, came in at 18%.

I will now discuss the outlook for Q4 FY '21. Our visibility has improved, but we remain on guard for risks related to the pandemic and geopolitical factors. Furthermore, as the semiconductor industry supply chain tightens, our operations teams remain vigilant in their efforts to manage manufacturing lead-times and on-time deliveries. We have seen some customers' request order pull-ins and orders placed within our lead-times. CV design activity is at record highs.

In our prior earnings calls, we estimated 2 professional security camera customers in China had pulled in $4 million -- excuse me, $10 million of revenue from -- into fiscal year '21 from fiscal year '20. In addition, the revenue for processor inventory is being reduced. We anticipate orders to remain weak, offset by some degree of a ramp in computer vision SoC orders.

During Q3, these 2 customers combined and remained in the mid-single digits as a percentage of revenue, which will likely be the trough for the foreseeable future. Based on these factors, and our best ability to estimate today, we anticipate total revenue for the fourth quarter ending January 31, 2021, to be in the range of $56 million to $60 million. Automotive revenue is benefiting from an early CV ramp and a post-pandemic rebound.

We anticipate auto revenue will be up at least 20% in both a sequential and year-on-year comparison. We anticipate security camera revenue will be up in the low double digits sequentially. Following a normal seasonal pattern, other revenue is expected to decline sequentially.

We anticipate Q4 nonoperating gross margins to be between 59% and 61% compared to 62.7% in the third quarter with product and customer mix, the primary driver for this change.

We expect non-GAAP OpEx for the fourth quarter to be relatively flat and in the range between $31 million and $33 million. Q4 other income should be modeled at $0.5 million. The Q4 non-GAAP tax rate should be modeled at approximately 10%. We anticipate our diluted shares for the fourth quarter to be approximately $36.3 million.

Ambarella is registered to participate in virtual conferences in Q4, including Wells Fargo, Imperial Capital, UBS, Barclays, MKM and Needham. Ambarella will also be hosting virtual CES meetings between January 11 and January 22. Please contact Louis for more details on these events.

With that, I'd like to turn it over to questions. Operator?

(Operator Instructions) Our first question comes from the line of Joe Moore with Morgan Stanley.

Congratulations on the quarter. I wonder if you could talk a little bit more about the funnel, the $600 million, I believe, is the number. Can you compare that to the $200 million, which was sort of secured design wins from a quarter ago, how much of that $600 million is secured? And then I assume that as you move forward, there's still opportunity to add incremental revenues to that funnel on the timeframe that you've suggested?

Yes. Thanks, Joe. You're right. The -- what we announced last quarter represented the design wins, won for the first 6 months of the year. What we're talking about today is the $600 million is for the design wins that are won and also in our pipeline. So based on what we have today and in our pipeline, we have $600 million of opportunity right now. To your point, that doesn't mean that we can't continue to add to that, we also give that a haircut, as you might imagine, related to what the customers think their revenue is going to be and also the probability of timing to market. So I think it's a relatively reasonable and conservative look at it. And there is, to your point, an opportunity to add to that.

Okay, great. And I assume that's mostly computer vision, I mean is there a video processing component of that?

There is, but there is also a healthy component for computer vision as well, but there is some vision-based system as well.

Let me add one more thing, Joe. I think in my script, I do mention that majority of that $600 million of our CV.

Our next question comes from the line of Tristan Gerra with Baird.

Looking at the design wins that you've announced in the security camera business. And at the same time, your 2 large customers not reordering. Is there a inflection point that you're expecting in terms of new orders coming from those customers, and is that when you expect inventory to happen?

Well, first of all, I think that we do see that the both customer were cutting -- are using their inventory right now. And also, we start seeing small orders coming from both of them. So that's why Casey just mentioned that we think we're in the trough of the revenue for those 2 companies at this point. However, moving forward, I think that I expect that data will be strong on the computer vision. And the video processor business will continue to be just flattish, but Hikvision is different though. Hikvision, we haven't announced any design win with TV, design win with Hikvision yet. So I would say we should treat these 2 companies separately, but we do believe that: First, the inventory is gradually being were down; two, the CV ramp in -- from Dahua will be important for us; and three, in the future, we're going to continue to work with Hikvision to see where we have opportunity work -- to work with them on the computer vision technology.

Okay. Great. And then a quick follow-up, which is as you expect your CV business to ramp meaningfully over the next fiscal year. I understand the higher ASPs, is there any impact or how should we look at the trajectory of OpEx as a result?

Yes. I think the modeling for CV, we've talked about is exactly what you just mentioned. And that is 2x, roughly 2x the ASP, the same -- a similar gross margin profile that, obviously, operating leverage as we're able to now leverage the $450 million that we've already invested in CV, along with taking our technology more broadly, as Fermi mentioned, to more and more customers. So I think the CV leverage on an operating basis and on a top line basis, is as you described it.

Our next question comes from the line of Matt Ramsay with Cowen.

For me, I wanted to ask, and I guess, Casey, feel free to chime in here, too, something from the prepared script, you reiterated CV as 10% of revenue targeted for this year, which, I don't know, $22 million, $23 million, I guess, based on the guidance for the fiscal fourth quarter. And then I just want to make sure that I heard this correctly that keeping camera -- sort of vision camera revenue flat, you expect 25% CV revenue contribution next year, which I guess implies a tripling of the CV revenue. Is that -- I don't know, I'm doing math here in the spreadsheet and probably screwing it up, but I just wanted to see if that's an accurate portrayal of what you guys said?

Yes. I think everything you say is correct, except 1 thing. We continue to expect 10% of CV revenue this year. And also, we provided guidance that we expect the revenue will be 25% of total revenue next year, but that's 2.5x, not 3x as you described.

Got it, Fermi. Just following up on, I think, Joe's question and maybe asking in a different way on the auto pipeline number. I just want to really get precise about -- there was a $200 million in 1 business that you talked about last quarter. And just within that, $600 million, what's the progression over the last, I guess, 3 months of the business one, the same metric that you gave us last time? I understand that you'll be giving the full pipeline number going forward, but just since you gave us 1 data point, maybe you could give us a second one so we can try to draw a line?

Well, we are not -- based on our current announcement, you see that we signed this 2 new design wins this quarter, right? One is that functional safety chip for the ADAS -- sort of Level 2+ ADAS application; the other one is for the electronic mirror. So with this 2 issued share, that will definitely add more to the pipeline just this quarter. And we didn't disclose it because we -- going through this exercise is time-consuming, to be accurate. So -- and we plan to continue to update this funnel numbers at least once a year, so that we continue to provide you a data point, but also, every quarter, we're going to continue to talk about the design win we get, which we will not disclose the amount, but you can -- that will give you some indication how we continue to add to our funnel every quarter.

Got it. Just one last one for me, and I'll get out of the way. I guess if you are talking about forecasting CV revenue and the remainder of the business for fiscal '22, what assumptions are you making at Hikvision and Dahua about their ability to secure high silicon chips and what that might mean for your business or are you assuming that things stay steady state and whatever changes happen there may come in the future?

I think my assumption for those 2 customers are different. My assumption is going to ramp up with Dahua on CV chip, which we will talk about in this script. In [Hikvision] region, we are, in our current revenue forecast does not include any CV revenue from them with this version. And that's what we are working on hopefully that we can convince them to use our CV technology in the future.

Our next question comes from the line of Tore Svanberg with Stifel.

Congratulations on the solid results. First of all, it sounds like you're assuming the video process business to be flat in fiscal '22. Is that just because of CV cannibalizing or you're just being a bit conservative there, too, given the geopolitical tension still?

Yes. Well, go ahead. Casey, go ahead.

I was just going to say, I mean certainly, we always try to be conservative in the way we look at business going forward. It's going to -- we'll learn over the next couple of years how much is pure growth, and how much of it is replacement. There'll be an element of both, but I don't know that we have enough information to date to make that determination. Clearly, after you get to that time point, it's going to be more and more cannibalization because we're not doing new video only processors for people to continue to extend, so they're going to have to either continue to use the products we have or over time go to CB.

Let me add one more thing. I believe when we talk to all the customers out there, I think all of that planning to use CV to gradually replace their video processor technology because everybody view that, although it's 2 different technology, but it's a continuous of the product line. All the video processor market will be replaced by CV in a span of 3 to 5, maybe even longer years, but eventually, CV can already replace that. In my opinion, I really think that in 3 years, the 50% video processor market will convert to the CV market, which is great for us because it's not only help giving us a chance to collect more new market share, but also increase the ASPs.

That's great perspective. And I had a question on CV28M. It sounds like this is more of a consumer product for you. And I'm just wondering if this is something that goes after higher unit volumes? And if so, what implications does that have for ASPs and potentially margins?

Right. So first of all, you're absolutely right. We talked about in the past, we want to build a complete family of CV chip-based on our CVflow architecture. And CV28, this new member is redesigned to target at high-value market and so -- which is also very sensitive to the ASP and the cost. And -- but at same time, I want to point out the 2x ASP versus our video processor technology. It still applies to CV28 because our low-end video processor price is very low. Also, that's first thing. In terms of gross margin, I will say that this chip's gross margin will be as good as or maybe better than the video process equivalent video processor technology, but however, it's not going to be as high as our corporate gross margin. So that's a balance of this too. When you go to higher value market, you are competing a bit on the pricing side. So CV28 is designed for that.

Our next question comes from the line of Quinn Bolton with Needham.

I just wanted to follow-up on the auto funnel. First, just a clarification. Did you say that was extended through fiscal '27? I'm just trying to get the date right?

Yes. Correct. It was fiscal '26.

It's fiscal '27. It's a total of 6 years.

Okay. It's 6. Okay, 6 years. So and obviously, can you give us any sort of shape of that funnel? It sounded like you said the bigger ramps kind of happened in calendar '22, '23. I would assume sitting where you are today, there may not be a lot of business that's being added at this point for, say, 2025. So should we think about it that you ramp up to a steady-state level by, say, 2023 with this current funnel? And then it would be sort of flattish for several years? Or can you give us any idea how you're viewing that shape? Obviously, anything you add to the funnel in the future, will add to this, but just kind of wondering what the shape of that funnel looks like?

Yes, go ahead, Louis.

Yes. You're right about the shape. So there's some large programs that start to go into production in calendar year '22 and calendar year '23. And then it levels out in terms of the annual contribution, but over time, we'd expect those numbers to change as we win new programs or bid on new programs.

Got it. Understood. Okay. And then, Casey, I guess I'm -- maybe I missed something on the gross margin guidance. It sounds like Hikvision and Dahua or maybe just more broadly, China professional is going to remain at a fairly small percentage of revenue in the January quarter, yet margins are going to be down nearly 200 to 400 basis points. And so it doesn't feel like you have an adverse mix shift back to China professional security. So what's the cause for the margins to come back down in the January quarter?

It's an overall general mix between several different customers, but there is a fair amount of revenue that we will start to see, I think, in Q4 from those 2 customers, some of the vision-based and, as Fermi just mentioned, some of it being CV based around Dahua. And I think that will be a nice contribution to the top line, but it will also put pressure. It's not the only factor, but it certainly is one of them. As I mentioned, the consumer side of the business is always down this quarter. That's a seasonal thing. So that changes the mix, not only customer, but also product mix. And so there's more than just one thing going on, but there will be some recovery in that revenue. And again, as I mentioned, I think, while Hikvision will be relatively stable or not have as much growth, we will see that CV start to come in, in Q4, and that, while it helps revenue, will also have some margin impact.

Got it. Great. And I guess with that vision business maybe starting to come back a little bit, can we assume now that you're largely through that $10 million inventory purge or do you think that, that $10 million that some of that is still burning off in the January quarter?

This is Fermi. I think we continue to see that they are burning through this inventory, but we -- I don't believe we have seen the end of it. Among all the product lines of video processor product like Hikvision and Dahua ordering from us, we're only seeing several items coming back for new orders. But remaining still, we haven't seen many orders yet. So I think that's a side for 2 things. One is they continue to burn through the inventory, but for some certain product line, the pressure (inaudible), so we start seeing new orders. So that's why we are saying, in the next few quarters, we're going to continue to monitor how they're going to come back, but however, at the same time, I also need to mention that -- I also mentioned that Dahua CV will start ramping in Q4. That will help us on the revenue side, put some pressure on the market side.

Our next question comes from the line of Kevin Cassidy with Rosenblatt Securities.

Congratulations on the great results. As you're introducing the new CV devices, the CV28 in particular, now that all of your customers have done some CV designs, is there a change in the time to market from the time you introduce a product to the time your customer comes out with the product?

Yes. Kevin, this is Fermi. I think you're absolutely right. Our overall CV family chips from CV22, 25 to 28, they share the same software SDK. So as soon as the CD28 works -- the chip, the hardware was sent over to customer, people can port the existing SDK that they already have on to CV28. And that's why we think the time to market is much, much faster. And we also -- in my script, I talk about that we expect to see -- we receive orders in Q1 next year, and we expect we going to ship on time.

Okay. Great. Yes. And the -- on the OpEx side, it looks like it's under control, very flat quarter-over-quarter, even with 5 nanometers products coming in the pipeline. Are there any foreseen expenses coming up or should we expect this through FY '22 to be kind of tight controlled over OpEx?

Yes, I think you're going to continue to see the drivers in OpEx be 2 things: One is we're going to continue to be hiring, not only in engineering, but we're starting to do some hiring outside of that as we start to really deeply pursue some of these markets that we haven't been in, in the past. The second thing is, although there can be a little bit of lumpiness to it, there will be a build in the cost of designing, as you mentioned, in lower technologies, not only the CAD tools and foundry costs, but the overall cost of being on the front edge of this technology will continue to be a driver. Generally, that gets smooth a little bit the way we treat that over the quarters, but you can, as projects start and stop, have a certain amount of lumpiness to it, but I think those will really be the 2 drivers. The rest of the business in pretty good shape.

Our next question comes from the line of Charlie Anderson with Colliers Securities.

My congrats on a strong quarter and outlook. I wanted to start with some of the opportunities in automotive. Your software partner was acquired by -- in Europe was acquired by an auto makeup. I'm sort of curious how that maybe changes the complexion over there. And then I think there's been some reporting on potentially some mandates or some goals in China around partial driving and self-driving, and I wonder if you're seeing that manifest itself in any of the design activity in terms of some of the push over there? And then I've got a follow-up.

Yes. Thank you. So first of all, [Hila], yes, they are very important software partner for us and I think they were acquired by the VW and (inaudible). But as a follow-up. First of all, I think that the video VW side, we got more -- we have (inaudible) in there now. I think VW will be (inaudible) VW (inaudible). At this time, I also realize that we need to have new software partners for other business. We spent a lot of time to develop (inaudible). In terms of your second question, second question. Charlie, can you remind me your question?

Yes, sorry, it was about -- I know in China, it seems like there's a push toward more partial autonomy and full autonomy, some mandates potentially. I wonder if you're seeing any of that manifest itself in design activity and some of the planning?

Yes, we do see that. And first of all, I think that is more for the commercial vehicle part than the fastest. In fact flat some strong revenue resting in Chinese for the (inaudible) market. So I really think that, that's one market that we start seeing drivers from a government regulation point of view. And that, I think one will see. We haven't seen any division requirement on super vehicle. So if we hear anything, we'll provide updates.

As Louis mentioned, we're in 2 different sites, and it seems like we're being leases at this point, but we'll see what we can do.

Okay, great. And then I just had a quick follow-up. I was curious if you could maybe update us on the HiSilicon situation. I know last call, you talked about a lot of inventory being out there. Just curious how that's looking in terms of some of the opportunities just to capture some of those sockets?

Right. So in the last few months, we continue to monitor the situation. We believe the situation just like what we said last time. There's still plenty of inventory out there, but many stay in the big customers. I think, for example, Hikvision, I believe -- we don't have any proof, but we do believe Hikvision has a lot of inventory of high silicon chips so they can continue to ship, but however, for other smaller HiSilicon customer, I think they are looking for new -- newer suppliers. And that's why we are seeing opportunities that we can win, and we mentioned that we win a handful of Chinese professional semi-professional security camera customers in the last 3 months, including Dahua, and also, we believe they're going to gradually ramp up the production with us -- I mean Q4 moving to next year. So -- and also, we also see -- in other video markets, we see a lot of customers are looking for solutions. And we definitely believe we have opportunity and a technology that can -- we can serve them.

Our next question comes from the line of Ross Seymore with Deutsche Bank.

Congrats on the strong results and guide. I guess for my first question, just another clarification on the automotive funnel side of things. And forgive me if I'm cutting this a little too finely, but was that $600 million a combination of both existing design wins? And then I thought, Fermi, you said designs you're competing for. Is that kind of like a SAM analysis? Are those design wins that you're competing for that you think you're going to get? I just want to clarify how kind of concrete those design wins are.

Ross, this is Louis. For that $600 million funnel, as Fermi said, a majority, think of 70%, is business that's won, and that business has been discounted based on our confidence in the revenue forecast that's been given to us at the time we won the business. The other $200 million are opportunities that we're bidding on with 2 discount factors. Discount factor 1 is the probability of winning that business; discount factor 2 is the confidence in the revenue forecast that's been provided to us at the time we're bidding on the program. So it's a 6-year funnel. And again, as Fermi said, the majority of that business is won.

So in another way to say, with the $200 million you gave last quarter that at that point, you had year-to-date, you pretty much added another $200 million until today. And then the final $200 million is kind of the part that you hope to win?

No. Last quarter, Ross, what we provided was $200 million of business won in the first 6 months of the year. So there's an additional $200 million on that, that we had won before that period of time or more recently.

Got it. Okay. Then maybe one for Casey. A lot of good questions have been asked and answered already, but you gave great granularity on the sequential guide by your 3 primary segments, but you just gave a directional guide of what they actually were in the prior quarter. So whether it was for the fiscal third quarter or if you want to do it for the entirety of the fiscal year, how do your revenue split between the IP security, automotive and then consumer and other buckets so we just have a baseline to work these percentages off of going forward?

Yes. For the full year, we've said and it remains true that the auto side of the business is probably going to be between the 15% and 20% of total revenue, that the consumer business was probably going to be close to 20% of total revenue, and then the balance would be IP secured. So we were saying in the past kind of 20%/20%, but it's changed a little bit, obviously, as the course, but not a lot.

And then as you go into the fourth quarter, this goes back to the gross margin question I think either Tore or Quinn asked about. The revenue mix side of the equation, I get that -- it sounds like Dahua and Hikvision are going to be a little bit more of the business, still mid-single digits, maybe a little bit more. What is the other mix-related headwind? Are there positives in the consumer segment that were big tailwinds in the fiscal third quarter that seasonally go away or is there something else going on?

Like I say, it is a mix. The only correction I would make is that for the security camera revenue, we said that it will be -- for the fourth quarter, we said that we'll be up in the low double digits sequentially. So it will be up. And with Dahua that we've talked about and Fermi's talked about as well as some HiSilicon. That's part of it, but not all of it, but that's part of the mix.

Our next question comes from the line of Andrew Buscaglia with Berenberg.

I wanted to follow up on -- you talked about the computer vision as a percentage of sales, around 25% next year. So that's much -- I was modeling out, that out and I came to just under [20]. So I'm wondering, it seems like you guys are pretty optimistic here. And I'm wondering if you could provide maybe some sort of breakdown of what that comprises of next year? I imagine it's mostly professional security, but do you have any other like color you can add as to how that -- the makeup of that 25%?

Yes. First of all, I think you are right, a big part of that 25% is from a professional security camera. However, we also mentioned that we believe we control our consumer -- sorry, our home security camera business will ramp up with CV -- CV product next year. And we also mentioned that we think it could be material next year for the revenue point of view, but also, more importantly, on the CV -- on the automotive side, I think our -- there is some CV revenue -- short-term CV revenue opportunity that we can ramp up, particularly in China, commercial vehicle and also the [bases]. While outside China, we talk about fleet management, all the fleet management customer outside China are using our CV chip for ADAS as well as driver monitor applications and also believe that, that business will ramp up. So I think they are all the potential business opportunity for us for fiscal year '22, but I think you're absolutely right. The professional security camera will be the biggest part of it.

So presumably, though, you have -- you're talking pretty positively on what's going on in automotive, too, with that funnel. So presumably, the real driver for computer vision will -- even comes beyond that 2020 -- calendar year 2022 and beyond. I would think that would be an even bigger driver for that computer vision segment.

I agree with you on that.

Yes. Okay. Interesting. And then maybe last one for me. You said -- you had some interesting commentary around the security market over 3 to 5 years adopting computer vision. Are you essentially saying products that you see being sold 3 to 5 years from now will have to -- I guess, what's prompting people to upgrade to these new CV capabilities? Is -- what are the drivers behind that? Because that could be a pretty big upgrade cycle.

Well, for professional security camera, the biggest driver is the following: In the past, all of the computer vision is down the server side. And that requires to transmit all the video to the server and store there and do the video analytics or computer vision off-line, which is okay for maybe 1 million units of camera, but if you want to scale to media, many tens of millions or hundreds of millions of camera that add to the huge pressure, not only on the bandwidth infrastructure, but also storage cost. So -- and when we talk to customer, they all agree, the only solution for that is to apply computer vision on [age] so they only send -- analyze the data back to the server and only apply some analytics on video data on this cloud when it is required. That will dramatically reduce the requirement of the transmitting of video and the server storage and server computation. And that's a huge investment for saving for our customer. That's why we believe that we're going to continue to see the trend, but on the other side, however, the difficult part is that the current CV camera is a lot more expensive video processor camera, but however, that's why I said in 3 or 5 years, the economy will bring down the cost of the camera, will make this more attractive to our customer to replace that.

So there's a huge driver purely based on the technology reasons, but I think the financial reason will come in 3 to 5 years to make this happen. Our internal forecast, we believe that our CV revenue will continue to go up at a big percentage just because of that. And also in the auto side, we also see the same thing: in the future, all of new design win that we are winning, majority will not CV related because even you don't need CV, but -- CV in a car today, but people want to put CV processing in there just so they can have a future upgrade. And this is a future-proof criteria really help us to design a lot of CV chip for the automotive opportunities.

Our next question comes from the line of Vivek Arya with Bank of America Securities.

Congratulations on the strong growth. For me, I had a few conceptual questions. First one, when do you think automotive CV becomes more than 10% of your sales? Is that something that can happen in '22, '23? Just conceptually, when should we expect automotive CV to be more than 10% of your sales?

I [remind] to give some additional color to the revenue and certainly, the emergence of some of these markets. It is early time, as Fermi, I think, mentioned. And so we try to go guide as best we can quarter-by-quarter. I don't know that today, we are ready to try to commit to timelines. Conceptually, we think that CV is going to continue to drive all of our revenue platforms. And so we're very bullish. We're seeing the development activity very strong. That's auto-included, but how that plays out in these early times during the next few quarters, I think we'll be able to comment on that more broadly as time goes on, but today, I think what we've done is try to characterize it as best we can.

I see. Okay. And for my follow-up, in the wins that you're having in automotive with CV, how many of them are exclusive where you are the only CV processor in the car, where you are one of many processors, and if it is the latter, are there certain applications where you have tended to do better? Are there certain applications where the customer is preferring to go with the competitor? So it's more a technology question as to whether you're exclusive or not. And where are applications where you tend to do better versus the competition?

Well, for the -- maybe it's easier to look at from different automotive application point of view from like (inaudible) or ADAS or the coders when they use a CV chip, usually, there's only one end process in those applications. So if we are chosen, then we are only processor, CV processor in that device by definition. But however, like you said, there's other applications like Level 2+, and people are using multiple chip to CV. In some of the applications, we are the only one to provide multiple chip into application, but there are cases also that there are some functions being done by auto chips, and we implement portion of applications. It all depends on the customers' preference as well as the current market situation. So for example, for Level 2+, Mobileye is dominating in the market. So you can imagine that Level 2+ there some customers say, "Okay, we don't want to change my phone camera, which is Mobileye, for other camera," then they can use other solutions, which we can definitely be considered, and we want some of the demand we make.

Got it. And just lastly, how should we think about operating expense growth as you're starting to grow your top line at a faster pace in these emerging markets? I understand that you have done a lot of upfront R&D, but is there a simplistic model to drive your operating expense growth from here? Will it grow at half the pace of sales growth? What is the right way to think about your OpEx growth for the next 1 to 3 years?

I think definitely, you're going to see, as I mentioned, leverage coming in the model. As I mentioned earlier, that's partly because of the CV in the 2x ASPs, which drives the top line, but you're not going to see that get taken out in the operating expenses. You're going to see our operating margin growth continue over the next couple quarters or next couple of years even, but I think you're going to see the leverage in the margins -- sorry, after the margin line come back into the model. We were doing operating margins at 20% to 25% historically. And I don't think there's any reason over the next year or 2, you don't see that come back into that -- the target range.

Our last question comes from the line of Suji Desilva with ROTH Capital.

Looking past the January quarter and the growth there, is there some elements of seasonality looking into the fiscal first quarter '22 or are your program ramps, you think, and maybe restocking perhaps going to carry? What's the dynamic looking into the next 6 months?

Well, looking at -- historically, as you comment, going from Q3 to Q4, we were typically down 10% to 15%, and going from Q4 to Q1, we were typically down around 10%. Part of that was driven by the consumer nature of the business historically, but part of what we're seeing now is good growth in that -- or start at the beginning, it's early, but the beginning is a good growth in the CV, which because of the profile -- the top line profile, I think that will take some of those -- some of that seasonality out. I think it still exists in the model, and we're going to have to see how that matures out and how CV matures out over the next few quarters, but clearly, we didn't see or we're not guiding to see that from Q3 to Q4. And when we get out to Q4, we'll guide into Q1, but some of that seasonality by the nature of the change in markets and products is going to go away.

Okay. And then maybe for Fermi. On the China Hikvision-Dahua dynamic here, it seems to me if Hikvision struggles to get high silicon chips and doesn't use you guys, I'd imagine some of the Tier 2 you're starting to work with in China might start to grow to a share position that's similar to those guys. Is that how this market is going to play out or are those 2 still going to dominate the market? Or is one of the Tier 2s kind of coming up the curve to become as big as them for you potentially?

Well, I need to be very careful about this because you sell the Tier 2 to you (inaudible) my customers. But however, in China, my personal belief that Hikvision, Dahua and the Huawei now become the third largest security camera provider, those 3 will continue to have majority market share. Just because their size and the momentum and the product portfolio they already have. I think it's -- all the others will have a chance to increase the market share, but you will not get to the point that they can challenging the market share of Hikvision, Dahua or Huawei.

This concludes today's question-and-answer session. I would now turn the call back over to Dr. Fermi Wang for closing remarks.

And thank you, everybody, for joining us today. And I really think that we made a great progress on CV side, and I'm looking forward to provide you more updates next time. Thank you now.

Thank you.

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