Ambarella Inc (AMBA) 2019 Q2 法說會逐字稿

Good day, ladies and gentlemen, and welcome to the Ambarella Second Quarter Fiscal Year 2019 Earnings Conference Call.

As a reminder, this call is being recorded.

I would now like to introduce your host for today's conference, Ms. Deborah Stapleton, Investor Relations.

Ms. Stapleton, you may begin.

Thank you, Liz.

Good afternoon, and welcome to Ambarella's Second Fiscal Quarter 2019 Financial Results Conference Call.

Thank you for joining us today.

Our speakers will be Dr. Fermi Wang, President and CEO; Casey Eichler, CFO; and George Laplante, Executive Vice President.

The primary purpose of today's call is to provide you with information regarding our fiscal 2019 second quarter results.

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 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 are 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 that we filed with the SEC, including the annual report on Form 10-K that we filed on March 30, 2018, for the 2018 fiscal year and the Form 10-Q filed on June 8, 2018, for the first fiscal quarter of 2019.

Access to our second quarter results press release, historical results, SEC filings and a replay of today's call can be found on the Investor Relations portion of our website.

I'll now turn the call over to Dr. Fermi Wang.

Thank you, and good afternoon, everyone.

Our Q2 fiscal year 2019 revenue was $62.5 million, representing a decrease from $71.6 million in the same period of the prior year.

Revenue from our target markets, including IP security camera and the OEM automotive car recorders, grew year-over-year.

Revenue from consumer applications, including sports, VR and the drone cameras, continue to decline.

And as you can see from our weak Q3 guidance, we expect revenue from these consumer applications to continue to be under pressure, reflecting lower demand for non-GoPro sports camera, the VR market remaining nascent and the drone market leader, DJI, mostly using its own silicon solutions.

The limited size of these markets, combined with these market dynamics, further reinforces our decision to focus on the growing IP security, automotive and the new robotic market opportunities.

During the quarter, we continue to make steady progress on the development and the delivery of our computer vision solutions based on our CVFlow architecture.

We benchmarked our CVflow SoCs against competing solutions from our primary silicon competitors running leading public domain neural networks.

These benchmarks show that we have a significant performance and the power advantages versus competing camera SoCs, application processors and the GPU-based solutions.

The efficiency of our CVflow architecture also gives us a big die size advantage.

When combined with the major advances in our image processing and highly efficient video encoding, we believe our new CVflow SoCs are very well positioned to win in our target markets.

We have now released our CVflow Software Development Kits, supporting our new CV2 and CV22 SoCs, including tools for the optimization of a Convolutional Neural Network or to our customers.

Customer feedback has been very positive in terms of the CV performance, power consumption and programming flexibility.

In many cases, customer are seeing performance results that are more than 10x faster using our CV chips through our target domain network than with competing solutions.

As a result of these customer evaluations, we now have a multiple leading professional IP camera manufacturers in design with our computer vision SoCs and expect this customer to enter best production in the first half of next year.

We believe that security cameras with computer vision capability will continue to grow as a percentage of total professional IP camera market, including those for new installations as well as for replacement of existing cameras.

This will provide the opportunity for us to increase our value contribution for camera and to grow our total revenue.

Professional IP camera customers are also continuing to introduce a wide range of cameras based on our existing SoC solutions.

For example, in May, European IP security camera leader, Axis Communications, introduced its Q6125 PTZ camera using Ambarella's S3L SoC that enables surveillance in total darkness up to 200 meters.

In the IoT home monitoring camera market, we saw a number of new product introductions from service providers spanning both traditional security companies and cellular operators offering security services.

U.S. security giant, ADT, introduced its Doorman Service based on an integrated doorbell camera and a door lock that can provide both remote viewing and a door opening.

The camera used Ambarella's S2LM camera SoC for high-quality HD video and leverages its advanced de-warping capability to provide a full 180-degree field of view.

Also during the quarter, alarm.com introduced its V622 indoor camera with full HD video, wide-angle 180-degree field of view.

Based on Ambarella's S2E camera SoC, the camera also includes IR night vision, digital pan/tilt/zoom and two-way audio calling and Bluetooth music streaming.

While Ambarella's IoT home monitoring revenue has been driven largely by the U.S. market in recent years, we are now seeing new opportunities for growth in Europe.

During the quarter, European mobile service provider, Vodafone, launched its smart home IoT service powered by Samsung SmartThings.

Featuring an HD camera based on Ambarella's S2LM SoC, the service has initially been launched in Spain, with other European countries to follow.

In the OEM automotive market, we continue to build customer momentum for our computer vision solutions.

Feedback from customers is that they are looking for a platform that provides more flexibility to add value and differentiation than the current market leading solution.

Furthermore, alternative, legacy automotive chip suppliers are failing to meet CNN processing performance requirements as well as being unable to operate within the limited power and the thermal constraints of the camera design.

Ambarella is meeting these requirements, delivering higher CNN performance, superior low-light performance and lower power operation while also offering video recording capability to support machine and human vision in a single design.

In the front camera ADAS category, we are being evaluated by leading Tier 1 camera suppliers as well as third-party software suppliers to deliver ADAS solutions, including those meeting Euro NCAP requirements.

To accelerate time to market while initially focused on customers and the partners that are developing their own algorithms, requiring high-performance CNN processing and low power, with Ambarella's providing an open, programmable hardware platform and SDK.

In the emerging level 2 through level 5 autonomous vehicle categories, Ambarella's operating a flexible, open perception platform, including both short- and a long-range stereo cameras, enabling customers to develop their own autonomous software stacks.

Additionally, we see opportunities to partner with customers in further development of our own perception modules for future integration into their autonomous architecture.

During the quarter, we continue to deliver CV1-based SuperCam stereo cameras to customers for evaluation.

The SuperCam cameras deliver the 360-degree short and long-distance viewing capability required for advanced perception and precise self-location.

They provide a perception range of over 150 meters for stereo obstacle detection and over [180] (corrected by company after the call) meters for monocular object classification.

We also demonstrated our new CV2 SoC that combines CNN processing that is up to 20 times as powerful as CV1 and includes more advanced image processing as well as enhanced stereo processing.

Our testing confirms that CV2 consumes less than 5-watts of power while simultaneously performing neural network processing, image single processing and HEVC encoding at 4K resolution.

On a processing performance per watt basis, this is more than an order of magnitude better than leading GPUs can achieve even when they only perform neural network processing.

For level 2 to level 5 autonomous applications, the CV2's combination of CNN performance and low power enables development of camera perception hubs that offload CNN processing from the main computer.

And in China, the world's largest automotive market, Ambarella currently enjoys a leading position in OEM car recorder market, is building upon the relationships established with camera makers and third-party software companies to target future designs that require computer vision technology.

This include forward-facing cameras that combine recoding with ADAS functions as well as electronic mirrors that integrate recording capability or add blind spot detection.

In the OEM automotive market, we continue to see new car models introduced in Asia with car recorders integrated as a pre-installed option.

During the quarter, Toyota introduced 2 new HD drive recorders in its C68 family.

The models are both powered by Ambarella's A12A SoC and include HDR processing, low-light nighttime recording and a wide-angle viewing enabled by A12's on-chip dewarping hardware.

While the models support 2 external small form factor camera modules for front and rear-viewing, with connectivity to the head unit supported by Ambarella's B6 serializer/deserializer link chips.

In China, carmaker, SAIC, introduced its Marvel X electric SUV with a full HD car recorder, including WiFi capability based on Ambarella's A12 camera SoC.

Additionally, Chinese carmaker, BYD, introduced its Tang SUV with an integrated full HD car recorder also based on Ambarella's A12 SoC.

The design leverages the A12 on-chip Ethernet capability for connectivity to the vehicle's infotainment system.

In the market dash cam category, Korea's Fine Digital introduced its GX2000 model based on Ambarella's A12.

The dash cam include a front and a rear Quad HD resolution cameras, ADAS functions, including lane detection warning and forward collision warning and a 3D noise cancellation for excellent nighttime image quality.

Ambarella is also engaging with aftermarket and fleet camera makers with a full camera design based on our CV22 SoC that combines traditional dash cam recording capabilities with advanced ADAS algorithms.

In a consumer drone market, Parrot, Europe's leading drone group and the world's #2 consumer drone maker, introduced its ANAFI model, setting a new benchmark for video features and quality at a sub-$700 price.

ANAFI is ultracompact, foldable design with a 4K HDR camera and a 25-minute fly time.

Based on Ambarella's H22 SoC, the camera includes advanced HDR processing and offers unique video and photo shooting capabilities, thanks to its 180-degree tilt gimbal and 2.8X lossless zoom.

Also during the quarter, Chinese drone maker, Yuneec, introduced its Mantis Q consumer drone, a small form factor designed for both outdoor and indoor use and capable navigation inside buildings.

The drone features an integrated 4K camera based on Ambarella's A9AC SoC and includes electronic image stabilization and a voice control.

In summary, while our revenue continued to be impacted by the decline of sales into specific consumer markets, we remain confident that our decision to focus on computer vision for IP security camera, automotive and the robotic market is correct and bearing fruit.

We are delighted by the customer response to our new computer vision SoC in IP security camera market and move forward to customers being in mass production in the first half of next year.

In the automotive market, our revenue from car recorders continues to grow, and we are making steady progress as we engage customers with our computer vision technology for ADAS and autonomous vehicle applications.

Finally, having successfully delivered a 10-nanometer CV2 and CV22 CVflow SoCs in the first half of this year, we are also pleased to announce that we have taped out an additional family member to further extend the range of price performance options available to our customers.

Before I turn the call to George for his portion of the call, I would like to introduce our new CFO, Casey Eichler, who has been on board a total of about 3 weeks.

We are also pleased to have him to replace George, who, as you know, is retiring at the end of September.

And also, I want to sincerely thank George for his many years of service to the company as his dedication, guidance and friendship have been invaluable to me and the entire team of Ambarella.

And we all wish him the very best in his well-deserved retirement.

Now I'll turn the call over to Casey, who would like to say a few words before George discuss our Q2 results and the Q3 guidance.

Thank you, Fermi.

I'm excited to join Ambarella, and I look forward to working with Fermi and the entire Ambarella team.

George built an excellent team and has demonstrated tremendous leadership while at Ambarella.

I appreciate the opportunity to work with him over the next month while he transitions to a new chapter in life.

With that, I'll turn it over to George to go through our Q2 results and Q3 guidance.

Thank you, Casey, and good afternoon, everyone.

Today, I will review the financial highlights for the second quarter of fiscal 2019 ended July 31, 2018, as well as the financial outlook for Q3 of fiscal year 2019 and update the outlook for the full fiscal year.

During the call, I'll discuss non-GAAP results and ask that you refer to today's press release for a detailed reconciliation of GAAP to non-GAAP results.

For non-GAAP reporting for Q2, we have eliminated stock-based compensation expense adjusted for the impact of taxes.

In line with our previous guidance, our Q2 2019 revenue totaled $62.5 million, which represents a decrease of 12.8% from the $71.6 million of revenue in the same period of the prior year.

Non-GoPro revenue declined in the quarter by approximately 8.3% from the $67.9 million in the previous year.

In Q2, we saw year-over-year growth in IP security and auto OEM, offset by a substantial decline in revenues from drones and wearables, which also includes non-GoPro sports and VR cameras.

In IP security, year-over-year growth was led by consumer security revenues, while professional security revenues were flat.

Solid revenue increases from OEM automotive video recorders in Japan and China resulted in year-over-year growth for the auto market.

Consumer drone revenues in the quarter were down substantially from last year, with the majority of our revenue coming from the new Parrot drones and the high end from DJI.

Wearables declined from last year as we experienced weak sales across most consumer camera categories that we believe are due to weak end market demand.

Non-GAAP gross margin for Q2 was 61.4% compared to 61.8% in the preceding quarter.

Gross margins continue to reflect the shift in market mix to lower-margin China security revenues and a drop in higher-margin drone and wearable camera revenues.

Non-GAAP operating expenses for second quarter were $29.9 million compared to $31 million in the previous quarter.

OpEx decreased from the previous quarter, primarily as a result of a decrease in the employee bonus costs as a result of lower revenue projections and nonrecurring employee costs incurred in Q1, partially offset by increased headcount costs.

Non-GAAP net income for Q2 was $88.5 million or $0.25 per diluted ordinary share compared with adjusted non-GAAP net income of $19.3 million or $0.56 per diluted ordinary share for the same period in the previous year.

The non-GAAP effective tax rate in Q2 2019 was approximately 7.1%.

The non-GAAP effective tax rate for fiscal '19 includes a change in the allocation of stock-based compensation across the company's tax jurisdictions to improve alignment of the non-GAAP tax rate to the GAAP tax rate.

For Q2 2018, non-GAAP net income was adjusted from $16.5 million to $19.3 million, and EPS was adjusted from $0.48 to $0.56 per share for comparative purposes.

The change has no impact on GAAP net income or tax rate.

In the second quarter of fiscal 2019, the non-GAAP earnings per diluted ordinary share were based on 34.2 million diluted shares as compared to 34.6 million diluted shares for Q2 of fiscal 2018.

Total headcount at the end of Q2 was 741 compared to 733 at the end of the previous quarter, with about 82% of employees dedicated to engineering.

Approximately 71% of our total headcount is located in Asia.

We ended Q2 with cash and marketable securities of $376 million, adding $6.9 million of cash from operations in the quarter.

During the quarter, the company repurchased 1,119,193 shares at an average price of $40.18 for total consideration of approximately $45 million.

As of July 31, 2018, approximately $66.1 million remained available for repurchases under the $100 million repurchase program through June 4, 2019.

From August 1 to August 24, we repurchased an additional 598,892 shares for total consideration of approximately $23.1 million.

Total accounts receivable at the end of Q2 was $29 million or 43 days sales outstanding.

This compares to account receivable of $25.7 million or 40 days sales outstanding at the end of the prior quarter.

Net inventory at the end of Q2 was $30.8 million compared to $23 million at the end of the previous quarter.

Q2 days of inventory increased to 100 days in Q2 from 96 days in Q1.

Although inventory is higher than we normally carry and our intention is to reduce inventory value in the second half, we do not anticipate any valuation issues as the inventory consists of high-volume chips for our primary markets.

We had 2 10% plus revenue customers in Q2.

WT Microelectronics came in at 62.7% of revenue compared to 67.1% in the second quarter of the prior year.

Chicony came in at 15% as compared to 8.7% in the second quarter of the prior year.

I will now discuss the outlook for Q3 and update you on the fiscal year 2019 outlook.

We expect revenues for our third quarter ending October 31, 2018, to be between $55.5 million and $58.5 million, representing a decrease of between 37.7% and 34.3%, respectively, from Q3 of last year.

We expect non-GoPro revenues at the midpoint of guidance to decline approximately 20% from last year, with GoPro revenues being immaterial in the quarter.

In the quarter, revenues will be negatively impacted by the substantial decline in demand for drones and consumer products -- camera products during what is normally our high holiday selling season.

Demand in consumer wearable markets, including non-GoPro sports, VR and other wearable cameras, will be substantially below our earlier expectations and last year's volumes.

Sales in these markets so far this year are below our expectations, and based on our customer feedback, we see this year-over-year decline accelerating through the normally strong holiday selling season.

We estimate Q3 non-GAAP gross margins to be between 59% and 60.5% compared to 61.4% in Q2 of 2019.

The decline of higher margin drone and wearable revenues as a percent of total revenues continues to have a negative impact to margins in the quarter.

We expect non-GAAP OpEx in the third quarter to be between $30.5 million and $32 million, with the increase from Q2 primarily coming from increased engineering headcount.

We estimate our diluted share count for Q3 to be approximately 32.9 million shares.

For the second half outlook, the substantial weakness in consumer cameras, particularly non-GoPro sports cameras, along with the previously discussed decline in consumer drones, is expected to continue through the remainder of the year, negatively impacting our fiscal 2009 outlook -- 2019 outlook.

In addition, we believe the auto market will be below our previous growth expectations in the second half as delays in launching new customer and products will reduce our year-over-year growth expectations in this market.

As a result of these changes, we now see non-GoPro revenues being minus 10% to minus 12% as compared to fiscal '18 non-GoPro revenues of $258 million.

In addition, as discussed on our last call, we expect GoPro revenues to be immaterial to the total revenues for the remainder of the year as compared to approximately $37 million for the full fiscal '18 year.

As a result of the U.S. ban on purchases of security cameras from Hikvision and Dahua, we are seeing some near-term reduction in their orders, particularly in the high end of the product range, normally associated with their export business.

We have seen an offsetting increase in orders from non-Chinese customers, but we will not see the positive impact to revenues until later in the year.

As it relates to other trade disputes, we are not currently impacted by any tariffs that have been put in place by non-U.

S. government but continuing to monitor the situation.

Other impacts from the trade dispute remain a risk, such as China customers shifting to domestic suppliers, but at this time, we have not seen these affect our business.

In addition, we do not see the dispute impacting the development efforts of new security cameras based on our CV chips at our China customers.

We expect gross margins to remain within our previous forecast of the low end of our target range of 59% to 62% for the year.

We have lowered our OpEx estimate for the year from the high end of the range of $120 million to $130 million to the midpoint of the range.

Lower revenue expectations, along with a reduction to the budgeted hiring rate, are driving the lower OpEx levels.

Due to the forecast reduction in the consumer markets, we are transferring some of these headcount resources to the development of our security, automotive and new robotics market, thereby, reducing the need to add as many new employees.

Before I turn the call over for questions, I would like to extend my deepest thanks to the rest of the executive team and all the Ambarella employees that have supported me during the last 7-plus years of my time with the company.

I believe Ambarella has some of the most talented and innovative minds in the industry, and I've been lucky to have been a part of the team.

Thank you for joining our call today.

And with that, I will turn the call over to the operator for questions.

(Operator Instructions) Our first question comes from the line of Kevin Cassidy with Stifel.

Congrats, George, on your retirement, and great working with you, and welcome, Casey.

My first question is on the CV devices.

Do you have any estimate yet of when you might see the first revenue for your CV devices?

Yes.

So, Kevin, this is Fermi.

We do believe that we will see our first CV revenue for the IP security camera first half next year.

And, in fact, we are tracking the progress very well.

Like I said, we have several design activities with several international professional IP security cameras at this point.

And with the progress, we do believe that we have a chance that -- to deliver first CV revenue first half next year.

Okay, great.

And on the automotive, you mentioned some delays in product ramps.

Is that related to the three-way camera systems, maybe from Gentex?

No, it's -- for the -- I think you're referring to what George said about the second half total revenue.

I think that really relates to our recorder business in Japan.

There is some product delayed and promotion by -- from 1 key customer that delayed the promotion.

So I think we start seeing some drop on the forecast on the auto recorder side, but I don't think that's a long-term effect.

Okay.

And maybe if I can -- let me just follow up with the Gentex.

Is there forecast on Gentex yet?

So we haven't told about Gentex forecast yet.

We are working with the e-mirror solution, 3-channel e-mirror solution with Gentex today.

And like I said, our first -- our first 3-channel e-mirror revenue came from China, and we haven't given any guidance on the Gentex revenue yet.

Our next question comes from Joe Moore with Morgan Stanley.

I have 1 shorter-term question and 1 longer-term question.

In terms of the quarter that you're guiding for, can we just get a sense for how much consumer business is left there versus the surveillance and automotive?

And, I guess, as you think out to next year, I feel like surveillance and automotive should both grow.

And so, I guess, could we be getting close to the end of these consumer headwinds?

Just give us a ballpark of that exposure.

Yes, I think if you look at the second half and the year-end total, I would think security in total is going to be at least 60% of revenue, and then auto will be in the low-20s.

So that leaves wearable and a little bit of infrastructure for the remainder.

So yes, it's getting to a relatively small percentage of our revenue at this point.

Okay, that's helpful.

And then longer-term, when I think about the automotive opportunity, I feel like a lot of people think that you guys are sort of competing directly with big companies like NVIDIA, and yet it seems like when we talk to people who -- both at Ambarella and other places who are working on this, that there's opportunity for either coexist with some of those existing solutions.

And I think you (inaudible) that the right way.

And is there a role for sort of multiple vendors and for you to sort of do perception offload in those vehicles that are already established prototype customers for people like NVIDIA?

Right.

So you referring to the -- we competing with NVIDIA is really referring to (inaudible) the market levels 2 to level 5 autonomous driving, because for the front camera ADAS category, I don't think we are not -- I don't think we are competing with NVIDIA on there.

So let's focus on level 2, level 5 at this point.

I think for that particular market, because we are offering a really flexible open perception platform, we are focusing perception [module] at this point.

And we have shown our customer that we have much better CNN performance per power -- per watt and also that we can do a very high-quality video processing, high-performance CNN performance, and that really helped our position to become a perception solution provider for those level 2 to level 5 guys.

Of course, in the long run, we want to have a complete solution, but today, with the CV2 and CV22, we are really providing a perception solution for those customer, and it's been well received.

And I will say that we have several very serious engagements in this autonomous vehicle category right now.

And I also want to add another point, which is in the front camera ADAS category where NVIDIA is not there, I think our main competitor is really Mobileye.

And, in fact, in order to accelerate (inaudible) market, we kind of focus on customer and partners that have -- developing their own algorithms.

And so we are really a hardware and platform supplier and working with the Tier 1 or OEMs with their own algorithm or third-party algorithm to enable their front ADAS cameras in this category.

And we have made a significant progress with several OEM, Tier 1 targets and engaged in that evaluation with multiple other parties globally.

So, really, that if you look at -- there's 2 market (inaudible), the front camera ADAS category as well as the level 2 to level 5 side, I think both have made progress.

And just to clarify that.

I mean, the perception module that you're talking about, I mean, you actually, in some cases, could be sitting alongside NVIDIA and complementing the functionality of NVIDIA solution?

That's correct.

In fact, I think some of our engagement is in that kind of category.

Our next question comes from Ross Seymore with Deutsche Bank.

George, thanks again for all the help over the years, and welcome aboard, Casey.

I guess, the first question in the near term, George, you updated the full year guidance.

The implied fourth quarter is down kind of, I don't know, 5%, 10% quarter-over-quarter again.

I suspect the same general trends would be the reason for that incremental drop-down, but just want to confirm, is it continuation of the trends and nothing more specific to the core IP security and auto businesses?

That's correct.

Actually, the wearables section is causing a decline to the whole second half.

Actually, in Q4, when you take out the GoPro revenue from last year, the decline year-over-year in non-GoPro revenue is going to be less than it was in Q3.

So a smaller decline in Q4.

Got it.

And, I guess, switching over to the CV side of things and maybe just want to talk in the first half of next year or even more long term.

But for me, as we think about the implications of what that can bring, I know there's a couple of different end markets that you guys are targeting, but if I recall correctly from some of your prior comments, there's a good ASP uplift if that business roll through as well.

Can you just talk about how we should think generally about that mix between units and ASPs in the various markets?

And is it fair to assume that the ASPs being higher is a good proxy for it being a tailwind to gross margins?

Or is there a COGS dynamic within those CV chips that makes that a poor assumption?

Right.

So I think that most CV22 and CV2 will increase our ASP dramatically.

And we -- although we are just starting engaging on the price negotiation with our customers, our feeling is CV22 will be 2 to 3x higher than our current average ASPs, I mean, the security camera market.

And CV2 is probably even higher than that.

And from the gross margin point of view, I think we are targeting to having the gross margin stay within our corporate gross margin.

And that's still the 59% to 62% range, I believe?

That's correct.

Got it.

And then, I guess, the last question, just bringing the whole tariff dynamic in, and thank you for your detailed comments on that.

But is the biggest risk here that the non-China-based folks just don't offset the China-based folks in the near term?

Or do you believe that what DJI's doing, et cetera, bringing it in-house is something that could occur outside of the drone market and occur in the IP security side of the equation as well, if, in fact, tariffs would be something that would be a catalyst for those companies to move that direction.

Yes.

I think the impact on the -- let's say, the U.S. government ban is going to be relatively small in the second half, a little bit more in the near term as cancellations will be in the books before actual shipments into the new customer.

Longer-term, I actually think it will benefit revenue.

The -- first of all, our ASPs are higher to the non-Chinese customers, so they will be fulfilling demand at a higher ASP and a higher margin.

So I think there's some longer-term benefit.

In addition to that, I believe the view of the HiSilicon chip outside China is becoming more negative.

So that becomes less of even a competitor in the marketplace outside China.

So I think that will also benefit us.

Well, in terms of our customer continue to do their own silicon, I think that's a trend a lot of (inaudible) company are taking on because DJI and GoPro, their ASP is so high.

In fact, we know that we have much cheaper -- smaller die size and much better product, but they don't really need to care about die size.

But we come to a competitive market that (inaudible) I knew that Hikvision and Dahua has to have the -- working on their chip, but throughout the years, we continue to be their supplier.

So I won't rule out the possibility that our customer continue to work on it.

That's something we'll continue to deal with.

But I think when the CV chip comes out, we really raise our bar (inaudible).

When people catch up our video-only solutions, I think that CV will create a bar for us, provide our barriers of entry for a while.

(inaudible) Go ahead.

Go ahead.

Just one quick housekeeping one for George.

And you provide a lot of details about the taxes currently and looking backwards, but if we think looking forward on a pro forma basis, how should we think of the tax rate?

I think the tax rate will probably -- if I was to estimate it now, it's probably around the 10% level for next year.

For the rest of this year, probably slightly below that.

Our next question comes from Matt Ramsay with Cowen.

I'm going to follow up on some stuff that Ross was just asking about guidance with respect to the security market.

Maybe you could sort of remind us what the mix of that business is today, both Chinese customers versus Western customers, or maybe camera shift into the China market versus cameras externally.

However you want to divide it up, I think just level-setting those mixes within the business would be helpful to investors.

Yes.

I think the mix right now, and we can just use dollars, is probably about 55% into China customers, 45% into -- sales into non-Chinese customers.

Got it.

Got it.

That's helpful.

Just to kind of level set that.

And then a question that, I guess, I get most often now from investors is really around the CV engagement in the automotive market.

You guys have obviously done some demonstrations at the Analyst Day around some driving platforms, and there's been lots of discussion about engagements.

But I think investors are just looking for some more formal engagement announcements with partners in Tier 1, and maybe you could update us on what kind of time lines we should expect around more formal engagement announcements that sort of give some guideposts to where investors are thinking about time lines for revenue in the automotive space.

Okay.

Thank you.

So like I said, since we already have some significant progress with some of the OEM Tier 1 discussion, ADAS as well as some engagements in autonomous vehicle, we are still targeting -- have some announcement at the end of the year.

That's our target right now.

And in terms of revenue, like you know, it's going to take a while.

We start working on recorder business for OEMs 3 years ago, and we start seeing revenue just last year.

So just showing you even a relatively simple solution like a recorder take 2 years in China to be in production.

So we expect that CV will take longer.

However, I do believe that the requirement and performance requirement and the power consumption limitation in the automotive space is -- will benefit us, and we assume that we have a very significant advantage (inaudible) solutions.

So I remember optimistic about our chance to making it into market.

Got it.

And then just the last one I could sneak in.

George had talked about sort of the new guidance range for this year, and it seems like the majority of the consumer businesses will be out of the run rate here exiting the year.

Is it too early?

Or is there an opportunity here that sort of set expectations for where you guys are forecasting the business for fiscal '20 just on the top line to keep folks level set on expectation?

That will be really helpful.

Yes.

I think it's a little early to do that.

We really need to get a handle on launch of the new products in the first half.

And then once we understand the business a little better, I think it'd be better to wait until those facts are in or at least a better picture of those facts before we talk about next year.

But as you said, pretty much most of the consumer business will be down to a minimal level by the end of the year.

So we'll be running pretty much a high percentage of security business, 60-plus percent, plus 20-plus percent in the auto space.

Our next question comes from Suji Desilva with Roth Capital.

Casey, welcome.

So in terms of the end market for the CV products, I'm wondering how you would size the addressable market for auto versus non-auto and if the robotic security camera portion of CV might come in sooner than the auto opportunity.

Well, first of all, I think IP security camera will come first.

I think the industrial machine learning type of robotic application will come second, and auto will come last.

In terms of the size, I really think security camera, because the audio (inaudible) the total market size is more than [100 million units] already.

And based on our conversation with customer, they have different expectation.

But I won't be surprised that as soon as the very cost-effective CV-based security camera will be available in the market, the transition from current video-only solution to CV-based solution will be fast because the demand is real and there's real use for those kinds of CV function.

So I think that we're expecting a quick transition from video-only solution to CV for security camera.

And for the robotic application, I think you can see that there are many company working on it, OEM -- several company is big enough to generate meaningful revenue at this point, but however, we do believe that this is a great opportunity for us because our CV2 (inaudible) highest possible performance.

And also that -- and also, we have a stereo processing, which is important for machine learning, as well as the CNN performance and power consumption.

So I do believe that the industrial level of robotics or even robotic by consumer level is a huge opportunity for us moving forward.

Yes.

Just 1 comment on the robotics.

I mean, since our launch of the CV2 chip, now we've been able to actually start engaging in discussions with robotics companies.

So I think the initial response is positive, and I think the architecture of what we're delivering fits their need.

Interesting.

Are you seeing revenues already for CV from IP security cameras?

Or is that -- if not, when will that be coming?

First half of next year, as we've said in the past.

It's still on target.

Got it.

Okay, that (inaudible).

And then maybe a more technical question in terms of the software and how you provide that with a CV.

Are you guys -- do you guys have your own software stacks (inaudible) customers to have fast time to market?

Or are you entirely relying on customer software stacks and (inaudible) programmability?

Or is it a hybrid approach?

I was curious.

I heard some comments from (inaudible).

Right.

In 1 sentence, it's a hybrid, but let me be more clear.

If you take out the new (inaudible) portion first, it's just -- if it's just camera by itself, we provide complete software stack so that a customer can leverage it without wasting too much of their time to build the camera.

But for the new (inaudible) portion, we have our tools to (inaudible) their neural network for other platforms to our inference engine on our chip.

And that's how we help several customer already -- in fact, many customer already to put their network on our chip.

That's how we confirm our performance not only running the public domain neural network but also running customer neural network.

However, we also have our in-house neural network.

The development with that is try to prove the concept.

We have very high-performance, low-power solution.

And if customer choose to use our now neural network, we will sell them the complete turnkey solution.

But that's not our main business model.

Our business -- our main business model is to using our SDK, helping our customer build camera and helping them to put their own neural network onto our platform.

That's our key business model.

Our next question comes from Quinn Bolton with Needham & Company.

George, just wanted to say it's a pleasure working with you, and welcome, Casey.

George, just wanted to come back.

If you look at the third quarter guidance, I don't know if I caught this or if you mentioned it, but will both security and auto be up year-over-year?

I believe auto will be flat, but security will be up.

Okay, okay.

And then you mentioned a couple of times that the consumer business is probably down to an immaterial level by the end of the January quarter.

Do you guys have any sense what fiscal '20 might hold?

I mean, is that a market that you expect to rebound?

Do you think, for modeling purposes, we should just assume that it's based at an immaterial level?

I mean, any thoughts just directionally how that business may or may not recover in calendar '19.

Yes.

The business is running -- will be -- the non-security and auto business, so taken those 2 out, you're going to be running other businesses in the high teens, maybe 20% of revenue.

I don't see it expanding from that base.

And even if the revenue dollars stayed flat, they would probably start declining as a percent of overall revenue as we roll out CV next year.

Just a clarification, George.

It sounds like it may be high teens for those businesses in fiscal '19, but your exit rate is going to be likely in the single digits.

Correct?

No.

It will be higher than that in Q4, still be -- it'll still be in the teens in Q4.

But what I'm saying is, as a dollar value, I don't see that expanding going forward.

So it will become -- the percent of revenue will decline next year as the other markets expand in CV.

Got it.

And then I just want to come back to the professional security market.

You talked about a very high ASP increase on the CV22 family of 2 to 3x increase relative to that current data processor you're shipping today.

What percent of the professional security market do you think will be able to afford that kind of price premium?

I mean, are you talking about selling the CV22 really only into, say, the high 10% to 20% of the market?

Or do you think you can see better penetration?

And then a follow-on question.

You sort of hinted at (inaudible) nanometer based out of a cost-reduced chip.

Is that cost-reduced chip for the security camera market?

And will that help drive adoption of CV in the security market, as Fermi alluded to in a previous question?

Right.

First of all, I think the percentage of penetration really depends on how aggressive our customer want to price their camera.

And with our [C22] ASP, I do think that you're performing to the category probably top 20% of the market right now based on the current -- the price of the security camera system price.

But you're absolutely right that we are expanding our performance price point (inaudible) 10-nanometer chip because we do expect that for us to continue to expand the transition from the video to CV solution, our customer will demand lower price but competitive solution that we need to be prepared for that.

Fermi, just for that new 10-nanometer (inaudible) I assume that will still carry some ASP premium to your current video processor.

It's just probably not going to carry a 2 to 3x premium.

Is that a fair assessment?

That's correct.

Our next question comes from Adam Gonzalez with Bank of America Merrill Lynch.

Just wanted to focus in on the first half '19 revenues in the -- for CV and security.

Just wondering if you could perhaps quantify what the impact will be if those revenues full ramp into the rest of the year and if you could just give us a rough percentage of what CV will be as a percentage of security next year.

I think it's a little too early to do that until we get the launch dates and determine how many cameras each of the customers are going to build and launch.

I think we have better visibility towards the fourth quarter of this year to be able to give you that kind of data.

Okay.

Got it.

And then on OpEx, just can you remind us -- I don't know if you put this number out there, but what percentage of your OpEx, your R&D do you spend on CV development?

Yes.

The -- that's been increasing.

At the last number we gave, it was around 60% of our engineering dollars are spent on CV or -- and/or auto.

And I think this year that will shift, probably closer to 70% to 75% of our engineering dollars will be shifted over there.

Our next question comes from Charlie Anderson with Dougherty.

My congrats to George as well, and welcome, Casey.

Great to hear about the CV wins on the security camera side.

I wonder, given the split you mentioned with China, non-China, do you have any China customers among those wins?

And then I've got a follow-up.

Yes.

Actually, we're being very successful in China with the CV products.

They're probably ahead of the non-Chinese customers as far as end camera product development and testing.

Right now, outside of China, we have multiple customers in evaluation and starting to build cameras, but China actually is ahead of non-China customers.

Great.

And then for my follow-up, Fermi, you mentioned that 10x faster processing.

I wonder if you could maybe help us with how you're defining that.

Is that an image recognition kind of thing.

And is that based off CV2?

And then as you're comparing it, what are we comparing it to?

Is that the HiSilicon chip?

Or are we also talking about NVIDIA and internal solutions?

(inaudible) any more color there would be helpful.

We compare to HiSilicon.

We compare to other CPU-based solution.

We compare to (inaudible).

There has been the frontrunner in the IP security (inaudible) in the past.

And we also compare with some of the traditional automaker guys who has shipping some kind of a DSP-based solution out there.

And the way we compare it is twofold.

One is (inaudible) public domain neural network on our CV22 or CV2 and also running the same network on those platform I mentioned, if we can get access to it.

So, for example, there are platforms out there you can buy, that you can run the same network at the -- and then you also -- because some platform has public domain information about their performance, you can collect all of that to benchmark ourselves.

That's 1 portion of the data we collect.

The other data collection is really that our customer doing similar evaluation based on the (inaudible) network, the public domain neural network running our platform and our competitive platform at their (inaudible).

And we collect some of the result from that, get feedback on that.

The third thing we did is, really, we also helped our customer port their own network running on CV2 or CV22, compare to the same network, the customer network running on their current existing hardware platform based on the other competitive solutions.

So we collect all the data.

And when I say 10 times (inaudible) comparing to CV22.

And CV2 is even higher, better advantage over that.

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

George, congratulations on your retirement.

And, Casey, I look forward to working with you.

I guess, my first question will be on the consumer security since I don't think there's been any real questions on this.

I think you talked about some growth in that business and talking about some growth outside of the United States.

If you can give us a sense first on sense of scale on that business.

Talked about security being, I think, roughly 60% a little bit more.

Maybe give us a sense of how big consumer is.

And then what kind of growth path can that deliver?

Is that a 10% growth or a little faster?

And can we see some contributions even outside the U.S. and Europe?

Well, I'll take the split, and then I'll let Fermi talk about the -- outside of the U.S. Right now, it's about 1/3 consumer cameras, 2/3 professional cameras, and that's based on dollars.

And then I'll answer the other one.

As far as expansion of the home security market, we're actually starting to see Europe.

We have several, as Fermi announced (inaudible), we have several customers in Europe now are starting to market the products into the home there.

So we think that will help growth in that marketplace.

We still don't have really clear visibility in Asia, but we do know both HiSilicon -- Hikvision and Dahua are producing home security cameras.

We just don't have necessarily a good split as to professional versus home security.

So we think that, that marketplace can continue to grow and will be -- the growth will be assisted by some geographic expansion in customers.

Okay.

Perfect.

I'll ask one more question, jump out of line.

You talked about a fairly fast transition within the IP professional security market from non-CV to CV solutions.

Any sense of how fast you expect that to be?

And do you expect that to happen with (inaudible) kind of the top Tier 1 OEMs out there?

Well, I think you got different feeling when you talk to different customers.

I will say that our Chinese customer are the most aggressive one because they have real demand to switch from the current video-only solution to CV1.

When you talk to our international customers, non-Chinese customers, some of them are very aggressive, some of them a little bit behind because they think they can use different way to solve these problems (inaudible).

So that's become a more controversial discussion for the CV solution.

But nonetheless, and just -- for example, Hikvision has been telling -- in their own conference call saying there should be 1 million units of CV-based camera today, and that one's very high priced compared to the current -- very, very high priced compared to what they're selling right now.

So let's just give you an indication.

Even at that kind of price, they still try to push many units of a CV-based security camera in China, that show you that there's real demand for this kind of functionalities.

I think one thing we have to take into consideration is also the Chinese government who are pushing very heavily for more and more information, and security cameras are one of the areas they're pushing.

And computer vision helps actually develop some of the data they would like to collect.

So I think one of the reasons there's more acceptance in China is because the government push to install these cameras.

I'm showing no further questions in queue at this time.

I'd like to turn the call back to Dr. Fermi Wang for closing remarks.

Thank you.

And I would like to thank all of our colleagues for their outstanding effort to (inaudible) CV platform so that we can establish ourselves a competitor in the space.

Thank you very much, and I'll talk to you next time.

Ladies and gentlemen, thank you for your participation in today's conference.

This concludes the program, and you may now disconnect.

Everyone, have a great day.