Ambarella Inc (AMBA) 2020 Q1 法說會逐字稿

Good day, ladies and gentlemen, and welcome to the Ambarella's First Quarter Fiscal 2020 Earnings Call.

(Operator Instructions) As a reminder, this call is being recorded.

I would now like to introduce your host for today's conference, Mr. Louis Gerhardy, Director of Corporate Development and Investor Relations.

You may begin.

Thank you, Skylar.

Good afternoon and thank you for joining our first quarter fiscal 2020 financial results conference call.

Our speakers today 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 the first quarter of our fiscal 2020.

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 file with the SEC, including the Annual Report on Form 10-K filed on March 29, 2019 for fiscal 2019 ending January 31, 2019.

Access to our first quarter fiscal 2020 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, Louis, and good afternoon, everyone.

Before I address our financial results, I would like to comment on the geopolitical factors injecting higher levels of uncertainty and volatility into our business.

Although there is currently no regulation in place, there is speculation our largest security camera customers in China may become subject to U.S. government regulations limiting or restricting our shipments to them.

In addition, some of our customers, both in China and outside of China, will be facing higher tariff when their products are imported into the U.S. During these uncertain times, market share may shift between our customers, and such transition could be subject to disruptions.

Furthermore, our non-security camera customers in China may also be subject to the unfavorable geopolitical forces.

Our largest competitor in the security camera semiconductor market, HiSilicon, a unit of Huawei, may also face headwinds of their own.

As a result, one can see the environment is volatile with a wide range of potential outcomes, for example, in the last two weeks we have received order pull-ins from security camera customers in China.

In terms of our financial performance in Q1 fiscal year 2020, the revenue of $47.2 million was slightly above the midpoint of our guidance and declined 8% from the $51.1 million in the prior quarter.

During the quarter, we demonstrated solid progress toward the multi-billion-dollar expansion of our addressable market.

Our deep domain expertise in video processing, integrated with the computational horsepower of our proprietary CVflow AI processor, yields a highly optimized computer vision solution enabling our customers' products to become more intelligent.

An important milestone for us is that we started mass production of our second CV SoC, CV25, in Q1 with shipment scheduled in Q2.

Also during Q1, we continued the early ramp of CV22 shipments and maintained strong design activity in our core automotive and security camera market.

I will now talk about our progress in our target security and automotive markets, starting with security.

During April's ISC West security convention, Ambarella demonstrated its full range of IP security solution including our CVflow family of computer vision SoCs as well as announcing our new S6LM SoC.

At the show, many leading professional camera suppliers demonstrated their new AI-enabled cameras based on our CVflow SoC, in advance of formal product announcements to be made over the coming months.

This included a leading Korean supplier Hanwha Techwin, which demonstrated professional IP cameras based on Ambarella's solutions for the first time.

Our CVflow SoC are in design at 9 of the 10 leading professional security camera suppliers.

We are continuing to displace SoCs from HiSilicon and computer vision co-processors, including those from Moviduus and FPGA-based solutions, in new designs.

We also believe that around five professional IP camera makers should be in mass production of cameras based on our CVflow SoCs in the second half of this year.

During the show, we introduced our new S6LM camera SoC, targeting both professional and home security cameras.

It includes Ambarella's latest High Dynamic Range (HDR) and low-light processing technology, highly efficient 4K H.264 and H.265 encoding, multi-streaming on-chip 360-degree de-warping, cybersecurity features and a quad-core ARM CPU.

Fabricated in 10nm process technology, the SoC has very low-power operation, making it well suited for small form factor and battery-powered designs.

Fully software compatible with our CVflow family of chips, the S6LM expands our IP camera product portfolio to address the most cost-conscious applications with no compromise on image quality.

In addition to our ongoing investment and progress in AI processing, we are continuing to advance our image processing technology, which remains a critical differentiator in both security and automotive applications.

At ISC West, we demonstrated the low-light capability of our new SoCs by generating full frame rate color video in less than 0.07 lux lighting conditions, or near total darkness.

At these light levels, previously cameras have had to turn on power-hungry IR-LEDs for illumination, which provides only gray scale images.

This new low light color capability is available on all of our new CVflow and S6LM SoCs.

We have now worked with over fifty customers and partners to successfully port their deep neural networks to our CVflow SoCs, with our advanced tools typically enabling network porting in a few days.

A number of leading third-party computer vision software developers joined us at ISC West for demonstrations.

These included AnyVision's face recognition, FFGroups vehicle and license plate recognition, Pilot.

AI's person and vehicle detection, TrueFace's access control, IntelliVision's license plate recognition, and XNOR's person detection and recognition.

In the consumer home monitoring market, Vivint, a leading smart home company owned by Blackstone Group and serving over 1.4 million customers in the U.S. and Canada, introduced its first Ambarella-based product, the Outdoor Camera Pro.

The camera is sold as part of Vivint's complete integrated home security system and 24/7 monitoring services.

Based on Ambarella's S5L SoC, it includes a 4K sensor, 1080p video with HD zoom, night vision, 2-way audio and ability to both detect and deter threats.

I would now update you on our continuing progress in automotive markets.

In April, Ambarella exhibited at the Shanghai Auto Show for the first time.

We hosted a significant number of leading Chinese automotive OEMs and Tier 1 suppliers, while Dr. Alberto Broggi, General Manager of Ambarella's VisLab team, delivered a speech highlighting autonomous driving technologies to over 600 attendees.

Our technology demonstration included our partner HELLA Aglaia's front camera AI-based ADAS software running our CV22, CVflow SoC.

We also demonstrated driver monitor solutions from partners EyeSight and Readsense.

The Readsense camera run on Ambarella's new CV25, CVflow SoC.

And yesterday, we announced that we are working with leading Chinese Tier 1 supplier, [Shenzhen Longhorn Technology] (corrected by company after the call), to develop automated parking system based on a combination of visual and ultrasound technologies with sensor fusion software.

The platform uses Ambarella's CV22AQ SoC throughout multiple AI algorithms simultaneously to enable free space detection, parking detection, vehicle identification, obstacle type identification and emergency braking in a full 360-degree panoramic parking system.

Yesterday we also announced a collaboration with [Shanghai] (added by company after the call) Baolong, a leading Chinese Tier 1 supplier to the automotive sensor and ADAS market.

Baolong's DMS AI algorithm run on our CV22AQ SoC create a high-performance, low-power and a compact form factor driver monitoring systems.

The driver monitoring system solution captures the driver's facial information, including characteristics of the eyes, head and mouth from an image sensor, and the CV22AQ provides, real-time both video processing and AI computational horsepower to determine the driver's level of fatigue.

In April, we announced that we are working with China-based Momenta.ai on a collaborative HD mapping platform for autonomous vehicles.

The combined solution leverages the broad CV22AQ SoC and Momenta's deep learning algorithm to provide HD map solutions, including localization and map updates through crowdsourcing.

The solution can create a closed feedback loop of big data, AI and HD map update.

We expect to expand our partnership with Momenta to include development of ADAS and Level 2+ autonomy solutions in the future.

Also in April, we announced that we are working with Chinese Tier 1 automotive supplier, Shanghai OFilm, a leading provider of smart driving solution on automated valet parking, or AVP, hardware platforms.

The platform uses Ambarella's CV22AQ SoC and OFilm's intelligent driving algorithms to perform high precision image recognition, free space detection and obstacle detection, and enable emergency braking and map construction.

In the OEM automotive market, we are continuing to win new designs in car recorder applications, as this kind of product increasingly becomes a pre-installed option in new vehicles.

In the last two years our drive recorder SoC have been incorporated into 18 new car models with 10 different OEMs.

These OEMs include Geely, Honda, Toyota, Nissan, SAIC, Dongfeng, Changan and the BAIC.

During Q1, a number of Chinese automotive customers introduced new car models with recorders based on Ambarella's A12 SoC.

This included Geely's Jiaji MPV, with a car recorders supplied by Tier 1 Longhorn; Chery's TIGGO 7 Fly SUV with the car recorder supplied by Tier 1 Skyworth, and BYD's Song Max MPV.

In the automotive aftermarket, Garmin announced its new "mini" family of compact dash cameras consisting of four new models ranging from $129 to $249.

All based on Ambarella's A12 SoCs, the new cameras include features such as 1440 resolution video, 180-degree field of view, parking security recording and the ability to link up to four cameras together to form a synchronized network that provides full coverage around a vehicle.

And in April, Ambarella's Italian team, VisLab, obtained authorization to test autonomous vehicle on public roads in Italy, the first authorization to be issued by the Ministry of Infrastructure and Transport, under the new Smart Road Decree.

The authorization was obtained based on evaluation of VisLab's technology, its proven experience in testing in real-world environments and the vehicular safety records.

Commemorating the occasion, Ambarella's EVA autonomous vehicle was demonstrated to the Italian Deputy Prime Minister and Minister for Economic Development on the iconic Torino Ligotto test track, built on the roof of the former Fiat factory.

And in the consumer camera market, drone market leader DJI introduced its OSMO action camera, raising the bar for features and image quality in the action camera market.

Priced at $349, and based on Ambarella's H2 SoC, it offers 4Kp60 video recording, full frame-rate 3-axis electronic image stabilization called, "Rocksolid" video HDR processing and excellent low-light imaging.

The waterproof camera also feature a front facing color display to eliminate guesswork when taking selfies.

In summary, despite the geopolitical headwinds, I'm pleased with our execution and the foundation we are building for the secular growth opportunity in AI and computer vision market.

With multiple CV devices now in production, with more CV design wins, and with new technology and products in development we see three CV waves taking shape.

CV revenue from the professional security camera market has started, and we expect it to become material in calendar 2020.

The second CV wave is expected from the consumer security camera market, expected to commence in the second half of calendar 2020 and become material in calendar 2021.

The third CV wave, automotive, is expected in calendar 2022-23 with the opportunity to be the largest wave yet.

I would like to recognize the Ambarella team for keeping their heads up and executing a plan despite all the macro uncertainty.

Thank you to all of our stakeholders, including our 750 employees, customer, vendors and shareholders for your trust.

Now I will turn the call over to Casey for the first quarter financial details and the guidance for our second quarter of fiscal 2020.

Thank you, Fermi, and good afternoon, everyone.

Today, I'll review the financial highlights for the first quarter of fiscal 2020 ending April 30 and provide a financial outlook for our second quarter of fiscal 2020 ending July 31.

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, we have eliminated stock-based compensation expense adjusted for the impact of taxes.

Consistent with our previous guidance, our Q1 revenue totaled $47.2 million, which represents a decrease of 8% from Q4 and a decrease of 17% when compared to the same quarter of the prior year.

In Q1, revenue by market was in line with expectation, automotive and "other" product revenue was up sequentially, and security was down sequentially, with consumer security increasing and professional security declining.

Non-GAAP gross margin for Q1 was 59.6% compared to 60.6% in the preceding quarter and slightly above the midpoint of our guidance.

Non-GAAP operating expense for the first quarter was $29.9 million compared to $29.5 million for the previous quarter.

OPEX increased primarily due to an increase in payroll taxes and other payroll-related expenses.

Other income of $2.2 million was higher than anticipated as we received a $346,000 onetime government subsidy.

Non-GAAP net income for Q1 was $266,000 or $0.01 per share compared to non-GAAP net income of $4.5 million or $0.14 per share in the fourth quarter.

The non-GAAP effective tax rate in Q1 was approximately 36%.

In the first quarter, the non-GAAP earnings per share were based on 33.5 million shares as compared to 32.8 million shares in the prior quarter.

Total head count at the end of the fourth quarter was 750 with about 81% of our employees dedicated to engineering.

Approximately 70% of our total head count is in Asia.

In Q1, we generated positive operating cash flow of $5.3 million.

Cash and marketable securities were $366 million up from $359 million at the end of the fourth quarter.

We did not repurchase stock during the quarter under the $100 million repurchase program authorized through June 4, 2019.

On May 29, the company's Board of Directors authorized the repurchase of up to $50 million of the company's ordinary shares through June 30, 2020.

Total accounts receivable at the end of Q1 were $26.5 million or 50 days sales outstanding.

This compares to accounts receivable of $26.2 million or 47 days sales outstanding at the end of the prior quarter.

Net inventory at the end of the first quarter was $17.6 million compared to $18.3 million at the end of the previous quarter.

Days of inventory declined to 85 days in Q1 from 93 days in Q4.

We had two 10% plus revenue customers in Q1.

WT Microelectronics, a fulfillment partner in Taiwan, came in at 59% of revenue; and Chicony, a Taiwanese ODM who manufactures for multiple customers, came in at 18%.

I will now discuss the outlook for the second quarter of fiscal 2020.

Before the recent geopolitical news, we were comfortable with the market expectation for our second quarter fiscal year 2020 revenue outlook.

However, due to recent order pull-ins, we are now raising our revenue guidance above the market's expectation.

We expect total revenue for the second quarter ending July 31, 2019 to be in the $51 million to $53 million range.

We anticipate automotive and security revenue to increase sequentially, while "other" revenue is expected to decline.

We continue to expect "other" revenue comprised primarily of customer electronics products, to remain volatile and decline as a percentage of revenue in the next 2 to 3 years.

Although forecasting is difficult in the current environment, we continue to expect our revenue will increase in the second half of the year when compared to the first half.

The global economic and political environment is generating a significant amount of uncertainty with our customers' outlook and our customers' supply chains.

Overall conditions remain extremely volatile.

A number of factors impact our gross margin outlook.

In anticipation of export restrictions that might be implemented, our professional security camera customers in China, have pulled in orders, which are lower-margin products.

In addition, rising tariffs may have a greater impact on our customers' ability to sell their products incorporating our higher-margin devices.

There is also some supply chain congestion in the packaging and test area that may make it difficult and/or more expensive for us to pull in orders.

As a result, we estimate Q2 non-GAAP gross margin to be in the 57% to 59% range as compared to 59.6% in the first quarter.

We expect non-GAAP OPEX in the second quarter to be between $29.5 million and $31.5 million due to increased engineering head count, payroll tax accruals and other engineering expenses.

The Q2 tax rate should be modeled at 32%.

The tax rate is a result of our inability to use tax losses in certain jurisdictions, and the tax rate should return to the high single digit as we become more profitable.

We estimate our diluted share count for Q2 to be approximately 34.1 million shares.

Ambarella will be presenting at the Bank of America Merrill Lynch 2019 Global Technology Conference on June 5, the Stifel 2019 Cross Sector Insight Conference on June 10 and participating in CES Asia June 11 to June 13.

Please contact Louis for more details on these events.

Thank you for joining our call today.

With that, I'll turn the call over to the operator for questions.

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

Congratulations.

Just to be clear on the pull-ins, these were because of HiSilicon and concerns of them not being able to ship.

It's not a concern that the U.S. may ban surveillance products or your products being shipped into surveillance products.

Do I have that right?

Well, it's hard for us to have perfect visibility into that.

But I would say, primarily, the pull-ins are related to customers' concern about a ban or potential ban.

And so I think we're seeing some order patterns as we've seen over the last quarter or so move around a little bit, and we think that we're already seeing some pull-ins into this quarter.

And that's what we're referencing.

From a HiSilicon standpoint, we may see some increased orders due to concerns around HiSilicon, but it's hard for us to have visibility into that.

And so should we be thinking of the second half -- I know you're saying it's still higher than the first half but maybe not as high as you were first thinking because the first half is going to be higher than expected.

I think what we're going to see is movement kind of quarter-on-quarter over the next 2 to 3 quarters.

We're still -- we're seeing the type of traction we were expecting, especially across the CV products and some of the other products, but as all of this gets digested and we try to get some resolution, it's a little difficult for us to have as much visibility as we typically do.

But what we're seeing right now is pretty good demand and certainly demand better than we saw in the first half of the year.

And so that's why we're guiding second half over first half.

But on a quarter-by-quarter basis, we're going to have to keep you posted as we move forward because it's difficult for us to have good visibility right now.

Our next question comes from Ross Seymore with Deutsche Bank.

All these uncertain times, I appreciate you guys trying to give us a little more of a crystal ball.

To that point, I know, Casey, you said you were comfortable with what The Street expectations were before.

Just trying to size the pull-ins here.

Is it fair to say that if The Street was at $49 million to $50 million, you're thinking kind of $2 million to $3 million to hit the midpoint is the pull-in side of the equation?

Or is there other things that are moving, new design wins, something else that's helping your business above the current expectations?

Yes.

I mean most people had us starting to have some revenue growth in the second quarter, thus the $48 million to $49 million you were talking about versus what we just did.

And the income level above that, again, it's hard to have perfect visibility, but clearly, a good piece of that is pull-ins that we're starting to see into Q2.

And so exactly how much of that is pull-ins is again hard to have exact beat on, but it's clear to say that a large portion of that additional above where The Street was is pull-in related.

Got it.

And then I guess any more color you can give on the confidence you have in the back half of the year still being bigger?

Obviously, you have a lot of drivers there, but it seems like it will be less so, to Kevin's question, given that you have the higher base in the second quarter.

Can you just talk about some of the things that you have confidence in that'll create that increase into the second half?

Well, as I said, our CV products remain on target and online.

We've got some of the other products and some of the things that Fermi discussed in the call today that we think are going to really start to be contributors there.

And generally, some of the markets that have been soft are starting to come back a little bit.

Auto kind of lightened up a little bit.

Security growth had sprung a little bit.

And if you listen to our customers, they were seeing growth start come back a little stronger in the second half.

So that's all good signs.

Now again, when you get disrupted in the visibility is that all of a sudden we're not allowed to ship products or if we can't get the supply chain straightened out for our customers, obviously, that's very disruptive, and we can't control that.

But if -- barring something like that happening, I feel pretty confident and pretty good about what we're hearing from our customers today.

And also, this is Fermi, and let me add one more thing.

Before the discussion of this ban -- potential ban of Hikvision, Dahua, Casey and I feel comfortable with our previous guidance, which means that Q2 should be a little better than Q1 and second half will be better than the first half.

We were comfortable even before this order discussion or excitement of our Hikvision, Dahua situation.

Our next question comes from Joe Moore with Morgan Stanley.

My question's on the pipeline of automotive wins particularly in China.

You guys have talked about the Shenzhen OFilm (sic) [Shanghai OFilm], Momenta, Baolong, Longhorn now.

Seems like there's a lot more design wins coming from China than I would have expected.

Can you talk about why you're getting so much traction in that region?

And obviously, the HELLA win in the West was pretty important as well, but it's sort of harder to get the companies that you're working with in the West to press release those activities.

Right.

So first of all, I really think that HELLA's announcement help us in China even because a lot of Chinese customer are waiting to get a platform they can use.

And with HELLA -- basically, with HELLA's endorsement of our CV22 platform gives the confidence to the Chinese vendors who are looking for solutions.

And there was really like we talked about in the past in China, people are more aggressive and they really want to get some product they can ship in 2020 and 2021 calendar years, and there are not many good solution out there.

So with the endorsement of HELLA as well as we continue to demonstrate the progress we made on our CV chip and performance and low-power consumption, all of that -- and image quality we just recently demoed, all of that help us to establish.

I think that, at the end, the momentum we see in China really reflect that our Chinese automotive industry is more aggressive to try new solutions as well as their time to market and time to revenue is shorter than outside China.

Okay.

Great.

And I guess a follow-up.

Is the competitive dynamic in the Chinese market different than here?

I mean here we talk about companies like Mobileye and FPGAs and things like that.

In China, is it the same?

We certainly hear about a number of AI inference-related startups.

So is that a different competitive set there?

Yes.

First of all, I think between the Chinese and non-Chinese companies, there are probably 50 kind of different AI company or start-up company trying to inference engine like we are doing.

But I still think that our main competitors is the people we talk about in the past, mainly Intel, Mobileye and NVIDIA and also the traditional company like NXP, TI and FPGA companies.

I think when we show up with CV22 and CV25, we believe will establish our performance, especially our CNN performance and power consumption better than the competitor, which I talked about, and also that because our existing sales structure and customer relationship in China that definitely help us to get momentum going in China.

And on top of that, I think Mobileye and NVIDIA, although they are very big and they have very strong sales there, but their price is much higher than us in China.

All of that, I think, help us to getting there.

So we believe, in China, we are competing with, like I said, Intel, NVIDIA plus TI, NXP and the FPGA companies.

I think that's our main competitors.

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

Just a real quick one.

Can you just quantify what your overall exposure to China is?

I think we have a bigger sense what it is in security, but just trying to think beyond security and automotive and consumers as well.

Sure.

In the security question of the exposure to China, if you think about our professional security revenue, security overall is about 60% of our revenue, and the professional side of that is about 2/3 with consumers -- security being 1/3.

Of that 2/3, if you look at our revenue in the security space alone, you're going to have a majority of our revenue coming from the export business.

And the export business -- we have some internal as well, but the export business is the biggest piece for us, but for them it probably represents 20% to 30% of their business.

So when you roll all of that up, there's a decent piece of our business is in the security professional security business that's sold by our Chinese customers.

But most of that is export, and it's a smaller piece of their business.

The business inside of China is primarily our competitor, HiSilicon.

And also in auto space, I think that we talked about in our OEM car TV business, OEM recorder business really Japan and China, I will say, Japan -- sorry, China, Japan, Korea and I really think China is probably roughly 50% of that business.

So that basically gives you idea of China exposure.

And then on auto, Fermi, you gave some color on the timing of the 3 waves to the CV ramp that you see over the next few years, and you said autos would really be material in calendar '22 and '23.

Just wondering if you can give us an update on what you expect some of the near-term drivers to be in autos between the -- I know it's 100% or mostly 100% auto recorders, drive recorders today.

What can help bridge the gap between where we are today and the material CV ramp 3, 4 years from now?

Right.

So when we talk about 2022, '23, we talk about the non-China CV ramp.

And as I just said, I think the -- there will be a Chinese CV ramp earlier than the 2022, '23 time frame we talk about.

For example, a lot of our Chinese engagement we announced that the time frame is really 2020, '21, although the value is not high but definitely give us momentum moving toward that.

But even before that CV wave will -- for ADAS, there's another wave, which is DMS, driver monitoring system.

We continue to see huge momentum behind this market and they -- all of them moving to using a CN neural network with our CV25 chip for the next generation.

So I really think that we're going to see another wave of the revenue coming probably 2020, 2021 based on the DMS momentum.

Before that, I would say it's all about video products recorder would be the biggest business, which continue to grow.

We believe the total penetration rate's still low in China, and we expect that the Chinese video recorder business will continue growing in the next 2, 3 years.

In addition to that, I think that we, in this time, we talk about we have this surround view and also the smart parking system, which is using -- leveraging the surround view system with -- plus CV22 object detection, parking solution.

So all of that will be incremental to our revenue in the next couple of years.

So when we say the auto wave is really referring to that, I think our western ADAS customer can go into production, and that will give us the biggest impact on our revenue.

But before that, I think all of these applications we just mentioned and also Chinese CV revenue will help us bridge the revenue.

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

I was just trying to sort of better understand your sort of thoughts around the impact of the tariffs and the potential bans on security cameras and the current ban on Huawei.

You'd said prior to those events, you were comfortable with sort of the annual consensus estimates, once those events happened, you've seen -- it sounds about $3 million pull-in into the July quarter, which I assume would have otherwise been recognized later in the year.

So I sort of understand that pull-in effect, but I guess my question is do you not see any risk to the second half just from overall tariff uncertainty, the supply chain tightening inventories as a result.

And any potential ban of security cameras to the extent that those entities are placed on the U.S. entity list?

Well, that's definitely a concern.

That's why we didn't give any guidance on the Q3, Q4 second half because all the risks you just mentioned.

The only thing we said so far is before this discussion, before this potential ban being talked about, we were comfortable.

And after that, the only thing we comment was that we see Q2 pull-ins.

We assume that if this geopolitical situation continues, we will see continued pull-in in Q3 until the ban starts.

It probably won't, but it if starts, it will definitely -- will put more risk on our Q3, Q4 outlooks.

But with so many uncertainties, at this point, we just don't want to give any guidance for Q3, Q4 at this point.

Got it.

And then [pushing] the question on sort of the exposure to some of the Chinese security OEMs that could potentially go on the entity list, I think in aggregate that could be roughly 25% of total sales.

You had mentioned in prepared script, I think, for me, that maybe it was only Hikvision that could go on that list.

Could you -- to the extent that it's only one entity that goes on the list rather than multiple entities, would you be willing to share what the exposure specifically is to Hikvision versus Dahua?

Or would you prefer not to get to that level of specificity?

Well, we are not making any assumptions at this point.

Both Hikvision, Dahua we mentioned in those media press, and we are doing all kinds of calculation assuming 1, 2, none or both.

But at this point, we haven't given any guidance about potential impact if any of the scenario happens.

We want to wait.

Well, in fact the biggest question is entity list is one thing to look at, but there are many other ways that can -- that potentially can come from.

So we just want to wait until the situation become clear.

Then, we will address the potential impact to us.

Just add to that, should companies like Hikvision or Dahua get added to the list, the first thing we have to determine, and obviously, we're not waiting, we're looking at that right now, is whether or not the Ambarella solutions, which are developed and manufactured outside of the U.S., are they subject to the restrictions.

And without knowing exactly what's going on, we're digging into that.

But 2 things would have to happen.

One is they would have to be put on this entity list, and then we would have to make a determination whether or not that impacts our products based off of our [back patterns].

So...

Got it.

And then not to make all my questions about the potential of a ban, wanted to shift over to professional security.

Do you guys have an estimate as to the number of AI-based cameras that would ship perhaps in calendar '19 and calendar '20 just to give us kind of a unit TAM, so we can start thinking about what that ramp might look like?

I know you expect material revenue from CV in 2020 in professional cameras, but just wondering if you might be able to give us kind of a unit TAM as you're thinking about the market today.

Right.

So right now only Hikvision has shipped mainly AI-based camera.

In calendar year 2018, they say they shipped roughly 1 million units.

In calendar year '19, their forecast shows that they want to ship roughly 3 million AI-based camera, and they haven't really given any indication for calendar 2020 yet.

And so between Hikvision, Dahua, we believe they will be aggressive on AI camera side.

On top of that, outside China, we only talk about one of our customers that will be ramping up their AI camera, and also we talk about totally we believe 5 of our CVflow customer will start shipping professional security camera by the end of the year.

So I hope this will give enough momentum on the supply point of view.

But from demand point of view, I really think that security depends on the pricing of the customer product, the aggressiveness of them pushing the technology and position the product.

So we're definitely waiting for more visibility from our customer about their sales rollout of those products.

Our next question comes from Matt Ramsay with Cowen.

Fermi, just quickly on the longer term, you guys are continuing to announce new CV chips and also obviously based on the CVflow architecture.

And I just wanted to maybe get a little bit of an understanding with the engagements that you've gotten so far what level of software investment those customers are making on top of the CVflow architecture, whether that's your own investment, the customer's own investment, any NRE payments that you might be getting to do that investment collaboratively and how sort of a software development on top of your platform and architecture is going for CV.

Well, that's a good question because one thing we noticed about all of the AI or CV technology development, software is probably the biggest investment at this point for both customer and us.

In this -- in my script, I mentioned one important thing.

More than 50 customer and partners have port their neural network onto our CV platform using our tools.

And most of those network -- in fact, for each customer, they probably port anywhere from 2 to 10 their own development neural networks onto our platform.

And all -- most of that network will be -- can be ported in a couple of days.

So that just show you efficiency and a number of our customers are looking for this technology.

And in fact, the more importantly, after they port their neural network onto our platform, they convinced themselves that our message today in terms of power consumption performance and video quality are matching our -- how we said in our spreadsheet.

So I think that's very important proportion of it.

And the other one is in the last -- 2 calls before,

I mentioned that HELLA ported their software onto our platform in less than 3 months, including neural network and other applications on CV22 in less than 3 months.

That just show you how it's much easier than our customers' expectation in terms of our software structure and the probability for their software control or hardware platform.

The third thing you asked was NRE.

For most of our automotive engagements, we are definitely asking for NRE to exchange for the support, software support, hardware support and maybe the ongoing system support for these long-term automotive engagements.

And that -- some will definitely show up in our financial guidance and will continue to show up in our financial guidance in the future.

That's encouraging and really helpful color.

I guess as a unrelated follow-up, my -- maybe my naïve understanding of the CV opportunity was really in professional security cameras and then in the longer term in auto, and it was interesting to me today to hear you guys talk about the home security market as sort of the second leg of CV adoption.

Can you talk a little bit about what opportunities you're seeing from a home security point -- standpoint and like how close we are to those transitioning on to CV?

Right.

So if you look at all of the reviews of the current consumer security camera, it doesn't matter who is building with Ambarella solution or other solutions.

The biggest concern is the accuracy of those alarms.

Most people try to sell service by providing alarms, sending e-mails to the customer when they see something in the camera and say, "Oh, there's an alarm.

You just look at it." But most of the alarms are false.

In fact, there is some stat just shows that 95% of the alarms sending to the consumer through the consumer security camera were false.

And so really the alarm -- the false alarm becomes so -- such a problem.

In fact, at the end, most of the consumer, including myself, turn off those alarms.

And the reason for that is the AI using -- streaming a video to the cloud with a very small video resolution didn't give you the accuracy and the response time you need.

So we all believe -- when we talk to our customer, we all believe that pushing the AI or computer vision to the [agent] so that you can apply CV on the high-resolution video without much any delay and making real-time decision is the way to reduce the false alarms.

We believe that with our AI technology, we can help our customers easily eliminate 90% or more false alarms with our technology.

With that, I think the alarm system will become useful and our customer can start really monetize their service from there.

So that's the -- one of the reasons I think that our consumer IP camera will start adopting AI.

But at the same time, we also know that a consumer IP camera can go 2 extremes now.

You see a lot of $20 consumer IP cameras get sold in Amazon.

Sometimes, you see people selling $199 with AI functions.

So you're going to continue to see these two trend.

One trend goes to high-end provide complete AI service and the other end, which is a very low end, nothing, just pure video streaming.

So you're going to see consumer IP camera go to those 2 extremes and definitely our CVflow is targeting to serve the higher performance and the high gross margin product that we are working on.

So in terms of timing of this product line, we start seeing a lot of design win activities.

And last month, we mentioned that we're probably going to see -- start seeing people ramping up calendar year 2020 -- second half 2020 and the material revenue coming in 2021.

Our next question comes from Suji Desilva with Roth Capital.

Just a housekeeping question.

What percent of revenue at this point is the consumer legacy?

Is it de minimis at this point or still material?

Yes.

We have said that we're entering the year in kind of the high teens and that, that would be rolling down over the next, like I say, 2 to 3 years.

And so you can think that, again, there's going to be ups and downs.

It's going to kind of lumpy because there's less and less products in there, but over the course of the year, you can see that decline from that point down.

So you can assume that it didn't make a dramatic change in the current quarter.

Okay.

And my next question may be a little unfair.

I mean there's a lot of uncertainty but also a lot of guidance on the call about computer vision.

But are we at the point now where we can talk about in the kind of calendar '20 time frame what percent of revenue that might be or the run rate or qualitatively?

And if not that, what are the 2 or 3 things we should watch out for that can drive CV as a material part of revenue in '20?

I think that it's too early because our customer haven't gained mass production yet.

I really think that the milestone you need to watch is when our customers start announcing their products, that's really when they start shipping and how many of them, what's the pricing point, and that's something we are watching anyway to give us indication the revenue stream we can expect next year.

And Fermi, when do you expect that wave of announcements to hit?

Holiday season or back to school?

It's going to later this year, early next year.

CES might be a good time to look at it, but I think I expect a lot of announcement at the -- late this year.

Our next question comes from Charlie Anderson with Dougherty & Company.

Congrats on all the CV progress.

I have just 2 quick ones.

So one, you mentioned sort of obviously the change in behavior from the Chinese security camera customers, the pull-in activity.

But wonder are you noticing a change in behavior or rhetoric from the customers who are using HiSilicon who could potentially cut over to you.

That's question one.

And then question two is, if the appetite increases for pull-in activity from the Chinese security customers, can you maybe help us understand how much you can maybe serve that increased appetite.

You did mention there were some constraints on the supply chain, so just want to understand the dynamics there.

Okay.

So the first question is about HiSilicon.

I think that we only start thinking -- for example, we talked about Hanwha Techwin today, which was a HiSilicon customer, and it's transitioned to us not only on the CV product but also on the video product.

So that's just one example.

I think that outside China, we do believe that we are taking over some of the HiSilicon accounts, and we're going to ramp up second half this year.

In fact that -- the other trend we definitely notice but we haven't seen big impact is that Huawei had made it clear they want to become one of the largest security camera vendors in China, which means they're going to compete with Hikvision, Dahua.

How does that impact their relationship and HiSilicon relationship with Hikvision, Dahua is something I try to understand and see how [that further is going to] help us in the long run.

That's something we -- it's not clear.

We haven't seen a clear indication yet.

So the second question is -- sorry, I forgot the second question.

Yes.

It had to do with your capacity to ramp up if the appetite increases from more pull-in activity.

Yes.

Thank you for reminding.

For the -- for example, you can see that Casey talked about we still have [85] (corrected by the company after the call) days inventory that definitely help us to address portion of the pull-in requirements and also that I think we have great suppliers that are willing to help us to do some urgent pull-ins from the wafer and packaging testing point of view.

So so far, we are -- we think we can handle more pull-ins, but hardware, it really depends on whether this a geopolitical situation continues and whether our customer continue to want more pull-ins in the next several months.

So we're watching that very carefully, but we do have some inventory that can handle it and also can prepare more if things continue to deteriorate.

On top of that, I think the things we are watching the most is whether the packaging testing, house capacity and whether the cost will go up if we have too much higher demand, and that's something we're watching.

We haven't seen that yet, but we worry that it might be the case.

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

Fermi, I think I caught your comment regarding your success in auto recorders.

I think you say 18 wins with the 10 OEMs.

Can you give us some perspective of that market so far in terms of maybe your share, the volume you're seeing with those customers?

Are they add-ins?

Or are they default selections, things like that?

Just give us a view of how that's going so far.

Well, we mentioned 10 OEMs and 18 models, all of them are either option or default inside of the OEM cars, and that's definitely one thing we are focused -- one market we are focusing on.

Our estimation is total penetration of that market is probably rough [5% globally] (corrected by company after the call) based on the number we can see.

So we do believe there's a potential market growth -- market share to gain and so market growth that we can do in China, particularly that we continue to hear that government might put a regulation out to ask for some commercial car, even consumer car to have this function as default.

If that happens -- when that happens, I think that will really help the market's growth.

And our advantage on this market, in addition to our traditional video quality and the power consumption, I really think that the quality of silicon that we have auto-grade chip and also not only just temperature but also get to AEC-Q level, and that which is a very good quality history with our current product line, I think that's really one of the benefit we enjoy in China right now.

Okay.

Great.

That's helpful.

And my follow-on question, Fermi, if you want to look a little bit bigger picture longer term, looking at the level 4, level 5 opportunity has been a lot of discussion -- well, there has been for a while but certainly this year about the interesting usage of lidar versus non-lidar approaches for level 4, level 5. Wondering what you're hearing about those or the discussion out there in the industry and then also maybe as that plays into the engagements on an interest in stereo versus monovision.

Right.

So first of all, we continue to believe level 4, level 5 will be a long-term market for us and we -- although we're still continuing to engage the OEMs who are doing level 4, level 5, but our expectation is that is more of a small volume production and also that some of the R&D project that we get involved with NRE and also building relationship and technology together.

In terms of lidar and non-lidar, [we don't take a stand on it] (corrected by the company after the call).

We believe that stereo cameras are a good technology complementary to the other mono camera systems, and we try to prove the concept that without lidar you can have a level 4, level 5 autonomous car.

However, like I say, if our customer coming to us says they want to use our CV chip with lidar, we will definitely work with them because this is a trend that we are not going to compete in market trend.

If market trend says lidar plus mono camera, we'll work with this customer.

Customer comes and says they want stereo plus mono, we will work with them.

We try to stay neutral on this discussion.

However, we definitely try to prove the concept, the following concept that when the stereo processing work properly and with our technology, want to show the people that stereo processing can be as powerful as lidar or even more powerful because the density, the cloud point density that we can provide with stereo cameras.

So this is a job that we will continue to improve the quality of our stereo processing and hopefully, that we can demo in the near future that combination with stereo processing plus our CNN mono camera can be the main sensor used by the level 4, level 5 cars in the future.

Our next question comes from Gus Richard with Northland Capital Markets.

Casey, in terms of the margin pressure in the current quarter, can you give us a sense of how much of that is increased packaging cost versus mix?

This quarter, I think most of it is mix.

Going out into the future, if there continues to be more pull-ins [and if there is supply chain congestion] (added by company after the call), then I think we have to go back and see what our opportunity is to get that done without any extra cost.

And so the extra packaging cost and test cost would really be a Q3 and forward issue.

This quarter, it's really all mix.

Got it.

And then when you think about the opportunity to gain share from HiSilicon versus the loss of opportunity because of tariffs and banning of customers, can you kind of size those 2 relative to each other?

Is there more opportunity for you to gain share or more opportunity for you to lose business?

Or are they about the same?

Well, this is Fermi.

As you can imagine, I ask this question in the company many times, and we have many different models show different results.

So at this point, I really think that it depends on your assumption.

It comes to total different results at this point.

I think the way we approach this problem is when we get more clarity in terms of whether the ban is happening to whether the ban applies to our products, then we will really have more -- we can give you a more precise and definite answer to your questions.

Do you have a range?

That's quite wide.

At this time, I would like to turn the call back over to Dr. Fermi Wang for closing remarks.

Yes.

Thank you all for joining today and looking forward to talk to you next time.

Goodbye for now.

Thanks.

Thank you.

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

This does conclude the program.

You may now disconnect.

Everyone, have a great day.