Ambarella, Inc. (NASDAQ:AMBA) Q1 2024 Earnings Call Transcript

Ambarella, Inc. (NASDAQ:AMBA) Q1 2024 Earnings Call Transcript May 30, 2023

Ambarella, Inc. beats earnings expectations. Reported EPS is $-0.15, expectations were $-0.21.

Operator: Thank you for standing by, and welcome to Ambarella’s First Quarter Fiscal Year 2024 Earnings Conference Call. At this time, all participants are in listen-only mode. After the speakers’ presentation, there will be a question-and-answer. [Operator Instructions] As a reminder, today’s program is being recorded. And now, I’d like to introduce your host for today’s program, Mr. Louis Gerhardy, Vice President, Corporate Development. Please go ahead, sir.

Louis Gerhardy: Thank you, Jonathan. Good afternoon, and thank you for joining our first quarter fiscal year 2024 financial results conference call. On the call with me today is Dr. Fermi Wang, President and CEO; and Brian White, CFO. The primary purpose of today’s call is to provide you with information regarding the results for our first quarter of fiscal year 2024. 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 that we filed on March 31, 2023 for fiscal year 2023 ending January 31, 2023. Access to our first quarter fiscal 2024 results press release, transcripts, historical results, SEC filings and a replay of today’s call can be found on the Investor Relations page of our website. Fermi will first provide a business update for the quarter, Brian will review the financial results and outlook, and then we will all be available for your questions. Fermi?

Fermi Wang: Thank you, Louis, and good afternoon. Thank you for joining our call today. Our Q1 results were slightly ahead of our expectations. Despite the significant headwinds from the ongoing semiconductor industry cyclical downturn, we are not allowing this difficult environment to distract us from further developing our AI business. Before I talk about the details of the quarter, with all the cross currents in the market together with the all the exciting developments in the AI market, I thought this would be a good time to review our strategic vision. Simply put, our transformation into an AI company is well underway, with AI already representing 45% of our total revenue last year and an estimated 60% this year. Now with our CV3 platform, we are expanding into a new phase of AI market development.

The AI market is at a very early stage, it is also dynamic with many technologies and applications emerging. With all the excitement about AI, the key to our continued success will be our focus and the degree to which we can leverage our unique core competencies. Even with our focus, our current serviceable available market, or SAM, is sizable, exceeding $4 billion this year and approaching $10 billion in fiscal year 2028. So, what are we focused on? Ambarella is focused on deep learning AI processors and software, which are replacing the legacy and less powerful traditional machine learning approaches. Within the deep learning market, the AI processor market has been dominated by training processors used in servers typically for the cloud, data center or enterprise.

Our focus is on AI inference, which is where AI models get deployed and are practically utilized by end users. As the AI market begins to mature, most third-party research firms forecast the size of inference AI to surpass training AI. We have already demonstrated how we can leverage our rich heritage in human perception, also known as video processors, into AI. Our CV2 family was our first move into AI, and it targets inference AI perception processing at the edge where cameras are the principal sensing modality. We continue to expect the CV2 family to be approximately 60% of total revenue in fiscal year ’24, and represent a material portion of our operating profit dollars. The CV2 SoC integrates our camera perception expertise with our proprietary second-generation CVflow AI architecture.

The incremental processing to enable AI causes our CV2 blended average selling price, ASP, to be greater than two times a video processor. This contributed to an over 20% increase in our firmwide ASP in fiscal year ’23. This year, the CV2 family is expected to become the dominant driver of our revenue and remain the key driver for several years. The solid stream of operating profit from video processors and the CV2 family of edge AI processors is now being reinvested into the significantly more powerful CV3 platform targeting mobility applications. The CV3 platform builds upon our CV2 family experience and utilizes our proprietary third-generation AI inference processor. For the typical Level 2 Plus application, the CV3 SoC provides the perception processing for all the camera and radar sensors, as well as the processing required in the fusion and planning layers.

The significant amount of incremental processing is expected to facilitate a CV3 SoC ASP to be five to 20 times higher than a CV2 SoC. It is also very important to understand CV3 is a platform, as the SoCs in the CV3 family can capture incremental value by running our own autonomous driving, AD, software stack IP and/or radar perception software IP. We aim to bundle this software IP with our CV3 SoCs in a platform approach, providing our customers with the flexibility to pick and choose exactly what they need. Regarding our Autonomous Mobility partnership with Continental, I am pleased to share that we extended our partnership to Level 4 system development and confirm the first business award of our jointly developed stack as a complete Level 4 Fallback System.

The system will be supplied to Continental for a customer in the commercial vehicle industry. To be clear, the CV3 platform is a major leap forward in term of our value proposition and it brings a new list of targeted customers; automotive OEMs. We are still in the early stages of building out the CV3 SoC portfolio and developing the market. However, we are not doing this alone, with leading Tier 1s like Bosch and Continental porting their software to CV3, validating our superior efficiency, jointly marketing to auto OEMs, using their scale and bringing more credibility to our CV3 market development efforts. Additionally, for the AI server inference market, we have already evaluated running large language models, LLM, on CV3-AD High, which has been sampled for nine months, and we believe the LLM performance on this existing SoC to be as good as the Nvidia A100 with much lower power consumption and a superior total system cost.

We are now establishing a software development effort as well as a business development program to engage with customers. Turning to new products and customer engagement in the quarter. In March, at the ISC West security show, we announced our CV72S for mainstream enterprise and public class security cameras. CV72 utilizes the same third-generation CVflow deep learning AI accelerator architecture utilized in the CV3 SoCs. This CV3 derivative SoC bring to the IoT market the highest AI performance per watt, the fusion of radar and camera data and it includes support for the latest transformer neural networks. Furthermore, CV72S offers six times the AI performance of CV2 family, enabling it to run Ambarella’s groundbreaking neural network-based image signal processing software for 4K color, night vision and HDR with plenty of headroom for additional concurrent neural networks.

CV72S is now sampling to leading IoT camera companies. In IoT, there were a number of new enterprise and public security cameras introduced, including: Motorola, who introduced the H6SL camera line based on CV25, as well as the V700 body camera based on our S6LM SoC; and Verkada introduced its TD52 video intercom featuring a five mega-pixel camera based on our CV25; iPro, formerly Panasonic and Japan’s largest security camera supplier, introduced multiple new product families based on our CV2, CV22 and CV25, including dual and quad multi-imager models; and European market leader Axis, part of Canon, introduced its 3905 rugged dome models designed for surveillance on board vehicles, such as buses based on our S6LM; also in Europe, Dallmeier introduced Domera E series cameras which use our CV22 AI SoCs to enable imaging in total darkness utilizing adaptive IR illumination; in the home monitoring market, Alarm.com introduced its ADC-780 battery powered doorbell based on our S5LM.

I’ll now talk about progress in the automotive market. As mentioned earlier, our new CV72S SoC is an important CV3 derivative for the IoT market. However, it is expected to also be an important derivative product for the automotive market, and in April at the Shanghai Auto Show, we announced and demonstrated CV72AQ. This SoC targets multiple automotive applications including Level 2 Plus and other applications with up to six cameras and five radars running on the same SoC. CV72AQ demonstrations at the show included an ADAS plus parking system with a five camera configuration including an eight megapixel front camera and multiple three megapixel fish eye cameras running YOLO v7 neural networks on each camera. We also demonstrated, versus the leading GPU solution, superior performance and lower power consumption of CV72AQ running transformer networks.

We received very positive feedback on CV72AQ from Tier 1s and OEMs in China. Also at the Shanghai Auto Show, a number of other Tier 1s demonstrated CV3-based systems. This included Continental which showed a 10-camera live demo with multiple neural networks running on each video stream. And Hyperview demonstrated its GT-HyperMax platform featuring a sensor suite of 11 cameras plus one lidar and three radar in a car, providing City Navigate on Pilot advanced functions and leveraging the latest transformer networks. In March, China’s GAC introduced its electric AION Y Younger L2 Plus ADAS SUV with an intelligent 1V1R driving assistance system based on our CV22AQ AI SoC. And in April, Geely Zeekr introduced its Zeekr X electric SUV with a face recognition access control system based on CV28AX AI SoC.

In summary, a majority of our new customer engagement activity continues to be for our AI products. AI is expected to be a majority of our revenue, for the first time, in F2024, and AI should continue to grow as a proportion of our mix. To bring our AI strategic vision together, first with CV2 and now again with the even more significant CV3 platform, we have leveraged our core competencies, cumulative knowledge and unique approach to establish a strong presence in the AI deep learning domain. Our investments yield differentiated products that are very different — very efficient, and on open platforms that are scalable and flexible. The CV2 family is already very profitable and we are well into the development phase with the CV3 platform. In summary, there is still a lot of work left to execute to our strategy and the ongoing semiconductor industry cyclical downturn pressures our near-term financials.

However, we are confident in the long-term secular growth opportunity for edge inference AI, we do not intend to stray from our strategic vision, and we are continuing to invest in our differentiated AI strategy. I will now turn it over to Brian to discuss the Q1 results and the Q2 outlook in more detail.

Brian White: Thanks, Fermi. I’ll review the financial highlights for the first quarter fiscal year 2024. I’ll also provide a financial outlook for our second quarter ending July 31, 2023. I will be discussing 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 and acquisition related costs, adjusted for the impact of taxes. For fiscal Q1, revenue was $62.1 million, in line with the mid-point of our prior guidance range, down 25% from the prior quarter and down 31% year-over-year. As expected, total Automotive revenue was approximately flat sequentially, while IoT revenue was down sharply driven by customer inventory reduction actions.

Non-GAAP gross margin for fiscal Q1 was 63.1%, in line with the mid-point of our prior guidance range of 62% to 64%. Non-GAAP operating expense for the first quarter was $46.2 million, up $200,000 from the prior quarter and below our prior guidance range of $47 million to $49 million. The lower operating expense was driven by continued expense management and the timing of spending between quarters. We remain on track to our internal product development milestones. Q1 net interest and other income was $1.3 million. This was higher than our original forecast driven by a higher cash balance and returns on cash invested. Our non-GAAP tax provision was $300,000, or minus 5.5% of pre-tax income. This was slightly lower than our original forecast, driven by the mix of pre-tax income across tax jurisdictions.

We reported a non-GAAP net loss of $6 million or a $0.15 loss per diluted share. Now, I’ll turn to our balance sheet and cash flow. Fiscal Q1 cash and marketable securities increased $20.5 million to $227.4 million. DSO improved significantly from 57 days to 43 days as the timing of shipments throughout the quarter normalized, after being back-end loaded in the prior quarter. Ending inventory increased slightly, up 1.8%. However, days of inventory increased more significantly, from 116 to 151, due to the sequential reduction in cost of goods sold on lower revenue. Cash from operations was strong at $22 million, driven by the decrease in accounts receivable, and capital expenditures for tangible and intangible assets were $2.3 million. Free cash flow, defined as cash from operations less CapEx, was 31.7% of revenue for the quarter and 6.4% on a trailing 12-month basis.

We had two logistics and ODM companies represent 10% or more of our revenue in Q1. WT Microelectronics, a fulfillment partner in Taiwan that ships to multiple customers in Asia, came in at 49% of revenue. Chicony, an ODM who manufactures for multiple IoT customers, was 16% of revenue. I will now discuss the outlook for the second quarter of fiscal year 2024: Customer feedback on end-demand remains generally healthy. However, at the same time, customers also continue to aggressively manage down their inventory levels. Considering these factors, we estimate that our fiscal Q2 revenue will be flat to Q1 and in the same range of $60 million to $64 million that we guided for the prior quarter. By end market, we expect that both automotive and IoT revenue will be approximately flat sequentially as well.

We expect non-GAAP gross margin to be in the range of 62.5% to 64.5%, up slightly from Q1. We expect non-GAAP OpEx in the second quarter to be in the range of $48 million to $50 million, with the increase compared to Q1 driven by higher R&D tied to new product development activities. We estimate net interest income to be approximately $1 million, our non-GAAP tax expense to be approximately $700,000, and our diluted share count to be approximately 39.7 million shares. Ambarella will be participating in TD Cowen’s Technology, Media and Telecom Conference on May 31st and June 1st, Bank of America’s Global Technology Conference on June 6th and Rosenblatt’s Age of AI conference on June 7th. Please contact us for more details. Thank you for joining our call today.

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

Q&A Session

Operator: Certainly. [Operator Instructions] And our first question comes from the line of Gary Mobley from Wells Fargo. Your question, please.

Gary Mobley: Hey, guys. Thanks for taking my question. I wanted to ask about inventory drawdown with customers. You mentioned that customer and demand appears to be healthy, but obviously, your [undershipping in] (ph) demand. Could you give us a sense of by how much and — by how much or how close we are to inventories being back down to a normal level? And then, maybe you can comment as well specific to China-related demand?

Fermi Wang: Yes, this is Fermi. I think like Brian said, we haven’t seen a huge change from the customer side. For example, last quarter, we talked about customer that has healthy growth based on our silicon, but our silicon revenue from them is strong 15%, 20% year-over-year. And that situation continues. And I think the customer continue to confirm that their growth and we continue to forecast lower revenue this year. So, I think from that point of view, I think the situation is very similar to last quarter and we have not seen anything — any indication that this inventory correction will end. So, I think what we are looking for is really that the ramping up of new orders consistently from different customers, that will probably give us indication that it’s recovered. We haven’t seen that yet.

Gary Mobley: And as a follow-up, I wanted to ask about your win that you captured in conjunction with Continental. Is that automotive-grade win? And maybe if you can give us a sense of where automotive — other automotive-grade wins may stand?

Fermi Wang: Right. So that’s automotive-grade win. And like I said, it’s a Level 4 car and it’s going to be auto-grade chip. And it’s going to be — its first design win that we work with — after the announcement, working with Continental. And in this design, we involve not only our CV3 SoC, but also our software IP, the software stack that we — Continental and us are co-developing. So, I think it’s combination of software and SoC win. If we do — we are working on other design wins. For example, we talk about, first of all, we’ll continue to work with Tier 1s like Conti and Bosch on any potential design wins. At the same time, we also mentioned that in China, I think we start seeing that a lot of opportunity in Level 2 Plus cars, especially at the Shanghai Auto Show, we introduced CV72AQ.

It make us believe that we have plenty of opportunity there and we are optimistic that we’re going to close on design win this year there. And also, the time to revenue is much faster with those potential design win in China.

Gary Mobley: Thanks, Fermi.

Operator: Thank you. One moment for our next question. And our next question comes from the line of Ross Seymore from Deutsche Bank. Your question, please.

Ross Seymore: Hi, guys. Thanks for letting me ask the question. Fermi, you mentioned that you haven’t seen any signs of end of the inventory digestion that’s going on and that growth would really resume when some of the new products kick in. Can you, I guess, dive a little bit deeper into that? And so, the two-part question would be, where do you believe the revenue level would be for your company versus the $60 million to $64 million if you are shipping to true-end demand? And to the extent it’s dependent upon new products kicking in, when do you believe that occurs?

Fermi Wang: Right. So, first of all, I think that we are now waiting for new product to kick in. We believe the current inventory correction, when they finish, the existing product line will come back to live and it will go back to original level. So, we are now counting on new product design win to fix this inventory correction problem. And for your first question in terms of level, last quarter, when we look at just one example of a customer, we think that we’re probably like a 25%, 30% below the realistic level. So, I think we still hope believe that’s a level the differences we’re looking at. And hopefully, when the inventory correction finished and all the customers went back to normal, I think that should give you indication where we think the normal level of revenue is.

Ross Seymore: Got it. And I guess for my follow-up on the automotive side specifically, it’s good to see the design win with Conti turned into products, et cetera. That business has been basically flat sequentially, I think, for four quarters now; three quarters you’ve reported and it looks like you’re guiding it relatively flat. When is the timing where we should start to see that business picking up? You guys have talked about this investment. I know it’s a longer-term strategy for the company, but it seems like one that should yield some pretty strong tailwinds off the size of company you’re currently running at. So, just wondered on the timing that we should look for and what the drivers of that growth should be.

Fermi Wang: Well, I definitely think that’s flat. If you look at few quarters before the inventory correction and now you’re comparing to the inventory corrections period. So, I think the last two quarters definitely been impacted by inventory correction in the automotive sector. So, I also believe that as soon as inventory correction finish, auto should show some revenue growth from that point of view. But like you said, the really big auto growth should come with the ADAS market and also at Level 2 Plus market when they go into production.

Ross Seymore: Great. Thank you.

Operator: Thank you. One moment for our next question. And our next question comes from the line of Tristan Guerra from Baird. Your question, please.

Tristan Guerra: Hi, good afternoon. Just wanted to have a follow-up on the inventory correction and also try to tie this with market share shift. So, it’s no secret that some Chinese companies have tried to diversify away from U.S. supply, either because there is push from that — for that — from the Chinese government or because they’re concerned about potential future sanctions. And I know you very much derisked your surveillance camera exposure to China in the past, but you still have exposure in automotive. So, I wanted to know if there is any signs that perhaps the ramp in China is not expected at the pace that you thought will happen a year ago? Are you getting any feedback? And also just to the extent that the inventory correction that you’re describing seems to be a bit more pronounced than some of the other companies where it’s been really more smartphone and PC centric for other companies. So, any elaboration around that would be great.

Fermi Wang: Right. So, in terms of the geopolitical situation, I think for automotive market, it’s much less severe than security, right? Security is really being viewed as the safety of the country so that — that’s why I think people trying to avoid U.S. components. But in automotive, in fact, if you look at the middle and high end auto components in Chinese market today, all of them are U.S. components. So, I think that’s because in automotive processing, I think that our solution among other U.S. components still have better performance efficiency and we haven’t seen a similar impact on Chinese government on mandating the Chinese automotive OEM use exclusively the Chinese component. So, I think that the two things add together, I think, I still believe that we won’t see a severe downturn on the Chinese automotive business.

Tristan Guerra: Okay, great. And then, as my follow-up question. Obviously, you’ve made that software acquisition, you have the sensor fusion chip. So, do you think you have all the pieces you need to move into L2 Plus and L3 application? Are you getting any feedback about customers looking at your company size versus, say, larger supplier, or is it purely based on chip performance where obviously you excel? Is there any other consideration that — and how you — in terms of getting design wins and how you’ll address that? And that question will also tie to the product roadmap and whether customers are kind of wondering where will you be five years out in terms of product roadmap?

Fermi Wang: Right. So, I think, from a product roadmap point of view, I think for Level 2 Plus, Level 3 car, I think we have all the contents that we need to go after this market. Because from the hardware and software point of view, I think we can offer a complete solution. But from the strategy point of view, as we said before, we are not bundling hardware and software together. We are trying to offer a software platform that customer can pick and choose and we can help our customer to build their own software stack and working with Conti is probably the best example. And in terms of scale, it’s always a problem, right? Trying to compete with a bigger company is always a disadvantage for us. But I think that’s the reason we continue to try to work with a bigger Tier 1s. And using — with their scale and expertise, that will help us partially further to address this problem.

Tristan Guerra: Great. And clearly, the Conti and Bosch design win speak to that effect. Thank you very much. Very useful.

Fermi Wang: Thank you.

Operator: Thank you. One moment for our next question. And our next question comes from the line of Tore Svanberg from Stifel. Your question, please.

Tore Svanberg: Yes, thank you. My first question, Fermi, could you just talk a little bit about the main difference between the CV3-AD and the CV3-AQ? Whether it’s functionality or ASPs or any other color you can share with us?

Fermi Wang: Right. So, CV3-AD and CV72AQ, first of all, they are all based on the same architecture, CV3 architecture. All of them using the third-generation AI inference — AI processor. The difference — main difference in the CV3-AD is designed for the auto-grade chip level [ASIL] (ph), and CV72AQ is designed for the system-level auto grade. So, I think that’s a main difference. So, I think the — for example, CV72AQ platform is target for Chinese market where people willing to accept system-level ASIL system versus chip-level ASIL system. That’s the main difference.

Tore Svanberg: Very good. Thanks for clarifying that. And my follow-up question, you announced the design win for the software IP modules. Again, I was just hoping you could elaborate a little bit more on that. And it’s surprising to me when I hear software IP module, right, because I think hardware and software. So, it’s like how exactly is the accounting for this particular design?

Fermi Wang: So, interesting — I think I see as we announced this is software partner with Conti, so basically the idea is that we are contributing a portion of software solution and we work with the Conti software team to integrate those modules into a complete software stack, leveraging the strength of both sides. For example, Ambarella’s more — strength is on the perception side, particularly today is video perception and radar perception, but Conti definitely has a lot more system-level solution and famous for their auto-grade system software. So, I think that’s where we see that we can leverage both sides of our strength and build a software stack based on leveraging the both side of strength. So I think that’s an approach that we’re different than the customer. And also for OEMs that if there are anybody who want to do similar business model, we are open to that too.

Tore Svanberg: And on the sort of revenue accounting for — I mean is this a module sale or IP revenue?

Fermi Wang: Yes, I see. That’s basically a software revenue split. We need to decide how to share the software revenue together.

Tore Svanberg: Understood. All right. Thank you very much.

Operator: Thank you. One moment for our next question. And our next question comes from the line of Kevin Cassidy from Rosenblatt Securities. Your question, please.

Kevin Cassidy: Yes. Thanks for taking my question. Maybe a similar question to what Tristan had about automotive, but in the AI server, as you move to the adjacent market, AI server inference, and you said you have about the same performance as NVIDIA A100, but much less power. Can you say like you have all the tools you need to move into the server market or into the cloud inference market?

Fermi Wang: Right. So, obviously, the similarity between our current automotive market and the new AI server market is, they are really running a new network on our chip. So, from tools point of view, that we always develop many, many software to help our customer to port a neural network onto our chip. But obviously, LLM is different beast, because they are much larger than the typical neural network that we are working with. So, definitely there is optimization cycle needs to work on. But the reason we decide and we think we have a great opportunity here is, first of all, we have a working silicon in demo; two, we have a bunch of expertise and the software tools available that we build for other market; three, we just need to fine tune and optimize the current software for this LLM to achieve the best possible performance that we can get.

So, I think from that point of view, the effort for us is well — is limited and also I think that we also believe that in the market, very few people can claim what I just said that we have a working silicon, can show real performance and real low power consumption and also demo to the customer. So, I think that’s our advantage. And also believe that the extra resource we need to put on is something that we can handle.

Kevin Cassidy: Great. And what would be the go-to-market strategy? Are you looking for a few maybe flagship customers to lead the way, or are you going broad with lots of different customers?

Fermi Wang: No, I think we have to be focused. I think we need to identify the sweet spot. I think, we should talk about strategy later because we are in the process of talking to customers, but I think we need to focus on where our strength and also focusing on companies that has — can leverage our chip and our software immediately. And so — I think that one thing we learned is to work with a customer who has really — they feel the most painful experience with current solution who are most likely to work with us and that’s where we’re going to focus on.

Kevin Cassidy: Okay, great. Very interesting.

Operator: Thank you. One moment for our next question. And our next question comes from the line of Quinn Bolton from Needham & Company. Your question, please.

Quinn Bolton: Hey, guys. Thanks for taking my question. I guess, Fermi and Brian, maybe just your best guess. I mean, we’ve been working down this inventory now for a few quarters. Certainly doesn’t sound like it’s — you haven’t seen any green shoots yet in terms of the orders. What’s your best guess as to how many more quarters you think it will take to work down this inventory? Do you think we’ll be pretty clear by the end of your fiscal year or the end of the calendar year? Do you think it could take longer?

Brian White: Yes, Quinn, this is Brian. A quarter ago, we said that our guidance for fiscal Q1, which was $62 million at the midpoint. We thought that, that would represent the bottom as it related to impacts associated with inventory adjustments and then it would likely not get worse from that point. And we’re kind of sitting in the same place we were 90 days ago from the standpoint that we still believe that’s the case. What becomes challenging is forecasting when this thing lifts off again and at what slope. We certainly have visibility to backlog. But that backlog has been shifting, right? We’ve had re-schedules, cancellations that we’ve had to deal with. So, while in normal times, we can look at that backlog and have a lot of confidence as to how revenue might shape up, say, in fiscal Q3.

Because of the movements that we’ve seen, our confidence in providing a forecast would be lower in this cycle than kind of a normal time. So, we don’t see it getting worse, but the visibility to the second half and just how that recovery plays out, I think it’s hard for us to talk to at this point.

Quinn Bolton: I understand you’re not guiding to the fiscal second half. Historically, you’ve seen some stronger seasonal trends in the second half. I mean, can you just provide any framework how we might be thinking about it? If the inventory doesn’t get any worse and you see normal seasonality, that would imply a lift? Obviously, if you start to see the inventory correction and that would imply a lift? I mean, are you sort of suggesting, “Hey, keep it in the $60 million to $64 million range until you see inventory clear,” or do you think you can see some seasonal upticks in the second half?

Brian White: Well, normally, we would see some uptick in the fiscal third quarter in particular. And we would hope that we do again. But we’re just at a point where we don’t have the visibility and the confidence to put a number out there and try to give you some magnitude of directional increase at this point in time.

Quinn Bolton: Got it. Understood. Thanks, Brian. And then, I guess, for Fermi, I guess, I was a little surprised to see your first win with Continental being a Level 4 win. I think the initial partnership you’d announced back late last year, I think was for Level 2 Plus. And so, can you just sort of maybe talk about how the Level 4 win came together? I know you had expanded the relationship with CES and so maybe that was on the fast track. But when — were you surprised that the Level 4 came before Level 2 Plus with Continental?

Fermi Wang: Well, I think that the engagement definitely after we announced this software collaboration between Conti and us, which happened at CES, and things goes really fast after that. So, I think definitely, I’m also surprised how fast this developed. And also — and I need to thank Conti that putting this whole thing together because the one important thing is that to sell that joint software stack and get a confidence with customer is important for us, and I think that’s a great win for us. But at the same time, I want to say again on a Level 2 Plus, I think that with — we look at two things. One is the momentum with the Conti and Bosch still — that we’re still working on and still there. But more importantly, I rethink now with Shanghai Auto Show and with our CV72AQ announcement, and the sampling the component and the software to a customer recently, that gives a confidence that we’re going to see CV72AQ Level 2 Plus design win this year and maybe quick revenue returns.

And in China, you know that the design cycle is not four years; usually, it’s less than two years. So, we are hopeful to see a really quick revenue return from the CV72AQ. And also that we have a roadmap continue to address this market. So, I think — overall I think Level 2 Plus design win is — will continue to be our focus and we think we’ll continue to deliver what we think that we can do.

Quinn Bolton: Perfect. Thank you, Fermi.

Operator: Thank you. One moment for our next question. And our next question comes from the line of Brian Ruttenbur from Imperial Capital. Your question, please.

Brian Ruttenbur: Yes, thank you. Can you give me, first of all, housekeeping D&A and material [depreciation and] (ph) amortization?

Brian White: Depreciation and amortization, is that the question?

Brian Ruttenbur: Yes, that’s affirmative.

Brian White: Yes. For fiscal Q1, I believe it was $5.8 million.

Brian Ruttenbur: Okay. $5.8 million for the first quarter. Also, in terms of DSOs, do you anticipate DSOs stabilizing here? So in other words, trying to understand your cash situation probably will just go down, you probably won’t have stair step event again, or do you anticipate DSOs continuing to go down?

Brian White: No. I mean, what we saw in fiscal Q1 was a normalization of the timing of shipments throughout the quarter versus, say, fiscal Q4, where shipments were very back-end loaded. So, we had a couple of quarters; fiscal Q3, Q4 were both very back-end loaded quarters. Q1, normalized, DSOs came down. That provided about a $22 million benefit to cash flow in the quarter. As we move forward, we would expect DSOs to remain at similar levels. Thus, we would not expect to get a large benefit in a future quarter from another stair step down, for example. So, as we move into Q2, obviously, at the revenue level and the other metrics that we gave you, that would be a forecast for a non-GAAP loss. And so, you’ll have lower free cash flow in fiscal Q2 versus Q1.

And as we go through the year, cash flow is just going to be highly dependent upon the revenue levels. We’ve talked about the fact that we just — we don’t have great visibility at this point in time to the second half revenue.

Brian Ruttenbur: Great. Thank you very much.

Operator: Thank you. One moment for our next question. And our next question comes from the line of Suji DeSilva from ROTH Capital. Your question, please.

Suji DeSilva: Hi, Fermi. Hi, Brian. Good to see the Continental win. I don’t know if I missed this, but did you say the L4 commercial vehicle customer, what geography that was?

Fermi Wang: We didn’t. And we — our customer doesn’t want us to disclose that yet.

Suji DeSilva: Okay. Fair enough. And then, I think when we talked about a range of ASPs, five to 20 times, can you just talk about what drives the delta there? Is that more compute hodge — horsepower in the chip, or is that compute plus software? Any color there would be helpful.

Fermi Wang: I think that comment is purely for silicon, not — does not include software. So, the difference is really from the low end to the high end. So, for example, the lower end — for the high-end chip, we talk about $400 plus. And the lower end, for example, our 655 chips that we’re talking about are probably in the $100 range. So it’s really that level of the different performance of price range that we’re talking about. I also believe that for automotive roadmap, you need to have a family of chip to address different performance level for Level 4, Level 3, Level 2 Plus. Even multiple layers of Level 2 Plus require different ships. So, I think that’s what we’re actually talking about.

Suji DeSilva: Okay. Thanks, Fermi.

Operator: Thank you. One moment for our next question. And our next question comes from the line of David O’Connor from BNP Paribas. Your question, please.

David O’Connor: Great. Good afternoon. Thanks for taking my question guys. Maybe Fermi, just going back to the question on the AI inference opportunity. Can you just give us a bit more detail there on what type of customers are potentially kind of you could engage with on the AI inference side? Is that [data price] (ph) or enterprise, or any particular vertical that you think may be open to you? And I know it’s early days, but as you mentioned, you have a working chip and you need to rework the software, but what kind of timeframe do you think you could potentially get to revenue there? Is that kind of three to five years now? Is it kind of before that? And anything around kind of content opportunity there would be helpful. Thank you.

Fermi Wang: Yes. So, it’s a lot of question. So, I think like I said, the target market is really where is — I think our first target is really on the edge server side, which — where you can focus on the enterprise and the people who are running their own neural network. For people running open AI or other very large model, that might not be the best customer at this point for us. But there are plenty of other different software [SVs] (ph) and that use — driving different neural net model either for the own code or they are running the enterprise level. Those are the probably sweet spot for our chip because we know that, just like I said, the scale is — definitely matters in this market too, and we need to pick the best market we go after.

And we are still in the process to figure out, but I definitely think that’s, what I just said, it’s along our current thinking. But I think in terms of the content, I think is obviously even better than our automotive ASPs, because the competition out there is starting at a much higher level. And also, we have great advantage on both. On the power consumption side, we are not talking about 5%, 10%, we’re talking about significant difference in terms of power consumption, therefore — as well as the total system cost. So, I think from that point of view, it will help us to get a healthy content there. I think there is another question, I forgot.

Brian White: Time to revenue.

Fermi Wang: Time to revenue. So, we — I think we need to get [indiscernible] running, then our software too running, so that a customer can port, and then we can talk about design win, then we talk about revenue. So, I think if you ask me today, I would say that’s a 24-month process totally. That’s a wild guess, yes.

David O’Connor: It’s quite helpful to frame that. Thank you, Fermi. And maybe just a follow-up on that for Brian. Just on the ramping that software development team, just to clarify, that fits in with the current OpEx envelope? Are we going to see some kind of step up there to fund that new team? Thank you.

Fermi Wang: Yes, our plan is to use our current expertise and team to facilitate this activity. The idea is simple because we have — our internal team is — has helping many other customers to port their neural network onto CV3. And we understand [LLM] (ph), because it’s so large and it takes an extra effort, so that — the best way to do this is fund it and put internal resource on this project. Obviously, it will take a trade-off, right? We don’t plan to add much of the resource into the company, so I think that we need to really focus on the area where we think it’s important. I think it’s very important for us is definitely security camera to provide cash for us, maintaining our CV3 momentum with the current design wins like Bosch and Conti, and try to secure a CV3 design win with OEMs and also try to find a resource to fund this LLM, and everything else is a trade-off that we need to consider.

David O’Connor: Very helpful. Thank you.

Operator: Thank you. One moment for our next question. And our next question comes from the line of Vivek Arya from Bank of America. Your question, please.

Unidentified Analyst: Hi, this is [indiscernible] on for Vivek. Thanks for taking my question. Just first, on the cash side. Just curious, I know you mentioned talking about increased R&D investments over the next few quarters. Just with — at your current cash flow, are you comfortable there? Or do you see any need to raise incremental cash in the future?

Brian White: No, we don’t see any need to raise incremental cash. We have a strong cash balance, no debt, and we’ve got a history of being positive from a free cash flow perspective. So, no need to raise additional cash.

Unidentified Analyst: Great. And then quickly as a follow up. Just given the current revenue levels, we’ve seen a relatively substantial [indiscernible] kind of split between IoT and auto is now roughly 65%, 35%. Gross margins are still kind of held up relatively okay and above this the long-term range of 59% to 62%. So, just kind of curious if you can give us the puts and takes on to the gross margin side and maybe beyond Q2 [indiscernible] level?

Brian White: Yes. I think, we’d stick with that long-term model that we provided previously. In the recent history, we’ve been delivering higher gross margins than that, and we would expect that that would continue until we get into the impacts of potentially very large automotive opportunities, and that’s why that long-term model provides for a slightly lower gross margin if we need to get there to secure those design wins. But for now, in the foreseeable future, we should be at recent gross margin levels and probably a little bit higher once we get through this inventory correction and get back to slightly higher gross margins that we were posting last fiscal year.

Unidentified Analyst: Thank you.

Operator: Thank you. One moment for our next question. And our next question comes from the line of Martin Yang from Oppenheimer. Your question, please.

Martin Yang: Hi. Thank you for taking the question. [indiscernible].

Fermi Wang: I’m sorry, I cannot hear you well.

Martin Yang: Can you hear me better now?

Fermi Wang: Yes.

Martin Yang: [indiscernible]

Louis Gerhardy: Hey, Jonathan, let’s go to the next question, and then we’ll give Martin another chance after this question.

Operator: Okay. Understood. One moment for our next question. And our next question comes from the line of Matt Ramsay from TD Cowen. Your question, please.

Josh Buchalter: Hi. This is Josh Buchalter on behalf of Matt. Thanks for squeezing me in. It is great to see that the CV3 win with Continental for commercial applications. I was wondering if you could provide some initial feedback on how it’s going on the consumer passenger vehicle side? I recognize it’s only been a few months since the partnerships have been announced, but you’re going against some entrenched and large competitors in the central ADAS domain. And I was wondering how Conti and Bosch are positioning your CV3-based solutions to win in that market. Thank you.

Fermi Wang: Yes, I think first of all, we have a strong relationship with both Bosch and Conti, and the Conti has even further collaboration because of software relationship, and you can see that software relationship already starting paying off, not only on the Level 4, but also we start getting our engineers in collaboration on that. So, I think that activity definitely is very helpful. And also on the business development side, both sides are working with the Conti and Bosch definitely help us to address the scale problem partially. And also, we believe that working with Conti — we continue to believe that working with Conti and Bosch is the right thing for us to do to get design wins in the U.S. and Europe. I think that said, that I also believe that our first Level 2 Plus design will come from China, like I said, because the momentum we see after the Shanghai Auto Show is real and that we not only demoed a powerful chip, but also we demoed something that very few people can do, which is running transformer neural network in a lower-end chip that just nobody out there can demo.

Like, the company you mentioned, none of them, the low-end chip can demo transformer efficiently. And the transformer becomes such an important neural network and being used as benchmark everywhere, so particularly in China where AI and neural network performance is very much appreciated. So I think that’s a momentum where we definitely enjoyed. And I think — I hope we can get several design wins in China this year.

Josh Buchalter: Appreciate the color. That’s actually a nice segue to my follow-up. Can you talk about — I think this is the first time you’ve mentioned auto being a source of inventory correction despite the results coming in in line with your expectations. Did that get any appreciably worse during the quarter? I know we’ve all heard about weakness in the China EV market. It would be helpful if you could help us understand your exposure there? And if that worsened and drove, I guess, some incremental weakness during the quarter? Thank you.

Fermi Wang: No. What we said before is we think that our inventory control in auto is much less than the inventory control in IoT. That’s what we said. We didn’t say auto was not impacted. We definitely see a few — several customers got impacted, but not as bad as IoT space. So, from that point of view, I think we still think that — I still expect that when the inventory correction finished, auto should go back to growth.

Josh Buchalter: Got it. Thank you.

Operator: [Operator Instructions] Mr. Yang, your line is now open.

Martin Yang: Yes. Sorry about my technical issue. Can you hear me now?

Fermi Wang: Yes.

Martin Yang: I have a question on CV72. So, do you expect your automotive and IoT customers to — most of them to opt and adopt the Oculii integration that comes with CV72?

Fermi Wang: That’s a good question. So, I think, for the IoT, the adoption of radar system will be slower because the core market, the IoT market is moving towards radar, but not as fast as auto. Radar in auto space is basically everywhere. Everybody realizes that they need to have a radar solution in any Level 2 Plus systems. So, I think in terms of adoption that the radar will go to auto space first. For CV72AQ, I think that the radar integration will come later because right now, we’re focusing on winning the video side. But however, at the second phase of software development, auto — radar integration will come also.

Martin Yang: Got it. Thank you, Fermi.

Operator: Thank you. This does conclude the question-and-answer session of today’s program. I’d like to hand the program back to Dr. Fermi Wang for any further remarks.

Fermi Wang: Thank you very much for you to joining us today. Looking forward to talk to you next quarter. Thank you.

Operator: Thank you, ladies and gentlemen, for your participation in today’s conference. This does conclude the program. You may now disconnect. Good day.