Ambarella Inc

NASDAQ:AMBA

Good day, and thank you for standing by. Welcome to Ambarella's Fourth Quarter and Fiscal Year 2024 Earnings Conference Call. [Operator Instructions] Please note that today's conference is being recorded. I will now hand the conference over to your speaker host, Louis Gerhardy, VP Corporate Development and Investor Relations. Please go ahead.

Thank you, Olivia, and good afternoon, and thank you for joining our fourth quarter and full year fiscal 2024 Financial Results Conference Call. On the call with me today is Dr. Fermi Wang, President and CEO; and John Young, CFO. The primary purpose of today's call is to provide you with information regarding the results for our fourth quarter and full year fiscal 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 based on currently available information and subject to risks, uncertainties and assumptions. Should any of these risks or uncertainties materialize or should our assumptions prove to be incorrect, our actual results could differ materially from these forward-looking statements. We're under no obligation to update these statements. These risks, uncertainties and assumptions as well as other information on potential risk factors that could affect our financial results are more fully described in the documents we file with the SEC. Access to our fourth quarter and full year fiscal 2024 result 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. The content of today's call as well as the materials posted on our website are Ambarella's property and cannot be reproduced or transcribed without our prior written consent. Fermi will now provide a business update for the quarter. John will review the financial results and outlook, and then we'll be all available for your questions. Bernie?

Thank you, Louis, and good afternoon. Thank you for joining our call today. In the fourth quarter of fiscal our revenue increased about 2% sequentially, and we slightly exceeded the midpoint of our guidance range. Thanks to the early actions we took to help our customers navigate their excess inventory, our business continued to stabilize and is beginning to recover. For the full fiscal year '20, our revenue declined 32.9% year-over-year as our customers digested inventory resulting from the industry-wide semiconductor cyclical downturn. Looking ahead to fiscal year '25, we continue to expect both our automotive and IoT business to grow as the cyclical challenges went and the secular growth of our AI strategy emerges. Our customers currently have a cumulative installed base of more than 20 million AI influence SoCs, all from our 10-nanometer CV2 family and the 5-nanometer CV5. This is based on approximately 280 customer products that have reached production on a cumulative basis. The CV2 family is expected to continue to be the key driver of our revenue growth in fiscal year '25. Our AI influence business all-in aged applications represented approximately 60% of our total fiscal '24 revenue and was the key factor in the meetings presented year-over-year increase in our blended ASP. The trend to a richer mix of AI revenue and higher averaging selling price is expected to continue, in particular, the CV3 SoC family enters production. At this time, virtually all of our customers' new design activity involves our AI influence processors. In fact, this was the first year where all of our SOC demos, more than 30 were based on our AI influence products. Fiscal 2024 was certainly challenging for most of the industry. However, there were key industry developments and the company's specific achievements that we believe leaves us very well positioned for growth as the market recovery plays out. for the industry. In the past, the AI process opportunity has primarily been represented by training GPUs in server located in data centers, and this is a market that we do not serve. However, in the last year, the important role and opportunity for inference processors in particular at age has become better understood, and this is exactly where we have been focused on. Internally, we achieved 4 key milestones during the last year. First, we have now shipped more than 500,000 units of our first 5-nanometer SoC and we expect our shipments in fiscal year 2025 to approximately double. Most of CV5 volume is currently in our IoT business, although we expect an automotive OEM to start production in the second half of the year. The fact that we have already achieved high-volume mass production at 5-nanometer helps pave the way for our other 5-nanometer SoC such as the CV3 family. Second, the automotive market we sample both the high-end production version of our 5-nanometer CV3 as well as a 5-nanometer version for China. At the high end, we sample CV3-AD685 targeting L3 and above autonomy. And this central domain controller is currently at is in evaluation at multiple OEMs and Tier 1s globally. So far, we are finding success in L3 and above commercial vehicles. For the basic highway L2+ opportunity in China, we introduced a CV72AQ, and we have numerous Tier 1 design wins and OEM discussions underway. Third, we introduced our generative AI, generative AI strategy for the age of the network, and we are assembling our 5-nanometer [indiscernible] 1 processor targeting applications range from IoT devices to edge servers. Fourth, we'll continue to build out the CV3 automotive platform to offer our Tier 1 and OEM customers turnkey options with our software stack and our centrally processed HD RADAR [indiscernible]. We started the new year at the Consumer Electronics where we hosted over 200 customer meetings and made a number of significant announcements for automotive, gen AI and our new corporate development platform. We were pleased to receive a CES Innovation Award for the second year in a row, this time for our centralized radar processing architecture. In December, we unveiled our first -- our latest software stack for Level 2+ and the higher autonomous driving applications. This software is optimized and can scale across our entire CV3 processor family, enabling OEM to get to market faster and reduce development costs. The new software stack, including the perception, fusion and planning layers is primarily deep learning based, which allows software development to scale more easily, resulting in a more accurate solution. Finally, most important, we rely on high-resolution camera and the radar perception data to create a real-time map inside the vehicle. And for this reason, we eliminate the use of stored HD maps that may contact stale data, which result in improved results and reduce costs for OEM. If needed, the software stack is available in modules and can be combined with an OEM's own software intellectual property. During the CS show, we demonstrated a stack running on a single CV3 automotive AI domain processor in our own autonomous vehicle, successfully completing over 150 accounts right. the demonstration integrate our [ aculite ] radar algorithm for the first time. We also announced the expansion of our CV3 processor family with addition of our CV3AD635 and the 655 SoCs. The new CV3-AD635 supports a sensing suite that includes multiple cameras and radars to enable mainstream Level 2+ feature set such as highway autopilot and automated party. In addition to meeting the GSR 2 and uncapped centers. Additionally, the 655 enables advanced Level 2+ with urban autopilot as well as support for additional cameras, radars and other sensors. With the previously announced flagship along with the China-focused CV72AQ SoC, the CV3 family of 4 processors now covers the full range of AD and ADAS solutions from mainstream to premium passenger vehicles. The new CV3-AD SoCs were endorsed by our partner, Continental. Kodiak robotics, a leading autonomous vehicle company focused on tracking and the defense announced that it had selected our CV3-AD685 AI domain controller for its next-generation autonomous vehicles. In IoT markets during CES, we announced we are bringing Gen AI capabilities to the edge through the introduction of our 9 processor for on-premises applications. This SoC supports up to 34 billion parameters, multimodal large language models, LLM with low power consumption enabling QA-for-ege occasions. We demonstrated multimodal RLM running on the new 1 processor at a fraction of the power per influence of leading GPU solutions. And broadly aims to bring in AI to a wide range of age application including video security, robotics and industrial applications. Quanta Computer announced it was partnering with Ambarella to develop products based on our CV3-AD685, CV72 to a new online processor to address cutting-edge AI devices. This offering address the growing market demand for a diverse range of neural network and RMs and the well-empowered business accrue sectors, including autonomous vehicle, smart surveillance, robotics and health care. Qualcomm demonstrated PCIe add-in cards based on our as well as showing automotive ECUs based on CV3-AD68. We also introduced and demonstrate our new Cooper developers platform. Cooper offers seamless integration of software, hardware, state-of-art fine-tuned AI models and the services that provide universal support of Ambea's entire portfolio of AI SoCs. We have now successfully deployed Cooper to some of our IoT customers worldwide. I will now quickly highlight some of the custom products announcement made during the last quarter. In the Chinese automotive market, we continue to expand our position in this important market. During the quarter, GAC Auto announced [ INX MAX ] passenger car with combination driver monitoring and in-cabin sensing base on our CV25AQ automotive AI [indiscernible] processor. GAC also introduced this [ Tronc ] M8 passenger car with driver monitoring and multichannel occupancy monitoring also based on our CV25. And in January, [indiscernible] unveiled its [ 89 million, ] including an electronic mirror system based on our A12 automotive SoC. In the enterprise IoT market, career market leader, [ HanhaVision ] introduced multiple models based on AI region SoC, including 4K and the full channel multidirectional cameras based on our CV2 SoCs and AI thermal camera based on our CV22 SoCs. All Korean camera supplier Idis introduced a 2-megapixel voice over IP video intercom based on our CV28 SoC. And Taiwan-based Vivo Tech also introduces new 87-V3 family of IP camera based on our CV22 AI SoCs and featuring fixed form and bully models with advanced AI capabilities. In the home monitoring market, Canadian service provider Telus announced its home view doorbell camera based on our CV28M SoC and featuring advanced AI detection. In summary, looking forward, our key objectives to restore revenue growth and profitability while continuing to drive our strategic R&D priorities for AI inference process opportunities at of age. To achieve this goal, we are highly focused on commercialization of technology and products we have developed. And in particular, converting the multiple RFIs and RFQs we are currently working on for CV2 and CV3 into awarded business. Furthermore, returning our IoT business to its positive secular growth projections is very important, and this includes our early business development from new Gen AI [indiscernible]. In conclusion, we have not been distracted by the prolonged industry-wide cyclical downturn, and we see the secular trends we address: safety, security and automation remaining very strong. The increased market attention on inference processing, in particular at a is aligned with where we have been investing. In the new year, we are very excited about the opportunities we are working on, and we look forward to moving more business into 1 color, and I'm excited about what we have -- what we will achieve in the years ahead. With that, John, will now discuss the Q4 and the full year fiscal year 2024 results and outlook in more detail.

Thank you, Fermi. Before I begin, I would like to say that I'm honored to assume the CFO role. I've been working with the team for 7 years, and I'm very excited to help the company as it pursues growth in its target markets. I'll now review the financial highlights for the fourth quarter and full fiscal year 2024, ending January 31, 2024. I will also provide a financial outlook for our first quarter of fiscal year 2025, ending April 30, 2024. 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 along with acquisition-related and restructuring costs, adjusted for the impact of taxes. Fiscal year 2024 revenue decreased 32.9% to $226.5 million. IoT revenue was about 2/3 of the total revenue and declined about 40% for the year. Auto revenue represented the balance of revenue and declined about 14% for the year. From a product point of view, a large majority of our fiscal 2024 revenue decline was from our human viewing video processor SoCs. For fiscal year 2024, non-GAAP gross margin was 63.3% versus 63.9% in fiscal 2023. Non-GAAP operating expense increased 3.9% for the year versus 17.6% in the prior year. Ending cash and marketable securities totaled $219.9 million up from $206.9 million at the end of the prior year. For fiscal Q4, revenue was $51.6 million, slightly above the midpoint of our prior guidance range. up 2% from the prior quarter and down 38% year-over-year. Non-GAAP gross margin for fiscal Q4 was 62.5%, in line with our prior guidance range. Non-GAAP operating expense was $44.1 million, approximately flat with the prior quarter and below our prior guidance range of $45 million to $48 million, driven by continued expense management and the timing of spending between quarters. We remain on track to our internal product development milestones. Q4 net interest and other income was $2.1 million. The Q4 non-GAAP tax provision was approximately $119,000. In fiscal Q4, we recorded a onetime GAAP noncash tax charge of $22.7 million, establishing a valuation allowance on certain U.S. deferred tax assets that were deemed more likely than not to be unrealizable in the foreseeable future. This valuation allowance was excluded from fiscal Q4 and non-GAAP tax provision, consistent with our historical practice for changes to tax valuation allowances. This adjustment is a noncash tax charge required by GAAP based on the proportion of taxable income in the United States. We reported a non-GAAP net loss of $9.8 million or a $0.24 loss per diluted share. Now I'll turn to our balance sheet and cash flow. Fiscal Q4 cash and marketable securities decreased $2.4 million from the prior quarter to $219.9 million. Receivables days of sales outstanding increased from 42 days in the prior quarter to 44 days, while days of inventory decreased from 145 to 131 days. Inventory dollars declined 6% sequentially and declined 28% from a year ago. Operating cash outflow was $4 million for the quarter. And for the full year, we generated operating cash inflow of $19 million. Capital expenditures for tangible and intangible assets were $1.9 million for the quarter and $12 million for the year. We had 2 logistics and ODM companies representing 10% or more of our revenue in Q4. WT Microelectronics, a fulfillment partner in Taiwan that ships to multiple customers in Asia, came in at 55% of revenue for the fourth quarter and 53% for the full fiscal year 2024. Chicony, an ODM who manufactures for multiple end customers, was 14% of revenue for both the quarter and the full fiscal year 2024. I'll now discuss the outlook for the first quarter of fiscal year 2025. Our early actions during the cyclical downturn in the semiconductor industry have helped our customers navigate their high inventory balances, and these actions are now enabling our business to stabilize and begin to recover. For fiscal Q1, we estimate our total revenue will be in the range of $52 million to $56 million. We expect sequential growth in both IoT and auto. We expect fiscal Q1 non-GAAP gross margin to be in the range of 61.5% to 63%. We expect non-GAAP OpEx in the first quarter to be in the range of $46 million to $49 million, with the increase compared to Q4, driven by new product development costs and employee-related expenses, which we were able to delay in previous quarters. We estimate net interest income to be approximately $1.5 million. our non-GAAP tax expense to be approximately $500,000 and our diluted share count to be approximately 40.8 million shares. Ambarella will be participating in a fireside chat in hosting one-on-one and group meetings on February 29, in New York City at Susquehanna's Technology Conference. We will also be participating in Morgan Stanley's TMT Conference in San Francisco on Monday, March 4. On March 18, we will participate in the ROTH Conference in Southern California. We hope to see you at 1 of these events. 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.

[Operator Instructions] Now first question coming from the line of Quinn Bolton from Needham.

This is Neil Young on for Quinn Bolton. So you said you were seeing project delays from Tier 1s and OEMs as well as volume reductions in planned projects, which you called out more of an inventory issue. How's that inventory improvement -- I should say, is that inventory improvement progressing ahead of where you thought it would? And if so, are you starting to get the sense that these projects will resume soon? And then I had a follow-up.

So you are referring to what we said a quarter before?

Yes.

So I think in November, we provide I think in early December, when we provide our final guidance for this year. I think we talk about there's a project pushed out and from OEM Tier 1 is also some decisions for new projects also got delayed and also, there are some inventory. I think what we are seeing is still consistent with what we have said in November and December last year, I think there's no new updates. I don't think we haven't seen new development in terms of a further project like delay or push out.

Okay. So on the auto side, regarding inventory, you aren't seeing any improvements?

We haven't seen any improvement, but we are not saying there is -- it's getting worse.

Okay. And then for my follow-up. So in the past, you talked about how the first CV3 revenue would come from China. I believe in your opening remarks, I heard you say you're engaged in discussion with multiple Tier 1s and already have multiple design wins on the way. If that's the case, when do you think you'll see first revenue from those wins? And then maybe just an update on the demand environment in China.

Right. So for CV72AQ, we expected that the first revenue from those design wins will be in calendar year 2026 that we have been -- we have talked about this in the previous call. And basically, that was a low -- CV72AQ is basically the low end of CV3 family and addressing a first level of Level 2+ for example, for the ADAS plus smart parking. So that's the new market in China, and we're working on we already have design with Tier 1 and working with OEM design wins right now. So -- but I think for the market development point of view, I think China continue to be 1 of the focus area that we are in because I think that I think everybody see the EV development in China, and we believe autonomous driving also what happened in China faster than in other areas. So that's definitely we believe we can monetize our CV technology in China faster than any other area.

Our next question coming from the line of Christopher Rolland with Susquehanna.

Just about your N1 product. Maybe any more thoughts on how large this could be for you guys. Have you considered or has anyone talked about combining multiple chips into a server or appliance? And then lastly, does this meet the Chinese compute restrictions for import as well?

Right. So first of all, in terms of 1, we definitely believe that -- first of all, we can -- technically, we can put multiple chips together is and to serve a high-end solution. But so far, we believe a single-chip solution has aged will meet demand for a lot of our current customer, maybe even new customers. But I do see your point, if you want to go to the a server side that with multiple chip will provide a better solution. Definitely, that's the direction we are looking at. And the current solution that, for example, we demoed with our partners building PCIe card today is a single-chip solution, but it can be multiple chip in the future. In terms of the American regulation, I think because of our architecture, although we can provide high performance at very low power consumption, but our total top number as well as the bandwidth is much lower than our competition. And that's our strength on our architecture that we can use smaller top number and lower bandwidth to achieve a similar or higher performance.

In terms of the market size, we've had many discussions at CES and afterwards with customers on our Gen AI and LLM products, and we see really good feedback about what these products can do. And many customers, we found out just were not aware that gen AI models like Lava could run so efficiently on a sub-50 watt SoC. And so this has triggered a lot of discussions with our customers and how they're going to use the product. And we're going to wait to put some market sizing figures out until we're a little bit further down that process. But the feedback is really good, especially doing this on the sub-50 watt SoC. Do you have a follow-up, Chris?

Yes. Maybe around the kind of AI and camera opportunity. Maybe if you could describe that, I mean, there's so much focus on auto, but next gen, like security cameras with all this AI functionality, like what are growth rates for that market? Do you have now visibility into a funnel to kind of refresh that in to revigorate that market? And what kind of growth could we be talking for kind of that edge market as well with your products?

Well, for our SAM, we haven't updated it for Gen AI kind of like the prior discussion. So we're still sizing that up -- but the prior SAM CAGR, if you will, that we talked about, was in the low teens range, thinking of a 5-year SAM CAGR for that market, but that does not include the Gen AI products, and we're going to take a little bit longer to put those numbers in terms of kind of the insight into building momentum in this market and any sort of funnel, I'll pass that off to Fermi.

Yes. In fact, although we talk a lot about auto because there's a huge opportunity, but we never underestimate the importance of a security camera market for us. This is really a big portion of our revenue, and we continue to believe that AI application for security camera is important for us, and we continue to develop new platform. For example, we announced the CV72, and we'll announce new chips for this market in the near future. So I think we believe that the AI performance demand in security camera continue to grow, and we want to be -- continue to be the player and the dominant player on the mainstream high-end product line.

And our next question coming from the line of Matt Ramsay with TD Cowen.

I guess, Fermi, I wanted to follow up with you on some of the initial feedback on the L1 from an inference perspective. And I guess, it's not a surprise to me given the engineering and architecture team getting good feedback on low-power inference. So I guess my question is, as you get that good feedback and you're interacting with customers that could potentially ramp this product over time, given kind of where the P&L is for you guys right now during the correction, what's the business model over the next 12 to 18 months to start to really build the business around this and get something that could ramp at scale given the software investments that you need, et cetera? Are you -- are customers willing and are you willing to do sort of NRE payment arrangements, are people willing to invest the long side you on software? I'm just trying to figure out. I can see big potential here, but there's also some limitations on capital given where the business is. And I'm trying to understand what the discussions are to get you from point A to point B if this is going to be a big product.

Right. So I think you make a good point. I think for the L1 development, it's going to be significant for us. But that's why we are open for any kind of business model, including from partnership to NRE numbers. I think with L1, we already can address some of the customers, particularly our existing customer demand. And also on the software, in fact, we can deal show you that our investment on the software and tools and the silicon can be leveraged for our first generation ship. So from that point of view, I think our majority of our investment for 1 is done. So the real question is what's our roadmap moving forward. And we -- for example, if we look at the Cooper development, although we defined Cooper for other purposes, but definitely directly apply to our L1 development. So let's talk about for our LLM or Gen AI roadmap. I think that's where the difficulty is, right? I think it's -- from a PR point of view, if we want to do this, we need to continue to invest in R&D for new chips and maybe even new software. So from that point of view, I agree with you that we have to look at all the possible scenarios, including a partnership as well as an NDA. So for the -- so some of the NRE payment for us to pay for the current cost. But I think based on the feedback, it's become very clear that it's not -- LLM is not only for the data center. LLM will penetrate to the device and our current existing customer and future customer or want LRM as a part of the road map. So I think that we need to be flexible to develop a road map for our customer. And we have to figure that out sometime this year.

Thank you for all the thoughts there for me. I guess as my follow-up question, where the revenue levels are right now, you guys have been consistent the last couple of quarters that you're working with the customer base to burn through inventory that they had built, and you're clearly under-shipping in sell-through by a pretty significant margin to do that. So I mean I asked this last quarter, and maybe it was too early to ask, but now that we've had 3 more months. Do you have a feel now as to what the steady straight sell-through revenue level of the business is currently just with the designs you've won, particularly in the security camera businesses? What what's sell-through and what's the market size right now after we've gone way up and then way down on the inventory correction? What's kind of the steady state sell-through that you're under-shipping to burn through inventory? Do you have an estimate for that?

Yes. So we are trying very hard to understand numbers. So let me give you my thoughts. I think when I look at the number at the peak, we shipped probably $92 million a quarter at the bottom, we shipped roughly $50 million. And when we do all of statistics and the numbers that the model we built, we feel the midpoint of that 2 number is probably the comfortable level for us, and we are definitely working hard to go to that to reach that level. So I think roughly in the $70 million range is probably the number we are shooting for when everything get equalized.

And our next question coming from the line of Tore Svanberg with Stifel.

My first question for me, so you talked about fiscal '25, you expect to see growth in both auto and IoT. I was just hoping you could give us a little bit more sort of the puts and takes on how you think the year to progress. Obviously, there's still probably some linger in inventory, especially on the auto side. But yes. Any more color you can give us as far as the growth you're expecting in both segments this year.

You are talking about CV5 or?

No. Overall.

You mentioned you expect both segments to grow this year. So if you could just more sort of the dynamics. Yes.

Right. So I think for the -- let's talk about IoT first. I think for IoT, it's pretty clear that with the CV2 product line that we have been growing CV revenue from close to 60% last year, and we're going to -- we believe that the momentum of CV2 family will continue, particularly after the inventory problem is behind us. So I think at that point, I think CV2 family will drive the growth for us. But more importantly, I think in the script, we talk about C5 will start ramping. Last year, we did 0.5 million units. And this year, we're probably going to double it. And that also, if you consider ASP, that could be meaningful growth for us. So I think that's square on the IoT side. On the automotive side, I definitely think that First of all, we continue to announce the CV2 design win in ADAS, in OMS, CMS, on electronic mirror and recorders. Those are continue to be a big portion of our revenue, but also we have announced -- announcing some partnership with CV3, early customers that we start delivering samples and also partnership with NREs, that will definitely play a role in our CV3 revenue -- sorry, our automotive revenues in there. So I think overall, although that automotive market continued to be weak based on the feedback from the market. But I still believe that we are a small player in the automotive space, and we try to be a big one throughout the process. We're looking at more along the line, our growth with the current design wins. So I think that's how we feel comfortable that automotive will also have growth this year.

Yes, Tore, from a product point of view in fiscal '25, our AI inference products, it's almost all CV2 will be more than 100% of our growth. That means the video processor business will -- which was down substantially, as John mentioned, in fiscal '24, it dropped about $80 million. That rate of decline in video processors will begin to really taper off in fiscal '22. Do you have a...

Yes, That was very helpful. As my follow-up, I was pretty impressed with the Cooper development platform when I saw your samples at CES. And I was just wondering how that development platform is helping you secure more business activity because it does seem like it was an important piece of the pie that was missing, but obviously, now that you have it readily available.

In fact, all our existing customers are eager to get a hands on Cooper tells me a lot about how much they like this development because now it's become very easy for them to put software through 2 different Ambarella platform but different silicon lens. And also, it's easy to transfer the software and the function or AI algorithm from chip to chip. So this whole development is important not only for us but also for our customers. And I think for the existing customer that make their development work even more comfortable and faster. So it will help us to keep both customers. But also for the new customer, even on the in the LLM part, I think that we are -- we can provide an environment for customer quickly can convert their software algorithm to run on our chip is important for our designers.

And our next question coming from the line of Ross Seymore with Deutsche Bank.

When I think about the ASP that you mentioned for going from CV2 to CV55 or even backwards looking due to CV2 itself. Can you just walk us through again kind of orders of magnitude or pricing ranges? How much for ASP as a tailwind in calendar year '24? And what do you expect them to be in calendar '25? .

Right. So for CV2 family, I think we talk about the price can be anywhere from the high single digit to the probably $30 range, and that's the average ASP probably high teens. That's a CV2 family. And the CV5, we're talking about anywhere from the low 30s to high 40s in that range. And that's -- and with our run rate, we think that we can maintain a very healthy not only ASP, but also gross margin in that product line. Then CV5, the effect we have CV72 that we mentioned, the price range is similar to CV5, but for AIoT, it's a different product line. So I think -- and then we talk about CV3, the ASP is anywhere from the $40 to $400 from CV72 to CV3-685. So let's just give you an idea of ASP changes.

Great. And then I guess you talked about the year growing in both sides of the business. Obviously, we have the first quarter guidance and talked about a little bit of the trajectory in a prior question on both your 2 sides of your business. But if we think about the kind of the second half versus the first half, it seems like you need some relatively sizable sequential increases on a percentage basis to get to that sort of number. Do you think you will be well within those kind of those average of roughly $70 million in true sell-through numbers? And if so, is that kind of a second half dynamic? And I guess is that more just about shipping to demand, so the inventory headwinds abate? Or is it about new products ramping?

Right. First of all, we didn't guide any quarter to be $70 million in our guidance. We talk about -- we believe that we're going to have growth this year and also believe that our Q1 guidance. But overall, I think when I look at the number that Street is predicting, I think it's reasonable. And also that based on what we have seen with our customer demands and as well as our booking, I feel comfortable with the current Q1, Q2 guidance. Of course, Q3, Q4, we haven't seen enough booking, but however, the momentum is there. So I think I'm comfortable that we're going to grow. And in terms of our quarter to high quarter-to-quarter growth, we haven't provided any guidance on that yet.

Yes. And Ross, just to follow up on the ASP question. Our ASP in fiscal '24 grew about 15% year-over-year. and looking into the next year, it really depends on the mix of video processor versus CV. But even within the CV2 family, the ratio of CV5s to some of the lower end CV2s. Then of course, we won't have CV3 revenue contributing in fiscal '25. So it should be some increase, but it's just hard to say how much now.

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

Congratulations on the strong results. Just on your one, as you're talking to customers about it, what is the competitive landscape? What are some of the alternative designs that they're looking at? And I guess, is the GPU still being considered even as an edge processor?

Well, some low NGP being considered, but as an a processor, you really need SoC with very low power consumption. And with that, GPU is much less considered. But however, I do believe that Qualcomm definitely have an ambition to come to this market. And when we compare to them, just like when we compare to them in the automotive space, I think we can deliver higher performance and lower power consumption. That's consistent to be the case. So I do believe we are looking at very similar competitors like our automotive market.

Great. the getting a lot of leverage out of the 5-nanometer process. You've got lots of parts, price performance ranges with the 5-nanometer. Is there anything in your road map looking to go below 5-nanometer now?

Yes, we have to. I think there's no chance we stay at final metal force for long. But however, I think it's really driven by 2 things. One is whether we can justify the cost and also whether that the performance requirement. But I definitely believe that you'll start hearing us talk about our next-generation of process selections in the near future.

[Operator Instructions] And our next coming from the line Joe Moore with Morgan Stanley.

Great. Fermi, you had alluded to some OEM wins for CV5 that start to ramp in the second half of the year. Can you talk about what applications you're addressing there?

It's an EV truck in Western space. And we definitely -- we have been working on this case for several years, and customer doesn't allow us to talk about yet just yet. But I think that since we are close to announce their product, and I feel that we should -- we feel comfortable to share with this news, but not to mention the customer names.

Great. And then I guess as far as the L1 product goes, you guys have kind of always shied away from doing anything in a phone because you don't want to become a feature and a chipset. But obviously, a lot of the potential large language model inference could be in devices like phones. So can you just talk about what -- are there opportunities around that to do coprocessors -- or where do you kind of draw the line of your participation?

Right. Since both Qualcomm and NVIDIA -- sorry, Qualcomm and MediaTek are very eager to come into introducing products in the phone space for LLM. I feel that our opportunity is limited because our -- my idea is that even LLM on the phone because you have 5G connectivity you might be able to use some LLM at age, but still leverage the 5G, so you can connect it to the cloud to run most of the LLM functions on the server side. So with that, cellphone becomes a limited opportunity for us, not only because Qualcom, MediaTek has advantage in terms of market share there, but also the usage model is really not purely age. It's a combination of ange and the cloud. So my feeling is we are going to look at pure edge devices that focusing on the battery sensitive and also the latency-sensitive applications just like what we had before.

And I'm showing no further questions in the queue at this time. I will now turn the call back over to Dr. Fermi Wang for any closing remarks.

Yes. And I want to thank all of you for joining us today and looking forward to talk to you in a different conference or next time. Thank you.

Ladies and gentlemen, that does on our conference for today. Thank you for your participation. You may now disconnect.