Ambarella, Inc. (AMBA) Earnings Call Transcript & Summary

Thank you for standing by, and welcome to the Ambarella's First Quarter Fiscal Year 2025 Earnings Call. [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 and Investor Relations. Please go ahead, sir.

Thank you, Jonathan, and good afternoon. Thank you for joining our first quarter fiscal year 2025 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 first quarter fiscal year 2025. 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 first quarter fiscal 2025 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. The content of today's call as well as 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. Then John will review the financial results and outlook, and we'll be available for your questions after that. Fermi?

Thank you, Louis, and good afternoon. Thank you all for joining our call today. Our Q1 results were 1% above the midpoint of our guidance range with revenue increasing 6% sequentially. As expected, both auto and IoT revenue increased sequentially, and the AI products were about 2/3 of our total revenue. As previously discussed, our customers are in the midst of recovering from a cyclical inventory correction, and the favorable impact from this is expected to carry into the second quarter. We continue to expect our fiscal '25 revenue to grow year-over-year, driven by both the cyclical tailwinds and the secular growth in our AI portfolio. The combination of the cyclical and secular forces is expected to enable our AI inference revenue to grow more than 30% in fiscal 2025. Zooming out for a minute. The significant capacity being added to the AI training network infrastructure globally bodes well for the ultimate deployment of AI inference at the edge, where we participate in the market. The deployment of AI inference at the edge enables end user to more practically take advantage of so many different AI breakthroughs. As focus on AI at the edge of networks increased, we see AI inferencing proliferating in multiple areas, and we believe we are well positioned to take advantage of this. In fact, we are already in the early stage of demonstrating how it may play out for us. In Q1, for example, our customer engagement includes our first passenger vehicle win for our 5-nanometer CV3-AD family of AI central domain controllers. We added another CV3-AD win in the commercial vehicle market, secured multiple enterprise-class AI inferencing wins, and even in other IoT cameras, we are reporting additional wins for our 5-nanometer CV5 AI processor. In the midst of this great change, our opportunity and challenge is to develop AI technology and products that not only are extremely efficient for edge deployment but also flexible enough to execute a very wide range of AI workload across all these disparate applications. We are already in mass production with our AI products for video-intensive CNN networks such as detection classification, fusion, planning, stitching, matting, tracking, framing, auto editing and the new network image signal processing. Now our third generation of AI technology integrated into our CV3-AD and CV7 series of SoC can support transformer networks for a variety of generative AI applications. I would now like to describe customer engagements that can offer some indication on how it can play out for us. As you know, we have made significant investment into our CV3-AD family of automotive AI domain controllers, and we expect the CV3-AD family to be a major revenue driver. So we are pleased to announce our first CV3-AD family wins in the passenger vehicle market, which complements our ongoing wins with the CV3-AD family in the commercial vehicle market. In April, during the Beijing Auto Show, we reached a strategic collaboration agreement with a battery electrical vehicle company in China. This company will use our CV3-AD AI domain controllers in new passenger vehicle models. This is an important development for us in several regards. First, measured by the number of EV delivery in calendar 2023, this OEM is considered 1 of the top 5 new EV companies in China. While most of the OEM delivers -- deliveries were in China last year, the company has an impressive plan for the globalization of its business. Second, Chinese automotive OEMs are aggressively and successfully adopting next-generation technology into their vehicles, which aligns with our strategy to intersect the safety and autonomy domain with next-generation technology, in particular, Level 2+. For example, to improve accuracy, many OEMs in China have aggressively adopted the BEVFormer AI framework for 3D perception tasks in its autonomous driving software stack. And this high-performance framework can leverage to a high degree the unique capability in our CV3-AD AI inference processors, including the ability to process transformers at low power. Third, the efficiency and scalability of our CV3-AD family portfolio is also a major factor in the collaboration with Ambarella as the OEM can reuse the software on low, mid and high-end vehicles. In April, we added another CV3-AD family win in the commercial vehicle market with the announcement of a strategic collaboration with SANY corporation. SANY is one of the world's largest engineering machinery manufacturers providing heavy-duty commercial vehicles to the global market. SANY intend to leverage Ambarella's CV3-AD family of automotive AI domain controllers to develop advanced automated driving solution on its next-generation commercial and special-purpose vehicles. The companies will collaborate on the joint development and the promotion of high-performance and highly integrated automated driving solutions with SANY's goal of achieving start of production for at least one model by calendar year 2025. At the Beijing Auto Show in April, Tier 1 Neusoft Reach announced a strategic partnership with Ambarella. The companies plan to expand on their existing relationship to jointly promote and explore AI-based product technology and market development in areas including autonomous driving and driver monitoring. Neusoft's third-generation forward-facing intelligent camera X-Cube 3.0 is powered by Ambarella's CV22 AI vision SoC to target Level 2 and Level 2+ autonomy levels and has already been mass produced and deployed by automotive OEMs. We are pleased to announce that Seeing Machines, the leading provider of driver monitoring system software, selected our CV25 for its own aftermarket system for commercial vehicles. The Guardian Generation 3 meets the European Commission's general safety regulation for drowsiness detection, a requirement for all the new cars, vans, trucks and buses across Europe. And in April, at the ISC West security exhibition, we successfully demonstrated the latest generation AI technology running vision language model, VLM, on our N1 and CV72 SoCs. Our demonstration includes using the multi-modal VLMs to search video recording to detect objects defined by text and provide near instantaneous results without the need for training specifically for that object. This capability opens up a whole new range of AI-based search capability for enterprise cameras and premise-based AI systems. Our third-generation AI inference technology includes specific support needed to efficiently run these new classes of networks. During the ISC West, we also announced and demonstrated our new 5-nanometer CV75 AI SoC, which provides the performance required to run the latest VLMs as well as AI-based IQ enhancement. This capability to very efficiently run these cutting-edge AI technologies is highly sought after the -- sought after for cost and power-constrained AI cameras used in enterprise, smart city, retail stores, robotic access control and AI-based consumer devices. At the Enterprise Connect conference in March, Poly, a leading global provider of workplace collaboration solutions and wholly owned unit of HP, launched its Studio E360. This center-of-table system utilizes a single 5-nanometer CV5 for 4 8-megapixel cameras with AI inference processing, choosing the optimal framing angle for in-room participants. And also in Korea, the Korea activity for our AI SoCs remained high with the leading security camera maker Hanwha introduced new AI model based on our CV22 and CV2 AI SoCs and IDIS introduced CV25-based models and CPRO introduced new sensor AI camera based on CV22. In the other IoT market, we are pleased to see our 5-nanometer CV5 AI inference processor being utilized in another consumer camera to significantly improve both image quality and to automate a certain user interface function. Insta360 introduced its X4 action camera in April with dual sensors for 360-degree 8K capture. In conclusion, I mentioned last quarter our goals are to restore revenue growth and profitability while continuing to drive our strategic R&D priorities. Q1 represented a step in the right direction, most significantly highlighted by our first passenger car -- passenger vehicle win for the CV3-AD family, but the expanding breadth of CV customer engagement stands out. And we expect this to build a broad foundation upon with multiple applications can drive revenue growth and result in positive earnings leverage for shareholders. The further commercialization and monetization of our technology and product is of utmost importance to us, and I'm excited about the opportunity before us and what we will achieve in the years ahead. With that, John will now discuss the Q1 results and outlook in more detail.

Thank you, Fermi. I'll now review the financial highlights for the first quarter of fiscal year 2025 ending April 30, 2024. I will also provide a financial outlook for our second quarter of fiscal year 2025 ending July 31, 2024. I'll 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 costs adjusted for the impact of taxes. For fiscal Q1, revenue was $54.5 million, 1% above the midpoint of our prior guidance range, up 6% from the prior quarter and down 12% year-over-year. Non-GAAP gross margin for fiscal Q1 was 63.4%, above the high end of our prior guidance range by 0.4%. Non-GAAP operating expense was $46.7 million, approximately $2.6 million higher than the prior quarter and $800 million -- excuse me, $800,000 lower than the midpoint of our prior guidance range of $46 million to $49 million, 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 $2.3 million. Q1 non-GAAP tax provision was approximately $607,000. We reported a non-GAAP net loss of $10.5 million or a $0.26 loss per diluted share. Now I'll turn to our balance sheet and cash flow. Fiscal Q1 cash and marketable securities decreased $16.6 million from the prior -- million dollars -- from the prior quarter to $203.3 million. Receivables days of sales outstanding increased from 44 days in the prior quarter to 47 days, while days of inventory decreased from 131 days to 123 days. Inventory dollars declined 2% sequentially and declined 31% from a year ago. Operating cash outflow was $15 million for the quarter. Capital expenditures for tangible and intangible assets were $1.1 million. We had 2 logistics and ODM companies representing 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 62% of revenue for the first quarter. Chicony, an ODM who manufactures for multiple end customers, was 13% of revenue for the quarter. I'll now discuss the outlook for the second quarter of fiscal year 2025. We remain confident our business is continuing to recover from the cyclical correction led by our AI inference products. For fiscal Q2, we estimate our total revenue will be in the range of $60 million to $64 million. We expect sequential growth in both IoT and auto. We expect fiscal Q2 non-GAAP gross margin to be in the range of 62.5% to 64%. We expect non-GAAP OpEx in the second quarter to be in the range of $47.5 million to $49.5 million with the increase compared to Q1 driven by increased head count and engineering-related expense, including our first 2-nanometer SoC project. We estimate net interest income to be approximately $1.8 million, our non-GAAP tax expense to be approximately $600,000 and our diluted share count to be approximately 41.1 million shares. Ambarella will be participating in Bank of America's Global Technology Conference on Wednesday, June 5, in San Francisco; Mizuho's Technology Conference on June 12 in New York City; and Rosenblatt's Virtual Age of AI Conference on June 13. We hope to see you at one 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] And our first question for today comes from the line of Christopher Rolland from Susquehanna.

Can you hear me?

Yes.

Yes.

You brought up VLMs on the call. Just wanted to know a little bit more about that application. Does that increase your TAM overall? Is this additive to what you're doing in, for example, LLMs? How should we think about that?

Right. So we start talking about LLM and start demoing at CES this year. And then we continue to watch the potential application that we can use with our solution for our customers. And we identify vision language model, which is really a large language model that can respond to text for search of different objects or using that to describe different video things, and we have a live demo. In fact, we welcome to -- you to give -- to call -- to set up live demo for you, how the vision language model used in our -- with our current enterprise customers or other potential customers in that space. So this definitely is a brand-new application being identified by the market and the customer, and we're engaging with our current customer, particularly on the enterprise and security camera market. And they showed extremely interested on this product because you can imagine that this is really a model that can replace the people who use to monitor the different cameras in a control center. And you can use a large language model not to give you early warning but also have -- using hacks to do a search of different objects without really training for those objects. And I think that's really a breakthrough for enterprise side, and we expect that some of our customers will kick off projects with us with VLMs in the near future. And we can use both N1 as well as CV72 for that product.

Great. And then on the auto opportunity, making some progress here, sounds like, on the Chinese EV side of things. Perhaps you can talk about the pipeline, kind of how you see expanding engagements of late here that you're seeing and whether you think there's going to be more to come and whether any of these could actually be implementing your software as well?

Right. So first of all, I definitely believe there will be more to come. And I also believe that the first design win really show that our -- we have been talking about our strengths, which is we can run -- really transform our most advanced network very efficiently in low power. And also, we have a scalable architecture that can really address using a sensor to address from low-end to the high-end vehicles and also that the power consumption that were much lower than our competitors give significant bond advantages in the design win. So that really is the reason we won. But also, I think it's important to point out that just like the EV market, Chinese EV automakers are really aggressive by pushing technology envelope, and for one of them adopting our Level 2 product solution will help us for other potential OEM design wins that we're bidding on.

And our next question comes from the line of Ethan Potasnick from TD Cowen.

This is Ethan Potasnick on for Matt Ramsay. Congrats on the great results. I wanted to ask, is the passenger vehicle win you guys discussed tonight likely to get CV3 to the tipping point with regards to wins at other OEMs? Are there -- and are there other OEMs waiting to see maybe another one kind of go first? Maybe if you guys could discuss those dynamics.

Right. So first of all, I think the lifetime value of this design win is meaningful and also material for us and also that we are very aggressive to target MP date at the end of '25, early '26. And that will help us to really take our CV3-AD platform into production and therefore provide a mature solution to other companies who is considering our current solution. So I definitely believe that not only this design win is help to our pipeline but also going to be a major help for us to secure other design wins in the near future.

Okay. Understood. And as my follow-up, you guys gave a lot of good information around the AI inference opportunity with revenue expected to kind of grow 30%. I think you guys said in fiscal '25, and you called out some progress. I think it was enterprise applications. I was wondering how these -- how the early stages of this opportunity is playing out. What sort of impact some of the announcements you guys made at CES, I think it was the N1 SoC and the Cooper development platform or -- what's kind of driving that? How is the road map progressing? Any early feedback? And any additional insight would be helpful.

Right. So just maybe refresh our gen AI strategy. We definitely intend to leverage our current existing third-generation AI inference engine into gen AI and LLM. And we demoed at, like you said, CES, N1 with Lava demo. And then at the ISC West, we demoed with VLM running on both N1 and CV7. So our strategy for current generation has become very clear. First of all, we're going to focus on edge application, both for edge camera as well as edge servers. Two, we'll focus on the workload that we can -- really can take advantage of our current silicon. And so that's really enterprise market as well as some adjacent markets that can use those Lava model as well as VLM models. And we think -- we continue to believe that our opportunity for gen AI revenues in 2026. We are not only demoing. We're sampling N1 and CV72 to our key customers who are interested in gen AI application. So that's our first generation. And also, as you can see, I think John mentioned that we are investing on 2-nanometer process node. You can think that's a preparation for our next-generation gen AI solution, although we didn't talk about our plan for that yet. We are trying to work on that. But we believe that our 2-nanometer technology development will significantly help us on the technology side for our road map.

And our next question comes from the line of Joe Moore from Morgan Stanley.

In terms of the passenger car win in China, can you talk about -- I think you mentioned it as an L2+, if I heard you right, type of opportunity. Can you kind of describe how many cameras per car and what capabilities you could offer there?

Yes. So first of all, it is an L2+ car. And because they have different models, it's anywhere from 7 to 10 -- 7 to 10 cameras plus radar configuration. And it's designed to do it from the highway level to CD levels of autonomous driving.

Great. And then was this customer -- I know you have a probability weighted funnel. Was this customer somebody that you had anticipated had some probability within that funnel?

Yes. And now I think with the next funnel discussion, they should be in one category instead of probability of categories.

And our next question comes from the line of Kevin Cassidy from Rosenblatt Securities.

Congratulations on the great results and the design win momentum. And maybe, Fermi, maybe you hinted -- I was going to ask more questions about the 2-nanometer process technology. Do you have a time line for when you'd have first silicon? And are you staying with Samsung? Or will this be with TSMC?

Right. So this 2-nanometer will be Samsung 2-nanometer, and we plan to tape out second half of next year and get ready to go into production with 2026.

Okay. Great. And maybe just a follow-up on Joe's question. You mentioned that it's 7 to 10 cameras plus radar. Is that your radar in the design?

We haven't talked about that yet.

Okay. All right. Maybe I'll go to an easier question with just your backlog visibility. If you're -- you're saying you're going to see growth in fiscal year 2025. It -- can you talk about your backlog visibility?

Yes. Our backlog visibility, getting much better than before. In fact, that's the reason we feel comfortable that -- with the Street consensus of $250 million revenue forecast this year. And we -- our lead times are really in the 24 weeks, and that tells us that we should have a clear visibility to middle of Q4. Reminding you that our Q4 is till January. So we see clear booking for Q3 and Q4. That's why we continue to feel comfortable with our guidance. However, let me add one more sentence. I think that we are comfortable with the $250 million consensus level. We believe there is a seasonality in Q4. So I think that is something we should consider. While the -- while we see the visibility, we definitely believe that we're going to continue to see the trend of Q2 to Q3 growth -- from Q3, Q4, just like the normal seasonality before the COVID, we should have seasonality in there.

And our next question comes from the line of Quinn Bolton from Needham & Company.

This is Nick Doyle on for Quinn. Maybe just zooming out a little bit, could you give us a pulse on the Chinese auto market overall? How do you see current EV demand and supply? And then you mentioned 2 CV3 wins in the passenger and commercial markets. Can you size those opportunities at all?

Right. So first of all, we don't play on the EV market. We play on the autonomous driving market. But they are very much correlated. I think what happened in EV is that Chinese automakers continue to aggressively push the technology envelope on the EV side. We see the same trend on the autonomous driving technology side. I think, for example, that almost all the RFQ that we're bidding in China, all of them require most advanced network like BEVFormer, very L2+, CV-level autonomous. All of that is really showing me that they are very -- Chinese automakers are very aggressive to push the technology envelope. And we believe that's played to our strength, and we believe that we can take advantage of that, although that EV market show some slowness in terms of total market size. But I think that from an AD market point of view, we still believe that the growth has continued to be there, and we expect that majority of the EV cars will have a Level 2, Level 2+ system in the coming years, near term.

Okay. The margins -- the gross margins were a bit of a bright spot in the quarter and guide. Could you expand on what's driving that?

Yes. So essentially, it's mix. We had little bit higher contribution than we had forecast for customer NRE projects that came through in the quarter. That's the main driver.