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

Thanks, Carmen, and good afternoon. We appreciate you being here for the conference call on our financial results for the first quarter of fiscal year 2020. Joining me on this call are Dr. Fermi Wang, our President and CEO, and John Young, our CFO. What we mainly want to do today is give you information about how we performed in the first quarter of fiscal year 2027. During this discussion and when we answer your questions, we will make forward-looking statements about our expected financial results, our financial prospects, market growth, and demand for our solutions, along with other matters. Such statements rest on information available at present and carry risks, uncertainties, and assumptions. If any of those risks or uncertainties come to pass, or if our assumptions turn out to be wrong, then our actual results might differ materially from what these forward-looking statements suggest; we have no duty to revise these statements, and these risks, uncertainties, and assumptions, together with additional details about possible risk factors that might impact our financial results, are described more thoroughly in the documents we submit to the SEC. You can find our press release covering the first quarter of fiscal year 2027 results, transcripts, historical results, SEC filings, and a replay of today's call on the Investor Relations page of our website. Today's call content, along with the materials posted on our website, belong to Ambarella and may not be reproduced or transcribed unless we give prior written consent. Before we begin the call, we would like to see you at a few of the investor events listed below, which are set for our second fiscal quarter. On June 2, we will be at the Bank of America's TMT Conference in San Francisco, and on June 23, at Northland's Virtual Equity Capital Markets Growth Conference. On June 23 and 24, we will host investor meetings in Baltimore and Boston, and on August 18, at the Rosenblatt AI event. When you plan your calendar for our third fiscal quarter, please keep in mind that we are a sponsor at the AI Infrastructure Summit in Santa Clara on November 15 to 17, and we look forward to seeing you there, where we will head the physical AI track with several AI product demos in our exhibit area. Fermi will now give a business update for the quarter, John will go over the financial results and outlook, and after that we will take your questions. Fermi?

Thanks, Luouis, and good afternoon. I appreciate everyone joining us on this call today. In our first fiscal quarter, we hit the key targets we had set for ourselves on revenue, gross margins, and operating expenses. Even more significant is that we keep strengthening our leadership position on the HA platform through advances in technology and products, which lets us tackle both established and newly emerging use cases. As a company recognized as a leader in HAI, we are now moving into an important new stage of market development, marked by long-term agreements with customers that can create a steadier revenue flow and carry lifetime revenue potential well beyond anything we have achieved before. Let me share a few thoughts on how the market looks right now; in Q1, our revenue landed at the top of the typical seasonal band and a bit above the middle of our guidance. The demand signals and the long-term secular growth picture for HAI stay extremely robust, and I have great confidence in our capacity to capture that opportunity, especially as AI workloads grow more intricate. LT applications accounted for roughly three-quarters of total revenue and declined on a seasonal basis, as our enterprise security camera market rose by high single digits quarter over quarter, offset by a double-digit quarter-over-quarter drop in our consumer IoT business. Automotive revenue hit a record high for us, with growth in the strong double digits, and the reason is that AI has quickly taken hold in commercial vehicle telematics — markets that are big and still getting bigger — plus automotive safety uses. After we spent a number of years expanding our AI training capacity inside data centers, the AI market's focus is increasingly turning toward AI that acts, and as we see it, the inferencing side — the processing — is spreading out more and more. At our company, that processing is moving to the edge as well as to the fiscal AI layers within the network hierarchy. As the age market shifts toward Gen Ai and especially Agentic AI, our standing only improves, and I'd like to dig into that further. Before I get into what sets Ambarella's positioning apart, I want to remind everybody about the Advantage HAI work as it relates to the data center. HAI processing reduces latency, draws less power, lowers commission costs, and improves both privacy and security — so what is it that makes Ambarella's HI platform so favorably positioned? For starters, Ambarella's HAI platform is comprehensive and well established, yet it keeps expanding and evolving to stay current with emerging AI trends. Our view is that a broad AI platform with a high degree of programmability is needed to serve a large variety of use cases, which lets customers become more efficient through software and by scaling their business. The software platform we offer is now open, simple to work with, and accommodates many kinds of AI models — over 200 distinct AI model architectures have made it into production. Altogether, we've delivered in excess of 46 million AI SoCs, and 12 AI SoCs are already on the market, delivering as much as hundreds of top life performance. One additional element places us in such a powerful position regarding generative AI and agentic AI: our software tooling along with AOCs unify every function of an accelerated computing system within a single platform. In data center environments, functions such as aggregating data, speeding up processing, running CPUs, and handling other system tasks are usually handled by multiple discrete SoCs purchased from different suppliers. But to win at the edge, our AI SoC brought every one of those capabilities together. Perception fusion, AI acceleration, CPU processing, and other system-level functions were all integrated onto one chip. Our differentiation comes not just from custom processing blocks and sophisticated integration through advanced packaging — it also stems from our proprietary algorithms, the complete Hasta software stack, and AI agent frameworks that unify whole systems as workloads become increasingly complex, such as generative AI reasoning across multiple models and workflows driven by autonomous agents. Our deep expertise spanning the whole accelerated computing stack, tuned essentially for a deployment, is becoming both increasingly uncommon and strategically valuable in this industry. Put another way, since our customers demand maximum performance from their AI applications, only an extremely small number of companies are capable of delivering this — and even fewer are as proven and established as we are; we're now being seen as one of just a handful of companies able to tie everything together as AI workloads get more complex; we're moving into a new phase of HAI and fiscal AI market development, in which we're holding many discussions with customers eager to build deeper relationships, multigenerational commitments among them. Those could take the shape of long-term agreements, or they might involve our usual products, our semicustom AI SoC tailored to a particular customer's workload, or both. Relative to the customer relationships we currently have, LTAs pave the way for long-term partnerships potentially built around a structured contract that addresses volume and pricing and typically runs 5 years or longer. Further ahead, we expect LTAs to turn into a major driver of revenue growth, improving visibility, generating less volatility and giving our revenue greater predictability. The first LTA illustration we have involves our initial 2-nanometer chip along with the semi-custom HAI SoC, which was taped out in January. Known as CV, this AI SoC is set to serve both consumer and enterprise applications within the IoT endpoint market. Through this multiyear arrangement, we've taken on the job of building a semi-custom ASIC for a client that needs help handling specific, complicated AI workloads; that same AI ASIC will be sold as a standard offering to a broad set of additional buyers in other sectors. Later today, we also unveiled a second significant one — this time involving Hanhwa, based in South Korea, aimed at the enterprise CapEx portion of the IoT space; under the Hanhwa deal, the LTA encompasses procurement and joint creation of Ambarella's AI technology throughout Hanhwa's product portfolio and the wider industry, reaching into physical security, operational automation, life sciences, robotics, plus other industrial markets. Revenue that could come from this agreement is more than $800 million over a stretch exceeding 10 years. It also stands as among the biggest deals ever signed by Ambarella, and among the earliest of its type within the HAI semiconductor sector. Because this partnership spans multiple generations, the two firms should be able to coordinate their technology road maps together, speed up product development cycles, and launch new categories of AI-enabled products into the market at scale. Standard AI SoCs that we offer to customers across many markets will be part of this relationship. Past those initial two LTAs, talks are underway with additional companies. I'll also give an update today on robotic edge infrastructure and the automotive sector, both of which hold substantial market opportunity for us. I'm glad to share that we've crossed 15 or more design wins in robotics — aerodrones included — and those have brought in more than $100 million of lifetime revenue, while our robotics pipeline now holds over 30 customers. By combining strong AI compute, advanced computer vision, and very low-power operation within a single architecture built for the edge, our AI SoC serves as the foundational platform enabling robotic systems to execute vision language action, or VRA, models aboard drones. The CD5 chip is what drives platforms such as the antigravity A1, providing things like 8K imaging, real-time perception, accounts navigation, upscaling on land, and on-device AI inferencing that works without needing an always-on cloud connection for those tasks. As the product evolves from a flying camera into autonomous aerial robots, Ambarella's CV flow AI crater architecture lets manufacturers achieve lower latency decision-making — improved latency, extended flight time, and greater economy at the edge.

The robotics sector is split among many players, and we're securing design wins in numerous additional robotic use cases, such as industrial automation, autonomous mobile robots (AMRs), and delivery robots. Our AI SoC spans everything from delivering perception, sensor fusion, and AI processing to handling decision-making plus the complete autonomy required for real-time robotic awareness and action. This coming together of premium imaging, AI acceleration, and autonomy-enabling VRA models puts us in a strong spot as a crucial enabler of the wider physical AI and embodied AI ecosystem. As I noted before, Q1 saw our automotive business set a quarterly revenue record, and it's tracking toward a fresh fiscal-year record. According to third-party research firms, worldwide automotive production should shrink 1% to 2% during this year. Yet since semiconductor content per vehicle keeps climbing, market research firms likewise expect the automotive semiconductor market to expand 10% to 15% this year. We anticipate that our automotive revenue growth will exceed those numbers thanks to our wins in commercial-free < > and safety applications. The commercial fleet telematics space keeps presenting compelling and thrilling growth opportunities, given an installed telematics base above 100 million vehicles expanding at roughly 10% CAGR — however just about 10% of that installed base is thus far AI-enabled. We're working alongside industry AI market leaders, who increasingly want AI SoCs capable of handling not merely additional sensors but also heavier AI workloads. And our lineup of 12 HAI SoCs fits them extremely well for scaling within this market. I'll also share an update about building out our indirect sales channel, which we announced to supplement our current direct-to-customer operation. Growing our indirect sales channel matters not just for helping us serve the fragmented robotics market, but also for supporting our emerging edge infrastructure business. Since launching our developer zone at CES in January, we've already brought on board six ISVs across vertical industries like retail, industrial automation, transportation, health care, and smart cities, with additional ISVs anticipated to join before this fiscal year closes. In March, for the very first time, we presented an embedded award in Nienburg, Germany, where we ran a live demonstration showing how Embraer AI SoC, software stack, and development tools create a competitive edge across many applications — from AI agent automation and orchestration to physical AI systems deployed in real-world settings. One of our current design partners showed a real-time industrial quality inspection solution running on CV72 and 1655. Several new ISV partners will appear at our booth, among them one presenting a retail AI solution for in-store and drive-thru optimization. Another ISV showcased continuous monitoring for a high-speed rail network, and a third demonstrated warehouse robot solutions. In March, we also hosted an invitation-only event at ISC West highlighting how HAI powers the next generation of intelligent security and physical AI systems. Taking center stage in our exhibit was the CV 7 HAI vision SoC we've just introduced, which combines sophisticated imaging with AI processing done right on the device, paired with the 1655 HAIC age infrastructure aimed at enterprise security uses needing low power and high performance. A demonstration came from one of our ISP partners, showing a smart city security solution built around 75 and 165. Now I'll give a quick rundown of representative customer engagement during Q1, and it's worth noting that for the first time, every example draws on our HAI OC — 3 from the CV2 family and 8 from our new CV7Xpund; within enterprise security, fiscal security is still the main force driving this market, while we observe customers building AI application software that lets their products deliver operational efficiency to their business. Instances include predictive maintenance, supply chain optimization, and automated customer support; over time we anticipate operational efficiency becoming a significant new growth shoot of what we call enterprise security today, especially as Gen AI and Agentic AI get deployed at age. March brought an important milestone: apron Panasonic unveiled the first endpoint camera able to range AI locally, leveraging the transformer capability inside our CV72-AISOC; we also recorded a number of other CV75 and CB72 wins during the quarter, among them < > in South Korea, access in Sweden, < > in Germany. And with a major communication equipment company in the Americas — notably, we added another CV2 with < > South Korea that likewise uses our AI imaging signal processing software; we also took a CV22 platform win with < >, plus another CV win with the major communication equipment company in the Americas; over in the industrial market, we secured one more AI-based barcode reader project built on CV28, this time with < >, extending its reach past the traditional physical security market. In automotive, engagement activity with our safety and panamotic customers stays strong. As an example, we're happy to announce Lite, an industry leader in the commercial and public sector telematic market, has designed CB75 and CV72 into multiple platforms. For the in-cabin pre-installed safety market, we have 2 B72 wing with South Korean-based Tier 1 Era and the CV22FS win for a Western OEM in China. Momentum behind our new products remains very strong, both for fiscal '21 revenue generation and for new products that haven't yet begun generating revenues. Although our 10-nanometer CV2 family of HCA processor for CN location keeps landing design wins, our new 5-nanometer CV75 and B72 — capable of both the < > and the transformer-based gene as well as Agenetic AI — sorry, as well as Agentic AI < > a production ramp and are expected to drive material incremental revenue this year. Beyond that, we continue to expect our new CV 7 AI processor to enter production by the end of the year. And in the first half of fiscal 2028, or less than a year from now. Our 2-nanometer CVA AI SoC should kick off manufacturing. Every one of these fresh items I've just outlined, plus the additional AIA SoCs not yet announced that are sitting in our pipeline, aim at heavier AI tasks and point toward an average selling price far beyond the 15 SoC ASP we posted in Q1. As you can see, plenty is happening across technology, products, markets, and customer work. Let me wrap up this quarter's call with three takeaways. Number one, the AGM sector has barely begun, and energy is gathering across several fronts. Number two, Ambarella stands out plainly as an AI technology platform and a product leader, and we are already, yes already, firmly in place. My sense is our standing keeps improving as AGM workloads grow trickier, and the count of firms able to fold every edge accelerated computing function onto one chip keeps shrinking. Number three, buyers see the first two items and now seek deeper, wider involvement with us. Take LTA deals, which can forge tighter ties and land us in fresh arenas such as robotic, whereas the indirect channel sales ecosystem delivers added scale.

So to wrap up, with all of these pieces falling into place, our goal is to create shareholder value through robust top-line expansion and financial models that are both more varied and more forecastable, giving our investors significant potential for operating leverage. On that note, I'll hand things over to John, who will walk through the details of our first-quarter results along with our second-quarter outlook. John?

Thanks, Fermi. Now I'll walk through the financial highlights for Q1 of fiscal 2027, which closed on April 30, 2026. I'll also give you our outlook for Q2 of fiscal 2027, which ends July 31, 2026. What I'm covering here are non-GAAP figures, and I'd point you to today's press release, which has a full reconciliation between GAAP and non-GAAP results; in our non-GAAP presentation, stock-based compensation and acquisition-related expenses have been removed, with a tax impact adjustment applied. Turning to fiscal Q1: revenue came in at $100.4 million, a bit above the midpoint of the $97 million to $103 million range we'd guided to previously, which works out to a 0.5% decline quarter over quarter and a 16.9% gain versus a year ago. Sequentially, automotive revenue, led by commercial vehicles, saw a robust double-digit percentage rise that was above seasonal norms, whereas IoT revenue declined in line with seasonality. Non-GAAP gross margin in fiscal Q1 landed at 59.9%, a touch above the midpoint of our earlier guidance range of 59% to 60.5%. Non-GAAP operating expense for Q1 totaled $56.4 million, somewhat under the midpoint of the $55 million to $58 million range we'd previously guided. Net interest and other income in Q1 was $2.1 million. Our Q1 non-GAAP tax provision came to roughly $740,000. Non-GAAP net profit for Q1 was $5 million, or $0.11 per diluted share. Let me move on to the balance sheet and cash flow. Cash and marketable securities at the end of fiscal Q1 stood at $277.8 million, a drop of $34.8 million from the previous quarter, though up $18.4 million compared with the same period a year earlier. That quarter-over-quarter decline in cash and marketable securities largely reflected higher inventory levels, which we built up to serve customers better amid a number of new product cycles. Receivables days sales outstanding came in at 35 in Q1, unchanged from the prior quarter, while days of inventory rose from 99 to 145 days. Operating cash outflow for the quarter was $25.6 million. Capital expenditures on tangible and intangible assets totaled $4 million for the quarter. Free cash outflow for the quarter was $29.6 million. In the first quarter of fiscal 2027, we bought back 47,798 shares of our stock, for total consideration of $2.4 million, which works out to an average price of $51.04 per share. In the second fiscal quarter, Ambarella's Board of Directors approved a new $50 million repurchase program running through June 30, 2027, which takes the place of the program expiring on June 30, 2026. This repurchase program doesn't require the company to buy any specific number of ordinary shares, and it can be halted at any time at the company's discretion. One logistics company accounted for 10% or more of our revenue. WT Microelectronics, a fulfillment partner based in Taiwan that ships to multiple customers across Asia, represented 6.7% of revenue for the quarter — that's for the first quarter. I'll now go over the outlook for the second quarter of fiscal 2027. Our outlook for the fiscal second quarter is for a seasonally robust period, with revenue projected to land between $105 million and $111 million, which puts us at $108 million at the midpoint. On a sequential basis, we anticipate that both auto and IoT revenue will rise, driven by growth in consumer as well as CapEx-driven markets. For fiscal Q2, we expect non-GAAP gross margin to come in somewhere between 59% and 60.5%. Non-GAAP OpEx for the second quarter should fall in the range of $56 million to $59 million. We project net interest and other income of roughly $1.9 million, non-GAAP tax expense of about $800,000, and a diluted share count of approximately 44.3 million shares. We appreciate your participation on today's call. With that, I'll hand things over to the operator to take questions.

Earlier, you brought up how the auto segment is expanding at a quicker pace than the end market on its own. My recollection is that you put that end market at 10% to 15%. Previously, you provided revenue guidance for the fiscal year — on your prior call, I believe the figure you gave was 10% to 15%. Now, for this year, what is your thinking?

Yes, so across the full year, our view remains that it's likely 10% to 15%. That is not being changed. And automotive is doing better — it expands more rapidly than the rest of the market.

All right. Turning to the long-term agreements, including the one you just disclosed this evening in the 8-K — the $800 million spread across a period of time — but thinking more broadly, what's your framework for these? Will they represent guaranteed revenue? I believe your wording was potential revenue. How should we incorporate that into our models as we look ahead for the company?

Sure. To start, we've had a run rate with Hanhwa going back 15 years, and we recognize that we capture only a portion of their existing market share. With this LTA, our expectation is that we'll pick up share against their yearly run rate, and beyond that, this represents a commitment spanning multiple generations on both sides, so that this is — we'll be discussing two generations of silicon that the two companies will develop jointly. From that standpoint, we see it as a lengthy commitment period as well as an opportunity to take share from the Haha. At the same time, when you consider our present ASP — our overall corporate ASP sits at $15, yet our CV ASP runs considerably above that figure. So when you add everything up, that's the basis for arriving at this potential $800 million.

Ross, I’ll give you a bit of context on Hanhwa: it’s a large multinational conglomerate, pulling in over $60 billion of revenue each year. Its operations span aerospace, defense, robotics, physical security, life sciences, industrial, Ocean solutions, chemicals — quite a few areas — plus retail services. So a key element of this relationship, looking to the intermediate to long term, involves expanding past what has simply been the physical security tie that Fermi characterized as our present run rate. That way, we’re able to take share in that business. Yet at the same time, the press release mentions, alongside physical security, areas such as operational automation, life sciences, robotics, and other industrial markets. That represents one more highly significant dimension of this relationship.

Yes. Perhaps a follow-up regarding the LTA, and not specifically the Hanhwa one, but how ought we to consider these types of arrangements folding in here across the coming few years? Clearly, these might be rather large. I'm certain you could respond affirmatively to each of them. Is there additionally going to be possibly some OpEx sharing with a few of these customers whose LTAs you signed? Any further detail you might provide us concerning how we ought to consider the scale and how it will be financed across the next 3 years?

So first, I think I'd like to go somewhat high level and say a lot of — the bulk of LTA conversation rests on two things. One is genuinely, when you examine the entire fast AI trend and our customers keep examining the new AI model, the info the greater performance requirement, reduced power efficiency. So to satisfy this sort of AI demand, while for the majority of AGI plication power efficiency is likely the most crucial thing. So it's becoming tougher and tougher to construct a silicon platform capable of addressing all these new applications; so I think that trend truly assisted our customers in thinking about how to collaborate with somebody able to construct a silicon platform that can assist not just on the security enterprise security, but additionally other associated market they are attempting to address. I think that's probably how the LTA began. And within this sort of LTA, the bulk of conversation will involve and — but we'll also develop a product mutually beneficial, correct? So I think those are two things we certainly want to ensure we develop a software platform capable of going across our present silicon platform that provides a complete road map to our customer; and in exchange, they were prepared to assist us in funding those platforms with, particularly not just on the silicon side but additionally on the software side.

That's really useful, and for my follow-up, maybe tying into that platform strategy. Thinking about who we're up against, clearly there are large processor makers. There are analog firms too, and so forth; but how adaptable can your software platform truly be? Since at the edge, the applications will differ a lot from one case to the next.

Right; I believe our distinctive architecture — you recall we've discussed < >. Our — figuring out how we speed up things like image processing, and our CVflow AX reader has — truly been through numerous battles and has been demonstrated to our customers as not just power-efficient, but also programmable enough to handle a wide range of applications and various AI models; for instance, as I mention, our CVflow architecture supports 200 distinct AI — pardon me, model architectures, not 200 models; we mean 200 model architectures, and every one of them went into production. Just to illustrate < > our accelerators. Yet even more crucially, with the newest agenting AI method, everything must be integrated. So there, that covers not just the ISV and the CV flow or impact income H2 for encoder; beyond that, you must bring together the entire SoC for CPU plus IO and DRAM activity to deliver a very — not just power-efficient but also cost-effective answer for our customers. That's precisely where we've essentially built a reputation for shipping such products reliably over the past 20 years; and considering the competitive field, with the DLC quota, certainly, but I honestly don't see anyone else introducing a single complete silicon platform like ours where 12 HAIC can range from very low performance up to a few hundred top performance while staying power efficient. Beyond that, every piece of silicon is supported by one unified software development SDK, so our customers, if they build one product on one SoC, can readily move to another SoC, offering a different price-performance point within their product platform. And that sort of flexibility along with the breadth of our product format, I doubt there are many others who can equal it.

So for my follow-up, I'd like to ask about the Hanwha LCA. In that press release, you mentioned the in-house SOC that Hanhwa keeps working on. That makes me curious — it sounds as though you're jointly designing several generations of ships. Could there be a chance to capture a larger share of the pie from that in-house SoC, or do you believe that in-house SoC will still occupy a portion based on requirements? Then I have one more question after that.

My sense is that both sides intend for more of this codeveloped platform to be used across additional product lines. As a result, we absolutely expect to gain additional market share from Hanhwa through this new development.

Great. So, Fermi, you folks have been discussing the fleet telematics sector for several years now, and on the script it seems like it might at last be beginning to inflect — you brought up an installed base exceeding 100 million units, or plus one of those being AI. But what do you believe is propelling that inflection? Can you identify anything specific that's behind the rather robust growth in the first and second quarter?

Yes. Quinn, this is Louis. Regarding the market opportunity, telematics, there are roughly 100 million subscribers, and third-party research firms such as ABI or Gartner expect it to grow perhaps 10% CAGR. Inside that $100 million subscriber base, perhaps just 15%, 20% of the market is employing AI and AI video as an extra ARPU-generating feature within their platforms. So what you have is the overall telematics market expanding, you've got AI video growing into that. And simultaneously, there's demand for more sophisticated AI workloads and multiple sensors, which is pushing our ASP, the demand for more sophisticated AI chips, higher. So those are the dynamics that we're facing. And we're doing a very good job with share. For instance, this quarter, we announced < >, as Fermi mentioned, multiple platforms with CV72,/CB75, which is they're one of the leaders of this market. So any follow-ups on that?

I'm curious whether shortages of DRAM or storage have any effect on you folks in the surveillance markets or in consumer markets. Are you noticing any consequences from any of that?

Yeah. To begin with, clearly, in the same way as everyone else, there's an effect on us, though it isn't direct — rather, we're causing a direct effect. During the prior quarter, when we discussed this topic, we certainly mentioned that customers are facing considerably greater DRAM prices and flash prices too, plus a possible shortfall in the back half of the year. That condition has remained the same. What hits us most directly is that the majority of our customers are coping with this shortfall by doing the following — first, they attempt to find a different DRAM supplier or to reduce how much DRAM each product uses; we essentially work to optimize how much DRAM a chip uses so they're able to employ a smaller DRAM option; and every one of those efforts means we must provide our AP support to assist our customer. We're glad to do so, since that's certainly something we have to do to help our customer. Yet simultaneously, I'd say that is likely the largest direct effect on us when it comes to engineering resource needs.

All right. Then to follow up on that, regarding the inventory, you mentioned it was there to support product ramps, but that was a pretty sizable jump. I just — could you provide any additional detail, help us get more comfortable with that level of inventory build?

Joe, sure, I mean, if you go back a couple of quarters, we were operating relatively lean. And during the past few quarters, we've examined the opportunities coming our way given the strong growth we experienced last year, and we wanted to place ourselves in the right product categories so we could keep serving our customers. And so, from a fairly low number of days of inventory at certain points last year, we wanted to build a bit of inventory as we headed into this year to be able to do that.

Joe, this is Fermi. I’d like to give you one more detail here. Sensor has told us officially that its supply is becoming more constrained. Of course, we’ve locked in what we need, but we believe it makes sense to accumulate a bit of inventory in case we encounter other supply problems. So from that angle, I think—I think you’ve also heard that supply chain has turned into an issue for everyone. We’re simply trying to be careful and stock up some inventory just in case.

Congrats on everything you've accomplished so far. In that same vein, you're expanding your indirect sales channel. I believe you said 6 teams have been brought on board. Could you perhaps give a bit more geographic detail about these? Also, do they include application engineers? Yes, I'll leave it there with those questions.