I'll sit here, and then you guys sit here.
Okay.
Okay. We're going to get started. Good afternoon. I'm Tim R. Curry. I'm the Semi and Semi-Equipment Analyst here at UBS. Very pleased to have Ambiq with us, and we have Scott Hanson, who's the founder and the CTO, and we have Fumihide Esaka, who is the CEO. So thanks to you both.
Thank you so much.
Thank you, pleasant and happy to be here.
So I just wanted to start, since you guys are recently to the public market. So I thought, Scott, maybe you can run through, just give us five minutes of just the story, why you started the company, what the competitive advantage is, what the moats are. Would be great.
Yeah. Yeah. Yeah. Absolutely, so first of all, happy to be here today. It's been a great day of interactions. Ambiq was a company that I founded 15 years ago with the objective of putting intelligence everywhere, and when I say that, I really mean putting chips and everything that goes with it everywhere: in our clothing, and the paint on the walls, and the roads that we drive over, everywhere, and this coincides really nicely with what I think is probably the most exciting technology or computing revolution we've seen, and that is really the dual revolution of AI breaking out in the data center, but then breaking out of the data center and out to all these devices around us.
And so the chief problem is when you start taking AI and intelligence and you put it in your clothing and the paint on the walls everywhere else, it requires energy. And that's the core problem that we solve. So we build chips. We build microprocessors that consume two to five times less power than what's out there. And we've got a proprietary technology platform that we call SPOT. This was my PhD research back at the University of Michigan. We spun it out. We founded Ambiq. And we're using that technology to reach four primary markets. So the first one of those is what we call personal devices. It's what I'm wearing on my wrist here. It's anything body-worn. It's AR glasses. It's smart rings. And frankly, that's where a lot of the volume is today.
That's a space where they were early adopters of intelligence everywhere, of AI everywhere. But the other three markets that are key to us are medical and healthcare, industrial, and smart homes, smart buildings. And look, the common thread with all of those seemingly disparate markets is they all want intelligence. They all want compute where none was possible. And it's this central problem that we solve. So we've done quite a bit of volume in these spaces collectively, closing in on 300 million units shipped. And that's set the scene: strong technology platform, strong customer base, strong macro tailwinds. Set the scene for a very strong 2025 for us. So every quarter, we've grown sequentially. We just guided Q4 that was 35% higher than consensus. I know there were a lot of questions about exactly what drove that.
And there's a couple of factors behind that growth this year and behind the Q4 growth in particular. So number one, we're seeing a diversification of the types of personal devices that our customers are driving. two named and notable customers, adding a third. We, of course, have many other customers, but three that have now been disclosed. And we've got a fourth coming soon. So we're seeing a diversification of those customers and of the types of devices that they build. But probably the more important thing is the lines are blurring between what used to be consumer electronics devices like the ones I'm wearing and healthcare devices. And that's evidenced by the fact that you can use HSA and FSA funds to buy these devices today. So that's driving volume. As I mentioned, we had two big-name customers last year in addition to a bunch of smaller customers.
We added the third customer this year, and they were a big driver of revenue for us in the fourth quarter, driving some of that 35% upside in particular, and we believe, after looking closely and talking with them, that should be a sustained trend, so really excited about what 2025 represented. One of the other exciting trends we're seeing toward the tail end of 2025 is a diversification of the types of customers that we serve, so if until now we've been kind of a personal device-oriented company, we're seeing a lot of designs pop up in things like water leakage sensors that get used in an industrial setting, machine health monitors that are used to detect imminent failure, smart headsets that are using voice recognition to understand various commands. Excuse me. And so what that sets us up for is really strong growth for the coming years.
As I look as the CTO, let's say, what excites me is what comes next. And there are two really exciting growth drivers from my perspective. One is that diversification. And I talked to our head of sales before coming here. Tim, I knew you weren't going to let us get away with sharing no new data. So one data point that I'll share is, as we look at the 2027 funnel, if I look at specifically designs that are going into production in 2027, 25% of those are non-personal device products. And that's in terms of dollar value. So this is a good trend. I would like that. As I was telling a lot of investors earlier today, I wish it was this year that I was telling you that. But these designs take time. It takes time for industrial and medical devices to ramp.
What I like is the trajectory that that suggests. Those markets come with higher margins. They will be driven in large part by our existing Apollo family that we've got today that will continue to expand. The other major growth driver for 2026, 2027, or for 2027 and beyond in particular is our new Atomic product. This is our first NPU and AI-first product. Our existing product portfolio is kind of general-purpose microprocessors that happen to be used extensively for edge AI. Atomic was designed from the ground up for AI. It's got a neural processing unit. It's got DRAM in package. I'm really excited about this. This is where I spend a lot of my time. What I'll tell you is that that product tapes out in the latter half of next year. This is one that's deep in design.
We're engaged with customers. We've got our first customers kind of lining up timelines for when they need samples. Samples will go to first alpha customers end of next year, early 2027, first customer ramps in 2027, and then meaningful revenue in 2028. So really excited about how that mix leaves us for the next couple of years. And with that said, I guess I'll pause. That's the Ambiq story at a high level. Happy to answer any questions you have, Tim.
Did he say everything about Ambiq already? I don't have anything else to say.
I'm sure he's got something. Fumihide, do you want to add to that?
No. Again, he said almost everything, so we're very excited. And when we did IPO back in July, we were cautiously optimistic about the edge AI market, but today, I think we're very, very excited about the growth opportunity of our edge AI market. I think it's growing much faster than even we expected, so we're very excited, and I think that we're looking forward to having 2026 and 2027 to be a great year for us.
So when you win, who are you winning against? And what's the primary driver as to why the customer is choosing you?
So let's take in personal device markets today. So whether it's a watch or it's a band like this, we tend to win against companies that have a similar feature set. So it's an Arm-based processor with a similar core, similar memory size. But the customer is looking at it going, "Okay, Ambiq is delivering two to five times less power." One of our big three customers recently swapped us out for another chip. They got actually a 10x power reduction when they moved from the other chip to us. At a system level, this meant three times better system power. And for them, that translated to more functionality, smaller battery size, longer battery life. That's compelling. That's really compelling.
Just so I've asked you this before, but that particular case where they did swap it out and they got such a big benefit, why haven't they done that in everything they sell?
In the case of this particular company, they have, and we have a nice, let's say, roadmap of design when it's happening.
Coming.
Yeah. And I would say with most of these customers, they have to start. You got to start somewhere. It's a big effort to go and port the software. They got to start somewhere, prove that it's all good. I would say that the barriers there are improved now that we're a public company. So as a private company, you got to justify your existence. You got to justify what does your balance sheet look like? And there's all these disclosures that are necessary. Well, we've got this new era of credibility that comes with being a public company.
Yeah. I guess where I'm going with that is, why wouldn't all wearables use Ambiq processors if they can enable such a longer battery life?
Yeah. Well, obviously, I agree with you. And there's a large number of companies that have made exactly that decision. And they've propagated us throughout their portfolio. And it's just a matter of every quarter they launch a new product and we're in it. At the low end, there are designs that we have not won. And that tends to be companies that are doing no processing on board. They're basically collecting data, sending it up to the cloud via radio. And that's kind of the old model. Actually, that's what all wearables used to look like until they realized, "We got to do some analysis here on the edge.
“We got to process this data here,” and so what I see happening is that the barrier between what we win and what we don't win is moving in the direction of us winning more and more of these as they decide to add more intelligence, do more local compute. At the high end, to give you an example in the AR glass space, we miss certain opportunities because, for example, we don't have camera support today. That will be fixed with the forthcoming Atomic product. But you fix these things and they start adding you into more and more sockets. In that particular space, for example, we just actually, one of our partner companies just launched a new AR glass product, but it doesn't have a camera. So there's some that we're boxed out of, but that's easy to solve.
Again, that's why we raise money, initial public offering, to fund the development of some of these new products.
There's a very large consumer electronics company based in Cupertino that doesn't use you. Can you just talk about that? Why would they not?
I'm going to let Fumihide speak to that because he has deeper experience.
We talk to them frequently, especially our shareholders do talk to them frequently. I think that they have a philosophy of they need to build, they like to have all their own chips. They would rather have, I believe, that our technology transferred to them. As a startup, it was very difficult for us to handle their requirements. We kind of stayed away from serving that company for now.
Something that you might do, though, in the future?
I'd love to. Yeah.
Okay. Let's just talk about software tools and field support. On the software side, you focused on shortening customer design cycles with tools and reference stacks. Where are you putting the incremental R&D to sort of make Edge AI deployment on your platform simpler, and can you just talk about how you're just making it easier for customers to come to you and adopt the SPOT platform?
We've got a big software platform that we've developed over the last 10+ years that is stable and growing. We've got a big team building that. Where a lot of the new investment is going is on the AI side. I would say our customers are falling into one of three bins. They're using AI effectively today. They're developing a next product or they're coming to us for help to add AI to a future roadmap product. My goal is to have software tools that make any one of those customers successful and get to market quickly. We have AI software being developed at three different levels. At the highest level, it's actual AI models that our customers can use mostly as templates for what they want to build.
We've got runtime engines, AI runtime, which is kind of the execution engine for the AI model, and then below that, we've got kernels. These are basically the library functions that implement all the layers in the AI model, and we're developing all these things and iterating actually quite quickly, so think our AI team is basically iterating on almost a weekly release schedule, so it's a really fast-moving space. That's really hard to comprehend for a lot of chip people because they think about yearly cadence, not weekly cadence, so I would say post-IPO, that's where a lot of the spending is going to be. It's going to be growing that team, and unlike chips, it's adding people. There's no giant CapEx that's required there. It's mostly just adding people that can go out and build out this roadmap we've got in mind.
Typically in a consumer business, you have some seasonality, but you also have new customers ramping. Are there any sort of how to think about those two factors? Will the new customer ramps just paste over any seasonal effects as we look into next year?
Usually when you serve the consumer market, definitely Q2, Q3 peaks out and then get turned down a little bit for the Christmas and Thanksgiving time after their sales is done. But Ambiq, and we did plan to see Q4 going down based on that. But because of the fact that Edge AI market is growing so fast and new company, new application, new product is coming up, so at least at Ambiq, we are overcoming this seasonality and keep on adding more and more customer and more demand. So my charter to my sales team, they have proven that 2025 we were immune to this seasonality. 2026, I expect them to do the same.
Great. And Scott, you talked. Well, both of you talked for a while that you'd like to eventually license this and develop this as a licensable platform. That's a big lift. You have to invest a lot of money to get there. Can you just spend maybe a few minutes on that? I mean, that seems to be really. You could really take the company to a new level if you can license this.
Yeah. Big opportunity. We think about it in a three to five-year timeline, and we've broken it down into three phases, so if I was going to license, first of all, I would be looking at two primary markets. One is automotive. We've had a lot of outreach from there. Fumihide has deep experience in that set of markets, and we're excited about what we could do there. The other is data center. Clearly a market that cares deeply about power and about energy per operation, and it's one, frankly, that's difficult for a company of our size to go after, so three phases for engaging those markets. First is I've got to prove that I can take our SPOT technology from 40 nanometers and 22 nanometers down to 12 nanometers, which is actually our first FinFET node.
That is ongoing basically for him, for my boss over here, the CEO. That's the gate. Prove that 12 nanometer SPOT works and works effectively, and then we can extrapolate to a 2 nanometer node or something beyond that. Phase two is to go engage in business development activities with these potential customers. Fumihide is basically sending off some of them right now, and we need to, for any IP engagement, there's a hard IP piece of it, meaning it's tied to the technology node, and then there's a soft IP piece, and so we need to talk to these various customers. We need to understand what nodes they're in and then go tailor the IP toward that, so there's going to be power management IP, and there's going to be standard cells and CAD flows that all have to be tuned, not reinvented, but tuned for the specific node.
And so phase two is kind of figuring that out, starting the development, doing some test chips. And then three is the revenue phase. It's where we go and we choose one or two alpha customers and really get after it. So I'm personally, it's going to be one of my top focus areas for the next couple of years.
When you think about how much money it's going to cost you, just the amount of money that you have to invest, the development cost, how do you marry that and be able to fund that alongside of keeping the existing customers happy?
Yeah. I think for at least the next 12 months, there's no difference. We're basically on one path. I can double dip on I'm doing 12 nanometer development. That's critical for the Atomic development. And I can look at investments. So then at least the risk is kind of deferred another year or so. And then after that, your spend starts small, meaning I've got to do test chips. I've got a relatively small team of specialists that's going and developing this. I've got the core of that team right now working for me. And I would say where things get a little more expensive is when you've committed to a customer or two. And you've got a target node in mind. You've got to commercially develop it. But I would say that in that time frame, we're a larger company with larger scale, stronger balance sheet.
But also there's a potential to put some of the cost burden on the target customers. So we believe that there will be strong enough demand that you can execute an early license agreement. You can get them to pay some extra NRE. So there's ways to offset that.
Yeah. Potential customer has already approached us. They're well willing to pay for some of the development we need to do for them. So I think we can cover those costs. I think the benefit is tremendous for those customers as well.
Can you talk about margins a little bit and how the mix is impacting margins? Because we've seen a bit of a step down in margin, even though you have this new product ramping. You have Apollo 5 ramping. But it's not just the new product. There's also new customers coming on. And sometimes you'd think that a new customer, when their initial ramp, they get a better deal. That's my words, not yours. So how do you think of the puts and takes on margins?
So, Q1 actually for 2025, we had a really great quarter with a great margin because we had some manufacturing credit. But Q2, Q3, Q4, we're continuously improving our gross margins. And yes, we did plan to have a little bit of a Q4 gross margin going down because Apollo 5 is in the initial stage. And when you launch a new product and customer wants our product faster than what we planned to, there's definitely the timing of yield and fine-tuning issue that we had. So now that Apollo 5 was ramping up in Q4, we expect the gross margin to be a little bit going to get hit a little bit. But because of the additional new customer with us enabling them with edge AI, they give us a value for that service.
So we believe that even in Q4, I think our margin is going to be pretty stable.
So if I look at revenue next year, whatever it comes out to be, how much of that revenue do you have visibility on? I don't mean you know how many wearables will be sold next year because you don't. But how much visibility do you have in terms of programs? Do you have 100% visibility?
I believe that we have about the next six months of visibility. But some of the design-in activity that we're engaging in right now, we know that we're working with. So we have a fair amount of visibility and we're very optimistic about 2026.
Yeah. I guess ask a different way. So you had the first two customers. You had the third. And you're adding four and five. How important are four and five to revenue in 2026?
Four and five, well, one, two, three, it definitely is a significant portion of the revenue. But yeah, four definitely is going to contribute to our growth. But fifth one is one mixture of various customers. But yeah, I think they're also an important part of our growth.
Great. And then maybe we can, Scott, you had talked about Atomic. You have the test chip back. What are the performance metrics that you're going to give us? Because I think the investment community and us on the sell side, we all want to hear, okay, here's the milestone. And what can you tell us?
Okay. So a couple of different things. First of all, we're going to be rolling out some of the technical details in the coming nine-ish months about Atomic. So CES in a month, we'll have our first drop of technical data. It will focus on the neural processing unit, talk about some of the performance metrics. These will be metrics actually that you guys are pretty comfortable probably hearing from other companies. So I think you'll be able to put it into context. We will talk about the types of neural networks that Atomic will be able to take on. We'll attempt to quantify that for everybody. We will, over the course of the year, talk about broader parts of the system. We'll talk about the memory subsystem in one of our focused releases.
We'll also give some insight as to what SPOT looks like at 12 nanometers, what voltages we're using, what some of the energy numbers are going to look like. And that will all then culminate toward the end of the year, beginning or maybe even CES 2027 in the formal announcement of the product. At CES in a month, the goal is to do as best as we can to really, again, quantify the types of neural networks that we can take on. And we'll talk about a couple of the different markets. There's three markets that we'll be targeting, as I alluded to earlier, smart cameras, AR glasses, and some wearables. And we'll talk about specific AI models in each use case that are aggressive, that are, let's say, out of the realm of possibility for Apollo, but within the realm of reason for Atomic.
Great. And so I wanted to go back in terms of just so people know. So when you're competing for a design, it really is against the mainstream MCU guys. It's against all the MCU guys that we know and love. Is that who your primary competitor is? Or are there any other companies that you are competing against?
Yeah. It's a lot of the mainstream MCU companies. And they sell MCUs that are increasingly becoming AI SoCs. They're adding NPUs. They're sometimes adding Linux-grade cores. And so yeah, these are the types of companies that we're up against. At the high end, we see some competitive overlap with some of the Linux-grade core vendors, some of the high-level OS guys that can run Android, etc. Sometimes the competition there is that they've got a sensor subsystem sitting in the corner of their die. And they say to the customer, "Hey, use this. And you don't need this other chip over here." In that case, we win on power. We're coming in and going, "Look, we're going to give you capabilities you cannot achieve with this other solution." And there's a big disparity in power consumption there.
Let's talk about cash burn. So you're burning a little over $2 million a month, roughly, so far this year. Probably maybe gets a little worse before it gets better. Can you just talk about cash burn? And these are all great ambitions. How do we fund all these ambitions?
Yeah.
Two points. Definitely great ambitions. Definitely we started seeing a lot more demand than we expected, especially when we did a modeling back in June. I think that we expect a certain amount of revenue income is higher than expected. Also the spending definitely is in line with what we planned, but more let's put it this way. We were cautiously optimistic. We were cautious about the spending. Now that the real customer and really demand is coming through, I think we will look at how much we can invest to serve all the customers that are coming to us asking for more products. Let me give you one example. I think we planned one product development in 2026.
Now we're thinking about potentially about three product IP purchases that we will do so that we can have a product launch of three products, one after the other, not waiting for one year at a time. So we believe that we can do that. And definitely, I mean, cash is something that we have to be very cautious about. As you know, we planned to raise $75 million originally. We did $100 million. So we believe that it would increase the revenue opportunities and what we need to do. We should have enough. However, when there is a way to increase shareholders' value and strengthen our balance sheet, we will definitely take a look at to see what additional funding that potentially we can raise.
So you think you can fund all your ambitions. You can get to being profitable without having to come back to the market. Not to say that you, I mean, you might. You might not. But you don't have to.
At this point, yes.
Okay. And then we haven't talked about China, which you de-emphasized China actually the year before last, but it really was this year. So can you talk about that decision and what went into that and sort of how important China or how not important China will be to you going forward?
China did value our new technology. And also that was phase one of our company, how's the battery life? We can extend the battery life for those customers. However, edge AI, I think that the application is more growing outside of China faster than we expected. And you know that the geopolitical risk was extremely high. So once we decided to go public, it was very important that we minimize shareholders' risk. So we kind of tuned down to about 10% maximum so that anything happened to the geopolitical situation, shareholders' value will not go down as some of the policy takes place. But we do still do some business in China.
There are definitely a lot of Edge AI products coming out of China that we want to make sure that we understand what they are and something that potentially be exciting. That's why we keep about 10% of our business in China. Who values our technology and we can add value to their new application, we stay and we enable them. We just want to make sure that shareholders' value is not going to be at risk by staying in China more than 10%.
So not to say you wouldn't tap back into the China market. You just, for now, because you wanted to get through this IPO process, that was a decision you made. But you might go back into China. And I mean, that's a potentially very fertile market for you. So as you get larger, why not tap back into that market, I guess, is my question.
Again, we will if the geopolitical risk has become minimal. But again, we're not the one that can judge that, I think.
But that 10% limit grows, right? So as the company's overall revenue grows, 10% becomes a larger number in absolute terms. And I'll say right now that there's some really cool work coming out of the industrial side, out of the medical side, out of the AR glasses side. So we're firmly committed to building customers there. It's just there's a cap on how much.
Yeah, of course. And then maybe just last thing. So Scott, maybe can you talk about what are the next big frontiers in terms of applications? What's the big Apple? If you can get into that market, what is the big opportunity? Is it autos? Is that it?
Okay, so I've touched on a lot of markets. There's one more market that I want to talk about briefly that we haven't talked about, and that's robotics, so clearly, MCUs have been serving that market for a very, very long time. We're seeing the emergence of humanoid robots. You're seeing hands that need to articulate and sense, and they have latency constraints. I would love to see Atomic get into those use cases. We're having some discussions on that front with a couple of different parties, so I love that. That's not to say that that's the only one. I'm equally excited about AR glasses and some of the medical use cases, but that's another one worth mentioning that I think is capturing a lot of attention as we walk around that floor here.
Got it. Well, we're out of time. So thank you.
Appreciate it.
Thank you so much.