Applied Optoelectronics, Inc. Aktie

Hi. Good morning, good afternoon. Welcome to the Rosenblatt Age of AI Tech Summit. I am Mike Genovese, the cloud and communications equipment analyst here and I'm very, very pleased today to be joined by the Chief Financial Officer and Chief Strategy Officer of AAOI, Applied Optoelectronics, Stefan Murry. Nice to see you, Stefan.

Thanks for having me, Mike. It's great to see you again.

Perfect. And we're going to have a 45-minute fireside chat here go over some key industry and company questions. And for the audience, which I'm sure we'll have a nice one here who, if they want to ask questions, on the upper right-hand of the screen, there's an ask a question button. And if you type something in there, it will come to me, and I will ask the question during the session. And in fact, we've already got some questions as I look. So I'll fold these in. I'll start to fold these in.

So Stefan, again, thanks for joining. Let's start with, I mean, AAOI in the data center revenue today comes from transceivers, right? So you sell transceivers, but I would argue that at your core, you're a laser maker and an indium phosphide laser maker. And that's why companies want to buy your transceivers because you're lasers. I mean, that's one of the key. It's not the only key reason.

I think there's 2. First of all, yes, I think you're right. That's an astute observation that technology-wise, the foundation of the company is in the lasers. I mean that's where we started. We -- for many, many years, we made lasers and didn't make transceivers or other modules. It's only been in the last 15 or 20 years really that we've been making modules and the lasers.

So I think that's a very astute observation that technically, the laser capability goes back further than even the transceiver capacity. And customers do definitely appreciate the fact that we have our own laser fab because it represents a differentiated supply chain for them. It's an additional source of lasers that others don't have access to that gives us greater supply continuity and just basically adds to the availability of lasers, which is currently one of the things that's constraining growth across the industry.

However, the other thing that customers, I think, really appreciate is the automation that we've built and developed for the production of the transceivers themselves. So we have the lasers on one side. On the other hand, we also have this highly automated production process for the transceivers, which, to some extent, I think it's fair to say customers don't necessarily care a great deal about exactly how your production process works.

However, with all the geopolitical tensions that we've seen, it now becomes very important to them that their supply chain for the transceivers has as high an integrity as possible, meaning they want production ideally to be situated in geographical locales that they view as more stable, less likely to be interrupted by any of the stuff that we've seen over the last 5 or 10 years, right, natural disasters, pandemics, government restrictions, like all these kinds of things come into play.

And so I think it's fair to say that most of our customers view the United States as being probably the most secure place in terms of supply chain in which to situate production. And by virtue of the fact that our production process is highly automated, we have a greater likelihood of being able to economically situate production here in the U.S. And that's another thing that's really important to our customers and they're probably growing in importance to our customers relative to where we were a year or 2 ago. So both of those factors, lasers automated transceiver production that can be located in the U.S. Those are the 2 big solid points.

Perfect. And do you find that customers are willing to pay a premium? Or does the price kind of equalize in the market between all the transceivers at 800 at 1.6?

No, they're willing to pay a premium for U.S. production.

Okay. And just because you touched on the geopolitical thing. I mean, we did the conference yesterday and the companies I talked to yesterday like Lumentum and things like that, VIAVI, they don't necessarily think that anything is going to change all that much, but I don't think anybody really knows what's going to happen. Do you have a view on that?

I think there's one person in this country who knows what's going to happen. He's not talking, at least not to me. So I think that's a fair statement. Nobody really knows what's going to happen. I do think there's enough concern about having the supply chain for such a significant component in our AI infrastructure being largely sourced by either from China directly or from companies that are Chinese companies. So I think there is a lot of concern about that.

What the government ultimately does about that situation, I can't plan to have any significant insight into. However, I would posit it to you that the actual -- because nobody knows, including our customers, right, to a large extent, it's the uncertainty that's driving their decision making at this point, right? Like to your point, they don't know whether there should be a ban coming, and there's a real possibility that, that could happen.

Again, I don't know when. I don't know what likelihood it is, but it's not 0 possibility that, that could happen. And against that backdrop, and by the way, if that were to happen, that would absolutely kill our ability to build new data centers in the U.S. if they were immediately banned from importing from the Chinese, right?

So I think against that backdrop, customers are increasingly looking at where their products are produced and by whom and how. And when you start to look at it in that light, AAOI U.S.-based production capacity, currently existing and planned growth becomes very attractive to them and more attractive now than it was perhaps just a couple of weeks ago, even if a ban isn't actually in.

Right. So going back to the lasers. I think what you're making, I believe, right, are CW lasers for silicon photonics transceivers. And when you compare -- I just -- when you compare yourself and your sort of laser making capabilities to, say, Lumentum, Coherent, Broadcom. And we're talking about 100-milliwatt laser, 70-milliwatt lasers. Do you think everybody is about equal? Or does the laser actually become a differentiation? Or are they all about the same?

Well, I mean I think there are certain technical advantages that our lasers have, especially at higher powers when you look at like CPO type lasers, which are in the 300 to perhaps even 400-milliwatt variety. What matters there is the amount of power that you can get into a very narrow line with not to get overly technical, but basically, they want a very tight power spectrum. And our lasers are among the best in that respect.

Now that being said, I think everybody that you mentioned which probably is an exhaustive list of all the companies that actually have 300 or 400-milliwatt lasers and it's not that big a list is my point. But I think all of us are able to make lasers that are good enough for the applications and probably sufficiently capable that it's not a decision-making criteria for our customers.

I think they'll buy anybody that meets the specification. We are relatively smaller and production capacity, especially for the larger chip sizes that are used in those high-power lasers. So that's the main differentiator for us is our production capacity currently is just a lot smaller than Lumentum, certainty Coherent probably and even Broadcom. So that's where we're working to try to improve is by increasing our capacity on the laser production. But from a design standpoint, from a specification standpoint, I think we're as good or better than anybody else, arguably better in some way.

So could you make the 300-milliwatt lasers and ship them today if you are -- if you have the capacity? I mean the technology is there. This is just about --

Absolutely. I mean we are shipping small quantities for evaluation and whatever to customers today. I mean, there's no problem with our ability to produce these products other than the fact that we just don't have enough capacity right now to meet some of the demand that we're seeing or hearing about.

And when you -- I think you said on the earnings, right, you said like we're not going to be the very first -- well, because, right, some companies are shipping for scale out this year, right, before the end of this year. And so that's not you. But then by the end of next year or early '28, there's the next phase for scale up. So is that what you kind of meant by that will be in the second phase, meaning scale up? Or do you also -- or do you mean the second phase of scale up?

Well, I think there'll be a second phase of scale out as well. I think to a large extent, a big reason why AAOI is not a bigger player in the first phase, what's going -- as we point out the early shipments, the stuff that's happening this year and early next year is just simply what I said, we don't have enough capacity to provide a meaningful market share to our potential customers.

So they have to work with somebody who is able to supply product regardless of whether they like us or how good our product is compared to anybody else's if you can't produce enough of it to be commercially useful to them, then they have to go elsewhere, right? As we begin to add capacity, and we have been adding capacity, but specifically into the latter half of next year when we see our capacity increasing meaningfully enough that we can supply all the lasers that we need for the transceivers and have additional capacity for the CPO or NPO or whatever type of laser that we're producing for those newer applications.

That I think that's where we really will start to see our participation in that market start to grow. And that's kind of what I meant. It wasn't specifically tied to scale up versus scale out, but it does sort of coincide to some extent with that scale up opportunity, which is what we're going to see, in my opinion, probably a very sharp increase in the overall market size for these optics. So it's timely that we're going to have additional production capacity right about the time that the market size is going to grow dramatically.

Have you looked at the content difference between scale out and scale up? Because I mean, what I'm seeing is that a scale-out switch might have 16 optical engines, whereas a scale out, right? The scale-up rack might have 5x as many optical engines. It might have 10x as many optical engines, Some optical engine has 8 lasers. So I mean it just seems like the content opportunity and scale, not only is it a bigger market, but people are saying, "Oh, it's double the content per rack." But I'm thinking it might be 10x the content per rack.

No, I mean we tend to think of it as an order of magnitude-ish increase as well. And then the other thing that you hinted at, but just to make it a little more explicit is the size, the die size of those lasers is much larger as well. So when you look at the capacity as an industry that we have to produce those lasers, I mean, you're talking about, let's just call it, roughly an order of magnitude increase in the number of lasers and 5 or 6x larger size per laser.

So when you look at the amount of wafer real estate that needs to be processed and fabricated into lasers, it's an enormous increase related to scale up compared to where we are today, which is why you see a lot of these investments from us and from our competitors in the laser space being made because there is not enough capacity in the industry right now to even come close to meeting the demand from scale up, much less the scale-out and regular opportunities that we see today. So it's a big amount of increasing capacity that's needed and the industry is responding to that along with us.

What size wafers are you using?

So we're a 4-inch right now. All of our equipment, all of our fab equipment that we purchased in the last, I don't know, a year or 2 is all capable of going to 6-inch. So we can turn on a 6-inch production at some point in the future when it's economically viable to do so. But right now, we're not at that point, so we're at 4-inch.

Actually like hearing that. I talked to Wupen about this yesterday from Lumentum. And he said that 4-inch is really just solid and the cost and the economics make sense. I mean the early guys working with 6-inch just the yields are bad and the economics are bad. So I think waiting is probably a good idea.

Yes. No, we would agree with that assessment. I think 6-inch sounds good on paper. And at some point, I think it will be economically viable. But for us, it's a very simple decision. When 6-inch is cheaper when it's basically cheaper to produce lasers at 6-inch that it is a 4-inch full switch to 6-inch, but that is not the case now.

And we don't have a defined time frame set in stone where on this date, we're going to switch over to 6-inch. We're going to evaluate the 4-inch yields as they evolve. We're going to begin work on 6-inch to start to get our yield metrics and laser qualification and things like that done. And then over time, as those 2 factors, as we learn more on both of those fronts when it becomes economically more advantageous to go to 6-inch and we have 6-inch substrate availability, which is another issue, then we'll switch over to 6-inch, but there's no particular time frame that we have for making that decision.

So stepping back from CPO here for a minute. And just getting back to transceivers in 800G. I think you said in the third quarter, you expect about a 5x increase in sales there, which would, I think, get you into kind of the mid-60s and $1 million in revenue for 800G. First of all, how many customers -- just because this is a question that the audience is asking, how many customers are contributing to that 800G revenue?

But more importantly, what changes between 2Q and 3Q? I mean what investments did you make how long ago that you now have the material, the capacity to be. And any more detail on sort of what's different this quarter versus last quarter, so that you can 5x the 800G?

There's 2 major hyperscale customers that are both purchasing meaningful quantities of 800G. And then there are several others that are smaller that are either purchasing or about to start purchasing smaller quantities. But really, the volume is being driven principally by 2 large hyperscale customers.

The question of what changed from 2Q to 3Q is simply the fact that we've added additional capacity. And we've talked about this for at least the last year that we are adding capacity. Now the capacity for this year that's being added is mainly in our Taiwan factory. We have a huge amount of real estate here in the Houston area that we are beginning to build capacity here -- but that really won't start to come online. I mean, the very first increment of it will start to come online later this year.

But really, that's more of a 2027 and even into 2028 thing where we're going to see more of that U.S. capacity coming upward. So right now, the additional capacity that we've added that allows us to go from a few million dollars in 800G sales a quarter to our projection that you mentioned for Q3 and into Q4, that ramp is coming mainly out of our additional capacity in our Taiwan plant. And that's the thing that's changed that's allowed us to grow that revenue.

Okay. I want to hit some of these questions in the audience that always -- aren't the questions that I would ask. So I mean sometimes, but I want to hit these. So to the extent that you can answer that makes sense. What is the capacity utilization or what -- sorry, what is the capacity of a single automated line? And what is the realistic capacity utilization?

We tend to think about our production planning in terms of 100,000 units per month capacity. It's -- that's kind of -- I guess that's what you would call a line. We don't really look at it that way exactly, but it's sort of a unit of production, additional capacity, and that's 100,000 pieces per month.

And as far as what's the realistic utilization, I mean it's going to be pretty high. There's some planned downtime for equipment maintenance and things like that. But the plan is to run those machines pretty much 24/7, again, with some downtime for necessary maintenance and things like that. It's the advantage of having an automated process is it's -- it can be run pretty much all the time, right? It doesn't require sleep and bathroom breaks and things like that.

So are you at about 200,000 units of capacity right now and then the end of the year target of 650,000. Is that correct?

Probably a little more than 200,000. We were 200,000 kind of at the end of last quarter. So we're probably a little more than that now. But we're on that pathway between 200,000 and 650,000.

So help us understand the difference between the capacity you have right this moment, but the actual -- the revenues you could ship this quarter. And then tied into that question, for the $471 million per month target for data center transceivers in the second half of '27. I just want to double-click on that, if that's a -- because you talked about it as a revenue target not a capacity target. And is that a real expectation that you would expect to have 100% of that revenue in that time frame?

That's our current expectation, yes. I mean it's not going to be -- I can say this. I mean, based on our discussions with customers and the demand is not the limiting factor. I mean we can actually do more revenue than that if we had more capacity at that point or sooner, right? So our customers are aggressively trying to work with us to expand our capacity as fast as possible because their demand is growing very, very fast. And it really honestly outstrips our ability to build capacity fast enough.

So yes, that is intended to be a revenue number, and it's basically gated by how fast we can add additional capacity and to some extent, supply chain, the expectations. I mean, there's not an unlimited amount of supply of everything that we need to build those modules, but those are the limiting factors, not demand.

Do you -- I mean, could you even take another data -- large data center transceiver customer, if another hyperscaler came to you and said, "We want your lasers, we want your U.S. production, and we want to buy lots and lots of transceivers for you." Would you bring them in the door or would you have to turn them away?

I mean it would be difficult right now, honestly. We would love to have a couple of other data center customers. I would say there's probably some large data center customers that would be very interested in working with us. But we have to be very careful to not overcommit. We're still relatively small in this industry. And we have to be careful not to overcommit. The last thing we would want to do is not be able to meet our commitments to our existing supplier or existing customers because we got ourselves out over our SKUs with new customer engagement.

So we're talking to them we're trying to find a way to intercept with their demand. Maybe it's CPO, maybe it's some other newer technology that's a little ways out that we can try to build some capacity and intercept for them. But to answer your question directly, today, if another customer came to us and some have had said, we need x number of units. We have to say, sorry, were really sold out through -- at least through second half of next year and beyond. Maybe we have some opportunities to add certain types of things, but not until then.

When we do this a year from now, do you think you'll be saying we're still relatively small in the industry because you're going to grow a lot, but the industry is going to grow a lot. So how do you -- what do you think a year from now, you'll say?

The way I see it with our capacity expansion plans and the way we see the market going, really the best we could hope to achieve is somewhere around 20% and market share. So I don't know if you would call that small necessarily, it's certainly a big improvement from where we are today. But it's not like we're anticipating being sort of the dominant player or whatever. We're still one of the smaller players.

That's big. I don't think any U.S. vendor right now has 20% market share. So I think that that's actually pretty big.

Fair enough. It depends on how you look at it, I mean there are certainly companies out there that are going to be a lot bigger than us, although they may not be U.S. companies, right?

Yes. And then 1.6, right? I mean I think that -- so obviously, one customer has -- there's been an announced $200 million order. I think you'll have some revenue this quarter, and you sort of talked about a $70 million to $80 million in the fourth quarter against that order. And then I think the rest kind of slips into the first half of next year, and then we get more orders from that customer.

So 2 questions, which to just flesh out what I just said, particularly the confidence in delivering and then what the new -- the new -- the next orders could be. But then also I think you probably have qualification trials going on now for other -- at least one other big 1.6T customer. And so when would be the timing where we could hear some news on another 1.6 customer? So I know that's a lot of questions ask all at once.

Well, yes, I won't really give an estimate on the timing for another 1.6 customer. It's hard to predict, but it's not based on our ability or inability to qualify those products based on their schedule on when they need those products, right? Right now, that customer is focusing on 800G getting ramped up on that. They're looking at 1.6, but not for a full qualification at this point. So I'll defer that part of the question.

The first part of your question is, when do we see additional orders. I mean, the orders are going to come regularly, and we're not going to be publishing every order that we get from here on out. The purpose of those first press release was just to let everybody know that, in fact, we are starting to get orders from 1.6T. But after -- as we satisfy those orders, we're anticipating getting continuous stream of orders to utilize the capacity that we're building. So those are just the getting phases, if you will.

Okay. Back to some questions from the audience, since we're talking about capacity, any more new fabs planned to expand the laser supply beyond what is being built right now? Or I know there's another more general question. So what are the plans for capacity expansion?

Yes. We will be adding fab capacity. We do intend to add capacity next year. We've -- we talked somewhat about those plans, not in a lot of detail, and I can't give you a lot more detail than that, but we do have some significant capacity expansion plans, largely around enabling our ability to satisfy some reasonable amount of the demand for CPO or NPO type laser devices, which we think, again, is going to be back half of next year.

So when you actually make the lasers in both Taiwan and then you'll be making them in Texas, or no?

No, the lasers are only made here in the U.S. in our Sugar Land facility. In fact, the building that I'm sitting in right now has a laser fab downstairs. We will be adding -- or we intend to add additional laser production also here in the Houston area, but we don't intend to put laser production overseas.

Right. And then you -- I mean, you surprised me on the last conference call with the amount of space that you have access to in Texas, I mean, 1.6 million square feet. How much of that is for transceivers and how much is for CPO? If that's the right way to think about it. I don't know if that's the right way to think about it.

I don't have an exact answer to that question. I have to look that up. But I don't know that, that's square footage is really the right metric anyway. But definitely more square footage is going to be related to transceiver production. Just because it physically takes more space than an equivalent amount of revenue generating capacity on the laser side, if you will. So -- but there'll be some significant square footage additions related to laser production, again, mostly last part of next year, latter half of next year.

And just substrates. Can you give us just an update on substrates. I mean, obviously, you need to buy them to make these lasers and the availability of them, the pricing of them, any other issues?

Yes. So we feel pretty good about our substrate situation, substrate availability situation through next year through 2027. We're actively working with a number of vendors on how to secure the capacity that we need for 2028 and beyond. Obviously, as our capacity goes up, we're going to need more substrate, more raw material, and we're working with vendors to see if they can scale up their capacity and what that would look like and how we can manage that. But at least for the next year, we're feeling good about our substrate availability.

As far as pricing goes, I mean, pricing has gone up a little bit, but it's not a dramatic impact on our overall margins. Yields for us are -- have room to improve on lasers. So that's one of the areas where if we can improve our yields, that has a very positive impact on our margins. So even if the substrate costs a little more, but we can improve the yield, it's a wash or even an improvement in gross margins. So I think I hit all your questions. Are there --

Because I'm actually -- I'm a little bit surprised in that I'm not out there -- I'm looking around and I'm not actually seeing like an overwhelming amount of demand for LRO, LPO type solutions. I would have expected more of that, but it's just across the board, we're not seeing that much of it. But my question is, given your -- given the size of your company and particularly if there are DSP constraints, I mean can you quickly -- I mean, do you have LRO, LPO products? Or how quickly could you respond to shifting demand if all of a sudden, the customers are asking for something different?

Yes, we do have -- we have LRO products for sure. We have some designs on LPO. I wouldn't say they're productized completely at this point. But yes, we're very customer-driven. So our production process and our ability to switch from one to another from a production standpoint, we've designed that in, so we can switch readily from one type to another.

Obviously, there are supply chain implications, right? If you need DSPs or you need other components to replace the DSPs, we may or may not be able to switched that supply chain on a dime, so to speak. But given our ability to source components, we could switch from one type to another our production process easily. But you're right. I mean customers, by and large, are still demanding full DSP solutions at this point.

Do you -- I mean, as the CFO, are you expecting to have LTAs with customers at some point?

Yes. I do expect that. However, we're being very cautious with those LTAs. I understand the investor interest. And again, put my CFO hat on in a way, I'm like, yes, it would be great to have a 3-year or 4-year commitment from a big customer to buy a product. But it does come with a price, right? I mean it's not -- you're giving up something in that process as well.

And so we want to make sure that we have a good a good deal when it comes to an LTA, one that's not only going to guarantee us long-term profitability, but also one that's not going to necessarily scare away other customers that we would like to have in the future, right? If we get an LTA with one company that's too big and not locks up too much of our capacity that I think there's a concern that other customers might not look at us as an available resource for them.

And we want to keep all of that in mind as we're approaching these LTAs. We do have several on the table, but until we get them into a position where we're confident they're going to be a long-term win for both us and our customers, we need to be appropriately cautious about getting to into that.

Okay. Great. Let's go back to the audience. There's so many here. I have to pick and choose a little bit. Let me just see. All right. I guess is -- I mean, there's a question here about M&A, but I don't really think about it as that as being part of the story. But the question is, how are you thinking about M&A to expand capacity or add additional IP somewhat like the Lumentum playbook?

So that has not been historically, the way that AAOI has grown, we've grown organically pretty much our entire history. So that's not to say that we wouldn't consider it. I don't think we need M&A to improve our technology or capabilities or and adding capacity through M&A is a little bit tricky because a lot of our production processes have been designed in-house.

And so acquiring somebody else, they're not going to have the same process we have. They're not going to have the same automated design capabilities that we have, their transceiver designs are not going to be compatible with our automated production. So I don't see that really as being a viable or necessary path to increasing our capacity.

Okay. Well, let's talk about margins. I mean -- so I think that I mean the target for the end of this year moved down a bit, right, because there's some timing issues on the -- on how much 1.6 we'll have by the end of the year. expedite costs, there's higher component prices. There's substrate increases that you just referenced. So talk about kind of how long some of these expedite issues will persist where you'll be at gross margins by the end of the year? And then how you expect them to trend through the different milestones in '27?

Yes. So I would say our expectation for margins exiting the year is probably going to be in the low to mid-30% range. So probably not all the way up to 35%, but getting somewhere in that range, maybe 32%, 33%, something like that. There's a couple of factors.

You mentioned some of them on the cost side. The other factor for us that's kind of a headwind that was a little bit unexpected was the 100-gig business for us is actually going down over the next couple of quarters, likely because one of our customers was unable to secure enough memory to meet all of their build-out needs and so they're reallocating what memory they can get to higher speed node, so 800 gig, 1.6T type of nodes, meaning they're applying relatively less -- they're deploying relatively less 100-gig links over the next probably, most likely a few quarters.

And so that affects our 100-gig business, which was a pretty high-margin business for us. So that's a short-term headwind as we get 800 gig at 1.6 million growing, that will help pull our margins back up. But in the short term, the next quarter or 2, that 100-gig is a little bit of a headwind for us as well. And then some of the cost pressures that you mentioned are real, although I think those are short-term in nature as well. We're not going to be paying expedite fees forever.

A large part of the expedite fee for us is just the fact that our volume on 1.6T has come up much faster than -- I wouldn't say it's necessarily faster than we expected, but it's faster than our suppliers thought was actually going to happen. I think to some extent, they gave us a little bit of a discount in terms of what they thought we were going to be able to do. And now we're saying, no, we actually do need it and they say, well, we're a little tight on capacity right now, but that will iron itself out over the next few months. And I think there won't be too much of a factor beyond Q4 or so.

So I mean, I guess -- you said low to mid-30s by the end of the year, but where would you think ballpark middle of '27 and then end of '27, they could be?

I mean we're still targeting 40% of as a target long term. And I think that's really achievable. Our thought process has been that we think we can get there by the end of next year, end of 2027. Again, there's some of these short-term headwinds makes that look like maybe it's a little bit more challenging now than we thought it was before. But honestly, I think most of these things are short term, and I think we can still get back to that 40% by the end of next year.

If there's a lot of CPO in '28, could they be higher than 40%?

Yes. Yes. I mean the margins on either ELSFPs or laser diodes, depending on what we end up selling or what combination of those, we end up selling for CPO, margins should be meaningfully above 40%. Certainly, on the laser ship, they'll be above most likely about 60%. Modules will fall somewhere in between. So there's definitely an argument that depending on how fast CPO or NPO or some other XPO type application, depending on the speed at which that actually comes up and contributes as a proportion of our revenue that, that could pull the margins up even from that 40% range. But for now, we want to sort of think more about getting to that 40%. We'll worry about upside beyond that.

How many different customers are you talking to about CPO right now? Are any of them talking about NPO with you? And then -- if we take this to '28, like do we expect to have multiple -- I mean, I guess starting with one would be great, but do you imagine having multiple CPO customers in '28?

So we've got 5 companies that we're talking to about -- some of those are also talking about NPO. So we do have very real discussions going on NPO right now. And do I expect to have more than one customer? Yes. I don't think we're going to target just one. I don't know if we get all the 5, that's probably the stretch because we probably don't have enough capacity to meet everybody's demand there. But certainly, I would expect more than one.

Okay. So can you talk about CapEx? I mean it was a very big number in the second quarter. Is it going to continue to be a very big number quarterly? Or was that -- did you get a lot of MOCVD machines and other things in that order or that is going to buy every quarter.

It's going to continue to be a pretty big number. We're going to continue to make those necessary investments. We said that the back half of the year will be at least as big as the first half of the year. So again, it's not going down. But all of that CapEx is really going into production equipment and machinery as a little bit in R&D, but pretty much the vast majority of it is going into production equipment and machinery and the real estate to support this additional production capacity, which, as we talked about, is really needed by us and by the industry.

So we think those investments are very appropriate, very well placed. Return on those investments is very short. I mean, less than a year. Certainly, right now, most of the investments that we're making are like 9 to 10 months ROI. So as long as we continue to see demand and we have capital availability, I'll make those kind of investments all day long, twice.

So as long as demand continues, then you'll make, I mean, similar type of investments in '27 and then maybe '28.

Why wouldn't we?

Yes. Why wouldn't you? Yes. Okay. So at what point does sort of -- can you organically fund that versus raising money?

Well, I think there's 4 different ways that we look at funding, right? Obviously, right now, we've been leaning on equity. That will become less a factor for us in the future as we start to generate cash from operations, that will help pay some of it. We think that it's appropriate for customers to make some sort of contribution as well. And we've talked about this for a long time, but there are still discussions ongoing about what that could look like.

And then there's various debt structures that we could take on that would be ultimately still us raising money, but it would be less dilutive or in some cases, not dilutive. And then we're also seeing some government subsidy income. I think certainly, one of the conversations that we're having around the potential Chinese ban is, well, look, if the government is serious about trying to ban Chinese competition, and they want to do it sooner rather than later.

There are vehicles in which they could accelerate the growth of U.S. industry by making some appropriate investments in that area. And we've seen that in rare earth. We've seen that in magnets and certain mining, minerals and things like that. So it's not unprecedented for the government to make a commitment where they see a national security risk or a desire to grow U.S. industry in certain areas. So all of those things, I think, could potentially contribute and certainly raising everything on the back of equity for the foreseeable future is not our plan.

Well, I haven't asked you about your cable TV business yet, so I should get some questions in there. I mean, the performance has been very, very strong recently. You've got multiple customers. You're talking about $350 million a year at least in revenues. Does that number increase in future years? Like is this cable TV a growth business beyond $350 million?

Yes, I think so. for another year or 2 probably. I mean the cable investment cycle is cyclical. It will not grow forever. Frankly, no business grows forever, as we all know. But certainly, the cycles there are a little more longer term. So the investment cycle is longer and then the period between investment cycle tends to be longer than we would see in the data center business, for example. So I think it will grow for a few more years and then there'll be a period of leveling off, and then we'll see what the next cycle looks like.

Yes. We just have a few minutes left, a couple of minutes left. I guess I want to ask you what else would you like to emphasize on this call for investors to -- or even people who aren't on the call to read later, watch, rewatch. What did we miss that you want them to understand?

Well, I think what -- and I think I've already kind of gotten this message across, but just to make it explicit right now, our ability to grow revenue and profitability is largely dictated by our ability to add production capacity, right? And to some extent, supply chain. So those have to go hand in hand.

But it's not limited by demand. I mean we're hearing numbers from our customers, really credible numbers from our customers that are much larger than we're likely to be able to produce any time in the near future. And so I think it's important for investors to understand when they look at things like CapEx or they look at the growth rate that we're experiencing and they say, wow, that seems like a really aggressive plan.

It is, in one sense, but in another sense, we're really -- if we don't make those investments, we're going to end up such a small player that we won't be relevant. So we're not over -- we're not -- these investments are making while they may look big and they may look very aggressive. The industry is even more aggressive than we are. And so we're just trying to keep pace with that basically.

Right. Like the biggest risk is that you don't grow fast enough that your customers want to work with you?

Yes. I mean that's kind of -- as I mentioned earlier, when we were talking about CPO, if we would have had more capacity a couple of years ago, we'd probably be a bigger player in CPO right now. And so we don't want to -- now that we have the opportunity to try to add capacity, we need to do it intelligently and reasonably but we do have to do it. But otherwise, we'll just be irrelevant, and that's certainly not where we want to end up.

I mean this is probably the most impactful game changer that we've seen in the optics industry, possibly ever, but certainly since -- in the 30 years I've been in this business, this is the biggest opportunity set that I've seen by far. And that goes all the way back to the dot-com age, which I don't want to bring up that sector because I think there's a lot of differences there, but there certainly were some boom times that we've been through before.

But this is by far and away, a more solid and longer-term growth prospect than I think we've ever seen. And we certainly aim to -- we haven't spent the last 30 years getting ourselves in a position only to miss out on this opportunity. We're really excited about what we're seeing. We're excited about our trajectory here.

Fantastic. It's my pleasure to follow your progress through it and cover this really exciting time. So thanks, Stefan, for doing this. Thanks to everybody in the audience. And we'll see you all down the road.

That's great. Thank you, Mike.

Bye.

Good afternoon. I will be your conference operator. And at this time, I would like to welcome everyone to Applied Optoelectronics' Second Quarter 2026 Earnings Conference Call. [Operator Instructions] And then also please note that this call is being recorded.

I would now like to turn the conference over to Lindsay Savarese, Investor Relations for AOI. Ms. Savarese, you may begin.

Thank you. I'm Lindsay Savarese, Investor Relations for Applied Optoelectronics. I'm pleased to welcome you to AOI's second quarter 2026 financial results conference call.

After the market closed today, AOI issued a press release announcing its second quarter 2026 financial results and provided its outlook for the third quarter of 2026. The release is also available on the company's website at ao-inc.com. This call is being recorded and webcast live. A link to the recording can be found on the Investor Relations section of the AOI website and will be archived for 1 year.

Joining us on today's call is Dr. Thompson Lin, AOI's Founder, Chairman and CEO; and Dr. Stefan Murry, AOI's Chief Financial Officer and Chief Strategy Officer. Thompson will give an overview of AOI's Q2 results, and Stefan will provide financial details and the outlook for the third quarter of 2026. [Operator Instructions]

Before we begin, I would like to remind you to review AOI's safe harbor statement. On today's call, management will make forward-looking statements. These forward-looking statements involve risks and uncertainties as well as assumptions and current expectations, which could cause the company's actual results levels of activity, performance or achievements of the company or its industry to differ materially from those expressed or implied in such forward-looking statements. In some cases, you can identify forward-looking statements by terminology such as believes, forecasts, anticipates, estimates, suggest, intends, predicts, expects, plans, may, should, could, would, will, potential or thinks or by the negative of those terms or other similar expressions that convey uncertainty of future events or outcomes. The company has based these forward-looking statements on its current expectations, assumptions, estimates and projections. While the company believes these expectations, assumptions, estimates and projections are reasonable, such forward-looking statements are only predictions and involve known and unknown risks and uncertainties, many of which are beyond the company's control.

Forward-looking statements also include statements regarding management's beliefs and expectations related to the expansion of the reach of its products into new markets and customer responses to its innovations as well as statements regarding the company's outlook for the third quarter of 2026 and for the full year of 2026.

Except as required by law, AOI assumes no obligation to update these forward-looking statements for any reason after the date of this earnings call to conform these statements to actual results or to changes in the company's expectations. More information about other risks that may impact the company's business are set forth in the Risk Factors section of AOI's reports on file with the SEC, including the company's annual report on Form 10-K and quarterly reports on Form 10-Q. Also, all financial results and other financial measures discussed today are on a non-GAAP basis unless specifically noted otherwise. Non-GAAP financial measures are not intended to be considered in isolation or as a substitute for results prepared in accordance with GAAP. A reconciliation between our GAAP and non-GAAP measures as well as a discussion of why we present non-GAAP financial measures are included in the company's earnings press release that is available on AOI's website.

Before moving to the financial results, I'd like to note that AOI management is attending Rosenblatt's 6th Annual Technology Summit virtually on August 18. This discussion will be webcast live, and a link to the webcast will be available on the Investor Relations section of the AOI website. Lastly, I'd like to note that the date of AOI's third quarter 2026 earnings call currently scheduled for November 5, 2026.

Now I would like to turn the call over to Dr. Thompson Lin, AOI's Founder, Chairman and CEO. Thompson?

Thank you, Lindsay. And thank you for joining our call today. We are pleased to deliver started second quarter results that were in line with or better than our expectations, driven by robust demand in both our data center and CATV business. We generated our fifth consecutive quarter of record revenue, and we achieved an important milestone as we return to non-GAAP profitability in the quarter. Demand to support next-generation AI infrastructure remains so robust that our near-term revenue is funded almost entirely by production capacity and key component availability. Because of this, we continue to anticipate steady [indiscernible] revenue growth this year and continued non-GAAP profitability. During the second quarter, we delivered revenue of $191.9 million, and non-GAAP gross margin of 29.8% in line with our expected guidance range and our non-GAAP income per share of $0.06 coming above our expectations.

Importantly, during the quarter, we saw continued robust customer engagement around our 800G and 1.6 terabit products. In line with our expectation, we saw a strong volume ramp of our 800G product in Q2, which more than doubled sequentially. Looking ahead, focused demand continue to outpace our production capacity to mid-2027. We are working hard to add additional capacity to meet this demand. We continue to believe our 2026 revenue will be around $1.1 billion this year.

With that, I will turn the call over to Stefan to review the details of our Q2 performance and narrow for Q3, Stefan?

Thank you, Thompson. As Thompson mentioned, we are pleased to deliver solid second quarter results that were in line with or better than our expectations. We generated our fifth consecutive quarter of record revenue and we achieved an important milestone as we returned to non-GAAP profitability in the quarter. Our performance continues to be anchored by robust demand across both our data center and CATV businesses validating the power of our dual growth strategy and diversified revenue streams. Demand to support next-generation AI infrastructure remains so robust that our near-term revenue is bound in almost entirely by our production capacity and key component availability. Because of this, we continue to anticipate steady sequential revenue growth and continued non-GAAP profitability this year as more capacity comes online.

In Q2, we delivered revenue of $191.9 million, which was in line with our guidance range of $180 million to $198 million. We recorded non-GAAP gross margin of 29.8%, which was in line with our guidance range of 29% to 30%, and our non-GAAP income per share of $0.06 was above our guidance range of a loss of $0.03 to earnings of $0.03. Notably, we continued to make progress on our key priorities in the second quarter, which included: One, scaling our next-generation data center products, including both our 400G and 800G solutions by expanding our production capacity in a disciplined manner. Two, diversifying our revenue base; and three, strengthening operational execution to improve our margins and position us for long-term profitability. Today, that execution is directly translating into tangible business momentum. During Q2, we continued to see robust customer engagement around our 800G and 1.6 terabit products, particularly as AI-driven data center investments accelerate.

In line with our expectations, we saw a strong volume ramp of our 800G products in the second quarter. 800G revenue of $12.8 million or 11.9% of our total data center revenue increased more than tenfold year-over-year and more than doubled sequentially in Q2. Looking ahead, we expect revenue from our 800G products to grow by nearly 5x sequentially in the third quarter and expect continued strong growth gated by our production capacity and component supply in the fourth quarter.

During the quarter, in line with our expectations, we saw continued strength in our 400G business, 400G revenue of $48.4 million or 45% of our total data center revenue increased more than fourfold year-over-year and 27.4% [Audio gap] in the second quarter. As a reminder, in Q1, we announced that we received our first volume for our 1.6 terabit transceivers from another one of our long-term major hyperscale customers. We also announced that we have received two new volume orders from this customer for our 800G single-mode transceivers. We began delivering these 800G orders in Q2, and we expect full qualification of our first 1.6 terabit product by this customer within the next couple of weeks, followed by shipments of 1.6 terabit beginning later this quarter.

We continue to expect that shipments of these orders will return this customer as a 10%-plus customer for us in Q3. Looking ahead, forecast demand for 800G and 1.6 terabit modules are projected to continue to exceed our production capacity through mid-2027. We are working to add additional capacity and secure necessary key components to meet this demand. During Q2, we continued to make solid progress on our production capacity ramp, particularly for our 800G and 1.6 terabit products. Once complete, we continue to believe that we will have the largest AI-focused data center transceiver production capacity in the United States. As a reminder, our U.S. manufacturing footprint is anchored in Sugar Land, just outside Houston. Through a combination of real estate acquisition and leases, we have expanded our Texas manufacturing footprint significantly to over 1.6 million square feet in the Greater Houston area, and which are in various stages of development.

During the quarter, we made further progress building out our 210,000 square foot facility, which is just a few hundred yards away from our headquarters. We continue to expect to begin initial production in this facility late in the third quarter. As a reminder, this facility will be entirely dedicated to the manufacturing of 800G and 1.6 terabit transceivers. While this will not directly increase our indium phosphide wafer capacity, we plan to move the existing transceiver production from our current headquarters facility to this new building, which will allow expansion of our indium phosphide capacity. The facilities in Pearland and Houston will be built out to expand our production capacity for 800G and 1.6 terabit transceivers. We began construction on these facilities in Pearland recently, and we're proud to have received strong local support to meet our manufacturing needs. We are excited to expand our presence in an area with such a strong workforce, excellent infrastructure, and room to scale our operations and continue to expect these facilities to come online in early 2027.

Currently, our total manufacturing capacity is approaching 200,000 units per month, up from nearly 100,000 units per month of 800G and 1.6 terabit capacity at the end of Q1. Looking ahead, we continue to expect by the end of this year that we will be capable of producing over 650,000 pieces of 800G and 1.6 terabit products per month. By the end of next year, 2027, we continue to expect to grow our production capacity to be able to produce over 930,000 pieces of 800G and 1.6 terabit products per month, with over half of that output coming from Texas. These investments reflect measured scaling of our footprint while aligning with strong and growing customer demand and qualification progress across both 800G and 1.6 terabit products.

It's important to note, as a reminder, our 800G and 1.6 terabit products can be manufactured on the same production line with the same process. While our 1.6 terabit product will require a different final testing, our 800G automated manufacturing lines have been developed with an architecture that will allow us to support future high-speed products as customer demand materializes and evolves over time. Our automated manufacturing lines are engineered to scale efficiently from 800G to 1.6T with minimal incremental investment. The structural flexibility provides a powerful dual advantage. It accelerates our time to market for AI customers while expanding our long-term margin potential.

Looking ahead, we continue to believe that our 800G products will drive our near-term data center ramp, followed by our 1.6 terabit products, which are on track to begin to contribute to our overall revenue later this year with the bigger ramp beginning in 2027. At OFC, we also discussed our plans to increase our manufacturing capacity for our external light source or ELSSP. That's for co-packaged optics or CPO. This utilizes the ultra narrow line with high-power laser that we announced late last year. We have very limited production of these modules now, but we anticipate ramping production later this year and into 2027, ultimately culminating at about 400,000 pieces per month in 2028.

We believe our in-house laser capabilities continue to be a strategic advantage for the company. As we have mentioned before, we've been manufacturing lasers internally for many years. This has allowed us to avoid some of the shortages that have affected others in the industry. As we continue to expand our footprint in Texas, our in-house laser manufacturing positions us well to support both near-term customer needs and longer-term growth. We believe that in the future, CPO will continue to drive increased demand for high-power lasers, and we plan to continue to expand our laser manufacturing capacity in Texas in order to accommodate these future growth drivers. Notably, our expansion planning has been underway for several years. We have already secured orders for long lead equipment and are partnering closely with vendors on delivery. Crucially, our reliance on proprietary in-house developed machinery heavily insulates us from the broader equipment supply bottlenecks in the industry. There are exceptions to this, of course. But overall, we feel that our in-house developed technologies give us an edge in ensuring reliable supply of production equipment.

During the quarter, direct tariffs had a $1.9 million impact on our income statement. With the overturn of the IEEPA tariffs, we have received a refund of approximately $5.7 million. We are still evaluating the potential impact of the new tariffs recently announced in the U.S., but at the present time, we don't expect any material change from tariffs as a result of this announcement.

Turning to our second quarter results. Our total revenue was a record $191.9 million, which increased 86% year-over-year and increased 27% sequentially off a strong Q1 and was in line with our guidance range of $180 million to $198 million. During the second quarter, 56% of revenue was from data center products, 42% was from CATV product and the remaining 2% was from FTTH, telecom and other.

In our data center business, Q2 revenue of $107.7 million increased 140.4% year-over-year and 32.3% sequentially. Sales of our 100G products increased 31.3% year-over-year, while sales for our 400G products increased more than fourfold year-over-year and sales of our 800G products increased more than tenfold year-over-year. In the second quarter, 38.3% of data center revenue was from 100G products, 45% was from 200G and 400G transceiver products. 11.9% was from 800G transceiver products, and 4.4% was from 10G and 40G transceiver products.

We currently expect to see a decline in 100G business in Q3 and due to one of our customers' inability to source sufficient 100G switches to meet their initial forecast. We believe that the shortage of switches is related to memory shortage and expect that 100G weakness will persist until the memory supply recovers. Even with this temporary weakness of 100G, we continue to believe that by mid-2027, 100G and 400G revenue will be approximately $90 million. 800G revenue will be approximately $217 million and 1.6 terabit revenue will be approximately $164 million monthly. In total, this is about $471 million per month of data center transceiver revenue.

In our CATV business, we saw record CATV revenue of $80.6 million, which was up 43.8% year-over-year and 20.6% sequentially and was slightly above our expectations of $75 million and $80 million. Similar to the last couple of quarters, we shipped a significant quantity of 1.8 gigahertz amplifiers to our largest CATV customer in Q2. We also continued to see momentum with a newer set of MSO customers that we have talked about on our prior few earnings calls. We continue to see a broad-based appeal of our amplifiers and Quantum Link software across our potential customer base. During the quarter, we announced that Mediacom selected AOI as the primary vendor to accelerate its DOCSIS 4.0 network upgrades, driving multi-operator commercial adoption of AOI's next-generation 1.8 gigahertz Quanta bandwidth smart amplifiers and software solutions. We're excited to partner with Mediacom to deliver more reliable service while lowering operational costs.

Looking ahead to Q3, we expect our CATV revenue will be between $100 million and $110 million. Looking further ahead, we continue to expect to generate over $325 million annually in CATV. While the vast majority of our CATV revenue expectations for this year are related to our amplifiers, we do anticipate that we will generate some revenue from our software solutions this year.

For the second quarter, our top 10 customers represented 99% of revenue compared to 98% of revenue in Q2 of last year. We had three greater than 10% customers, one in the CATV market, which contributed 42% of total revenue; and two in the data center market, which contributed 26% and 24% of total revenue, respectively. In Q2, we generated non-GAAP gross margin of 29.8%, which was in line with our guidance range of 29% to 30% when compared to 29.2% in Q1 2026, and 30.4% in Q2 2025.

As we discussed on our last quarterly earnings call, while we do expect continued gradual improvement in gross margins, we continue to expect that the revenue mix and data center in the short term will be a slight headwind. We remain committed to our long-term objective of returning non-GAAP gross margins to around 40% and believe that this goal is achievable as our mix shifts towards higher-margin products and as we capture additional efficiencies across our operations. The revenue figures presented above are net of a contra revenue amount due to the accounting for warrants provided to customers. As a reminder, this amounts to approximately 2.5% of revenue derived from certain customers to whom AOI has provided warrants in exchange for future revenue. In Q2, the amount of this contra revenue was $1.2 million.

Total non-GAAP operating expenses in the second quarter were $67.6 million or 35% of revenue, which compared to $42.1 million or 41% of revenue in Q2 of the prior year. Our OpEx this quarter was higher than expected due mainly to higher shipping costs associated with the rapid ramp in CATV revenue in the quarter, combined with higher-than-expected R&D spending as we have been asked by customers to qualify new 800G and 1.6T products in the quarter. We believe that R&D spending will continue to be elevated. However, we do not expect additional shipping costs to recur in Q3 or subsequent quarters.

Looking ahead, we expect non-GAAP operating expenses to be in the range of $70 million to $80 million per quarter. Non-GAAP operating loss in the second quarter was $10.3 million compared to an operating loss of $10.8 million in Q2 of the prior year. GAAP net loss for Q2 was $22.8 million or a loss of $0.28 per basic share, compared with a GAAP net loss of $9.1 million or a loss of $0.16 per basic share in Q2 of the prior year.

On a non-GAAP basis, net income for Q2 was $5.5 million or $0.06 per diluted share, which was above our guidance range of a loss of $2.5 million to income of $2.8 million or non-GAAP income per share in the range of a loss of $0.03 to earnings of $0.03. This was largely due to foreign tax benefits and modest government subsidy income, which we expect to continue in subsequent quarters. This compares to a non-GAAP net loss of $8.8 million or $0.16 per share in Q2 of the prior year. The weighted average fully diluted shares outstanding used for computing the earnings per share in Q2 were $88.1 million.

Turning now to the balance sheet. We ended the second quarter with $508.8 million in total cash, cash equivalents, short-term investments and restricted cash. This compares with $449.4 million at the end of the first quarter of 2026. We ended the second quarter with total debt, excluding convertible debt of $92.8 million, which compared to $77 million at the end of last quarter. As of June 30, we had $278.8 million in inventory, which compared to $206.2 million at the end of Q1. The increase in inventory is primarily due to increased inventory of raw materials for near-term production as we ramp capacity.

As we disclosed in May, we initiated a new at-the-market offering. To date, we have raised $538.8 million, net of commissions and fees under this new program. We intend to use these proceeds to continue to make investments in the business, including new equipment and machinery for production and research and development use. We made a total of $565.5 million in capital investments in the second quarter, including $280 million in prepayments on equipment we have on order. These expenditures are mainly for manufacturing capacity expansion for our 400G, 800G and 1.6 terabit transceiver product. We expect CapEx intensity in the second half of the year will be higher than in the first half as we prepare for increased 400G, 800G and 1.6 terabit data center production. We expect to finance these investments through a combination of cash on hand, cash generated from operations and some equity sales, along with additional debt.

Looking ahead, we believe we are uniquely positioned to capture two distinct growth engines, the rapid AI-driven demand acceleration in our data center business. alongside a robust runway in our CATV business. Our current capital investments are designed to scale our advanced manufacturing footprint, structurally lower our long-term production costs and enable our path towards sustained profitability.

Moving now to our Q3 outlook. We expect Q3 revenue to be between $255 million and $290 million, representing 130% year-over-year growth at the midpoint. We expect non-GAAP gross margin to be in the range of 29% to 30.5%. Non-GAAP net income is expected to be in the range of $10.1 million to $24 million. And non-GAAP earnings per share between $0.11 per share and $0.26 per share using a weighted average diluted share count of approximately 92.8 million shares.

Looking more broadly at 2026. We believe our 2026 revenue will be around $1.1 billion. As we have discussed previously, this revenue level is limited by our production capacity and supply chain, not market demand, which we believe is much larger.

With that, I will turn it back over to the operator for the Q&A session. Operator?

[Operator Instructions] And our first question today will come from George Notter with Wolfe Research.

I wanted to ask about all the stuff that's been in the news of late around Chinese transceivers and the potential for U.S. ban on those transceivers being shipped into the U.S. Could you just talk about your perspective on that? What are you seeing hearing? How might that affect AOI? Does that change anything in terms of your capacity planning? Does it change anything with your conversations with customers, any insights would be great.

You know it's a little early to say that report that came out a couple of days ago was obviously somewhat speculative in terms of the fact that this ban or whatever it ultimately turns out to be isn't in place yet, and the details aren't really out there. I think U.S. manufacturing presence has been a very important, probably the most important element of our appeal to our customers. And clearly, that announcement tightens that appeal. As we said in our prepared remarks earlier, we believe that we are and expect to remain the largest manufacturer domestically of optical transceivers for AI. So certainly, anything that would heighten interest in that is good for us. But it's hard to point to any specific ramifications at this point since it's still kind of early.

In the -- this is really not news at all as this is discussion for quite a while. So as I mentioned, we have been working very good with the customer for the long-term agreement, especially AOI making [indiscernible], and we are maybe the only one company commit to really invest ever in U.S. for U.S. manufacturer, not only later, including the transceiver. So with some early, I would say, early feedback, the customer more aggressive, will give us much more share, okay, especially for U.S. manufacturing, yes. But how serious or how big it is, we were not because, as I said, do I now [indiscernible] already from now to Q2 next year. The best we can do is maybe we can be more aggressive adding more capacity maybe from Q3 next year, especially for the -- our U.S. manufacturer.

Got it. Okay. And then just as a follow-up. I was just curious about the ramp in 800 gig. It looks like the growth in the business right now is coming from 100, 200 and 400 just based on your comments about the mix of transceivers in the quarter. It seems like at this point, the 800 gig has got to be the driver on the growth in data center. And so kind of wondering like exactly where you guys are. Is all that tooling fully installed and ready. Are those laser datacom chips built and sitting on the shelf? Do you need to get qualifications on any of this any more you can tell us on the ramp and the readiness would be great.

No. I mean as we said in our prepared remarks that our ability to deliver revenue in general and specifically when it comes to 800G products, is limited by our production capacity right now. So if we could produce more, we could ship more right now. So to answer your question directly, no, not all the manufacturing capacity that we have or expect to have is online at this point. In fact, we're going to continue to add more and more capacity really most likely through next year, but certainly into next year. We've only begun. As we mentioned in our prepared remarks, we have 1.6 million square feet of manufacturing space here in Houston. A year ago at this time, we had about 65,000 square feet. So clearly, all of that manufacturing space has not been built out, doesn't have equipment in it yet. So there's tremendous room for expansion than we plan to expand. However, over the next couple of quarters, the first increment of that manufacturing does start to come online, which will increase our manufacturing capacity from roughly pieces a month as we mentioned earlier, to about 650,000 pieces a month by the end of the year. That's all 800G and 1.6, that's not -- and most of that is going to be 800G given the demand that we're seeing. So that's all incremental new capacity that doesn't exist today and then growing from there. So no, there's significant room for expansion, and we're only kind of getting started on the energy ramp now.

But by the way, I think, I want to emphasize one thing. Right now, the Q3, you can see compared to Q2 average we had 40% growth at [indiscernible] [ 35% ] growth, but the growth is mainly from 800G. [indiscernible] mentioned, because the memory issue for the switch, okay, actually, the 100G were decreased by $20 million to $25 million. Otherwise, our Q3 revenue will be [indiscernible] -- the cost should be more than 50%, as I mentioned. So -- but this is certainly that we believe the demand will come back to normal within a few months, okay, either by end of this year or early next year. And don't forget, and we mentioned by June, July, the [indiscernible] revenue will increase from today's number to $471 million per month, there's huge growth and our growth actually from 800G and 1.6T and 100G will go down. And all this is because the additional capacity we are building not only in Asia, most of the increased capacity will be in the U.S. But even though, let me say that, like I keep emphasize, that is not good enough for customer demand. The cost demand is 20% to 40% higher, all right? And actually, we are getting this kind of demand from several because almost every week, every month to speed up our delivery schedule. And we're doing everything we get, the other working vehicles with our key supplier because as you know, as in the PSP, PIA. Good news, they are making on laser other less spikes battling right now for the transceiver business. This is not the only one, okay? That's what we're working with very severe suppliers to increase the volume in the next few quarters to mitten our demand.

And our next question will come from Simon Leopold with Raymond James.

I appreciate you've given us a lot of guidance commentary and some quick arithmetic suggests that in the fourth quarter, you're anticipating the combination of 800 gig and 1.6T revenue in the neighborhood of $330 million. I want to make sure that I'm thinking about this correctly, first? And then I've got a follow-up.

Yes, that's about right, directionally.

Great. And then I recall in the spring that you were talking about the 1.6T ramp and having a commitment for $200 million through an order. And I'm trying to get a better sense of how to think about the timing of that particular project in that it sounds like it may start in the fourth quarter of this year, but maybe the majority of it is a 2027 event. How should we think about that $200 million order you had talked about for 1.6T in the past?

Actually, we'll start deliveries on that probably very late in the third quarter and then ramp into the fourth quarter. I think the bulk of it should get delivered in the fourth quarter. And then there may be a tail into the first quarter. But the important part about that is that's just the first -- the beginning of what we expect to be significant orders from this customer for 1.6T for the foreseeable future. So I wouldn't get too wrapped up on that particular order. That's just the very beginning of it.

Great. And then I want to ask a different China-related question. Apart from the potential regulatory issues, there's been -- we've been getting a lot of questions about the suggestion that there will be new manufacturing of lasers coming out of China. And just trying to get a sense of how you're thinking about that potential? And if that were to occur, if new Chinese manufacturing came online to make lasers, what could that possibly mean to your business?

Let me answer the question, okay. I think the investor may be for onetime how big is laser market. Let me say that, right now, for the CPO, okay, using [indiscernible] is [ 300 ]milliwatts compared to 70 milliwatts use at energy transceiver and 100 million used for the 1.6 terabit transceiver. But people don't understand not only the power is made several times bigger. But the size, okay, the size is about 6x or even -- or even more, but it is lower. So overall, you're talking about to meet the demand just for the -- I would say, the Phase 1 ESP is only the [indiscernible] [ 1310-nanometer, ] okay, for [ warrants. ] But I think the next generation of ESP will be using DWDM. So because warrants so tight, I think the year loss will be easily, okay, 40% to 50% compared to [ 1310 ] brands, okay? So that means to meet the demand for the CPO market, the operator market needs to be 8,000 to 10,000 bigger compared to today. This is now the AOI, Luminaire, [indiscernible], everybody will invest huge money to meet demand in the next few years because for laser from today by older equipment, will take minimum 21 months to 24 months, start manufacturing in high volume, okay? That's on [indiscernible]. [indiscernible] be the semiconductor process, build a long-term mix of the equipment visit. So I think, yes, there will be some new products in China, we are not surprised because it's a market need, but most of them still working maybe 70-milliwatt very few can really do [indiscernible]. But for 300 million [indiscernible], especially [indiscernible], we don't see that not in the next 2, 3 years, especially the demand is so big, okay. Even combined AOI [indiscernible] altogether, it's still very tough to meeting the customer demand in the next few years. And we are doing everything to speed out the process, to expand our capacity. So I think this, for me, in no effect at all because the demand is much bigger than the worldwide capacity even including oil company in Taiwan, China or other countries like Japan.

[Operator Instructions] Our next question will come from Ryan Koontz with Needham & Company.

Maybe just following up on the question about laser supply and thinking about your own constraints there for iindium phosphide, how are you guys feeling about substrates and other raw materials that you need to ramp? And is that a current bottleneck for your products and which products are the most challenging for you to ramp at the moment?

No. I mean, as Thompson mentioned on our last earnings call, and then the situation hasn't changed. I mean we've secured supply out into next year. So we're not currently limited by substrate capacity. And we've had a lot of discussions with substrate suppliers going back into last year and continuing even until very recently. We feel pretty good about the substrate in the by situation. I would say it's incrementally better than it was last earnings call and prior to that. So things, I think, are getting somewhat better, but we're feeling pretty good at least as far as we can see into the future for the substrate supply.

Yes. I think the specialty right now, AOI, we just moved into a 4-inch substrate volume the manufacturer. As I mentioned, we already had 2 suppliers in Europe, 2 supplier in Japan, cross-supply in China. So right now, we are very aggressive to have some kind of partnership with 2, 3 suppliers, even maybe the possible, I would say, potential joint venture, okay? Because as you can see, how much the capacity AI will increase in the next few years. Let me say that it's much, much more than a fact of 10, okay, to spare for the CPO laser market. So I would say right now, we inventory will enough supply until the end of next year. But what we are looking for is the -- I would say, the [indiscernible] we will need in, I would say, 2028, 2029. That's why we are very careful and very aggressive to working with all the suppliers for the expansion.

Great. Really helpful. And then maybe following up on George's question earlier about 1.6T. How are you feeling about your broad market traction with that product? Obviously, we're expecting a pretty big uptick in demand with Tomahawk 6 at the end of the year. How are you feeling about your traction with other customers besides the one order you have in hand now?

No, I think we have pretty broad-based interest among customers. I think as Tom mentioned, and we've talked about it pretty extensively, we're still in the process of adding capacity until we have sufficient capacity to service multiple customers, we have to be careful about taking too many too many orders. So we're trying to balance the capacity additions against the customer demand. As Thompson mentioned, the customer demand is a lot bigger than what we can provide, especially in the short term. As we get further out, then our capacity expands, and we have a little more breathing room, if you will, with respect to new comer orders, but we're trying to be careful in what we -- we definitely don't want to overpromise what we can actually deliver. So we're being appropriately circumspect on that.

And let me say, we always say at as of today, in AOI will be the fourth supply qualified by 1 big hyperscale data center customer for 1 positive terabit transceiver. And right now, I think the [indiscernible] will finish most of the qualification on the last stage, we should be at should be finished within maybe say 2, 3 weeks. So we can start delivered, I would say, by end of this quarter. And as you know, right now, we have more than $200 million already enhanced. So I would say right now, our target is to finish all the order by sometime Q2. How much we're going to deliver in Q4 is always said, yes, a good news, we have already. At the same time, we are working very close without the DSP and TI supplier. But even though we still believe we can deliver more than, I would say, [ $70 million ] of revenue in Q4 for 1.6T transceiver for sure, customer one. So right now, we are doing everything to speed up. So I think one of the big concern in Q4 for 1.6 terabit transceiver is the material supply. I think our measure capacity should be already was within our. But the over demand is very big. Right now, I think all based on the customer feedback. The volume [indiscernible] is more than 500,000 transceivers per month by end of next year. So then you talked about how much money per month, I would say, $300 million to $350 million. But as I said, we are working with customers based on this schedule, we don't want to overpromise, and we do want to be careful [indiscernible] is very important, especially more expansion in U.S., it comes. But I just say the demand of big, especially right now, not only, I would say, the overall industry demand is so big because AI, the other for sure is a U.S. manufacturer. That's a very important factor for the customer.

And our next question will come from Michael Genovese with Rosenblatt.

Guide for the full year is on track and the milestones for next year seem to be on track, but there has been a push out this year into 4Q. So could you just give us a little bit more color on the challenges of ramping up capacity that were like different than what you expected three months ago and that clearly are going to keep getting better as we go forward. But what are some of these specific challenges where 800G, for instance, wasn't quite as big as you thought it would be in 2Q.

Right now, let me say that as I mentioned, because the memory issues, okay? So in Q3, I say we lose about $20 million to $25 million of revenue for 100G single-mode transceiver. In Q4, I think [indiscernible] bear capacity, we able to deliver, I would say, more than $500 million revenue, okay? So you can see how big our goals will be just like 60% of growth from Q3 to Q4. But right now, the big change, as I say, is a DSP and TIA, okay, for 800G, 1.6T transceiver. And that's why we are working very close with the supply be, let's say, 1.6T. I've seen the whole supply chain is very tight. But good news is that in the supplier put for priority and a long policy. So we will not have very caused discussion with all the key supply almost every week or twice a week. It's much better than other competitors. Because the other competitors, their #1 issue is not the DSPs [indiscernible], and we don't have this problem, okay? I making our [indiscernible]. And that's why the customer come to AOI, especially with aggressive expansion plan in the next few years. indi ESP right now, we already have about 5 customers. They will come out some kind of very aggressive demand for the next 3, 4 years. And that's why our CapEx is so big because as I said, if I promise any customer or the last like Q3, Q4, 2028, I need to start to spend money by the agreement, by the build and build [indiscernible], that's the reason our CapEx is increasing so fast.

Okay. Great. That was great color. I guess my next question would be, given that 1.6 will be a lot more in the mix in the fourth quarter, do we still expect to exit the year in the mid-30s of the gross margin?

Gross margin will be -- we're still better. But because we need to pay some [indiscernible] for some key supplier, I would say, I don't know, 32%, 33%. The most important is how much is 1.6T because it's high gross margin product. The more we deliver for 1.6T, the high gross margin will be. For sure by Q3, Q4 next year, one was to deliver the CPU or ESP module for -- to different customers, the gross margin will be even higher because gross margin for this is about 55% to 65%. For ESP, the gross margin should be more than 50%, but that's more like Q3 next year. But in the short term, of course, improvement -- for sure, the most important factor is the percentage of 1.6 terabit transceiver because gross margin is very good. So that's why I said, it will start to improve on ramping [indiscernible] 1.6-terabit transceiver in Q3, especially Q1 next year. Let's say, right now, even customers want even in Q4, but right now, what we can commit maybe, I would say, $70 million to $80 million in Q4, not because of capacity, because of, I would say, material constraint. Q4 should be much -- Q1 next year will be much better, okay. [indiscernible] price for Q1 revenue. [indiscernible] overall revenue for 1.6 terabit transceiver revenue. Q1 next year will be double Q4 or more than double as our [indiscernible] right now.

Okay. Final question for me to just kind of follow up on what you said about CPO. I don't think that all the investors kind of know exactly that you're in the CPO market or necessarily have high expectations for you guys in CPO. So any kind of additional update in terms of number of customers that you're talking to and status that you're at with that program, I think, would be helpful. And that's it for me.

[indiscernible] is a laser company since day 1, okay? Maybe compared to other suppliers in U.S., AOI is a pure laser suppliers stay since 1997. This is some major technology, the core technology. We have been working very well with one -- [indiscernible] customer. But if you took on to be a high-volume manufacturer or more like the next Q3 next years. As you can see right now, we are adding more CBD. EP in stepper, everything in Houston. We even have -- we had second fab. And the size of current [indiscernible] will increase -- it is, I would say, the overall capacity will increase by almost 300% by Q3 next year. But that's not enough. So that's why we we are building a [indiscernible] of the second fat in Houston area. The size will be about 4x of the current facility, just give you some rough idea how aggressive [indiscernible]. But let me say even though they're not good enough for the customer demand in the next few years. So we are still working very hard to expand our laser including the manufacturer ESP module based on customer.

Mike, if I can just interject there. I mean when we've talked to several of the major CPO customers, they love our laser. We just can't make enough of them to be involved in their current first-generation deployments because there's just not enough capacity. We have to prioritize our ability to make lasers for our own transceivers first. And then as we expand the fab, like Thompson mentioned, then we'll have more capacity for the EL SP for the CPO-type laser. So it's not -- it has nothing to do with customer engagement. It has nothing to do with performance of the laser or not having the design, all of those things very good. In fact, I would argue our high-power narrow-line with laser is the best in the industry, the best in the world right now. but we just can't make enough of it and still manufacture enough lasers for our transceivers, which has to be the priority. So again, it's all about -- same story as we had on the transceiver side of things, it's all about our ability to add capacity. And Thompson just outlined kind of our plans on that.

As in investor maybe can understand has been working on high-power laser say, 6, 7 years ago for LiDAR. The [indiscernible] deman study is very even, even much higher than CPO laser. And that's why it's not tough for you to get into CPU as market. So right now, I would say for certain 10, CPU list and ESP, we -- stable customers qualify AOI. But the focus, as I said, is a [indiscernible]. And right now, this is what customers really want to focus because they'll be very changed. For DDM, CPO laser and the CPO module, like we call it ESP I would believe only due come U.S. can do this because it's very high spec, a lot of change. Maybe some company can do it, but I don't believe the year performance will be as good as like AOI to mention [indiscernible], okay? So I think that's very important. And that's the additional [indiscernible] capacity.

Let me sneak in one more question. I'm sorry to interrupt, but do you think -- how far away do you think the Chinese are from having 350-milliwatt lasers? And do you think they'll ever have them, or how many years away do you think they might be from having CPO lasers out of China?

You are talking about [indiscernible] year and not quality issue or performers issue, I would say at least 2, 3 years or even longer.

And this will conclude our question-and-answer session. I'd like to turn the call back over to you, Dr. Thompson Lin for any closing remarks.

Okay. And thank you for joining us today. As always, we want to extend a thank you to our investors, customers and employees for your continued support. It is an exciting time for our industry and for AOI. We continue to believe the fundamental driver of long-term demand for our business remain robust, and we are unique position to drive value for our risk opportunity. We look forward to seeing many of you at upcoming investor conference. Thank you.

The conference has now concluded. Thank you for attending today's presentation. You may now disconnect your lines at this time.

Good afternoon. I will be your conference operator. At this time, I would like to welcome everyone to Applied Optoelectronics First Quarter 2026 Earnings Conference Call. [Operator Instructions]. Please note that this call is being recorded.

I will now turn the call over to Lindsay Savarese, Investor Relations for AOI. Ms. Savarese, you may begin.

Thank you. I'm Lindsay Savarese, Investor Relations for Applied Optoelectronics. I'm pleased to welcome you to AOI's First Quarter 2026 Financial Results Conference Call. After the market closed today, AOI issued a press release announcing its first quarter 2026 financial results and provided its outlook for the second quarter of 2026. The release is also available on the company's website at ao-inc.com.

This call is being recorded and webcast live. A link to the recording can be found on the Investor Relations section of the AOI website and will be archived for 1 year.

Joining us on today's call is Dr. Thompson Lin, AOI's Founder, Chairman and CEO; and Dr. Stefan Murry, AOI's Chief Financial Officer; and Chief Strategy Officer. Thompson will give an overview of AOI's Q1 results, and Stefan will provide financial details and the outlook for the second quarter of 2026. A question-and-answer session will follow our prepared remarks.

Before we begin, I would like to remind you to review AOI's safe harbor statement. On today's call, management will make forward-looking statements. These forward-looking statements involve risks and uncertainties as well as assumptions and current expectations, which could cause the company's actual results, levels of activity, performance or achievements of the company or its industry to differ materially from those expressed or implied in such forward-looking statements.

In some cases, you can identify forward-looking statements by terminologies such as beliefs, forecasts, anticipates, estimates, suggest, intends, predicts, expects, plans, may, should, could, would, will, potential or things or by the negative of those terms or other similar expressions that convey uncertainty of future events or outcomes. The company has based these forward-looking statements on its current expectations, assumptions, estimates and projections.

While the company believes these expectations, assumptions, estimates and projections are reasonable. Such forward-looking statements are only predictions and involve known and unknown risks and uncertainties, many of which are beyond the company's control. Forward-looking statements also include statements regarding management's beliefs and expectations related to the expansion of the reach of its products into new markets and customer responses to its innovation, as well as statements regarding the company's outlook for the second quarter of 2026 and for the full year of 2026.

Except as required by law, AOI assumes no obligation to update these forward-looking statements for any reason after the date of this earnings call to conform these statements to actual results or to changes in the company's expectations. More information about other risks that may impact the company's business are set forth in the Risk Factors section of AOI's reports on file with the SEC. Including the company's annual report on Form 10-K and quarterly reports on Form 10-Q.

Also, all financial results and other financial measures discussed today are on a non-GAAP basis, unless specifically noted otherwise. Non-GAAP financial measures are not intended to be considered in isolation or as a substitute for results prepared in accordance with GAAP. A reconciliation between our GAAP and non-GAAP measures as well as a discussion of why we present non-GAAP financial measures are included in the company's earnings press release that is available on AOI's website.

Before moving to the financial results, I'd like to note that AOI management is attending the 21st Annual Needham Technology Media and Consumer Conference on Wednesday, May 13. This discussion will be webcast live and a link to the webcast will be available on the Investor Relations section of the AOI website. Lastly, I'd like to note that the date of AOI's Second Quarter 2026 earnings call is currently scheduled for August 6, 2026.

Now I would like to turn the call over to Dr. Thompson Lin, AOI's Founder, Chairman and CEO. Thompson?

Thank you, Lindsay, and thank you for joining our call today. We are pleased to deliver solid fourth quarter results, still in line with our expectations, driven by robust demand in both our data center and CATV business, we generated our fourth consecutive quarter of record revenue as we executed well to expand our manufacturing capacity. We continue to see accelerating customer demand needed to support the next wave of AI infrastructure deployments, and we anticipate a sequential revenue growth throughout the year with a significantly large rent expected started in Q3 as additional capacity come online.

During the fourth quarter, we delivered revenue of $151.1 million non-GAAP gross margin of 29.2% and non-GAAP loss per share of $0.07, or in line with anticipated guidance range. Importantly, during the quarter, we saw and continue to see strong customer engagement around our 800T and 1.6T products, particularly as AI-driven data center investment is late.

We completed our fourth volume shipment of our 800T single mall transceiver to one of our large hyperscale customers in Q1, and we continue to anticipate strong volume rate of our 800 product started in Q2. During the fourth quarter, we announced that we received our fourth value order for our 1.6T transceiver from another long-term major hyperscale customer along with 2 new value orders from this customer for our 800T single-mode transceivers. Looking ahead, focused demand continued to outpace our production capacity throughout mid-2027.

We are working hard to add additional capacity to meet this demand. Based on new demand and our anticipated capacity ramp, we now believe our 2026 revenue will exceed $1.1 billion. and we now expect to generate more than $140 million in non-GAAP operating income in this year.

With that, I will turn the call over to Stefan to review the details of our Q1 performance and outlook for Q2. Stefan?

Thank you, Thompson. As Thompson mentioned, we are pleased to deliver solid first quarter results that were in line with our expectations, driven by robust demand in both our data center and CATV businesses. We generated our fourth consecutive quarter of record revenue as we executed well to expand our manufacturing capacity. We continue to see accelerating customer demand needed to support the next wave of AI infrastructure deployment, and we anticipate solid sequential revenue growth throughout this year with a significantly larger ramp expected starting in Q3 as additional capacity comes online.

In Q1, we delivered revenue of $151.1 million which was in line with our guidance range of $150 million to $165 million. We recorded non-GAAP gross margin of 29.2%, which was in line with our guidance range of 29% to 31%.

Our non-GAAP loss per share of $0.07 was in line with our guidance range of a loss of $0.09 to breakeven. Notably, we continued to make progress on our key priorities in the first quarter, which included: one, scaling our next-generation data center products, including both our 400G and 800G solutions; two, expanding our production capacity in a disciplined manner to support anticipated demand, particularly in our Texas facility.

Three, diversifying our revenue base; and four, strengthening operational execution to improve our margins and long-term profitability. Importantly, during the quarter, we saw and continue to see strong customer engagement around our 800G and 1.6 terabit products, particularly as AI-driven data center investments accelerate. We completed our first volume shipment of our 800G single-mode transceivers to one of our large hyperscale customers.

Notably, 800G revenue in the first quarter was $4.6 million or 5.6% of our total data center revenue. Looking ahead, we continue to anticipate a strong volume ramp of our 800G products starting in Q2. During the quarter, in line with our expectations, along with the increasing demand for our 800G products, we also saw particular strength for our 400G products. Looking ahead, we expect continued strength in our 400G business, and we expect to ship nearly 4x the quantity of 800G compared to our Q1 shipments.

In Q1, we announced that we received our first volume order for our 1.6 terabit transceivers from another one of our long-term major hyperscale customers. We also announced that we have received 2 new volume orders from this customer for our 800G single-mode transceivers. Following product qualification, we expect to begin delivering these 800G orders in Q2 and the 1.6 terabit order as early as Q3 and to complete all of the deliveries by the end of this year. This hyperscale customer has been a key and valued customer of ours for many years. and we are excited by the increased engagement and meaningful discussions we have had as this customer boosts its network bandwidth for AI workloads.

We expect these orders to return this customer as a 10%-plus customer for us. Looking ahead, forecast demand for 800G and 1.6 terabit modules are projected to continue to exceed our production capacity through mid-2027. We are working to add additional capacity to meet this demand.

At OFC in March, we provided more color on our ambitious plans to increase our manufacturing capacity. During the first quarter, we made solid progress on this production capacity ramp particularly for our 800G and 1.6 terabit products. As a reminder, our U.S. manufacturing footprint is anchored in sugar land just outside Houston. Through a combination of real estate acquisition and leases, we have expanded our Texas manufacturing footprint to about 900,000 square feet. This includes 135,000 square feet of existing capacity at our headquarters.

Two new buildings of 388,000 square feet in Pearland, Texas, a 210,000 square foot facility, which is under development and a 154,000 square foot building in Houston, Texas. For those of you who are not familiar with the Houston area, all of these facilities are located within a 15-mile radius of our current headquarters facility in [ sugar ].

During the quarter, we made progress building out our recently leased 210,000 square foot facility. We expect to begin initial production in this facility in the third quarter. Notably, this facility is located just a few hundred yards from our headquarters and will be entirely dedicated to manufacturing of 800G and 1.6 terabit transceivers. While this will not directly increase our indium phosphide wafer capacity, we plan to move the existing transceiver production from our current headquarters facility to this new building, which will allow expansion of our indium phosphide capacity.

The facilities in Pearland and Houston will be built out to expand our production capacity for 800G and 1.6 terabit transceivers. We expect these facilities to come online in early 2027. As a reminder, internationally, we have 795,000 square feet across 3 facilities in Taiwan focused on optical transceivers as well as a larger 1.2 million square foot facility in Ningbo, China primarily dedicated to transceiver and cable TV manufacturing.

Exiting Q1, our total manufacturing capacity approached 100,000 units per month of 800G and 1.6 terabit capacity. Looking ahead, we expect to continue to rapidly expand our production capacity to approach $150,000 per month of 800G and 1.6 terabit this quarter. As a reminder, we expect by the end of this year that we will be capable of producing over 650,000 pieces of 800G and 1.6 terabit products per month with about 30% of that output coming from Texas as we expand into additional facility space and bring new production online.

By the end of next year, 2027, we expect to grow our production capacity to be able to produce over 930,000 pieces of 800G and 1.6 terabit products per month, with over half of that output coming from Texas. These investments reflect measured scaling of our footprint while aligning with our strong and growing customer demand and qualification progress across both 800G and 1.6 terabit products.

As a reminder, our 800G and 1.6 terabit products can be manufactured on the same production line with the same process. While our 1.6 terabit products will require a different final testing, our 800G automated manufacturing lines have been developed with an architecture that will allow us to support future high-speed products as customer demand materializes and evolves over time. While we continue to be encouraged by the conversations we are having with our customers pertaining to our 1.6 terabit product, we continue to believe that our 800G products will drive the near-term data center ramp. Our 1.6 terabit products are on track to begin to contribute to our overall revenue later this year with the bigger ramp beginning in 2027.

At OFC, we also discussed our plans to increase our manufacturing capacity for our external light source or EL SSP, that's for co-packaged optics or CPO. This utilizes the ultra narrow line with high-power laser that we announced late last year. We have very limited production of these modules now, but we anticipate ramping production later this year and into 2027. Ultimately culminating in about 400,000 pieces per month by the end of 2027.

As a reminder, we will be making the high-power lasers for these modules for the in-house production of the EL SSP. We believe our in-house laser capabilities continue to be a strategic advantage for the company. As we have mentioned before, we've been manufacturing lasers internally for many years. This has allowed us to avoid some of the shortages that affected others in the industry. As we continue to expand our footprint in Texas, our in-house laser manufacturing positions us well to support both near-term customer needs and longer-term growth. We believe that in the future, CPO will continue to drive increased demand for high-power lasers, and we plan to continue to expand our laser manufacturing capacity in Texas in order to accommodate these future growth drivers. We expect to further expand our laser fabrication capacity by around 350% by the end of 2027.

A central element of our strategy is a high process for transceivers, which allows us to deploy production capacity where it makes the most sense economically and geopolitically while scaling output quickly reliably and efficiently. As I mentioned, this automation platform is also highly flexible, enabling us to produce across multiple generations from 400G to 800G to 1.6 terabit using many of the same techniques and equipment.

In a fast-moving AI environment, that flexibility is critical as it allows us to rapidly ramp specific products and shift production in response to changing customer demand. This capability is the result of over a decade of investment in proprietary in-house design equipment and tightly integrated product and process engineering. The plans that we have unveiled have been evolving for some time. So while some of the required equipment does have long lead times, we've already ordered many of the key pieces of equipment and are working closely with our vendors to ensure on-time delivery.

Notably, equipment availability has not been a problem for us to date, which we believe is largely due to the fact that most of this equipment is developed in-house, which means that we're not generally in direct competition with other similar companies for supply of the necessary machinery and equipment to build our factories. There are exceptions to this, of course, but overall, we feel that our in-house developed technologies give us an edge in ensuring reliable supply of production equipment.

During the first quarter, direct tariffs had a $1.4 million impact on our income statement. With the overturn of the IEEPA tariffs, we have applied for a refund, which we currently anticipate will be at least $5.7 million. Our application for the refund has been approved, but as the process is still very new, we currently cannot estimate the time frame for recovery of these tariffs.

Turning to our first quarter results. Our total revenue was a record $151.1 million, which increased 51% year-over-year and increased 13% sequentially off a strong Q4 and was in line with our guidance range of $150 million to $165 million. During the first quarter, 54% of revenue was from our data center product, 44% was from cable TV products and the remaining 2% was from FTTH, telecom and other. In our data center business, Q1 revenue came in at $81.4 million, which was up 154% year-over-year and 9% sequentially.

Sales of our 100G products increased 36% year-over-year, while sales for our 400G products increased tenfold year-over-year. In the first quarter, 41.9% of data center revenue was from 100G products, 46.7% was from 200G and 400G product. 5.6% was from 800G transceiver products. and 5.6% was from 10G and 40G transceiver products.

In our CATV business, CATV revenue was $66.8 million, which was up 4% year-over-year and 24% sequentially. The and was at the high end of our expectations of $61 million and $67 million. Similar to the last couple of quarters, we shipped a significant quantity of 1.8 gigahertz amplifiers to our largest CATV customer in Q1 and based on recent conversations with customers, we believe demand will be somewhat higher than our initial projections for 2026.

We continue to see momentum with a newer set of MSO customers that we have talked about on our prior few earnings calls. Looking ahead to Q2, we expect our CATV revenue will be between $75 million and $80 million. Looking further ahead, we now currently expect to generate over $325 million annually in CATV. While the vast majority of our CATV revenue expectations for this year are related to our amplifiers, we do anticipate that we will generate some revenue from our software solutions this year.

Now turning to our telecom segment. First quarter revenue from our telecom products of $2.6 million was down 13% year-over-year and 50% sequentially. As we have said before, we expect telecom sales to fluctuate from quarter-to-quarter. For the first quarter, our top 10 customers represented 98% of revenue compared to 97% of revenue in Q1 of last year.

We had 3 greater than 10% customers, one in the CATV market, which contributed 44% of total revenue and, two, in the data center market. which contributed 26% and 25% of total revenue, respectively. In Q1, we generated non-GAAP gross margin of 29.2%, which was in line with our guidance range of 29% to 31% and compared to 31.4% in Q4 of 2025 and 30.7% in Q1 2025.

As we discussed on our last quarterly earnings call, while we do expect continued gradual improvement in gross margins, we continue to expect that the revenue mix in data center in the short term will be a slight headwind. We remain committed to our long-term objective of returning non-GAAP gross margins to around 40% and believe that this goal is achievable as our mix shifts towards higher-margin products and as we capture additional efficiencies across our operations.

That margin expansion, combined with increased scale, positions us to move towards sustainable profitability, which we continue to expect to approach on a non-GAAP basis beginning this quarter. The revenue figures presented above our net of a contra revenue amount due to the accounting for warrants provided to customers. As a reminder, this amounts to approximately 2.5% of revenue derived from certain customers to whom AOI has provided warrants in exchange for future revenue. In Q1, the amount of this contra revenue was $1 million.

Total non-GAAP operating expenses in the first quarter were $51.4 million or 34% of revenue which compared to $35.5 million or 36% of revenue in Q1 of the prior year and were in line with our expectations of $50 million to $57 million. Looking ahead, we expect non-GAAP operating expenses to be in the range of $50 million to $58 million per quarter.

Non-GAAP operating loss in the first quarter was $7.3 million, compared to an operating loss of $4.8 million in Q1 of the prior year. GAAP net loss for Q1 was $14.3 million or a loss of $0.19 per basic share compared with a GAAP net loss of $9.2 million or a loss of $0.18 per basic share in Q1 of the prior year. On a non-GAAP basis, net loss for Q1 was $4.9 million or $0.07 per share which was in line with our guidance range of a loss of $7 million to a loss of $0.3 million and non-GAAP income per share in the range of a loss of $0.09 to breakeven. This compares to a non-GAAP net loss of $0.9 million or $0.02 per share in Q1 of the prior year.

The basic shares outstanding used for computing the earnings per share in Q1 were $76 million. Turning now to the balance sheet. We ended the first quarter with $449.4 million in total cash, cash equivalents, short-term investments and restricted cash. This compares with $216 million at the end of the fourth quarter of 2025. We ended the first quarter with total debt, excluding convertible debt of $77 million which compared to $67.3 million at the end of last quarter.

As of March 31, we had $206.2 million in inventory, which compared to $183.1 million at the end of Q4. The increase in inventory is primarily due to raw material and work in progress needed for production, partially offset by a decrease in finished goods inventory as purchase orders to customers were fulfilled in the quarter. We made a total of $68.7 million in capital investments in the first quarter, which was mainly used for manufacturing capacity expansion for our 400G, 800G and 1.6 terabit transceiver products.

We expect to continue to make sizable CapEx investments this year as we prepare for increased 400G, 800G and 1.6 terabit data center production. On a quarterly basis, we expect our capital expenditures to be above the total that we spent in Q1. We expect to finance these investments through a combination of cash on hand, cash generated from operations and some equity sales along with additional debt.

Notably, in Q1, we increased availability under existing and new loan agreements by $13.4 million and added another $14.5 million in April. Going forward, we believe we are well positioned for sustained growth across both our data center and CATV businesses, and the capital investments underway are expected to fundamentally strengthen the company as we execute on these opportunities. Given the rising demand, we now believe that by mid-2027, 100G and 400G revenue will be approximately $90 million. 800G revenue will be approximately $217 million and 1.6 terabit revenue will be approximately $164 million monthly. In total, this is about $471 million per month of data center transceiver revenue with about 40% of this capacity in the U.S.

Moving now to our Q2 outlook. We expect Q2 revenue to be between $180 million and $198 million, accounting for a sequential increase in CATV revenue as well as a sequential increase in our data center revenue. We expect non-GAAP gross margin to be in the range of 29% to 30%. Non-GAAP net income is expected to be in the range of a loss of $2.5 million to income of $2.8 million and non-GAAP earnings per share between a loss of $0.03 per share and earnings of $0.03 per share using a weighted average basic share count of approximately 80.7 million shares.

Looking more broadly at 2026, we now expect to generate over $1.1 billion in revenue this year, with a non-GAAP operating profit of over $140 million. As we have discussed previously, this revenue level is limited by our production capacity and supply chain, not market demand, which we believe is much larger. Based on our planned capacity additions, we expect to see an acceleration in the second half of the year as new production capacity comes online and additional customer qualifications are completed and orders begin to ship. We believe that this is an ambitious yet achievable target based upon our customers' forecasts and what we know about the unprecedented investments that are being made in AI infrastructure.

With that, I will turn it back over to the operator for the Q&A session. Operator?

[Operator Instructions]. Our first question comes from Simon Leopold with Raymond James.

I wanted to dig in a little bit to understand the risk profile for ramping the capacity I appreciate the nuance that you do a lot of your own tooling and machinery and so that should put it in your control. But I wonder if you could reflect on sort of the prior capacity expansions what led to the -- any kind of timing or disruption? And help us understand sort of how to prioritize the risk for meeting your schedule? And then I've got a quick follow-up.

Sure, Simon. I think it's important to understand that the expansion that we're undergoing, while it's large in scope is not something that's brand new to us, right? We've built significant capacity especially in our Asian factories over the last couple of years. And now we're basically adding additional increments to that capacity, the same type of equipment, the same manufacturing process mainly here in the U.S., okay, here in Texas, as we talked about during the call.

So from a risk standpoint, the risk of doing something that you've already done is a lot lower than doing something that's brand new, right? As we mentioned on the call, on the prepared remarks, a lot of this equipment is developed in-house. So the risk of supply chain disruptions for the equipment I mean it's not eliminated, right, of course, but it's a lot lower than if we were relying on the same equipment that was being bid up by other suppliers and it had limited supply to begin with, right?

So I think those 2 risks are minimized because of the nature of the manufacturing process that we have. It's worth noting, too, that because the process for us is very highly automated. We're not hiring a lot of people. So the labor, the risk associated with quality control issues or being able to scale labor doesn't really exist to any great extent for us as well. So it's really just a matter of can we get the equipment in and can we put it into production on time. And so far, we're executing very well to that, which isn't surprising because we've done a good job of it over the last couple of years already.

Yes, just a quick follow-up. I want to make sure I understand and clarify the metric you shared with us towards the end of the call, the $471 million monthly of production by the middle of '27. I want to make sure I understand, is that a capacity number? Or is that a number that assumes a certain percent utilization of the total capacity available? How should we take that $471 million value? Is that a revenue forecast? Or is that a capacity capability we should assume some haircut to that for lower utilization.

This is Thompson, that's based on revenue. Actually, the actual capacity is higher. But you understand -- we've got equipment in the several months to hire people qualification. So let me space the order in hand or minimum commitment from a customer pros. The equipment has been fully qualified. So that means June, July, less there we believe we can deliver.

For sure, not only I think -- another risk is material. So this is why we are working with all the material suppliers to see -- keen material supply. That's a number we feel comfortable to commit at this moment. But if you say the actual demand could be even higher than this number, but that's the best we can do. And let me say that, the actual number from a customer is bigger. And actually, what they expect is April, not June, July. So we are -- every in the pure in.

And Simon, just to make it really clear, if you go back to our remarks in the last earnings call, that number was $378 million monthly. So that $471 million is directly comparable to that, and it represents almost $100 million a month of additional revenue in the middle part -- starting in the middle part of next year.

Next, we have George Notter with Wolfe Research.

Jaren on for George Notter. On the OSP business, can you talk a little bit more about the customer engagements you're seeing there? Who are you working with? Or how many customers are you working with any details would be appreciated.

Yes, we have a couple of large customers that we're working with. We haven't said who they are, but...

Let me say that right now, we are working on a long-term agreement with -- including laser, including ESP. So that's the number we are talking about. That's why now in the transceiver, we are expanding very fast about our laser capacity. And right now, we have been doing a 4-inch growth process. Our targets go to 6-inch by end of next year. So yes, I think we need to do more investment to meet the demand for CPO market.

As you know, the CPU laser is about 300-milliwatt to 400-milliwatt compared to 70-milliwatt for AND transceiver and 100-milliwatt for 1.6 terabit transceiver. The size is much bigger, okay? [indiscernible] maybe 5x or 6x bigger. That's why we need to go to maybe the low we already got from like 2-inch to 3 to 4-inch in the past 18 months, but we still plan to go to 6-inch by end of next year. including -- they will increase our capacity a lot. But at same time, we are adding a lot of capacity like MOCVD, EP and -- everything, okay?

Jarren, we see a shortage of indium phosphide laser manufacturing capacity across the industry right now, and we think that's going to persist and even get more acute with the advent of ELSS, as Thompson mentioned. That's why we see this need to really expand our -- phosphide fabrication capability pretty dramatically over the next 12 to 18 months.

Great. And then just to follow up on that. How do you see the ability to secure the substrate capacity for the aluminum phosphide?

Right now, we record -- supplier with some kind of discussion -- sorry, not much we can say. But 4 of them are outside of China. So I would say right now, we should have enough inventory minimal almost 1 year. But since the volume, we increased surface, we are making calls with all the suppliers.

I would say we've got good line of sight into how we think we can not see the shortage there. But we can't say too much about it specifically at this point because a lot of it is under discussions.

Our next question comes from Michael Genovese with Rosenblatt Securities.

Can you give us more granularity on when you expect qualification for 800G with this hyperscaler. It sounds like we'll be your third hyperscale 10% customer. But when in the quarter, exactly do you think you'll have this qualification. And then does your guidance derisk it, meaning that if you got it see or if things went to plan, would there be upside in the quarter?

Well, as we mentioned in our prepared remarks, we've already started shipping. So I'm not sure what the qualification question really is referring to.

So we had 2 big cost, one is qualified, another one almost qualified. The one give us a large order for, I don't remember, 100, I don't know, 40 million in they are negotiated with AOI with some kind a long-term agreement with a very big volume. The qualification is pretty smooth. I think we started shipping some value in next month.

For another customer, we've been working for a long time, has qualified. We increased the capacity in this month -- this quarter too. So we started shipping volume to big customers, not console.

Got it. Okay. And then your guidance for the year, you've you're doing about 1/3 of the revenue for the year in the first half. And then obviously, you expect a big sequential growth in the third quarter. Would then we have more big sequential growth in the fourth quarter? Or is 3 and 4 more linear. Like how should we think about the shape of the second half?

That is a very great question. Right now, as I say, because dynamic expand to you -- order equipment and qualification, installation, everything and sometimes we liability. Even in Asia, it will take 5 to 7 months. In the U.S., it's adding another 2 months because of shipping, okay? So that's why the ramping from the Q3, not Q2, okay, even we've got some equipment in already, but still need to go through a lot of process which will take several months. So right now in Q3, we can see compared to Q2, 60% to 80% increase. Q4 should be similar, and you can figure out the number. And let me say that the actual demand is not 1.1 billion, it should be 1.4 billion, 1.5 billion. So our target still go to 1.1 billion or 1.2 billion, but we still need to work very hard, like the supply chain -- manpower -- but right now 1.1 billion is the number we feel very competent and this increased from 1 billion we commit in the last quarter. But our internal number is high.

Mike, just to summarize what Tom has said, right, the limiting factor for deliveries is our ability -- the manufacturing capacity that we have available. So once that capacity that we've been building, we talked in detail about the real estate that we have and the number of square feet that we've added and the equipment, Simon asked some very detailed questions about our equipment capacity and how confident we are in that once that starts to come online, it's not going to be a linear type of thing. It's going to be another large increment and then another large increment in Q4, as Thompson outlined. But that's why it's not you can't extrapolate from the first half and go, well, there's only a certain growth rate. No, when you have new factories coming along, that adds capacity very quickly.

And that said, even you got the equipment, okay? It is still take easily in -- type ton. It takes at least more than 3 months or even longer to deliver revenue, okay? Because sometimes costs will need to do another onset audit, some clarification. So we got a light equipment in, but cloud rerailing is more like Q3. So that's what I told you. Yes, I think Q2, we had maybe 30% growth. Let's limit backout capacity. But Q3, Q4, we took now 60%, 70% or even 80% of growth in every quarter or actually even Q1 next year -- in the next few quarters, of course will be very fast because that's when we can start the deliver to the customer.

Our next question comes from Ryan Koontz with Needham.

Great. I just want to ask about get back to the Indian phosphide topic here and where you are in terms of that capacity relative to your demand and the different fab equipment you need to support that growth. Can you maybe kind of walk us through some of the major milestones we should think about for the laser supply internal here over the next couple of quarters?

Right. Great question. So as I said earlier, I think indium phosphide capacity is critical right now. The fact that we have our own in-house laser manufacturing capability is one of our key advantages. Certainly, when you talk to customers, that's one of the big things that they like about us, especially now that we're seeing shortages across the industry. So our fab expansion is well underway.

As Thompson mentioned, we've got a number of critical pieces of equipment and coding machines and others that are in various stages of either being delivered or being qualified. It does take a pretty extended period of time to qualify a new piece of laser manufacturing equipment, as you can imagine, you don't want to take a risk of having an unknown quality issue there.

So a lot of that is already here and already undergoing qualification or it's very close to being here. And that's why we can be pretty confident that our capacity is going to be where we need it to be. It's just a matter of going through that qualification process internally, which is, by the way, different from the transceiver qualification here, I'm talking about our internal qualification of new equipment as it comes in.

Let me say, it is very -- transceiver, the latest front day, you pressed order to our equipment supply. You take minimum 18 months or even longer. Even right now, I think with the equipment delivered take 21 to 24 months for us to start to deliver later to the customer, okay? Because sometimes the custom require 3,000 hours or even 5,000 hours of reliability data.

So we placed a lot of orders every -- more than 50 suppliers, let me say that. We got a commitment from the supplier, and we're getting some equipment in-house already almost every month. And let me say that by end of next year, we should be a minimal top 3 laser worldwide okay? I can't tell you how many equipment we had is competent. That's why we are working, we serve a customer for, now needed for laser, not only for transceiver, including laser for ESP. As I said, ESP is very challenged. This is very high spec and very high power as within control. I would say the tranche is more than 10x of like 70 or 80, 100 -- for transceiver, this total different ball game.

That is our focus. And you know AOI has been doing laser since day 1, including my previous users has been doing it laser since 1990. So we know how to do a good job.

If I can -- a quick follow-up there in terms of your margins and how we should think about that and the mix. As you production mix of 800 moves up here? Should we think about that the tailwind for margins? Maybe can you unpack that for us just a little bit how to think about the mix.

Yes. The margins get a lot better as we expand the capacity. Right now, what's going on is we're in this shifting mix between 400 and 800 and between predominantly cable TV and predominantly data center, right? So as we see that continue to shift and as 800 takes precedent you'll start to see growth in gross margin in the second -- primarily in the second half of the year.

So we go to 35% gross margin by end of this year. And at the same time, in Q1, Q2 since we started ramping up under the no, we need the time to fine-tune the process, okay? Efficiency size as good as what we expect. But I think within 2, 3 months, I think with a fully automated manufacturing line, we can crew out efficiency in it very fast. That's the major advantage of automation. By Q3 or by Q4, the gross margin, the whole company should be, I would say, more that 40% or special with the laser in that will kick in Q3, Q4 next year.

[Operator Instructions]. Our next question comes from Tim Savageaux with Northland Capital Markets.

First question is trying to understand where you are capacity-wise versus what you're forecasting. So I think in the release, you talked about 100,000 units a month exiting Q1 in 800 gig. And that puts your capacity revenue-wise well over $100 million, right, a quarter. We've got orders in hand for $124 million of 800 gig. The capacity theoretically to ship those quarters, and yet you're guiding to, what, $18 million, $20 million in 800 gig revenue. What I'm trying to understand is that delta? And what's driving that apparent disconnect? And I have a follow-up.

It's just timing on how long it takes to do the manufacturing process really. Not all that $100,000 was online in the middle of the quarter and then you add the cycle time to it. It puts the real production output for that closer to middle to even tunes the way through the quarter. It's just the timing of the manufacturing lead time.

So that's what I said on [indiscernible] why we're talking about [ $471 million ] for June, July, the next year. That's why I said that is the revenue. Not capacity, the capacity is much higher because as I said, when capacity, you need have added more than 1 month on a future cycle time of 6 weeks. For us, maybe customer needs to do -- auditing, quite -- and this all kind of requirements. So that you have -- even you install [indiscernible] everything, it still easy to take another 2, 3 or 4 months to realize revenue or even some of the customers, even have like some kind of light hub, all kind of -- processes -- why I made clear when we're talking about $471 million -- capacity and we're talking about equal to about [ 780,000 ] of transceiver per month by mid of next year, but actually, it could be high. Actually, it's high.

Okay. Got it. Yes. I mean, incidentally, that would make you about the same size as coherent after kind of a multiyear ramp over there. sort of numbers on a yes -- another question.

Yes, go ahead.

Speaking of competition, earlier this week, we had a prominent contract manufacturing space, announced 2 deals whereby they would be making transceivers for hyperscale customers directly. How would you assess the competitive and margin impact of that development on AI?

We don't really know. But anyway, right now, I think the demand is more support we get delivered in this one. Let me say that the LTE, we are negotiated with these 3 customers the 3-year number is quite high, okay. And it depends, let me say that for money more, okay, it's easier, maybe you can use like fiber or whatever or even for like [indiscernible] it's easy to manufacture, but it will be very tough for the 800T or 1.6T to buy FFO because you need for laser. But key is the sensing, can you get late or not? Even there's many laser transceiver supplier by how quick these laser from. Right now, dementia to cable book even broken -- so with all that, I can make intent.

Got it. And last one for me. This goes back to the 1.6 comments where Stefan, I think you talked about some revenue contribution later in the year and a bigger ramp in '27. And yet, I think it was my understanding that the big order that you announced was -- was that to be shipped completed in '26? Has there been some change there? Or what's the...

It means that order is just a small order compared to what we're going to see in 2027.

Much bigger. I think the bar slide [indiscernible].

All right. So we got to define our terms. $200 million is not a big amount. Okay. I got it. Exactly.

Next year, we are talking about more than $2 billion, 1.6 transceiver -- have much more than $1.6 billion we need to deliver in the next year.

This concludes our question-and-answer session. I would like to turn the conference back over to Dr. Thompson Lin, Founder, President and CEO, for any closing remarks.

Again, Singapore joined us today, as always, we want to extend a thank you to our investors, customers and employees for your continued support is an exciting time for our industry as for AOI. We continue to believe the fundamental driver of long-term demand for our business remains robust, and we are unique position to drive value from this opportunity. We look ahead to see my view at upcoming investor conference. Thank you.

The conference has now concluded. Thank you for attending today's presentation. You may now disconnect.

Good afternoon. I will be your conference operator on today's call. At this time, I would like to welcome everyone to Applied Optoelectronics Fourth Quarter and Full Year 2025 Earnings Conference Call. [Operator Instructions] Please also note that this call is being recorded today.

I'll now turn the call over to Lindsay Savarese, Investor Relations for AOI. Ms. Savarese, you may begin.

I'm Lindsay Savarese, Investor Relations for Applied Optoelectronics. I'm pleased to welcome you to AOI's Fourth Quarter and Full Year 2025 Financial Results Conference Call. After the market closed today, AOI issued a press release announcing its fourth quarter and full year 2025 financial results and provided its outlook for the first quarter of 2026. The release is also available on the company's website at ao-inc.com. This call is being recorded and webcast live. A link to the recording can be found on the Investor Relations section of the AOI website and will be archived for 1 year.

Joining us on today's call is Dr. Thompson Lin, AOI's Founder, Chairman and CEO; and Dr. Stefan Murry, AOI's Chief Financial Officer and Chief Strategy Officer. Thompson will give an overview of AOI's Q4 results, and Stefan will provide financial details and the outlook for the first quarter of 2026. The question-and-answer session will follow our prepared remarks.

Before we begin, I would like to remind you to review AOI's safe harbor statement. On today's call, management will make forward-looking statements. These forward-looking statements involve risks and uncertainties as well as assumptions and current expectations, which could cause the company's actual results, levels of activity, performance or achievements of the company or its industry to differ materially from those expressed or implied in such forward-looking statements.

In some cases, you can identify forward-looking statements by terminology such as believes, forecasts, anticipates, estimates suggest intends, predicts, expects, plans, may, should, could, would, will, potential or thinks or by the negative of those terms or other similar expressions that convey uncertainty of future events or outcomes. The company has based these forward-looking statements on its current expectations, assumptions, estimates and projections. While the company believes these expectations, assumptions, estimates and projections are reasonable, such forward-looking statements are only predictions and involve known and unknown risks and uncertainties, many of which are beyond the company's control.

Forward-looking statements also include statements regarding management's beliefs and expectations related to the expansion of the reach of its products into new markets and customer responses to its innovations as well as statements regarding the company's outlook for the first quarter of 2026 and for the full year of 2026. Except as required by law, AOI assumes no obligation to update these forward-looking statements for any reason after the date of this earnings call to conform these statements to actual results or to changes in the company's expectations. For information about other risks that may impact the company's business are set forth in the Risk Factors section of AOI's reports on file with the SEC, including the company's annual report on Form 10-K and quarterly reports on Form 10-Q.

Also, our financial results and other financial measures discussed today are on a non-GAAP basis, unless specifically noted otherwise. Non-GAAP financial measures are not intended to be considered in isolation or as a substitute for results prepared in accordance with GAAP. A reconciliation between our GAAP and non-GAAP measures as well as a discussion of why we present non-GAAP financial measures are included in the company's earnings press release that is available on AOI's website.

Before moving to the financial results, I'd like to note that AOI management is attending the Susquehanna Annual Technology Conference virtually tomorrow as well as the Raymond James Annual Institutional Investors Conference on March 3. Additionally, management will host an investor session at OFC on Tuesday, March 17 in Los Angeles. This discussion will be webcast live, and a link to the webcast is available on the Investor Relations section of the AOI website. Lastly, I'd like to note that the date of AOI's First Quarter 2026 earnings call is currently scheduled for May 7, 2026.

Now I would like to turn the call over to Dr. Thompson Lin, AOI's Founder, Chairman and CEO. Thompson?

Thank you Lindsay. Thank you for joining our call today. We are pleased to deliver record fourth quarter results that were in line with or better than our expectations and which came off the strongest year in our company's history. Our results were driven by robust demand in both our CATV and datacenter business.

In 2025, total revenue increased 83% compared to 2024 to a record $456 million compared to the revenue of $196 million, increased 32% compared to 2024, while our CATV revenue nearly tripled to $245 million in the same period. We entered 2026 with strong momentum. And due to the software investment we have made, we have materially expanded our manufacturing capacity. We believe this position us well to meet increasing customer demand and we lead to accelerating growth this year.

During the quarter, we announced that we received our fourth 800G volume order from one of our major hyperscale customers to support its AI datacenter growth. This was an important milestone in our next-generation datacenter roadmap and followed the successful clarification of our energy products by the customer. It also reflects both the strength of our product portfolio and the deep relationship we have with these hyperscale customers. We continue to work with this customer to finalize the firmware use in this module to ensure inability across their network, which we believe will be completed in March. We have begun ramping up production of this 800G module in anticipation of a strong volume ramp starting in Q2. We focused demand for 800G module are projected to exist our production capacities through mid-2027, and we are working to add additional capacity to meet this demand.

During the quarter, we saw particular strength for our 400G products with this customer, which more than offset our 800G revenue which came in below our expectation of $4 million to $10 million. Due to the ongoing firmware optimizations, I mentioned about looking ahead we expect continued strength in our 4G business. Also 800G is expected to dominate our revenue beginning in Q2. As a reminder, our Taiwan facility was already qualified for production of several 800G product types from this hyperscale customer during 2025. Our tax facility was also qualified for production of some of our 800G products. During the quarter, we made investment with qualifying additional product from our Texas facility with this customer and expect full qualification by Mid East.

We expect that we move throughout the year to ship an increasing amount of 800G products from our tech facility as we expand our capacity. In addition to this for major 800G customers, we have had indications from another existing hyperscale customers that they intend to begin to order 800G from us soon. Finally, a new hyperscale customer has become a discussion about qualifying our 800G and 1.6T products just within the last few weeks. So we feel increasingly confident about our trajectory in 800G and 1.6T receiver with multiple customers. During the fourth quarter, we delivered revenue of $134.3 million, which was in line with our guidance range of $125 million to $140 million.

We recorded non-GAAP gross margin of 31.4%, which was above the high end of our guidance range of 29% to 31%. And on non-GAAP loss per share of $0.01 was narrowed than our guidance range of a loss of $0.13 to a loss of $0.04. Total revenue for our data center product of $74.9 million, increased 69% year-over-year and 70% sequentially. Sales of our products increased 54% year-over-year, and sales for 400G products increased 141% year-over-year. Total revenue in Q4 in our CATV segment was $54 million, which was up 3% year-over-year and in line with our expectations, was down 24% sequentially from a later Q3, similar to the last couple of quarters, we see a significant quality of 1.8 gigahertz amplify to our largest CATV customers in Q4, and demand from then continue to be robust. In addition to these customers, we continue to see momentum from a new set of MSO customers.

With that, I will turn the call over to Stefan to review the details of our Q4 performance and outlook for Q1. Stefan?

Thank you, Thompson. As Thompson mentioned, we are pleased to deliver record fourth quarter results that were in line with or better than our expectations and which capped off the strongest year in our company's history. Our performance was driven by robust demand in both our CATV and data center businesses.

We entered 2026 with strong momentum. And due to the thoughtful investments we have made, we have materially expanded our manufacturing capacity. We believe this positions us well to meet increasing customer demand and lead to accelerating growth this year. Throughout 2025, our focus remained on a few key priorities. One, scaling our next-generation data center products, including both our 400G and 800G solutions; two, expanding our production capacity in a disciplined manner to support anticipated demand, particularly in our Texas factory; three, diversifying our revenue base; and four, strengthening operational execution to improve our margins and long-term profitability. I'm pleased to report that we made significant progress on each of these fronts and these will continue to be key priorities in 2026.

Importantly, we saw and continue to see strong customer engagement around 800G and 1.6 terabit products, particularly as AI-driven data center investments accelerate. In 2025, total revenue increased 83% compared to 2024 to a record of $456 million. Datacenter revenue of $196 million increased to 32% compared to 2024, while our CATV revenue nearly tripled to $245 million in the same period. Additionally, we expanded our gross margins and made progress on our path to profitability.

Turning to the quarter. In Q4, we delivered revenue of $134.3 million, which was in line with our guidance range of $125 million to $140 million. We recorded non-GAAP gross margin of 31.4%, which was above our guidance range of 29% to 31%. Our non-GAAP loss per share of $0.01 was narrower than our guidance range of a loss of $0.13 to a loss of $0.04. During the quarter, we announced that we received our first 800G volume order from one of our major hyperscale customers to support its AI data center growth. This was an important milestone in our next-generation data center road map and follows the successful qualification of our 800G products by this customer. It also reflects both the strength of our product portfolio and the deepening relationship we have with this hyperscale customer.

We continue to work with this customer to finalize the firmware used in these modules to ensure interoperability across their network, which we believe will be completed in March. We have begun ramping our production of these 800G modules in anticipation of a strong volume ramp starting in Q2. Forecast demand for 800G modules are projected to exceed our production capacity through mid-2027, and we are working to add additional capacity to meet this demand. During the quarter, we saw particular strength for our 400G products with this customer, which more than offset our 800G revenue, which came in below our expectations of $4 million to $8 million due to the ongoing firmware optimization I mentioned above.

We expect continued strength in our 400G business, although 800G is expected to dominate our revenue beginning in Q2. As a reminder, our Taiwan facility was already qualified for production of several 800G product types from this high-scale customer during 2025. Our Texas facility was also qualified for production of some of our 800G products. During the quarter, we made advancements with qualifying additional products from our Texas facility with this customer and expect full qualification by midyear. We expect as we move throughout the year to ship an increasing amount of 800G products from our Texas facility as we expand our capacity. Given the strong demand, we have continued to invest in our manufacturing capacity to support current and future demand.

During the fourth quarter, we made solid progress on the production capacity ramp we outlined last year at OFC. Over the past several years, we have purposely developed and scaled automation across key elements of our production process from laser fabrication to transceiver assembly and testing. This automation not only improves yield but it also supports rapid scale up with greater flexibility in terms of geographic location of production and lower geographic index labor cost relative to many of our competitors who rely on traditional more labor-intensive manual operation. As we continue to bring new automated lines into production, we expect this differentiation to increasingly translate into execution strength and significant revenue expansion.

As we discussed at length at OFC last year, our focus remains on scaling manufacturing capacity for our next-generation transceivers, particularly 800G and 1.6 terabit products, and we remain on track with the milestones we previously discussed. As we exited the year, we neared our target of 100,000 units per month of 800G capacity, with approximately 90,000 units per month of 800G capacity at year-end with roughly 31% of that production based in the U.S. We made tangible progress during the quarter through facility expansion and equipment installations, both of which are critical steps as we prepare for higher volume production.

Our production capacity in the U.S. is currently in our existing footprint in Texas. During the fourth quarter, we announced that we signed an agreement to lease an additional building in Sugar Land. We began construction on this new facility earlier this month and are working hard to scale our production towards the middle end of this year to achieve our 2026 targets. Looking further ahead, we expect that by the end of this year, we will be capable of producing over 500,000 pieces of 800G and 1.6 terabit products per month with about 1/4 of that output coming from Texas as we expand into additional facility space and bring new production online. These investments reflect measured scaling of our footprint while aligning with strong and growing customer demand and qualification progress across both 800G and 1.6 terabit products.

Further, we have recently had dialogue with another large hyperscale customer who has been a long-term customer of ours, and it was eager to begin qualification efforts for our 1.6 terabit products. This customer has also indicated a desire to purchase potentially significant quantities of 800G products from us in 2026 and 2027. We continue to discuss capacity availability, and expect orders for 800G from this customer soon. It's also important to note our 800G and 1.6 terabit products can be manufactured on the same production line with the same process, while our 1.6 terabit products will require a different final testing, our 800G automated manufacturing lines have been developed with an architecture that will allow us to support future higher-speed products as customer demand materializes and evolves over time. While we are encouraged by the conversations we are having with our customers pertaining to our 1.6-terabit products, we continue to believe that our 800G products will drive the near-term data center ramp and our 1.6 terabit products are on track to begin to contribute to our overall revenue later this year.

Before moving on to our fourth quarter results, I'd also like to reemphasize our in-house laser capabilities, which we believe continue to be a strategic advantage for the company. As we have mentioned before, we've been manufacturing lasers internally for many years. Having these capabilities has allowed us to avoid some of the shortages that affect others in the industry. As we continue to expand our footprint in Texas, our in-house laser manufacturing positions us well to support both near-term customer needs and longer-term growth. We believe that in the future, CPO will continue to drive increased demand for high-power lasers and plan to continue to expand our laser manufacturing capacity in Texas in order to accommodate these future growth drivers.

During the fourth quarter, direct tariffs had a $1.2 million impact on our income statement. As it relates to tariffs, as I have previously mentioned, while we do utilize some imported components in our transceivers, many key components like our relationships are already manufactured in the United States. Importantly, in our 800G and 1.6 terabit transceiver designs, less than 10% of the value of these components used is currently sourced from China, and we have a path to further reduce that exposure to near 0 that we have discussed on our prior earnings calls. Given the recent court decision on IEEPA tariffs, it's worth noting that AOI acted as the importer of record for many, if not most, of the tariff shipments we incurred in 2025.

Turning to our fourth quarter results. Our total revenue was a record $134.3 million, which increased 34% year-over-year and increased 13% sequentially off a strong Q3 and was in line with our guidance range of $125 million to $140 million. During the fourth quarter, 56% of revenue was from data center products, 40% was from CATV products and the remaining 4% was from FTTH, telecom and other. In our datacenter business, Q4 revenue came in at $74.9 million, which was up 69% year-over-year, and 70% sequentially. Sales of our 100G products increased 54% year-over-year and sales of our 400G products increased 141% year-over-year. In the fourth quarter, 51% of data center revenue was from 100G products, 41% was from 200G and 400G transceiver products and 8% was from 10G and 40G transceiver products.

In our CATV business, CATV revenue was $54 million, which was up 3% year-over-year but was down 24% sequentially from a record Q3 and was in line with our expectations of $50 million to $55 million. Similar to the last couple of cores, we shipped a significant quantity of 1.8 gigahertz amplifiers to our largest CATV customer in Q4, and demand continues to be robust. In addition to this customer, we continued to see momentum with a newer set of MSO customers that we have talked about on our prior couple of earnings calls. Looking ahead to Q1, we expect our CATV revenue will be between $61 million and $67 million.

Looking further ahead, the broad-based appeal of our CATV amplifiers and software solutions has been evident in these customer engagements, and we see software as an increasingly important part of our CATV offering. Our QuantumLink software suite is designed to provide operators with enhanced remote management, visibility and control over HFC network element, reducing operational costs and improving service quality. If current momentum continues and while it is still early in the year, we still believe that it's feasible that we could generate nearly $300 million annually. While the vast majority of our CATV revenue expectations for this year are related to our amplifiers, we do anticipate that we will generate some revenue from our software solutions this year, and we will share more on the amount and timing as we progress throughout the year.

Now turning to our telecom segment. Fourth quarter revenue from our telecom products of $5.1 million, was up 45% year-over-year and 37% sequentially. As we have said before, we expect telecom sales to fluctuate from quarter-to-quarter. For the fourth quarter, our top 10 customers represented 96% of revenue compared to 97% of revenue in Q4 of 2024. We had 3 greater than 10% customers, 1 in the CATV market, which contributed 39% of total revenue and 2 in the datacenter market, which contributed 31% and 21% of total revenue, respectively. Of note, one of these data center customers became a 10% customer for the first time in a long time and is a U.S.-based large hyperscale customer.

In Q4, we generated non-GAAP gross margin of 31.4%, which was above the high end of our guidance range of 29% to 31% and was up from 31% in Q3 of 2025 and 28.9% in the prior year quarter. The year-over-year increase in our gross margin was driven primarily by our favorable product mix and our cost reduction efforts. Looking ahead, we expect continued gradual improvement in gross margins, although we continue to expect that the revenue mix in data center in the next few quarters will be a slight headwind. We remain committed to our long-term objective of returning non-GAAP gross margins to around 40%, and we believe that this goal is achievable as our mix shifts towards higher-margin products and as we capture additional efficiencies across our operations. That margin expansion, combined with increased scale positions us to move towards sustainable profitability, which we currently expect to achieve on a non-GAAP basis beginning in Q2 of this year.

The revenue figures presented above are net of a contra revenue amount due to the accounting for warrants provided to customers. As a reminder, this amounts to approximately 2.5% of revenue derived from certain customers to whom AOI has provided warrants in exchange for future revenue. In Q4, the amount of this contra revenue was $0.73 million. Total non-GAAP operating expenses in the fourth quarter were $49.3 million or 37% of revenue, which compared to $31.5 million or 31% of revenue in Q4 of the prior year and were in line with our expectations of $48 million to $50 million. Looking ahead, we expect non-GAAP operating expenses to be in the range of $50 million to $57 million per quarter.

Non-GAAP operating loss in the fourth quarter was $7.1 million compared to an operating loss of $2.5 million in Q4 of the prior year. GAAP net loss for Q4 was $2 million or a loss of $0.03 per basic share compared with a GAAP net loss of $119.7 million or a loss of $2.60 per basic share in Q4 of the prior year. On a non-GAAP basis, net loss for Q4 was $0.6 million or $0.01 per share, which was narrower than our guidance range of a loss of $9 million to a loss of $2.8 million or non-GAAP income per share in the range of a loss of $0.13 to a loss of $0.04. This compares to a non-GAAP net loss of $1 million or $0.02 per share in Q4 of the prior year. The basic shares outstanding used for computing the earnings per share in Q4 were $70.3 million.

Turning now to the balance sheet. We ended the fourth quarter with $216 million in total cash, cash equivalents, short-term investments and restricted cash. This compares with $150.7 million at the end of the third quarter of 2025. We ended the fourth quarter with total debt, excluding convertible debt of $67.3 million compared to $62 million at the end of last quarter. As of December 31, we had $183.1 million in inventory, which compared to $170.2 million at the end of Q3. The increase in inventory is primarily due to raw material purchases or increasing production. We made a total of $84 million in capital investments in the fourth quarter, which was mainly used for manufacturing capacity expansion for our 400G and 800G transceiver products.

In 2025, we made a total of $209 million in capital investments, which was above the CapEx projections we gave on our Q4 call last year of $120 million to $150 million for the full year. This was primarily due to increased customer demand projections. In Q4, the direct tariff impact on capital equipment was $3.1 million. As we have mentioned before, while we will continue to do our best to minimize any impacts. Tariff rates and equipment import mix may cause future results to vary materially. Notably, we source equipment from all over the world, including for both domestic and international locations.

Going forward, we believe we are well positioned for sustained growth across both our data center and CATV businesses, and the capital investments underway are expected to fundamentally strengthen the company as we execute on these opportunities. Given the recent surge in customer inquiries and apparent rising demand, we believe that by mid-2027, 100G and 400G revenue will be approximately $90 million. 800G revenue will be approximately $217 million and 1.6 terabit revenue will be approximately $71 million monthly. Altogether, this represents $378 million in monthly revenue for transceiver products. However, we believe that the customer demand is even larger than this. In order to accommodate this expected surge in demand, we plan to more than triple our laser manufacturing in Texas. We are evaluating our CapEx projections for 2026, and we intend to share those at a later date.

Moving now to our Q1 outlook. We expect Q1 revenue to be between $150 million and $165 million, accounting for a sequential increase in CATV revenue as well as the sequential increase in our data center revenue. We expect non-GAAP gross margin to be in the range of 29% to 31%. Non-GAAP net income is expected to be in the range of a loss of $7 million to a loss of $0.3 million and non-GAAP earnings per share between the loss of $0.09 per share and breakeven, using a weighted average basic share count of approximately 76.4 million shares.

Looking more broadly at 2026. While it's still early in the year, we expect to generate over $1 billion in revenue this year, with a non-GAAP operating profit of over $120 million. This revenue level is limited by our production capacity and supply chain, not market demand, which we believe is much larger. Based on our planned capacity additions, we expect to see continued strong sequential revenue growth in the first 2 quarters with an acceleration in the second half of the year as new production capacity comes online and additional customer qualifications are completed and orders begin to ship. We believe that this is an ambitious yet achievable target based upon our customers' forecast and what we know about the unprecedented investments that are being made in AI infrastructure.

With that, I will turn it back over to the operator for the Q&A session. Operator?

[Operator Instructions] We will take our first question, which will come from Simon Leopold with Raymond James.

First, just a very quick clarification, if I might. I missed the value you mentioned on 800 gig revenue, I know you had a little bit of a software firmware glitch and said it was below the guided for the $4 million or so. But what was the value of 800 gig in the quarter?

We didn't break out exactly, but it was below $4 million.

A lot below or a little below?

Below is delayed to Q1. But we have emphasized the revenue will be big in Q2 next year across the target. Our revenue in this year is $1 billion.

And I wanted to really focus on the trajectory for gross margin improvement. And I want to maybe first start with understanding how much of your laser production is in-house today versus external merchant lasers. And I guess I appreciate that typically, as you ramp production, there's sort of a learning curve of improving yields and things like that. So I'd like to make sure we have a good understanding of really the time line or trajectory to achieve that target, you mentioned of 40%. So sort of where are we now? And when do we -- what's the roadmap?

Because as I said, the gross margin for [ 1.6 ] is much, much higher than the other product. And I would say that, okay, as we say, I've seen by Q2 or the monthly revenue sometime in Q2, like June, July or something like that, the revenue is like $378 million. So it depends on the portion of 1.6 transceiver. So at the time, I believe the gross margin should be 35% to 38% overall gross margin for all the transceiver revenue. So I believe we're going to achieve 40% growth margin by late Q3 or Q4 next year.

By the way, Simon, just to make sure you're on the same page, the revenue figures that Thompson mentioned, were 2027 in Q2, not this year.

Yet, not this year. Yes, yes. It's very important, okay, not misstated.

Because I wrote down '26. So you anticipated my mistake.

Okay. It's impossible.

No, I appreciate that. And then before I pass maybe just a quick check in on the cable TV side of the business in that it sounds like you remain very confident in the trajectory However, the outlook offered by big cable operators was not as inspiring. Can you sort of help folks triangulate between the CapEx forecast and your involvement in cable TV upgrades?

Sure, Simon. I mean, I think as we've said consistently, where the money -- I mean, the overall CapEx numbers are one thing, but where the money gets spent is another thing. And I think they're spending this year and next year a significant amount of their spend is going towards the amplifiers, the outside plant, part of the network, and that's where we play. So there are some other parts of the network in the nose and other things that maybe are a little slower to ramp, although I think those are also ramping pretty significantly as well. So really you have to look at it on a kind of granular basis. The other thing is we've got a lot of new customers that we alluded to in the call earlier, and we'll start to see some significant contribution from those newer customers as well. So it's not just the 1 or 2 or 3 top largest MSOs, but also a wider swath of smaller companies that are contributing to our revenue.

And Simon I'll let you know, by end of this year, I would say more than 95% unless the because there's a huge issue of laser shortage and actually even some supplier to us, want to get it from them, we need to add at least 1 year or even longer. So that's why we announced we'll invest $300 million in Texas. The purpose as we said, we need to triple even more than triple our laser magnet capacity by Q2 next year, and that's to fulfill our transceiver demand. And let me emphasize, actually, transceiver demand is much bigger than what we projected. Right now, the number we said, the $378 million of transceiver revenue in June, July next year, not this year is limited by our capacity and the supply chain is not limited by the customer demand.

And our next question will come from Michael Genovese with Rosenblatt Securities.

So it sounds like instead of having a steady kind of ramp on this 800G, we're expecting to come in was really big numbers starting 2Q this year and then huge numbers next year by 2Q next year. I guess for the ramp in 2Q this year, could you just go over some of the milestones, maybe talk about the issue on the sub, but also the ongoing qualification milestones that you have to hit to kind of have that ramp in 2Q?

So as you mentioned, in order for the ramp to start in earnest, our 800G products have to be interoperable with all the different platforms that are out there at this particular customer. And there's a lot of them. So the firmware has to be modified to work with all those different platforms. So hardware-wise, everything is fine. No problem with that. Firmware is good on most of the platforms, we just have to make some tweaks to get it to work across all of the end platforms that they have. And that's really -- the customer and us have agreed that, that should be done in the middle of next month. So a couple of weeks, 3 weeks from now, something like that. And then that's basically the last hurdle to kind of unleash the ramp.

As we talked about, we've already started manufacturing products for that ramp. So from a manufacturing standpoint, we're gearing up just to touch on your -- the beginning part of that question, too, about the kind of nonlinearity of the ramp. That's because what you're seeing is our production capacity comes online, right? It's not gated by demand. It's gated by our ability to produce, and that doesn't come on in a linear fashion, right? You build a production line and you get a step function, not a smooth ramp.

Yes. So let me emphasize because every customer has some many different switch, transceiver supplier, so many different kind of ASIC. And when the customer adding more switch, the change the firmware. We are supposed to get a green light to ship already in December last year. The delay is not our problem or whatever. It's because how come build AI, the whole is much more complicated than before. That's why it takes much longer and do now, I think we feel very comfortable. And then I never said, okay, we have got almost 2 years of loading forecast from more than one customer, let me say that okay, 800G. And right now, let me say that more on customer at least 2 or even 3, they would like to buy all the transceiver we can make for 800G, 1.6T because AOI laser. And right now, it's limited by our capacity in the manpower and the supply chain. So that is why we are trying everything to ramp it take time. They take time. This say, right now this year, we said [ 1 billion ]. And let me say that demand is much, much bigger than [ 1 billion ]. But that's the number we feel comfortable.

At least we see a minimum 99% competent we can deliver. Otherwise, the raise much bigger than that way. Same thing for the 1.6T transceiver for the June, July next year is still limited by supply chain. So that's a lot of issue we need to solve. And the other thing I can tell you, in the short term, we should receive more than $100 million of 800G transceiver within a few months, maybe 1 month, 2 months for sure less than 3 months, we should receive more than $200 million or 1.6T transceiver.

Transceiver orders. We're not buying the transceivers, we're receiving the orders.

So for sure, that's how great is the market. But something is slot, okay? It's very comprise the whole is working great. This is good -- but we are never say that. All right. .

I got it. Sounds great. I guess is it fair to say that it sounds though when you say demand on the 800G side is already very high. Does that mean orders? I mean, do you have that level of orders already in for 800G?

Coming through at least from 2 customers because they want to make sure we commit to our promise. You not the problem, they will give us a loading forecast. But to make sure we guarantee what we promised, then sometimes we need to allow agreement and send down for sure they give us order, okay, at least by year or something like that.

Okay. Perfect. And then last question for me. On the 500,000 units, I think you said by second quarter next year, and I think that's 800G and 1.6...

[indiscernible].

End of this year, okay. Just the mix between Taiwan and the U.S. I mean it sounds like you're expanding capacity in both places. Is it, I guess, harder and more expensive to do it here, which is why only a quarter of the capacity will be here. Would you prefer to have more here if you -- if it was easier, like what's the decision making that on that where to do it?

Take much longer time to spend in U.S. takes us great since it's -- I would say the best in the U.S., let me say that. So -- but let me say by end of next year, I would say more than 55% will be manufactured in the U.S. or even 60% or 65% for 800G and 1.6T because that's why we just -- you know that we just got breaking a few weeks ago, it take time, we need more green more space, it's quite expensive and it take time to give the care they don't get heavy equipment, then we do a quite vacation training. It take time while it's catching up. All right? So the number will change a lot. Let me say that, more than 80%, 85% of investment will be in Texas.

Okay. Perfect. Really exciting, looking forward to following us more and seeing you guys at OFC.

And our next question will come from George Notter with Wolfe Research.

Really impressive conversation here in terms of the demand profile. I guess I'm just curious about what you're seeing on tariffs. Obviously, we've had some moves on tariffs recently, 15% across the board tariff. I'm not sure if transceivers are going to be exempt or what the situation is. But can you just talk about kind of the tariff situation, maybe the perception your customers have on tariffs and how that may or may not be translating into orders?

Yes. I guess there's 2 ways to say that. First of all, I mean, I think anybody that's telling you they confidently know exactly what the tariff situation is going to be throughout the year is probably not being truthful I certainly don't -- we have a viewpoint on obviously, on the current tariffs, it's pretty much in line with where we've been in terms of tariffs. I mean I don't expect -- if things stay the same as they are now, I don't expect it to dramatically change kind of the tariff picture that we outlined on the call earlier. That being said, we are looking at the pro in terms of the IEEPA tariffs. Those have been outlawed. So at least there's some pathway where we might be able to recoup some of those.

The other thing that I've said pretty consistently, and I think it ties in with Thompson's earlier comments, while it takes a while to build capacity in the United States, the one thing I could say is -- the one place where I'm pretty confident in saying it's not going to be tariff, it's a product that's made in the U.S., and that's what we're scaling up to do. So the more -- as time goes on, the more we can manufacture in the U.S. and the more that we can attract other supply chain partners, which we are doing to move their production to the U.S. as well. that will help us in the long term, that's going to be the solution for really minimize the tariff impact.

Got it. And then the comment about recouping tariffs. I think you said you were the shipper of record. Is there -- how much could that be? What is the potential upside you guys could get if indeed, you can recoup that?

I mean, sorry, if we could recoup all of it, we had about $4.6 million, I believe, just last quarter in tariffs. We probably paid last year $7 million or $8 million in tariffs overall. Again, we're still analyzing exactly how many of those are IEEPA related, not all tariffs that way. So there's a lot of nuance there, but I mean it's not going to dramatically change our picture, but -- but it certainly would be a welcome cash flow development for sure.

[Operator Instructions] Our next question will come from Ryan Koontz with Needham & Company.

Just maybe stepping back a little bit. As you think about the ramp in 400G with your large customer and 800G, which is pending, maybe can you compare kind of the production and demand view compare and contrast those 2 that gives you confidence in executing your own capacity and visibility from your customer for those 2 different product lines?

Sure. I mean, the 400G products, as we said, are going to continue -- I think there's going to be a continued strength in the sales of those products, driven by a couple of large customers, pretty much the same ones that we've already been shipping to, although we're seeing increased demand from at least one of those customers. But as we said in our prepared remarks earlier, 800G is expected to dominate those sales starting in Q2 of this year. So we'll see more revenue in 800G in Q2 than we did in 400G and then moving through the year and into next year, I think we're going to continue to see a very strong ramp in 800G because that's most closely associated with AI, right? That's the closest to the AI computing clusters at least until we get to 1.6 terabit later this year.

That's helpful. And then on your laser supply indium phosphide, we were down to your facility in the fall, where are you in terms of the equipment you need and kind of lead times with regards to expanding indium phosphide production? Any color you can give us there? And in terms of building out your new facilities and acquiring necessary equipment?

Yes. I mean, as Thompson mentioned, we're planning to triple our production of indium phosphide-related devices here, laser devices made on new phosphide materials by the middle part of next year. And we have line of sight into all that equipment. Some of it is already -- I mean, it would be a very long conversation to go through every piece of equipment and how -- what the schedule is. But the bottom line is, when we talk about tripling our capacity, that includes the equipment that we either have on order or have line of sight into order that will be delivered in time to accommodate that ramp. .

Great. And maybe just one last if I can squeeze in. In terms of cable TV, you mentioned another customer. I assume that's a large U.S. customer moving forward of 1.8 gigahertz here in terms of ES 4.0 East E.

Yes, it is. We have a number of customers. I would -- again, I want to caution, none of those customers are as large as our largest customer, okay? But in aggregate, I think there could be a significant contributor to the revenue, which is what I was trying to outline earlier in my response to Simon's question.

And our next question will come from Tim Savageaux with Northland Capital Markets.

A couple of questions I wanted to follow up on. It looks like given the increase in cable in Q1, you expect data center revenues up about $10 million. I guess what's driving that if you don't expect to 800 gig to ramp until Q2?

Well, I think we're going to see 2 things. We'll see a growth in 400 gig continued growth in 400 gig. And then, we're also -- we also do expect to see some revenue in 800 gig, just not the dramatic ramp that we expect to see starting in Q2.

Okay, great. So principally 400-gig. And you mentioned some, I guess, near-term gross margin headwinds driven by mix, I think you said, but you do have cable TV up in Q1. So I want to get a little more color on what's happening gross margin was there in Q1.

Yes. As we said, I mean, at the end of the day, if you look at our guidance, it's kind of a wash in terms of gross margin, we're seeing a little bit of headwind coming from the product mix, especially 400G, as I mentioned earlier, is going to continue to grow in Q1 until later when 800G started to take over. Meanwhile, in cable, gross margins there are better and they're actually expanding. So that's kind of the put and take on that. That's why it ended up being kind of a wash.

As we said, with the [indiscernible] 800G, we need time to fine-tune the production in the year. That's why this is early stage of volume mix future 800G -- that's what I say, by Q2 next year will be the overall gross margin will be 35% to 38% just for transceiver. By the end of next year, we believe we can achieve more than 40% to see gross margin for order transceiver by Q4 2027.

Okay. Maybe just 1.5 more here. You mentioned expectations for 800 gig to, I guess, dominate revenue? Trying to get a sense of what that means in Q2. You should have about in the $40 million range for 100 gig probably will be in the $40 million a quarter range for 400 gig. Would you expect 800 gig to be larger than both of those combined in Q2? Or how might you frame that?

What we're saying is it will be our largest segment within the data center. And the largest -- it will be the largest contributor to revenue of those 3 800 gig, 100 gig and 400.

I seem to be more than $25 million or $30 million, I don't know something -- as I said, -- the issue in the demand, okay? The Q1, the Q4 delays due to the firmware optimization, but Q2, Q3 is limited by our capacity, all right? It's not demand issues. Let me say that as we said, we got demand from a customer from 2 customers. Even I would say we got the order from that pretty soon, we use in a few weeks. Then the next issue is the supply chain and our manufacture capacity. So when I would say, I'm very common $25 million, but customers could be $35 billion to $40 billion. So this we just say what we see come right now. There is -- all the numbers we say here is not customer demand issue. It's AOI, manufacture and supply chain issue.

Okay. Great. And then I guess last question for me. Talk about the potential for $1 billion in revenue in calendar '26. I think, are in total, I wonder from a customer standpoint, I mean, would you -- I guess, how would you expect customer concentration to look in that scenario? I know you got a big guy on the cable side, I'm principally talking about data center. Do you think primary customer will be half of that or what have you?

So if you break down the revenue, right, if you just take a round number of $1 billion, right? So track the [ 300-ish ] that we have in cable TV, that gives you $700 million-ish left over. Right now, I would expect that's going to be dominated by -- most of that is going to be 2 large hyperscale customers. And they'll probably be roughly equivalent exiting the year. We'll see how that plays out. That's -- it's pretty early to say exactly how the timing on that is going to go. But I would expect at least 2 to be sort of comparable in size, let's put it that way. And then obviously, the third one that would be smaller in scale, still significant.

So I would say we were 2 hyperscale datacenter customers to be more than 10% or much more than 10% for the whole year.

And at this time, we have no further questions. I'll turn the call now over to Dr. Thompson Lin for any closing remarks.

Okay. And thank you for joining our call today. As always, we want to extend a thank you to our investors, customers and employees for your continued support we continue to believe the fundamental driver of long-term demand for our business remain robust, and we are uniquely positioned to deliver to drive value from those opportunities. We look forward to see many of you at upcoming investor conference as well OFC. Thank you.

The conference has now concluded. Thank you for attending today's presentation. You may now disconnect your lines.

Good afternoon. I will be your conference operator today. At this time, I would like to welcome everyone to Applied Optoelectronics Third Quarter 2025 Earnings Conference Call. [Operator Instructions] Please note, this call is being recorded.

I will now turn the call over to Lindsay Savarese, Investor Relations for Applied Optoelectronics. Ms. Savarese, you may begin.

Thank you. I'm Lindsay Savarese, Investor Relations for Applied Optoelectronics. I'm pleased to welcome you to AOI's Third Quarter 2025 Financial Results Conference Call. After the market closed today, AOI issued a press release announcing its third quarter 2025 financial results and provided its outlook for the fourth quarter of 2025. The release is also available on the company's website at ao-inc.com. This call is being recorded and webcast live. A link to the recording can be found on the Investor Relations section of the AOI website and will be archived for 1 year.

Joining us on today's call is Dr. Thompson Lin, AOI's Founder, Chairman and CEO; and Dr. Stefan Murry, AOI's Chief Financial Officer and Chief Strategy Officer. Thompson will give an overview of AOI's Q3 results, and Stefan will provide financial details and the outlook for the fourth quarter of 2025. A question-and-answer session will follow our prepared remarks.

Before we begin, I would like to remind you to review AOI's safe harbor statement. On today's call, management will make forward-looking statements. These forward-looking statements involve risks and uncertainties as well as assumptions and current expectations, which could cause the company's actual results, levels of activity, performance or achievements of the company or its industry to differ materially from those expressed or implied in such forward-looking statements.

In some cases, you can identify forward-looking statements by terminology such as believes, forecast, anticipate, estimates, suggests, intends, predicts, expects, plans, may, should, could, would, will, potential or think or by the negative of those terms or other similar expressions that convey uncertainty of future events or outcomes. The company has based these forward-looking statements on its current expectations, assumptions, estimates and projections. While the company believes these expectations, assumptions, estimates and projections are reasonable, such forward-looking statements are only predictions and involve known and unknown risks and uncertainties, many of which are beyond the company's control.

Forward-looking statements also include statements regarding management's beliefs and expectations related to the expansion of the reach of its products into new markets and customer responses to its innovations as well as statements regarding the company's outlook for the fourth quarter of 2025. Except as required by law, AOI assumes no obligation to update these forward-looking statements for any reason after the date of this earnings call to conform these statements to actual results or to changes in the company's expectations. More information about other risks that may impact the company's business are set forth in the Risk Factors section of AOI's reports on file with the SEC, including the company's annual report on Form 10-K and quarterly reports on Form 10-Q.

Also, all financial results and other financial measures discussed today are on a non-GAAP basis unless specifically noted otherwise. Non-GAAP financial measures are not intended to be considered in isolation or as a substitute for results prepared in accordance with GAAP. A reconciliation between our GAAP and non-GAAP measures as well as a discussion of why we present non-GAAP financial measures are included in the company's earnings press release that is available on AOI's website.

Before moving to the financial results, I'd like to note that the date of AOI's fourth quarter and full year 2025 earnings call is currently scheduled for February 26, 2026.

Now I would like to turn the call over to Dr. Thompson Lin, AOI's Founder, Chairman and CEO. Thompson?

Thank you, Lindsay, and thank you for joining our call today. We successfully deliver revenue, gross margin and [indiscernible] per share in line with our expectations. [indiscernible] recorded the highest quarterly revenue in our history, driven by strong demand [indiscernible] achieved revenue in the third quarter. The strength we saw in our CATV business more than offset our [indiscernible] revenue is trending [indiscernible] with our expectations, largely due to the timing of certain shipment at quarter end.

In [indiscernible] were approximately $6.6 million in shipping our of [ 400G ] transceiver to a large [indiscernible] customer, who was not able to return into revenue during the quarter due to very shipping and receiving delays and [indiscernible] in Q4. Despite this delay, our financial results clearly highlight the range of having diversified revenue stream as a result of a revenue on a combined basis increased 15% sequentially and [ 22% ] year-over-year.

Probably, we continue to make profits on customer colocation [indiscernible]. As we mentioned last quarter, we believe we are [indiscernible] the final stage of qualifications with several customers. We expect qualifications in the near-term [indiscernible] that we are having with our customers, and we continue to believe that we will produce meaningful shipment [indiscernible] in the fourth quarter. [indiscernible] would be the revenue of [ $118.6 ] million [indiscernible] in line with that kind of a range of $150 million to $127 million. We recorded non-GAAP gross margin of 31%. This was in line with [indiscernible] range of 29.5% to 31%. And on [indiscernible] $0.09 [indiscernible] in line with our guided range or a loss of [ $0.10 ] to a loss of [ $013 ].

Total revenue for our [indiscernible] of $43.9 million increased 7% year-over-year, but was down sequentially. Revenue for our [indiscernible] products includes 32% year-over-year, while revenue for our [indiscernible] product was [indiscernible] year-over-year or $7.1 million. [indiscernible] due to the timing of certain shipment this quarter and [indiscernible].

Total revenue in Q3 in our [ DSD ] [indiscernible] was a [indiscernible] [ $17.6 million ] with more than triple year-over-year and was up 26% sequentially from a strong Q2. This increase is due to the continued ramp in orders for our [indiscernible] products for both existing as well as new customers.

With that, I will turn the call over to Stefan to review the details of our Q3 performance and outlook for Q4. Stefan?

Thank you, Thompson. As Thompson mentioned, we successfully delivered revenue, gross margin and a non-GAAP loss per share in line with our expectations. In fact, we recorded the highest quarterly revenue in our history, driven by strong demand in the CATV market, which also achieved record revenue in the third quarter. The strength that we saw in our CATV business more than offset our data center revenue, which came in a touch below our expectations, largely due to the timing of certain shipments at quarter end.

In particular, we had approximately $6.6 million in shipments of 400G transceivers to a large hyperscale customer, which was not able to be turned into revenue during the quarter due to various shipping and receiving delays and which we have booked in Q4. Despite this delay, our financial results clearly highlight the advantage of having diversified revenue streams. As a result, our total revenue on a combined basis increased 15% sequentially and 82% year-over-year. In Q3, we delivered revenue of $118.6 million, which was in line with our guidance range of $115 million to $127 million. We recorded non-GAAP gross margin of 31%, which was in line with our guidance range of 29.5% to 31%.

Our non-GAAP loss per share of $0.09 was also in line with our guidance range of a loss of $0.10 to a loss of $0.03. We continue to make progress on customer qualifications on our 800G products. As we mentioned last quarter, we believe we are near the final stages of qualification with several customers. We expect qualification in the near term based on conversations that we are having with our customers, and we continue to believe that we will produce meaningful shipments of 800G products in the fourth quarter.

For the third quarter in a row, we did record immaterial revenue for our 800G products related to deliveries for customer qualification activity. As we have mentioned before, our schedule on ramping up our production is largely constrained by our ability to build and qualify production capacity. On that front, we are pleased to report that we made good progress on getting our production ready during the third quarter and remain nearly on track to achieve the targets that we laid out at OFC.

As a reminder, we expect this will culminate later this year with what we believe will be the largest domestic production capacity for 800G or 1.6 terabit transceivers, approximately 35,000 transceivers per month or roughly 35% of our overall capacity for these advanced optical transceivers. Notably, we will be able to accommodate this expansion in our current Texas facility footprint. Further, by mid-2026, we continue to expect to be able to produce over 200,000 pieces per month with the majority produced in Texas. As I just mentioned, we made solid headway towards these targets for next year.

As you may have seen, we announced last week that we signed an agreement to lease an additional building in [ Sugarland ], Texas. We will begin construction on this new facility later this year and are confident in our ability to scale our production towards the middle to end of next year to achieve our 2026 targets.

It's also important to note that AOI has had an in-house laser manufacturing capability for many years, and we have been expanding and improving this capability recently. While we have heard talks about laser shortages, having a laser production capability in-house gives us an advantage. And to date, we have not experienced a shortage of lasers that has affected our ability to deliver products according to our customers' requests. We've spent years developing our automated manufacturing capabilities, which gives us an advantage in the ability to do manufacturing virtually anywhere in the world that we would like to and makes building out another facility in a cost-effective way in Texas possible.

The message from our customers is consistent. Many of them have a strong preference for production in North America. And so that's what we have been and are currently focused on. When we talk about adding capacity, the lead time for us to add new equipment and add machinery to our production process is typically less than the lead time it would take to hire and train the types of skilled operators that are needed to do the manual processes that are used by most of our competitors.

Just to reiterate, we currently have 3 manufacturing sites, one here in [ SugarLand] , Texas, where our headquarters is and which will soon involve 2 facilities, 1 in [ Ningbo] , China and 2 in Taipei, Taiwan with an additional one under construction. As you may have heard me say at OFC, we expect to increase the total production of 800G and 1.6 terabit products by 8.5x by the end of the year, and we are on track and dedicated to achieving this goal.

During the third quarter, direct tariffs had a $1.1 million impact on our income statement. As it relates to tariffs, also, as I mentioned on our prior couple of earnings calls, while we do utilize some imported components in our transceivers, many key components like our laser chips are already manufactured in the U.S. Importantly, in our 800G and 1.6 terabit transceiver designs, less than 10% of the value of the components used is currently sourced from China, and we have a pathway as we scale production to further reducing this China content, ultimately to near 0.

We are also in discussion with several key suppliers about onshoring their production to the U.S. to support a robust domestic supply chain.

Turning to our third quarter results. Our total revenue was $118.6 million, which increased 82% year-over-year and increased 15% sequentially off a strong Q2 and was in line with our guidance range of $115 million to $127 million. During the third quarter, [ 60% ] of revenue was from CATV products, 37% was from data center products, with the remaining 3% from FTTH, telecom and other.

In our data center business, Q3 revenue came in at $43.9 million, which was up 7% year-over-year and was down 2% sequentially. Sales of our 100G products increased 32% year-over-year, while sales for our 400G products decreased 65% year-over-year or $7.1 million, which was primarily driven by the timing of certain shipments at quarter end that I previously discussed. In the third quarter, 83% of data center revenue was from 100G products, 9% was from 200G and 400G transceiver products and 7% was from 10G and 40G transceiver products.

Looking ahead to Q4, we expect a substantial sequential increase in our data center revenue, driven by growth in 400G revenue as well as layering in some increased 800G revenue.

In our CATV business, we saw exceptionally strong demand in Q3. CATV revenue in the third quarter was a record $70.6 million, which more than tripled year-over-year and was up 26% sequentially from a strong Q2. This increase is due to the continued ramp in orders for our 1.8 gigahertz amplifier products. Similar to last quarter, we shipped a significant quantity of 1.8 gigahertz amplifiers to Charter in the quarter and demand continues to be robust. On our last earnings call, we had discussed how in addition to Charter, we had 6 other MSO customers who had already begun to order and deploy our 1.8 gigahertz products or are in various stages of qualification of these products. We were pleased to see continued momentum with these new customers and are excited to see the broad-based appeal of our amplifiers and [ QuantumLink ] software.

During the quarter, we announced the addition of 4 new software modules to our [ Quantum Link HFC ] remote management solution which offers our customers actionable intelligence to optimize network performance, reduce operational costs and improve the broadband experience. The new suite of software modules are add-ons to our existing [ Quantum Link ] Central, providing telemetry, adding unified visibility, predictive diagnostics and automated controls to our remote amplifier management platform. Most software features will be available this quarter. The feedback we are hearing from our customers is very positive.

In September, we attended the Society of Cable Telecommunications Engineering Expo. We had great interactions with customers and potential customers during the expo. And as I just mentioned, feedback from our customers continues to be very positive with many noting that our amplifiers are groundbreaking in terms of performance, ease of setup and control and monitoring capabilities. As cable operators prepare for substantial upgrades to their infrastructure to meet increased spectrum and bandwidth demand, it's clear that the deployment of next-generation amplifiers and related equipment has become essential. Looking ahead to Q4, we expect strength in our CATV business to continue, although we expect revenue in this business to moderate to between $50 million and $55 million next quarter, following this quarter's exceptionally strong results.

Now turning to our Telecom segment. Revenue from our Telecom products of $3.7 million was up 34% year-over-year and 93% sequentially. As we have said before, we expect telecom sales to fluctuate from quarter-to-quarter. For the third quarter, our top 10 customers represented 97% of revenue, up from 96% in Q3 of last year. We had 2 greater than 10% customers, one in the CATV market, which contributed 66% of total revenue and one in the data center market, which contributed 24% of total revenue.

In Q3, we generated non-GAAP gross margin of 31%, which was in line with our guidance range of 29.5% to 31%, and was up from 25% in Q3 2024 and compared to 30.4% in Q2 2025. The year-over-year increase in our gross margin was driven primarily by our favorable product mix. Looking ahead, we expect continued gradual improvement in gross margin although we expect that the revenue mix in data center in the next few quarters will be a slight headwind. We remain committed to our long-term goal of returning our non-GAAP gross margin to around 40%. The progress we have made so far demonstrates that we're on the right track, and we continue to believe that this goal is achievable.

The revenue figures presented above are net of a contra revenue amount due to the accounting for warrants provided to customers. As a reminder, this amounts to approximately 2.5% of revenue derived from certain customers to whom AOI has provided warrants in exchange for future revenue. In Q3, the amount of this contra revenue was immaterial at about $50,000.

Total non-GAAP operating expenses in the third quarter were $47.1 million or 40% of revenue, which compared to $27.9 million or 43% of revenue in Q3 of the prior year. While operating expenses increased this quarter and were a bit higher than our forecast, this rise was largely driven by increased shipping costs related to increased business activity in our CATV business this quarter. Looking ahead, we expect non-GAAP operating expenses to be in the range of $48 million to $50 million per quarter.

Non-GAAP operating loss in the third quarter was $10.3 million compared to an operating loss of $11.7 million in Q3 of the prior year. GAAP net loss for Q3 was $17.9 million or a loss of $0.28 per basic share compared with a GAAP net loss of $17.8 million or a loss of $0.42 per basic share in Q3 of 2024. On a non-GAAP basis, net loss for Q3 was $5.4 million or $0.09 per share, which was in line with our guidance range of a loss of $5.9 million to a loss of $2 million or non-GAAP income per share in the range of a loss of $0.10 to a loss of $0.03. This compares to a non-GAAP net loss of [ $8.8 million ] or $0.21 per share in Q3 of the prior year. The basic shares outstanding used for computing the earnings per share in Q3 were $63.3 million.

Turning now to the balance sheet. We ended the third quarter with $150.7 million in total cash, cash equivalents, short-term investments and restricted cash. This compares with $87.2 million at the end of the second quarter of 2025. We ended the third quarter with total debt, excluding convertible debt of $62 million compared to $54.3 million at the end of last quarter. As I mentioned on our prior earnings call, earlier this year, we announced a revolving loan facility with BOK Financial of $35 million, which we intend to use to meet some of our working capital needs going forward. As of September 30, we had $170.2 million in inventory, which compared to $138.9 million at the end of Q2. This increase in inventory is almost entirely due to purchases of raw materials to be used in production of our products over the next several months.

During the quarter, we initiated a new ATM program. We completed this program during the quarter, raising $147 million net of commissions and fees, which we intend to use mainly for new equipment and machinery for production and research and development use, including the earlier mentioned production expansion in Texas. We made a total of $49.9 million in capital investments in the third quarter, which was mainly used for manufacturing capacity expansion for our 400G and 800G transceiver products. On our last few earnings calls, we have discussed our plans to make sizable CapEx investments over the next several quarters as we prepare for increased 400G, 800G and 1.6 terabit data center production in 2025.

To date, this year, we have made a total of $124.9 million in capital investments and we are tracking at or above our CapEx projections we gave earlier this year of $120 million to $150 million in total CapEx. We have noted on our prior couple of earnings calls that these costs could be impacted from tariffs, but that given the evolving nature, it is difficult to predict what type of impact or by how much.

In Q3, the direct tariff impact on capital equipment was $1.9 million or roughly 4%, but tariff rates and equipment import mix may cause future results to vary materially. We source equipment from all over the world, including both from domestic and international locations. We have and will continue to do our best to minimize any impacts. It's clear that U.S.-based production is a priority for our customers, and we remain fully committed to expanding our capacity to meet that demand.

Moving now to our Q4 outlook. We expect Q4 revenue to be between $125 million and $140 million, accounting for a sequential decrease in CATV revenue as well as a more substantial sequential increase in our data center revenue. We expect non-GAAP gross margin to be in the range of 29% to 31%. Non-GAAP net income is expected to be in the range of a loss of $9 million to a loss of $2.8 million and non-GAAP earnings per share between a loss of $0.13 per share and a loss of $0.04 per share using a weighted average basic share count of approximately 70.3 million shares.

With that, I will turn it back over to the operator for the Q&A session. Operator?

[Operator Instructions] Our first question comes from Simon Leopold of Raymond James.

I'm going to ask two and start with the cable TV side. So clearly, a strong blowout number here this quarter. So the moderation makes sense. And I guess where I'd like to go is to understand how you're thinking about the broader outlook for CATV in that I recall last quarter, we talked about the potential to do over $300 million in 2026. If we sort of run rate out what you're doing, you're certainly on that trajectory. But I want to assess this given the lumpy nature of cable TV.

Yes, thanks for bringing that up. So yes, I mean, we do think $300 million plus in cable TV revenue is still achievable next year. As you pointed out, we're kind of approaching a run rate there in this quarter. What I think is significant to point out, though, and we pointed this out on the last earnings call as well, that a lot of that growth is going to come from new products that we've announced, and we discussed in our prepared remarks a few minutes ago about the great success that we had at the Society of Cable Telecommunications Engineering Show, showcasing some of our new products, including the software products that we highlighted.

So yes, I think the $300 million plus mark is achievable next year. However, it's not likely to come just from the amplifier products, although again, we expect strong results in the amplifiers, but the additional revenue that we expect to see from those other products should get us up to that $300 million mark.

As we said in the script, we expect the cable TV revenue in Q4 will be reduced to maybe $50 million to $55 million. So that means the data center growth should be a lot, okay? Since the revenue increased by about [ 10% ] compared to Q3. So revenue will increase by $25 million to $40 million in Q4. Because...

Thanks. And then...

Yes, sorry. Go ahead.

Yes. No. So that's why I wanted to follow up on the data center, particularly around your comment about 400 and 800 gig being up, given 800 gig is key right now, I'd like to unpack that a little bit because I don't think you've announced certifications, qualifications on 800 gig yet. It sounds like that's somewhat imminent but I don't want to over interpret.

So maybe just drill down specifically to how you think about 800 gig in that 4Q? And then, of course, how should we think about the timing of when to start thinking about 1.6T? I understand that's not in 4Q, but should we be thinking about that for next year?

Yes. So 800 gig, we do expect meaningful shipments in the fourth quarter, as we said in our prepared remarks. Now the growth in Q4 is largely going to come from 400 gig, but we do expect meaningful revenue from 800 gig in this quarter. And as you pointed out, that would require product qualification to be pretty imminent, which is what we believe.

With respect to 1.6 terabits, yes, we do think that we'll see revenue from 1.6 terabit later next year, but it's not going to be a factor in Q4, as you pointed on, and probably not in the first half of next year.

[indiscernible] especially [indiscernible]. The 1.6T [indiscernible] transceiver we have right now would say [indiscernible] So we are working very hard [indiscernible] by end of this year early next year. [indiscernible] more like I would say, June, July next year for 1.6T. And we have several [indiscernible] products for 1.6T [indiscernible] transceiver and we will announced previously [indiscernible].

Also, Simon, I just want to reiterate something that we've mentioned repeatedly and talked about a little bit on the call. But just for clarity, the factory that we're building both here as well as the increased capacity that we've been adding in Taiwan is capable of manufacturing both 800G and 1.6 terabit on the same production line. The only difference really is in the final testing in terms of the type of equipment that we need.

The next question comes from George Notter of Wolfe Research.

I was just curious if you could tell us more about the shipping and receiving delay at the end of the quarter. I'm just wondering what that was. And can you confirm that it was a single customer? Was it multiple customers? Any insights there would be great. And I've got a follow-up, too.

Sure. It was a single customer, a single hyperscale customer, which is a relatively new customer addition for us. And as a result of that, some of the shipments at the end of the quarter, I can't go into too many details because it's obviously nondisclosure agreements and such. But let's just say that not all the systems that were -- all the inventory management systems and all that have been properly configured at that point to be able to receive those goods in time for us to book them as revenue in the third quarter.

So we resolved that in the first few days of the fourth quarter and have booked that revenue since then. So it wasn't anything that we expect to recur or anything like that. It was just kind of unique to this, I would say, sort of start-up business, if you will, with this particular large hyperscale customer.

Got it. Okay. I'm sorry. So the products were delivered to the customer, but it sounds like ownership couldn't transition because it hadn't been through their inventory management system. Is that the right view?

Yes, basically. I mean there's a system integrator that's involved, again, without going into too many details, but it essentially comes down to just a timing issue with the computer systems on all sides that need to be [indiscernible].

Okay. Got it. And then can you give us an update on the capital spend? I mean you said you're tracking ahead of the $120 million to $150 million for the year. What does that look like now when you layer in Q4? And then how about 2026? Do you have an initial view on what CapEx would look like next year?

Yes. So we don't have -- a lot of the Q4 guidance comes down to sort of timing on when we're going to receive a lot of the equipment. So we're still looking at that. It's probably going to be ahead of that $150 million top end that we said before, but it's unclear at this point exactly how much of that equipment will really be able to be delivered in this quarter. So we'll get back to you on that.

Similarly, for 2026, we're still working on the CapEx plans for 2026. So I would expect it to be above what we're seeing in 2025, but I don't have a precise number yet on that. We're still going through this time.

[Operator Instructions] And our next question comes from Michael Genovese of Rosenblatt Securities.

I guess for 400G with that customer becoming a run rate business. And if I'm not mistaken, I was thinking about 100,000 units per month. So maybe you could update on that. But is that the right way to think about it? And is it getting there in the fourth quarter? And then we should think about that customer being at the same level of 400G per quarter all year in 2026. Am I thinking about that the right way?

So it is approaching -- I mean, it is on track to becoming sort of a run rate business, as you mentioned, that is -- when I hear the term run rate business, so I'm assuming you mean is sort of capacity limited, right? That is we can -- we'll be selling a relatively consistent amount every quarter based on our capacity. So it's on -- it is moving in that direction. We will not be fully at capacity in Q4, not to mention the fact that we're continuing to add some capacity, especially in Taiwan, like we talked about earlier. So it won't reach its maximum potential in Q4 by any means, but it will be a meaningful contributor to revenue.

As Thompson mentioned earlier, if you think about the guidance that we gave, right, cable TV has an implied decline of, let's just say, $15 million roughly, give or take, while the overall revenue is going to be up roughly $15 million, again, give or take, within the ranges that we specified. And so the rest of that growth is going to come from data center. And most of that is going to come from 400G as we discuss earlier. A little bit of 800G, but not a lot. Most of it is going to come from 400G.

[indiscernible] limited our capacity. So right now, for 400G single transceiver in Q4, we can maybe close to 60,000 per month. Then Q2, [indiscernible] to 110,000 to 120,000 per month. So it depends on our capacity [indiscernible] to spend [indiscernible] Taiwan and U.S. The other [indiscernible] laser capacity. it's very important as you know, it's a short news we have laser capacity. That's why AI right now is doing 3-inch, 4-inch next year. At the same time, not to mention the other laser our 25G [indiscernible] for the silicon photonics, we are talking about our target by December next year is at least more than 2 million per month. Where we spend all the CapEx, [indiscernible] all kind of stuff. And for sure, we are working on 6-inch wafer, but it will be more like a [indiscernible] project. But I think 4-inch probably is ongoing and pretty...

Okay. Great. And then it sounds like you've got pretty high confidence of an 800G qualification coming soon if you're putting some in the fourth quarter guidance. So I just want to double click on that confidence. But then also if we just back up 3 months ago, is this process going the way you expected? I know matter of a couple of weeks on either side of no big deal, but did you think you have it by now? Or is this kind of going the way you thought it would go?

I think we should get pretty soon. We will send [indiscernible] 3 to 4 weeks or even sooner. [indiscernible] So finally, they get into volume, but really not [indiscernible] because [ the 150,000 per month or even 200 ] a month. But now we are talking about maybe 10,000, 20,000. So it's not a of [ 100,000 ] per month. By end of [indiscernible], [ 100,000 per month]. By end of June next year we have 20,000 per month. We spend money is based on customer commitment, okay? It's not based on our wish list, let me say that, okay? And don't forget especially in U.S. [indiscernible] space, the capacity, [indiscernible] money. [indiscernible] based on very strong commitment [indiscernible] order...

[indiscernible] directly answer your question about relative to expectations. I think we said that we expected the qualifications to be coming in, in the late Q3 or early Q4. And so we're still in that range, but I would consider what we expected and we're basically still on track for that schedule.

Okay. If I can ask another, what should 100G be doing in '26 versus '25? And just when you compare the ASP of 100G to 800G are we at an 8x multiple? Or is it even higher than that for the ASP of 800 versus 100?

It's not an 8x multiple. It's less than that. With respect to what 800G -- or sorry, what 100G will do next year, I think it's going to be pretty consistent. I don't see a big falloff for sure. It could even go up a little bit. There continue to be new deployments of 100G. So -- but I think the best scenario that I would model in is sort of a flat 100G business next year.

Okay. If I could just ask one more. How are you guys feeling about the -- with the CapEx plans and the expansion plans, you've gone to the market a few times this year. Are you at a good place for your spending next year? Or do you think you're going to have to do more fundraising?

I think we're going to continue to raise the capital that we need to fund the CapEx. I mean our plans continue to expand. What we're hearing from our customers is more and more bullish in terms of the volume that they need and particularly the volume that they'd like to have out of U.S.-based factories, which we don't have yet. I mean we just announced the lease a week or 2 ago of the new facility, which still has yet to be built out. So that's basically a 2026 event. So we're going to continue to add capacity as we see that demand coming from our customers, and it's very strong right now.

But let me say point. One, we have some discussion with 1 or 2 major customers for the U.S. capacity. I think maybe customers will invest AOI maybe [ $200 million ] under discussion. Number two, we are working very close [indiscernible] fund rasing some support, including the U.S. government [indiscernible]. So I think maybe we can get some good money from both Texas and the U.S. government. But number three, don't forget next year, we should be profitable quite a lot. I would say no surprise, our net profit should be more than $150 million next year or even higher. So some of the expansion can be paid by our profit.

The next question comes from Ryan Koontz of Needham & Co.

On the follow-up on the transceiver products here and it sounds like you guys are doing well in silicon photonics. I wanted to ask your view on kind of the macro of [ Sipho ] versus EML versus VCSELs and what you're hearing from your customers about their interest as the data rates move up here to 800 to 1.6?

Well, I would say, first of all, what we're hearing from our customers may just be a function of where they view us in terms of our technological capabilities and what they need. So that is to say, I'm not saying that, for example, when I tell you that our customers like [ Sipho ], they like our [ SiPho ] solution for sure. That's not to imply that they're going to switch all their products over to [ Sipho ]. They have other vendors that are working with EMLs, for example.

So -- but all that being said, I think broadly speaking, I think [ Sipho ] is seen as a technology that has more scalability in terms of its ability to go to higher data rates in the future. I think we're at the early stages of implementing silicon photonics in terms of volume manufacturing and all that. So it's going to take some time for them to become sort of comfortable and let that technology ramp up, but it certainly has more legs in terms of higher data rate than...

Especially you need less laser for [ Sipho ] [indiscernible] a very serious problem shortage of lasers, especially for EML, okay? Not to Mention 200G [indiscernible] EML. So for EML, [indiscernible].

Great. And maybe shifting gears to cable. How are you feeling about share there at your larger customers? Do you feel like the uptick in demand here for cable is this share gain? Is this higher deployment rates? Any view on how you feel about share versus customer spend?

I would say it's share gain primarily. We the customers' plans continue to evolve. But as I mentioned, we've had some very, very successful interaction with our major customers, including the larger MSOs like we talked about with Charter and others, but also with a number of smaller operators. The Society of Cabled Telecommunications Engineers show that we were at was really very, very positive for us. So I think we're taking slacks that could have potentially gone somewhere else and gaining that share.

So besides Charter, we have 6 other customers, and we sized order from 2 of them. So total, we have 7 customers right now for cable TV, 1.8 gigahertz, okay, including 1.2 gigahertz. But next year, we're talking about another 10. So that by end of next year, [indiscernible] 17 [indiscernible] in North America, Latin, Australia, even Asia. So now [indiscernible] customer...

Great. And when you talked about the new products coming in cable, are you referring to nodes or the software products for the amps that you mentioned earlier?

Both. Yes.

Don't forget software is pretty good. That's very important quantum customers really like a lot of problem of the customer. They solve a lot of issues. They can take a lot of operating expense, and that's why they like it. That's why I would say, become the #1 supplier in cable TV. It's only hardware, but integration of hardware and software and system.

[Operator Instructions] Our next question will come from Tim Savageaux of Northland Capital Markets.

A couple of questions, but I want to start with what we've been hearing pretty much all week here is about a pretty dramatic kind of step function increase and really across a lot of the different areas in AI optical, including inside the data center for modules. Focused on 1.6 to some degree, but pretty broad-based seeing.

My first question is, are you seeing that in terms of your conversations with customers about overall levels of transceiver demand, just I don't know, in the last 4 to 6 weeks.

Just look at our I'm sorry, I cut off there. I didn't hear the first part of your question.

No, all good. Go ahead. Sorry.

Yes. So yes, I mean, we're seeing very strong increase in demand. If you look at our guidance, again, just kind of go back to the segment guidance that we gave, it implies a dramatic ramp in data center revenue in the fourth quarter. And we didn't give annual guidance for next year, but we certainly believe that's the beginning of a sustained ramp. So I think we're exactly in sync with what you described. We're seeing that ramp first at 800 gig. But as we talked about later next year, we expect 1.6 to be a strong contributor as well.

Does that answer your question, Tim?

Yes. And I wanted to follow up on your capacity targets exiting the year I think at 100,000 units a month. And Thompson had mentioned before commitments from customers, I guess. And I want to kind of dig into that a little bit more, which is would you be in a position to ship that full -- given there's -- we're running out a year a little bit here, but would you be in a position to ship that full capacity in the first quarter? And do you have either orders on hand, commitments, however you want to describe it to kind of cover those type of volumes starting in Q1 next year?

I would say more like Q2. Don't forget the Chinese New Year and the cycle time is 1.5 months. So we got all ready, we are doing the pilot right now, both in Taiwan and U.S. [indiscernible] going to maybe 90,000 to 100,000 pieces per month of revenue, more like Q2.

And to answer your first question, right now [indiscernible] crazy number, okay? A total more than 300,000 of 800 [indiscernible]. So for sure, we spend money until we got the commitment. So yes, it's true. Right now, all the hyperscale data center customers are really serious. So it's not [indiscernible].

Tim, I just wanted to touch on one thing that your question asked earlier. I want to make sure we're on the same page. So you mentioned the capacity of 100,000 per month. That is our 800 gig or 1.6 terabit, but again, 800 gig primarily this year capacity. In addition to that, we also have capacity for 400 gig. Those are not shared, right? So the 400-gig capacity, as Thompson mentioned, should be 120,000 pieces or more early next year. And we do have customer commitments that would cover that.

At this time, we have no further questions, and I will turn the call back over to Dr. Thompson Lin for closing remarks.

Again, thank you for joining us today. As always, we want to extend a thank you to our investors, customers and employees for your continued support. We continue to believe the fundamental driver of long-term demand for our business remain robust, and we are unique position to drive value from these opportunities. We look forward to welcome some of you to our Texas factory tour next week and seeing many of you at the upcoming investor conference. Thank you.

The conference has now concluded. Thank you for attending today's presentation, and you may now disconnect.