Welcome to day two of Citi's 2026 Global TMT Conference. My name is Kelsey, one of the analysts here covering U.S. semis. Very pleased to have Steve Litchfield with us, from MaxLinear, CFO. Perhaps, Steve, how about we start with giving the audience an overview of the company's heritage, product portfolio, and also how MaxLinear is benefiting from the AI compute build-out.
Great. Sounds good. Well, thanks for having us, Kelsey, and thanks for seeing everybody joining us today. A little bit of background on MaxLinear. The company's really founded as a mixed signal semiconductor company. We started an effort in infrastructure about five to six years ago, and that's been a labor of love in the beginning, and we really started to get some solid traction here in the market over the last few years. This most recent quarter, it was over 50% of our revenues. That's been the focus of our data center business, as well as our storage accelerators and some of our wireless infrastructure. We also have several other end markets, broadband connectivity and industrial markets. But over the last several years, as far as R&D dollars and product focus, it's been around the data center and the infrastructure business.
How should we think about the tailwinds within the infrastructure build-out? Is it the 400G, 800G, 1.6T transition or scale-up, scale-out? How are you guys positioned within the networking transition?
Sure. We're participating in all those scale up and scale out. Let's maybe step back a little bit and talk about where revenue comes from today and then where it's going. This year, the mix between 400G and 800G, primarily DSPs. That business is growing nicely this year. The midpoint of our guidance for the optical piece of it is $220 million. The mix of that business today is probably 70% 800G and probably 30% 400G. Getting to the end of the life on the 400G. I think most of the market data says that will slow down next year, but 800G, I feel like will grow nicely this year as well as next year. Then some debate on what happens in 2028.
Our new product, Rushmore, customers are qualifying that product, working with those solutions today, expected to go to market mid-next year. We're working with tons of customers right now with that particular ramp. If it starts mid-next year, it probably goes into full scale production in the following year. I guess one thing I would add there is we also have our Washington TIA that sells alongside of that. Just from a pure ASP standpoint, if you look at 800G, most of the market data will say ASPs are running, I don't know, $35- $50 probably, versus a 1.6T, which is probably $80- $100 plus the cost of the TIA, which pushes it above $100. Right now, the market environment is such that pricing is holding very good.
I don't expect that to change right now, given the constraints that we're seeing throughout next year. I don't anticipate pricing coming down much, between now and the end of next year anyway.
On the optical data center piece, you guys have raised guidance, I believe the last three quarters. How should we think about that trajectory? Or what changed during this period that resulted in that consistently higher guidance over time?
Right. No, it's a good question. Yeah, because going into the year, I think we had said that we'd be north of $100, and then each quarter we've raised. So what's changed? I think it's a few things. The market continues to grow. You guys have all probably seen the growth in transceivers this year versus the beginning of the year. So that's clearly one. But I think our share that we were anticipating has gone up more than probably what we had expected. We've seen new customers come on that were not planned for. I just think it's a handful of things. I don't think it's one in particular, but several things.
We've also tried to be a little bit conservative entering the year, just from the standpoint that with supply constraints being what they are, we don't want to over-commit to investors, nor do we want to over-commit to customers. I think we've tried to be just conservative on those expectations.
How has the backlog visibility improved since earnings or even the last two quarters?
Sure. Look, visibility is very good. Again, back to, I think there's a big gap between supply and demand here. Visibility is strong from the standpoint, customers are coming in. We quote 26- 28 week lead times for our optical products. That's fairly consistent, but I will say that customers have come in placing orders throughout all of next year, or at least at a minimum, given us forecasts so that we can go out and plan. As all of you know, TSMC and all of the other folks in the supply chain need forecasts, need planning. In order to secure that capacity for next year, we need to see those orders upfront. You're in an environment where you're getting those orders right now.
Got it. Could you talk a little bit about MaxLinear differentiation? In the DSP market, it is dominated by a couple of players. How do you see that competitive positioning change or evolve as the market transitions to 1.6T?
Sure. Yeah, look, there's two big players in this market that we've competed against for a long time. I started out talking about our mixed signal expertise, but that expertise is really what allows us to bring technically unique solutions that are very differentiated. Our current 800G solution runs about 20% less power than a comparable module from one or the other of those two competitors. You have to start there, right? We've always had that technical differentiation, and we continue to have it. I think a couple of things that have changed or that have allowed us to maybe further gain share, further penetrate the market is, look, over the last couple of years, we've done a lot in broadening the portfolio itself. So now we have our retimers, we have our Washington TIAs, we have drivers.
We're broadening out the portfolio, and that makes a big difference to the customer base because they want to work with less suppliers at the end of the day. So they want to see more capability out of us. I think the third thing was just simply that we've been in production for now 18 months, mass production, and I think that's, it's always a bit of a challenge, this chicken and the egg thing. They don't want to design you in, especially big customers
Yeah
without seeing you in production. We've passed that hurdle, and they're very comfortable with it. It makes a big difference, especially as I look out into these 1.6T ramps and where we are with the ramp, the knowledge that we have, the visibility that we have, the customer comfort, I guess, if you'll call it that we have, that we didn't have at 400G or even 800 to a certain degree. So, I think that technical differentiation, I think it always starts with that. We've also bolstered up our systems expertise. I think that's something that's changed a little bit in the environment. When you're dealing with large data center customers or hyperscalers, they're not so much worried about component performance, they're worried about system performance.
Over the last, call it 12- 18 months, we've hired a number of system engineers, located them very close to the big hyperscalers and made sure that we're able to bring systems expertise, and I think that's something else that's important going forward.
Got it. How would you describe your customer concentration today within that?
Sure. It is fairly broad. I am sure most of you guys know in the audience, we have to think through all the data center customers as well as the module customers, right? MaxLinear, we had always focused on the data center customers, and we probably paid a little less attention to the module guys. That is flipped over the years. We have put a lot of effort into the module customers recently just because those are the guys that are working on qualifications, interrupt, trials, firmware changes, things like that, solving customer problems. So those folks are really important. As far as concentration, we have had pretty broad adoption, and we have got engagements across all the major data center guys as well as all of the module players. We do not have any one customer that is dominating that.
As I look out into the, naturally we will see if market share is 5%- 7%, I think we have aspirations to get up to 20%, 25% over the next couple of three years, and I think we have a great chance to do that. But I do not think it is necessarily dominated by one guy. I think there are three, four, five guys and many, you will be familiar with all of them, of course. So I do not think it is dominated by one person, but I do expect to see more concentration next year and certainly into 2028.
Got it. What needs to happen, Litch, that market share aspiration from where we are today?
Well, first of all, it is happening, and when I say that transition is kind of already happened. We are seeing those share gains. We have got these qualifications done. We have got to continue to support those customers. We have got to continue to ramp. We have got to continue to build out our own portfolio. We have got to continue to get supply, right?
Right.
I think we are confident on that. We have got to continue to bring some differentiation, whether it be on the technical side or even innovativeness on the supply chain. With our 1.6T, as many of you know, we moved to Samsung. The industry is somewhat constrained by TSMC, and so this move to Samsung has caught a lot of people's attention, and they are very interested in the fact that it gives them some diversity. Most of your module guys selling into this space, they are trying to get some independent supply chains, which is not very easy to do given some of the constraints on all of the components. So, I think we have got to continue to do innovative things like that. Not just at the chip level, but even in the ecosystem we have to think through.
Could you walk us through, how do you think about the manufacturing strategy, like I said, and things are getting tighter?
Right
At the advanced nodes, how do you go about securing that capacity? What negotiation you have to?
Sure
to work with?
Yeah. How do we go about it? Maybe I'll follow through on the Samsung. Our strategy is always to have multiple foundries. We've always had multiple foundries over the years. Call it four or five years ago, we would alternate between UMC and TSMC, and then UMC kind of stopped investing on advanced nodes, and then we were kind of stuck with TSMC for a period of time. But we had been kind of talking to Samsung and, they clearly have advanced node capability, and so we ultimately chose to move forward with that. It really wasn't done just for optical. It was done for the company and to keep some diversity. You'll see other products and other product lines or business units that will be done on Samsung as well.
Right
I think overall, it's a long-term strategy that we have as a company. It's not just for optical. As far as securing capacity and what we're doing other than kind of keeping diverse supplier base, yeah, we're negotiating, you're putting in orders. A lot of folks are talking about LTAs. We would try to avoid that for the most part, but from time to time, we've done those, just to make sure we secure supply. In some cases, we're prepaying or you put letters of credit. There's other ways that you can secure capacity as well. We have to use everything in the toolbox right now in order to make sure we get that.
Because you guys are fabless, and it seems like as demand continues to upside, much better than our expectation, do you foresee any challenges, or based on a relationship that you have had with the foundries, that is not so much a problem?
Look, I think we have a great relationship. It has been forged over many, many years. I think we have seen great response from them in tough times like this, so I am confident that we will continue to get more supply. I think your question alludes to more, how much more can you get? I think we can get a lot more. But we do have to continue to keep a really close relationship. We have to give them visibility and help them, and we do that in a number of ways, not just giving them POs. But we also think through from a technical standpoint on how we can improve that throughput as well, whether it is IP or chip size. There are all kinds of ways that we can do this in order to increase output.
They are not easy, and sometimes they take months and some R&D dollars, but often that will make sense for the long term. So we have a close partnership, and we will keep that up for sure.
Got it. I will open it up to the floor for any questions. Go ahead. One second.
Thanks for sharing. I have a quick question regarding your products' market share. As the technology evolves, what do you think, for example, like DSP, like your portion or like a dollar value, as the technology advance, would it be higher or lower, and how do you think about the product mix going forward?
Yeah. Good question. As we go, just like we have seen from 400G- 800G, ASPs have gone up. Same thing at 1.6T. I mentioned there is a big jump at 1.6T. I think as you go to 3.2T, you are going to see another big jump, and 6.4T naturally. These performance levels keep increasing. Talk about the component, like what exactly increases. I think of some of the simple pieces like a TIA, the driver, the DSP, which includes the SerDes. Those are kind of the three big pieces, right? As you advance, and especially as you go to 3.2T, architectures are changing quite a bit and you are seeing some of those parts become a little disaggregated.
More than likely you are going to see a standalone TIA, you are going to see a standalone driver, and then you are going to see the SerDes itself, which really is going to be incorporated differently. Maybe it is not in a DSP, but it certainly exists and will very much be needed. There is only a handful of guys out there that have their own SerDes at these performance levels, and so that is something that is very unique to us, and I think we can continue to increase our value proposition to the customer. I absolutely think ASPs go up at each transition for MaxLinear. I cannot give you an exact number of what that ASP is because we do not know what 3.2T or 6.4T is going to look like, but each of the pieces may look a little different, right?
Your TIAs and drivers at 1.6T, the ASPs are quite a bit higher than they were on 800G. Quite a bit higher than they were on 400G. And because those pieces end up becoming more valuable, as does the SerDes.
What is the metric that drives that content increase? Like what do customers sort of look at before they are willing to pay a higher dollar for that?
Well, that's a tough question. Because I think if you think it from a customer's eyes, they're building entire systems, right? And they're trying to be more efficient, effective.
Yeah.
I started out talking about our differentiation being on power. Their power costs are tremendous, right?
To the extent they can lower that, they lower the overall system cost. You can simply put, even we were talking the DSP portion, if you end up doing a half retimer solution, if the signal integrity is such that you can reduce one side, then you have to increase the other. So I could reduce some content here, but I've got to increase. For example, I could decrease the DSP over here, but I increase the TIA and the driver on the other side.
Yeah.
Right? So I don't know that there's one perfect metric, because every one of these guys, their architecture's a little bit different. And I think what's important to us, and actually I think it's a very unique value proposition that we have, and as you mentioned, the two other competitors in the space, there's not many people that have these IP blocks that are able to address the problem statement here, if you will. Right? You can't just come in and develop SerDes at these performance levels. Right now, the entire market is short on these resources today.
Yeah.
I think that's something that really stands out, and it's why customers come to us.
Okay. Got it.
Anyone else?
Just as you move to 1.6T and utilize Samsung more, just curious what kind of market share opportunity do you think that opens up, and whether or not there's significant slack capacity there?
Well, I guess I don't think of it. I think if your question is about market share as a whole at 1.6T, I think it's whether we are on TSMC or Samsung, I think we have equal opportunity. That being said, with Samsung, given the constrained environment, I think, we're seeing a lot of people show up going, "Yeah, we would love this." It's something else that differentiates us. So, we've long held getting to that 20%-25%. I think we stand by that today. It's not going to happen overnight. Naturally, you do see bigger jumps at the different transitions. So I do think that we can take a sizable jump at 1.6T. I think that's what we're counting on and looking towards. With regard to getting wafers, we don't anticipate that being a problem.
I don't mean that they have infinite supply, but, naturally, we've worked with them on the forecast and potential ramps, and we feel very comfortable that we can meet the demand of the customers.
I asked about your product mix going forward as the technology advances. Eventually, we're going to have NPO and then CPO, right? You'd have your products to offset those potentially DSP less risk, right? Can you talk about that area and the plan?
I'm sorry, offset what risk?
If DSP will be less reliant-
given, it's eventually-
When you say reliant, you mean?
Currently, it's pluggable, and in optics, eventually we go to NPO or CPO if they use less DSP.
Okay.
That's a lot of your product's revenue, and you can use other products' revenue to offset that.
Right.
For that technology advancement, and what is the strategy?
this is a little bit of what I had shared a little bit earlier about the increased content, right? in an NPO or a CPO application, that content would increase. what increases? The TIA increases, so ASPs increase, performance levels increase, drivers increase. In fact, today, the driver's integrated into the DSP. Now you need a standalone driver. then the SerDes ends up being in a chiplet. you take all of those versus the DSP and a TIA today, I'm saying that the content is higher in a CPO or an NPO application.
Anyone else? Perhaps, shifting gears a little bit, I believe the team has talked about a long-term growth of 2X the semiconductor industry rate.
It's a fairly ambitious target given that this year, if you look at the semiconductor market, it's on track to double 100% year-over-year.
Right.
Just want to get your views. Is that the right growth rate this year, next year, or what is the normalized growth rate?
For better or worse, we are a smaller semiconductor than a lot of these other guys. Our infrastructure business, I think a lot of the analysts this year have it growing somewhere between 150% and 170%. I think we are doing a pretty good job on that front as far as gaining more share. I think you will likely see some of that continued growth in 2027 and 2028 based on some of the products and the share gains that we have. Yes, if we can continue to outpace this market, and yes, the market is very strong right now. I think we have got a lot of new products with our retimer, which we did not talk so much about, but like the retimer AC applications, our next generation, Big Sky, which addresses 3.2T. We have got a number of products.
We are a new kind of player, albeit we started this five or six years ago, now we are on a third-generation product, so I think we are getting a lot of traction. Customers acknowledge that. Our Panther accelerator-
Yeah
another big one. We have a lot of new products that just came to market this year that start to ramp in 2027. That's all going to be what gives us the fuel to grow and outpace the market.
Double the industry growth rate is still the expectation?
Yeah. It's pretty amazing. We've all-
Yeah
Talked about data center growth. It's been very substantial, but I think our content, our share gains. I get this question, and I've been in meetings all day. What's underappreciated? I think this is one of the things that's frankly underappreciated. Granted, it's a little bit what's different. This is that third generation, that second generation. We got the traction. We proved ourselves. Now they're comfortable moving forward with us, where they had not been before. They had not seen that. I think also, just the simple fact that the market's gotten a whole lot bigger. They want to see an alternative supplier. You've also seen the two big guys kind of focus less on this.
Right. Yeah.
I think that's an opportunity as well. And not to say that they don't have other big things to focus on, but that opens up the door for MaxLinear.
Right. Especially improvement of power performance.
Yes.
Got it. Could you talk a little bit about your Panther? You mentioned that it's on track, it's doubling this year.
Doubling this year.
Potentially double again next year.
Right.
This is still pretty small, as I said, with entire portfolio. Talk about a growth rate, talk about a customer traction, and also what's your longer term expectation?
Sure. You're right. It's somewhat of a smaller product line. It only did $15 million-$20 million last year, but do expect it to double this year. I think we're on track to double again in 2027. It's a storage accelerator, which is an offload capability. So where CPU utilization is very high, the ability to offload with our solution, it's using a compression technology, enables them to use less CPUs. That was the original purpose. We focused on the enterprise storage market. We had done a lot of work with Dell and a number of other guys and had a lot of success. Continue to work with that entire community today.
as CPU utilization continues to go up and get more expensive, as NAND and just memory prices and availability continue to go up, customers have come and said, "Hey, if we tweak the architecture, change this firmware, could we do some things differently?" The answer was yes. We've been working on that. That's part of what's going to drive growth next year, and I think you're going to see more out of us. One other thing that changed, but another engagement, we had a relationship with AMD. Maybe a brief bit on the competitive landscape here. The other alternative to this accelerator technology is a product called QAT, and it's from Intel. And Intel has a software capability that compresses data on top of their CPU. Now, a lot of our customers, that compression capability is not sufficient, and they will use our Panther technology.
the other big CPU player, AMD, came to us and said, "We don't have that capability, and customers would like to have your capability." So they did a reference design with us. We've done a couple of keynotes at FMS, and so we've seen a lot of progress as they have sent customers our way. And so some of those customers are going to be driving that growth next year. And then you'll also start to see the product portfolio evolve a bit. You've seen a few press releases on some of the newer products, but there'll be more coming on that front.
How should we think about that driver of that business? Is it tied to CPU unit shipment?
I would say today, CPU utilization, okay? And then I think soon-to-be memory and storage. If we can reduce or offload that usage, it's not a controller. I've had this question a few times. It's not a controller, but it is a capability that we bring that is pretty unique.
Okay. Got it. Let's shift gear a little to gross margins. The company's on track to achieve 60% gross margin and has a long-term target of 65. Could you walk us through the key drivers to get there? Also, could you stack rank the gross margin profile of our different businesses?
Yes. I'll give some color there. Gross margins, this most recent quarter. Historically, MaxLinear kind of run in the low to mid 60s. Last down cycle, it kind of dropped down to 58. We've been in the process of recovering, making good progress on that front. Most recent quarter, upped the guide to 60%. Good traction. That's amidst some of the backdrop of the increasing wafer prices.
Yeah
The like. Making good progress. The big thing here is the mix, right? Our infrastructure business runs above the corporate average, and so we do expect gross margins to expand from here as our percentage of business on the infrastructure side continues to increase.
Yeah.
Been a little cautious, frankly, just to up that guidance too aggressively because we're seeing big increases in whether it's wafer premiums or if it's just wafer prices in general.
Yeah.
we don't want to get ahead of this, but I do see a general tailwind on the gross margin side because of the mix.
Got it. Are you able to pass through majority of that
Yeah. that's a good question. look, pricing is very firm, no doubt. You've seen prices increase this year, and so for the most part, a lot of that pricing has been able to be passed on to the customer. I think that'll continue next year. I'm not worried about prices coming down next year, at least in some of these areas where there continue to be constraints. the pricing environment is very good.
Okay. in terms of pricing, do you price just to offset the input costs, or is it slightly higher than input costs?
I don't know who's in the room. Just joking. Certainly. look, at the end of the day, our job is to get the value for the products that we provide. In all seriousness, that's what we want to focus on, and so where we can bring down system costs, for example, we want to make sure that we're getting compensated for that.
Okay.
Where we see input costs go up, and it needs to be passed along, unfortunately, from time to time, we do have to do that.
Got it. I know we're out of time, but one last question. what is the key takeaway that investors needs to know from this fireside chat?
Look, I think we have a tremendous opportunity with our data center business. I think you've seen us kind of ratchet up the guidance, but I think we're really hitting our stride. I think this is just beginning.
I think the share gains that we've seen are really early days, and so I think we've just got to continue to stay focused. I think you're going to see that over the next couple of years as these newer products hit the market.
Got it. Thank you.
Great.
Thank you very much, Steve.
Thank you. Thank you, everyone.
Thank you for joining us.