Good.
Great.
Thanks a lot, folks, for joining us. My name is Simon Leopold from Raymond James. We're here at our annual tech conference here in New York, and we've got a fireside chat session with Alan Lowe, the CEO of Lumentum. Alan, thanks for joining us.
Thanks for having us.
Pleasure. We know it's a heavy conference season, and so you're bouncing back and forth, and so appreciate you making the trip to the East Coast. This is where the investors are, so it's good to be here. So I think it's sort of nice in terms of a December conference to sort of think about the next year. So I know you're on a fiscal year ending in June, but how do you sort of, what's sort of your expectation, your high-level view of what's in store for us in calendar 2025?
Yeah, well, it's a super exciting time for us at Lumentum, and I think, you know, we made a very strategic acquisition and pivot a year ago, November of 2023, to get us one step closer to the hyperscalers and AI, and so we see very strong demand both from our components, our EML laser chips that go into transceivers, into the hyperscalers, as well as the transceivers themselves, and over the last couple of earnings calls, we've made announcements about new customer wins, and we're doing quite well in our strategy to diversify our customer base, as well as to ramp our production outside of China, which is going very well, and then continue to innovate and drive innovation in photonics to enable the next generation of AI and machine learning, so that will be critically important to our customers, not just in 2025, but also 2026 and beyond.
So what do you think will be kind of the main characteristics of your business if we sort of look out that three years? How's it going to be different than today?
Yeah, I think a lot's going to be driven by machine learning and artificial intelligence, and that's not just inside the data center, but we've already started seeing strong signs of data center interconnect driving telecom revenues for us. And that's, you know, things like 400 GB ZR modules, as well as 800 GB soon to be coming ZR modules and the components that go into them. As data centers continue to get built and are limited by the ability to get power to them, they've had to move them further and further apart, but they still need to connect them. So that's really what's driving rapid growth and what traditionally is service provider telecom. And so a lot of our business is being driven by those hyperscalers having to put data centers where the power is and then having to interconnect them. So that's pretty exciting.
So take us back to the present. We've gone through a period where there was excess inventory at your customers, and I think it affected primarily two lines of your business, if I'm thinking about it correctly, as the telecom components and the industrial lasers. Help us understand where does that stand today and what does a recovery look like?
Yeah, I mean, it's been a long spell of bleeding off that inventory. I mean, if you go back three years when supply chain constraints limited our ability to meet customer demand and customer demand seemed almost endless at the time, we built up inventory at Lumentum, at our customers, and at their customers, and so the telecom slowdown has been the longest of what I can remember, and I've been in the industry for 17 years, so typically the boom and bust cycles of our industry had been two to three quarters. This has been two years. I'd say we're almost to the end. There is still some inventory out there, but products that we weren't in production with a year and a half ago are seeing signs of strengthening.
Our high-speed Coherent components for 1.2T and 1.6 T Coherent transmission, as well as high port count ROADMs that are new, are really fueling that growth. As I said before, the data center interconnect is quite strong, and we're seeing the need for actually adding capacity above and beyond our peak levels for components like narrow linewidth tunable lasers.
Now I want to maybe unpack a little bit about what you've said about your long-term financial model and how we get there. So I think in the most recent quarter you've done just over $300 million in quarterly sales, gross margins, I think in the 30s. But you've talked about getting gross margins back up towards the 40% range and operating margins back into the 20s. Yeah, I reflect back at the sort of the prior cycle when you were big with 3D sensing, we had operating margins.
North of 30.
High 30s, right?
Yeah.
So what's sort of the moving parts to get to those milestones?
Yeah, I think there's two things. One is operating leverage, so growing the top line, and we're on track to do that quite rapidly through calendar 2025. And what we said is that by the end of calendar 2025, we will be at a $500 million run rate again, and you know, with growth expectations beyond that. So the other thing we've said is that at $600 million, we expect to get back into those levels of gross and operating margins, so high 30%-40% gross margins and 17%-20% operating margins at a $600 million quarterly revenue run rate. And I think, you know, we've shown that we have expectations to get to the $500 million by the end of 2025.
You know, clearly with these hyperscale wins that we've been talking about the last two earnings calls, there's no reason in my mind that we can't get to $600 million shortly thereafter.
Now to get to those kind of higher revenue rates, higher margin rates, what's the planning in terms of factory capacity, footprint, investment, and when do you opt to outsource? So in the past, you've used contract manufacturers. What's sort of your build/buy decision process?
Sure. I mean, we go through that all the time, but I'd say for the build process, EML lasers, which are really being driven by machine learning, build-out data centers, and artificial intelligence, we are growing output from the June quarter of 2024 to the June quarter of 2025 by 40%, and then another 40% by the end of the calendar year. So basically a doubling of EML output from June of last year to the end of next year. So that's really more like $100 million of EML revenue above corporate average standard gross margins. And then, you know, with the wins of the data hyperscalers that we talked about, you know, that can grow dramatically. We're not planning on a huge amount of telecom service provider growth, but data center interconnect is growing quite nicely, as I talked about.
In industrial lasers, to your point, there is a buildup of inventory that still needs to come down in the laser business, and that we're not expecting a lot of growth there. So we don't need a lot outside of the data centers to get us to north of $500 million. And you know, if we do get other areas of growth, you know, we'll get there faster.
So one of the areas that is just dominating our inbound discussions. It's all about AI. So maybe talk a little bit about the specifics of the AI opportunities for Lumentum. I think we've generally assumed that when you sell an 800 GB transceiver, that correlates directly to AI. Can debate whether or not your data center interconnect is related to AI. How do you factor this into your own planning?
Well, I mean, we're working. Well, let me back up. A year ago, we were working with all the transceiver manufacturers in the world, and they were, you know, we were one step removed from the hyperscalers, and now with the acquisition of Cloud Light 13 months ago, we're at the table with the hyperscalers, not only talking about our adding capacity, but also talking about the roadmap for future technology, so I think we're gaining a lot of mind share with the hyperscalers, as well as the AI infrastructure providers to give us insight into where we need to be placing our R&D dollars, as well as our capital dollars to make sure we're keeping them satisfied, and I think we're doing a very good job right now, and you know, we'll see as we continue to grow our business over the next 12 months and how we do.
What's your sort of best guess or, you know, third-party research on how do you size the TAM for that? So I assume we've got the optical switches, the transceivers, your EMLs. What's the opportunity?
Yeah, well, we look at the transceiver market as $6 billion-$7 billion today and growing 30%-40%. So five years from now, it's north of $20 billion. So there's plenty of room for growth for us as well as the industry as a whole. And in that market, we expect that the laser portion or the indium phosphide portion of the content of those transceivers to be somewhere between 10% and 15% of the BOM content or the selling price content of those transceivers. So you can see going to a $2+ billion indium phosphide type of TAM. And so I think, you know, there's plenty of market there. And then, you know, the DCI is going to come into, you know, where do people get power, right?
How far away do they have to place these data centers to be able to get the power to run these data centers? I think, you know, it's only going to get tougher. So the data center interconnect and then the transport networks connecting these data centers is going to continue to drive a large portion of our telecom growth. It's really data center driven.
Maybe take the opportunity to educate folks a bit. We're trying, but we often get these questions around VCSELs versus silicon photonics versus EML, and the questions often sound like who's the winner, as if something's going away and the other one wins. How do you think about this? Because you've played in all three flavors. How would you explain to a financial audience how to sort of factor that market? What's important?
Yeah, I think it's a tool in our toolkit to provide the best solution for a customer need. And so I think there's going to be need for each of the solutions. And in fact, you know, carrying that forward to linear optics as well as co-packaged optics, there's going to be a home for all of the above. I think we've seen a shift from multi-mode or VCSEL-based transceivers for design work to really more single mode, whether that's silicon photonics or EML-based. I think that over time, the shift will be more towards EMLs, frankly, over time as we get to the multiplexing of multiple wavelengths over a single fiber. So I think today, you know, there's a combination of silicon photonics and EML-based transceivers, but I think over time, and we've shown the 400 GB EML at ECOC a few months ago.
So there's, you know, next life, if you will, of EMLs. So it's, you know, the investments we're making today for design wins two and three years from now are going to pay off. And it actually brings the hyperscalers' engineering teams to our factory and to our labs and to their labs to work on how do they make sure that these GPU clusters are really going to be able to get the data they need at the time they need it, at the speed they need it.
So I would appreciate, because of your position in the EML market, you might have an EML bias.
Yes, I do.
But Cloud Light products have been mostly silicon photonics. So how do you explain that sort of difference? Why or when would you migrate that to an EML, or is it a matter of sort of bill of materials? What's the logic behind that?
Yeah, I mean, it's part of its legacy and what were designs in the works when we acquired them 13 months ago. We are working on designs for EMLs. Right now, the demand for EMLs far outstrips the supply. And so we expect that we will be launching transceivers with EMLs-based designs in the second half of calendar 2025. And so that will actually help our transceiver margins dramatically, especially as we look at the multiple different wavelengths being maxed together over a single fiber. EMLs are a clear differentiated winner in that kind of a transceiver design. And so, you know, given our margins on EMLs being, you know, higher than corporate average, putting those into our transceivers will really help drive vertical integration and the margin expansion of our transceivers.
Now, this industry's, I think, always been a little bit crazy in that you've got a lot of co-opetition. So you sell to your competitors, you buy from your competitors. How do you explain that to somebody who's new to following this space of you're selling EMLs to companies that are selling transceivers up against your transceivers? Why does that work? Why is that a good strategy?
It's a good strategy because there's a lot of reliance on each other. So my biggest U.S. competitor is also a good supplier of mine, and I'm a good supplier of theirs. And I think there's this reliance on each other that really seems to work well. And you know, there's a lot of trust that has to happen at the CEO level. And I think the partnership we have with, you know, our biggest competitor is actually quite good and one that we built some trust around. And so I think that's going to continue to be the way of the future. I think the way that the demand is today, there's plenty of room for competitors to grow dramatically and us to enable them to grow with our technology and our componentry.
Now, you mentioned a couple of sort of leading-edge technologies like co-packaged optics, which I feel like have been on the horizon for.
Ever.
Ever. And we're hearing more and more about Coherent technology getting use cases inside the data center. How does this affect your business? How does it affect your strategy? What do you need to do or how would it affect you?
Yeah, I think we are working on co-packaged optics and with very high-power lasers to be able to split into multi-lanes of interconnects within a data center. That's probably more like 18 months out from really being in volume. But at the same time, it puts us in the labs with the engineers around how are they looking at the architecture of the future. And so I think from that perspective, we have a very differentiated laser technology around that. So, well, it may take away from some transceivers. I think we will gain a good share of these what we call ultra-high power lasers if or when they actually go into production. Because to your point, it's been talked about for many years, and there's camps on both sides that say it's not feasible and that it is feasible. And you know, it does drive down power.
So there are areas of the data center that it might be very attractive for, but it's not done yet.
In those cases, you need to partner with or align with either a network processor or GPU, right? So you'd have to work with some basically semiconductor company. There's not something you could be doing on your own there.
Yeah, I think that there's, you know, within the infrastructure, you need to be able to work with everything. And so, you know, we are working with an AI infrastructure provider on that, and we provide a critically important part of that, and they provide the other part. So I think that there's, you know, good cooperation there.
Now, am I correct in remembering that you had a DSP effort for Coherent optics and you scrapped that? Is that correct?
We decided to use our funds in other areas for growth, and I think, you know, when we looked at what it costs to do a three-nanometer Coherent DSP versus what we could buy it on the open market for, and then take those R&D dollars and put it towards transceivers where there's rapid growth and we can differentiate with our vertical integration, we decided to stop that and focus those R&D dollars more on the more rapidly growing area and then rely on our third-party partners to provide the DSPs because they're making the investments already.
Which makes sense to me. So you seem apologetic. I'd rather you grow faster and get better earnings than throw money where it's not needed. So I don't think apologies make sense there, but nonetheless. So how do you sort of play into that market if Coherent starts getting inside the data center? How does that affect your business?
It's still going to need a lot of tunable lasers. And I think, you know, I think the challenge is Coherent inside the data center. There may be applications for it at 1.6T. I don't know. The cost of a or the price of a 1.6T transceiver is pretty low, and the power consumption is not terrible. So I think there may be an application for it. There may be more application for it at even higher speeds and next generation. So we'll see. And it makes a lot of things in the data center a lot easier to deal with when you look at optical circuit switching and things like that. It's just better with Coherent. The question is the cost, and can the cost come down fast enough and far enough to be able to really compete and replace the traditional telecom or datacom transceivers?
So maybe use this to sort of pivot to the telecom discussion. So that market has been in sort of an excess inventory situation for some time. And I guess there's some debate as to what a recovery looks like. So some of the discussions we've had with some of the OEMs, your customers, have said, "Hey, you don't need to replace the ROADMs every year because they're speed agnostic." So maybe this market doesn't go back to the prior peaks between the usability and between, you know, overbuying. What's sort of your thinking of how to forecast out that overall market?
Yeah, I'd say that, you know, the traditional service provider market is a mid to high single-digit grower. But the problem with that is that it's becoming a smaller part of the growth of telecom because the data center interconnect and the hyperscalers building out their own networks is growing very rapidly. So, well, say 25% of our telecom revenue today is for data center interconnect and hyperscalers today. That's growing probably at 30% per year, whereas the other is growing single digits. So eventually, the data center interconnect and hyperscaler networks will be half of our overall revenue in telecom, but it's really driven by data com. And so I think from that perspective, you know, we don't need a big pickup in service provider spend to get back to the levels of peak revenue before, but we do need some.
And how has this changed your customer mix in that generally you've sold into OEMs, manufacturers of systems. Now maybe you have more direct sales to operators, the cloud guys. How's that snapshot look today, and how's it going to change over time?
Yeah, we still sell a vast majority of our telecom products to the network equipment manufacturers, to your point, even the data center interconnect, because, you know, whether it's a component like a tunable laser that goes into one of their ZR modules that ends up at a hyperscaler or a next-generation ROADM that we sell to a network equipment manufacturer that goes into a hyperscaler network, you know, it's all through the NEMs. We sell very little outside of the data center directly to the hyperscalers. Inside the data center is where we sell the transceivers, will be selling the optical circuit switches as well as other stuff, you know, that interconnects these things.
How has sort of the changes in architecture going to these open line systems affected your business? Are you seeing opportunities to sell directly to telcos now where you can sell them a ROADM line card, or did that never really play out?
It didn't really play out for us. You know, partnering with a network equipment manufacturer seemed to make more sense. Now, a lot of them are coming to us and say, "Hey, can you make this multi-rail amplifier?" Because as the Shannon Limit limits the amount of data that can go over a single fiber, there's more and more multi-fibers that are going over a single route. And so amplifying those or switching those and having a multi-rail application of our ROADMs or our amplifiers really goes a long way to helping them, whether that goes through the NEMs or directly through to the hyperscalers yet to be determined.
How do you feel about the submarine market? Your amplifiers are generally there. Just the other day, my mom sent me an article about all these submarine cables being cut out by Africa, which I had missed. We keep hearing about these constraints and this argument that there's a major build cycle for submarine. What's your thought on that market dynamic?
Yeah, I mean, we've seen good, strong demand for the pumps that go into the amplifiers at the bottom of the ocean. They're what we call high-reliability submarine pumps for two reasons, one of which is data centers are being built all over the world and they need to be interconnected. So if you go back 10 years ago, these cables were being bought and deployed by a consortium of carriers. Now it's a hyperscaler putting in a cable between two continents. And because of the challenges with respect to, you know, the seaworthy traffic, the cables are getting longer because there's some routes that the boats will not go to lay the cables. So that's on new cables. So more amplifiers, more pumps because the cables are longer. Now, the whole damaging or purposely cutting the cables, that's just a big headache for everybody.
It's very, very costly to pull up a cable from the bottom of the ocean, resplice it, put a new amplifier on it. That for us is not a driver of a lot of business. It's really more the new cables that are connecting continents together or the fact that they're going longer.
It also speaks to the importance of route diversity, though.
Yes, true.
So how big a business is that for you, the amplifier side of it that's in telecom? Is that material or?
You mean submarine amplifiers?
Yeah.
It's probably less than 5%, but because of the limited competitors and the stringent nature of how reliable they need to be and they have to work for 20 years, the margins are quite good.
It's basically a duopoly, is my impression.
The competition is pretty limited.
Okay. And then the ROADM market, you've talked a bit about some new opportunities with L-band. Can you help us really frame out how material that is? How does sort of the next cycle of upgrades compare to the previous cycles?
Yeah, I'd say, you know, we're seeing strong demand for integrated C-band plus L-band, so a ROADM that handles both C-band and L-band to be able to put more and more wavelengths through a single fiber. That's primarily in China today, but we've seen some interest outside of China, and so, you know, given our limited ability to sell to the Chinese network equipment manufacturers, in fact, one of them, that's not a huge part of our business, but it's growing nicely off of a relatively small base, so we have seen desire from customers to get more into the L-band or integrated C-band plus L-band to really be able to put more over a single fiber.
And how about, I want to pivot to the commercial laser business. I personally, it's been frustrating because it always feels like we're talking about the road to $200 million in sales, getting to $400 million in sales, and then something happens. So we've had a couple of disappointing quarters of late. We've had some product transitions. Just help us reflect on maybe lessons learned in that it's always looked like there's a lot of promise on the horizon and then some obstacle pops up.
Yeah, that damn pandemic. I would say there's a lot of interest in precision manufacturing using our ultrafast lasers today. We have a very good share of solar cell manufacturing using our picosecond laser. We have a next generation of picosecond lasers that should help grow that business and grow the margins. And then we're also working on femtosecond, so even faster lasers for things like solar, rather semiconductor processing, as well as display processing. So there's markets out there. It takes a while. It just takes a long time to design in a new laser into a new tool that goes into a new process, but we're making good progress there.
I guess I've developed a better appreciation of the complexity of that industry. When I first started covering, I just sort of industrial lasers are one thing. I guess it would help to maybe help, you know, frame who are your competitors? Because that maybe sets some context because there are companies like IPG and nLIGHT and Coherent and then Raycus in China. It seems like a crowded market, but everybody's kind of got their thing. What's sort of where you're competing?
Yeah, so we have one large customer for our kilowatt fiber laser class lasers in Japan. The challenge with the kilowatt laser business, it's very differentiated. So our customer uses the differentiation to win in the market, and they have a very high-end laser cutting tool. China, to your point, Raycus and others in China have really dropped prices. And so it's very tough to compete inside China. And now whether they get outside of China or not, it's yet to be determined. But that business is tough. And we're having to innovate and drive costs down to be able to have our customer compete in that global market against the nLIGHTs and IPGs, but more importantly, the Chinese manufacturers of these lasers. On the ultrafast laser side of the business, we compete with MKSI and with Coherent. And so that's a market that is highly technical and highly differentiated.
That's an area I think we're going to be growing dramatically as our products are really being well received by the customers. Now they just need to put them into their tools and qualify and then start ramping.
How long does that normally take? What's kind of the?
It's usually 18 months, so we're, you know, into that process, so, you know, I can imagine by the end of next year being in production on those, but I'd say for the next two quarters, it's going to be kind of flattish revenue.
Okay. And we're just about out of time. So I always like to close with the same question. You know this is coming. So what do you think is the least appreciated aspect of either Lumentum's story or the stock? So give you a chance to sort of leave the investors' audience with a sound bite.
Okay. I think it's around our operating leverage. And as we grow revenue, especially things like EML chips, the variable margin on EML chips is very high, as well as in the telecom business. So, I mean, you saw it on the way down as the revenue went down through the pandemic, but you'll see it on the way up as well. I mean, margin expansion and operating leverage is very solid on the upswing. And I think we've laid out a very solid roadmap for revenue growth, and the margins will come.
Great.