Good afternoon, everyone. I'm Samik Chatterjee, and I cover the hardware and networking stocks at JP Morgan. I have the pleasure of hosting Lumentum for the next fireside chat. With us is Chris Coldren, Senior Vice President of Strategy and Corporate Development. Chris, thank you for the time to come to the conference, and Kathy, thank you to you as well. Chris, I'll start you off with a question we've been asking all of our companies to just sort of give us their forecast of what the end markets look like 12 months from today. So you obviously have exposure to telecom, datacom, some of the industrial exposure as well. Just help us maybe project forward 12 months from now, where do you expect demand to be for each of those end markets?
Sure. Thank you, Samik, and thanks for having us. We've had a lot of good meetings so far, a little later in the day, and a lot of productive discussions. So, you know, as we look to the next 12 months and kind of go through the end markets that you mentioned, I'd say that we're probably most optimistic in terms of the data center cloud market, where, as many investors are aware, photonics are now, you know, penetrating into the compute fabric, if you will, not just interconnecting servers the way they have for years, but actually interconnecting GPUs, TPUs, et cetera. And, you know, that's driving a lot of growth and a lot of expectations of growth.
And so over the next 12 months, we can see, you know, or maybe put it this way, calendar 2025 versus calendar 2023, I expect the market for high-speed optics to be multiples of what it was in 2023. So that's really a lot of upside. And frankly, as I look to the telecom business, I mean, it's the market is very difficult right now. It's quite slow, but I'm optimistic about the next 12 months, given we are shipping well below end market demand. So even a you know, partial recovery, let alone getting back to where we were historically, would be significant growth from today's run rates. Industrial is kind of the same situation, going through an inventory correction, and I think that'll work its way through the next 12 months.
Probably the only area where I'm maybe more cautious on a 12-month timeframe, but still excited long term, is when we look at some of the consumer market, perhaps automotive, LiDAR. These markets have a lot of potential, but a lot of the new product designs at the customer levels have pushed out a year or two, and new designs are willing to drive growth for us. So I don't think you're gonna see a lot of growth from those markets over the next year. But certainly industrial, the telecom, and then really headlining is the data center market.
Okay, good. Relative to the data center market, and obviously a lot of interest in Datacom, but, and relative to just AI infrastructure build, before we get into that, just maybe share your thoughts of what are you thinking in relation to AI adopting it internally within the organization w here can you drive more tangible improvements in the operations itself, timelines of getting some of those improvements before we get into the product areas?
Sure. Probably like everybody that operates in enterprise, we've got the same things of meeting notes, and summarizing, and translating in different languages. As a global company, wouldn't underestimate how important that is, but maybe more important, as somebody that's a developer and manufacturer of hardware, we do have and benefit from, you know, coding, copilots, et cetera. So that's well-covered space, so I'll let others opine on it more. But I think that's something that I've been really pleasantly surprised about. Early results that I think will scale up is employing AI when we're looking at test data or visual inspection. You know, we make tens of millions or more of, you know, laser chips, as an example, per year.
Being able to use AI to be able to correlate back, you know, hey, a chip that looks like this or test data that looks like that, how does that correlate back to performance and yield? And really being able to accelerate internal efficiencies that typically would take several years to figure out, having AI be able to replace that sort of long human learning cycle, is, I think, gonna pay big dividends. And I'm sure the same is true for other manufacturers of a wide range of products in the world. So it makes me very optimistic that we ourselves can see tangible benefits, and I think others will as well.
Yeah, got it. Turning to Datacom, you obviously, that's an area where you're seeing already some tailwinds because of the AI infrastructure build, but maybe hit on AI-related demand more broadly across your portfolio, like beyond Datacom, how do you think over the spillover effects to telecom, for example, or other areas of your business?
Sure. Yeah, I mean, certainly, as you highlighted, it's caused a significant increase in demand in the data center products that go directly into data center, as well as accelerating the demand for very high-performance products, perhaps, you know, adopting them earlier than they would have been now initially, and also shortening the cadence of when the next speed and technology will come in. The reason I am bringing that up is that then the knock-on effect is outside the data center. You're certainly. We have an expectation that that will be a driver of sort of telecom products, for lack of a better word, i.e., data center interconnect transceivers, optical infrastructure that, you know, going hundreds of kilometers or thousands of kilometers, and that will be a driver of demand there.
I mean, one of the bigger challenges our understanding of building data centers today is being able to actually deliver enough power, and being able to do that 'cause, or the lack of the ability to do that, is causing customers to decide to separate data centers physically, which then means they need to have a lot more interconnects in between them over long distances, driving traditional telecom-like products for data center applications.
And then in the industrial space, I think it kinda goes back to what I was commenting, how we're using it, that we're seeing a lot of customers that, you know, we, we sell, for example, laser products to folks who build manufacturing equipment, and that manufacturing equipment is really taking things into, you know, vision systems that are then feeding back into AI engines to be able to either process control or quality control built into the machine tools, if you will. That then drives more value and will drive a cycle of purchasing of new enhanced machine tools that then drive the need for more of our products.
Yeah. Okay. Let's dive into Datacom specifically a bit, and I know we've asked you this question before, but still want to revisit it because you have increased your presence in the Datacom module business with the Cloud Light acquisition after exiting the business a few years ago. When you now look at how the module business is ramping at Cloud Light, I mean, still maybe, go through how do you think the opportunity is different from what you had a few years ago when you decided to exit the business? Is the AI-related demand the primary driver of deciding to revisit this module business, or what does that also assure you in terms of pricing which was a major concern back when you exited the module business?
Yeah. So we had a Datacom business that was, you know, basically built around supplying into the historical Ethernet market, the enterprise equipment market, maybe more precise. At the time we exited was when there was a big transition between enterprise really driving the market and then cloud taking over. In that transition, that changed, you know, what customers desired. It introduced a brand-new level of competitors that, at the time, we said, "Wow, either we need to sort of redevelop or adapt our team," and that would take multiple years and a lot of money, or, you know, do we wanna fight that war or sell the bullets?
Essentially, we decided to exit selling enterprise-type transceivers and trying to move to selling cloud-based or cloud-oriented transceivers, focused on building up a significant component business at the highest end. And what we've seen since then is, you know, technology, the speed, the component level technology was essential, but it wasn't as critical as now, as the speeds have gone from 100 Gb to 800 Gb, going to 1.6 Tb. That component level technology really matters a lot more. The number of competitors, or at least the competition, is probably consolidated into a few leaders, if you will. And then as you bring in this infusion or step function at the highest end of the market in AI, drove us to decide, maybe this is the right time to reinsert ourselves.
And then per an earlier question, also the realization that, the same customers buying these transceivers are also increasingly going to be customers for telecom components or telecom modules, and other gear that we sell or could sell directly to the cloud operator. So we felt having a more direct relationship was also an important point, as, you know, at some point, the cloud operators will probably be the largest consumer of telecom gear, and important for us to have that direct relationship.
Got it. What about the pricing side? I mean, how do you get comfort that pricing pressure is not going to return to what you were seeing in 100 Gb back at that time?
Well, I think that there will always be intrinsically a competitive market, but that said, at the time when you have, you know, a gaggle of new market entrants coming in that were really focused on using price as a leverage to get into the market, I don't see that today. I think that there's, if anything, the transition to higher speeds has sort of concentrated the supply, who's really able to supply those products, into a handful of players. It doesn't mean there's gonna be, you know, healthy competition. Prices will come down.
But I think, you know, previously, I would say it was more for lack of a toxic competition, as opposed to, I think the way that one has to really think about this is price comes down to enable more compute capacity to be able to be delivered to the market. And so in the end, it ultimately means you're gonna ship more product, but I don't think you're gonna see the same level of toxic competition, let's say, that was previously seen, because you have to be able to deliver the needed products. At the same time, you gotta remember, we also supply components into the market, and the components, there's even, you know, fewer competitors that are capable of supplying the highest-end components.
What we can see there is certainly prices do come down with volume, and that's necessary for the whole scaling of the ecosystem, but it doesn't have kind of the kinda same level of price competition that you're talking about, you know, eight or nine years ago at the transceiver level, because it's a, you know, technology matters. There is differentiation ultimately.
Okay. Okay, got it. Cloud Light's primary customer is now a 30% customer for the company in the latest quarter. Now, Lumentum historically has had its challenges with customer concentration, first in 3D Sensing. You had also Huawei in the telco business as a sort of a large customer. How are you strategically thinking about diversifying from the customer concentration you have now to this sort of one Datacom customer? How realistic also is it that you can diversify beyond that?
Yeah, I think, it is a very top focus. It was a major sort of element of the thesis of the acquisition that we could help the Cloud Light team get into accounts that as a private company were a startup company based in Hong Kong were unable to probably access. And so far, you know, it's, you know, we've got a lot of very good traction with major customers. That traction has to translate to revenue. But, you know, it's been six months since the acquisition, and you kind of have to time your, y ou have to have a product that matches the customer's adoption cycle if you will.
And so that's why we've highlighted that we do expect, as we get into later this calendar year, and then much more so into next calendar year, that's when the opportunity for new customer traction to translate into, you know, significant revenue increase. And I believe, you know, we have a lot of value we offer customers between what Cloud Light was offering on their own in terms of efficiency, speed, very good products, now part of a U.S. headquartered company, vertical integration on component technology. We're, or will be, we're getting ready to do this ramp up in Lumentum's factories in Thailand. All these things are sort of top of mind of customers, and there's a, you know, again, a limited set of,
We know the transceiver competition because they're actually our customers for a lot of the components, and that suggests that there's room for at least another supplier to be able to deliver on the volumes and capacities. I mean, again, the 800 Gb and above market, we're talking revenue dollars, is probably 3-4x in 2025 what it was in 2023. So from units, you're talking about something that's even more. That's a lot of capacity the industry has to put in place, and so I think us being able to win a few more customer sockets is something I feel a lot of confidence in, but we've got to go do it.
Okay, got it. So you've acquired Cloud Light, so now you have a traditional Datacom component business and a module business in Cloud Light. How are you thinking about the growth outlook for the two different businesses over the next few years? What sort of- How do the drivers differ between the two?
Yeah. Yeah, I think that they both supply into the same end market, so I certainly think both have the same strong tailwind driving them forward. Perhaps, maybe the differences were a little bit more scale or have more aggregate market share at the component level. So there is continued opportunity to gain market share at the component level, but I would argue, at the transceiver levels, probably a lot more opportunity to gain market share via share gains, not just growing with the very strong tailwind of the market. So that's why I'm optimistic about both growing quite significantly. But I think from a revenue standpoint, there's more long-term upside in the transceiver business, just given more limited market share exposure over the last 12 months, if you will, than at the component level.
Got it. Got it. Vertical integration, when should we start to see that play out a bit more on the margin? And what are the sort of steps you have to take from here to drive that?
Yeah, I mean, as acquired, Cloud Light was not using a lot of Lumentum components, so there's a large opportunity for component in-feed. I would say, the products that we're launching late this calendar year and into calendar 2025 will have a certain level of Lumentum component in-feeds, as well as a certain level of manufacturing in Lumentum factories, so those drive synergies there. But at the same time, I think over the next two, three years is really the timeframe where we'll be able to more deeply penetrate sort of through the entire portfolio. Because there's certain products where we may not have the component already developed, and components do take, you know, a year or two years, depending on the nature of the component, to develop, to intersect the roadmap.
I think that will be an unfolding story that will see upside into historical performance in each of the next several years.
Got it. Got it. Investors seem to be trying to figure out when we put 800 Gb, when we put 1.6 Tb together, how much of the technology will be dependent on silicon photonics or CW-based lasers? How much is EML versus VCSELs? What's your perspective on which one will emerge to be the leading technology, within these three, and where can Lumentum differentiate the most?
Maybe a little tongue-in-cheek, but all in all, and I guess my point is that each of those exists for a reason. They have different pros and cons. I mean, VCSEL-based technology is more cost-effective for shorter-reach applications, shorter-distance applications, but that shorter distance, or the distance with which you can go, kind of shrinks as speeds go up. So that's why the other technology, silicon photonics and the indium phosphide EML-based solutions, exist for longer distances or higher speeds. There is perhaps, so I don't see a lot of competition between VCSEL-based and indium phosphide or silicon photonics. They kind of naturally fall into the application when you determine the speed and the distance they need to go. So we're working on all the above.
When it comes to silicon photonics and EMLs, there's certainly some overlap in, in capability, but for that reason, we have both in-house. We have silicon photonic-based transceivers, EML chips we supply in the market. We are developing our own next-generation silicon photonic component-level technologies. We're also developing EML-based transceivers. And really, certain customers have views of which technology they want to buy. But I would say, maybe simplistically, and this is more from a technical standpoint, that silicon photonics tends to be good when there's a lot more parallel things. It's more of an integration platform, so you have eight or 16 or more wavelengths that, or channels in parallel, maybe more importantly, that you're transmitting. But if you're down to a fewer set of channels at much higher speeds, then there's some optimum in, in EML performance.
Again, I go back to having the whole suite of technologies. I think is really important, so that when we sit down side by side with customers, we're not trying to just push what we have. We have the whole range, and then we can determine, in partnership with the customer, what is the right technology to solve their problems? I think that's, you know, very important from both a relationship standpoint, but usually the right solution ultimately wins. It's just who figures it out first, and that puts us in a great position to be able to figure it out first.
Yeah. Maybe let me follow up in sort of a different way. Will your vertical integration be similar across all three, or, how do you think about sort of where your vertical integration is higher than sort of relative to the other platforms?
I still think there ultimately, we make VCSELs, we make EMLs, we make in a silicon photonics-based solution. There's a high-power laser that's also needed. Now, in the silicon photonics, most people like us, we outsource or use a foundry, but we are designing our own silicon photonics. So, I think we have, you know, the ability to be equally high, vertically integrated across each of those, products ultimately. Now, in the near term, it depends. You know, that's one of those examples of we have, you know, the most bleeding-edge, EML technology today, then followed by the high power, CW lasers. Our VCSEL technology is very adequate for what is needed today, and we're working to catch up to be, you know, a leader at the bleeding edge.
Thank you. Last earnings call, you highlighted capacity expansion that you're doing for the Datacom business. Just maybe flesh that out a bit more, where exactly sort of which areas are you expanding capacity on, and how to think about the timing when those sort of when those capacity come online?
Yeah, I think there's really two pieces to think about. There's, at the component level, being able to have sufficient wafer fab capacity to be able to meet, you know, not only internal vertical integration needs, but also to continue to serve the external market. And with a market that's growing as rapidly as it is, that's no easy task, meaning that expanding wafer fab capacity can take time, and so we're well underway with increasing capacity. I think you heard on the call, we talked about being able to increment up, you know, maybe 50% or more additional output out of our wafer fab, once the next slug of capacity comes online. We didn't really put a precise timeline, but you're gonna see that probably out in the calendar 2025 timeframe.
When it comes to transceivers, you know, we have been kind of running at a certain run rate, which is essentially our capacity. And now, as we have new customer opportunities or frankly, even growth within our existing customers, will require additional manufacturing capacity. And the way we're doing that is adding it in Thailand to add more geographic diversity to our supply base. And certainly, customers have concerns of supply from other geographies, China included. So that's why focusing on adding significant amount of capacity in Thailand, and you know, there's sort of several tranches of that relating to you know, where do we have strong customer traction? We know we can fill that capacity, and then if we, t hat's sort of one way of thinking about it.
Another way of thinking about it is, hey, if you look at the market out two years, three years, how big is it gonna be? If we have a certain market share assumption, wow, that means we're gonna need a lot of capacity. What are the tall poles, and what do we need to do today if it's got a two-year timeline? Let's go do that. So that's kind of how we're working through the capacity additions so that we can add sort of multiples of current revenue run rates ultimately over the next several years.
Okay, got it. Also on the last call, you talked about doubling or more than doubling the Datacom business by the end of calendar 2025. How should we think about the opportunity to double the component business as well as the module business independently within that same time frame? How should we think about what customer footprint you need to be able to achieve those targets?
So I would say, maybe start with the last first, which is to say that I think we generally have engagement with all the needed customers in general to achieve those kinds of growth rates. Doesn't mean that there's not more customers to go ultimately, but to double the transceiver business as an example, it doesn't mean we need to add, you know, 10 customers. It's one or two ultimately, at most need to double. I think there's more opportunity to do ultimately more than double, if you look at that on a little longer time horizon. And the same thing is true on the component business, that we have a wide range of customers there. So it's...
There are some customers to go add, but we have engagements with them, and potential design and activities underway. So I think the customer footprint is, you know, we have to go in, but I think it really gets down to operationally, can we add the capacity in the needed timeframe? And what we discussed about the transceiver capacity, I think, is well in line with being able to have capacity to support a doubling of revenue.
And I'd say probably even the same thing in the laser chip side, that not only are we adding increments of capacity, but there'll be some probably favorable product transitions as we go to higher speeds, where even for similar wafer output, you might be able to get more revenue per wafer, given you're bringing more value to the equation with higher speed lasers. So I think the both are well. I mean, the market demand or market opportunity is there. We have the customer footprint or are engaged with the needed customers today for that to happen. Now, there's, you know, got to get the products developed, got to get the capacity finalized, and get a win, right? I mean, it's not all done yet.
How does 1.6 Tb, and I guess how does 1.6 Tb figure in your plans to double the module business? And for the component business, how should we think about 200 Gb EMLs factoring in that?
Sure. I would say that certainly getting outside of our current largest customer at 1.6 Tb would, you know, help a lot, if not far exceed the, a goal of doubling. But I would also say that there are plenty of other opportunities, whether they be at four, eight or 1.6 Tb, that will feather in over the next two years to be able to hit those kinds of revenue goals. On the component level question, yeah, I mean, that's kind of what I was referencing, that as we go from 100 Gb to 200 Gb per lane, that certainly brings more value to, to, it's a more valuable chip, so therefore, more dollar content per wafer effectively coming out of the wafer fab. That'll also help with growing revenue to being able to double effectively.
Okay. You talked about the capital investment being made into the capacity. How should we think of the magnitude or the change in the capital investment that you've generally had, to support that plan?
Yeah, I mean, you know, for better or for worse, we've obviously had a slower business in the sort of telecom end markets over the last year, and therefore, there hasn't been a lot of capital investment in that space. So within the same kind of spend envelope, we're able to re-steer the spend towards the data center investments. That said, obviously, if there's a more aggressive outlook ultimately for telecom, then we'll need to invest a little bit more. But at this point, you know, we last left off in telecom at a substantially higher revenue run rate than we are today, so we've got a lot of capacity in place for that.
So I don't see that as a major driver of dramatically increasing our CapEx spending, unless other parts of the business start to increase a lot more in a multi-year, you know, sort of outlook, if you will. Another sort of point on some of our capital spending that has been elevated for a little while is, you know, we did close a significant acquisition with NeoPhotonics a couple of years ago, and there was significant capital investing in consolidating facilities, if you will, that we're still have more to go, but in terms of achieving synergies, but that's also an area that's kind of more rolling down in spending that than the data center can sort of leverage the same spending, that spending rate that we were doing.
Okay. Okay. Let me just pause and see if anyone in the audience has a question to ask. Any questions? Okay, let me continue here. So let's move to the telecom business for the last few minutes here. Firstly, just break that down in terms of what you're seeing in the telecom business. How does that compare to purely telco equipment demand versus DCI demand, which effectively is, to some extent, cloud demand, right?
Sure.
What are the trends there?
Yeah. So, you know, what's burdening our telecom business is that, you know, all end customers through the pandemic built, you know, purchased and a lot of inventory out of security supply concerns. And so, you know, about a year ago started burning off that inventory, and so therefore brought demand in general down. In fact, we first saw it with the cloud customers, cloud end customers, that is, and then that moved towards our traditional telecom network equipment manufacturer customers. And so, you know, fast-forward to now, what we're starting to see is telecom products that go into the cloud folks starting to grow. We sell some product directly, so that's actually grown nicely. We also more so sell products to other folks who then sell to the cloud guys.
That part is starting to move a little bit in the positive right direction, but there's still a little bit of inventory. So certainly we see a positive dynamic where the cloud, whether it be DCI or their own investments in general, long-haul metro networking, improving. And you know, we being sort of more down, I would say, than most in that space, just given where we play a role in the supply chain, is supplying two layers below the right? We're supplying the components. And so, we're starting to see those components tick up.
Whereas what we talked about on our earnings call and what we've observed more recently is that in the true telco portion of the telecom business, that's seeing a little bit of an incremental leg down, either due to slower carrier, cable MSO CapEx that either extends inventory burns at network equipment manufacturers that were in an inventory burn situation, or those that were not, then they're just seeing less in network demand, and therefore have less demand for our products. But at the end of the day, I think that's still a temporary situation, that network bandwidths and telcos are continuing to go up. And when you sell components, we, you know, we're kind of more purely exposed to bandwidth growth, if you will. That's essentially what our products are playing into.
So as long as bandwidth continues to grow, that inventory will come down, and eventually we'll be shipping more products. In fact, what we also highlighted on our earnings call that's important to note is, we're starting to see some new, newer products starting to tick upwards, whether they be higher speed components for new modulation or, you know, 1.2 Tb, 1.6 T system level, or some of the, you know, multi-band, or wider band optical switching products. And one, there's not inventory of those products, so you're starting from zero base, but two, they deliver bandwidth more cost effectively than historical products, and therefore, that does have a demand-creating effect when you can now get something that delivers more for less than what you have historically, and you still need it.
Okay. Good. I mean, maybe just to follow up on that, you mentioned it's a temporary headwind, but we've had some of your peers talk at the, or present at the conference this morning, and some of them have hinted at telco being maybe impaired more structurally this being more than a temporary headwind. How do you think about there being a more sort of shift in spending towards the cloud, and then telcos structurally spending lower over the medium term?
Yeah, I guess maybe I would concur that certainly cloud folks are spending, growing their spending more quickly on telecom equipment, data center and telecom, as you got to interconnect the data centers with telecom gear. So yes, but I'm not sure that's necessarily at the expense of telcos. I think it's just that they spend more slowly and grow more slowly in general, and are now dealing with inventory situations. I think, again, I go back to what we're selling is bandwidth growth, and if there's bandwidth growth in telco networks, then there's gonna be growth in our components going into telco networks.
If there's structural changes in the industry, where certainly we know there is a more of a focus on things like, I don't know, pluggable modules and things like that, but for us, we sell components to guys who build pluggable modules. We sell pluggable modules, so that transition is something that's less impactful, I guess, at our level of the supply chain.
Thank you. Good. Since we are close to the time here, I'll wrap it up here. Thank you, Chris.
Oh, thank you.
Thanks for the conference. Thank you
Thank you for having us. Thanks for tuning in, everybody.