Viavi Solutions Inc. (VIAV) Earnings Call Transcript & Summary

Hi, everyone. Good afternoon. Welcome to the Rosenblatt Age of AI Technology Conference. This is a fireside chat with Viavi Solutions. I'm Mike Genovese, the cloud and communications equipment analyst. Super pleased today to be joined by the management team of Viavi Solutions, Oleg Khaykin and Ilan Daskal. And we also have Vibhuti Nayar from IR on the panel as well, just off camera right now. So we've got the whole team and hi, guys. Great to see you.

Mike, good to see you again. Been a long of time. A whole week.

Well, bear with me here because my first questions tend to be a little bit worry and have a little bit of wind up to them. But I just want to level set the audience here because you just reported fourth quarter, as you said, fourth quarter '26, the June quarter. And you said that data center within your NSE segment, your big segment, was about 50% of the revenues, aerospace and defense, about 17% and telco is the rest. So given that, I mean, it looks like to investors that 800G test transceiver-type testing is probably the highest volume lab and production driver. And there's a surge in transceiver production across North America and Asia. So what do you view your market share is in 800G and Ethernet testing today? How does that differ between the Western and the Asian transceiver makers? And how much tailwind remains in upgrading from 400 to 800 before 800 peaks?

[ Let me ] remind before Oleg addresses the first question, just [ reading ] in the safe harbor, [indiscernible] we get in front of [indiscernible]. Obviously, we'll be discussing today some forward-looking in nature's data. And everything is subject to recent uncertainties, as we all know, and everything that we discuss today is for today is only, and we undertake no obligation to update these statements. [indiscernible].

Thank you. Your line is also our general counsel [indiscernible]. So your question again in terms of market share monitoring, means, in testing measurement is kind of a difficult number because there's no third-party sizing of this thing. But I'll go the following. So if you take top 5 transceiver manufacturers in the world, we'll be working with all of them. The question really becomes to what extent and how deep we are. I would say our level of engagement is very high with probably 4 out of 5, 1 of the 5, they may be doing some internal engineering so that a bit less. But I would say from #1, #2 and #3, we are very much in all of their labs, so they use our equipment to design it and then they buy our test equipment to build it out. But it's no longer just a -- when we talk about production, test and measurement our market has been expanding significantly. In the old days, when you did a 400 gig, a, just in production, you measure a deal. Is your laser working? How good is your dispersion? Power in, power out. And kind of [indiscernible] and it's done, right? As you get into 800, you now worry more about lasers, so you need a more precise, better laser measurement. You're looking at the insertion losses, you're looking at some maybe transceiver test, you do more. As you go to 1.6, it's a whole new ball game, you're getting much more optical and you have a more extensive test and now you're actually doing a lot more functional test on the line. So the content for test is actually increasing, not only there is an intensity of test, but there's also more equipment you have to buy to test these things. And that's a traditional market. But then there's a whole thing with the fiber with the [indiscernible], multicore, multi -- like ribbon cable that you plug in like 8, 16, 32. Before we never really cared about it. Now the customers worry about make sure that connector is good because if 1 fiber is bad, you throw away the entire connector. And these are very expensive connectors. So you're looking for how clean is your surface? What is the interferometry on the quality of the surface because you're at the higher the speeds you are using any kind of imperfection can impact the quality of service. So we're opening all new markets now in manufacturing for fiber. And now the last but not the least, that's probably going to be the biggest segment when we now start talking about CPO, that's a completely different ballgame. We were never in the semiconductor test but now we are being pulled in right there with the likes of Advantest, Teradyne, and form factor in that setup to do the [indiscernible] testing. So I mean, it's no longer just a play of transceiver test.

Yes, there's a lot of different things to follow up on there. So we'll go one at the time. Well, so a transceiver company that I follow on their earnings call said, "Well, we've built our capacity so that we can make 800G and 1.6 transceivers on the same equipment, really all we have to do is we have to buy new test equipment for 1.6." Which I think is what you sell, which is a great thing. And I also think that while the transceiver price maybe as you move from 800 to 1.6 is as much as double, what's the ASP up rise for you moving from 800 to 1.6. But where are we? I mean, how early are we in 1.6? How much have you seen so far? And as you look at the next 1, 2, 3 quarters, is there an inflection in 1.6 happening right now?

Well, we've been selling 1.6 equipment for now for over a year, mainly to the developers. It's more or more of a lab system vendors, module vendors, silicon vendors. The production, I think, started, I'd say, probably early production beginning of this year. And it's still, I'd say, maybe third inning for 1.6, second, third inning depending on what we're talking to. And remember, 1.6 is an interesting thing because as far as the test equipment goes, it's always backward compatible. 1.6 can test 800 gig, and it can test 400 gig. So with 1.6, you can test 4 400 gig modules or 2 00 gig or 1.6. So it's very good in terms of economics. The cost per bit is much lower. Now the cost per box is much higher. So often what they do is, now, they are going to throw away 800? No, they probably will move them to test 400 gig and things like that, but they will buy new stuff because the production line stays the same, but now with 1 high-performance tester, they can service multiple lines at the same time.

Got it. And the ASP outlook, I mean, I think you said in the past about 50% that box [indiscernible]?

Generally, you see about 50% price increase. And that's because I'm very simple. I mean, the components are much more expensive. So early on, you always kind of -- I mean, what you say is a cost per gigabit goes down. Your cost for the whole box goes up, about 50%. And it's not different from what you see with wafers. When TSMC goes from 1 node to the next node, they don't leave the price per wafer for the same. It usually goes up about 1.5x

Now. I think you've also -- you said in the past that your market share of testing for optical circuit switching, OCS is extremely high. And just this last quarter, we had very, very positive guidance on OCS saying this current quarter will be a $100 million quarter. And then I actually just spoke to Wupen, so this is Lumentum. You're old...

Our cousins, yes.

Your cousin, yes, separated it later in life, not at birth, but separated later in life. But Wupen been said, look, over the next several quarters, it's going to double, it will be $100 million plus this quarter and then double that the next quarter and double that the next quarter. And I assume that you'd be seeing [indiscernible] related to that.

I just say -- the saying goes, his mouth to god's ears. And the more the area I love these guys. And I hope all of that comes through because clearly, it's a great news for us. And it's not only the -- then because remember, there's also some hyperscalers who do it directly. And I think it's catching on people are realizing there's not enough power in the world to power all these data centers more and more, people adopt more optical architecture rather than electrical to do more in optical domain. And you could say that from a sustainability point of view, that's a much more environmentally friendly approach. I think -- and that's going to be the case. And I mean when you're doing this very big switches, you always ask yourself, what's the cost of failure that 1 of those port payers doesn't work. Well, the cost of failure, you switch is back. So I mean, they test everything to the ampdegree making sure that everything leaves the factory, every connection works and that requires significant investment in test infrastructure. And that's why we said, as you go into higher speeds and you go to more of these complex products like OCS, our content is actually growing per test versus the traditional simple pluggable transceivers.

So can you be -- just on that market share comment that you made in the past, I mean, is it only the OVI doing testing in this market? Or are there others involved?

Well, there are some people may be making in-house equipment. But generally, I mean, anybody who is buying merchant equipment, we are the dominant supplier.

And is there a way to think about this question, which is -- now the OCS market, 1 company, Coherent has talked about a $4 billion TAM. Lumentum, I think, actually said, we think it's going to be a $10 billion TAM by 2030. So let's just say it's somewhere in between those numbers. Is there a way to think about it? Is the percentage of the TAM that would then be the testing opportunity?

Well, give me a couple of quarters. So I have more than 1 a couple of data points. We can tell you the trend because 1 quarter doesn't make trend. But generally, I think when you look at the size of the market, depending on the capacity, you say 10% of the annual size is what the test and measurement market ought to be.

So that's a remember I've been using. But I've been using that number more for OCS and for CPO that has more complexity. And then for transceivers, I've been using a lower number. Is that...

It's a lower number for transceiver. I'm talking about the higher-end stuff.

Yes. Okay. Great. Good.

And who Knows? I mean, because initially, when they start just everything, over time, as the yields get better, they may do sample tests or every other port or things like that. Although for 100 terabit switch, I don't think anybody is going to skip any tests.

Great. Well, before we get into CPO and NPL, I wanted to ask a question because it seems like in your business, right, I mean, so data center and aerospace and defense, 2/3 of NSE revenue is now more than probably half of the revenues of the whole company, just hitting that to be more than 50%. I mean, it's growing very, very fast. And then just people think of, okay, telco is a low-growth market. But I'm wondering because I keep hearing more and more and bigger numbers and more bullishness about scale across. And maybe it's a little bit early, but we're just -- it's just maybe a low billion-dollar market now, but people are talking about cross going tens of billions of dollars within a few years. And so I'm wondering, could that be a driver of your telco business, if you're...

No. Absolutely. So when we say maybe there's -- let's clarify some definitions, right? When we say data center ecosystem, if we are selling to like a Tier 2 player like, say, Lumen or Zayo who are interconnecting all those data centers, right, we count it as a data center revenue, right? Because to us, it's who pays for all that equipment. If Verizon sells to data centers, will count their revenue as a scale across. When we say service provider, that's your traditional metro fiber to the home, the wireless carriers, these kind of laps, right? So today, I mean, that business is growing low single digits. Although I'd say wireless is fairly anemic, but the fiber business is doing very well. Cable is actually going to have some really nice pop-up quarters as they go to the DAA architectures. So even that business is actually doing pretty well, but it cannot compare what we've seen with the data center ecosystem for us, it's semis module systems, hyperscalers and the people who interconnect them, right? So that's what we say when we say data center ecosystem.

Got it. So basically what I'm asking about scale across -- I mean, obviously, it is big data center operators building these things that is counted. And then -- if you're selling to Ciena for that application to a hyperscaler, I mean I should -- or Ciena Nokia, Cisco, that kind of company, that kind of OEM company, you would also include that data center because...

It counts as a data center. Exactly.

Got it. Got it. Okay. Helpful. Super helpful. Okay. Well, let's talk about CPO NPO. Investors were freaking out that CPO is delayed. But now everybody is reported and all the vendors and everybody said, "No, it's right on track." And it's -- the demand is higher than before. And CPO is basically NVIDIA but there's all of these other companies that want to do NPL and NVIDIA wants to do NPL as well. So what are you seeing on the timing for NPO CPO and when and how that hits your revenues? And then we'll have follow-up questions on all these insertion [indiscernible].

Why do people want to do CPO? It's much more complex and all that. Why not do NPO? Why not do just pluggable? Well, it's very simple. It's power and performance. I mean, in a nutshell, if you really want to get really high level, if I have n-1 no silica, and I put it into the advanced heterogeneously integrated module with a CPO. I get a performance of the end node, right? So if I have a 3-nanometer silicon, and I use all this clever packaging technology, I get the same performance as a 2 nanometer without paying the 2-nanometer mass cost and all the development, it's a huge economic stimulus for me to be able to do it. But yes, if I do have a 2-nanometer silicon, and I still also do all this other stuff. I get a performance of 2x, 4x better performance than otherwise. So that's what it is. So this is the cost is not an object because you're -- it's all about making the most of your power budget that you have and making the most your performance that you have. So it's your bandwidth and the thermals that you generate. And also you need fewer lasers if you're using CPO, right? So I mean that's what it's all about. Now it's a very complex technology, and there's 2 caps. There are haves and have nots. There are companies who are far ahead on CPO, who have been doing it for years, and you named 1 of them. And then there are those who have been a laggard thing, it will happen when it happened. Well, the ones who don't have it, they are [indiscernible] a lot of it. But I'll tell you 1 thing, those who are ahead of it. They're going full speed, and they're going to have a performance advantage that the others cannot dream of. And for us, it's actually I mean, I want to improve disclosure, the more we move towards CPO, the greater our content per port goes up, right? I mean, pluggable is ANA power dispersion insertion loss. As you go to CPO, it's multiple insertions. It's a lot of functional test, much more advanced test and fundamentally, you have much tighter customer intimacy, working directly with them from the point when they are laying out their silicon and thinking about packaging to all the way until when they do the deployment in the field.

Right. Right. So yes, look, I think you mentioned some of the players that you -- that are also doing various parts of testing measurement. And I think in CPO NPO, it's a larger group of the test and measurement players versus, say, transceiver modules. But the question is, is the content are the insertion opportunities, the numbers of times you test and the complexity of the test. Does that actually make your opportunity in a CPO world larger with a smaller share then versus transceivers. I mean how do you think about that?

Well, I don't know, actually. I think as it gets to a point of CPO, we should probably have a higher share because what we have is an integrated platform that is automatable and we have 2,500 SKUs for any kind of measurement modules you can put in a mix and match to make it. So you're generally going to have an IC tester in the configuration companies like Advantest to Teradyne. You're going to have a prober form factor. And you're going to have optical rack that has all the instruments. And they are going to be interconnected and they are going to be testing different parts of the equation.

That's helpful. That's a good way to...

It's not a point solution kind of like you put a little box here and there. It's all got to look very automatable and it's got to be all integrated. So there are companies that do nothing but integrating all these pieces together at behalf of our leading players to provide them with the platform to test our products.

Can you kind of walk us through the process of creating the optics, creating the laser, what tests are needed and then you build an optical engine and what tests are needed and then you build an ASIC and [indiscernible]?

Look, we talk about yield management. It might -- in our past life, I ran an offset player, right? And whenever you talk -- you heard the customer say, I want a system in a package today, they call it heterogeneous integration. He had a cringe because the question was who is going to take the yield loss. And the more things you pack into this thing any 1 thing goes wrong and this module is bad. The question becomes who is liable for the throwaway module. So what you do in these kind of things is you do extensive known good start with that. You take a long, good die for electronic IC. We take a good known good guy photonic integrated circuits and you just see each of them, make sure they are good. Then that's your first insertion. Then you put them together and you build an optical engine, okay? Now typically, electronic IC should not change. Normally, it's come through many cycles through reflows and all that. So it's going to work just fine. The problem you have with optics, you can change the characteristics of optics. So then now you've got to retest their optical engine, make sure it's still good, and you get the first fall out, then you throw those away. And you only pass through the non good optical engines. Then you take all these optical engines and combine it with a very complex substrate, right? So you put them all on it and you connect all the cables, everything. And now you test this full contraction, make sure that you have a known good interconnect platform, okay? And only when that thing is good, do you put down the ASIC and memory on it because now the price of that module skyrockets. And when it's all said and down, you're talking thousands of dollars, right? And then you test the whole system. So if you don't do all these other things ahead of it and make sure that all the pieces that you're putting together and the final product is good, and you just go ahead and build it and they put ASIC, you will be throwing away tens of millions of dollars, right? So that's why it is so critical because optics is more like an analog rather than digital. If you don't handle it right, you will change the characteristic of optical components.

Yes. Makes sense. Okay. And then on the timing of -- I mean is there -- some of the companies that are making lasers and things in this quarter's earnings such as well, we'll see maybe NPO demand 1 quarter before we see CPO demand. Are you able to see things -- are you seeing that similar kind of thing?

So generally, customers make a bet when they order equipment. So that -- I mean, for example, we've been working on a CPO test now for well over close to 2 years, right? So we're now starting shipping, I'll say even this quarter is currently shipments. So generally, the capacity is being facilitized and it's being put in place. At what point, usually you put a test capacity ahead of the demand. And then as it gets to about 60%, 70% utilization at the next incremental capacity. So I think lasers, I mean, they are more like quarter-to-quarter forecast driven.

Okay. Is there anything else on the sort of the data center business that you want to highlight before I move on to...

I think the -- when we talk about level production that has been the biggest beneficiary of the data center growth. But 1 big surprise for us was our traditional field instruments. I mean today, that business is over 40% data center [indiscernible]. And that's really advanced stuff as fiber monitoring. Because what data center operators have realized is you need to police people who build your network because it's like -- to paraphrase Forrest Gump, "Life is like a box of chocolates. You never know what to get." Well, the same thing was true for the -- you can use a quote on your note. If you're a data center hyperscale buying a fiber network ended up being like a box of chocolates. Sometimes the fiber work to buying a lot of dark fiber, you don't use it. But when you turn it on, you find it doesn't work. So they wisened up very quickly and said, "You know what, I'm not going to play this game. I'm going to monitor the whole -- I'm going to get really smart on the whole networking." And they extended they're putting fiber monitoring on the edge. And they are monitoring all the buildout. They're making sure that every fiber is truly connected and tested and qualified and they are monitoring it all through. So when they are ready to turn on more capacity, they can pick exactly the fiber that gives them the best performance. So that has driven significant growth for our field instrumentation business because now they really appreciate the whole automation and the workflow automation, workforce automation and monitoring and collecting all the data so they can predict any kind of outage in their network that may be coming based on the parametrics that they're collecting while they monitor the fiber. I mean, I think if you ask me who is the state-of-the-art fiber operator today, it will be the hyperscalers. They know more about fiber optic networks than the traditional fiber optic operators.