Applied Optoelectronics, Inc. (AAOI) Earnings Call Transcript & Summary

communications and networking equipment at Cowen. It's my pleasure to have with us today Applied Optoelectronics. Stefan Murry, who is Chief Financial Officer and CSO, Chief Strategy Officer. You were thinking after what's been on now for 20 years...

Neither are the same grade back here.

Exactly. Well, it used to be 6 foot 2. But no, it really is my pleasure. So Stefan, let me start off, it's been a journey. I remember you guys when you were 85%, 90% leveraged to data center into cloud, and now you've got a cable business that's become, I think it's fair to say, meaningful and is growing, and it looks like you have a good outlook. And I like you to start off by explaining how the model is changing before we talk about what needs to happen for you to get back to profitability.

Sure. No, that's a great question. And thank you very much, by the way, for inviting me here. It's great to have the opportunity to get together with investors in person again. It's been a long time since we've done 1 of these lives. So it's good to see everybody not on a small screen, but real life. So nice to be back. So your question was having to do with kind of how things have been evolving in the business. And I think you hit on a very key trend for us, which is a greater diversification of our revenue. As you pointed out, at 1 point, not too many years ago back in the 2017, 2018 time frame. We were -- what I would consider today to be over leveraged to the data center or perhaps more specifically over leveraged to just 1 or 2 large data center customers. And so over time, we've made great strides in diversifying our product portfolio and diversifying our sales efforts to be able to attract the greater diversity of customers, not only between -- as you pointed out, our Cable TV segment and our data center segment, which is certainly the most significant trend that we've seen evolving over the last couple of years. But if you look a little below that, we've also made some significant inroads into 5G telecom for our telecom segment, small business, but 1 that I think has good long-term potential. And we've also made some investments in other areas, which haven't yet generated significant amount of revenue, but it's been a very deliberate strategy for us to try to invest in areas outside of our core data center and in particular, our core sort of large hyperscale data center customers and diversify out into other areas. Within our data center segment, we've also seen some shifts there, where we've also seen a lot of growth in some of the smaller data center operators, the Tier 2 type of data center operators, many of whom are sizable companies as well, but they're not necessarily that hyperscale Tier 1 type customer. I think that's partly a reflection of the efforts that we've made in attracting those customers. But frankly, it's also just a reflection in the fact that we're seeing much quicker adoption of newer, more advanced optical technologies by those so-called Tier 2 data centers, that is the Tier 2 guys really aren't Tier 2 anymore. They're starting to look a lot more like the Tier 1 guys in terms of their technology deployment, optics being a big part of that. So -- as the business has evolved, the Cable TV segment, which you hit on has been really growing very quickly for us in the last couple of years. The data center side has admittedly been a little bit slow. And I think that's because a number of factors there, but the biggest 1 for us is our focus has been on 400G. And in particular, a few of our large hyperscale customers have not yet really started to deploy 400G in the kind of quantities that we think they will be shortly. And so that part of the business hasn't really started to kick in. What has been kicking in, well, is the cable TV. So that -- the strategy that we've had to diversify our revenue streams certainly has been very effective in blunting the impact of the decline in data center as we wait for the 400G cycle to sort of kick in. Meanwhile, we've seen tremendous growth in the Cable TV segment. A question that we get a lot is that cable TV growth sustainable? And I think it is because I believe that we are in a sizable and probably lengthy investment phase in the cable TV industry. So if you kind of look back over the last 10 or 15 years, the cable MSOs have been able to upgrade bandwidth relatively easily by making sort of incremental improvements to their networks. So they basically increased the upper frequency content of their cable plant which they can do without really making a whole lot of changes to the network. Where we sit right now, the big concern among the MSOs, and it really has to do with the shift in consumer bandwidth usage is the ability for the MSOs to deliver upstream bandwidth return path bandwidth, it's that ability to deliver high bandwidth from the consumer back towards the cable TV network, which if you kind of take a long view on the history, that wasn't what cable networks were designed for, right? They were designed to transmit lots and lots of high-bandwidth, television video signals 1 direction, not really much bandwidth going in the opposite direction. And so the shift in consumerism where we're actually seeing much we would call it a symmetric network distribution where you're seeing as much data going upstream practice.

That's driven by real-time flows.

Yes, it's driven by real time sharing. I mean all the -- fortunately, we are in person here, but because this is easily been on Zoom where you're seeing your videos are being transmitted at the same speed roughly as the downstream bandwidth is coming in. And so you need a greater symmetrical nature to the allocation of the bandwidth. I don't want to get too much into the technology, but just say that the cable networks were fundamentally designed with a hard limit on how much bandwidth is utilized for the upstream versus the downstream, and there's a big difference between the amount of allocated bandwidth in the downstream and the upstream. So as the cable carriers now are looking to how do they improve their networks, this became a real issue in COVID in the early stages of the pandemic where there was a very rapid shift towards more symmetrical data usage. Over a short period of time, and it was across the board in parts of their network where they didn't normally see that, right? People are moving home and doing they're working from home. And so during patterns -- during parts of the day when they weren't normally doing a lot of upstream, all of a sudden, there is, and it really threw the MSOs networks into a disarray there for a little bit. They were really touch and go on their ability to deliver that bandwidth. And so that happened, that forced some sort of, I would call them, Band-Aid upgrades, certain things that they were able to do quickly which generated some revenue for us and for our competitors in the cable TV space. But it wasn't the ultimate solution. In order to fix the solution long term, they have to be able to allocate more spectrum, more frequency to the upstream bandwidth. And that requires real investment in the network, and it requires them to be able to make a network that can move the upstream -- or sorry, the downstream content further out, freeing up space for the upstream content and -- and those changes are fundamental to the network, and they're going to have to invest in new equipment to be able to do that.

Just for the benefit of those who may not know what you make the solution you're providing for those cable TV providers that play to them implementing greater symmetric bandwidth is what?

So we make a number of different products there. node products -- the cable TV network is -- many people may not know it's heavily fiber optics. So there's an office where the signals come in and get aggregated, and there's a fiber network that goes out to a variety of different network elements, hubs or eventually, it all ends up at a node, which is located in a neighborhood and node typically today might serve around 200 homes. And that's really where the network changes from being a fiber optic network to a coaxial network. So this moniker hybrid fiber coax or HFC that gets tossed around a lot really what that comes from. It's a hybrid of fiber and coax. But fiber makes up the vast.

And if I may, 1 way operators have increased bandwidth putting aside the symmetrical aspect is to constantly split nodes and which means we're really pulling the fiber deeper towards the end user and having decreasing the number of individuals served off of the net. .

Right. And that's actually what happened a lot during the COVID early stages that trend that I alluded to earlier, where they had to do these sort of band-aid fixes involved, node splitting. It doesn't necessarily, by the way, have to mean they actually put a whole separate node out there. It could mean that. But in a lot of cases, the nodes had the capacity to add additional optical to electrical converters within the node so they could put some additional modules in there and not have to replace the whole thing, which was very good for them because it allowed them to make those changes relatively quickly. However, once that's done, once you run out of slots in your node, then you really have no choice but to either replace the whole node or do something else. And right now, we're in that do something else phase, by and large. There's still node splitting going on, but it's a lot less than it was early in the pandemic. What's happening now is really trying to upgrade to what we would call a mid split or a high split architecture where it's actually moving the return path bandwidth, the amount of bandwidth, frequency content that's allocated to the turn path from 54 megahertz or somewhere in that range where it has been historically out to 200 megahertz or even 400 megahertz and DOCSIS 4.0, even envisions possibly moving out to 600 megahertz at some point in the future. And so that's multiples of the amount of bandwidth...

There's more spectrum available...

There is more spectrum which creates more bandwidth, right? And ultimately, you can only split nodes so far, especially economically, right? I mean at some point, it just becomes uneconomical to have very few customers connected to a node. So the only way to really practically increase bandwidth beyond that would be to allocate more spectrum to it.

Right. Before I ask about your datacom business, for those who may be concerned when they listen to EMCORE and what just happened with them, which I'm sure you're intimately familiar with the differences between what you do and what EMCORE does. It's not one and the same.

Right. No, exactly. AOI manufactures a wider variety of products. And in particular, we manufacture a line of amplifiers that are used in mid split and high split architectures so that newer technology. EMCORE's products, by and large, and I don't want to profess to be an expert on everything that they're doing. But certainly, the products that I'm aware of were largely targeted at traditional node splitting type of applications. And so as I mentioned earlier, early in the pandemic, no splitting was the only thing that they could do quickly. That no doubt benefited, it benefited us. I'm sure it benefited EMCORE and other players in the industry. But when that trend started to decline and the node splitting was no longer prevalent, that has a much greater effect on EMCORE than it does on us. And in fact, the newer architectures where they're going to mid-split, high split, are fundamentally much better for us because that's a much greater amount of dollars that they're going to have to spend to upgrade.

It's not a one set fits all. Let's talk about your datacom business. I want to discuss with you from 2 perspectives. You mentioned the fact that previously, for better or worse, you were highly, highly constant. There were 3 credit, you managed to win 3 web scalers the downside was you had dramatic customer concentration. And those guys, you probably know better than anybody, they drive hard bargains. And I'd argue the #1 issue from the deco has been the ability to drive acceptable margins and profitability in a competitive landscape with very challenging customers. And the good news is you're diversifying. They're smaller customers, but they could be good customers. And so my direct question to you is, what is the outlook? What are the hurdles to you getting to acceptable margin structure in that business? What needs to happen? What do you think is a realistic time frame of the opportunity?

So my observation on the data center business is that the -- just like any other business, it's not that unique, right? The stages of the life cycle of a product where you have a unique product or a product that has significant amount of technological content in there that other companies can't immediately duplicate are the most profitable stages of the business. We have not been in 1 of those periods for several years now. The 100 gig early on had reasonable margins, but it very quickly declined. And we're moving now into a 400-gig cycle, as I alluded to earlier, where because of the nature of the technology and the difficulty, frankly, in creating a good 400-gig solution for our customers, that's become -- it's a differentiator that gives us the ability, we think, to drive increased margins. And so that has not yet become a large part of our revenue. 200-gig and 400-gig business was something on the order of 1% of our revenue last quarter. So it's not a significant contributor now. But we did announce last quarter that we have a significant design win with 1 of our large hyperscale data center customers, who is about to start adopting 400-gig in the very near future, the next couple of quarters. And we expect to be a big part of that. The announcement was that were selected as 1 of their major vendors. And so that will be 1 event that we think can catalyze some margin growth. And certainly, as other customers start to adopt more of that 400-gig that will also drive margins.

For the bit for the audience who may not have listened to your call. That was 1 of the big 4 or outside of the big 4?

It was one of the big 4, 1 of the big 2, probably.

One of big 2. All right. The fact that not all of the hyperscalers are doing the same thing at the same time, defined by line rate throughput you've got Facebook doing 200-gig upgrade, Microsoft doing a 400-gig upgrade, correct me if I'm wrong, but I think Amazon and Google are doing 800-gig or about to. Does that create an incremental challenge for you In terms of having sufficient volumes?

Actually, I think it's good in a way or at least -- it can be good in the sense that we're not -- unlike some previous waves where it's been completely feast or famine, right? If you either hit that first wave of customer interest and time it precisely correctly, to have the amount of manufacturing capacity necessary right from the get-go versus being able to see more of a gradual sort of ramp and decline. I'll prefer the latter. It does create a challenge in discerning which technologies we want to develop and how much resources to allocate to the various different technologies that customers may want. And that's not always very apparent or in the early stages of discussion with these customers, right? I mean we are 1 of the major customers that all of these companies come to when they want to create a new technology, right? And so we have to make some decisions about. And at that point, it's not clear which direction they're going to go. I mean, they're looking to us and our competitors to give them options on technology direction, right? And so making that decision about, well, this customer is asking for this, but are they really going to deploy it and when, all those kind of things become more difficult in an environment where it's not clear that everybody is going in the same direction at the same time. So it's a bit of a -- it's got some puts and takes in terms of whether it's a good thing or bad. But all things considered, I think not having to ramp up very, very quickly and then the concomitant ramp down later on is certainly more healthy for the industry as a whole, I would say.

And before I ask the next question, I apologize. I know you and I have discussed about this in the past, but for the benefit of those who are improving in those conversations. Is the 400-gig landscape, is it meaningfully different than was the case at 140-gig just to set up again for those of may not be familiar, in sell after many, many, many years, I think, a decade of trying they finally appeared to have gotten it right many years ago and said they came into the market and just by virtue, even with Intel's relative challenges to the previous Intel, it's Intel. And so their balance sheet, their size, right, they do run losses on which even Finisar stand-alone or Today To 6, yourself, et cetera, nobody else can bear plus Intel light, which Google had helped both financially and technologically. And it also appears there's some other Chinese that maybe have closed the gap. But that changed the landscape. Is 400-gig is the minable landscape in there about pricing like -- and I recognize you still at the very inception, you mentioned 1%. Is it likely to meaningfully different, better from a pricing competitive landscape standpoint?

I don't see a lot of new entrants into the market at 400-gig. The barriers to entry get larger as the data rates get higher, generally speaking for a number of reasons. I mean the technology gets more difficult, just simple things like test equipment and things become much more expensive as the speeds go up. So both financial and technological barriers to entry become higher as we go to higher and higher data rates. And so I don't see a lot of new entrants coming in. There are a few companies in China. I mean China is a very -- as a country, it presents some formidable challenges in a lot of different ways. But certainly, in our industry, in particular, China has a stated goal of trying to promote their own domestic technology. So they have spent a lot of time and money from a government level and certainly individual companies that spend a lot of time and effort to try to compete in the 400-gig market or even the 100-gig market for that matter. That being said, I don't see a lot of new Chinese entrants. Some of the other Chinese companies that have been out there for a while are working hard to catch up in terms of the technology. That being said, I think there's a flip side to that, which is many of the western companies because of geopolitical events because of the way that China's economy shut down during COVID and other things that have happened. I think a lot of Western companies now are taking a hard look at how much reliance just like China is looking at how much reliance they want to have on Western technology. I think a lot of Western companies are taking a hard look at how much they're willing to source from China and whether they need to have alternative sources that are maybe less leveraged.

So you anticipate the clear suggestion being 1 or more of the Microsoft and Amazon, Google, Facebook and many others may be or if not totally exiting out the Chinese -- potential Chinese suppliers, they don't want to be 100% reliant on them?

I think that -- my guess, and I haven't had this discussion with all of those customers, I have had it with some of the customers. But my guess is all those companies are probably looking -- taking a hard look at how much it makes sense to be leveraged to China. They may all come to different conclusions. And I'm not saying any of them are going to be 100% non-China. But I think it's a much bigger factor than it was just a few years ago. So on the 1 hand, we have more competitors potentially coming in China. On the other hand, the barrier to them becoming a sizable supplier to some of these large data center operators in the West is higher than it was.

Right. And to your earlier point that China presents a unique set of challenges in that the Chinese government by virtue being an autocratic regime can do things that our capital societies like can't, correct me wrong, but last time I checked, an it's probably a lot greater than this number, but it was $43 billion of state money that was available to the optics industry. And you have a raft when I go to OFC, it seems like of the 200, 300, 400, whatever the number is, it seems like half of them are Chinese. Now 90% of those are 3 individuals and a dog. So I don't know that the real competition. But nonetheless, you've got folks like Hisense and Excellink or Acelink,I'm pronouncing it correctly, Acton a number of others that seem to have closed -- I don't know if they're able to ship to a 400-gig, but if they -- if the gap was like this, if they were 2, 3 years behind in 100-gig, it doesn't seem like they're 2, 3 years behind in 400-gig, hard to tell.

I think it's a little hard to tell. I mean, again, we're at the very early stages. So it's difficult to say who's able to really ship what at this point. So I think the jury is still out on that. I wouldn't want to make a judgment on that. I think it's probably fair to guess that the -- there is -- I mean, they're a little bit closer to just by virtue of having been involved in the 100-gig and gone through that another technology cycle, they're a little bit more mature in that. So it's reasonable to expect that they're probably a little bit closer. But is it -- are they a year behind when they were 2? Or are they 2 years behind when they were 3, I'm not really sure how to answer that question.

One more question on this line from a different perspective. You just had Juniper announce a JV with Synopsys where Juniper is putting Orion a private company that I don't think it's ever shipped $1 in datacom transceivers. But nonetheless, they created this JV in an effort to commercialize that, plus you had Cisco acquiring [indiscernible].

Well, they acquired a...

Product company.

A few years ago.

Yes, more recently, anyway, it was a datacom transceiver company, right? And so you've got some of the big boys now trying to ship their own datacom transceivers. Are you seeing any insight you could share?

Well, okay. I think there's 2 ways -- or 2 things to think about there. It is my opinion, and I think this is shared by a lot of people in the industry that the primary motivating factor behind those acquisitions is the eventual need to further -- to integrate -- and this is more of a technological thing than an economic one, but the need to place the electronic devices, the switch fabrics or the GPUs or whatever closer to the optics. So it's just -- it's a physics thing that as frequencies go up, the ability to transmit over a distance goes down, and so you've got to put things closer together as frequencies go up. So it's my opinion that the primary driving force is they want to have the technology so that eventually, when the speeds get so high that they can't practically use, say, a pluggable optical transceiver because it requires too much of a distance from the transceiver to the CPU that they'll have an option right.

What I referred to is onboard optics.

Onboard optics or there's a whole slew of different acronyms that go around it, but it's basically the idea that you've got to put the optics and the electronics closer together at some point down the line when the frequencies become very high. So I think that's still the driving factor. That being said, they own these assets where they're developing these assets, and there isn't right now the need for onboard optics, right? And so there's not going to be a revenue stream from onboard optics immediately. So they're looking at opportunities to what can they do in the meantime to utilize those assets that they develop to generate new revenue streams. And I think that is going on. I'm not that familiar with the Juniper, the recent Juniper acquisition. So I can't comment on that intelligently. But the Cisco acquisitions, I think that's a fair take that they're -- they continue to sell optical components on a merchant basis, while no doubt internally using that technology to try to develop an onboard optics fiber.

Details aside, I trust you clearly believe that pricing can be sufficient, if not maybe robust is the wrong word, but at least sufficient that you could generate acceptable economic returns eventually in your data, which has not been a base in quite some time. But I would hope that you believe that otherwise you'd be knocking your head against the wall and a true definition of insanity.

Yes. No, we do believe that in particular, AOI's primary benefit on the -- really in all of our businesses, but especially in the datacom business is the fact that we have our in-house wafer fab and we have the laser technology and the photodiode technology and the integration technology and the manufacturing experience for those types of optical components. And that's where we expect to drive the margins really. I mean the assembly of optical modules for better or for worse, is not a high-margin business, right? The margins are driven by IP and our IP is primarily resident in our wafer fab. And so the greater content, the greater need for these high-speed lasers that are used to 400-gig bodes well for us. The -- some other areas of business telecom, which I alluded to earlier, also has a similar aspect, right, where the lasers are driving the need -- the technology is being driven by the advanced lasers, that's good for us as well. So those -- and those are the areas where we're going to drive the strongest incremental margin improvement.

1 more or 2 and maybe you can remind the audience to get back to cash flow positive and bottom line profitability, what gross margin do you need?

We need mid-20s.

And when do you think you can do that?

Well, it's a bit hard to say right now. We haven't put out a number publicly yet. I think it really depends -- the reason why we haven't been able to do that just yet is because the -- the trajectory of 400-gig, we know very clearly now we've got our initial design wins it's a little uncertain how fast that ramp is going to go, primarily because of all the supply chain disruptions that we have. We know how fast the customers would like to go, but remember, when we're doing a major network upgrade like 100-gig to 400-gig, all the parts of the 400-gig ecosystem need to be there, right? All the servers, all the racks, all the switches, all the cables, all the connectors, all the fans, all the fuses, right? Bel Fuse was here just a minute ago, talking about their -- I mean everything in there needs to be available to be able to offer those upgrades, right? And that -- it's not clear when that is going to be completely resolved. And so that's why we're hesitating to say I will say it's -- a lot of those things are not likely to be resolved by the end of this year, but I think we'll make good progress starting in Q3 when the 400-gig business starts to incrementally improve.

Yes. And just on my response, while there are different levers, different variables such as mix and volume, among others, I trust the key relates getting the volumes in order to get that 25%?

Yes, it's volumes, but it's also, again, the 400-gig in and of itself at this stage in the life cycle, it should have better margins, even at relatively lower volumes.

What revenue volume you think you need to get to 25%?

What revenue do we need to get to 25%? Again, that's a difficult question to answer because it depends on product mix. But $65 million, $75 million, somewhere in that range, we can probably do it.

Let me see anybody wants -- if anybody's got a question. I've got 1 last question for you. You all have very fine technology, but -- what's the chance strategically they say, you know what, we can sell our lasers. We have great lasers. We could sell them to third parties, it would be a better business than selling modules. Is that something you all have considered?

It's certainly something that we think about. As I mentioned earlier, we view our laser technology as our primary driver of sustainable competitive advantage and thus gross margin. And so the trade-off -- that being said, we also have great manufacturing technology around the transceivers. We've invested heavily in that. That -- to your point, as the Chinese continue to invest and develop new technologies, that's the easiest. That's the low-hanging fruit, right? Making the lasers is probably going to be the last thing that they're going to technologically be able to do. So the incremental technology advantage that we have in the assembly processes is probably going to be eroded quicker than the incremental advantage that we have on the laser side. And so that would probably be the first thing to consider.

In 10 seconds or less, you've had very low cash burn, you don't have a robust balance sheet, but you kept your firm very low. Any concern that you're going to run out of runway before you get to profitability?

No. I mean I think we've got a lot of levers that we can turn on that. So I'm not concerned about that.

All right. In the interest of not allowing anybody further, I want to thank Stefan and AOI. I want to thank all of you. As always, if I can be any help on Applied Opto or any other name, please do let me know it would be my pleasure in for which.

Thanks.