Andrew Schmitt created some excitement when he spoke about optical component and equipment market trends on a Piper Jaffray investor call on April 22. 83 investor clients were on the line, and some insightful questions were posed to Andrew as he gave his observations on a wide range of topics. The following is a full transcript of that discussion.
Troy Jensen: All right, thanks everyone, for dialing in. As everyone knows, we’ve got Andrew Schmitt on the phone with us. Andrew’s been a long-time contact, a good friend of mine. I’ve known Andrew for probably seven, eight years now, but anyway, Andrew just got back from China and Japan … I think it’ll be a very timely call. Andrew, why don’t you just quick start out … There’s going to be a lot of people on the call that may be new to you, probably a lot of people on the call a month or two ago that we hosted, but just quickly go through your background if you would.
Andrew Schmitt: Sure, Troy. I started out in engineering doing high-speed semiconductors for telecommunications back in the ’90s and worked with a semiconductor firm all the way through 2005, and then left and have been doing consulting for, first, people in the financial sector and since 2009 working primarily as a consultant to industry as a market research analyst. I worked for Infonetics and later IHS for about six-and-a-half years. I left that position at the end of 2015 and started a new market research firm called Cignal AI, but I’ve been involved in the optical business from a management and engineering and independent third party perspective now for over 20 years.
Troy Jensen: Perfect. We are going to open it up for Q and A two-thirds of the way through so you guys can ask questions if you want. You can also shoot me emails, and I’ll filter through them and ask also.
China 100G Coherent
Troy Jensen: Andrew, I just want to start with China. Obviously you just got back. I’d love to spend some time really focusing on that, so just let us know where you went, who you saw, and we’ll start out with what some of the key takeaways were.
Andrew Schmitt: Sure, I was in Asia for two weeks. I spent a week in Japan, and it’s kind of overlooked but there’s a lot of component suppliers and subcomponent suppliers in Japan that feed into China. It’s a really good place to get information on the direction it’s heading and what’s going on there. After that, I went to China and met with Huawei, as well as ZTE and really tried to focus the discussion on spending plans for 2016 surrounding 100G coherent. I was able to get information from the supply chain and information from Huawei in terms of what sort of quantities they’re looking at for the year, and everything was remarkably consistent in terms of the numbers. Particularly in the supply chain there was just profound shock at the magnitude of the business that’s coming from there.
Troy Jensen: Kind of quantify a little bit there if you could.
Andrew Schmitt: Yeah, let me put this in perspective. In 2015, Huawei shipped just over 50,000 ports of 100G coherent, and this was a mix of things that they built themselves using parts from NTT Electronics, as well as some of their own high-silicon components. Their plan at this point is to ramp that to over 100,000 units this year. In fact, the number that they’re talking about is 120,000, so more than doubling, and doubling from what was already a pretty big number.
Troy Jensen: So from 50-to-120,000, okay.
Andrew Schmitt: Yeah, and the way they’re modeling it is their … They see that in Q1, this quarter, or actually no, last quarter, they shipped a total of around … Let me check my notes here, 35,000 ports, and this is in support of a series of tenders from China Mobile, China Telecom, and China Unicom. Huawei taking the vast majority of this business. ZTE got a material portion of China Mobile in the western part of the country, but really it was primarily Huawei that took all of the business. ZTE and Fiberhome kind of split the rest.
Troy Jensen: Okay, that seems normal over there, but can you talk specifically about the projects, these metro upgrades and these core upgrades?
Andrew Schmitt: This was not … The stuff that was bid in Q4 and shipped in Q1 was more of the backbone, and Huawei is indicating that in Q2 they’re expecting a similar size of shipments to go into the regional portion of the networks, so it’s selling down into each of the individual sub-states, the smaller domains that make up each of these networks. I think China Mobile is divided into 16 or 18 different regions, and each of those regions has some autonomy in how they pursue and build their networks. Huawei’s expectation was that they would also be bidding in Q1, buying in Q2, and supporting a similar amount of material. They saw the same cycle repeating again in the back half of the year, so that’s basically 35,000 ports per quarter flat through 2016. This is consistent with the guidance that they provided into their Japanese supply chain.
Troy Jensen: Okay, I got a question, an email in. Can you just clarify what you think the split is again? You say ZTE was 5% or 25%?
Andrew Schmitt: I’d have to go run through the exact numbers, but for instance, at China Mobile … Let me just run a quick number here. ZTE would have gotten about 15%, 14-15% China Mobile, and substantially less at Telecom and Unicom, but the majority of the business that was awarded in Q4 was China Mobile.
Troy Jensen: Tell us about sustainability of China. It sounds like these numbers imply that this wasn’t just kind of a first half project, correct?
Andrew Schmitt: Yeah, everyone’s been waiting for the Minsky moment in China, and it just never seems to come. I don’t see any reason at this point that … The guidance that they provided to me and that they have also provided to the supply chain comes from two different sets of people. I was speaking with a lot of the marketing folks at China, and the supply chain and engineering guys are the ones who talk to the component guys. In the past, I’ve communicated with ZTE and seen that the numbers don’t necessarily line up and it’s inconsistent, and that always raises some flags, but when you talk to the folks who are selling into Huawei, these are the same kinds of numbers that they’re talking about. At least Huawei internally believes that they can do this kind of volume.
Troy Jensen: Okay, I apologize if you answered some of this, Andrew. We had some problems with the dial-in number, so I was emailing some clients here, but as far as the projects, the China Mobile primarily is done sort of the core refresh?
Andrew Schmitt: There’s a core refresh that was bid out in Q4, and Huawei is saying that shipped in the Q1. They’re think that that’s going to significantly swing their market share in China as a result. Then in Q2 they’re expecting to ship a similar amount into the regional portions of the network. China Mobile still would be purchasing, but it wouldn’t be going into the backbone that connects all the different regions. They would be selling into the individual regions.
Troy Jensen: Think of that as metro upgrades, then?
Andrew Schmitt: Yeah, for lack of a better term you could call it metro, but it’s not necessarily within the city, but it’s connecting the suburbs within the … You could use the term metro, sure.
Troy Jensen: Okay, I get you, so key products. Help us out with like vendors that are best positioned in China Mobile, and I want to do the same thing with China Telco and China Unicom.
Andrew Schmitt: Well, I don’t think there’s too much of a difference in terms of what the three guys are buying. If you look at what Huawei is shipping, they’re shipping the same products to those three guys. On the ROADM side, which maybe we can get to later, it’s a little bit different, but as for 100G coherent, Huawei is doing things a little bit differently than the folks in the west. There’s been a big move towards CFP2-ACO among the guys like Cisco and Ciena.
In China they like to use what I call communist engineering. If you’ve ever been through Eastern Europe and seen the kind of stuff that was built in the ’70s and ’80s, it’s just fortress-like architecture. That’s how they’re building their 100-gig networks. They are not using any … They’re using very little 200-gig. In fact, most of the stuff that China’s shipping at 200-gig is going to go overseas. It’s basically brute-force, dirt-cheap 100-gig coherent, and they’re using pluggable CFPs. Some of this has been provided so far by Acacia in lower quantities, and I’ve done some work to go break out exactly what Acacia’s revenue stream looks like into ZTE and into China, but what they’re working on now at Huawei, and this is the guidance from the Japanese component suppliers, is they’re basically building their own pluggable CFP using their own DSP silicon and then buying components like coherent receivers and modulators from various suppliers.
I would think that NeoPhotonics is front and center in this effort because of their … I think they do something like half the revenue of Huawei, if I remember right.
Troy Jensen: Yeah …
Andrew Schmitt: It’s pretty material.
Troy Jensen: 68% was China last quarter so …
Andrew Schmitt: Yeah, so lasers, ICRs, I don’t know who’s actually supplying them to Huawei, but NeoPhotonics must be selling something, so I would think that they’re front and center in this effort.
The other thing that’s happening … Basically, to summarize, China is going forward with 100G coherent; 200G is really not important for within China at this point. That’ll change maybe in 12 months, but they’re building their own CFPs. They want to put those into production this year and then use those as their way to deliver 100 gig, the CFP-DCO, so they’ll be pulling from a number of component suppliers to build those on their own.
Troy Jensen: Okay, let me hit that. On the CFP, so it’s CFP not CFP2, you’re saying?
Andrew Schmitt: Correct, and that’s digital CFP.
Troy Jensen: Digital CFPs, okay.
Troy Jensen: With Huawei, right, kind of legacy products they start to try to in-house more and more?
Andrew Schmitt: That’s right. Their objective is to have at least 30% of the stuff built in-house, and like at 10-gig, it’s already well beyond 30%, so I think they’re going to be looking to take this as far as they can in terms of building in-house. At this point, there’s not too many people who sell pluggable CFP coherent. Acacia has that technology and some of the Japanese have it, but Huawei won’t make a major move towards using this format until they have their own internal supply chain turned on. I don’t think it really has much of an impact on the supply chain. Instead of them building their 100-gig line cards by soldering everything down to a board, they’re going to solder everything into a box, and then ship that little box and plug it in with their equipment, the CFP.
Troy Jensen: They’re just going to create the density by getting it into a smaller and smaller factor but still buy all the components through the suppliers?
Andrew Schmitt: Yeah, that’s their strategy. It’s a little crazy compared with what the West is doing, but it’s in line with what I like to call their communist engineering strategy.
Troy Jensen: I did get a question from a client here on China, and we’ll jump ahead. It’s a ROADM question, but it said, “Is China using ROADMs within their domestic networks, and assuming capacity increases, increased wave lengths require more sophisticated ADM so ROADMs are needed?”
China 100G Clients
Andrew Schmitt: I’ve got some good detail on a ROADM because things have really changed over in China and I was educated, but before we move to that, just to finish on the 100-gig, the CFPs …
The other thing that’s starting to happen and what Huawei has indicated is that there’s less and less 10-gig client and more 100-gig client. They’re looking to buy things like CFP2s to go with the 100-gig coherent that’s being shipped. Before the 100-gig coherent might handle 10 10-gig clients, they’re seeing a greater mix shift towards 100-gig clients, so that’s going to drive demand for pluggable CFP2 client-side transceivers, the kind of things that Finisar, Oclaro, and Lumentum make, so that’s creating some demand there. You talk about where the demand is going to hit, it’s going to be the coherent components that go into these CFPs and the client interfaces that are going to be paired with them in CFP2.
Troy Jensen: Okay, so in general it sounds like the same guys we saw in Q4 that had the strength. It was Lumentum. It was Oclaro. It was NeoPhotonics and also Finisar to some extent.
Andrew Schmitt: Yeah, I don’t see anything changing going forward.
Troy Jensen: Okay. All right, then, so how about … You want to focus on the ROADM opportunity now in China?
Andrew Schmitt: Yeah, I’ve been meeting with the folks in Beijing every year for a while. You ask them about ROADMs and why they’re not doing it, and it was just never really … They didn’t want to use the technology, and it never really made sense. They like, again, the brute-force approach of using these OTN switches and electrical switching and then manually cross-connecting things. I didn’t meet with the folks in Beijing this time around, but I did talk to Huawei and I talked to the guy who basically runs their whole transport business about ROADMs. He said that things have really changed and that the carriers there now have rotated over towards looking at ROADMs because they have completely blown out of these electrical cross-connects. They just can’t make them big enough anymore.
The other thing that’s happening is that the carriers like China Mobile and China Telecom are perceived to be extremely slow in servicing the needs of a lot of data center and the enterprise companies like Baidu, Alibaba and Tencent. They’re putting these ROADMs in place also in the metro networks so that they can more rapidly reconfigure the networks to sell services to those guys.
There’s two demand vectors that are coming in. China Mobile right now is the one that’s moving most aggressively towards using the ROADMs, and they’re just starting to look at deploying them this year, so I don’t think there’s going to be a lot of volume this year, but I think this is going to be a major trend as the year is closing out and going into next year.
Troy Jensen: I just want to point out that’s a key point because at OFC and what we had heard from Finisar and Lumentum is that it was China Telco was the one that made the first decision to do a ring-around Shanghai, and that if China Mobile goes, they could be multiples bigger with respect to ROADM demand.
Andrew Schmitt: Yeah, according to Huawei, China Mobile is the one that’s moving now, and Telecom and Unicom are really just evaluating the technology. That’s what they communicated to me.
Troy Jensen: How much bigger is China Mobile versus China Telco or Unicom when it comes to capex and kind of content?
Andrew Schmitt: On an overall basis, I’m not sure. Everything has really been skewed over towards China Mobile simply because they deployed 100 gig so much more aggressively than everyone else, and the amount that they’ve been purchasing is disproportionate to the size of the company. I think that their spending on 100-gig has been a greater proportion of their capex than Telecom and Unicom because Telecom and Unicom relied on 40-gig technology for a very long time, and they deployed that aggressively where China Mobile did not. There’s been a bit of a spending mix shift away from Telecom and Unicom toward Mobile because Mobile sort of under spent in the 40-gig domain.
Troy Jensen: All right, understood. Then on the ROADMs, obviously Finisar and Lumentum, any of these other players in that space? There’s really no one else, right?
Andrew Schmitt: Yes, talking to the guy at Huawei, I said, “Look, you guys really going to go forward with ROADMs when you only have two, two-and-a-half, suppliers?” I got to believe that Huawei is going to try to light up some more capacity in that area somehow, either…
Troy Jensen: Can Huawei internally manufacture ROADMs, or is this a more … Haven’t they tried for years and not been able to do it? How long of a window do you think these vendors before they can potentially be competitive with their own product?
Andrew Schmitt: I think they have a pretty significant window, but it’s just hard for me to believe that Huawei will … If in fact ROADMs are start having a major role in all three of the Chinese carriers, it’s hard for me to imagine that Huawei will continue to use just two suppliers. I did mention CoAdna to Huawei, but it didn’t really seem like they were seen as a significant supplier for this new effort. It was really Finisar and Lumentum. I think that was the order in which he said the names of the companies.
Supply Chain Security
Troy Jensen: Yeah, that’s fair. That’s consistent. I got two kind of questions got associated here. It has to do with the ZTE issue, so the question is … Let me read it here. “Does China need to secure or acquire optical components supply after the ZTE issue?” The other question that’s kind of similar is, “Do these Chinese guys end up displacing the likes of a Finisar or NeoPhotonics in favor of Chinese suppliers or Japanese component suppliers?”
Andrew Schmitt: The second question, I don’t know. The first question, that was the first thing that I asked the guy who ran the transport business. I said, “Look, did the ZTE thing spook you and is it going to adjust your whole supply chain strategy?” And he said, “Not at all,” because they’ve always made a major effort to try and internalize as much as they can, so that what’s happened to ZTE isn’t really affecting their component strategy because they’ve, in essence, already had a strategy to try to avoid this sort of problem as much as they can. He said they’ve been working on some stuff with silicon photonics to try to move some of their ICRs over to silicon photonics.
They have an effort underway right now to go build internally their own CFP2-ACO. They don’t have any products that are designed to use that yet, but they’re going through the exercise of putting that together and trying to make something like that work by next year just as a way to have that sort of technology in-house. I don’t think that what happened with ZTE is going to have Huawei act any different.
I think they’ve already been aware of these sorts of risks, and not just with optics, but I think the central company is Broadcom, and that was really one of the major reasons why they started high silicon is to provide a way to have an internal second source against other components that maybe sold first. That’s why I asked the question about the WSS because that is one area where there’s not a lot of supply for many different companies and is certainly is an item that they would be concerned about.
Troy Jensen: All right, I just want to spend a couple more minutes on China and then move to some other topics. To me, the key takeaway is demand is greater than you’d thought, and what is the visibility under this band and then the quality of the visibility?
China Wrap Up
Andrew Schmitt: The word I would use talking with some of the Japanese guys was it was a combination of disbelief and panic. Even on the 10-gig side, Huawei has been absorbing a lot of the capacity of their laser fabs to build some of their 10-gig lasers. Huawei is just inhaling components at this moment from what I can see, and everyone is just sort of waiting, like, how long can this really go on because the magnitude is so big.
Troy Jensen: Awesome. I think you said in your email to me, “mind-blowing demand,” and just your thoughts on the sustainability of it, though. I guess that’s what the audience wants to hear.
Andrew Schmitt: Just to borrow from a different area, which is fiber to the home, if you go back five years ago, Huawei, ZTE, they were purchasing tons of fiber-to-the-home equipment, and people were having this exact same conversation. It’s grown by a factor of four or five since then. Every time I’ve ever bet against Huawei buying or not being able to sustain the demand, I’ve been wrong. The numbers are enormous. If in fact there’s a 100-130,000 or so ports shipped into China this year, I have to go back and look at some of my projections, but just going from my gut, that’s got to be half of all the 100-gig in the entire world.
Troy Jensen: Yeah, okay. All right, so this year is a no-brainer, solid demand. We’ll see if it continues, right? There’s never any promise beyond a couple quarters.
Andrew Schmitt: The only thing I will say is it doesn’t seem like it’s just 100-gig, because the same thing is happening in 10-gig. They have started to absorb a lot of the 10-gig fab capacity for lasers, and it’s causing a shortage of 10-gig EMLs. One of the Japanese guys said they have a one-year backlog in place now from Huawei for 10-gig EMLs.
Troy Jensen: Anything else you want to highlight on China, or should we shift to takeaways from OFC?
Andrew Schmitt: Yeah, I think that’s about it.
Troy Jensen: Okay, perfect. I guess I want to hit first … Probably biggest announcement out of OFC was the Inphi-Microsoft announcement. Some of the vendors don’t believe you can go 80km. Can you just talk about what that announcement was and maybe just start off with what were your key takeaways from OFC?
Andrew Schmitt: Sure, I think a couple things hit OFC at a high level. Silicon photonics is starting to get real. What is happening is a lot of the original products that were built in silicon photonics essentially clone the architectures that were used with previous generation processes. The best example would be Cisco and the CPAK. They built something that was very comparable to the CFP2, similar performance. It didn’t really demonstrate the value of silicon photonics. It didn’t really bring anything new to the table. What’s happening now, we’re starting to see products built in silicon photonics that essentially could not be built using Indium Phosphide, and the Inphi product is one of those. There really was no other way to build that product except using silicon photonics. That’s one trend that I took away from OFC.
The other trend at OFC is there’s a lot of effort now by equipment vendors to move toward dis-aggregation. Fujitsu, Coriant, even Adva, are building these modular types of systems that allow vendors to pick and choose individual components, and then they’re selling the software to stitch it all together. A lot of that is being driven out of the Webco world and also some of the directives coming from AT&T, and that’s certainly why Fujitsu’s been very aggressive in this area.
Inphi & Microsoft Announcement
Troy Jensen: Can you touch on the Inphi-Microsoft announcement?
Andrew Schmitt: My feeling is that I think Microsoft threw their incumbent suppliers under the bus. I was actually on the panel when they made the announcement and did the presentation. If you listen to what Microsoft said, what they said … I’m not saying that this is actually accurate, but basically everything under 80 kilometers is going to go to this Inphi technology, but then if you talk to some of the incumbent suppliers who sell coherent, like Infinera, they said that that’s absolutely not true.
There’s definitely a gap in terms of what one company saying and what the supply chain is hearing. I think that Microsoft might have been a little bit too aggressive in promoting this technology, and it didn’t really reflect the reality that there’s still going to be a major role for coherent. Anything beyond 80 kilometers and applications sub-80 kilometers, there’s going to be cities where the fiber’s not going to be able to support the Inphi technology. There’s going to be customers who don’t have the level of expertise or they don’t have the scale to justify the work that goes into using that sort of tech.
I think Inphi’s got a great product. I think they’re going to be very successful with a small set of customers, but I don’t think that this is going to kill off coherent in the shorter reaches. What I see happening is you’re going to see, let’s say for instance, Microsoft in some cases they’ll use the QSFPs from Inphi running at 100-gig. In other situations, they’ll buy 80-kilometer 400-or-500-gig coherent interfaces and run them that way. The advantage that coherent has is as you start going to shorter and shorter reaches is you can get massive amounts of capacity. CFP2-ACO doesn’t run at 100-gig, it runs at 400 gig, so you’re getting more and more leverage on the same fixed cost the shorter the distance gets.
Troy Jensen: Okay, so Microsoft was pretty vocal clearly. Have you heard from any of the other Webcos on this issue versus traditional DCI?
Andrew Schmitt: No, I have not.
Troy Jensen: There’s a lot of debate about, or skepticism whether or not Inphi can really do 80km with this technology.
Andrew Schmitt: I haven’t done it myself, so I don’t know firsthand, but talking to folks like Adva, who are supplying the amplifiers because it requires a very specialized amplifier, they say it does 80 kilometers. Microsoft was pretty adamant that they can do 80 kilometers. Inphi was saying they can do hero experiments, 120, 130 kilometers.
The thing to remember is that if you buy an Infinera CX, you can plop one of these things down at either end of 80 or 120 kilometers of fiber, push the button, and then turn it up, and it’ll just work. With the Inphi solution, you’re going to have to put amplifiers in place mid-span. Then you’re going to be using dispersion compensation, and it’s like tuning a piano. Once the piano is tuned, then as long as you don’t touch anything, it should work. That’s Microsoft’s intent is they’re going to go implement these large upfront fixed costs on these links, and then they won’t touch it for several years.
Troy Jensen: Okay. How about Infinera? That was one we talked about in the past, you loved historically, kind of thought they lost their lead. They had a nice product announcement at OFC…
Andrew Schmitt: I don’t think I ever said they lost their lead. I think what I was trying to say on the last call was that the PIC technology was getting dated and that they needed to introduce something new, which is what they did at OFC.
Troy Jensen: Any thoughts on the new product?
Andrew Schmitt: It’s pretty fantastic. They were a little bit cagey in terms of when this is actually going to ship into products, but I’m hearing that it’s supposed to be early in the second half. If that’s the case, they should be very well positioned. If you look at what they’re doing with the PIC technology, they’re pairing it with some pretty significant advances in the electronics as well to do things like remove noise from adjacent channels, which I’m not aware of any other company doing that at this point. It’s pretty impressive.
Troy Jensen: I want to hit kind of three hot product segments. How about CFP2-ACOs first? I guess there’s some news at OFC. We saw NeoPhotonics getting the space, but if you can just give us your thoughts on the ACO market.
Andrew Schmitt: Yeah, I think Neo is really pursuing this more as a target of opportunity because there’s so much demand, and the supply is kind of challenged. They looked at a way to build one of these Class 3 ACOs by cobbling together conventional off-the-shelf components that they have which are very advanced. They’re not using the monolithic Indium Phosphide approach that all of the other guys are using. Acacia also announced an ACO module using silicon photonics, and it’ll be interesting to see what the performance of that looks like.
The story hasn’t changed. There’s still a lot of demand for that type of format, particularly among the DCI boxes. If you look at all the DCI boxes, the last three or four of them that were announced, they were all based on CFP2-ACO. It really appears to be the technology of choice for having an incremental capacity method and high-density line cards.
Troy Jensen: Oclaro is still the clear leader in that space and just any thoughts on the guys trying to get into the market now?
Andrew Schmitt: Yeah, I think, from what I understand, Lumentum is still taking the long road. I don’t really see them playing a role this year. They’re playing the long game, which if you were to go and build a product like the Oclaro product but bring it to the market by the end of the year, you’d be a little bit late. Lumentum is probably doing the right thing and pursuing a product that is going to be market-leading when it comes out.
I’m not sure what to add on the ACO side. I haven’t heard any issues other than some of the suppliers are saying they can’t get ACOs. They can’t seem to buy them, but from what I understand, it’s because a lot of that supply has already been locked up and locked down.
Troy Jensen: I asked Finisar and Oclaro at my dinner … You were there … if they expect to be capacity constrained the entire year on ACOs. Both of them didn’t hesitate and said yes, but in near-term demand it’s all the DCI stuff. Then once we get the metro upgrades, specifically at Verizon, that kind of sustains the demand for those guys, so … All right, switching to … We hit ROADMs a little bit. I guess any updates on the ROADM opportunity at Verizon?
Andrew Schmitt: No. I’m not aware of anything changing at Verizon. They’re planning to deploy the large CD ROADMs in basically every node. They’ve taken a tack that’s different than what AT&T is talking about, which is disaggregating the network. Verizon is taking a much more traditional approach as one system from Cisco or from Ciena that can do everything from route wavelength to route packets.
Troy Jensen: Okay. How about QSFP28 update, Web 2.0 demand, anything you picked up interesting at OFC?
Andrew Schmitt: I have to think about that. In terms of QSFP28, the one trend that became clear to me at OFC is that it seems like everybody has it. Everybody is claiming they can ship QSFP28, but the capacity behind those claims is rather thin. There’s only a few players out there that have the depth to ship in quantity. There’s an illusion of widespread abundance when it’s a rather thin level of supply from many of these vendors. They haven’t made the investments in capacity to ramp to large quantities. That was the impression that I got, and I don’t …
Troy Jensen: You need to rank the top … Go ahead, finish. Sorry.
Andrew Schmitt: I’m not sure I could rank. The one thing I do hear consistently is that Lumentum is really the big guy in terms of the supply in those. Then occasionally I hear … I haven’t been able to verify with Oclaro, but I’ve heard that they have also been investing heavily in being able to deliver … not just announce product, but be able to ship capacity for QSFP28. The key question I would ask during this season of earnings calls is not how many QSFPs did you ship, but next quarter how much capacity are you going to have? How many could you ship?
Troy Jensen: They don’t want to answer it. I don’t know if it’s competitive reasons or acting aggressively. One guy said, “I’d have to check my notes. I don’t want to say anything on the call that might be misquoted.”
Troy Jensen: All right, well let me know if any come in, and I’m going to start going through the questions in my inbox here. Some of these you may or may not be able to answer, Andrew, but I’ll just fire them off here.
Macro Spending Trends
Do you have a view on Juniper’s miss and implications for guys like Ciena based on slower North American service provider spending?
Andrew Schmitt: I have no clue on Juniper. I can’t add anything. I don’t have anything intelligent to say there.
Troy Jensen: Okay, how about just on North American service provider spending for optics?
Andrew Schmitt: No, I don’t have a good picture of what macro spending looks like in North America outside of … The only thing is all of the real alpha… The alpha that’s been in the spending in North America for the last couple years has been coming from outside the traditional service providers. I don’t think there has been a lot of fluctuations in spend from the traditional guys. It’s been kind of monotonic. I haven’t heard anything that would make me believe that that’s changed.
Troy Jensen: I got two questions in from two different guys on the same question. You’ve kind of already answered it to me, but I’ll just let them hear this. Any impact on the Verizon strike and the metro 100G roll out?
Andrew Schmitt: That was sort of news to me. I don’t know. Obviously, the number one variable is how long this thing goes on. I think people are probably concerned specifically about the new metro roll outs that Verizon is planning. I actually sent an email out to somebody at Verizon about this, but they’re in the field working, so I haven’t heard back, but I don’t know. The material amount of work for the new metro projects is really in the second half of the year, so if this resolves itself in the next month or two, I don’t think it should affect that.
Troy Jensen: Okay, that’s fair, and just for everybody in the audience, I did hear from one of the management teams that they had heard that the Verizon projects have all been put on hold, so it depends on the duration of how long the strike goes. I think Verizon said something similar on a call yesterday that they don’t see any real major impacts on the strike, but it depends on the duration. We’ve just got to watch that closely. Views on Ciena? That was a short question.
Andrew Schmitt: I think they’ve always been … They’re the bellwether for the industry because they sell into so many different corners of it. Whereas someone like Infinera is a good barometer of spending from the DCI guys, as well as Adva, I think Ciena is really the best macro-level company. Nothing’s really changed in terms of my outlook for Ciena. I think they’ve got the broadest set of products and can do the most things for different companies. I don’t really have anything constructive to say, I guess.
Webco 40G to 100G
Troy Jensen: Okay, I’m just kind of firing these off, too. They’re jumping around here, but does Andrew have thoughts on the webco guys spending on guys like Lumentum, Finisar, and Oclaro?
Andrew Schmitt: The big story there is the rotation from 40G to 100G, and I don’t know how that’s really even factored in yet because a big part of all the 40-gig optics for the data centers that are shipped go into these web scale guys. As they rotate over to 100G, I don’t expect that demand for the 40G is going to continue to remain the same. This transition is going to be a big deal in the supply chain, so if somebody is heavily dependent on 40-gig business and then they don’t ride the curve on the 100-gig side, it’s going to be a negative event for them. Likewise, if somebody doesn’t have exposure with 40-gig and they’re well-positioned to ride the transition to 100-gig, they’re going to benefit a lot. Just as 100-gig coherent started altering the landscape on the equipment side, I think the transition to 100-gig in the data center is going to adjust the market share landscape among the suppliers of the components. That’s the big story there.
Troy Jensen: Yeah, I hear you. How quickly does 40 fall off is going to be the risk for them. Here’s a question to that point. What is at risk in the follow-up, Adva gave great revenue guidance but terrible margin guide, and then AAOI just had a total blow-up. What do you think is at risk here with these trends?
Andrew Schmitt: I think the overall risk that overhangs the industry is that at some point these DCI companies, as they’re growing capex so rapidly, at some point they’re going to over build. They’re driving so fast they can’t see out of their windshield. I don’t know when that event is, but it will happen, and it’s going to be a three-to-six-month period of time where Microsoft says, “Hey, you know what? We put enough capacity in. Let’s go fill it. Let’s go sweat the assets a little bit before we continue to spend.” That’s what I am constantly watching out for is that there will be some sort of a pause like that.
Troy Jensen: They always end bad, right? Just figuring out the timing’s the key here.
Andrew Schmitt: Yeah, but I don’t think the issue with AAOI is anything to do with that. From the little bit that I’ve heard, it’s isolated specifically to that company.
Troy Jensen: Okay, good. Here’s a good one. It says, “Following on the modular strategy you mentioned from Fujitsu and et cetera, are there competitive implications for leading system guys like Infinera and Ciena that’s an attractive approach for the big web guys?”
Andrew Schmitt: That’s a great question. Let’s just assume that this concept catches on, and that’s a pretty big assumption because right now the consensus is that this is not going to be a major trend when you talk to the supply chain. Let’s just say that they’re wrong and this ends up being a big deal. What it would do is put more power in the hands of the folks who own the fundamental technology.
Just like the commoditization of the PC put the strength in the hands of Microsoft and Intel and took it away from IBM, the commoditization of an optical network would put the hands in the Lumentums, the Finisars, and the component guys, as well as someone like Infinera. In essence they have their own component supply chain inside that company, and it’s very competitive. It puts more emphasis on the expertise that you have in the components and then it adds the expertise to the other side, which is the systems integration and the software. Adva’s been able to do a good job for a long time helping a lot of these enterprise guys do system integration and make things work.
My guess is they’re going to be at the front lines with some of these people who are trying to make this Inphi solution work because it is going to be difficult to implement. Ciena, that’s why they’re so focused on software, as is Fujitsu and Coriant. They’re making a major effort that if they do get disaggregated at least they want to have the controllers that can manipulate the other components that are coming in. Someone may buy ROADMs directly from Lumentum, but Fujitsu or Coriant is still selling the software and the control plane to manipulate them. That’s how they’re really coming at this problem.
Troy Jensen: Okay, a few more here I want to go through.
Andrew Schmitt: The one thing to keep in mind is this isn’t like the router business where they had 65% margins, and as it gets disaggregated it could be painful. The margins are already kind of compressed in this business, so if somebody’s disaggregating their network so that they can try and save money, it’s not going to work that well. People who are disaggregating their networks aren’t doing it because they’re trying to save money. They’re doing it because they want to have more control and more flexibility to implement exactly what they want. I don’t see this as being a really major risk for a lot of the incumbent equipment providers.
Wrap: CFP2-ACO, Neophotonics, Other
Troy Jensen: All right, so a couple more. I got a couple on CFP2-ACOs. One is, is Oclaro’s production ramp in China on track? Then, can you size the CFP2-ACO market this year and next?
Andrew Schmitt: Yeah, I think the size of the ACO market this year is going to depend on how many the few companies that are supplying can build. The number that I started out with was around 35, 40K, but the component guys kind of walked me back towards 20, 30K this year. I talked to a component supplier who doesn’t have a hand in the ACO game and asked them, what do you think Oclaro is going to ship out of that number? He said probably 60 to 70% of it, which I thought was awful high.
For next year, the ACO business should at least double or more than that. At that point the growth of the market is going to depend on how fast the pricing can come down. That’s really the challenge for ACO is, if you look at somebody like Huawei, they can buy discrete lasers and ICRs from NeoPhotonics or modulators from other vendors, solder it all down in a board, and their total building material cost might be, let’s say, $3,500 or $4,000. Then they look at buying an ACO, a pluggable which does exactly the same thing, that costs $6,000. For them there’s no reason to move to that technology, but as the ACOs get cheaper because of the photonic integration and Indium Phosphide, it’ll start providing a lower cost alternative to people building these things discretely on their own board.
The total market for ACO is really a function of how quickly the prices can ramp down because it will cannibalize the discrete market.
Troy Jensen: Understood. Can you quantify the size of the ROADM opportunity?
Andrew Schmitt: In China?
Troy Jensen: In aggregate or whatever you can would be helpful.
Andrew Schmitt: Unfortunately, I haven’t got my hands dirty on ROADM numbers in a while, but it seems as though we’re definitely on the edge of another cycle excluding China, just as there’s a refresh in the metro for 100G, there’s a push towards putting ROADMs in place as a result of rolling 100G into the metro. The China thing just sits on top of that, but I don’t think the China demand really arrives until 2016, so overall I think you’ll probably see more ROADM consumption as the year goes on, but the really big bump probably arrives around the end of the year.
Troy Jensen: Here’s one. It looks like he’s a little bit more semiconductor focused, but can you comment on the outlook for TIAs, drivers, and CDRs and the relative position for guys like Inphi, MACOM, Avago, and do you ever see GigOptix now?
Andrew Schmitt: Yeah, I know GigOptix, and they’ve been focusing on this area. I don’t have a good idea of the puts and takes in that market, who’s leading. No, I guess I don’t really have a good answer to that question.
Troy Jensen: All right, that’s fair. Let me see here. There was one you kind of answered, but we’ll just hammer it home again. Who are the biggest winners at Huawei given the 120,000 ports this year?
Andrew Schmitt: Just off the top of my head, with Neo selling so much into Huawei, it’s hard for you to imagine that they’re not somehow going to benefit from this. They would have to try really hard to not get hit by this wave. Then I think the … What’s that?
Troy Jensen: Since I’ve been talking to you, you added about $50 million of market cap to Neophotonics stock.
Andrew Schmitt: Okay, I think the second thing would be the folks that sell on the client side because the attach rate for 100-gig clients is increasing in China, so the folks that sell the CFP2s, they’re going to see more demand coming out of China there. That’s distributed among several different suppliers. I don’t know who really has the edge there at Huawei.
Troy Jensen: All right, here’s one more. I think this will be the last one because we’re kind of at the end here. Is Andrew seeing any slowdown from ZTE demand due to the government restrictions?
Andrew Schmitt: ZTE really, like I said, was … What was the number? Like 14% of …
Troy Jensen: Yeah, 15 from Oclaro.
Andrew Schmitt: Yeah. No, I mean in terms of the amount of business they won in China. ZTE has really been edged out hard by Huawei and Fiberhome, so they had a little bit of market share loss there. They don’t seem to be thriving — let’s just put it that way — in the Chinese market at this point. I’m sure that the recent changes are a distraction, but just looking at the tender that was issued in Q4, ZTE only got a small portion of that, so for things to change they’d have to have a much more dramatic type of win at the regional level, and I don’t know if that’s going to happen or not. Huawei is just, in general, a lot more aggressive in terms of building unique technology for the Chinese market like this in-house CFP. That’s what the carriers want.
Troy Jensen: I understand. Well, Andrew, I’ll wrap the end of the call here. I really appreciate your time. Your insight on China is incredibly helpful and incredibly timely right now given all these guys are going to be reporting either next week or the following, so we really appreciate you doing this for us.