Andrew Schmitt was the featured speaker on an investor call hosted by Troy Jensen of Piper Jaffray on February 21st. Forty-four participants listened in and posed their questions as Andrew gave his perspective on the latest events in the optical component and equipment market.
Topics during the call included:
- Anticipated impact of China’s MIIT plan on existing component suppliers
- The divergence in Chinese equipment revenue and supply chain
- The market outlook for ROADMs in 2018
- The 400G+ Coherent DSP and component outlook
- The evolution of the CFP2-ACO and DCO market
- Analysis of equipment companies outsourcing DSPs
- Update on Inphi’s ColorZ, modulator driver, and PAM-4 business
- Infinera’s market opportunity with the CX2 and ICE4 based products
- North American CapEx in 2018 and optical equipment forecast
- Ciena’s 400G market monopoly in 2018
Active Insight clients can access a summary of the key takeaways, a written transcript, an audio replay, and participate in Q&A.
Key Takeaways
- There is no significant near-term impact from China’s MIIT roadmap. Rather, the long-term impact is much more significant.
- The divergence in revenue from Chinese equipment companies and what component suppliers are shipping is a paradox that is not fully understood and requires further investigation.
- There is no change to the outlook for ROADMs in China. Operators in the region are waiting for next-generation technology. Verizon is not ramping hard in 2018, and it is unclear why component companies are communicating this.
- Component revenue for 400G coherent shipments will not be material until 2019, with the exception of suppliers to Ciena which is roughly a $20M opportunity in 2018.
- CFP2-ACO technology is being re-purposed for use in Huawei DCOs.
- Infinera is probably the next company to outsource DSP development.
- Inphi’s ColorZ direct detect technology will be eventually outclassed by coherent. The company still owns the commanding heights of the driver and PAM-4 market but many other companies are trying to enter these markets.
- The success of Infinera’s new GEN4 PIC based technology is difficult to predict. Cignal AI has not yet spoken with customers who have evaluated the technology. Infinera needs an edge in order to take share in what is already a tough market.
- Ciena finds itself in a great position during 2018 with the Wavelogic AI but Cisco and Infinera can use incumbency in some cloud and colo accounts to their advantage. Decisions on this technology are not only about who has the fastest & cheapest wavelengths – software integration plays a role too.
Operator: Now it’s my pleasure to turn the conference over to Troy Jensen, Senior Research Analyst at Piper Jaffray. Please go ahead.
Troy Jensen: All right. Thanks everyone for dialing in and thank you Andrew for taking the time to do this call with us.
Andrew Schmitt: Sure thing.
Troy Jensen: For those dialing in, we will try to open it up for Q&A, probably midway and towards the end of the call. But also, I know a lot of you guys just like to email me and I can ask the question. Whatever you prefer. Andrew, just wanted to kick off first on China. If you go back about a month ago, you were kind enough to forward me your report on the new five-year plan from China. I feel like I’ve asked every executive this quarter, all of them claim that it won’t really impact them but, to me, the intentions of what China’s talking about seem like it could be a big headwind for the suppliers. Could you touch on your thoughts on the five-year plan and what China’s trying to do?
China MIIT Component Plan Impact
Andrew Schmitt: I assume most of the folks on the call are familiar with what happened, so I won’t go into detail on the plan other than, a government agency issued a long document that identified areas of technical and commercial weakness within China, specifically surrounding optics. They talked about things like LCD displays and some other things like optical fiber, you know the business that Corning is in. Despite the fact that something like LCD displays is a huge business when compared to optical components, the majority of the document and all of the detail was focused specifically on optical components as an area that needs much greater focus and improvement within the country in order to be strong. I think the specific words were along the lines of, “You cannot build a house on a foundation that is not yours.” They identified, essentially, every single optical component that you would need in the supply chain to make these systems. That included telecom components, datacom components, they specifically mentioned DSPs. They specifically mentioned PAM-4 chips. So, everything.
I think part of the issue that’s causing some doubt on the part of the component companies is, when you see something like this, it does seem totally unrealistic that they’re going to actually come in and just do everything. The reality is probably closer to, there’s going to be some areas where they’re successful and there’s going to be some areas that they’re not, and all of this is going to take some time. So, in the near term, there’s really not any commercial impact.
But, in the long term, there’s, at least in my mind, a very high degree of uncertainty of what’s going to happen because this is the same game plan that they ran when they got into the optical equipment business and the telecom equipment business 20 years ago. I recall going over to visit ZTE in 1999 and they worked in buildings that look like they belonged in Beirut after the war. Fast forward to today and they’ve completely transformed the telecom equipment business, as everyone knows. I’m approaching this with a very high degree of concern because I’ve seen them do this in other businesses and there’s not a technical reason why they can’t accomplish the same thing in the components business.
Troy Jensen: It seems like a lot of the pushback I’ve been getting is that we hear that China’s always trying to go in-house for a lot of their production, and this is kind of the same game plan. But, you’re pretty adamant that this is a change, this is an acceleration on their intention to buy more products domestic.
Andrew Schmitt: Yeah. Huawei has already been doing this themselves because they recognized that they were such a large portion of the market. They build their own DSPs with help from western suppliers. They’ve been investing heavily in both Silicon photonics, as well as indium phosphide-based components but they haven’t put anything into production. Is all of this very, very hard? Absolutely. But, that doesn’t necessarily mean that they’re not going to be successful in some areas.
I would expect to see Huawei try to spin out some of these components, perhaps. Or, try to shore up HiSilicon as a more independent entity that can sell to other equipment companies in China.
Troy Jensen: The other comment we heard, even just this morning, is that they’re mainly focused more on just modules but they continue to buy the discrete components from the North American guys.
Andrew Schmitt: That’s true. The first place to do this is getting the low hanging fruit, right? One of the things, though, they talk about in the document is that … China’s already pretty strong in 10G shipments. There’s a lot of people building 10G components in China. Specifically, they’re purchasing the lasers and assembling these things. They call out in the document that just being an assembler of these modules and buying lasers from other people is not a solution. I would expect that that’s how they’re going to start and, for that matter, Innolight has been purchasing lasers from Oclaro as part of their direct shipments to a DCI customer. I would expect that there’s going to be several component companies in Wuhan and maybe one of them ends up being successful. When that happens, I don’t know but, it would appear that this is now a strategic effort within the country to enter this market.
Troy Jensen: Okay. I think you’d agree though, this is beyond just components. This is coherent DSPs. This is TIAs. This is everything involved in making the transceivers and modules?
Andrew Schmitt: Yeah. Everything. The only thing that was surprising to me is that they also didn’t identify things like ethernet switch chips. It’s probably beyond the scope of the document. Just to share a story, when I was in China a year ago, I met with Huawei. It was right after the ZTE penalty that was announced by the Department of Commerce. I can’t remember the number, but it was enormous in terms of the fine that was levied on ZTE. It was the first time I’ve ever heard anyone at Huawei have some pity or concern over ZTE. That was a real wake up call to the companies over there that they need to do something. Huawei had already started to insource a lot of their own manufacturing, but I think, after what happened with ZTE, they really recognized how vulnerable they were.
Troy Jensen: If you think about what’s most likely for China to in-house quickest and, which ones are going to be difficult, can they really do coherent DSPs? Can they do ROADMs? Talk about what do you think is most at risk right away?
Andrew Schmitt: I would say its – you highlighted it – the module assembly. They can build modules and purchase lasers from multiple suppliers. I think on the client side, that’s the lowest hanging fruit. That’s what I would identify as the area of most risk. It also happens to be the worst market, right now, in the optical area in terms of profit. So, it’s probably the least vulnerable. Going up from there, I would say, Huawei is already focused on building their own DSPs. I would expect to see them continue to do that and take that to higher speeds. Huawei having their own DSP and considering they’re two-thirds of the Chinese market, they’ve in effect, insourced a lot of that capacity. What I don’t know is, whether or not Huawei would consider spinning that out as an independent entity that could then sell to other Chinese manufacturers or there could be a second source within China to do DSP work. I haven’t heard of anything along those lines but, that was one area that was identified.
Troy Jensen: What are the other areas?
Andrew Schmitt: The tough ones are going to be things like tunable lasers. That’s going to take some time to put something like that together. There are not too many people out there who can build a high quality, narrow linewidth tunable sources today.
ROADMs are another one because it isn’t just having the WSS. Building the WSS by itself was already challenging. But, in order to really be a successful WSS supplier, you have to have a greater appreciation of the system level effects and what’s going on. So, that’s another difficult area for the Chinese to get into. I know that they have already been in discussions with a Japanese vendor of LCoS sources, several years ago. There are people who have LCoS technology that’s not currently being used in a WSS, who are trying to find a way to put it to work. The building blocks are out there but, the expertise required to build a system level ROADM is extremely challenging.
Chinese Equipment vs. Component Revenue
Troy Jensen: Okay, fair. Andrew, I want to just challenge you a little bit here. If you go back, I don’t know if it was Q2 or Q3 tweeted out that China port shipment was really strong, then I just read your Q4 Market Survey, and China is really strong. I guess what I’m trying to figure out is, has China already started to in-house some of this production and that’s why the equipment companies aren’t seeing a rebound?
Andrew Schmitt: I don’t think that’s it. If you look at even where Huawei is going in their next generation DCO this year, they’re still sourcing those products from the outside. The numbers that came out of Huawei and Q4 so far, it was a record quarter in terms of revenue. There’s been a big divergence in the revenue that these companies are reporting and the components that are essentially being shipped into by the western vendors. I don’t think it’s an insourcing issue. I haven’t been able to identify an alternative source. It’s a matter of trying to figure out and forensically understand why there’s such a gap. Is it just a question of the inventory that was shipped in was just so huge that it’s taking time to work through? Or, is it reductions in price? I haven’t really come to a conclusion of why there’s such a big gap.
Troy Jensen: Okay. All right, that’s fair. I got one inbound-
Andrew Schmitt: I will add one more thing. If you look at Q4, it was very much a Huawei centered quarter in China. If you look at ZTE, you haven’t seen the same hockey stick in revenue. ZTE, you can correlate pretty strongly with Acacia. It’s not like ZTE doubled their revenue or something like that, and Acacia was flat, which would be very strange. This is very specific to Huawei and their performance in Q4.
Troy Jensen: All right. I got one inbound question on this topic. It says, “Does Chinese policy make them more likely to buy US optical component companies? To reach these goals, could they buy NeoPhotonics straight and get a lot of capability, quickly?”
Andrew Schmitt: I don’t know. I think they don’t want to just own the company, they want to own the expertise. I don’t know to what degree that expertise, as a company like NeoPhotonics is here in the US versus China. I don’t know the ins and outs of these companies and where the expertise lies. If 100% of the expertise was in San Jose, I would expect the answer to that to be no.
ROADM Demand in 2018
Troy Jensen: That makes sense. Let’s shift gear to ROADM demand. I guess, starting in China, what are you hearing and ultimately, North America?
Andrew Schmitt: I haven’t heard any change out of China. I know that Lumentum was talking about upticks in demand in China. That might have been for export because, as we talked about Huawei had a good quarter. Domestically, I haven’t heard of any change of plans there from China Telecom and China Mobile. I think we’ve talked about this on previous calls that, in a sense, they’re waiting for the latest generation ROADM technology that Lumentum is working on in order to do widespread deployments. In the meantime, they’re just limiting what they put into their network. That’s my understanding of the situation. I haven’t heard anything else.
Troy Jensen: Can you talk about the ROADM that Lumentum has? I know it’s the MxN, but maybe just explain?
Andrew Schmitt: Lumentum hasn’t, I don’t think, announced anything about this but, Huawei has been pretty open about what they’re doing and Huawei’s customers have been talking about it. Lumentum is building a new WSS module that basically solves all the problems that exist in existing CDC based architectures – Colorless, directionless, contentionless. If you look at the design of a CDC ROADM, it’s complicated. There’s a lot of parts. There are certain network architecture limitations when yous tart using lots of super-channels and things like that. It’s sort of a substandard solution with some compromise in terms of cost and usability.
What Lumentum is working on, this MxN WSS, it’s challenging, but it addresses a lot of these complexities and could drive the cost of a CDC based ROADM architecture to be on par with more conventional ROADM architectures. If they can get this thing to work, it’s a real game-changer in terms of how people are going to build ROADM based networks because there’s really, at least at this point, nothing in the way of compromises that you have to accept, if they can solve the technical problems. Talking with Huawei, they’re planning to introduce this product in the second half of this year. As I understood it, the device is already in 10,000-hour testing. I don’t know if they’re going to have to do a rev.
Strategically, Lumentum seems to have a very strong relationship with Huawei in supplying these next-generation ROADMs. I haven’t seen or heard anything from Finisar.
Troy Jensen: Just to dumb that down, the MxN, it’s absorbing the multicast switch functionality in the ROADM, is that correct?
Andrew Schmitt: That’s right. Basically, the MCS goes away. Once this product exists, there’s really no need for an MCS anymore. A normal WSS has lots of inputs and one output. This has lots of inputs and several outputs. It basically takes over the function of the MCS in a superior way.
Troy Jensen: I think Lumentum and Neo are the two suppliers of multicast switching, right? So, that explains why Lumentum can get –
Andrew Schmitt: NTT Electronics is also a supplier.
Troy Jensen: Oh, and NTT. All right. So, if Neo was counting on a lot of multicast switch revenue, this could be an issue.
Andrew Schmitt: Yeah. They may not agree, but my opinion is once this product exists, in volume, at a fair price, the MCS business goes away.
400G+ Market Outlook
Troy Jensen: How about switching gears here, to the 400/600G market. Love to get your thoughts on the discrete components, who are best positioned? Modulators, receivers, transmitters.
Andrew Schmitt: Yeah. This is a very clear situation. In 2018, Ciena’s the only company that’s shipping 400G product in production. Acacia’s going to be prototyping things in the second half of this year. I think you may see things from other vendors. I’m not sure, exactly, where Inphi is but, they’re probably in that kind of time frame. In terms of real production shipments, Ciena’s the only game in town and that product, so far, is doing really well. In terms of Ciena’s commentary about 2018, it’s hard for me to reconcile that with the leadership position that they’ve put themselves in with this product. I don’t understand why they’re not going to have a better year than what they’ve said, looking at their annual projections because they’re in a very strong position with the WaveLogic Ai in 2018.
Now, going into the second half of this year, you’ll start to see products from Acacia. Acacia is working with several OEM equipment folks with that design. As I understand it, they’re playing to use their silicon photonics. But, I know of at least one customer who has asked them to de-risk this approach and use Oclaro technology so that, in case there’s a problem with some of the optical front end in silicon photonics there’s a backup plan to use proven types of technology. There’s always a risk with the DSP but it seems as though there’s some double coverage on the optics that sit in front of it. That DSP is a little bit later than what people have expected. I haven’t gotten a refresh on where it sits today but, I think it’s safe to say, given Acacia’s history of solid execution that, that’s going to be available by the end of the year.
Troy Jensen: How about some of the discrete component guys? Lasers, ICRs.
Andrew Schmitt: My understanding is that lithium niobate is being used for a lot of these first-generation systems, particularly if they want to try to get to some of the higher baud rates. I have to think about exactly who. Oclaro is a major lithium niobate supplier. Fujitsu is also one. But, I don’t have a good sense of who’s best positioned, relative to each other, for this next generation of products. I do know that Oclaro’s been successful in several designs, but I don’t know that other people have not.
Does that make sense?
Troy Jensen: Yeah, no, makes tons of sense. How about your sense of timing? Is it design wins in the second half with a bigger ramp in 2019?
Andrew Schmitt: I think the design wins have already, for the large part, happened. There are people who are positioning to use the Acacia DSP or they’re positioning to use Inphi or NEL DSPs. I wouldn’t expect any volume until the end of this year. It’s really going to be prototyping in Q3 and Q4. Like I said, Ciena’s the only one that’s shipping in volume this year. I would expect them to ship probably around 10,000 units in 2018.
Troy Jensen: Got you.
Andrew Schmitt: You assume 10,000 attach a 2K ASP for the front end, sitting in front of the DSP. It’s a $20 million piece of business.
CFP2-ACO and DCO Market Evolution
Troy Jensen: Shifting gears, just an update on the ACO market?
Andrew Schmitt: The ACO market got very interesting. It’s really turned into an annuity now for Oclaro supplying Ciena and Cisco. Cisco is still running very hard with their NCS1000, it’s continuing to be successful. That’s probably a major source of revenue for Oclaro as the NCS1000 continues to do well. Cisco’s confident that it can hold its own against this Ciena solution at 400G. They think they’ve got a good solution. They don’t see the bottom dropping out on that business the way the same thing happened to a few other vendors during these technology shifts. I think it’s also challenging because Ciena’s sole sourcing this 400G technology and that’s always a bit of a headwind to new technology adoption.
I don’t see the bottom dropping out on the ACO business. There’s nobody coming into the business at this point but, there’s also not a lot of upside growth. All of the ACO sockets that are out there have been designed and as part of this transition to 400G, ACO plays no role at all.
The interesting thing that has happened now is, Oclaro specifically, if you look at the technology that’s inside that ACO, as people go to these 200G DCO designs, or even 400G DCO designs, the guts of that ACO can be repurposed and used inside of a DCO. In fact, I’m pretty confident that is what Huawei is doing with their next generation DCO design that they’re doing internally. They’re sourcing the guts from an ACO, using that with their own DSP and packaging it in their own DCO platform. During Oclaro’s call, they talked about the ACO business ramping into 2018, which really surprised me. My understanding is that when they talk about their ACO business, it’s not just the module business, it’s also the parts that go into the ACOs. This business with Huawei, if Oclaro has secured it, could be significant simply because those guys are shipping tens of thousands of units a quarter.
Troy Jensen: Greg was with us this morning and he did say that the ACO bucket that he’s going to be reporting is going to be components that are going into other ACOs, DCOS. Makes sense.
Has Finisar or Acacia got any traction with their ACOs?
Andrew Schmitt: I don’t know about Finisar. Acacia, I think they have some traction in low power designs.
Troy Jensen: That’s fair. How about shifting gears to the DCO market, just quick thoughts? You already stated that you think DCO is going to dominate for the 400G space.
Andrew Schmitt: Acacia is just killing it in that business. They have a really good 200G CFP2 DCO. My understanding is that some of the other component companies were planning to build CFP2 DCOs using an external DSP supplier. Then, they looked at the performance of that DSP and they realized if they built a module around it, it wouldn’t come close to competing with what Acacia had. So, a lot of those programs were delayed. You hear Oclaro talking about DCOs in early 2019, but my understanding is Oclaro and others were planning to ship things this year, it’s just that they basically couldn’t go toe to toe with the Acacia product.
I think Acacia has a very strong product now. There’s a significant market for CFP2 DCOs, particularly in applications like switching and routing. Someone like Arista, the CFP2 DCO, that’s a perfect product for them. Juniper too. Portions of Cisco. That’s where the market really exists for that sort of product.
Troy Jensen: But, it’s your belief that all the guys that are coming out with the new DCOs, whether they’re partnering with Acacia’s DSP, or Ciena’s, or NTT’s, or Inphi, that they’re going to be doing CFPs and not CFP2s?
Andrew Schmitt: At this point, anyone who’s building a DCO, if you’re doing the R&D now, it would really need to be CFP2. There’s no market for a CFP DCO outside of what Huawei is doing – they’re going their own way. If you can’t build a CFP2 DCO, there’s really no point in building a DCO. I would expect people are working with Inphi, they’re working with NTT Electronics. I don’t think that Ciena has a product anytime soon that would enable a CFP2 DCO. I’m not aware of Acacia selling their DSP to component companies to create another CFP2 DCO. For example, I don’t think Oclaro can buy a DSP from Acacia and build a CFP2 DCO around that. Until we get effective DSPs from Inphi and NTT Electronics, there’s just not going to be a CFP2 DCO market.
Troy Jensen: A year ago Ciena announced their kind of open platform. They partnered with Lumentum and Oclaro and Neo to make DCOs with their DSP. I think you just said that they can’t do CFP2s?
Andrew Schmitt: No. That chip won’t do a CFP2 DCO. It just won’t work.
Troy Jensen: But, wouldn’t those three then, by definition, be making a CFP DCO?
Andrew Schmitt: They’re actually making MSAs and they’re not pluggable at all. They’re building modules that have to be pre-installed in the equipment. My understanding is that those MSAs are now under evaluation in China, that there may be some limited uptake for those MSAs, but this is not going to be a material success for Ciena. I think this is really about priming the pump for whatever Ciena is doing next. Whatever chip that may be, having some partners that they’re already working with would allow them to get a much wider penetration of their DSP technology into the market.
Coherent DSP Landscape
Troy Jensen: Shifting gears again now, you hit on some of this but, the coherent DSP suppliers. Just love to hear your thoughts on where the competition is in relationship to Acacia.
Andrew Schmitt: I do know that Inphi wasn’t where people wanted them to be with the previous generation. I think Inphi has talked about some of the changes that they’ve made in their DSP development. You’re going to see a lot of these companies going to 200G and producing 200G chips but, really 400G is the next battleground.
There are two types of 400G. There’s the 400GZR, which is the 80-kilometer application, very low power, specialized for shorter reaches. A lot of people talk about DCI applications, but it’s not just DCI, it’s basically every single link that’s under 80 kilometers, which is probably half the connections in the world. They’re still running a 10G today and there’s no upgrade path to coherent. This 400GZR platform is that upgrade path. I think you’re going to see 100G variants of it and the volume is going to be large outside of the DCI applications that everyone focuses on.
Then, the second is 400G, 600G, long reach and mid-reach, which is what Acacia’s announced with the Pico chip and what I expect you’ll see from Inphi and from NTT Electronics, although they haven’t announced anything.
Troy Jensen: All right. How about, have you heard any chatter about some of these proprietary DSP companies looking at merchant?
Andrew Schmitt: I haven’t heard any chatter. Cisco turned. They’re now an Acacia house. If you look at Coriant, they use Acacia. I think the ones that are kind of on the bubble would be Nokia, Infinera. If I had to name one that was likely to turn, it would be Infinera because they’re a very pragmatic company, they’re trying to accelerate the speed at which they release new products. It’s going to be difficult to release a new PIC every two years if they’re trying to do their own DSP. It wouldn’t surprise me to see them move to an external DSP supplier for the next generation ICE5. We’ll see. I could be wrong about that. But, to move those DSPs to a two-year cadence is a very big commitment.
Troy Jensen: What about Cisco? Cisco’s already using Acacia, right, for the Verizon metro?
Andrew Schmitt: Yeah. Pretty much. I think they still have a skeleton crew working on DSPs and I wouldn’t be surprised to see some targeted products come out of that. Maybe a 100G … I don’t know, I’m speculating. But, they’ve clearly partnered with Acacia going forward.
Troy Jensen: I got an email question here, it says, “Is Huawei using Oclaro’s components to build CFP DCOs or CFP2 DCOs in-house?”
Andrew Schmitt: CFP DCOs with the potential for CFP2 DCOs. Huawei is riding this CFP architecture as long as they can. Specifically, what they’ve said is they don’t want to move to CFP2 DCO until they can get equivalent performance to the larger CFP DCO. So, it’s going to be a function of when they can get a 200G DSP that matches the performance of their existing 100G DSP. Once that happens, then the whole company goes to CFP2 DCO. But, I would expect that on the optics side they would probably try to stay with the same design, going from one to the next because that Oclaro design can support any of those applications.
Troy Jensen: Part two of this question was, “Can ZTE make CFP2 DCOs in house?” I assume the answer’s no if Huawei’s not doing CFP2s yet.
Andrew Schmitt: I think the challenge there for ZTE is, if they’re going to build CFP2 DCOs, they need to have access to the standalone DSP from Acacia. I don’t know if Acacia’s willing to sell them that or not because Acacia’s in the CFP2 DCO business. It’s a complicated situation.
Troy Jensen: They decouple it, right and just sell them the DSP?
Andrew Schmitt: Yeah. Originally, it was the same way at 100G, is they only sold the modules to ZTE and then ZTE said, “Look, we just want to buy the DSP. We’re going to build our own optical front end.” There was some negotiation when that happened. Maybe the same thing happens with the CFP2 DCO. ZTE may not … I don’t know for a fact that they’re going full guns into this pluggable type of architecture that Huawei is doing. It may be that they’re going to do something different.
Inphi
Troy Jensen: I got another question here, emailed to me. It says, “Please ask him about Inphi, TIAs, drivers, near-term and then, the ColorZ opportunity.”
Andrew Schmitt: ColorZ, they talked about losing their second big customer. Maybe losing’s not the right word. But, the second opportunity isn’t coming through. My understanding is that was for the shorter reach version of the ColorZ that they announced at ECOC. I still firmly believe, and I would think that Inphi would probably agree with me, that long-term, ColorZ is going to be essentially put out of business by 400G ZR. Inphi knows this and they’re working very hard on the 400G coherent ZR business because that’s where the ColorZ stuff will migrate. In the meantime, they have this interesting business and they’re going to see what they can do with it to capitalize on the investment that they’ve made.
Troy Jensen: I think Microsoft, was their first customer. Would you agree Google was their second customer?
Andrew Schmitt: I don’t know.
Troy Jensen: Okay.
Andrew Schmitt: I think a lot of people are speculating.
Troy Jensen: Yeah, fair. How about just their TIAs and drivers, near term?
Andrew Schmitt: Well, I mean, they’re a good barometer for the coherent business because they have such a strong position. I’m not aware that Inphi’s position’s been compromised at all. I think the challenge will be protecting the margins because it’s a great business, there’s a lot of people looking at it. Inphi’s done a great job in terms of execution and moving up to higher baud rates and better throughput to support 400G. They are the barometer for the coherent business and I think if there’s one company in the world that knows what’s going on, in terms of coherent port shipments, it’s Inphi because they have the commanding heights in that business.
Troy Jensen: Last question on Inphi is just the PAM-4 opportunity.
Andrew Schmitt: They’ve been working on that longer than anyone else. The PAM-4 opportunity is huge. If you go into 2019, you start moving to 400G ethernet and also single-carrier 100G ports. The whole hyperscale market is going to move there in two years. I don’t have a good idea of who’s winning and losing in that market. The fact is that they have been working on it longer than anyone else. It seems to be a safe bet that they’re in a pretty good position there. But, I don’t know how they stack up versus the four or five other companies that are all trying to do the same thing.
Infinera ICE4 Prospects
Troy Jensen: Andrew, I want to ask you about Infinera. You’ve been very, very tuned into that business and nailed in on the upswing and downswing. Just kind of status on the new products right now, the Gen4 and the traction they’ve got in the market.
Andrew Schmitt: Listening to the most recent conference call, I was surprised just looking at the market reaction. I’m certainly not an expert in markets but, to me there wasn’t a clear sign that the company is completely out of the woods from a revenue perspective. So, I was a little surprised at what the stock market did. But, looking at the fundamental business, they’ve now started to ship ICE4 based product. If you look at from when they started first making shipments early last year, they just haven’t been able to ship it in any kind of big volume. I think what may have happened is that has now become unstuck. Whatever had that stuck has become unstuck and they can now roll that product out.
What I don’t know, and this is really the critical question, is how ICE4 competes against all of the existing 200G solutions that are now pretty entrenched from multiple vendors. When they came out with 100G, they were late but it came in at a cost point that was very attractive relative to the incumbent solutions. So, they were able to get some good traction.
But, if the ICE4, from a pricing perspective, for applications where people need multiple wavelengths, lots of capacity, and they can leverage the PIC capacity, then they’re got a good opportunity. I just don’t know where they are in terms of yields and costs on that particular product and how they’re going to face what is, at this point, a pretty entrenched 200G market that has already undergone, in 2017, some brutal price reductions.
We haven’t talked about this but spending from hyperscale in 2017, year over year, it’s not growing despite all the attention. The reason it’s not growing is that the pricing reductions were so brutal. I don’t see that happening again in 2018 because there’s not a similar technology transition. In 2017, we went from having two vendors who could sell 200G to five or six. In 2018, there’s no such event. There’s really no major key technology reaching the market in 2018. You have Ciena with 400G, but that’s sole-sourced. The other thing is you have Infinera now entering with the GEN4 pick. I wouldn’t expect the same pricing pressure in 2018 that we saw last year, at least I’m not modeling that. But, it really comes down to how Infinera’s solution’s going to stack up against what is already a pretty competitive marketplace.
Troy Jensen: What have you heard about them, technically? I think that was your issue when you kind of went negative on Infinera is that competition caught up with them and if ICE4 was effective they could take the technology lead again.
Andrew Schmitt: I hate being called negative on Infinera because if there’s one company I want to see succeed, it’s Infinera. They take the big risks and they’re doing very interesting work. They were late with their product. That was the problem. Now, it appears that it is in volume and they can ship it. You have Infinera entering the battlefield and it’s hard to understand how it stacks up against the existing solutions. I think technically it stacks up well. What I don’t know is whether it stacks up from a pricing perspective. I think they need a pricing advantage in order to really take share from the incumbents.
The last thing I’ll add is it isn’t just about price and performance, particularly with a lot of these hyper-scale vendors. They have very specific demands about how these systems integrate into their own systems and the types of data and telemetry that you send out. If you don’t have those sort of functions then, you can’t really design the product in at this point. Infinera was the first company to pursue this market with the CX. So, they do have some advantages in that market. They have an installed base with that product and that does give them a downhill fight.
North American 2018 CapEx
Troy Jensen: Can we spend the last 10 minutes, I guess I got two topics yet. Just carrier capex spending. What have you heard about Verizon or AT&T?
Andrew Schmitt: After the Lumentum call, I did a little digging around. I’m trying to put together my 2018 forecast for optical equipment. Based on what they said, I thought maybe I was being a little too conservative. But, after communicating with a couple of vendors, the Verizon business is not ramping this year. I got the impression from the Lumentum call that, that business was really starting to ramp into 2018 but I don’t think that’s the case. I think it’s more of a steady state business at this point.
The second thing is AT&T. AT&T is pulled back in Q4, particularly in long-haul WDM. They’re getting ready to roll out their new optical architecture. It’s been delayed. I don’t really see that happening until the end of the year. This is the new open ROADM initiative and the disaggregated terminals.
You take Verizon’s flat spend and AT&T is probably not going to start spending until the end of the year, that’s already a significant headwind for me to think that my forecast is going to be too low.
Troy Jensen: To that point, I think Ciena said on their call that they expect North American tier one to be down 2018 over 2017. That’s consistent. But I’d tell you, both Lumentum and Oclaro told me on my follow up calls that Verizon’s spending’s picked up. Maybe it’s just running through some inventory, right? Kind of getting back in the hole.
Andrew Schmitt: I can tell you that there were customers who heard the comment from Lumentum and were genuinely confused.
Troy Jensen: Right. Interesting. What about Century Link, Level 3? They got a big RFP out, do you think the stick with Infinera? Heard anything on when their spending bounces back?
Andrew Schmitt: I really don’t know.
Ciena 400G, Fiber-Deep, Applied Optoelectronics
Troy Jensen: What about anything else? You know, internationally, any deployments of significance in China Mobile, India-
Andrew Schmitt: Ciena’s had a tremendous amount of success, as has Nokia in India. You look at the revenue that’s come out of Asia, Asia ex-Japan, Asia ex-China, Ciena’s now sourcing almost $100 million a quarter out of that area. That’s been a big area of growth for them and I think they feel like they can continue to do well. It’s also impressive because they’re going head-to-head against Huawei in that environment. Nokia’s had a lot of success in the Middle East in 2017 going head-to-head against Huawei.
That’s matchup has been going well for both companies. One risk that I’ve identified in India is, if you look at where the spending has been coming from, a lot of it has been Reliance Jio. There was a merger announced between Jio and the old Reliance. Everyone knows what happens when carriers merge.
Troy Jensen: Not good.
Andrew Schmitt: All I have is that data point and then I have a history of what happens when this sort of event takes place. So, I am a little worried about that but, I don’t have any specific knowledge that this is definitely going to affect capex. People may not be aware that those companies are merging and I wanted to highlight that.
Troy Jensen: All right. Interesting. How about, the other thing we’ve heard this quarter too is on the cable side, seems to be bouncing back and hearing more about fiber deep architecture. Would you agree cable seems a little bit stronger? Who benefits off fiber deep?
Andrew Schmitt: I’m not aware of anyone shipping specific hardware into these fiber deep applications yet. What they are going to create, though, is a demand for greater backhaul bandwidth in more traditional systems. That helps drive demand. If you look at the cable companies, the one thing that they always do is deploy the latest high capacity technology. Everyone knows about the DCI folks, but the cable companies are very fiber constrained, so they want to put that stuff into play. I would think that’s going to be a big benefit to someone like Ciena because the 400G product would be perfect for their needs.
Troy Jensen: I got one more question here on email. It’s on the datacom side, so stop me if you’re not the expert. “Does Andrew have a view on Applied Optoelectronic positioning, what that looks like going forward? Then, what is Intel doing and what impact does that have on the market?”
Andrew Schmitt: No opinion.
Troy Jensen: On Applied or just datacom optics?
Andrew Schmitt: Everything went sideways last fall in that market around ECOC. I don’t trust any of my knowledge at this point. Everything changed very, very fast. Anyone who’s in charge of running that business right now, I really feel for them because it’s tough.
Troy Jensen: Agree. How about just any last thoughts? Any other topics we should have asked? We just got a couple minutes left here.
Andrew Schmitt: Let me look at my list I put together here. No, I think I wanted to talk about what was going on in China with the DCO designs and how that’s really impacting and putting extra life into the ACO market. We talked a little bit about 400G at Ciena, I think they’ve got a unique product there. We talked about the 400G situation with DSP suppliers. I really think, if you look at the next nine months, it’s a pretty static market on the telecom side, there’s not a lot of changes. It should be more predictable than usual.
Troy Jensen: Yeah, I’ve kind of said that too. It feels like we have to wait for the 400G market to ramp before these stocks can really work again. I did get one more, probably this will be the last one. It says, “How competitive is CX2 versus a wave server?”
Andrew Schmitt: I don’t know. I think a lot of it is going to come down to pricing. Infinera, as I said earlier, they have an installed base, they understand the needs of the hyper-scale folks, they were there first, so that’s got to help them to some degree. But, technically I don’t know how it stacks up. Like I said, it’s been difficult because until recently, there just weren’t any copies of this GEN4 PIC out in the field. You talk to carriers and you ask them, “Hey, how does the GEN4 stuff look like,” and they’d say, “Well, we don’t have it. We can’t get it.” That was the problem. Now that it appears that they’re starting to ship, more people will evaluate it and I can start to talk to folks and get an idea of how it’s looking. But, my understanding is that Infinera has been prioritizing all their shipments into all the hyper-scale folks, as well as some submarine applications. Bottom line is, I don’t know.
Troy Jensen: All right. With that Andrew, I think that is our time limit or thereabouts. Thank you so much for taking the time today. I guess I’ll probably see you at OFC.
Andrew Schmitt: I’ll be there.
Troy Jensen: Cool. All right, sir. Well, thanks again Andrew.
Andrew Schmitt: All right. Thank you, Troy.
Operator: Kind ladies and gentlemen, that does conclude our call for today. We thank you for your participation, everyone.
The transcript contains modification from audio to make it more readable and fix some annoying grammar.