Andrew Schmitt gave an update on the optical equipment and component market during an investor call hosted by Troy Jensen of Piper Jaffray on Tuesday, October 4. Sixty-eight investors participated. An extended Q&A session followed Andrew’s presentation, and topics of interest included:
- China 100G port volume trends for 2H16 and 2017
- Huawei’s upcoming 100G technology decisions and supplier purchasing shifts
- CFP2-ACO shipments and the component & equipment supply chain for this product
- Coherent DSP R&D developments and supplier updates
- Updates on direct-detect 100G and PAM-4 development
- Observations on Infinera, Ciena, Acacia, Inphi, and Clariphy and discussion of potential M&A activity
Following is a summary of the key takeaways, as well as a full transcript of the discussion. Cignal AI clients can also listen to an audio replay.
Register an account with your corporate email address to read the article.
- No drop off expected in Chinese component purchases for 2H16.
- 30% port shipment growth in China for 2017. All of the growth at Huawei will come from new CFP-DCO pluggable formats; suppliers must be designed into this new product to experience growth.
- Acacia likely not supplying CFP or DSPs for Huawei CFP-DCO effort. Instead Huawei is using an internal HiSilicon DSP and potentially some new NTT Electronics silicon. Acacia silicon photonics is of interest for current and future products at Huawei.
- Huawei will revisit ACO vs. DCO module debate in 2017 as part of a move to a higher density format than CFP-DCO.
- Approximately 5,000-6,000 CFP-ACO modules have been shipped through 3Q16. Oclaro has shipped most of these and it is the only company in production with substantial amounts of 200G capable parts.
- Fujitsu and Finisar are trailing behind, with production volumes 6 months in the future. Customer belief is that, of the two, Fujitsu is leading. Acacia is providing alpha samples of their own ACO module.
- Verizon is now taking delivery of significant ROADM nodes.
- 100G direct detect will be a niche product but things could change.
- A strategic combination of Clariphy and Inphi doesn’t make sense. Better to match Clariphy with a company like NeoPhotonics, Oclaro, or Lumentum to produce full optical subsystems, such as Acacia does.
- Infinera was late to 100G market and thrived. Possible to be late again and still succeed.
- Lots of R&D activity underway in ACOs and DSPs for lower power 200G and new 400G DSPs.
- CFP2s are being purchased by Webco’s as substitutes since QSFP28 modules remain scarce.
Troy Jensen: I know I talked to you briefly last week. You just got back from Europe, from the European Optical Conference and just thought maybe we’d start out with an update and your key takeaways from ECOC.
Andrew Schmitt: I was in Dusseldorf 2 weeks ago. Every year, there is a optical conference in Europe. It’s similar to OFC. It’s a little smaller in scale. It’s focused exclusively on components and there’s a little bit more of a technical bend to it. It’s a great opportunity to go meet with a lot of the component suppliers and also some of their customers who, mostly, are in the engineering side.
I think if you look at the high level takeaways from the conference this year, there were some interesting things that I learned on some specific areas, but the real big trends was everyone was talking about the Chinese volume and what happens next. We can talk about that today and how things are shaping up for 2017. There’s also a lot of activity going on with coherent DSPs from various players and people beginning to position themselves for the next technology leap which is going to be 400G coherent.
There’s also some activity around direct detect 100G opportunities, similar to what Inphi announced at OFC with the ColorZ product. We had a lot of discussions about CFP2-ACO, the ongoing ramp there as well as the end markets for that product and how it is starting to ramp in a lot of these new DCI systems. That’s about as quick as I could summarize it Troy. I don’t know where you’d like to start.
China 2016/2017 Projections
Troy Jensen: Let’s just start at the beginning. China has been the most topical and I can tell you from the analyst perspective, what Wall Street fears the most is that China bookings get so high so fast; that they might not be able to grow from here. I know all these guys talk about the 5 year broadband plan that they have and I agree, but will years 2, 3 and 4 be as high as years 1 was as far as the spending level. Your insight into China would be very helpful right now.
Andrew Schmitt: I think before we get into the specifics, I just wanted to share a little bit of my experience with what’s happened in China over the last couple of decades. You get these bursts in terms of spending and every time it happens, everyone says, “Oh my gosh! There’s no way that this is sustainable,” yet it goes on and on much longer than anyone would have expected.
The areas that I think of specifically are the DSL market. That was huge in China and went on for a long time and then later on, the fiber to the home market. This feels a lot like that to me, where there’s a sudden shift to a new technology and the expectation is that this is just a short term type of bump. In the end, it’s a long run for some time.
Now, specifically on China, I met with Huawei and I talked with a lot of their suppliers 2 weeks ago. I came away with the conclusion that there’s not going to be a drop-off in terms of 100G purchases in the second half of this year or in the next year. You’re certainly not going to see the same kind of year over year percentage gains that we’ve experienced in the last 12 months.
Troy Jensen: You said there’s not going to be a drop-off. You expect it be stable on a year over year basis out of China?
Andrew Schmitt: Well, the consensus is that it’s going to be at least 30% next year. Some people say 30%, others say at least 30%. When I talked to Huawei and I ran the 30% number past them, they agreed to it pretty quickly. There was no hesitation in terms of that kind of number when I spoke to the folks there.
Troy Jensen: You’re talking about 30% revenue growth next year?.
Andrew Schmitt: No. 30% port growth (in China during 2017). That’s an important distinction.
Troy Jensen: Okay, and you’re speaking specifically about 100G ports, right?
Andrew Schmitt: 100G coherent ports, correct.
Troy Jensen: All right, so expecting growth next year in the ports. Are you expecting ASP erosion to offset that or are we going to have revenue growth at this level?
Andrew Schmitt: That, I don’t know. I don’t have a good idea of pricing going into next year. That’s obviously the magic number but I don’t know how that’s going to shape up.
Huawei CFP-DCO Transition
I think here’s the important thing, is if you go look at, and this is an important detail that I learned, the growth that’s coming next year, this 30% unit growth, at least in the case of Huawei, what they guided me towards was that this growth would be mostly in the CFP-DCO format; is that they expected the shipments of their traditional, 100G line cards which are built primarily with their own DSPs as well as NTT Electronics, and discrete components from a number of suppliers including Oclaro and NEL Photonics.
They were saying that that business was going to be flattish going into next year, and that they really expected the unit growth, when it turns on, to come from the CFP-DCO market. They’re trying to convert their production over to using these pluggable, coherent modules. Unlike the rest of the world which is ACO, they’re really targeting DCO which is where the DSP is inside the CFP itself.
Troy Jensen: Okay, so 30% growth but that growth coming from DCOs, you’re saying.
Andrew Schmitt: I think the key thing here is, from a component supplier perspective, you really need to be designed into that new DCO module that Huawei is building if you want to continue to participate in the unit growth.
Troy Jensen: Got you.
Andrew Schmitt: Because other than that, it looks like things are going to get flattened out.
Troy Jensen: Who do you think has the most exposure to that form factor? I mean Acacia sounds like it would be the one.
Andrew Schmitt: Acacia is not playing any role in Huawei DCO as far as I can tell. They may be supplying a silicon photonics frontend in that module. It wasn’t clear to me if that was something they were doing for the existing CFP module, or if that was something that Huawei was going to be looking at a year from now when they cost reduce the CFP-DCO and go to CFP2-DCO. My impression was is that Acacia was not playing a role in that module.
DSP Vendors in China
Troy Jensen: Then the DSP they’re getting, it’s either from HiSilicon, they’re internal R&D or NEL?
Andrew Schmitt: They’re planning to use HiSilicon. I don’t know this for a fact, but I suspect that they’ve designed that in conjunction with some assistance from ClariPhy, but they’re also continuing to work very actively with NTT Electronics (NEL).
Troy Jensen: You are the one who’s also always told me that Huawei has a 1+1 approach. They want to internally develop and outsource. Is this going to be a component they outsource or is this going to be all internally developed with HiSilicon?
Andrew Schmitt: It was pretty clear talking with them that they’re planning to use their own internal DSP development and they’re going to be using NTT Electronics.
China Pricing, ROADMs
Troy Jensen: Then I did have a question from a client here just, “What was the 100G port growth this year in China?”
Andrew Schmitt: Well, the year is not over.
Troy Jensen: Yeah, you have to have a good estimate.
Andrew Schmitt: I don’t have my numbers in front of me but looking at the first half of this year, it more than doubled.
Troy Jensen: Then if I go back to my question about revenues, so yes, we’ve got 30% growth in units. Those units are shipped in towards DCO form factor, but for the most part, we’re not going to see any type of pricing erosion near that 30%, so you would expect China spending on optics grows next year.
Andrew Schmitt: Yeah. I’m not going to try to predict the pricing with Huawei.
Troy Jensen: I don’t think it’s any close to 30%, right?
Andrew Schmitt: The suppliers are in a much better position to give comments on that. I think Huawei knows that they’re purchasing a huge amount of components and they have a lot of leverage. I imagine that this transition to the CFP is … This is an opportunity for them to say, “Hey, you know, if you want to be … If you want to stay on the train, you’ve got to helps us out.” I don’t think it’s 30% but I’m sure there’s going to be some reductions coming, given the big jump in demand.
The flipside of that though is, as everyone knows, is there’s not all kinds of capacity sitting around. It’s a question of whether someone wants to sell off that capacity to Huawei at a lower price that they might get somewhere else.
Troy Jensen: Fair. All right, and then sticking to China, have you heard much on China datacom spending initiatives?
Andrew Schmitt: I have not. I mean I’ve had some discussion surrounding the Google and Microsoft equivalents there, the Baidu, Alibaba and Tencent. People there, call them the BATs. There’s been a lot of movement in terms of technology and vendor selection but they typically are much more free-thinking and aggressive when compared with their traditional carriers. Not unlike what we’ve seen here in the west.
My understanding is that they’ve been much more interested in CFP2-ACO technology than some of the more traditional telco customers. For example, they’re using Nokia and their pluggable ACO approach. Although, I don’t think Nokia is actually using true pluggables but those customers have a lot of interest in that type of approach because they believe there’s going to be a higher density that they can realize with it.
Troy Jensen: All right, perfect. I want to get one more in on China and then we can move to another topic. I’ve been told that China’s going to deploy ROADM, so do you have a thought on the timeline for when and then, more importantly, I guess the key question is are they going to do a contentionless architecture? Because if they do, that needs more WSS, to my knowledge, and it needs multicast switching.
Andrew Schmitt: I haven’t dug into the contentionless question with China. I don’t really have the primary knowledge on that. I also don’t have a good idea about the timing on the ROADMs, but in talking informally with some of the supply chain, I get the sense that both China Telecom and China Mobile are going forward. China Telecom is actually ahead in terms of volume, although, China Mobile is very active from an R&D standpoint.
The expectation is that they’re going to catch up. The other interesting thing that I’ve picked up that I haven’t really been able to investigate is when I talked to the folks from Lumentum, they always talk about China Telecom. When I talked to the folks from Finisar, they always talk about China Mobile. I’m just wondering if there’s some sort of relationship between the pairs, whereas one particular WSS is related to that particular customer, but that’s just anecdotal. I don’t know whether there’s a relationship.
Troy Jensen: Just so I heard that right, you said Lumentum talks more about China Telco and Finisar talks more about China Mobile.
Andrew Schmitt: Correct.
DCO vs. ACO Roadmap in China
Troy Jensen: Perfect. All right, so I do get a question from the audience here. It says, “Why DCO for Huawei? Doesn’t it have fundamental heat issues?”
Andrew Schmitt: Well, just like in every other industry, there’s not always rational decisions being made. This is something that they’ve (Huawei and China) been very focused on for many years. I recall having meetings with China Mobile, I think, 2 or 3 years ago and talking about ACO versus DCO. They were very opposed to ACO and some of the risks and the complexities of that approach. I think fundamentally, one of the things that they like in China is, with the DCO approach, they can pull out that module in 2 years and they can plug something else in that does whatever they need it to do.
Whereas with ACO, the DSP is inside and you really already determined what the function of that line card’s going to be. The DCO gives them some more flexibility that they like, particularly if they start looking at putting this into the Metro where maybe they’ll have a different CFP with better performance that they want to use at some point in the future. That’s one comment that I’ve heard from the Chinese and why they’re looking at DCOs.
Even speaking with Huawei, they’re very upfront that they’re going with DCOs but when they make the transition to CFP2, which they expect to do next year, they’re going to reevaluate the DCO versus the ACO and decide what to do.
Troy Jensen: All right, so next year, they’re going to make a decision about CFP2 form factor?
Andrew Schmitt: Yeah… Exactly.
Troy Jensen: ACO or DCO, okay.
Andrew Schmitt: They’re going to look into whether they continue to use DCO architectures when they migrate to a CFP2 which is twice the density of the CFP.
Troy Jensen: Got you. All right, that’s good. I did get another question here from a client. What sort of components will be used inside the Huawei or the DCO, specifically lasers, modulators, versus when they buy discrete components?
Andrew Schmitt: It’s pretty much the same thing except it’s inside of a box. There isn’t a tremendous difference between the 2 architectures. The biggest difference is that the DSP and the power envelope that that has to live in becomes a lot tighter. My guess is that Huawei probably had to make some changes to their own DSP in order to get it to work in that environment, or they’re planning to use something from NTT Electronics.
I think one of the things that’s a bit confusing is when you talk to Huawei, they’re very adamant that, going forward, they’re going to be using their own DSPs and make that a big part of their own push forward. If you talked around them, inside the supply chain, it’s evident that they’re still sourcing product from other sources, particularly NTT Electronics. It may even be possible that they will be purchasing some Acacia CFPs, but talking directly with Huawei, I did not get that impression.
Troy Jensen: Very fair. Then this is another question from a client. We hit it but just to go over it, the competitive environment for the DSPs right now, so it’s Acacia-
Andrew Schmitt: It’s an interesting environment for DSPs. You have companies like NTT Electronics who are still continuing to invest in multiple DSPs going forward. Acacia is definitely investing and ClariPhy is still out there looking to continue to provide some either design expertise to the folks who want to be vertically integrated as well as their own DSPs.
I don’t know of anyone else who’s waiting to enter the market beyond those 3 suppliers. Outside of that, if you look at the vertically integrated guys, you’ve got Infinera, you’ve got Nokia who just announced their own coherent DSP that’s got some legs still. I would expect that Ciena has got to make a decision on what they do for next generation DSP next year. We hear about what they’re going to do but the big story is really 400G and how you get to 400G and begin to address the shorter distances with a higher baud rate and higher modulation for more bandwidth.
CFP2-ACO Market Update
Troy Jensen: Okay, fair. Why don’t we switch gears a little bit now? We hit DCOs pretty thoroughly. Let’s jump into ACOs. It gets multiple questions there. Can you talk about competitive environment? Anyone else qualified yet?
Andrew Schmitt: Yeah. I spoke with a couple system vendors and got a pretty good lay of the land on ACOs. Putting together my own internal projections, I think that, to date, there’s been around 5,000 to 6,000 ACO modules shipped. The vast majority of those have been shipped by Oclaro.
Oclaro is really the only supplier right now that can provide 200 gig capable ACOs in any kind of volume. Fujitsu is now sampling 200 gig ACOs, but talking to their customers, the expectation is they’re not going to be able to supply significant volumes for at least 3 to 6 months. The same is true for Finisar. The expectation is that Fujitsu will be able to supply more volume before Finisar will.
Troy Jensen: Why is that?
Andrew Schmitt: Fujitsu is definitely … They’re bringing 200 gig to market but it’s going to be 3 to 6 months before they can really provide any kind of volume.
Troy Jensen: Let me challenge that, 3 to 6 comment, I mean to go from sampling to volume, if you think how long it took Oclaro, isn’t more like 9 to 12 months between sampling and volume?
Andrew Schmitt: Well, I think there’s a difference between supplying something and then actually shipping it for production. It may be that they can ramp their production faster than people can qualify it and begin using it.
Troy Jensen: All right, that makes sense.
Andrew Schmitt: Fujitsu, when you look at people who started on ACO years ago, they were one of the first ones to start. They’ve put a lot of wood behind the arrow and it’s just a matter of getting this stuff up and running. When you talk to their customers, when you’re dealing with Japanese suppliers, they take a long time to get the plane running down the runway but eventually, when it lifts off, it really goes.
Troy Jensen: Hey, what about continuing on competition for ACOs and any thoughts on Acacia’s ACO?
Andrew Schmitt: Yeah, so from what I understand, they’re sampling it now and the comment that I heard was that it’s an alpha product and it’s still a long way away from being something that people can really qualify.
Troy Jensen: Those are really considered the next 3, right, for the ACO market. I know there’s others that have announced but we don’t know.
Andrew Schmitt: Neophotonics has announced an ACO and it’s a very pragmatic type of product where it’s 100 gig only. They’re not using any kind of exotic technologies. It’s a brute force approach that they feel will produce a product that’s high yield and low cost in the 100 gig market. The downside is that they can’t really stretch it to do 200G. They’re just going to compete in that niche (100G only).
That being said, most of the … Really, all of the demand for ACOs is in the 200 gig side of things. That’s where it becomes compelling.
Troy Jensen: How big do you think the ACO market’s going to be in 2017?
Andrew Schmitt: I don’t have a specific number on that but I think you could begin to that out by look at Verizon and what they’re going to be purchasing, and then looking at the DCI market that’s being served by a number of different suppliers. The big guys are Infinera, Cisco, Coriant, Fujitsu and … Missing one, starting with those … Infinera uses their own in-house technology. Also, there’s ADVA but ADVA is using the Acacia product.
There’s a lot of interest and uptake in those platforms and I think the ACO is really the lead item for things like the Cisco NCS 1000 and the Coriant Groove platform as well as Fujitsu. Fujitsu, from what I understand, is starting to have some success with their product because they’re not just going after the DCI market but they’re actively targeting guys like the cable MSOs here in North America as well as AT&T.
Troy Jensen: All right. Well, let me go back to another question or a statement you had, is that 5,000 to 6,000 ACOs shipped. Is through end of Q3 or mid-year or is that for-?
Andrew Schmitt: End of Q3.
Troy Jensen: End of Q3, and so how much do you think Oclaro’s manufacturing currently?
Andrew Schmitt: I don’t know.
Troy Jensen: They’ve got to be 80%, 90% of that, 5,000 to 6,000, right?
Andrew Schmitt: I don’t know.
Troy Jensen: I know they’re guarding it pretty closely, but could this industry be a 250 to 300 million dollar market next year?
Andrew Schmitt: Well, let’s see, so if we … Let’s just say it was $6,500 a unit. Where does that put us in terms of total units shipped? It’s around 45,000 units. I’d say that’s probably on the upper end of where things would end up.
Troy Jensen: 45K, all right, and fairly, Oclaro are going to have a dominant share of that because these other guys got to go through qualifications still and ramp volume capabilities.
Andrew Schmitt: Yeah, and a lot of these things trace down into specifics like the new Infinera platform, the Cloud Express 2. If that comes out and captures a lot of the market, that eats into the ACO opportunity, whereas if they struggle to get that out and qualified and working, then it increases the ACO opportunity. Likewise for ADVA and their success with their new platform which is Acacia-based. Those are really some of the variables that impact the ACO adoption going into 2017.
Verizon and ROADMs
Troy Jensen: All right, so I’m going to just shift to next topic. I’m going to look through some of the questions too coming in, but let’s just hit quickly on WSS and ROADM line cards. I know we hit a little bit on China. Can you talk about status with Verizon and maybe status of Finisar at Cisco?
Andrew Schmitt: My understanding is that with Verizon, there’s already significant node shipments taking place from Ciena and from Cisco. While they’re not necessarily recognizing a lot of revenue, they are shipping out product, so I think that the supply chain for those components is starting to ramp now.
Troy Jensen: Any thoughts on Finisar getting certified, qualified at Cisco?
Andrew Schmitt: I have no idea. I don’t know.
Troy Jensen: We still don’t know yet either. All right, so ROADMs to DCOs, ACOs. Well, I think Ciena might have been the one company that you forgot to mention. Is that correct?
Andrew Schmitt: Oh, yeah, for ACO. They were really the ones that made the format happen and drove it. Cisco has extended this product all throughout their whole product line, and my feeling is that Ciena is in the process of doing the same thing. There is likely to be new products coming out from Ciena in the next 12 months that are based on the ACO platform. That’s my expectation.
100G Direct Detect & PAM-4
Troy Jensen: All right, interesting. Hey, let’s move on. Another topic you brought up, and I got a question on this, about the direct detect market. It’s interesting. If I go back, it seems like when it was first announced, Microsoft, Inphi, everybody denied the ability to go 40 to 80K, right? Then since then, there’s been more people announcing direct detect type of solutions, so just quickly, your thoughts on that market.
Andrew Schmitt: I think it’s an important niche market for the next couple of years. After 2018, after 2019, it remains to be seen whether it can hold off an incursion by coherent and the advances that that technology is making in terms of power and cost and density. There is a couple of key customers that are going to buy things next year, but I think it’s still the jury’s out on whether that becomes a widely adopted technology.
When I look at my models and how I’ve modeled out the adoption rate of direct detect is it’s going to be more of a niche used by a couple of select service providers rather than something that’s used in a widespread way. I’ll be the first to admit that there’s a lot of uncertainty in that kind of a prediction at this point.
Troy Jensen: To reword your comment, it’s a Band-Aid solution for now but coherent is more flexible and better for getting to 200 plus speeds, is that correct?
Andrew Schmitt: Yeah. I wouldn’t call it a Band-Aid. It’s a very specific point product that solves the problem that needs to be solved right now.
Troy Jensen: All right, so then outside of Inphi, who else is there and any thoughts on the size of that market maybe this year or next?
Andrew Schmitt: I think you’ve seen an announcement from RANOVUS, and they’re working with ADVA as part of their press release. I think ADVA is pretty excited about that technology as well, but it really remains to be seen how widely adopted that approach (direct detect) is. You have companies like ADVA which … They have a long history of taking some of these corner case technologies and successfully introducing them into their customer base. That’s one of the things that’s really made that company successful.
Whether or not you see guys like Ciena introduce it or Cisco, I think we’re still some ways from seeing that happen.
Troy Jensen: Then it’s interesting too. If you think about my thoughts, my observation is PAM4 was announced with this direct detect and everybody poo-pooed the idea, but since then, everybody’s had some announcement with PAM4. I think it’s just been more broader –
Andrew Schmitt: Yeah, but PAM4 is a vital technology for inside the data center, and 2 kilometer types of reaches. I think when you look at all the PAM4 announcements at the component level, the market that they’re really aiming at is 2 kilometers or less and getting more bandwidth out of the 25 gig capable lasers that are out there.
They’re not going after this 80 kilometer market. In terms of volume, it’s just going to be a pimple compared with what goes on inside the data center when they go from 100G to 200 or 400G.
Troy Jensen: That’s my understanding too, it’s inside the data center, a PAM4 should be broadly deployed, right?
Andrew Schmitt: Yeah.
Troy Jensen: Interesting, and as far as leaders in that space outside of Inphi, Broadcom, anyone else we should watch?
Andrew Schmitt: I don’t have a good idea of benchmarking or ranking the players but I know that APM is also in that market. I would expect that you’d see somebody like MACOM also having some sort of solution there to pair up with their optics. I think there’s still a lot going on there but Inphi was really one of the companies that really went hard and fast into this market first.
Clariphy Acquisition Thoughts
Troy Jensen: We get a lot of emails, just not a lot of live Q&A so I’ve got a few more I’m going to ask here. These are maybe out of order and jumping around a little bit Andrew, but just to hit some of the inbound questions. Here, it says, “Andrew mentioned Huawei might be using ClariPhy IP. Any thoughts on whether Inphi might acquire ClariPhy, you know, or maybe what is the strategic logic of that combination?”
Andrew Schmitt: I don’t know. The fact is if I knew anything, I couldn’t say anything.
Troy Jensen: Can you speak to the strategic logic of Inphi owning Clariphy?
Andrew Schmitt: Well, a DSP company is a major investment. Acacia has made this case that it’s a major investment and they’re right. For someone like Inphi to pick up a company like ClariPhy, obviously, it always depends on price. You’re talking about a business that’s going to burn a substantial amount of cash in the next couple of years. You need to make sure you have enough sockets, enough design wins in order to justify the continued development in that area.
Strategically, I think it makes sense for someone like Inphi. I would argue that probably, it makes more sense for an optical component company to take that approach, someone like a Oclaro, a NEL Photonics or a Finisar because then, they can, in essence, deploy the same model that Acacia has deployed. That is being able to closely couple all the optics and the DSP to sell a full subsystem, whereas Inphi can’t really do that.
I think Inphi would struggle to make the numbers work, and you would really need to, in essence, copy the same business model that Acacia has. The companies that are best poised to do that are the folks who also have the discrete components.
Extended Audience Q&A
Troy Jensen: Okay, fair. Hey, so I thought of another one on DCO that you just brought it up. We talked about competition for the DSPs. Now that we’re getting more DSPs qualified here, what’s the risk of new competitors having DCO offerings?
Andrew Schmitt: New companies introducing DCOs such as Acacia?
Troy Jensen: Yeah, why doesn’t Oclaro, why doesn’t NEL Photonics come out with the DCO products?
Andrew Schmitt: Oh, I think once NTT comes out with their low power DSP, you are going to see that. If, in fact, there is a market for DCOs outside of Huawei, which I think is a big question. Let’s just assume that there is, then yeah, absolutely. You would see people taking the NTT DSP and putting that into a module and selling it. At that point, it just comes down to whether it has the same kind of performance as Acacia.
Acacia has a track record of producing some very good product, so could somebody go and build DCOs? Sure. Could it be the same performance as Acacia? That, I don’t know.
Troy Jensen: A few more questions lighting up my inbox here. I’m gonna jump around here, but we hit this a little bit but there’s some more depth here. “How much demand is there for PAM4 related products from the hyper-scale players? Obviously, we know about ColorZ, Microsoft announcement, but when does Andrew expect single lambda PAM4 solution to come to market?”
Andrew Schmitt: Oh, I think you’re probably talking about inside the data center. I haven’t really spent enough time on that to form an opinion, so the short answer is I don’t know.
Troy Jensen: Don’t know, okay. Let me keep jumping, so now we’ve had a couple of guys shooting me questions about Infinera, and I know that’s a topic you know well. Could you just talk about … Here’s a question, “Does Andrew have any thoughts on Inphi’s new products for DCI and Metro in Infinera …” I said Inphi, “Infinera’s new products for DCI and Metro? Can Infinera catch up on DSPs and the peaks in time to participate in the Metro and 400G?”
Andrew Schmitt: I think the short answer to that is yes. Infinera was late to 100G and they ended up coming out with a product that was significantly better than what other people had at the time. The issue with Infinera is that their cycle time, in terms of R&D, is too long. They take several years to reload the gun and update their PIC. My understanding is that Tom Fallon and the folks there are now changing the way they approach this, so they don’t have this 4 year cycle times in terms of technology evolution.
If you look at the product that they have announced, it’s very compelling. If they can get that working and out to market, my feeling is that you could have a repeat of what you saw in 100 gig market where they were late but they came out with a great product. The types of systems that they’re building are well suited to where the demand is coming from, these very high density, point to point connections.
Plus, they’ve made a lot of changes to the peak architecture that makes it much more suitable for the Metro whereas before, it really wasn’t suitable for lower cost Metro use. Yeah, I think it’s entirely possible. It comes down to the ability to execute and deliver the product and I can’t comment on that.
Troy Jensen: There’s another one on Infinera. It says, “It seems Infinera won’t have 200G, 16 quam product out until next year. Is that a strong positive for ACO market?”
Andrew Schmitt: I don’t know the exact date when Infinera is going to have this product out. Regardless of whether they are successful or not, there’s going to be a good market for the ACOs because outside of China, that’s really the format that everyone’s adopted. People are either using an Acacia solution or they’re using their own DSP or an Acacia DSP and putting an ACO in front of it. Unless you expected Infinera to take over the market, there should be a substantial market for ACOs, and I don’t think anyone’s predicting that’s going to happen (Infinera taking over the market).
Troy Jensen: All right, that’s fair. Another one her, and someone’s been on multiple calls, I can tell by his questions. He said, “Previously, Andrew said 20,000 ACO shipments in 2016. For now, at 6K, is that 20,000 still possible or what’s your new projection?”
Andrew Schmitt: No. I think it was too high. I think there was demand for 20,000. I think the issue is there just hasn’t been the supply. If we were to rewind the clock a year ago, I think the expectation was that there’d be other people out there beside Oclaro providing product and it just hasn’t happened.
Troy Jensen: Do you have a new forecast for ACO shipments this year?
Andrew Schmitt: I don’t but it really comes down to what Oclaro can build in Q4. They’re going to be able to sell anything they can make.
Troy Jensen: Exactly. All right, any new thoughts on Lumentum’s ACO development?
Andrew Schmitt: I haven’t heard too much in terms of what they’re doing there. I think if you’ve been reading their conference calls, you’re familiar with … They’ve taken a much longer approach. My guess is that they may be bypassing the existing 100G, 200G ACOs and really targeting the next step which is going to be 400G or, at least, going to an architecture that can run a 200G and scale to 400G.
That’s really the roadmap for ACO, is this transition from 200 to 400G, starting to see discrete components of 400G. The next is going to be 400G ACOs in the CFP2 form factor. Then you’re also going to see 200G ACOs in an even smaller form factor called OSFP. That’s what people seem to be centering on. There was a proposal that was put together by Arista.
It’s reasonable to think that maybe in 18 months, 2 years, you’re going to be able to get 200 gigs of coherent capacity in something that’s only slightly larger than a QSFP28 today.
Troy Jensen: Well, all right. One thing I had written up for a question I wanted to ask you is I’m starting to hear more about CFP8 or the LR8 module. To my knowledge, that’s 400G but probably more of a router interface. Is that close? Is that something important to watch?
Andrew Schmitt: The very first modules, like very prototype modules, were actually on the floor at ECOC so I would expect to see a lot of activity at OFC next year, people making product announcements around that product. From a volume perspective, what’s really interesting about that is it is prototyping all the technology needed to really cut the cost of 100G Ethernet by using PAM4, because they leverage all the PAM4 and they leverage all the higher speed optics in the CFP8.
All the development that’s being done for CFP8 can be re-purposed for the 100G market, and that’s where it’s really going to be big in terms of volume. You’re going to see a shift in terms of the type of technology that’s used inside the data center from 25 gig per wavelength to 50 gig per wavelength. The CFP8 is just 1 vehicle for how that’s going to happen.
Troy Jensen: Is there anyone close? Is there someone that’s going to have a competitive lead here or is it still too early to tell?
Andrew Schmitt: I don’t have an opinion on that.
Troy Jensen: I was just asking because typically, if you get a time-to-market advantage like the ACOs from Oclaro, you can really help the business. Hey, a couple more questions here. I’ve got a bunch actually. “Can you please ask Andrew his view on Ciena’s business in the near term as well as over 2017? I know he commented on Verizon, we’re happy for Ciena now but curious about thoughts on Ciena’s overall business and supply chain.”
Andrew Schmitt: I think that if you look at from a macro perspective, you take away China, this year, actually, it really hasn’t been that good from a spending perspective. I think that there’s been, in essence, some pause as people wait for these new 200G technologies to come to market. Verizon’s an example, they are a little bit later than what people originally expected.
The cable MSOs really want to get their hands on a lot of this 200G technology. Once it’s now in production, I think that next year’s going to be better for the West, particularly in North America. Ciena is the big proxy for the overall market because they sell to so many different types of customers and to so many different markets. My feeling, without using any numbers, is I would think that Ciena’s business next year is going to be significantly better than this year.
Troy Jensen: That’s fair. I’ve got 2 questions about DCOs and what it means to NEL Photonics. Here’s one of them, they’re about the same but does the hard shift to DCO mean that NEL is a net share loser at Huawei and then this kind of news-?
Andrew Schmitt: Well, I think it depends on whether they’re designed into the CFP, and I don’t know the answer to that question. I’ve got a sense that Oclaro was playing a role in the CFP at Huawei. I would have a very hard time believing that Neophotonics was also not there given the relationship that they have with that customer. Obviously, Huawei is going to have more than 1 source for the components that are in there.
If I had to guess which 2 suppliers are going to be supplying the majority of the content of those modules and the components, it would be Oclaro and Neophotonics, and that’s a guess.
Troy Jensen: Okay, fair. Here’s the other question, sounds like a little bit of a clarification request, so is NTT supplying all the components for Huawei’s DCO and Huawei is using HiSilicon’s DSPs or is it NTT’s DSPs? Is there any opportunity for Neophotonics?
Andrew Schmitt: I’ll be clear. What Huawei has communicated to me is that they’re using their own DSPs in the CFP-DCO. NTT Electronics, they don’t really make the other components that sit in front of the DSP. They do make some optical components but I don’t think they’re playing a role there. I’m not aware of it. I believe that NTT Electronics is shipping some DSPs and will continue to ship some into Huawei, but I don’t know if they’re also going to be inside the CFP. According to Huawei, they’re exclusively using their own DSPs going forward.
Troy Jensen: Got you, understood. When do you think NTT’s DSP will come out?
Andrew Schmitt: I don’t know. I would expect that they have several in development. They’re going to be coming out with a lower power version of the products that they have now that, I think, would enable higher density, 200 gig types of solutions, maybe even 200 gig CFPs or CFP2s in DCO format, but they also have other products in development that I don’t know the details of.
Troy Jensen: All right. Let me-
Andrew Schmitt: I guess to be specific, my expectation is you’re going to see a lot of 400G DSP announcements between now and around OFC.
Troy Jensen: OFC, yeah, that makes sense. Has Andrew met ZTE, any rundown on their strategy, demand or how they’re evolving?
Andrew Schmitt: I haven’t had any discussions with ZTE of late. I will share something that’s a little bit interesting about ZTE. My understanding is that, well, I know that ZTE used to purchase DSPs from both NTT and from Acacia. One of the questions has always been – what’s the mix? How much is ZTE buying from NTT versus Acacia? I never really knew that. One of the things I learned in Dusseldorf is that ZTE has pretty much converted all of their production to Acacia on the DSP side.
NEL has essentially backed away from that client. I’ve seen behavior like this in the past in China, is that NTT has done some sort of a deal with Huawei to be a supplier. As a result of doing that, they’re not going to be providing things to ZTE.
Troy Jensen: ZTE is probably the other half.
Andrew Schmitt: My expectation is that, at this point, Acacia really owns a big part of the ZTE account. What I don’t know is how ZTE plans to deal with the shift to CFP. They do have a history of emulating what Huawei does so I would expect they’re also going to be looking at how do they start shifting CFP-DCOs? In their current position, their only real option is going to be to source this from Acacia, would be my expectation.
Troy Jensen: Here’s a timely question given you just got back from Europe, but we heard about Orange starting their upgrade, anything big going on with the European telcos and how would they ramp compared to the China ramp?
Andrew Schmitt: I haven’t really heard much on the European telcos one way or the other.
Troy Jensen: That’s fair. Do you have a market size for the DCO market in 2017?
Andrew Schmitt: Yeah. Basically, it’s going to go from something that’s pretty small. I mean I think Acacia talked about shipping, what was it, 8,000 units last quarter? The majority of those went to ADVA and so the consumption of the DCO module is outside of ADVA is pretty minimal. It’s growing from a small number and I would expect that almost all that market growth is going to come from Huawei and the shipments that they make.
I don’t think it’s all Acacia DCOs (includes in-house Huawei DCOs) but the DCO market could be… Let’s say 40-50,000 units next year, but it’s very dependent on what ADVA does and whether they convert from using the DCOs to the 400G modules from Acacia. My expectation is that’s what they’re going to do.
Troy Jensen: A couple more here Andrew, it’s going back to a question, so for the direct detect markets, has anyone other than Microsoft announced that they’re going to deploy this?
Andrew Schmitt: No.
Troy Jensen: Then I’ve got a couple questions here on QSFP28 demand and supply. I know we haven’t spent much time on that yet. I know it’s not as much your expertise.
Andrew Schmitt: There was an interesting thing on QSFP28. One of the things that appears to be going on is that the Webco’s., the big guys have actually been purchasing CFP2s because they can’t get the QSFP28, because supply has been so constrained. They’ve been buying some of the larger modules in order to get things to work. There’s obviously a lot of demand for QSFP28. I don’t have a number on that market but there has been some demand for CFP2s as a result of the supply being so constrained in the QSFP28 market.
Troy Jensen: Do you think these constraints on QSFP28 is helping out 40 gig transceivers too?
Andrew Schmitt: The short answer to the question is I don’t know. The long answer is if people can’t convert over to 100G because there’s not the optics, then yeah, they’re going to continue to keep buying 40G. That’s the logical conclusion.
Troy Jensen: Me too. All right, so then another one here. It’s, “How is Coriant share for 100G new builds?”
Andrew Schmitt: Coriant has a couple new products that are just coming to market. I don’t know what their … I mean their share and participation right now is pretty minimal in the 100G market. I think they have been actively pursuing upgrades at AT&T with their product, and my understanding is they’ve been very aggressive in terms of pricing out in the marketplace. They have new products that are coming to market that really represent the growth potential for 100G and that’s their Groove product and there’s a new long haul blade for their 7300 WDM system.
Both of those are based on Acacia DSPs with … In the case of the Groove, they use ACO optics and in the case of the WDM system, they have their own in-house discrete design.
Troy Jensen: All right, a couple more here. You may not know this but I’ll run this past. Have you heard any updates on Huawei’s issue with the US Department of Commerce, very similarly with ZTE, thoughts on the trade restriction stuff?
Andrew Schmitt: I don’t know anything about that. I’m only away of what was going on with ZTE.
Troy Jensen: Just a couple more here and I’ll let you go. Guy says, “Does Andrew have a view on Acacia for 2017?”
Andrew Schmitt: I think Acacia, their success is a function of the success of their customers. They depend on what happens at ZTE and at ADVA and at Coriant. Beyond that, their success is going to depend on their ability to penetrate other accounts and bring them in-house. One of those is Cisco. If I had to do the math, I don’t think that Cisco is going to be material next year, that would be my guess.
The other thing is the ACO module, and that’s going to be exceptionally interesting from an engineering perspective because we’ve never had a head to head, indium phosphide versus silicon photonic shootout. That’s what we’re going to get with their silicon photonic ACO.
Troy Jensen: Exactly. All right, I got 2 more that popped in then I’ll let you go. It says, “How are demand trends shaping up among cloud service providers? How much of an impact will hyper-scale players have on the optical market in the second half of this year and next year?”
Andrew Schmitt: I haven’t heard of any changes in demand from them. I think there’s technology shifts that are underway and the big one is the transition to 100G. It appears that that’s a function of how much volume they can get at what price. Once a QSFP28 hits a certain volume threshold at the right price, I would expect that those guys, they transition pretty fast, so I-
Troy Jensen: Yeah, and so if they can get QSFPs. You said they’re buying CFPs, CFP2s, right?
Andrew Schmitt: Right, so once that supply is in place at the right price, then I think you’re going to see a very big move to 100G away from 40G pretty quickly. The thing that’s different this time is when there was a transition from 10 gig to 40G, not everyone moved at the same time. Now, it’s evident that all these guys are ready to go as soon as they have the supply. I think that the transition is going to be more dramatic this time.
Troy Jensen: All right, last question I promise. What is Acacia supplying into Cisco and for what applications?
Andrew Schmitt: Acacia is supplying a DSP to Cisco and they use that in the NCS 4000 which is used for the Verizon business.
Troy Jensen: The Verizon business you said?
Andrew Schmitt: Yep.
Troy Jensen: Metro. Okay, but Cisco has their own DSP but is this a 1+1 strategy or are they acknowledging that Acacia is better than what they had?
Andrew Schmitt: I think Cisco decides to use whatever technology meets the requirements in the short term, and I guess that Acacia had a better product for the Verizon business. They’re using their own DSP for the DCI box which they can do 250 gig per wavelength. I think Cisco is probably going to revisit the decision of using internal versus external when they go to 400G.
Troy Jensen: Andrew, that was fantastic. I really appreciate the time and your insight’s great. I personally think we’re going to have a greater earning season here. There’s been just too much healthy demand and high margin business, and we’re going to see ROADM upside and healthy margin expansion, and it still feels like it’s early in the stuff. Give us a couple of weeks and you will start to see some earnings results if there’s not free announcements. Thank you so much for your time and we’ll be in touch.