Andrew Schmitt provided key OFC 2017 takeaways during an investor call hosted by Troy Jensen of Piper Jaffray. Sixty-seven investors participated. Topics of interest included:
- Impact of Ciena’s Wavelogic AI DSP Licensing
- Outlook for Coherent Port Shipments in China During 2017
- Market Outlook for CFP2-ACO/DCO and QSFP28 Markets
- Potential Vertical Integration by Intel or InPhi
- Timing of Metro Coherent and ROADM deployment in China
- Chinese SARFT Capex Guidance
- Roadmap for 400G and 600G Coherent
- Observations on Infinera, Ciena, Acacia, Oclaro, and Neophotonics
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.
Summary
- Ciena DSP licensing announcement was the big surprise but its too early to determine impact. It’s the first move in a long chess game that will be played.
- 30% port shipment growth in China during 2017 remains most likely outcome with situation becoming clearer over next few weeks.
- CFP2-ACO market is ramping, led by Cisco, with a total of 35k-40k units likely for 2017.
- CFP2 DCO market exists but not big for 2017, China transition at end of 2017 is a big opportunity.
- InPhi needs to become vertically integrated to get return on DSP investments.
- Coherent 600G announcements are interesting but 400G will be a the next stepping point.
- QSFP28 market is not near an overcapacity situation.
Ciena DSP Market Impact
Troy Jensen: As you know, we’ve got Andrew Schmitt here that’s going to give us our takeaways from the OFC Conference. Andrew’s still out in LA, sounds like he’s wrapped up three consecutive days of back-to-back-to-back meetings. So, lots of good content to go through. But thank you to Andrew, in advance for doing this.
Andrew Schmitt: Sure.
Troy Jensen: I would love to get your quick thoughts on some OFC highlights. How about just some quick takeaways to start?
Andrew Schmitt: I don’t think there was any real one big thing. I mean, every show has a blockbuster announcement. Last year, it was the InPhi ColorZ announcement. I think this year, the big surprise announcement was what Ciena was talking about, licensing out to DSP. We can drill into that in detail. It certainly was a big deal in terms of being a surprise, but I think it’s one of those things where it’s a chess move where it’s not apparent how it’s going to work out until some more time passes, some more moves take place. It’s certainly a very big deal, but I think it’s a little too early to call on whether it’s going to be very positive or not.
The one thing I do see with that announcement is there’s not a whole lot of downside. The people who … Lumentum, Neo and Oclaro, who have licensed the DSP from Ciena, they didn’t have pay any NRE, and I think that Ciena is working with them to make sure that they can get to market. There’s not a whole lot of cost on them, and so they get a free call option on the success of that product.
For Ciena, if it doesn’t work out, well then, there’s nothing lost either. It’s all potential upside but the magnitude of that upside at this point I think is difficult to figure out for some reasons.
Troy Jensen: Right. I had meetings with all three of the partners. All four with Ciena included, it sounds like it’s still very, early days, so it’s not going to be until the end of ’18 that they potentially get revenues from it. Everything I’ve heard about the DSP from Ciena is its world class, so its a very good product. Correct me if I’m wrong, but isn’t it starting off as a 5×7 form factor versus CFP?
Andrew Schmitt: It is. It’s a 5×7 form factor. One of Ciena’s goals was to try to get into China, and I think ideally if you want to get into China, as we’ve talked about on these calls, a large portion of what’s happening there is moving towards DCO. So, as a 5×7, it really cannot participate in that trend.
It can, however, participate in the long-haul portions of the network there. That’s where they have a shot, starting in 2018, to try and take some of the business … I think realistically, best case is ZTE. I have a hard time seeing Huawei migrate, but they can take some business potentially at ZTE in their long-haul network with that 5×7.
What it boils down to is they’re not going to be sampling this product until the first … I think I heard different dates, but the number that I’m going with is the beginning of 2018 when Oclaro, Lumentum, and Neo will have modules. In that time frame, there’s going to be other vendors such as Acacia, Neo, and InPhi that will all have 400G solutions.
I think in a parity situation, Ciena is going to lose every time because nobody is going to go with the Ciena DSP if they can use a more neutral supplier. If there’s-
Troy Jensen: That was a question in every meeting too, how the hell are you going to convince ZTE to put a Ciena DSP into their networking equipment?
Andrew Schmitt: That’s another concern that is yet to be resolved, but I think if ZTE has been relying on Acacia and their 400G DSP and for some reason that thing is not working, which is hard to believe because Acacia is the number one guy in terms of execution, well then, ZTE suddenly has other options that they can go and look at.
The other thing when I had meetings with Lumentum, Neo and Oclaro, and in all three of them, each one said that it’s not about the first release. This is an ongoing effort. I don’t even think that their expectations for this product are huge, but what they see this as is a way to get access to DSP technology that isn’t locked in at Acacia or InPhi. That’s whole other discussion, but I think ultimately, I’d be very worried if I was an optical component vendor outsourcing my DSP from InPhi.
Ultimately, I’d be very worried if I was an optical component vendor outsourcing my DSP from InPhi.
Troy Jensen: Fair. I think the other … Lost my train of thought. Sticking on the ACOs, DCOs would love to get your thoughts on the status of that market, your forecast for the space and maybe any competition. Start with the ACO, please, first.
CFP2-ACO Market Ramp
Andrew Schmitt: Sure. On ACO, that market is ramping hard. From what I’ve heard Cisco is inhaling product to support shipments of the NCS 1000. They’re also shipping in the NCS 4000 for Verizon, but the lion share of that product is going out in their DCI box. They’ve had some real success with that platform. I don’t know where that success is but it’s almost certainly one of the big Webcos out there, but that’s certainly creating a lot of demand for the ACO.
I think overall for the whole market this year, the number that I’ve been using is around 35,000 to 40,000 units. Cisco realistically could be at least half of that market.
Troy Jensen: As far as Oclaro and how they’ve could lock up contracts, have you heard anything as far what percentage of Cisco they’ve captured?
Andrew Schmitt: No, I don’t have any … I mean, I don’t have any visibility into the commercial side of things, but what I have heard is that on a technical basis, their product has set the bar very, very high.
There’s now a lot of companies out there with ACOs, for example, Coriant is making their own ACO, there’s Acacia, which is now going into production, Finisar is there with a decent product, and I’m probably forgetting somebody but there’s a bunch of suppliers. But if you can’t meet the performance specs that Oclaro has set then you don’t have a product.
You may have an ACO, but this isn’t like the client business where every single QSFP is all the same as a commodity. The ACOs are very performance … The performance of those varies from vendor to vendor. So, just because people have an ACO and its released to production, it doesn’t necessarily mean that they’re going to get a seat at the table.
Troy Jensen: Right. That’s fair. How about thoughts on Lumentum? Are they anywhere with their integrated chip ACO?
Andrew Schmitt: Yes, they had some stuff … I didn’t go into their back room, but I heard that they had some stuff in their back room that they’re getting close to releasing their ACO, but they haven’t provided any details on that publicly.
Troy Jensen: Have you heard whether or not it’s going to be able to scale up to 200G, 250G?
Andrew Schmitt: If they can’t do 200G, 250G then they don’t have a program. I don’t know that but knowing Lumentum, they wouldn’t be working on this today if it couldn’t support that. That’s my opinion.
Coherent DSP Market
Troy Jensen: Right, right. If you think about Acacia being vertically integrated with its DSP for its DCO, can these other guys compete with them on pricing with the Ciena DSP? We haven’t even heard of what pricing is yet but-
Andrew Schmitt: Potentially, sure because Ciena might be willing to sell that DSP for a pretty low price. It’s all upside for them. Ciena’s interest … I mean, if you think about why did they do this, they didn’t do this to make money, I don’t think. I think what they wanted to do is get the technology out into the market and put it up to the Pepsi challenge and find out exactly how good this stuff is. Internally at Ciena, you have the engineering teams, and they need to write a really big check to continue to be in this business, and so somebody says, “Why are we in this business? This is really expensive.” Then, the subordinate says, “Because our stuff is best in class, world-class.” Leader says, “Well if that’s the case, let’s get it out into the market and let’s put an alternative source of supply and let’s make sure that there is a diverse market for DSPs.” If in fact, this thing is world-class, then you’ll see some commercial success, but if they don’t meet with any commercial success, I think that calls into question whether Ciena should continue to develop their own DSPs.
I think what they wanted to do is get the technology out into the market and put it up to the Pepsi challenge and find out exactly how good this stuff is.
Internally at Ciena, you have the engineering teams, and they need to go and write a really big check to continue to be in this business, and so somebody says, “Why are we in this business? This is expensive.” Then, the subordinate says, “Because our stuff is best in class, world-class.” Leader says, “Well if that’s the case, let’s get it out into the market and let’s put an alternative source of supply and let’s make sure that there is a diverse market for DSPs.” If in fact, this thing is world-class, then you’ll see some commercial success, but if they don’t meet with any commercial success, I think that calls into question whether Ciena should continue to develop their own DSPs.
Troy Jensen: I get it, good point.
Andrew Schmitt: From a pricing perspective, I would think that that’s an easy thing to solve because it’s in the best interest of Ciena and it’s in the best interest of Oclaro, Neo, and Lumentum to make sure that there’s an alternative source of supply. Those guys are vertically integrated except for the DSP, and I’m sure that Ciena would be willing to help them along with pricing.
Troy Jensen: Right, I do too. I think Ciena wants to do this for a profit and thinks it’s going to be marginally creative, to think about where the gross margins are now at the mid-40s, if you can start to layer on just pure DSP sales on top of that I guess it would be-
Andrew Schmitt: Just about anybody in the world can make this margin accretive. It doesn’t require a whole lot of skill. It’s like printing money.
Troy Jensen: Okay, that’s fair. How about just DCO market, while we’re on the topic?
CFP2-DCO Market
Andrew Schmitt: Sure. The way to think about the DCO market is for some equipment vendors who are willing to accept a compromise in performance; they get all of the benefits of pluggability that DCO brings.
In the near term, Acacia has their 200G DCO … Feedback from people I spoke to is that it’s doing well, and they’re very positive on it, but it’s not clear to me that there’s a huge market for that yet. I think if you look at the Webco guys who are trying to optimize on a cost and performance basis, they’ve deployed dedicated DCI platforms, none of them use DCO. The market for DCO is the IP router guys and some of the lower volume situations where people want to have coherent interfaces on equipment they want more of the convenience and not necessarily the high density and high performance they can get from dedicated or ACO based coherent solutions.
The real question is, what’s going to happen in China when they migrate from CFP to CFP2 and whether they go to DCO or ACO? That’s the ultimate opportunity for the CFP2-DCO – what’s going to happen in China. That transition will take place around the end of this year, and at that point, there’s also going to be DSPs in place from both NEL, who’s already sampling theirs, and Clariphy, which I think is going to be sampling shortly.
Troy Jensen: It sounds like they are soon. As far as the timeline or status of the CFP2-DCOs from Acacia, they’re sampling now? Is that your best understanding?
Andrew Schmitt: Just talking with people, my feeling was that it was baked. It was ready to go. But I didn’t drill into that in a big way. I wouldn’t take that information to the bank.
China 2017 Forecast
Troy Jensen: Okay. All right, perfect. How about the topic du jour in Wall Street is China, I’m curious to know if you were able to glean anything about China tenders or statuses, some of these Chinese network upgrades?
Andrew Schmitt: Yeah, that was funny. I was at dinner on Wednesday night with a couple of the Huawei guys, and I asked them some questions about this, and the look on their faces was that they were just so sick of everyone asking them that question. The guy who runs the optical group at Huawei, I think he feels like he’s been felt up like 100 times by folks trying to figure out what’s going on.
I think it boils down to this- if everything happens as they expect it to happen, you’re going to see 30% port growth in China this year. Of the stuff that gets shipped, around a third of that is going to be DCO, which is going to be built internally at Huawei.
The only thing is there’s some activity in the next couple of weeks and, unfortunately, I don’t know the details of it. Some people call it phase 12. Other people, when you talk about phase 12 they don’t even know what that is, but there are some decisions being made that will lock in place what the demand is going to look like this year, and the numbers that I gave earlier are the expected outcome.
When I was talking to someone from Huawei, there was a guy there from China Mobile and he (Huawei employee) was frustrated from being asked the same question over and over so he just pointed his hand over to the China Mobile guy and said, “Ask him.” He (China Mobile) said, “Well, we’re in the process of finalizing our decisions.” So, something is happening in the next couple of weeks, but I wish I could say – this is the meeting, this is where it’s going to be held, this is the day, but I don’t have that.
Troy Jensen: That’s consistent. We’ve been hearing early April is when the tenders, phase 12 tenders, will be made public and then there will be a period before those tenders get to awards. I guess, my concern is … I don’t know the lead time between tenders and awards, but you still feel like the industry can shift 30% port growth to China if its back end loaded?
Andrew Schmitt: Yes. First, I’m not sure exactly how back-end loaded it is because from what I understand there was some inventory overhang going into Q1, so they may be building and shipping right now. They’re just doing it from stuff that they had on the shelves. I don’t know that … when you say back end loaded, it’s going to be that heavily skewed.
Troy Jensen: Okay. At a minimum, it feels like there are not many shipments into China for 100G and Q1, so it’s just three quarters to do it. That’s what I mean by back-end loaded-
Andrew Schmitt: Well, the component shipments. Component shipments. That doesn’t mean that Huawei is not making and shipping.
Troy Jensen: Sure, sure, they can be burning up their inventory that they have.
China Component Market
Andrew Schmitt: The other things that’s going on, and I think Greg (Oclaro CEO) talked about this at the dinner, is that there’s a change regarding the suppliers that are going into Huawei. I think to some degree that’s complicating the outlook on a vendor-by-vendor basis – there are some people that are emerging through this transition stronger than they were before.
Troy Jensen: Who do you think is emerging stronger?
Andrew Schmitt: I don’t know. Listening to Greg, it sounds as though they have benefited from this transition. I sat at the table with Tim Jenks from NeoPhotonics on Monday night at your dinner, and I asked him straight up, “As part of this transition to DCO, did NeoPhotonics benefit from a design win standpoint? Or, was it neutral or negative?” He said, “We benefited.” Everyone I’ve talked to has benefited, which I don’t think is possible but-
Troy Jensen: Exactly, that’s classical. Everybody denies they lost share but … If you look at Oclaro, it’s the only one guiding up, and they seem to be the most confident about China. What I’ve been told too is that they won share on the laser side, inside the CFP-DCOs.
…there’s a change regarding the suppliers that are going into Huawei … is there are some people that are emerging through this transition stronger than they were before.
Andrew Schmitt: I don’t know definitively who has won share, but the one thing that I do feel comfortable is there are going to be some winners and losers as part of this transition. That’s inevitable.
Troy Jensen: Let me ask, I just got an inbound question on this, how about as far as drivers? Have you heard any share shifts between Qorvo, MACOM, and InPhi at Huawei?
Andrew Schmitt: No. I mean, if I were a betting man I would think that InPhi is probably doing very well at Huawei right now because if you recall when they announced the Clariphy acquisition, they mentioned that this was the behest of a very large tier one customer. Do you remember that?
Troy Jensen: Yes, yes.
Andrew Schmitt: That tier one customer, if I had to guess logically, it’s got to be Huawei because Huawei was a major partner for Clariphy. That’s my understanding. As part of this deal, I would imagine InPhi probably benefits just in selling some of the surrounding components into this DCO. I don’t know that but-
Troy Jensen: I agree.
Andrew Schmitt: I can’t believe that they just did Huawei a favor by taking Clariphy under their wing and then Huawei is going to go and give all the driver business to somebody else. That doesn’t make any sense at all.
Troy Jensen: I’ve heard that speculation too, is that they pushed them to do it and they gave them share in some of the driver business to boot.
InPhi
Andrew Schmitt: I think the interesting thing about InPhi, and I was talking about this with some of the optical component companies, is before the acquisition was announced somebody asked me if someone like InPhi should buy Clariphy? My answer was no.
I still feel the same way that by having that asset … I mean, the amount of cash that it’s going to take to continue to be in the DSP business, especially to start going to things like 7 nanometers, the only way to get a return on that investment is by wrapping more stuff around it. Just wrapping the drivers around the DSP isn’t going to be enough. To get a return on the continued investment in the DSP business InPhi is going to have to get into the optical component business. It’s the only way.
When I talked about this at the dinner with investors at one of your events, they said, no way because the margins are lower. But I don’t see how they can justify continuing to invest in that DSP without also having the things to sell around it because that business is very, very hard. It’s a standalone business, and that is why Clariphy had to sell the company, they couldn’t get the return on the massive cash investment. Now that it’s inside a much larger company, that doesn’t change the fundamentals.
Troy Jensen: Who would be the logical buyer? Or, who would be the logical company that they acquire?
I think it isn’t just a little, tiny piece part type of acquisition, they need to be fully integrated end-to-end, and then InPhi is the Broadcom of optics
Andrew Schmitt: Well, what you want is somebody who had … I think you need a good business that’s doing very well and has all the components so that you can be vertically integrated end-to-end. When I think about really the premier companies there who don’t have some of the client business that would be Lumentum or Oclaro. I think it isn’t just a little, tiny piece part type of acquisition, they need to be fully integrated end-to-end, and then InPhi is the Broadcom of optics.
China Metro Upgrades
Troy Jensen: Interesting. I want to go back on a China question here. With China starting to do the metro upgrades if you think about this last cycle that began in November 2015 and went for four quarters, how long do you think it will take them to do a full-blown metro upgrade in phase 12? Can we have a timeline, or it this bigger more complex and more-
Andrew Schmitt: I don’t know what phase 12 is. A metro upgrade in China is multi-year … I don’t have an answer to that. It isn’t something that goes on for 18 months and then stops.
I think the real question is at some point the numbers are getting so big you cannot continue to grow 30% year over year, so eventually, you’re going to get to a point where you’re having flat to slightly down years, and I don’t know when that is.
It isn’t something that goes on for 18 months and then stops.
Troy Jensen: Okay. I’ve always thought that metro networks who were two and a half times the dollar spend. That’s what I’m trying to get to – is this magnitude bigger, ultimately in spending for China to-
Andrew Schmitt: I would say that’s a good guide. If you think about 2016, that spend was almost entirely long haul. The total spend could continue to grow, I’d say, over the next five years by a factor of … it could be 3x of what it was last year. (this was a mistake – Andrew meant port shipments, not revenue could be 3x what 2016 was.)
Troy Jensen: Perfect.
Andrew Schmitt: It’s like a 2021, 2022 that kind of target.
ROADMs in China
Then, don’t forget, you also must add in the ROADMs, which is a whole new thing for China. That was an area where I think things are starting to come together in China. I think the second half of this year, we should start to see some volume.
Troy Jensen: Perfect segue, that was next on my list, and I was getting ready to cue you up on that. What have you heard about timing? When do ROADMs get deployed in these networks once they start?
Andrew Schmitt: Well, like I said, I think it’s the second half of this year. Lumentum had an interesting announcement; they announced 1×9 WSS. I was talking to them and said, “Well, this doesn’t seem like a big deal. What’s the big deal about this announcement?” They said, “Well, it’s a cost-optimized WSS,” and I said, “Oh you mean it’s Chinese WSS?” They said, “Yes, exactly.” Lumentum has built this smaller WSS specifically for the Chinese market. I think they’re ready to go and deploy that.
The only area in China where I think there’s still some question is for the MCS. China Mobile is pretty committed to going to a full CDC architecture, which uses both the WSS and MCS. China Telecom, the research labs have qualified both colorless and directionless CD as well as the CDC contentionless architectures, but it isn’t as clear that the regional guys are going to deploy contentionless at China Telecom. But the ROADM business is starting to move at both of those carriers.
Troy Jensen: So, the ROADMs are a no-brainer everywhere. It’s just whether or not … Just so people on the call know too, contentionless means it has a multicast switch, correct, Andrew?
Andrew Schmitt: That’s correct.
Troy Jensen: Great. That’s going to be Lumentum and Neophotonics that dominate that space. Can you say for sure it’s China Mobile and the big carriers, and maybe not the rural networks have the contentionless?
Andrew Schmitt: I don’t have a good idea about the division between the rural networks and the core. My understanding is that China Mobile is heavily biased towards CDC, whereas at China Telecom it’s not as clear.
Troy Jensen: Why don’t we just take a pause right here, see if there’s anybody in the audience that wants to ask questions? I’m still going to dive into QSFP28 and some other stuff, but if there are any questions specifically on China and some of the stuff that’s been discussed, why don’t you just pull the audience quickly?
Q&A – InPhi
Operator: Your first question comes from the line of Scott Hogan from the PYDH Capital. Your line is open.
Scott Hogan: Hey, Andrew. This is Scott Hogan. We sat next to each other Monday. Just back to your insight commentary about them needing to vertically integrate, it’s my understanding that much of InPhi’s success has been from being Switzerland, so working with Oclaro and Lumentum and the platform guys like Huawei, they’ve got the TIAs going into the new DCO and all that. If they did vertically integrate with one of them like in Oclaro, wouldn’t that drive Finisar away or one of the other vendors away? Wouldn’t that drive other people away from them?
Andrew Schmitt: Absolutely.
Scott Hogan: Strategically, why do you think they would make that move if there would be that dis-synergy?
Andrew Schmitt: Because, I think ultimately, they’re facing pressure from one side which is coming from the switch of the DSP, and they’re facing pressure on the optical side. Increasingly, to get the better and better performance out of these optics, you need to have the full system approach.
Another way to put it is when you look at someone like Acacia, what if Acacia gets into the driver business? What if Oclaro gets into the driver business? Then, suddenly InPhi’s franchise is a lot less valuable. They’ve built a very, very strong technical expertise in this area, and it’s just my opinion, I can be wrong, but if their objective is to grow that business that’s the way to do it is to take that expertise and become vertically integrated like someone like Acacia.
If they weren’t in the DSP business, I would agree with you, but now that they’re into the DSP business, which is the most cash intensive portion of this whole supply chain, in order to get a return on that investment, I don’t know how they do that without being able to wrap a lot of other components around it the way that Ciena wraps sheet metal around their DSP or the way that Acacia wraps a sub-system around their DSP, especially as you start going to things like seven nanometers for 400G coherent. The investments become very, very large. Even if you’re getting 60% or 70% gross margins on those chips, the gross margin dollars and the magnitude of those gross margin dollars, it’s hard to get enough of them to cover the fixed investment costs of the DSP programs.
Scott Hogan: Okay. Thanks.
Troy Jensen: Andrew, I got a couple of emailed ones that I’ll rattle off now. Could you ask Andrew about MACOM’s acquisition of AMCC, and if what he said about InPhi, Clariphy applies to them as well?
Andrew Schmitt: That’s a good question. I think MACOM’s acquisition of AMCC, and I’m a little soft in this area, that was more for the OTN framer business and some of those components, I don’t think it was for the drivers. I think that that was already part of MACOM. That’s my understanding. I’m not sure that that’s related. MACOM has a lot of the optics that they picked up as part of an earlier acquisition, so I wouldn’t be surprised to see them move in this direction as well.
DSP Wins in China
Troy Jensen: All right, another question I got, let me find it here, I think I know the answer, but it says, do you think Clariphy has a good chance to get into Huawei for DSPs versus Acacia?
Andrew Schmitt: I think InPhi and Clariphy are already there. I mean, I think that’s why InPhi did the deal, and it’s just in the form of an ASIC, so it’s “Huawei’s Chip,” but my understanding is that Clariphy provides a lot of the key intellectual property for that chip.
Troy Jensen: Maybe to paraphrase what … I think what you’ve been saying is that Acacia could potentially get into Huawei because of their time to market advantage on CFP2-DCOs. Is that correct?
Andrew Schmitt: Yeah, and a lot of that is going to be dependent on what happens with NEL, which is the other key partner for Huawei. NEL has chips that are now coming out that should be able to do 200G DCOs just like Acacia is doing.
Acacia has a lot of very strong vertical integration to go with the DSP. They announced at the show this BGA-based photonic packaging approach, which is interesting. They’re doing a lot of different things right now that are very interesting.
400G and 600G Roadmap
Troy Jensen: Okay. Another question here, thoughts of Fujitsu’s 600G CFP2-ACO? The timing and when you think it’s available.
Andrew Schmitt: NeoPhotonics has also announced one, and some other vendors didn’t announce them that are working on them. I think that’s a real long, long-term product. I think there was a consensus that everyone is working towards 64 Gbaud 64 QAM. That’s the end game.
People that I spoke with at the component vendors said that there’s going to be a stopping point at 400G, 64 Gbaud 16 QAM or 56 Gbaud 32 QAM, or something because the challenges to getting to 600G are just huge, to make that work and make it deployable. There was a tremendous amount of difficulty just getting 250G to work at Cisco with the Oclaro products, and that was not easy, and so 600G is an order of magnitude more.
The timing on those ACO modules, I would expect to see people starting to use those maybe … I don’t know, mid next year, end of next year. But I would expect it’s probably going to be a 400G first, and 600G will be something that will come later.
Troy Jensen: Okay. I got one more question … There’s two more that came in, and then after this, I want to switch to QSFP28. Who poses the greatest risk to Oclaro and coming CFP-ACOs? Where is the ACO pricing heading?
Andrew Schmitt: Well, regarding pricing I don’t know where that’s heading. Obviously, it’s down, but it comes down to how many suppliers can go into Cisco and deliver a product with comparable performance. I don’t think there’s too many that can do that right now, so I wouldn’t expect that there’s going to be a huge pressure on pricing.
Regarding who that vendor is, my understanding is that the folks that are in the lead are Finisar. That’s what Greg (Oclaro CEO) said at the dinner on Monday, that was the first company that he named, but I’m not aware of who else is … The only thing that I’m troubled by is I just don’t know what happened to Fujitsu. They entered this market almost before anyone else, yet they’re still not here. It’s not clear to me what’s gone wrong with that company, and I wish I understood what happened.
Troy Jensen: Like you said earlier, it’s hard to ramp these things even once you’ve sampled.
Andrew Schmitt: My understanding is that before Oclaro made the decision to invest in all of the production capacity on the ACOs, is they found a way to make sure that they were back-stopped regarding delivering. I don’t know the details of that but, Greg has been adamant that they wouldn’t have gone forward with all this investment if they didn’t think that they were covered.
Troy Jensen: All right, fair. One more question and then we’re off to the data center stuff. What are your broad thoughts on Acacia now?
Andrew Schmitt: For the last couple of years I’ve been very focused on the importance of vertical integration by a lot of the OEMs is that someone like Ciena, Nokia, Infinera, they needed to own the DSP technology to be strong. But what’s happened in the last year has started to change that, you’ve got Clariphy moving under the wings of InPhi, which can make all the investments necessary, you’ve got NEL which is still in the game, and now you have Ciena which is throwing their hat into the ring for DSPs. There are other companies out there how are evaluating starting DSP developments specifically for the shorter reach 80km, 400G area. The number of people who are building DSPs and supplying them is increasing.
I think, Acacia, one of their challenges has always been when they wake up in the morning, 80% of their market was gone because those people are already doing their own DSPs. That’s why Acacia got into the ACO business, that’s why they started selling their own DSPs, is they had to change their business model to address these larger companies who didn’t necessarily want to buy a module and pay for the margin stacking that’s in that.
But now that there are other DSP vendors that are out there, there’s a lot more choice. I think in some ways, that’s better for Acacia’s business because now it’s more of a traditional component environment.
QSFP28 Market, PSM4 vs CWDM4, Intel
Troy Jensen: All right. Interesting. All right, so now shifting gears, as the market goes away from CFP, CFP2 towards QSFP28, who are the winners and losers?
Andrew Schmitt: Oh, boy. I don’t … Are you talking about people going 100G or did you say 200G?
Troy Jensen: No, no, no … Well, 100G. Just QSFP28 right now.
Andrew Schmitt: I don’t have a good idea of who are the winners and losers there. It appears there’s a lot of supply coming online. The one that everyone talks about is the potential for Intel to spoil the party if they come in. I’ve heard that they’ve had a very hard time getting this CWDM4 stuff to work in volume. I’ve also heard Luxtera has got some interesting stuff that they’re working on.
I don’t have an opinion on who the winners and losers are going to be QSFP28. I mean, as commodity products go that’s as commodity as they get. The winners and losers are going to be the ones that have the most volume, and have the technical benefits on the manufacturing side. I think those are just your traditional vendors that everybody knows. The one wild card is Intel and what they can make work.
Troy Jensen: Andrew, can you help us out for the QSFP28 market, how much of it currently is PSM versus LR4 versus CWDM? I mean, we just tier the biggest to the smallest market.
Andrew Schmitt: Troy, I don’t know. I don’t track that on a quantitative basis. I don’t think I can answer that.
Troy Jensen: Say on units, it’s got to be PSM on short reach, right? Because it’s in their data center stuff?
Andrew Schmitt: Like I said, I’m careful about trying to have an opinion on that because I just don’t know.
Troy Jensen: Okay. All right, how about on the silicon photonics stuff that you mentioned? I think Luxtera just does PSM for, I’m not totally sure of that, but how big of a risk is that then to these established guys?
Andrew Schmitt: Well, I think the risk is that if somebody has a breakthrough in terms of manufacturing costs and then they can bring the prices down … somebody has a breakthrough in manufacturing technology coupled with the ability to ramp it in volume; I think that’s where Intel is the one that everyone is watching because if they can do that, then that’s a … I don’t want to say risk because, at least from the equipment perspective, having more suppliers for the modules would be great, but it certainly would be a risk to a lot of the incoming guys and whatever pricing that they have.
Troy Jensen: On Intel, they do have PSM4 correct? You’re just saying that they’re struggling with the WDM?
Andrew Schmitt: I think everyone knows PSM4 … There’s not new PSM4 installs taking place now. All the cabling and things that were put in place … Almost now everyone moved to CWDM4. The PSM4 market isn’t going to get bigger.
Troy Jensen: I’ve heard that too. I had always heard that Intel was the one … excuse me, Microsoft was the one who had adopted PSM4 the most but moved away because the cable is so expensive.
Andrew Schmitt: Well, originally, Facebook was going to use it, and then they didn’t, and then Amazon was using it, and they started to move away from it. I don’t know Facebook’s status … I mean, I don’t know Microsoft’s status.
Troy Jensen: With PSM4, so if these silicon photonics guys are making PSM4 modules, that sounds like a dead end for them, right? They must get into DWDM.
Andrew Schmitt: Absolutely.
Troy Jensen: Intel has struggled with that. Have you heard where Luxtera is as far as CWDM4 development?
Andrew Schmitt: No.
Troy Jensen: Okay. That was short and sweet. LR4, how about the longer reach QSFP28 products?
Andrew Schmitt: Those are going to be cannibalizing the demand for CFP and CFP2, which goes back to some of the inventory stuff in China. I think Huawei is trying to transition to QSFP28 just like everyone else is because it’s a far more compelling solution for LR4 and below.
As that starts to come on and on in terms of demand, the CFP and CFP2, which was the only way to get LR4 until recently, the demand for those will cap out and decline.
Troy Jensen: Okay. Let me see here. I got another question from a client, who should Intel acquire?
Andrew Schmitt: I’ll tell you what Intel needs to buy, Intel should have bought Broadcom. They let that one get away. That was a mistake. That ship has sailed, but what Intel needs from a strategic … is they need an ethernet switch.
When Intel can start embedding those optics directly into the logic and cutting out all this stuff in between, that’s the holy grail of Silicon Photonics
There’s a bunch of these new Ethernet switch startups, and I don’t know who’s better than who, but that’s a business that Intel needs to be in. They have everything now around it, including the optics and, more importantly, the process technology to build those optics. But when they can start building optics right into an ethernet switch and how that’ll all be co-packaged, well then that’s when you need to start being worried about the QSFP28 business. But an Ethernet switch is the one piece that they’re missing if they can get all their optics stuff working. When they can start embedding those optics directly into the logic and cutting out all this stuff in between, that’s the holy grail of Silicon Photonics.
That’s the end game that Intel is working for in a way that some of these smaller guys like Luxterra and everyone else, they don’t have the size and scale to control the brains that these optics are going to need to be embedded into.
Troy Jensen: Fair. How about higher level on QSFP28? I mean, a lot of the guys that we talked to in the industry says its capacity constraint throughout the year and they’re adding 20%, 25% capacity on a sequential basis, do you agree with that? The size and sustainability of the demand right now for QSFP28?
Andrew Schmitt: I think in the short-term that’s right, but at some point, all this capacity is going to catch up to demand.
What do I know is that at the current pricing, the QSFP28 costs in very nicely. If you look at the reasons why you still see all this demand for 40G, it’s not because of pricing; it’s just because they can’t get 100G QSFP28s. At the price point that it’s at now, even when it’s supplied constrained, the economics of 100G work in the data center.
As more and more capacity comes online, it seems like we haven’t gotten close to that pricing elasticity threshold where … “Well, I bought 100,000 units at this price, but for me to want 500,000 units, it’s going to have to be half the price.” It doesn’t seem like we’re close to that.
I think the bigger risk is at some point all of the capacity is just get ahead of the natural demand. I’ve said this over and over again; the 40G market is going to be impacted because the reason that is continuing so strongly is because of the lack of product on the 100G side.
Troy Jensen: I agree with that, 40G stuck in much longer than we thought. I think all the shorts listening to this call are trying to figure out that point when the supply-demand and balance-
Andrew Schmitt: I don’t know the answer, and one of the things I said at the dinner is that figuring this out is next to impossible because there are so many moving parts, but I do know that at some point we’ll get there. It just doesn’t feel like it’s anytime soon.
Infinera
Troy Jensen: I’m just going to change gears a little bit. Andrew, I’ve always thought of you as my expert on Infinera, if you go back in time, you loved the company, about a year ago you turned negative on them and was dead-right, just your thoughts on Infinera currently.
Andrew Schmitt: The next quarter is crucial for that company… the future of that company is starting to turn on the events that’ll happen in the next three to six months. I don’t know what the outcome is going to be.
I spoke with a customer of Infinera, who is not a DCI customer more of a traditional system vendor, and they had not yet evaluated the GEN4 PIC base equipment in their labs. That makes sense because Infinera has been prioritizing the GEN4 PIC in their DCI interconnect system, CX2. It hasn’t quite made it to all the other systems. But that part must have performance and pricing that’s superior to a lot of these other solutions that are out there now like Cisco’s 250G NCS 1000, the new Waveserver from Ciena, and there’s a lot of these new systems out there who have pretty high density and pretty compelling solutions.
If the Infinera system is just at parity with those guys, they’re going to have some problems. If they’re able to do what they’ve done in the past and leap ahead of them, then this could be a huge turning point. I wish I knew the answer, but everything will become clear, and all will be revealed in the next three to six months.
Chinese SARFT Upgrades
Troy Jensen: Okay. All right, I got a few more questions I want to hammer out here. This is a good one that I meant to ask earlier. This week there was an announcement out of China, specifically the broadcast authority or I think the acronym is SARFT-
Andrew Schmitt: SARFT. The worst acronym ever.
Troy Jensen: It rolls off the tongue so easily. It sounds like the announced plans to invest 15 billion in network upgrades over a two-phase Capex plan. Can you explain this? Do you have any thoughts on this?
Andrew Schmitt: No, I haven’t been able to dig into that. I don’t know what the impact will be, but it will all go to Huawei, ZTE and Fiberhome.
Troy Jensen: But this is incremental, right? Everything we’ve heard of China is really for China Mobile, China Telecom, China Unicom. This would be a new China Capex spend?
Andrew Schmitt: Originally the cable system in China was a lot like the cable system in the United Sates 35 years ago. It was just basically mom and pop cable shops in the little neighborhoods that they are built in, and eventually these things were conglomerated and glued together. That process has been underway in China where you have a cable company that’s served only an apartment building, and so these things have been glued together, and now the government wants to put in place a second path to the home. Just like we have the phone company and have the cable company providing broadband to our homes, what China wants to do is make sure that the same situation exists there so that if you want broadband service, you don’t just have to buy it from China Telecom or Unicom, there’s another alternative.
That’s the overall mission, but regarding the magnitude and the timing, I don’t have any idea. This has been a very long, ongoing effort by the Chinese government and SARFT to ensure that there’s some competition at the broadband level for the consumer.
Troy Jensen: Yes, very good. Sounds incremental. I had two guys ask me questions about your view on AAOI.
Andrew Schmitt: I just don’t know those guys very well. I don’t want to pontificate because I don’t feel like I’m well informed about the company.
Troy Jensen: Okay, I knew that, but I thought I’d ask it. Another one it says, thoughts on what went wrong at Neo, aside from China.
Andrew Schmitt: Don’t know. Ask Tim.
3D Sensing Market
Troy Jensen: All right, here is another one … Here someone asked me this question, I know the answer, but I just wanted everyone to hear it, your thoughts on 3D sensing.
Andrew Schmitt: Well, it’s outside of my strike zone. I think everybody just loves the idea of getting into the iPhone. I’m assuming that – no one ever says that word. I think I can say that word because I’m not selling anything to Apple. But it’s also been hinted out that obviously, it’s not just them, it’s every other smartphone manufacturer in the Android ecosystem.
It’s a tool that once it’s put in place, all the software guys are going to figure out all kinds of cool things to do with it. It has the potential to be a part of the phone in the same way that the camera is a part of the phone that is now something that you can never take away.
To me, it seems like a huge opportunity. The trick is to avoid being in what I like to call the PortalPlayer situation. I don’t know if you remember a company called PortalPlayer, but they were the ones that were supplying the chip for the original Apple iPod and had a great business. Then, one day Apple decided to do their own chips, and then it was game over. The company just vaporized.
On 3D sensing, the trick is this is so important that a company like Apple might want to vertically integrate this kind of technology. That I don’t know. Everyone is so excited about 3D sensing, and they’re right to be excited, but if you wanted to have something to worry about, that’s what I would worry about.
Troy Jensen: That’s very interesting because Princeton Optronics just got acquired … I mean Apple could do that, just buy a design shop and create your own VCSELs and use some outsourced manufacturing, but it’s at risk in the future.
CFP to QSFP28 Transition
A couple more I want to hammer out here quick, Andrew. This is a good one. The transition to QSFP28, the CWDM4, and LR4, have a negative impact on light and Finisar give the price disparity between CFP and QSFP28.
Andrew Schmitt: I don’t think so because what happens is it’s not that CFP goes away, is it just stops growing because those are very Telecom focused types of formats. It isn’t like in the data center where things ramp up fast, and then they disappear very fast. It’s a much more gradual curve.
At a high level, I wouldn’t come to that conclusion that suddenly all of the CFP revenue is going to be displaced by QSFP28. I think that QSFP is additive to CFP and then the CFP decays. I don’t think that this is a bad event.
Troy Jensen: To me, this is just natural with optics, every form factor has a subsequent smaller form factor and-
Andrew Schmitt: Right. The one thing that’s always consistent in optics is these formats live a lot longer than anyone ever though they would, at least on the telecom side.
ADVA
Troy Jensen: I agree. Here’s another good one just came in, what happened to ADVA DCI? Microsoft?
Andrew Schmitt: Well, what I think happened, and then I’ve had somebody tell me I was wrong, so I’m worried about repeating it, but I think ADVA had some pretty significant business with the major North American Webco, and I’m not going to name who-
Troy Jensen: Amazon, I believe was the belief.
Andrew Schmitt: Well, that’s who I thought it was but somebody told me I was wrong, so I could be wrong on that, but it was definitely a North American Webco, and that business went away. The debate is whether it’s coming back or not, or if it’s gone for good.
The problem I have with the idea that it’s coming back is they were using a very old system. They were using the first-generation Acacia CFP and the FSP 3000 from ADVA, and there’s just a lot of other options out there now, including ADVA’s cloud connect which uses the Acacia 400g module, the 2×200, which is a great product. ADVA could get this business back, but I would expect it to return in a different system. But whenever there’s a transition from one system into another, it goes out to bid and then everybody has a shot at it. I don’t know whether ADVA is going to be able to bridge that business over to their new system or not. That’s a better question for them, but that’s ultimately what I think they need to do.
Transcript contains modification from audio to make it more readable and fix some annoying grammar.