Jeff: Coherent is a vertically integrated outfit. We actually have three different divisions. One is materials, which we use for the growth of our lasers and the growth of the VCSELs, the growth of the long wavelength. They have about five different fabs for production of these parts. Mostly, we use the output from Zurich, Johannesburg, Sweden for some lasers, Sherman, Texas, and Fremont, California. We also have a large lasers. We do metal working lasers. That's completely separate from our particular area. But our group is basically involved in the building of two areas. Telecom, which involves things like ROADMs or the parts that go into ROADMs, items like amplifiers on the telecom side, and then digital coherent optics. And then on the datacom side, it's the transceivers, everything from 100 megabits all the way up to currently 800 Gigabits per second.
Ciara: Jeff, you mentioned 800G there, which is a really hot topic at the moment. Can you tell me what 800G applications the optics will be going into?
Jeff: There are two mains applications, depending on whether that's a "digital coherent optics", which is for long haul transmission, that's from data center to data center or else from in the telecom network.
Then the separate one is the transceivers for data communication, which are generally used in data centers, the large web hyperscaler type people, it will gradually migrate down into enterprise networks. The reason for that is because it's being driven by the adoption of AI, the GPUs coming from NVIDIA, and eventually from AMD and Intel. Those drive a tremendous bandwidth to go from one side of the equipment to the other and interconnect. So there's a tremendous amount of short distance applications inside a data center, whereas the DCO is from a data center or a site which has all this data to another. So, you know, could be like financial things, sending information under the Holland Tunnel or something in New York City.
Chris: Where we're seeing that is 800G for that Data Center to Data Center Interconnect (DCI) is what's currently under development. I mean, what we have today in production is that 400G DCO version. And we talked about the bundle that we're doing here with the EPS and Edgecore and IP Infusion, where we are bundling together our 400G DCO with a white box switch. And that can be used certainly for that DCI to DCI interconnect that Jeff talks about. But where we're really starting to see some more traction out in the market is in that CSP area, where now your telcos are going to start to look at these white box switches and starting to adopt that. There's some funding in the US known as the bead funding. A lot of that's being planned out. I think this year, we hope to see that go into actual volume production in 2026 through 2028. So that's I think a lot of growth is going to occur on the 400G side with what's in production today.
Ciara: We're seeing a lot of interest as well and in other areas in the world. And you mentioned the BEAD program in the US, there's a similar project gigabit in Britain and similar program in Germany. It is driving a lot of interest to connect that last the last mile, as they call it in the UK, or the underserved and rural areas. So all of this technology enables that, which is great.
Chris, you mentioned that the 400G ZR+ and in relation to the bundle that we did launch in the market, but also there is a lot of interest in 100G ZR+ at the moment as well. And can you talk about the key features and benefits of that technology, please?
Chris: Yeah, absolutely. So there's 2 parts coming out for the 100G ZR+. The one that kind of came to production first is based on an electrical interface known as CMOS. So that provides the full benefit of being able to interface with a tunable 100G optic. So right now in the in the market today, we have 10G tunable is what's the industry standard out there. And then with the DC 100G, we can kind of get to 100G tunable. And so that's going to really bring out that edge connection into these rural areas and allow us to to use existing equipment. But the existing equipment that's out there doesn't use CMOS electrical interfaces, they use the SFF 8636 from the QSFP form factor. And so that's the next product that's coming out. So that product actually is going into production now. And we expect that to have a little broader use case because it kind of goes into the existing network equipment.
Jeff: And the advantage of the 100G product is number one, we built where we build the DSP that's in it, the guts of the of the equipment. We also built the lasers, but the DSP is actually our design. The overall design is designed for low power consumption. So it's a maximum of 5.5 watts, which means that if you have a QSFP switch, it can fit into that slot without having to do tweaks or getting more electrical power in there or anything else. It can go into an existing QSFP port.
Ciara: There's a lot of talk about power consumption and carbon footprints of data centers. So I would imagine that the less amount of power needed will work towards people's sustainability agendas as well.
Jeff: It's true, but since you'd only be using a few of these modules for the DCOs, it's not a huge influence. But the power consumption in the rack and the provisioning of power to it. The QSFP was designed around a maximum 5.5 or something like that power consumption. This part can fit into it. Other, for example, our 400 gigabit DCO transceiver and most other people's uses something like 20 watts. So this is one quarter of the power consumption.
Ciara: Okay, that's a big USP then for the Coherent parts.
So what macro trends do you see shaping the future of optical networking and how is Coherent prepared to service or to meet those changes?
Jeff: One of the areas that we see as a possible growth area is we've got the DCO type transceiver. That does part of the systems. It gets rid of a whole need for mux ponders and those sorts of things. But we've also come out with a part called POLS, P-O-L-S, which is a pluggable optical line system. And that allows you to do amplification of that signal, bundling a whole bunch of signals together optically. So you can basically do the features of a really big box system in a part that's the size of a QSFP. And we think that that's a way to reduce the complexity of the overall transmission system, especially if you're going, say, 50-60 miles / 80-100 kilometers. The people who want to go from Boston to San Francisco, well, I don't know why they want to go from Boston to San Francisco, but if they want to go, you know, our parts, these parts are really designed to go shorter distances. So if you want to go Boston to New York or something like that, it will work fine for that. And there's a lot of customers who have those sorts of needs rather than trying to go across the world or whatever.
Ciara: I would think that, in addition to reducing the complexity of the transmission, you're also lowering the CapEx. You're saving on upfront costs because you don't need the additional hardware.
Jeff: Exactly. That's correct. The DCO provides that advantage by eliminating the need for dispersion compensation boxes, which significantly lowers the overall system cost. However, it is somewhat more challenging because it requires navigating through the disaggregation process and software, but the technology...
Ciara: That's what we're here for! We try and simplify that journey for our users.
Chris: There's no question, IP over DWDM (IPoDWDM) is a mega trend for our telco business. That's really where we're seeing a lot of effort into this whole bundled effort, this disaggregated network, as we've been talking about, this idea of having a white box switch that is going now into a telco operation is made possible because of these DCO optics.
Jeff: Then, in terms of the data center optics, the hyperscalers are already trying to drive, they'd like to get 6.4 Terabits out of a transceiver. So we have to do fundamental research. We're currently using 100 Gigabits per lane, per laser, to get up to 800Gs in a by 8 type configuration. We are currently starting to introduce 200 Gigabits per second per lane. That will get us to, say, a by 4 configuration, reducing the cost, but also allowing us to go to 1.6 Terabit. We are working on further generations from that. I'm actually unclear about how we get to 6.4 Terabits, but I think black magic is probably the way to solve all that.
Chris: Good thing you live in Salem!
Jeff: Yes, there's an unlimited need within the data center for bandwidth. I don't believe there's any indication that that's going to slow for the next 10 years.
Ciara: Chris, in closing, what do you value most about your partnership with EPS Global?
Chris: We really value our partnership with EPS. They're very accessible. Your team is, for a small team that you have, I mean, everybody's always very accessible, very knowledgeable. I believe that that differentiation of providing solutions to customers is what really helps enable our partnership to succeed the way that it does. And so that's why we really enjoy working with EPS.