Laser production for telecommunications products. A very small business in sensing, optical sensing. Making lasers for chemical sensing applications, LIDAR and other things like that. Now the sensing business impact seem not significant and revenue wise it is a fairly small part of our revenue, but the ultra narrow bandwidth laser that we use for active optics applications was actually developed for sensing for LIDAR applications. We do have a significant amount of R&D going into that area, and sometimes that R&D can transfer over into other parts of our business. In terms of overall growth expected in the data center market, again, you know, really extraordinary growth I think going on in this business being driven by AI. This data here is from Omdia.
This was published just back in January, so a few months ago. They're predicting a $55 billion market for transceivers of 100 Gb or greater in 2029. Again, you know, a level of growth this industry has never really experienced. Most of that growth is being driven by 800 Gb and 1.6 Tb products, as you can see the breakdown. As far as our manufacturing footprint, I touched on a little bit of this, but I'll spend a couple of minutes just going through that as well. Our headquarters and wafer fabrication facility is located in Houston, in Sugar Land, actually just outside of Houston. There we have about 350,000 sq ft currently.
We just signed a lease on another facility about 150,000 sq ft. That'll bring our total footprint in the Sugar Land area to around 500,000 sq ft currently. 210,000 of that square footage was recently leased back in the fall. That's in the process of being built out now, and it should begin production, not at full capacity, but should begin production in the third quarter of this year. In Taicang, we have 795,000 sq ft of space there. That's located in three different facilities that are fairly close to each other. Mainly we're manufacturing optical transceivers there. We don't do any of the wafer fabrication or anything like that.
It's optical transceiver and a little bit of our cable TV products are manufactured there as well. In China, a much larger facility, 1.2 million sq ft, mainly focused on transceiver and cable TV product manufacturing. We have a small office in Atlanta that's mainly for R&D in cable TV and some sales support and other functions related to our broadband access division that facility has. I alluded to this a moment ago, but last fall in October, we signed a lease for a 210,000 sq ft facility and announced a $150 million expansion project associated with building out a brand new manufacturing infrastructure within that building. That will be for the production of 800 Gb at 1.6 Tb products.
We're going to expand our wafer fabrication facility as well in both the current building that we're in and also a new facility that we hope to secure in the near future, to be able to ramp up the capacity for indium phosphide laser growth as well. On the right-hand side, you can see a picture of that facility that we leased. It's very close to the current building, a couple hundred yards away. Again, it's 210,000 sq ft. To put that in some perspective, the current manufacturing facility, the building that you see there on the left is our headquarters building. That's comprised of manufacturing space, R&D, you know, overhead functions. The actual manufacturing dedicated space within that building is 65,000 sq ft.
The 210,000 that we leased recently will entirely be manufacturing, and it'll expand our manufacturing footprint. Just that one building will expand our manufacturing footprint by about 4x. We're, you know, very excited to get this facility online. That will be coming on later on in the summer. I mentioned this at the outset, but a key part of our story is the fact that our process for manufacturing these transceivers is very highly automated. That gives us the ability to place that manufacturing capacity wherever it makes sense, both economically and geographically. It gives us the ability to scale our production, reproducibly and relatively quickly.
It gives us the ability because of the way that we've designed the automation, the platform is very flexible between 400 Gb, 800 Gb and 1.6 Tb. We can utilize all the same production techniques, all the same production equipment across multiple generations of products, which gives us the ability to continue to innovate and to scale that innovation into production very rapidly, which ultimately is what our customers are asking us for. This AI market is very dynamic, as all of you know. Being able to, you know, rapidly spin production of one product and then on a dime kind of shift production from one product to another as the needs of our customers demand is really, really important. That flexibility and scalability is critical to our customers, and that's what our automation enables us to do.
The foundation that we developed is not something that just happened overnight. We've been working on this for about 10 years. A lot of the equipment that we utilize to manufacture the products was designed and built by us, designed by our automation team and built by us. That gives us the ability to customize the production process for the design of the transceiver that we have. Vice versa, the transceiver design needs to be optimized to work with the automation the way that we have it designed. There's no silver bullet, if you will, where if you just have this machine, you can start popping out a bunch of transceivers. You have to design the equipment, and you have to design the product itself to be automated by that equipment.
We spent a lot of time and effort focused on not only developing the design for the module of the equipment to manufacture those modules. They go hand in hand, and as I mentioned earlier, they're flexible across multiple generations of products. That as we move from one generation to the next, we'll be able to reutilize a lot of that same equipment. The phases that you see here really have to do with the extent to which the process is automated. The first phase, we you know began automating individual workstations. The second phase, we started to put more of those workstations together so that they were, you know, a partially automated process. At Phase 3, that's a pretty fully automated process.
The difference between the light and the full Phase 3 has to do with the method of transporting the product from one manufacturing workstation to the next. In Phase 3, the full Phase 3, which we're developing and are rolling out in Taiwan now, we'll be rolling out in our Houston facility later this year. That one is fully end-to-end integrated. From the beginning to the end of the process, all of the transportation, all of the flow of the product from one workstation to the other is all entirely automated. Human beings aren't touching the product anymore at that point. That's what we'll see in the Sugar Land facility that I mentioned earlier that's looking developed and put in place in Taiwan.
I mentioned this earlier, but when it comes to automation, it's not just about having, you know, a particular piece of equipment or a particular design. It's really got, you know, a number of different aspects to it. Having the correct product design for the manufacturing process is very, very important. Being able to implement sensible automated inspection processes. Again, designing the product so that it can be tested and automatically inspected is very important because again, you know, you wanna have the ability to do this production process and test these products in a way that's not involving a lot of manual labor. The online inspection, the in-process material handling, that's the difference between essentially the Phase 3 light and the full Phase 3.
That is the transportation of the products from one workstation to another is something that we spent a lot of time developing as well. An efficient test system as well. The final test is obviously a very critical quality control step for us. It's also a step that allows us to enhance our ability to learn and improve the manufacturing process. If we can intelligently test the products at the end, we can learn from that, recycle those learnings back into improvements in the process as we go forward. That's been a very critical part of our optimization of the manufacturing process and improving our quality control. The vertical integration that I mentioned earlier gives us a number of advantages, including our ability to rapidly innovate for our customers.
You know, I've had a number of customer meetings here at OFC, and there's a lot of talk about, "Well, we really think we're gonna need this product, but we might need some variation of it. We might like it, you know, tweaked a little bit this way or that." Those kind of things are things that have been built into our process from day one. We build in a lot of flexibility into the manufacturing process and into the design so that as our customer needs evolve, we can evolve with it very quickly and we're not suffering, you know, a lot of extra design cycles and a lot of downtime and manufacturing ramp up schedules. The advantage that I mentioned earlier is 100% utilizing our own in-house manufactured lasers.
That gives us an advantage in terms of supply chain continuity, as well as cost and our ability to, you know, scale that capacity at will, essentially. I mean, we're not beholden to somebody else or, you know, worried about allocation or shortages or those kind of things. That's within our control. The PCBA and the rest of the manufacturing process, again, is also made in-house. Having all of these things together is really what drives our ability to flexibly and scalably manufacture these AI-related 800 Gb and 1.6 Tb transceivers. I'm gonna skip through this slide really quickly.
It's not been in the presentation for a while, but I would be remiss if I didn't point out that, out of the 17 major process steps that are involved in the manufacturing of these transceivers, 14 out of those 17 pieces of equipment were designed by our in-house engineering team. Again, this is not something that you can just go out and buy off-the-shelf equipment, regardless of the lead time, regardless of, you know, capital investments or whatever. You really have to be able to design this equipment yourself and design the product to go with it. It's a very special manufacturing process.
Relative to where we were a few years ago with our manufacturing process, you know, large numbers of human beings at workstations doing the manufacturing, the automated production process that we have gives us better traceability, manufacturing process because we're not beholden to so many human beings in the manufacturing process. It gives us the ability to, with AI, to learn from the manufacturing process as it's happening and adjust our process flow in real time to maintain not only high quality but also high throughput on the manufacturing process. The result of that is that we've decreased our labor hours by more than 95%, correction, 90%. We've reduced our manufacturing cycle time by more than 35%.
We've reduced our defective parts per million, DPM, to well under 50 for multi-lane, single mode 800 Gb transceivers, which I think is a very, very good number in terms of quality control metrics. Really some very tangible benefits to this automated manufacturing process in addition to the ability to scale rapidly for our customers. I'm gonna focus on this slide for a few minutes. You know, those of you who've been paying attention to our story for a while, we've talked a lot about our manufacturing expansion plans. You know, today I want to spend a couple of minutes really talking about what that manufacturing capacity ramp really looks like for us.
This is looking at a snapshot, you know, biannual snapshot at the middle of the year and the end of the year, both in 2026 and 2027 for a variety of different types of modules that we manufacture. What you'll see is that the 100 Gb /s optics, those are what have been, you know, driving our production volume for quite some time now. That's gonna be relatively static at about 140,000 pieces a month. Most of those are made in our Taiwan factory, and we're not really expanding the capacity for building 100 Gb/s optics beyond sort of where it's at today. Currently on the 400 Gb CWDM4 optics, single mode transceivers, we're manufacturing about 140,000 pieces a month there as well.
That's gonna go up by about two and a half times over the next year and a half as we get to the end of 2027. We're gonna grow from about 140,000 pieces a month to about 310,000 pieces per month by the end of 2027. On the Active Optical Cables, those of you who've been familiar with that story, you know, we've had a number of customer engagements on Active Optical Cables, including some very early automation in the U.S. related to the production of Active Optical Cables. We do anticipate growing that production volume slightly from where we are today, ultimately about doubling it by the end of next year. With some gradual growth between now and then, mostly focused in the early to mid part of 2027.
Some growth in Active Optical Cables module. On the 800 Gb and the 1.6 Tb, I think, is where the growth rate becomes really interesting, and this is related specifically to those AI applications. The 100 Gb, 400 Gb optics are mainly for general purpose compute infrastructure for our hyperscale customers. The 800 Gb and 1.6 Tb are focused squarely on AI related applications. You know, the middle part of this year we'll be able to produce about 150,000 combined products of 800 Gb and 1.6 Tb. By the end of this year, we expect that to be about 650,000 pieces.
By the end of 2027, growing to close to, you know, potentially around one million pieces a year of those two modules combined. That represents compared to the last report that we had about a 30% or so increase this year and almost doubling of the capacity that we anticipated next year. That's based on recent customer interactions that we've had and demand that we've seen coming from these customers. There is a very, very large demand for 800 Gb and 1.6 Tb optics. Then the final module that I've added to this slide from what we haven't really talked a whole lot about in the past is our external light source, our ELSFP. That's for co-packaged optics, CPO. This utilizes the ultra-narrow linewidth, high-power laser that came out a few months ago.
We have very limited production of those modules now, but we anticipate ramping production later this year and into 2027, ultimately culminating in about 400,000 pieces per month by the end of 2027. Our anticipation is that we're gonna be shipping mostly those ELSFP modules. We will be making the high power lasers for those modules in-house, but we don't anticipate selling those lasers externally. We're gonna use them pretty much for the in-house production of the ELSFP. This is what the capacity ramp looks like over the next year. We put out some guidance in the earnings call a few weeks ago of $1 billion in revenue this year.
These numbers here that you're seeing in terms of the capacity ramp will definitely support that $1 billion revenue number and then some. Obviously in 2027, you know, continuing to see the expansion around the production capacity for these various modules. In terms of our product roadmap, and you know, the way that this data center market will move to even higher speeds beyond 1.6 Tb, you can kind of see this broken down. We've broken it down into a couple of different ways of looking at it. Number one, by GPU generation or switch type and also by data rate, like we were talking about, 200 Gb coming out of 400 Gb per lane.
You can see that we have an evolution pathway going from our current generation of 1.6 Tb and 800 Gb to 3.2 Tb at 400 Gb per second per lane. Ultimately 6.4 Tb with our co-packaged optics, our near-packaged optics and the ELSFP module that we manufacture. Those things will be coming out later on in this year in terms of samples and production volume in limited production volumes by late this year or early 2027. I think that I'll probably cut it off after this slide because I wanna make sure that we have plenty of time for Q&A. The automation that we have can then be leveraged also across multiple different product areas as well.
I mentioned earlier the fact that our light on our lasers became very useful when we looked at the need for ultra-narrow linewidth, high-power lasers for CPO applications. In the same way, a lot of the automation that we're developing or have developed can be leveraged in other types of products as well, whether it's for our cable TV group or that CATV telecom group. That automation or the knowledge of how to automate, you know, complex production processes is I think something that's not limited just to the data center market. Of course, in this environment where data center needs to scale up very, very quickly in terms of capacity, it's very critical for us to be able to scale that manufacturing technology as well. There's a few slides here on financial performance.
I'm gonna skip through them relatively quickly because I do wanna get to Q&A. You know, year-over-year growth in revenue, if you look at 2025 compared to 2024, it's about 83% growth, again, largely driven by growth in data center, although we'll point out the cable TV group, you know, handily as well. Looking at the balance sheet, I'll kind of end it at this. We ended the end of the year with $260 million in cash. We have since raised another $250 million in ATM funding. Those of you that pay attention to our filings know that we increased that size. We were able to raise that $250 million in a couple of days.
We went ahead and upsized that to allow us to, you know, generate some more capital. In terms of debt, we're relatively low on debt right now. We have $67 million, sorry, in debt, excluding convertible debt that you're in. We do have $125 million to convert to our due in 2030. Excluding that, we have $67 million in debt. Moving forward, as we continue to look at options for financing the growth that we're experiencing, we do intend for a more optimal mix of debt and equity going forward. I don't anticipate leaning entirely on equity like we have been.
We're also looking at, you know, our customers for, you know, contributions there, whether it be in terms of prepayments on orders or, you know, other ways of helping to de-risk some of this expansion for us. We've had very positive reception on the possibility of doing those types of investments by our customers in the future. You know, looking forward to the future, again, equity will be likely a part of our financing plan, but it's not gonna be the entire answer as it has been so far. With that, I'll kind of leave this slide up here and go ahead and turn it over for Q&A. If you have a question, just raise your hand, and then somebody will come around and hand you a microphone.
Please, do speak in the microphone so those on the webcast can hear you.
Thanks. [audio distortion] , BMO. The 400 Gb and 800 Gb transceivers that you're delivering in Q1 and planning in Q2, for what percentage of those are you making the laser chips inside?
They're 100% made in-house today. Yeah.
And, um,
That can be put in the Q&A with management. Yeah.
To achieve the targets on the finance, how much more capital do you need from here?
Yeah, we're still working through the capital plans. Generally speaking, there's two facets to the capital expense plans that we have. There's the expansion of our wafer fabrication facility, and there's the expansion of our manufacturing capacity for the optical modules themselves. They don't necessarily scale at the same rate, right? Fab capacity for the lasers, you have to add a bunch of capacity in sort of larger increments. For the capacity expansion related to the manufacturing of the transceivers themselves, that you can scale at a smaller increment, right? We kind of look at, you know, 100,000 per units per month type of capacity additions. Depending on what type of product you're talking about there, it can be anywhere from just for the equipment to do the manufacturing of the transceivers.
That can be anywhere from $50 million to maybe $80 million per 100,000 per month. As far as, you know, the wafer fabrication, we're still evaluating, you know, what those costs are going to look like, how many of those lasers we're gonna need to manufacture. We're getting a lot of calls for, you know, increasing our capacity beyond what our internal use is as well. We're evaluating some of those needs as well. We'll have more to talk about in terms of the capital expansion plans, you know, going forward as we kind of flesh out those plans.
Hi, Christian from Craig-Hallum. I was wondering if you could clarify the manufacturing footprint. It looks like you have about 2 million sq ft, you know, in China and Taiwan.
Yeah.
You're gonna have 500,000 sq ft in Sugar Land.
Mm-hmm.
It seems like the maybe 50% of your transceiver manufacturing will be in Sugar Land. Can you just kind of square that with that statement?
Yeah. I mean, it's a little bit of a complicated picture. There's quite a bit of interdependence between our two Asian locations in terms of, you know, manufacturing over there. It's not like one place that's fully dedicated to one thing or another. The manufacturing space that we're talking about in Houston is solely dedicated to two things, the wafer fabrication and the chip manufacturing and the production of optical transceivers. Okay. In Asia, that's using our full Phase 3 automation capabilities. Okay. In Asia, some of the older stuff, particularly the 100 Gb and 400 Gb, still utilizes older, less fully automated manufacturing space. That's why it's, you know, square footage wise, it consumes a larger footprint in terms of square footage than what we have currently in the Phase 3 automation.
That's why on per square foot per unit basis, the U.S. is more optimized, if you will, more efficient in that sense.
Alex.
Sorry, Alex.
Thanks. Stefan, Glenn here. The ELSFP module, so that those modules each do multiple frequencies, so that means we're using wavelength division multiplexing to get out of the
Yeah. There's a couple of different versions of ELSFP, one using multiple different wavelengths and another using, you know, similar wavelengths on parallel optical fibers. Both of those. We're able to make both of those. The laser design is very similar. The frequencies aren't that far apart, so there's a little bit of optimization in terms of, you know, the design for particular wavelengths, but that's not a big deal for us. We can manufacture both.
Those are made as a pluggable module because people wanna be able to swap that out in case it fails?
Yeah, I think, you know, historically, the concern of customers about having the optics embedded directly in the switch pattern has to do with the relative reliability or at least the perceived relative reliability of the optics, the lasers in particular, versus the rest of the system, right? Lasers use a very high electric field, have very high, you know, optical density and get very hot locally. The reliability profile of the lasers tends to be less than some of the other components in the system. The thinking is if we can take the lasers outside, those can be swapped out as necessary while maintaining, you know, the integrity of the internal optical modulator switches inside, as well as the switch fabric and the rest of the infrastructure of the system.
I had a question also on the ELSFP.
Yeah.
Are you assembling the module or do you also have internal capabilities for the laser? I guess that's my first.
Both. Yeah. We have the internal capabilities for the laser. That's this 300 or 400 mW ultra narrow line width laser that's used in. There was something that we developed a number of years ago. That will be built in-house as well as assembling the module.
Okay. This is a high power CW laser that could be a second source to a Lumentum module, for example?
Yeah. I mean, it's our intention to manufacture those lasers in-house and anticipate utilizing, you know, Lumentum or somebody else. I mean, it certainly is possible, but, given the state of affairs right now, I think it's difficult to source those lasers externally, so we plan to make all those lasers.
Have you been qualified by some of the large CPU switch vendors on that product?
We're working on qualifications now. It's a little bit early. I mean, what we're hearing from most of our customers is they probably won't need, you know, big volumes of these products until sometime in 2027. It's a little bit early for those qualifications, but we are starting to work with them on what that qualification looks like.
Excellent. Then also in that slide, you had NPO and OBO, near-packaged on-board optics. Is that just related to that product or is that a different product as well? Are you gonna be like doing more level designing or what exactly is that?
Yeah. Right now we're looking at the ELSFP. That's the primary optical module we have. We also have a design for on-board optics. The internal modulator optics basically could be going inside the module. We have that as well for 6.4 Tb. I mean, it's under development. We have samples available now, but we'll have, you know, more wider product availability later in this year. Ultimately we will be making both the laser light source as well as the modulator, you know, mirror born optics that go on the inside of the module.
Thank you.
Stefan, thanks for hosting. Just a quick clarification on the DSP side. Do you take those on consignment your customers or you purchase and resell the DSPs?
Currently we're using DSPs that we procure ourselves. We have had a lot of discussions with various customers about their desire to possibly consign or somehow manage the supply of DSP. Not so much because individual companies and their supply chain can't manage that themselves, but they're a little concerned about being able to allocate correctly across multiple different vendors so that there's not, you know, multiple ordering, over-ordering of DSPs or sort of a land grab of trying to, you know, I should say, artificially follow the market by controlling the supply of DSPs. We're currently purchasing them, but that may not be the case forever.
Sure. Great. Okay. A quick follow-up. On the robotics equipment, I know from our business share line that you've talked about some lead times there. How are you feeling on visibility to procure the equipment needed to scale this on the robotics equipment side?
Yeah. So far the equipment's been coming in on schedule. This plan that, you know, we want to build today and we've talked about, you know, last year and kind of has been evolving for some time, this has been something that we've been talking to our customers about for quite a while. A lot of the key equipment. Some of the pieces of equipment that are necessary to build this can be long lead time, 12-18 months even. Those types of equipment that have those very long lead times, we've already ordered those equipment. Some of them—in some cases are starting to come in. In other cases, you know, they'll be expected to stay on schedule. So far equipment availability hasn't been a problem.
I will emphasize too, because a lot of this equipment is developed in-house, we do have it manufactured in many cases by a sort of a system integrator type partner. We're not competing with other transceiver companies, for example, for a lot of this equipment because, you know, we're buying common types of parts and having them integrated separately. We're not really in competition directly with a lot of the other companies that are doing similar things.
Clay.
Clay at Invesco. Stefan, if you think about, like when you get to scale for 1.6, call it fourth quarter 2027, how will the fully loaded cost of manufacturing, you know, in the U.S. compare to what you would do internally in Asia and also relative to competitors?
That's a great question. It's a harder question to answer relative to our competitors because again, I think the fact that we have this high degree of automation makes that a little bit more favorable for us than I suspect that it would be for others. For us, when you compare our cost to manufacture in the U.S. at scale, not yet, but at scale, we think it will be probably around 10%, maybe 15% more expensive to manufacture in the U.S. compared to what it would cost us to do that same manufacturing process in Asia at that same scale. Okay. Again, I obviously don't know the cost structure, you know, of our competitors as well as I know our own cost structure, so I can't completely comment on that.
I would suspect that the differential for a competitor who might be utilizing, you know, more of a manual manufacturing process is probably higher. It would probably cost $30, $40, $50 or something. I don't know the exact number, but that is one of the reasons why our customers have come to us is they correctly, in my estimation, judged that we were really the company that was most able to onshore or nearshore this production rapidly for them without incurring, you know, dramatic cost increases for them. So far, the reception to, you know, a modest cost increase for U.S.-based manufacturing has been very well received by the customers. We haven't really received significant pushback that, you know, "No, I'm gonna just pay a little bit more for these products." They like the security of the U.S.-based supply chain.
They like the vertical integration that we bring and our ability to scale manufacturing quickly using the automation that we have, and they're willing to pay, you know, a small cost add for that.
Alex?
That's price already after the tariffs, so that's 10%-15% increase including the tariffs.
That's including the tariffs. Yeah. I mean, the tariff situation, of course, is ever evolving, so it's a little hard to comment specifically on that. It's taking into account what we know the tariffs to be today. I mean, that may change in the future with increased tariffs.
If you're gonna make 400,000 of the CW modules for solid state products a month, it's a lot. It's really a lot.
Yeah. It's awesome and amazing.
How many wafers do you need to make for the lasers in order to do that? Is that ramp gonna be pretty difficult?
It's a big number. I won't disclose the exact number because, well, there's a number of proprietary things that go into that yield and wafer die size and such. But it's a pretty big number. It will require some substantial scale up in our manufacturing capacity in the fab. That's something that our customers have been very interested in us scaling up. I mean, the commentary that I'm getting from a lot of our customers is, look, you know, what is it gonna take to get you guys to scale up? We like your technology, we like the manufacturing capacity that you have, but we'd like it to be a lot bigger. That goes both on the wafer fabrication side as well as on the manufacturing, you know, assembly side for the transceivers.
It's definitely a large increase in the amount of wafers that will have to go through the fab.
I think that ELSFP module must be a pretty expensive thing. It's a lot of expensive stuff that's going into this. Will that be more than if you have three lasers now, and just to balance that three fibers coming out, is that gonna be more than like three 1.6 Tb /s ?
Likely. We're still evaluating, you know, pricing this out with the customer, so we haven't set pricing or anything on that yet. It's definitely gonna be a fairly expensive module. Again, that plus the modulator section that goes on the inside is replacing substantially all the pluggables that would go on front. It's still a more economical, you know, cost of ownership for our customers.
I would think maybe it's higher margin for you, too, because you're not, you don't buy a DSP and resell the DSP. It's entirely your technology instead of buying something expensive.
Margins on those parts we do expect to be substantially better than, you know, what you're seeing on pluggables.
Thank you very much.
We have a question in the back, Stefan.
Hi. [audio distortion] from Jefferies. Just a question on sort of how you think about sort of rapid capacity expansion in the industry which has a mixed history. In other words, I guess what gives you the confidence by the end of 2027 industry substantially oversupplied if there's not excess capacity in the market?
Yeah. That's a great question. We've had a lot of discussions with our customers. We are making these investments incrementally today, as our customers begin to place and contemplate longer term arrangements with us. We do expect that we'll have significant, let's say, agreements with customers about their long term needs and commitments to those long term needs, you know, before we make really the complete investment that we're talking about here. In other words, you know, unless we're voluntarily committed, we're not gonna build to this level necessarily until, you know, we see those commitments coming in. So far, the commentary that we've had is that they're very interested and excited about supporting this kind of ramp and they'll put their commitment behind it.
We have a question in the second row. Yeah.
[audio distortion] from Genstar Capital. I have two questions. One, if you could go to the next page, it's page 20.
You got it.
Stefan, if I'm reading this correctly, it looks like you're gonna ship 1.6 Tb ahead of 800 Gb and do it in Q2. Is that-
No. The 800 Gb, this is a newly developed version of our 800 Gb, so there is an 800 Gb version that we'll be shipping before that. The 1.6 Tb later this year, I think you're referring to sort of the second line, maybe the 1.6 Tb OSFP. That'll be shipping late Q2, early Q3 of this year. Yeah.
Got it.
We've already started shipping some 800 Gb. We're gonna continue to ship that 800 Gb. What we're showing here is a newer developed version of that.
Got it. Thanks. The other question I had is around, you know, the indium phosphide supply. You know, just can you discuss your current situation and, you know, obviously you're gonna ramp, you know, into the future.
Yes.
Talk about how you're gonna source that and secure it.
Yeah. I think you're referring mainly to the substrate availability, the indium phosphide substrate availability. That is something that we're, you know, monitoring. We have very good long-term relationships with many of the major indium phosphide suppliers. And you know, that's something that we have a good line of sight into how, you know, how to address any capacity issues that may arise in that indium phosphide. I mean, I can't say too much about it right now, but we're feeling fairly comfortable with our ability to address those supply constraints or perceived supply constraints.
It's worth mentioning too, I wanna say this, I think many of you in the room probably know this already, but the situation with indium phosphide, to the extent that there's a shortage or a perceived shortage, it's not so much that it's not like rare earths, which are, well, rare, right? Indium and phosphorus are some fairly common materials. What we're seeing right now is sort of a geopolitical question of China, you know, sort of artificially restricting certain exports and things like that. It's not that the industry capacity overall isn't or can't be big enough to do it, but it's gonna take a realignment of resources, because of you know some of the activities that China's doing in terms of regulating the export of those indium phosphide wafers.
It doesn't necessarily from the perspective of a company who's utilizing the indium phosphide, that doesn't necessarily make it better in the short term, but it does mean that it's a in some sense a more tractable problem to solve than something where you're dealing with a material that is fundamentally rare and difficult to come by, and difficult to process. That's not really the case with the indium phosphide. It's just the, you know, current situation is a little bit in flux, so we think it will be resolved.
Can you go back to the capacity chart?
Sure.
Yeah. So what do you think your CapEx will be for this year and to get to, you know, 2027, you know, next year? What would the split between wafer fab versus transceiver?
Yeah. We haven't given out the CapEx numbers just yet. We're still refining exactly. It has to do with the timing of when certain equipment is gonna come in and be placed in service. We don't have a precise number on that. I don't wanna give you any guidance on that just yet. We are working through that until some, you know, some more to say on that as we move forward.
I think earlier you said like $400,000 per month, you said something like $120 million. Do we multiply that by whatever the total is?
It's not uniform across all the different product lines. That was mainly focused on the 800 Gb type of production line, and that would account both for the capacity of the lasers as well as the capacity of manufacturing. Again, those don't scale at the same rate. That's kind of an average rate over, you know, a full production expansion. It doesn't necessarily mean that quarter by quarter you can kind of correlate that back to CapEx. It's a longer term average.
Go ahead.
Hi. Mike Kennedy, [audio distortion] . My question is on this slide, the capacity ramp, the Q2 2027, is that more or less how you built up to $378 million a month? You also had this new product that we were just discussing a couple of weeks ago. It seems like $378 million.
There's a little bit of upside to that.
I guess just I'm just quickly wondering, 10 years ago, what was the motivation to start deal automation and things?
That's a great question. It was not because we foresaw the explosion in AI demand and the geopolitical tension with China at that time. At about the same time we did a lot of our manufacturing in China, and we started to see, first of all, some competition, some pretty severe competition in price from Chinese suppliers. We also were experiencing pretty dramatic wage inflation in China at that time. We sort of reasoned that if we could begin to automate the process and, you know, control the rate of growth of our labor costs by automating more and more of it, that we would eventually be cost competitive with the Chinese suppliers who were relying on more labor intensive process.
I would say that was modestly successful. I mean, we were very successful in controlling the rate of growth of our labor costs. However, those costs themselves eventually kind of level out for everybody. I'm not sure that our differential advantage was as big as we had hoped it would be. What we did see, and what our customers more importantly saw, especially during COVID, was that as supply chains became snarled, particularly in Asia, really particularly in China, if you remember the lockdowns that occurred in China, really snarled supply chains there. Many of our large hyperscale customers noticed, and we certainly helped them to point out that while other customers were actually missing delivery schedules and slipping, you know, deliveries for them, we were able to deliver throughout the COVID period. We were able to deliver on time repeatedly.
That lesson didn't go unnoticed by our customers. They started to ask, you know, "Why are you able to make this work and the other guys are having issues?" When they really started to investigate it, we talked to them, it was largely because of the amount of automation that we have. It's because we relied on less people. All the people things that happened related to COVID, you know, weren't as severe for us. It wasn't that there were no problems at all. I mean, we certainly had our challenges as everybody. That particular challenge of having to manage a large labor force during a pandemic wasn't as acute for us as it was for others.
Our customers noticed that, and that's really when we started to get more interest in, okay, well, if that works for maintaining supply chain continuity in the face of the pandemic, can it also work for, you know, geopolitical supply chain risks or other things? That's where those conversations really started and culminated ultimately in OFC.
Go ahead.
Back to the technology roadmap slide.
Could you talk a little bit about your 400 Gb per lane, you know, the strategy there, the different technologies? We've seen a number of demos on, you know, differential MZM at 400 Gb per lane. You know, this morning there's been discussion of a view that the differential MZM at, you know, 400 Gb per lane might be the dominant technology. Could you talk about is that something you'd be able to do or just your view past, you know, 5 Gb there?
Differential MZM is something that we definitely can do. We've had some, you know, R&D work associated with that. We think we've got a lot of experience with lithium niobate, in particular thin film lithium niobate, from other applications in the past. We think that's a pretty suitable technology, and one that would be a hybrid type of integration, but I think it's something that we have a lot of experience doing. We think that makes a lot of sense. I've also mentioned, you know, the potential of indium phosphide modulators as well. That may be a part of this picture for us also. We certainly have a lot of experience with indium phosphide. We haven't committed specifically to one or the other of these technologies at this point.
It's very much, you know, kind of off in the future. We're evaluating these things and trying to see which one, you know, makes the most sense. From a manufacturing standpoint, we have experience with all of these technologies, and I think whatever ultimately proves to be the most suitable is the one we're gonna go to production.
We have another question, second row. Go ahead.
Yeah. Just, relationship to that, is that on the 2-in wafer at this point? When do you transition to three and four?
No, it's mostly 3-in today, going to 4-in. Our equipment, most of our equipment is capable of doing 6-in wafers as well. Not all of it is. There are a few pieces of equipment that we have to replace to go to 6-in. I don't see us necessarily going to 6-in for a while. Certainly not across the mainstream laser production. There may be some products that are in really high demand that, you know, 6-in might eventually make sense. I think we're gonna be primarily 4-in, and maybe some 3-in for the foreseeable future.
Scott.
Two questions. Number one, in terms of the fab expansion, is that something you can do with your existing footprint, or would that require a new site? I think you-
It'll likely require a larger real estate footprint as well. You know, we have line of sight into suitable facilities in the area that are available for us to acquire, you know, undertake that expansion.
Got it. Then secondly, in terms of your plans for automation and sort of onshoring, you talked about, you know, your customers. You have to get the customers on that for a long time. Some of your peers have adopted some of these co-investment models where access rights and so forth. You know, could you maybe talk about, you know, is that a potential option and how that would figure in your financing?
Sure. I think the question really has to do with, you know, to what extent, would customers be willing to, in some method, some way, you know, de-risk at least a portion of the investments that, we would be making in these technologies? The answer is yes. I think whether that's through some sort of co-investment, you know, prepayment of orders or, you know, significant long-term commitment or all of the above, you know, those are all options that I think customers are that are on the table with a lot of our customers. Certainly, some of the recent announcements in the industry of other companies that have had, you know, various creative, investment types of deals certainly increases, I think, the probability of getting those deals done with our customers as well.
Any other questions?
Yeah. Just wanted to get you clarification on just the capacity builds on page six.
Sorry. Yeah.
Yeah. Based on what I have here, I think you're sort of fully funded for everything sort of up till second quarter of 2027 except for ELS. Is that fair?
Approximately. I mean, you know, again, the schedule at which we're gonna undertake, particularly the wafer fab portion of the expansion, is a little bit up in the air in terms of how fast we're gonna do that right now. You know, put a precise timeframe on exactly how long the capital that we've raised or anticipate raising will last is a little hard to say. It's in that 2027 timeframe.
Based on the unit volume, the unit capacity that you've laid out here.
Yeah. Correct. Yeah. Again, when we actually decide to, for example, like if we're gonna scale up the production capacity in the fab for Q4 of 2027, we can't wait till Q4 of 2027 to do it because it takes a significant amount of time to qualify the new fabrication equipment and make sure that it's all there. We're gonna have to add that capacity well ahead of that Q4 of 2027 timeframe. How far in advance and when exactly that CapEx would need to be incurred is one of the things that's kind of a moving piece in the puzzle right now. That's why I'm not being more specific with the CapEx plans.
That's helpful. Just one last question on the ELSFP, the ultra-high power capital intensity, just kind of ballpark figures. What is that? I know it requires, you know, more wafer areas, but-
The high power ultra narrow linewidth laser is a larger, physically larger device size than, you know, our typical pump lasers for silicon photonics. You know, on a per die basis, it will require more fabrication. It'll be more fabrication intensive. You know, again, I don't have a precise number on that that I could share today. We're still undertaking, you know, kind of a study on exactly how we're gonna do that. For example, is it gonna be on 4-in? Is it gonna be on 3-in? Might it be on 6-in? I mentioned earlier that we don't really see 6-in, but it's still something that we're considering.
There's still a few, you know, when it comes to really the manufacturing method and how we're gonna scale that up. In particular on, you know, what size die and what's the capital intensity associated with that's something that we're working on putting final point on. I mean, it's worth mentioning it's 18 months or so away, so we've got a little bit of time to undertake that. I will anticipate having an answer for you at some point. We're just not going to be it.
Yeah. Eric Kirby, Raymond James. I wanted to clarify. Would the goal be to in-house all the EMLs and the CWs over time, ultimately 100%, if you could? If that is the goal, when do you think you may get there?
Yeah. That's a great question. From a risk standpoint, we don't like to be 100% insourced. Okay? It makes sense for us to have, you know, other sources where we can draw from for at least some portion of our production. I wouldn't say that that's our goal necessarily to be 100% insourced. On the other hand, given the current situation that we see in the industry with laser shortages and other things that being really acute, and especially, I don't think they're getting any better anytime soon. Based on, you know, what we're seeing and what we anticipate seeing in terms of the future capacity from, you know, our competitors, the market sort of writ large, I think we're gonna have to be 100% insourced.
That's not necessarily what I would optimally wanna do, but that's certainly the plan that we have is to be 100% insourced. When we talk about wafer fab capacity, square footage, capital intensity, all the things that you guys have been asking, when we start to communicate those plans, that's gonna be based on being 100% insourced, even though that may not necessarily be the exact optimal situation. Great question. With that, guys, I think we're pretty close to our time. I'd like to just wrap it up and say, you know, kind of when I started this presentation, it's a really, really exciting time to be in the optics industry. Fundamentally, the applications for the optics, interconnections in AI data centers are growing at a phenomenal rate.
Nobody, you know, I don't think anybody questions that at this point. We're very, very excited to be here, and we're very excited to have, you know, the capability to manufacture a lot of these devices in-house. We're excited to talk to you about our plans and looking forward to having the opportunity to speak with you in greater detail at a future date. Thank you all for coming today.