Applied Optoelectronics, Inc. (AAOI) Earnings Call Transcript & Summary

Greetings, and welcome to the AAOI OFC Virtual Investor Tech Talk. [Operator Instructions] As a reminder this conference is being recorded. It is now my pleasure to introduce our moderator, Monica Gould, Investor Relations for AAOI. Thank you, Ms. Gould. You may begin.

Thank you, Diego. I'm Monica Gould, Applied Optoelectronics' Investor Relations, and I'm pleased to welcome you to AOI's OFC Virtual Investor Tech Call. This call is being recorded and webcast live. A link to the recording can be found on the Investor Relations section of the AOI website and will be archived for 1 year. Please note that there is a presentation that accompanies today's call on the Investor Relations section of our website. You can find the presentation at investors.ao-inc.com under the Events and Presentations section directly under the webcast link. Joining us on today's call are Dr. Thompson Lin, AOI's Founder, Chairman and CEO; and Dr. Stefan Murry, AOI's Chief Financial Officer and Chief Strategy Officer. There will be a question-and-answer session following our prepared remarks, where we will poll questions from the audience on the audio call as well as the webcast. You can submit a question by entering it into the chat window on the webcast screen. Before we begin, I would like to remind you to review AOI's safe harbor statement. On today's call, management will make forward-looking statements. These forward-looking statements involve risks and uncertainties as well as assumptions and current expectations, which could cause the company's actual results to differ materially from those anticipated in such forward-looking statements. In some cases, you can identify forward-looking statements by terminology such as believes, anticipates, estimates, intends, predicts, expects, plans, may, should, could, would, will, or think and by other similar expressions that convey uncertainty of future events or outcomes. Forward-looking statements also include statements regarding management's beliefs and expectations related to the expansion of the reach of our products into new markets and customer responses to our innovations as well as statements regarding the company's outlook for the first quarter of 2020. Except as required by law, we assume no obligation to update forward-looking statements for any reason after the date of this earnings call to conform to the statements to actual results or to changes in the company's expectations. More information about other risks that may impact the company's business are set forth in the Risk Factors section of the company's reports on file with the SEC, including the company's annual report on Form 10-K for the year ended December 31, 2019. Also, with the exception of revenue, all financials discussed today are on a non-GAAP basis unless specifically noted otherwise. Non-GAAP financial measures are not intended to be considered in isolation or as a substitute for results prepared in accordance with GAAP. A reconciliation between our GAAP and non-GAAP measures as well as discussions of why we present non-GAAP financial measures are included in our press -- in the slides available on our website. And with that, I'd like to turn the call over to Dr. Stefan Murry. Stefan?

Thank you, Monica, and thank you, everybody, for joining us today. In light of the situation with the virus, I'm really pleased that everybody can join us here virtually for this tech talk. Normally, we would have posted this, of course, at the end of the Optical Fiber Communications Conference, but thank you very much for joining us and allowing us to present to you today. So on the slide, as we go through, I'm going to spend the majority of the presentation this morning, which should last about 30 minutes, discussing AOI's technology, specifically the technology around our 400G products and touching a little bit on the 5G products as well. Before I do that, though, I do want to try to tie all these technological discussions that we'll have during the rest of the presentation into the investment highlights that you'll see here. So really, the technology for us is a building block that allows us to -- enables us to have these advantages that we've highlighted here, advanced optical technology, allowing us to serve these very dynamic markets, marquee customer base, which we've added to recently with a number of the design wins that we've had and the proprietary manufacturing that we bring to enable us to manufacture these products with the technology that we have as well. We'll talk a little bit about some of that proprietary manufacturing technologies in a couple of slides. Before we start to dig in too much to the technical content in the presentation, I did want to spend just a moment to give everybody a short update on the COVID-19 situation with AOI. I anticipate this is top of mind for most everybody on the call, as we've seen a number of companies reporting various issues. So we had our earnings call a couple of weeks ago and gave you all an update on where we were at, at that time. I just wanted to provide a little more detail on some updates, as you'll see here on this slide. So our China manufacturing operation, of course, was the first of our operations to be affected by the virus. We are not located in Hubei, near the epicenter of the virus. We're located some distance away from there. However, like many companies in China, we were shut down for a period of time by the government, following the initial discovery of the coronavirus. The shutdown lasted about 2.5 weeks, mostly in the month of February. Right now, we currently have about 80% of our employees back to our Ningbo factory. So I would say our manpower is slowly recovering. Every day, we get a few more people who are coming back to work. We still have no anticipated supply chain issues in Q1. As I noted on the earnings call a few weeks ago, at that time, we had built up a significant amount of inventory into Chinese New Year, which is typical for our operations. But that inventory will help us to recover operations more quickly than if we had to procure a bunch of inventory from suppliers in China. So we still do not anticipate any supply chain issues in Q1. We also have no issues so far with shipping product out of China. I know other companies, some of them have reported difficulty getting products shipped out of China. That has not been an issue for us so far. When it comes to risks and uncertainties, I think at this point, the main things that we're managing or working to manage here are the fact that the cycle time for the production of many of our products, especially our 100G products, is really making this quarter kind of a race to the finish. The Q1 revenue and margin are going to be difficult for us to predict until the last few days of the quarter because again, restarting manufacturing operations and getting that cycle time down to the point where it allows us to ship before the end of the quarter is the challenge that we have. Again, the more employees that we get back, the more production lines that we're able to bring up, the better that situation becomes. So we're working very hard to manage that at this point. We are also giving a priority to our data center deliveries. That is the deliveries of our 400G, 100G and other data center products. So our cable TV shipments are likely to be negatively impacted in Q1. This is somewhat expected because we do have typically a downward seasonality in the first quarter of the year. But we did want to highlight that there could be some differential impact on the cable TV relative to the data center products, just during this quarter. We don't think that's demand -- representative of demand. I mean, the orders are holding up more or less as we had expected. But because of the manufacturing emphasis that we're placing on the data center products, Q1 will probably be -- you'll probably see more decline in cable TV than we might otherwise expect. In Taiwan, not too much impact from the virus to our operations. We've added overtime shifts, and we've moved to a 24/7 production schedule to attempt to offset some of the loss of the China capacity. However, I would note that the wage manpower cost in Taiwan is more than it is in China. And the fact that we're doing overtime and other things also will increase our wage expense. So that will probably negatively impact the gross margin in this quarter and perhaps into Q2, depending on how fast the recovery in China goes. But we also have no expected issues related to supply chain in the quarter in Taiwan as well. Sort of a similar story in the U.S. Our operations here, we've added overtime shifts. We've added 24/7 production schedule, again, to partially offset the loss of capacity in China. We've taken back some of the manufacturing for certain laser products and other things that used to be done -- used to have been done in China and we're doing more of that in the U.S. as well, which will have similar impact to the gross margins, as I mentioned earlier, with Taiwan. Obviously, the labor rates here in the U.S. are more than they are in China. And so we'll see some negative impact on gross margin because of that. But also in the U.S., no expected supply chain issues in Q1 related to the virus. So that kind of gives you a snapshot of where we are with respect to the virus. I'm happy to take any questions on that later on as well. Before, again, we move to the technology, I want to kind of bookend the discussion of the technology with some general demand drivers that we see across our markets. As I mentioned earlier, at the outset, we're going to spend most of the time discussing our technology related to data center products, specifically our 400G and 100G products. But in the context of what's really driving the overall demand picture, data center is 1 of these 4 markets that we see. The data center demand right now, of course, is being driven primarily by 100-gig deployment. At this point, 100-gig and 40-gig for AOI are relatively close together, as we've noted in our last 2 earnings calls. We do see 100-gig continuing to grow, and we see the gradual decline of 40-gig business, as we've noted again in our previous remarks on the earnings call. That outlook hasn't changed. On the cable TV market, as I mentioned, broadly on the market, we do see a cyclical market that's currently at a relatively low point in its cycle. We do see -- we do expect to see improvement coming later in the year around new deployments that are either aimed at deploying DOCSIS 4.0 or in getting ready for future DOCSIS 4.0 deployments. We have started to see some new orders coming in for cable TV customers related to some of those upgrade projects. So we do have reasonable confidence that we'll start to see some improvement in the cable TV market in the back half of the year. On the telecommunications side of things, we have started to see some new orders related to 5G deployments for networks. Many of those are related to China customers. And so we believe that a lot of those are probably destined for deployment in China. Obviously, the coronavirus situation in China will probably affect those deployment schedules, although I will note that we've started to see some new orders coming in from customers that we haven't previously had a great deal of business with that we think are related to concerns over the supply chain of products that might be available from other suppliers. And so that's something where we're also working hard to meet those orders in addition to our sort of normal order backlog to try to make those new customers happy as well. And on the fiber-to-the-home side of things, no change here from where we were in the earnings call, just continued work being done on the technical side of things to prepare new products for the fiber-to-the-home deployments that we expect to come in future periods. The data center market as we begin to focus in, we noted on this slide, which hasn't changed from the previous presentation, but I did want to highlight for everybody who may not have been involved in some of our previous discussions that we are in a sort of inflection point now where we're starting to see the final stages of the qualification and the early ramp of 400-gig products. We expect 2020 to be a year where a lot of the decisions around 400-gig will get made in terms of what the supply chain for our customers is going to look like. We do not expect a huge ramp in 400-gig deployments this year. We think it's more likely to come at the end of the year or into 2021, and this chart here from Ovum Research kind of indicates that trend as well. However, at the same time, while we expect growth in our 400-gig products and indeed, in the 100-gig products as well, we expect to see continued sales of 40-gig and even 10-gig products. We expect those sales to diminish, as we've noted in previous remarks and also as illustrated here on the right hand of this slide, but we do continue to expect to see those products to sell. Customers still have needs for networks that are architected around some of these older products. And for that reason, we think these products are going to have a relatively long revenue tail over the next several years. So with that sort of market background, I wanted to spend the next 15 minutes or so really going a little deeper on the technology. So this talk was originally planned around the Optical Fiber Communications Conference. It is one of the largest technology conferences related to fiber optic communications. So I think it's appropriate for us to spend some time taking a little deeper dive into the technology. So the way I want to approach this is to start by talking a little bit about 400-gig in general, some of the standards that apply to 400-gig and then try to really focus that down into what technologies are available to enable 400-gig in the various types of reaches and other standards-based approaches and then really focus from there on the laser technology and some of the assembly technology, and how AOI is approaching this market relative to some of our competitors. And specifically, we're going to talk a little bit about AOI's discrete approach versus some of the silicon photonics approaches that are out there. So we'll spend a little bit of time trying to compare and contrast those 2 technical approaches. So that's the direction that we're going to head the rest of this afternoon. So to start out, I wanted to start out with a very simple, high-level taxonomy here of when we talk about 400-gig, what does that really mean? What -- first of all, it's worth noting that 400-gig is not one single product. Just like 100-gig and 40-gig before that, it's a family of products that are -- that differ primarily by reach, that is the distance that they're designed to go. And there also can be various different form factors or physical size packages that are used in various applications. So -- but the first order, you'll see, if you start to research this a little bit, you'll see mentions of things like 400G-FR4, for example, which I highlighted here. So the first group of digits there indicates the data rate, 400G, in this case, that's fairly obvious. The middle section there, the FR, that indicates what type of standard body promulgated that standard. So in the case of 400G, there's 2 principal standards bodies that are meaningful today. There's the IEEE, the Institute of Electrical and Electronic Engineers, and there is a Multiple Source Agreement group, an MSA group, that's founded by AOI and a number of other companies. I think there's about 20 or 21 companies at this point that are in this 100G Lambda MSA group. And this is an industry consortium that's gotten together to write standards around 400-gig and indeed, 100-gig as well, and I'll talk about that in a minute because they are related. So that's the -- so any standard that you see related to FR is coming out of this Lambda MSA. And if it says DR in the middle, that's the IEEE. And then the last digit there typically indicates the number of optical channels in either the transmit or receive side, so in this case, a 4 channel transmitter. So a 400G-FR4 would be a 400-gig transceiver that is meeting the Lambda MSA standards and has 4 optical channels, each of which is at a rate of 100 gigabits per second, okay? So that's just kind of level set to give everybody an idea of some of the various different types of standards that are out there. I'll talk about what those specific standards are in just a second. But I think it's illustrative to go through just a moment here why there are 2 standards, why do we have DR and FR? For those of you who've been around this industry for a while, you'll remember that at 100-gig and even 40-gig, we had parallel single-mode type optics and CWDM or Course Wavelength Division Multiplex optics. The main difference there, of course, is that parallel single-mode means that you have your however many optical channels you have, typically 4 optical channels at 100-gig. Those 4 optical channels would be carried on 4 separate optical fibers, and then there'd be an additional 4 fibers for the receive. So you'd have a total of 8 optical fibers for the parallel single-mode implementation. In the context of 400-gig, that same parallel single-mode type optic would apply to the DR modules. So the DR modules contain 8 fibers total, 4 for the transmit, 4 for the receive, and they're designed to operate over the same plant -- same type of fiber plant as the 100G PSM. So to first order, if you're a customer, a data center operator that's deployed a lot of 100-gig PSM, all things being equal, the easiest path for you to upgrade to 400-gig would be to go with a DR type -- DR type transmitter or transceiver. If you, on the other hand, have a lot of CWDM deployed in your networks for whatever reason, typically, that would be because you have a larger physical data center size. If your 100-gig and 40-gig infrastructure has been deployed largely over CWDM, then you might choose an FR module because the FR modules are, like the CWDM modules before them, are 4 different optical channels, 4 different wavelengths that are carried -- that are multiplexed together and carried on a single optical fiber. So -- and again, 1 fiber for the transmit and 1 fiber for the receive. So the FR is 2 fibers total for the 400-gig, and it's designed to operate on the same fiber plant as the 100-gig CWDM. So again, that's the primary difference between the DR and FR in terms of the implementation. DR is analogous to PSM, and FR is analogous to CWDM, okay? So as we go through some of the standards on the next few pages, just keep that in mind that, that's the principal difference and the principal reason why these standards exist. The next slide, I tried to break down this issue of the various different transceiver types in a couple of different ways. So the first way I looked at it is to map out for you the various different reaches or distances that the transceivers are designed and then try to map that on to either an IEEE standard, a Lambda MSA standard or a data center unique type of standard. There are some standards out there that are not necessarily standardized by an existing standards body, but customers have gotten together and said, "Well, this will be useful for us so even though it's not standardized by a standards body, we think we can use this." So for example, the 400-gig DR4-plus is not a -- at this time, it's not a recognized standard that's been published by a body, but there are customers that are interested in it. So it's a non-MSA type of application there. The ones with the asterisk, by the way, in this slide are ones that AOI is either currently producing or is sampling or intends to sample in the near future for customer use. So those are the ones that we're supporting at this point. And so you'll see the IEEE standards generally cover shorter distances, 500 meters or below, and the Lambda MSA standards generally cover the longer distances, from 2 kilometers to 10 kilometers. To kind of map this out in a little more detail, as we move down into more what we would call the physical layer, that is the lasers that are utilized in these modules, I try to take those standards then that you see on the previous page and give some context for the optical wavelengths that are used, whether it's one single optical wavelength or multiple wavelengths in the case of the FR standards. And then I've also indicated here the specification reference, the IEEE or Lambda MSA reference that would apply. That's mainly so that if you want to do some research off-line, you have the ability to go look at these standards and see exactly what applies to these different modules. So I guess the main thing to note here is, again, the DR standards, as I mentioned at a couple of slides ago, the DR standards are analogous to the 100-gig and 40-gig parallel single mode. So they are multiple different fibers in the case of 400-gig. The DR and FR standards also both have single 100-gig specifications that accompany that as well. So for example, a 100-gig DR is 1 channel, 1 fiber of that 400-gig DR4. And so you could have a 400-gig DR4 module on one side that breaks out into 4 100-gig DR modules. And you can do the same thing with the 400-gig FR4. If you'll note there, there's 4 different optical wavelength: 1271 nanometers, 1291, 1311 and 1331. Those can be mapped to the 100-gig FR specification as well. So you can have either -- any one of those optical wavelength could operate in the 100-gig FR mode as well. So again, the purpose of this is really just to begin to map the standards that we've talked about in the previous slide into physical devices, the lasers that enable those various different standards to be the physical layer that applies to those different standards. So then to go 1 step further, we -- I tried to break it down into the types of lasers themselves. So the previous slide indicated the wavelengths that would be used. As we talk in the next few slides, we'll talk a little bit about some of the differences between AOI's discrete approach and some of the silicon photonics approaches. And that -- in order for that to make sense, we kind of need to understand what types of lasers can be used in these various applications. So the shorter distance reaches the 70-meters and 100-meter type transmission distances. Those products are going to use vertical-cavity surface-emitting laser or VCSELs. Across the top of the slide, you'll see the various different laser types, right, the vertical-cavity surface-emitting laser, directly modulated laser, electro-absorption modulated laser and then silicon photonics. So the directly modulated laser can be used at distances from about 500 meters to 2 kilometers. AOI's demonstrated directly modulated laser technology that works all the way from below 500 meters to 2 kilometers. You can also use for those same applications, 500 meters to 2 kilometers, you can also use an electro-absorption modulated laser. And you can see the standards that could apply for the EML as well. In addition, the electro-absorption modulated laser can be used at 10-kilometer distance where the directly modulated laser, AOI believes, is not applicable to that 10-kilometer distance. And finally, on the far right corner there, far right column, rather, you can see the silicon photonics approaches. So you'll note here that the 2 single channel, the 100-gig DR and 100-gig FR, I've not indicated that those are appropriate for silicon photonics. You can, of course, use silicon photonics approach for those applications, but it really doesn't make sense because those are single-channel applications. And so it would be a cost-effective way to manufacture a single-channel module in most cases. But the silicon photonics approach certainly is applicable to the multiplex standards. So the multiple channel standards, the DR4 multiple 1310 and the FR4 CWDM-type approach. And also, I didn't talk about this in the previous slide, but there is some work being done to come up with a standard for 400-gig LR4. And we think the EML and the silicon photonics approach would both be applicable for that standard when it is ultimately published. So I mentioned a little bit about the -- these 3 approaches. Then excluding the vertical-cavity surfacing-emitting laser that's used in the short reach, if you kind of focus in on the intermediate and longer reach, which is AOI's primary market, the 3 approaches that I think are widely applicable across those standards are the directly modulated laser or DML, the electro-absorption modulated laser or EML and the silicon photonics approach, which is not strictly speaking, a laser, it's a laser combined with some modulation circuitry. So -- and the way to think about these 3 different types of optical module design is if you think about the transmit side of a transceiver, it has 2 principal functions: Number one, it has to generate light; and number two, it has to modulate or encode information on that light, okay? Those 2 functions can be combined into 1 device, for example, it will be directly modulated laser, or they can be -- the generation of light and the modulation of that light can be separated in 2 different devices. So for single-mode fiber optic communications, the generation of light is always done using an indium phosphide-based laser diode. So indium phosphide is a particular semiconductor. It happens to be from a class of semiconductors called III-V or compound semiconductors. And it forms the basis of pretty much all of the laser diodes that are used in single mode communication, whether it's a DML, an electro-absorption modulated laser or the laser portion of the silicon photonics module. All those are going to have lasers that are based on this indium phosphide chemistry. There have been -- just for completeness' sake, there have been some efforts to generate light using other materials, for example, silicon, but these really haven't been proven to be superior to indium phosphide. And in fact, they're not really commercially viable at this point. So a directly modulated laser is the simplest approach or the simplest conceptual approach, I think. Basically, this involves modulating the light intensity by varying the electrical current that's supplied to the laser. So you have your laser and if you apply more current to the laser, you get more optical power or, in some sense, a brighter light. So when you're dealing with on/off key, where you're talking about just simple 1s and 0s, if the on state or 1 bit would be transmitted by a bright flash, if you will, from the laser, the directly modulated laser, and a 0 bit would be represented by a dark or less bright time swatch, in this case, with the directly modulated laser, there's no need for an additional modulation device of any kind. You simply turn the current up and down to the laser, obviously very fast if you're doing this at 100 gigabits per second, but no modulator -- separate modulators required. The information is encoded just by that increasing and decreasing of the current that's applied to the laser diode. In contrast, both the electro-absorption modulated laser or EML and the silicon photonics solution separate those functions of generation and modulation of light. And the principal difference really between the electro-absorption modulated laser and the silicon photonics modulated device really has to do with where is that modulation done. In the case of the EML, the modulator is built into the indium phosphide chip. And in the case of silicon photonics, the modulation is done on the silicon chip, okay? That's -- in a lot of ways, in the interest of time, this is kind of a gross oversimplification of things, but that's basically the difference between silicon photonics and EML approach. Both are using -- both separate the generation and the modulation of light. And if you do the modulation in the indium phosphide chip right next to the laser, it's an EML. If you do it on the silicon photonics -- separate silicon photonics modulator, that would be appropriate for the silicon photonics approach. So graphically, this little cartoon is kind of what that looks like. So an electro-absorption modulator has 2 sections on the same indium phosphide semiconductor chip. So they're grown typically together, using the same epitaxial technique to grow both the laser section and the modulator section. And on the silicon photonics approach, you have a separate laser that's manufactured, again, indium phosphide-based, and that laser is bonded or somehow aligned with the silicon chip that would contain the modulator. So the modulator is made using a silicon fabrication technique. The laser is made based on indium phosphide and the appropriate processing techniques for indium phosphide, and the 2 of those are combined by somehow bonding them together or aligning them together on a separate substrate. And the first order, that's really the technical difference then between the electro-absorption modulated laser and the silicon photonics approach. The other important element that comes into play here when you're designing a transceiver. So you not only have your laser itself, but you have to modulate that laser. You have to be able to drive that modulator element, whether it's a directly modulated laser, in which case, you need to have an integrated circuit that can take our incoming 1 and 0 bit stream and translates that, if you will, into a current waveform that can be applied to the device to make the current go up and down and modulate the intensity. Or in the case of silicon photonics or electro-absorption modulated laser, you need some integrated circuit, some driver IC that turns that modulator on and off and actually encodes the information onto the light stream. So AOI believes that the lowest power consumption and the lowest cost can be achieved by utilizing a directly modulated laser approach. As I noted in the earlier slide, that's sort of the simplest approach. There's no separate modulator needed. There's no alignment procedures like there would be in silicon photonics. There's no separate indium phosphide section in the case of the electro-absorption modulated laser. It's simply taking the laser and modulating the current that goes into it to encode the information on there. AOI has demonstrated DML performance suitable for 100-gig per channel or 400-gig in aggregate. Distance is up to 2 kilometers, meeting the DR 4 and FR 4 specifications. So the problem -- the current limitation, I would say, on the directly modulated approach is that driver ICs, these high-speed PAM4 encoded directly modulated laser diode drivers, are not currently commercially available. We've demonstrated that the technology works. The manufacturer of driver ICs for this is nontrivial, and the driver ICs just aren't widely available yet. We're working with various IC vendors to develop these ICs. But when those are available, we believe that this approach with directly modulated laser will provide the lowest power consumption and the lowest cost. And those are the 2 main drivers that customers care about when it comes to selecting a technology for their 400-gig optical transceivers. So because the directly modulated laser diode drivers are not widely available for these speeds, the first generation of 400G modules are likely to utilize either silicon photonics or electro-absorption modulated lasers. Both of these technologies, silicon photonics and EML, have very similar -- the electronics, the driver electronics that work with either of these types of approaches, are relatively similar. They're both commercially available, and both the silicon photonics and EML approach can utilize these drivers. So there's no limitation, I would say, on the commercial availability of the driver electronics. And I would say that for Gen-1 devices, we've looked at the costs and compared the production costs. And I would say, in general, there's not a large cost difference one way or the other. Some modules -- the cost of the silicon photonics module and the cost of the EML type module are relatively close together. And again, it depends a little bit on the type of module. But there's not a major difference between the 2 production costs. And we think that, that scales pretty well. I would say this equation, the cost equation, is a little bit different for AOI because we can manufacture the EMLs in-house. On the silicon photonics side of things, we can also make the CW lasers that are used in silicon photonics. So for us, our cost structure probably looks a little bit different from some of our competitors. But if you really go down to the base level and say, "What do all these component costs to manufacture?" There's not a big difference between the cost of silicon photonics versus CML. And AOI has approaches that are based on both silicon photonics and EML for our 400G products, especially the Gen-1 products. But we do believe longer-term that the directly modulated laser approach is the least expensive, least complicated way to manufacture these modules. And when the driver electronics become available, we think that will be a very interesting opportunity for us moving forward. So I wanted to make one note before we move on and talk a little bit about 5G, as we're getting close to the time here. I did want to talk a little bit about the other -- besides the -- so the standards that I talked about earlier, the IEEE standards and the Lambda MSA standards, those specify optical interfaces and various higher-level performance specifications and that sort of thing. Overlaid on that is a number of physical form factors. So you'll see 4 of them mentioned here, the SFP-DD, the QSFP-DD, the OSFP and the COBO module. These modules can, in principle, incorporate any of the optical standards that I mentioned earlier, the Lambda MSA, the IEEE standards. You could have a 400-gig FR4 module in a QSFP-DD or you could have it in OSFP. You could have it in the COBO. It probably wouldn't fit in an SFP-DD. But the point is that there's -- the optical standards aren't necessarily directly related to one physical type of device. You can meet those optical standards in various different kinds of devices. Today, I would say that the 2 principal contenders for 400-gig pluggable optical modules for most applications are the QSFP-DD and the OSFP. And the major difference there, the sizes are a little bit different, but the major difference has to do with the amount of thermal capacity. The QSFP-DD dissipates somewhat less thermal energy than does the OSFP. And so depending on your thermal budget and what type of device you're trying to put in there, you may need a higher thermal budget, in which case you would choose the OSFP. If you can get by with a lower thermal budget, 7 to 14 watts, then the QSFP-DD would likely work for you. But that's the major reason why you would choose one over the other. But in principle, any of these can apply to the optical standards that we talked about earlier. So before I kind of wrap up, I wanted to spend just a moment to talk about the 5G because now that we've spent some time analyzing kind of the 400G, as I mentioned at the outset, the 400G is multiple channels, typically a 50-gigabit per second lasers that are PAM-4 encoded to give you a net throughput of 100 gigabits per second per channel. Multiplex with 4 channels, that gives you the 400 gigabits, okay? But if you back that off and say, "Okay, each of these channels is going to be a 50-gigabit per second channel or PAM4 encoded a 100-gigabit per second channel." Then you kind of compare that to what are the needs for 5G optics. And in the fronthaul and midhaul, we're looking at data rates that are typically 25 gigabits, 50 gigabits or 100 gigabits per second over distances that are a few kilometers in the fronthaul and many of the midhaul applications. So the point I'm trying to make is to -- when we consider all of the advantages, all of the technology that we talked about at 400-gig for the data center, a lot of those same advantages in technology is being utilized for the fronthaul and midhaul optics. The main difference there or the biggest challenge in moving from data center to 5G is really the fact that the 5G modules are typically designed to operate in harsher environmental conditions. They're going to be deployed in antennas that are mounted outside. And therefore, they have to be wider operating temperature ranges. They have to be sort of more ruggedized, if you will. But the main optical technologies, whether it be an EML, silicon photonics or directly modulated laser, generally can -- the advantages of those approaches to the data center optics also apply to the requirements for the front- and midhaul optics. And so that's why one of the reasons why AOI is pretty excited about the advent of 5G technology is because we do believe that a lot of the technology that we've developed for the data center really is applicable to this telecommunications generation. And that hasn't always been the case in previous generations of technologies. So the other thing that, again, before wrapping up, I wanted to just very quickly go over some of the automated manufacturing that we have. I mentioned earlier that the -- that AOI employs a kind of discrete approach, that is we're putting multiple different lasers together with optical receivers and PCB, printed circuit board, that contains the driver ICs and what have you. One of the reasons why we think that this approach scales well even with relatively high volume is because we've invested a lot of money and a lot of energy over the years in developing a highly automated manufacturing process. And here, you can see just a couple of snapshots of some of the automated assembly and test equipment that we have put in place. There's a lot more, if you ever get a chance to come to China or Taiwan and see where a lot of this is done. We'd be happy to show off a little more of this to you live. But suffice it to say that in the old model, where a lot of these modules were kind of assembled by hand some years ago, this discrete approach, it was a reasonable guess that this discrete approach might not scale very well with volumes. Because of the attention that we've paid to the manufacturing and the time and energy we've put into automating this manufacturing process, we think this actually scales pretty well to high volumes. And we've been able to demonstrate that during the 40-gig transition to 100-gig, where we've been able to kind of flexibly change as demand from the customers ebbs and flows for various different kinds of devices. Our automated manufacturing process was able to be efficiently utilized both for manufacture of 40-gig as well as 100-gig. And we think that this same automated manufacturing approach plays pretty well at the 400-gig generation as well. And with that, I see we've already gone a few minutes over where I hope to end up here. So I'll just go ahead and throw the floor open for questions.

[Operator Instructions] Our first question comes from Simon Leopold with Raymond James.

At first, just a quick clarification. Stefan, in your opening comments, you talked about how you've shifted work into Taiwan and Houston facilities and the added cost that brings back. I guess, the end of last month, on your earnings call, you had forecast gross margin of 23% to 25% for the March quarter. Is that reflecting those shifts? Or would there be additional pressure on the margin? And then I've got a follow-up.

Yes, Simon, that's -- the guidance that we gave in the earnings call is our sort of best guess at this point. I did note in our prepared remarks this morning or this afternoon earlier that there's still a great deal of uncertainty around that. I wanted to give a little more information to everybody about what the nature of these additional costs are, a little more than we did during the earnings call. I'm not necessarily trying to indicate that our guidance is changing. There's just a great deal of uncertainty around it, and a lot of this is going to come down to kind of the wire, if you will, at the end of the quarter. So that's all I was trying to say. We're not backing away from the guidance at this point. Just wanted to give a little more color on it.

Great. No, I appreciate that. So in terms of the trending question, I wanted to see if we could get a better understanding of how you view the 5G market opportunity overall in terms of how big a market is it. And specifically for you guys, is AOI's strategy one of really selling lasers into Chinese transceiver manufacturers? Or is there an opportunity to participate on the transceiver side for 10- and 25-gig fronthaul, midhaul in that particular opportunity in 5G?

Yes, Simon, I appreciate the question. We're looking at both approaches, that is we are likely to sell laser diodes for -- to other companies to manufacture transceivers for certain applications. And in other applications, we'll also be working on selling transceivers. So it's not an either/or for us. We're going to look at the economics of the opportunities that we see, and we can work in either mode. One of the big advantages, I think, that AOI has, of course, is the fact that we have our own internal laser fabrication facility. Relative to many of our competitors, I think that facility is operating at scale, that is we're able to make a fairly high volume of lasers. And those can be utilized in our own transceivers, if it's economical to do so, or they could be sold to other transceiver manufacturers, if that makes more sense. So we'll be kind of trading off margin and revenue, depending on whether we sell the laser diodes as discrete components or if we sell them incorporated into an optical transceiver.

And any thoughts on the size of the market opportunity from an industry perspective?

At this point, I don't have a good number on that right now. I mean, it's certainly a very sizable opportunity. I mean, the numbers that we're hearing out of China are enormous in terms of the number of optical modules that they're going to need. It's a little difficult to flesh out the timing on that, which is why I'm reluctant to give a number at this point. But it's certainly a very sizable opportunity for the optics industry in general and I think AOI, in particular.

[Operator Instructions] There appears to be no questions at this time. I'll turn it back to management for closing remarks.

Thank you, everybody, for joining us this afternoon. We appreciate your time and attention. And I'd like to keep in mind as we go through this COVID virus situation, everybody, our colleagues and friends, I wish everybody the best of luck in dealing with this. And we'll be back in touch as soon as we can to give you all an update on how things are going with AOI. Thank you very much for your time.

Thank you. This concludes today's conference. All Parties may disconnect. Have a great afternoon.