Applied Optoelectronics, Inc. (AAOI) Earnings Call Transcript & Summary
November 5, 2026
Earnings Call Speaker Segments
Lindsay Savarese
executiveYou can find it on the company's site, ao-inc.com. This call is being recorded and is also being webcast live. A link to the recording is available in the Investor Relations section of the AOI website, and it will stay archived for 1 year. With us on today's call are Dr. Thompson Lin, AOI's Founder, Chairman and CEO, and Dr. Stefan Murry, AOI's Financial Officer and Chief Strategy Officer. Thompson will walk through an overview of AOI's Q3 results, and Stefan will cover financial details along with the outlook for the fourth quarter of 2026. After our prepared remarks, we'll open things up for a question-and-answer session. Before we begin, please take a moment to review AOI's safe harbor statement; on today's call, management will present forward-looking statements. Such statements involve risks and uncertainties, as well as assumptions and present expectations, which might lead the company's actual results, activity levels, performance, or achievements—whether of the company itself or its industry—to differ materially from what those forward-looking statements state or suggest. At times, forward-looking statements may be recognized by words like beliefs, forecasts, anticipates, estimates, suggest, intends, predicts, expects, plans, may, should, could, would, will, potential, or things, or through the negation of such terms or comparable expressions that signal uncertainty regarding future events or outcomes. These forward-looking statements reflect what the company currently expects, assumes, estimates, and projects. The company considers those expectations, assumptions, estimates, and projections to be reasonable, but forward-looking statements are merely predictions and carry known and unknown risks and uncertainties, and many of those sit outside the company's control. Forward-looking statements also cover what management believes and expects about extending its products' reach into new markets and how customers respond to its innovation, along with statements about the company's outlook for the fourth quarter of 2026, for the full year of 2026 and for 2027. AOI is under no duty to revise these forward-looking statements later—whether to reflect actual outcomes or shifted company expectations—after this call's date, unless the law demands it. For further detail on additional risks that could affect AOI's business, see the Risk Factors portion of the company's filings with the SEC, which include its Form 10-K annual report along with its Form 10-Q quarterly reports. Additionally, unless we say otherwise, every financial result and other financial measure covered today is presented on a non-GAAP basis. Non-GAAP financial measures should not be viewed on their own, nor should they serve as a replacement for figures prepared under GAAP. You can find a reconciliation of our GAAP to non-GAAP figures, along with an explanation of the reasons we report non-GAAP financial measures, in the earnings press release the company issued, which is posted on AOI's website. Before I turn to the financial results, one scheduling note: AOI management will be at the Raymond James TMT and Consumer Conference on December 9. That discussion will be streamed online for you to watch, and a link to the webcast will be placed in the Investor Relations section of the AOI site. And one final item: the call to discuss AOI's Fourth Quarter and Full year 2026 results is currently scheduled for February 25, 2027. With that, I'll turn things over to Dr. Thompson Lin, who created AOI and serves as its Chairman and Chief Executive Officer — Thompson, go ahead.
Chih-Hsiang Lin
executiveThank you, Lindsay, and thanks for being with us on today's call. We're glad to report solid results for the third quarter, powered by strong demand across both our data center and CATV businesses. Third quarter revenue was more than double what we posted in the same quarter a year ago, and this marks the fifth quarter in a row of record revenue. We also got back to GAAP profitability during the quarter while non-GAAP profitability kept climbing. In October, we signed a multiyear long-term agreement, or LTA, with a Tier 1 hyperscale data center customer. We see this deal giving us meaningful visibility into revenue and demand over the long term, which strengthens our ability to add capacity, deploy capital, and put money into next-generation technology with greater confidence. The < > speaks to our technology leadership and our differentiated capabilities in advanced optics. One thing worth highlighting is the AI infrastructure ecosystem; this agreement reinforces our confidence in the long-term opportunity ahead, since demand keeps rising for high-bandwidth, energy-efficient optical connectivity products, and we're well positioned to scale manufacturing and support the next wave of AI data center deployment. On that point, I'm very proud of everything our team has already achieved, since we've deliberately and swiftly scaled our manufacturing capacity to meet this unprecedented demand. Our Phase I expansion is scheduled to start operations in December at our new facility in < >. That said, our Phase II capacity expansion is now expected to slip by several months because of anticipated longer-than-expected time to bring in the necessary electrical power, along with other construction-related issues; we believe our goal of a revenue run rate of $471 million per month is achievable toward the end of 2027, which accounts for this delay, and we expect our U.S.-based manufacturing capability to grow as this new facility comes online. We still anticipate a significant revenue ramp in Q4 and over the next several quarters, as we keep seeing benefits from our expansion in Taiwan and from expanded laser manufacturing capacity in < >. While we're disappointed about the anticipated Phase II construction delay, the underlying demand remains strong, and we remain committed to growing revenue and profitability, albeit on a slightly delayed schedule. Turning to the third quarter: we posted record revenue of $269.9 million, in line with our expectations, and non-GAAP gross margin of 28.7%, slightly below expectations due to product mix, which Stefan will cover in more detail. Our non-GAAP earnings per diluted share of $0.17 also came in line with our expectations. Importantly, the volume ramp of our 800 product carried on in Q3, with 800 revenue nearly quadrupling sequentially to $47.5 million. Our 400G revenue also almost doubled sequentially, building on a strong Q2. Finally, demand for our 1.6 terabit product is robust, and our new LTA covers 1.6 terabit transceivers exclusively. With that, I'll hand the call to Stefan to walk through the details of our Q3 performance and the Q4 outlook. Stefan?
Stefan Murry
executiveThanks, Thompson. We're happy to report solid results for the third quarter, powered by healthy demand across both our data center and CATV businesses. Third quarter revenue was more than double what we posted in the same quarter a year ago, and it represented our sixth quarter in a row of record revenue; we also got back to GAAP profitability during the quarter while non-GAAP profitability kept growing. In Q3 we posted record revenue of $269.9 million, which fell inside our guidance range of $255 million to $290 million. Non-GAAP gross margin came in at 28.7%, a bit under our guidance range of 29% to 30.5%, because of product mix, which I'll go into more detail on shortly. Non-GAAP diluted earnings per share came in at $0.17, which falls within the guidance range we provided of $0.11 to $0.26 per share. Healthy demand across both the data center and CATV businesses continues to underpin our results, underscoring the strength of our dual growth strategy and our diversified revenue streams. As Thompson mentioned, during October we signed a multiyear long-term agreement, or LTA, with a Tier 1 hyperscale data center customer. In our view, this agreement provides meaningful long-term visibility into revenue and demand, which bolsters our ability to expand capacity, allocate capital and invest in next-generation technologies with greater confidence. The entire scope of this agreement is a multiyear commitment for 1.6 terabit transceivers. We've achieved significant strides in expanding our production capacity to keep up with surging need for AI-focused optical technologies; this deal showcases our technological leadership, our distinctive strengths in sophisticated optics, and the key part we play within the AI infrastructure ecosystem. I want to provide a few more specifics regarding this arrangement. This buyer is already a customer of ours, and the LTA exclusively encompasses 1.6-terabit transceiver items; we recently completed successful qualification of our inaugural 1.6 terabit product with this particular customer. Provided certain requirements are met, among them our capacity to deliver, this customer has committed to sourcing no less than 10% of the 1.6-terabit transceivers it needs from AOI across the three-year span running from June 2026 to May 2029. Right now, our expectation is that this LTA will bring in revenue of $1 billion to $1.5 billion over this period—though that figure could shift considerably based on how much demand the customer actually has for 1.6-terabit products. On our side, AOI has committed to providing that volume at agreed pricing, which includes certain price reductions across the period. No capital commitment was made by the customer under this LTA. Thanks to our automated manufacturing capabilities and in-house laser technology, we think we are well positioned to serve this customer and others as 1.6 terabit deployments speed up. In the quarter, we kept advancing our main priorities as well. These were: first, growing our next-gen data center offerings — among them the 800G and 1.6 terabit solution — by adding production capacity in a measured way; second, broadening where our revenue comes from; and third, tightening operational execution so margins improve and we're set up for profitability over the long run. Now, a closer look at our third quarter results. Total revenue was $269.9 million, representing a 127.5% increase from the same period last year and 40.6% growth versus the prior quarter. During Q3, 58% of revenue came from data center products, and CATV products represented 41%. Third-quarter revenue in the data center segment hit $155.6 million, which was up 254% from a year earlier and 45% higher sequentially. Thompson noted that 800G revenue nearly quadrupled from the prior quarter, reaching $47.5 million, and 400G revenue almost doubled sequentially to $93.4 million. Demand also remained strong for our 1.6 terabit product, and as I mentioned, the new LTA we signed is exclusively for 1.6 terabit transceivers. Looking at data center revenue by product in Q3: 100G products made up 7.5%, 200G and 400G transceiver products made up 60.2%, 800G transceiver products represented 30.5%, and 10G and 40G transceiver products accounted for 1.4%. Although 800G growth was solid, it fell somewhat short of what we had expected, mainly because our LTA customer moved its demand from 800G to 1.6 terabit faster than we anticipated. That shift should carry into Q4. For the fourth quarter, we project data center revenue in the range of $270 million to $300 million, which works out to 281% year-over-year growth at the midpoint. Demand for our data center products remains exceptionally strong, but we do expect fourth-quarter data center revenue to land somewhat under our prior expectations, since — as I just said — our LTA customer is moving orders from 800G to 1.6 terabit more rapidly than we had forecast. Looking at the longer horizon, we see this transition as beneficial, since 1.6 terabit commands better pricing and margins compared with 800G. The accelerated timing, however, leaves us short on enough 1.6 terabit raw materials in Q4 to meet the higher demand in full. That customer returned as a greater than 10% customer in Q3, which was what we had anticipated, and we expect that status to continue across the entire LTA term. Looking ahead, we anticipate that demand for both 800G and 1.6 terabit modules will exceed our production capacity throughout all of 2027; we continue adding capacity and securing the key components needed to serve that demand. As Thompson mentioned, during the quarter we made substantial progress in ramping data center production capacity. As a reminder, our U.S. manufacturing base is anchored in Sugarland, just outside Houston; we hold more than 1.6 million square feet across the Greater Houston area at various stages of development, and all of it will go toward expanding capacity for 800G and 1.6 terabit transceivers, ELS FP modules and the lasers needed to build those products. The first of these major expansions is the 210,000 square foot facility, situated only a few hundred yards from our headquarters. As a reminder, that facility is devoted entirely to manufacturing 800G and 1.6 terabit transceivers. Initial operations there should start in December, in line with our plan to begin operations by the end of 2026. Our plan is to shift the transceiver production we have today out of the headquarters site and into that new building, and that opens up room at HQ to grow our indium phosphide laser capacity. As our footprint in Texas keeps growing, making lasers in-house puts us in a strong position to serve near-term customer demand as well as longer-term growth. We think CPO and related technologies are going to keep fueling rising demand for high-power lasers down the road. To support those future growth drivers, we intend to expand laser manufacturing capacity in Texas even further. In the quarter, we got encouraging preliminary test results from a number of CPO customers regarding how our lasers performed, and we're continuing to work alongside those customers toward full qualification of our lasers and, in time, ELS FP modules. Overall manufacturing capacity now stands at close to 330,000 units per month, compared with roughly 200,000 units per month when Q2 ended. By the close of this year, we now expect to be able to make over 500,000 units of 800G and 1.6 terabit products each month, versus the 650,000 units per month we previously expected, because of construction delays, which I'll get into shortly.
Stefan Murry
executiveLooking ahead to the end of 2027, we expect to be capable of producing more than 1 million units each month of 800G and 1.6 terabit products — an increase from the 930,000 units monthly we had forecast earlier — and that reflects greater demand from our customers. We anticipate that nearly 30% of this output will be manufactured in Texas. These production figures are projections of the capacity of our qualified production equipment; actual output will vary from them based on factors such as customer qualification schedules, staffing levels, raw material inputs, manufacturing cycle time and other considerations. The expansion we're planning for capacity amounts to a deliberate, step-by-step increase in our manufacturing footprint—one that matches robust, rising customer demand as well as the qualification headway we're making on both 800G and 1.6 terabit products. It's worth recalling that the identical production line and identical process can build both our 800G and our 1.6 terabit offerings. Our 1.6 terabit products do require a separate final test, yet the automated manufacturing lines for 800G were designed with an architecture enabling us to accommodate upcoming high-speed products as customer demand emerges and shifts over time. Those automated lines are built so they scale up efficiently from 800G to 1.6 terabit while requiring only minimal added investment. That flexibility in our structure provides two strong benefits at once: it shortens how quickly AI clients can reach the market, and it broadens the margin opportunity we can capture over time. Going forward, we anticipate that 800G will stay the biggest factor behind our data center ramp in the near term; 1.6 terabit should start adding to revenue during the fourth quarter, with a bigger ramp getting underway in 2027. In addition, we foresee 400G revenue continuing to rise, helped by the manufacturing capacity we are adding. Just like we anticipated and talked about during the previous quarterly earnings call, we recorded a drop in revenue from 100G transceiver sales this quarter, since one of our customers hasn't been able to obtain sufficient 100G switches to fulfill its original forecast. We think the switch scarcity stems from the memory shortage affecting the whole industry, and we still anticipate that 100G softness will last until memory supply rebounds. Despite this short-term 100G softness, we anticipate that our data center revenue for the fourth quarter will rise substantially, on both a sequential and a year-over-year basis. To recap our expansion plans. Our Sugarland Phase 1 capacity expansion is still mostly tracking to plan, though as we noted before, we now expect to come in a bit under our prior target for production readiness at year-end; we still anticipate this facility starting operations by the close of 2026. As Thompson said, we now expect Phase 2 to slip by a number of months, and we believe hitting a monthly revenue run rate of $471 million is achievable near the end of 2027; until then, we look for a substantial revenue ramp in Q4 and across the following several quarters, driven by gains from our Taiwan expansion and our Phase I U.S. On the expansion front, we still see our top line as constrained by how much we can produce and by supply chain, rather than by demand in the market. Turning now to CATV. In Q3 we posted record CATV revenue of $110.9 million, up 57% from a year earlier and 37.6% quarter over quarter, which came in slightly above the $100 million to $110 million guidance range we had given. As in recent quarters, we shipped a large quantity of 1.8 gigahertz amplifiers during Q3 to our largest CATV customer. At the FCTE Tech Expo in Atlanta at the end of September, we hit a major milestone — shipping our 1 millionth 1.8 gigahertz amplifier. Beyond that, we showcased what connectivity could look like going forward by running a live demonstration on a 3 gigahertz network, and we also marked the first appearance of the dual-mode Quantum Wave 25G tunable module along with fresh autonomous features powered by AI within our Quantum Link software. Customers have responded very positively, and we're excited about the potential of these emerging technologies. As for Q4, we anticipate CATV revenue coming in at $105 million to $115 million. The bulk of that Q4 CATV revenue will come from amplifiers. We look for software solutions to start contributing some revenue beginning next year. In the third quarter, our top 10 customers accounted for 99% of revenue versus 97% of revenue in Q3 of last year; we had 3 greater-than-10% customers — one in the CATV market, contributing 41% of total revenue, and two in the data center market, contributing 28% and 23% of total revenue, respectively. As I noted, one of those data center customers, a large U.S.-based hyperscale customer, returned as a 10% customer this quarter. Non-GAAP gross margin for Q3 came in at 28.7%, a bit below our 29% to 30.5% guidance range, versus 29.8% in Q2 of 2026 and 31% in Q3 of 2025. The difference from our guidance reflects a greater proportion of 400G within our data center segment, which stems from the same shift in customer demand—away from 800G and toward 1.6 terabit—that I mentioned earlier. Our long-term aim remains returning non-GAAP gross margins to about 40%, and we believe we can get there as our revenue mix tilts toward higher-margin offerings and we realize additional efficiencies throughout our operations. The revenue numbers I've cited are net of contra revenue associated with the accounting for warrants granted to certain customers. As a reminder, that works out to roughly 2.5% of revenue from the customers to whom AOI has issued warrants; during Q3, this contra revenue totaled $1.93 million. Non-GAAP operating expenses for Q3 totaled $70.3 million, representing 26% of revenue, compared with $47.1 million, or 39.7% of revenue, in the same quarter last year. Going forward, we anticipate non-GAAP operating expenses will land between $68 million and $75 million each quarter. Non-GAAP operating income for the third quarter reached $7.2 million, up against an operating loss of $10.3 million in Q3 of the previous year. GAAP net income for the third quarter was $6.3 million, equal to $0.07 per diluted share. This compares with a GAAP net loss of $17.9 million, or $0.28 per basic share, in Q3 of the year before. Looking at non-GAAP, net income in Q3 reached $14.8 million, or $0.17 for each diluted share, and that landed inside our guidance range for non-GAAP net income of $10.1 million to $24 million, or non-GAAP income per share of $0.11 to $0.26. That's versus a non-GAAP net loss of $5.4 million, or $0.09 per share, in Q3 a year ago. Weighted average fully diluted shares outstanding for the Q3 earnings per share calculation were 87.8 million. Moving on to the balance sheet now. We closed the third quarter with $621.5 million across total cash, cash equivalents, short-term investments and restricted cash. That's against $508.8 million at the close of the second quarter of 2026. At the end of the third quarter, total debt excluding convertible debt stood at $148.1 million, versus $92.8 million at the close of last quarter. As of September 30, inventory totaled $378 million, compared with $278.8 million at the end of Q2.
Stefan Murry
executiveThat rise is mainly due to buying more raw materials and having more work in process, which supports the revenue growth we expect. In early October, we announced that we had finished the $600 million at-the-market equity offering program we had previously disclosed. That completes our third ATM program this year successfully. We plan to use the proceeds to keep making strategic investments in the business, including new equipment and machinery for production and for research and development use, as we keep scaling our advanced manufacturing capacity. In the third quarter, we made $501 million in capital investments in total, including $142.8 million in prepayments on equipment we have on order. Those expenditures are mainly for expanding manufacturing capacity for our 400G, 800G and 1.6 terabit transceiver product. We expect capital expenditures to rise in Q4 versus Q3 and to rise further in 2027. Customer demand guides our capital investments, and purchase orders and long-term agreements support them; we are putting capital toward investments that scale our advanced manufacturing footprint and enable our path toward sustained profitability. We still expect to finance these investments through a combination of cash on hand, cash generated from operations, equity issuance and additional debt. Turning now to our Q4 outlook: we expect Q4 revenue to be between $385 million and $420 million, which represents approximately 200% year-over-year growth at the midpoint; we expect non-GAAP gross margin in the range of 29.5% to 31%. Non-GAAP net income is expected in the range of $45.9 million to $60.4 million, and non-GAAP earnings per share between $0.48 and $0.63 per share using a weighted average diluted share count of approximately 96 million shares; with that, I will hand it back to the operator for the Q&A session. Operator?
Simon Leopold
analystThanks for taking my question. The first thing I wanted to check on was really to get an update on sort of what's baked into that 4Q 800 gig and 1.6T, given the customer transition you mentioned as well as the manufacturing capacity delays. I know we discussed this last quarter. I just want to get refreshed on what's included in your expectations for 4Q; and then I've got a follow-up.
Stefan Murry
executiveRight. So as we said, we expect to see continued growth in 800G, continuing that trend we've seen over the last several quarters. And we do expect meaningful contribution from 1.6T. The commentary we gave about the 1.6T was simply that we have a lot more orders for 1.6T than we can supply this quarter due to material limitations. I mean when the customer switched from 800G to 1.6T faster than we expected, we couldn't turn on a dime and just create those additional materials — we had what we had and maybe a little bit more, but that's the revenue shortfall we were discussing in the prepared remarks.
Simon Leopold
analystThen I get that construction problems caused the delay. What I'd like is a clearer picture of how capacity ramps up, maybe, because earlier we figured about $471 million of capacity would come around midyear, and now it sounds like the end of the year. Can you update us on where you expect to be at the midyear point?
Stefan Murry
executiveI'd kind of spread that growth evenly across the year. Essentially, we'll keep adding capacity as the year goes on. Even with the construction delays, we do expect to get some of that capacity online. I mean, it isn't linear — it's not an all-or-nothing switch; so maybe I'd smooth that revenue out across the year.
Chih-Hsiang Lin
executiveSo essentially, looking at Q4, data center revenue is around roughly $100 million each month, and that grows to $470 million in monthly revenue by the close of next year. So in rough terms, you could describe capacity expansion as a linear function, since—as we keep stressing—customer demand remains really strong. The real constraint comes from all the manufacturer capacity along with the key material supply for 1.6. At this point, I think we already have—I'm not sure exactly—well over $400 million in orders on hand. So—but at present, I think because of the manufacturing cycle time, and I think the key material supply, that's why we can perhaps deliver—I don't know—under $100 million of revenue in 1.6T during Q4. So we're continuing to collaborate with the supplier to address that supply problem.
Simon Leopold
analystI appreciate that. And then perhaps one final, fairly open-ended one: clearly, the entire AI ecosystem trade has looked bad, plus worries regarding the election and worries regarding data centers. I realize not much time has passed, but I'm just curious about your view on that, along with any conversations or discussions you've had with customers that might hint they're reconsidering their own investments?
Stefan Murry
executiveRight. So I'll address that second question first. I mean, no, we haven't held any conversations with customers that would suggest they're having any second thoughts about their investment plans. If anything, we keep hearing that they simply can't obtain material quickly enough, right? Now some of the — there are shifts in that demand, such as what we discussed regarding 800G moving to 1.6T and similar things; so there's movement beneath the surface, but overall demand, as we stressed multiple times during our prepared remarks, stays really strong. Regarding the election itself and what followed, if you will, I mean, it's quite early to tell and I can't say much about that. I think taking a serious look at what makes sense when it comes to data center build-out makes sense. But I think the bottom line is that data centers still have to get built. AI is a tremendous tool for enabling greater efficiencies in the workplace and every kind of scientific and engineering advance. So I think those investments will be made. Certainly, our customers will make them. Whether they'll be built in the same locations as originally planned or how much of that will be — it remains to be seen. I think it's also worth noting that our customers are solely the big hyperscale operators; we aren't working with Tier 2 operators in the data center space, for instance. And I think those big players largely have their projects for the coming several years already underway. And they clearly possess the capital to make those investments. So my inclination is that the hyperscalers, the biggest hyperscalers we work with, will probably be less impacted or not impacted at all by these political changes, but that's merely a guess at this point, it's very early.
Chih-Hsiang Lin
executiveLet me just say this: sure, there could be some kind of delay or cutback because of certain government matters. But on that point — even by the end of next year, right now at < > and with other key suppliers — we're only able to satisfy under 80% of customer demand. So even if some kind of reduction does happen, the demand out there is larger than what we'll be able to produce going forward. We're expanding as quickly as we can over the coming, I believe, 12 months or 18 months. So I don't foresee any impact on our revenue, at least over the next 18 months.
George Notter
analystStefan, I'm just wondering how you see the financing side of things? I did hear your monologue about mixing equity and debt. But clearly, these ATMs < > a big impact. With the share price sitting where it is right now, if that keeps up, would that shift how you think about funding the business down the road? Do you have any contingency plans?
Stefan Murry
executiveWell, I think we're constantly examining, as we've mentioned many times, the most efficient route for financing the investments we intend to make; so the share price is naturally one of those variables we take into account. That could influence, to some degree, how we make decisions there. After the quarter closed, we secured a debt facility—something we covered in the subsequent events section and disclosed via an 8-K, so the specifics are available there. As we've said many times, our approach now favors debt over equity more than it did previously, and we'll keep evaluating which balance of debt and equity is appropriate going forward.
Chih-Hsiang Lin
executiveSo as you can see, we just closed on $1,200 million of that bank loan within the past week; and honestly, if it's needed, I believe we'll grow the U.S. bank loan even larger — something like $400 million or $500 million over the coming 12 months. At the same time, I think during this year we raised the bank loan within our Asia operation to back our capacity expansion; and keep in mind, we've started turning profitable. On that basis, for next year, I think the revenue projection means we ought to see a substantial amount of profit. If you actually work out the figures, the data center should climb from, I believe, under $100 million this quarter to $470 million by the end of next year, with a better gross margin. That means the profit margin will be roughly, what, 20% by the end of next year. So we're able to — I think we can generate a substantial amount of profit on our own; so that will support our operation as well.
Simon Leopold
analystIs there any chance to return to customers and secure prepayments? I realize you didn’t pursue that during this latest LTA, but could there be an opportunity to shift some of this burden onto your customers? Then my other question concerns CapEx for next year. You said it would be higher, but any indication of the size would be helpful.
Chih-Hsiang Lin
executiveNext year won’t be higher; what we said was that Q4 would exceed Q3. We never claimed next year would be higher. CapEx next year will depend on customer demand. And as we mentioned, between the U.S., there are several months of delay because of the electricity issues. So of course, we’re delaying payment as well; we never said CapEx would be much higher next year. In fact, I believe it could be lower.
Stefan Murry
executiveOn your question about customer financing—that topic comes up in every conversation we have with customers, so I'd never claim it's off the table. As customers' own balance sheets get scrutinized—and that's been a major theme in the recent commentary surrounding the data center trade—they're also taking a hard look at how they put that capital to work. So it's probably less likely now than it once was, though it's definitely still on the table; and I'd say it's more likely to come up with some of the newer technologies, such as CPO and NPO, than with 800G or 1.6 transceivers.
Ryan Koontz
analystStefan, I'd like to ask about supply and where things stand with phosphide substrates and DSDs — I assume the 1.6 DSPs you brought up earlier are giving you trouble? Would you say those sit at the very top of the company's supply worries and priorities, or is there something else you'd flag?
Stefan Murry
executiveYeah. As we've mentioned in the past, on indium phosphide, through 2027 we're covered with the new phosphide. So that isn't really the present bottleneck — we're clearly taking steps to lock in supply further downstream, particularly as we begin concentrating on CTO and NPO and getting those ramped, which consumes far more indium phosphide than our current usage. So indium phosphide isn't an issue right now. It's simply a future item on our radar. The DSP constraints are genuine, above all at 1.6T. That said, it's driven more by priorities shifting and customer demand expectations changing than by an outright shortage. Put differently, had we gotten a few months of warning on some of these customer shifts, we likely could have lined up extra DSPs; so it's tight; we do expect this to be resolved, or at minimum improve substantially, in Q1 of next year. For the time being, though, the DSP is a major constraint on 1.6T.
Chih-Hsiang Lin
executiveRegarding the 1.6T DSP, I believe we are qualifying a supplier. I think we're collaborating with that supplier at present, since the DSP is going to be a risk factor over the coming years and over the next few months. So I think we grasp the problem. That's the reason we're qualifying additional supply at this time. And the good news is that our customers accept the qualification. That matters a great deal, because as I've stated on many occasions, sometimes the difficulty isn't on AOI's side — the customer has to approve, and a qualification has to be completed for the new supplier, okay?
Ryan Koontz
analystGot it. Regarding the power question tied to the new facility, is there anything we can learn on that front when it comes to Texas and ERCOT, along with political changes? I mean, how would you describe the environment there? Is it simply — is this more a logistics hurdle? Or is it more of a macro-level challenge when it comes to the power...
Stefan Murry
executiveThis goes back before anything related to the election — it has nothing to do with that. And we aren't derisking it at all — nor are we altering our view on it after the election; it's simply logistics around how you obtain transformers and cables and everything else that must be delivered to supply that power. Obviously, at some level we're competing with all those data centers out there. I think there's some — I mean, I suppose there could be somewhat of a silver lining, so to speak, in that if Texas somehow enacts some kind of moratorium or slowdown on data center growth, that might let us secure that power a bit sooner than we'd previously expected; but that's definitely not in our planning right now. I mean that — whether it happens or not, we'll be seen over the next 3 to 6 months or so...
Ryan Koontz
analystThat is clear. And perhaps one more quick question just to wrap things up. Do you have any high-level thoughts regarding what this Lumentum and Marvell combination means for the industry? How do you see that affecting the coming several years?
Stefan Murry
executiveYes. I mean, look, I believe it underscores how large the opportunity is that we are discussing when it comes to optics and data centers, and particularly AI data centers. I think the consolidation it reflects within the industry, and drawing semiconductor companies and optics companies nearer to each other—that is something the industry has sort of toyed with over the years. I think this stands as a very big testament to that; and it emphasizes the value on both sides of the equation, right, the DSP, the silicon side, as well as the optics, and being one of the remaining independent suppliers of optics, I think places us in a good position as consolidation, whether that accelerates or carries on, I think it is a good position to be in.
Michael Genovese
analystThis seems to involve a customer who's already with us—a familiar name, roughly 1.6, that's entering into the LTA. Typically, when we consider LTAs, there are trade-offs involved: some features work in your favor and others don't, relative to simply using stock supply agreements. So could you elaborate a bit on what makes this one an LTA and why it works to your advantage?
Stefan Murry
executiveThat's a fair question. I'm not certain there's really—actually, I was hashing this very question out with someone earlier. I don't think a universal definition exists for what counts as an LTA and what doesn't. In our view, an LTA means a commitment spanning multiple years—that, I'd say, is the key characteristic. When I say LT, I mean a long-term agreement—what sets it apart for us is the multiyear structure, and that's uncommon, quite uncommon in our case and, I'd argue, for most firms in this sector. That said, you're right. As with any contract, these things usually aren't one sided; so you get some favorable elements and some that are less favorable. On our end, we pointed those out, I believe, already during the prepared remarks. Still, the three-year term and the sizable share of the customer's business that we're bringing in at a minimum are definitely favorable for us. It provides visibility into demand over the long term. On the plus side, it lets us scale more comfortably and put the CapEx into building out the manufacturing infrastructure this customer and others require — those are the benefits. Of course, the flip side is that we give up some ground on price, so pricing is fixed. And should market pricing rise substantially, we'd forgo part of that extra upside owing to the LTA, since that rate is locked in. Those are the trade-offs.
Michael Genovese
analystAll right, that's great. So even with the capacity additions being delayed somewhat toward the close of this year and into the middle of next year — by year-end your goal, as I recall, is somewhat greater. You mentioned 1 million a month; previously I believe it was 950,000. I suppose I'm just hoping to get a better sense of how confident you are that the demand exists and that whatever you produce can be sold. How do you feel about it at this point?
Stefan Murry
executiveYeah. As we've mentioned a number of times during the prepared remarks, demand is exceptional and unprecedented. It's hard to find the right words — I'm not sure what superlatives fit how much demand we're seeing in the industry. I can't really put it any other way. It is very, very strong. As for the demand — the demand we can see — it's likely that we won't be able to satisfy it. Thompson touched on this — possibly just 80% of that demand gets met even into next year. What that implies is we'd likely be able to hit 1.2 million units per month, or possibly beyond, by the close of next year, provided we could build at that pace, and even then we'd only just be meeting what customers want. And that reflects the demand we're seeing right now, correct? That demand could shift further down the road. So demand looks good. The thing that limits how much our top line can grow is how quickly we can bring additional capacity online—and in the U.S. in particular, securing enough power is proving somewhat harder than we projected earlier, which is essentially what's behind this schedule slipping; even so, as you pointed out, we anticipate exiting the year at a stronger run rate than the one we had previously projected for mid-year, so on balance, despite some lag, the overall mood remains quite favorable and upbeat.
Michael Genovese
analystAnd then one last thing from me. Something we on the investor side tend to do is sort of rank what we understand about lasers—about 400-milliwatt ultra-high-power lasers—and who's got solid lasers versus who doesn't. And you folks are somewhat of an enigma. So is there any testing feedback at all, or just confidence you can share regarding how far along you are on that product and how it stacks up against the ones currently out there in the market? And if you were competing in that market head-to-head, where would you be at this point?
Chih-Hsiang Lin
executiveYes, you may be referring to the CPO NPO market.
Michael Genovese
analystExactly.
Chih-Hsiang Lin
executiveSo for the MPO laser, that's roughly a 100-milliwatt to 200-milliwatt laser, while for the CPU market it's a 300-milliwatt to 315-milliwatt laser. Based on what customers are telling us, AOI currently delivers the best power efficiency among all suppliers — and you're quite familiar with the other 2 suppliers, so there's no need for me to name them. So I believe that's the situation we're dealing with, and it isn't just the < >; it also spans the next generation of < > high-power lasers. So right now things are progressing fairly smoothly. As I mentioned, we've been running qualification with, I believe, 4, 5 customers for a substantial period already. That explains why we invested so heavily in the laser expansion — that's where the money goes — and exact emphasize detects more than 18 months, nearly 24 months. Starting from day 1, you pay the prepayment until the day you can begin delivering laser involving to customers, okay? But the good news is the laser carries a very good gross margin, okay, above 60% gross margin, with very few competitors in a very big market, particularly for the MPO and CPO market. < > sits beside the 4 LTA; I believe the second ALT is in the final stage, and we expect to buy within the few weeks. The site is far, far bigger than the first 1 — it will even be 5x bigger or perhaps more than that. They were including 800G and laser and other stuff for the second customer, and 1310.
Tim Savageaux
analystI have a couple of questions, and congratulations on the ramp. You announced a substrate agreement — I believe earlier this week, though with how things have been moving, it could have been a month ago. My question was this: it's a pretty substantial quantity, I think around 15,000 wafers a month, and is that enough to support your target run rate level as you exit '27, up near $500 million a month? Or it seems like it could support something higher than that, just from that substrate agreement alone; so I'd like your comment on that. And then, when does that agreement go into effect — when do you expect to start purchasing those substrates?
Chih-Hsiang Lin
executiveYes, it will be much bigger; and this — actually, the agreement is more < > for 2028, the other supplier is still expanding the capacity. < > a quick example, okay. For the new MOCVD, the G10 MOCVD, every month, every < > about 600 4G substrates; so it depends on how much substrate, okay? Then you know how many subs you need per month, okay? Number two, for every good wafer you can produce about more than 8,000 to 9,000, the good 100 < >, per substrate; for high-power laser for < > it is much bigger, so you divide by a factor of 5 or even 6. So you can figure it out. Yes. You know now, as we emphasized, we have 6 suppliers for substrate, but we are working with all of them. But by next year, we have enough supply, including our inventory, because of the potential risk factor, just how we signed the LTA with this supplier. And actually, you are working with other suppliers sometimes, including 2 in Japan and 1 in Germany.
Tim Savageaux
analystAll right, great. I do appreciate that added detail. A few more questions. First, I suppose, how many customers — how many hyperscale customers are you currently shipping 800-gig to? And then, I'm not sure whether you've addressed this already, but do you have some kind of revenue targets in mind for Q4 for 800 gig and 1.6?
Chih-Hsiang Lin
executiveYes. At the moment, we have 2 big customers, but we have, for other customer to — you mean the hyperscale data customer. Yes. Currently, I believe we have 2 major ones. But we do have other smaller customers, 4, 5 of them. For < > just at 1.6T is emphasized. Right now, the 1.6T, at present, we have — $415 million, something like that; but even in Q4, we can only deliver just $50 million to $70 million, okay? And our customers aren't very happy, but we're doing everything we can, and we're working very closely with the DSP supply. I think that's as in 1 of the bottlenecks.
Tim Savageaux
analystAll right. So it seems like — well, I'm not sure whether we'll witness any steep climb in 400 from this point. But is it fair to think that 800 gig will account for the bulk of what you're forecasting as an almost doubling in data center revenue?
Stefan Murry
executiveYeah, 400 gig will go up somewhat, though the majority of that expansion will come from 800 gig. And after that, as Thompson said, 1.6T is going to contribute a significant amount.
Chih-Hsiang Lin
executive$50 million to $70 million, depending on material supply or even more.
Chih-Hsiang Lin
executiveAlright, and thanks for being with us today. As we always do, we appreciate our investors, customers, and employees for their ongoing support; this is an exciting moment for our industry and for AOI. And as we move forward with the growing demand required to support the next phase of AI data center deployment. We look forward to meeting many of you at upcoming investor conferences, along with the next facility tour we are planning for mid-December. Thank you.
Operator
operatorThat brings today's call to a close, everyone. Thanks so much for joining us. You're welcome to disconnect at this time.
Read the full transcript via the API
You're viewing the first half of this call. Get the complete Applied Optoelectronics, Inc. transcript — plus 131,000+ transcripts from 7,000+ companies, speaker segments, AI summaries and full-text search — through the EarningsCalls.dev API.
Get the API View API docs →This call discussed
For developers and AI pipelines
Programmatic access to Applied Optoelectronics, Inc. earnings transcripts and 131,000+ others is available through the
EarningsCalls.dev REST API. Plans from $39.99/month — full transcripts, speaker segments,
full-text search, and the recently-added /api/v1/transcripts/recent polling endpoint for ETL pipelines.
More tools for stock research
OptionsBell: an unusual options activity scanner that tracks options flow across all US stocks and emails you when large trades appear in Applied Optoelectronics, Inc. or any ticker on your watchlist.
StockMarketScan: a stock screener with alerts: 24 pre-built screens refreshed every 5 minutes, watchlists and email notifications. Both offer API and MCP access and fair monthly pricing.