Tower Semiconductor Ltd. (TSEM) Earnings Call Transcript
August 4, 2026
Earnings Call Speaker Segments
Thanks, and welcome, everybody. Good to have you with us for Tower Semiconductor's call covering the financial results from the second quarter of 2026. Joining us on this call are Mr. Russell Ellwanger, our Chief Executive Officer, and Mr. Oren Shirazi, our Chief Financial Officer. Before we get started, I want to remind you that some of what we say today could be forward-looking, and such statements carry risks and uncertainties that might make actual results come out materially different. Those risks get spelled out in our SEC filings — Form 20-F and 6-K — along with documents filed with the Israel Securities Authority, and all of them can be found on our website. Tower has no obligation to update any forward-looking statements. The results for our second quarter of 2026 were put together following U.S. GAAP. A portion of the data we present might involve non-GAAP financial measures, which SEC Regulation G defines. You can find a reconciliation to GAAP figures, plus complete explanations, in the press release issued today and in the financial tables. As a reference point, a supporting slide deck sits on our website and is built into this webcast. Having said all that, let me hand things over to Mr. Russell Ellwanger, our CEO. Russell?
Hi, everyone, and thanks for being on the call today. I'm genuinely thrilled to walk you through where Tower stands, how we're progressing, and what the future might hold. Q2 was a really big one, setting major company records on every key metric. What these results show is that our technology portfolio keeps gaining value, and there's potent operating leverage built into how we run things. Revenue is climbing with real momentum, and that's flowing through to earnings at an accelerating pace. Q2 revenue came in at $460 million, and profitability was especially strong: 30% gross margin, 20% operating margin, and 20% net margin — all records for the company if you exclude non-recurring accounting items — and those represent 58%, 55%, and 55% quarter-over-quarter contribution from the added revenue, respectively. These outcomes mark just the first stage of the steady margin expansion we anticipate in the years ahead, powered by market-leading customer partnerships that call for a very rich product mix, supported by solid operational execution. Looking forward, for Q3 of 2026 we're guiding mid-range revenue of $520 million, which works out to an annualized revenue run rate above $2 billion; we said at the start of the year that Tower would have a very strong second half as previously announced capacity investments get qualified and turned into shipments. Kicking off the second half of 2026 at a $2 billion run rate opens the door to multiple exciting new chapters for this company. The strength of customer demand, our expanding partnerships, our demonstrated execution abilities, and the strategic investments we've made and keep announcing together form a powerful catalyst to speed up our growth in the short, mid, and long term. Because of direct and growing customer demand — which reflects the size of the market opportunity and, more importantly, our capacity to seize it — we've revised our 2028 model to $3.6 billion in revenues, $1.63 billion in gross profit, or 45% gross profit margin, and $1.2 billion in net profit, or 33% net profit margin. Please take a look at slides four and five. Very importantly, the profitability gains from Q2 aren't a one-time thing — they're just an initial step toward profitability expansion and cash generation, as shown in our updated 2028 model. Just as importantly, as slides four and five show, this growth comes with even greater efficiency. Right now we run at roughly a 10 percentage point gap between gross margin and operating margin — a highly efficient structure for a company investing heavily in future growth. As revenue grows, we drive more efficiency, with operating expenses as a percentage of revenue dropping to approximately 7%, which is 30% lower than current levels. This improvement doesn't come from limiting investment in R&D. Quite the opposite. The model includes an increase of R&D investment by over 40% against present levels. Instead, this improvement reflects the enhanced efficiency achieved through our support functions as we grow scale. Speed and execution are one of the primary differentiators for high-tech success; we must continue to invest in R&D. Also, we need to concentrate on simplifying context across every SG&A function. That's what we're doing, and probably with figures that match or beat what anyone else in the industry achieves at their best. Efficiency and faster execution naturally produce solid margins, yet even more crucially, they allow the business to keep succeeding over time thanks to that speed. Twenty-one days back, we disclosed a dual-track strategic expansion of 300-millimeter capacity in Japan covering our silicon photonics, silicon germanium, and advanced optical packaging abilities, and we secured backing from Japan's government via the Ministry of Economy Trade and Industry, METI. By merging Tower's specialized technology leadership and our finest worldwide workforce with Japan's unmatched manufacturing know-how and quality results, its top-tier research institutions, and a deeply dedicated workforce, we are creating a strategic platform that will fuel innovation, economic growth, and semiconductor leadership for many decades ahead. The first track of this dual approach brings substantial new 300-millimeter silicon photonics capacity, with complete production readiness anticipated in Q4 of 2027. It involves converting the Arai facility, previously Fab 6, into 300-millimeter silicon photonics capacity plus advanced packaging capabilities, while also maximizing Tower Semiconductor's Fab 7 300-millimeter output in Uozu. Track 1 is what drives the update to the 2028 business model. Track 2 will begin alongside the first track and involves building another 300-millimeter manufacturing facility next to Fab 7. That facility should deliver a 4x boost in our Japanese 300-millimeter manufacturing output, centered on silicon photonics, silicon germanium, and associated advanced optical packaging, putting Tower Semiconductor in position to keep serving our growing customer demand for emerging AI and data center applications, propelling next-generation optical connectivity needs, and is designed to offer a smooth route for Tower Semiconductor and our customers to keep growing beyond 2028. Turning to our businesses, please check slide six for the Q2 revenue breakdown. Our RF infrastructure revenues in the second quarter accounted for 49% of corporate revenue, with roughly 43% growth quarter-over-quarter and more than 140% growth year-over-year. Silicon photonics revenue on its own rose by more than 60% quarter-over-quarter and over 270% year-over-year, reaching a Q2 annualized run rate above $680 million, aiming for a $1 billion run rate in Q4 2026 as the capacity expansion previously announced keeps getting qualified. The complete ramp of wafer starts from these investments should happen within Q4 of 2026, generating wafer start capacity more than three times greater than the second quarter silicon photonics revenue shipments, with the full financial effect expected in Q2 of 2027. After starting the year with multiple breakthrough technology milestones, the second quarter was focused on further advancing Tower Semiconductor's leadership position in silicon photonics as an extra foundation for the next growth phase. We concentrate on growing capacity, pushing our technology capabilities forward, and strengthening our engagements with strategic customers in line with their long-term roadmaps. A basic shift is taking place in the data center sector, since AI performance is no longer judged only by compute; more and more, it comes down to how well data travels among processors. Optical connections that offer high bandwidth, low latency, and good energy efficiency are now an essential enabler for AI infrastructure. At the same time, spreading deployments across geographies has grown just as important to hyperscalers and to the communities involved. These developments map squarely onto the core strengths of Tower Semiconductor. As AI clusters grow from thousands of XPUs to hundreds of thousands, electrical interconnects are fast nearing the ceiling of what they can practically do. For 800G and 1.6T pluggable optical interconnects, silicon photonics has become the foremost platform, and for the most part those links have already pushed out copper when it comes to scale-out connections beyond the rack. What comes next is bringing optical interconnects to scale up, whether inside one rack or spanning several racks. Here too, silicon photonics paired with Tower Semiconductor is in a strong position to drive this shift, with multiple Near-Packaged Optics, or NPO, rollouts slated in the coming year and a great many additional ones still in design. Relative to pluggable optics, NPO provides far higher bandwidth density while using less energy per bit, and even so it draws on the very same mature ecosystem that pluggables use, which addresses hyperscaler and data center operator concerns regarding reliability, serviceability, or flexibility in multi-sourcing.
We shared that we have customer agreements in hand worth roughly $1.3 billion in silicon photonics revenue for 2027, and growth should be even stronger in 2028. Based on what we've said regarding Track 1 investment, a meaningful amount of extra 300-millimeter capacity is going to come online over the course of 2027. That additional 300-millimeter capacity has already been requested and committed by a number of lead customers, which shows how much confidence they have in Tower's capacity not just to ramp manufacturing quickly, but also to keep delivering on the highly differentiated technology roadmaps that future networking architectures will require. To support rising long-term demand, we unveiled Track 2 — the largest buildout of our Japan manufacturing operations to date. That project should increase its 300-millimeter capacity by more than four times, which positions Tower well to handle both the pluggable optics that do the heavy lifting today and our fast-expanding Near-Packaged Optics, along with the emerging co-package optics market down the road. Still, having capacity alone won't get us there. Future growth will also stem from the next wave of foundational technologies now moving through development. Looking ahead one to two years out, a number of these technologies should move into high volume manufacturing. That includes, in particular, heterogeneous integration of III-V materials onto silicon photonics, which enables integrated lasers, advanced modulators, and optical signal processing. To back this roadmap, we signed a multi-year epitaxial wafer supply agreement with IQE, which locks in a strategic supply of III-V epitaxial material, and at the same time we keep internalizing key manufacturing steps that improve performance as well as supply chain control. Our collaboration with Marvell, which has run for a long time, hit an important milestone: SiPho-based coherent optical modules moved past the multi-million mark. That result shows Tower can produce some of the industry's most complex silicon photonics ICs at high volume production scale. In the future, we anticipate that coherent photonics will become more and more vital for scaling AI architectures, in which several data center campuses function together as one AI factory and spread out power, cooling, and infrastructure needs over locations that are geographically dispersed. Growing demand for low latency, low power, efficient analog drivers, and transimpedance amplifiers across traditional pluggables, plus linear pluggable optics and linear receive optics architectures, keeps benefiting our silicon germanium business. Across all three 200-millimeter fabs, our 100G per lane and 200G per lane products are in high volume production, while we move toward Near-Packaged Optics and 400G per lane solutions that call for tighter electrical IC, photonics IC integration, and co-design. We're seeing a lot of demand from customers for the 300 millimeter silicon germanium platforms in our next generation. RF mobile revenue made up 12% of corporate revenues in the second quarter. During the prior quarter's discussion, we noted that our RF SOI business is going through a strategic shift from 200- millimeter manufacturing to 300- millimeter manufacturing, which allows for better performance, more integration, and greater value for our customers. In addition, we are relocating our 300-millimeter RF SOI production to Fab 7, which frees up Fab 10 and also frees up capacity at Fab 7 for our SiPho and silicon germanium business, and both of those are growing rapidly; because of these actions, revenues from our 300-millimeter RF SOI declined 14% compared with the same period last year. Looking ahead, momentum for design wins on our 300-millimeter platform is quite strong, particularly when it comes to premium smartphones. A roadmap packed with best-in-industry figures of merit has sparked enthusiasm and involvement across the market, which is fueling an expected tripling of RF SOI 300-millimeter wafer starts by the middle of 2027 compared with Q2 2026 shipments. For the second quarter, power management revenue made up 14% of total corporate revenues, posting year-over-year growth alongside solid demand for our BCD offerings in both 200 millimeter and 300-millimeter. By concentrating our technology on power delivery for high-performance computing, we hold a leadership position in LDMOS devices featuring low gate charge and low RDSON. Our latest power technology generation lets customers create DC-to-DC converters with high frequency and high efficiency for many growing markets. During this quarter, orders increased from customers we already had, plus we achieved very strong acquisitions of new customers across our entire power portfolio. That momentum largely originates in growth areas where thermal efficiency and power density are essential. By collaborating tightly with our key customers, we continue pushing forward our next generation power management roadmap. Sensor display represented 12% of company revenue in the second quarter. For our image sensor business, revenue is mostly unchanged compared with the same period last year. Even so, demand has risen dramatically—most notably within machine vision, where high-end, high-resolution sensors are needed for semiconductor inspection. That spike comes from faster construction of DDR and HBM memory assembly lines. Automotive is another area seeing the same effect, above all for EV battery inspection. Over the coming two years, this robust demand should keep expanding. With products our lead customers created on our cutting-edge global shutter 300-millimeter platform, we have a strong position to meet it. Now to utilization. While capacity was ramping heavily, the 200-millimeter fabs—Fab 2, Fab 3 and Fab 9—operated between 80% and 85%. Utilization at Fab 5 in Japan stood at 75%. Oren, please take over. Fab 7 is still running at full capacity and remains well beyond the 85% utilization threshold we use in our model. I'll now hand things off to our CFO, Mr. Oren Shirazi. Appreciate it.
Thanks. Good day, everyone. We put out our financial results for Q2 2026 earlier today. Now I'll walk through the key points from those results, along with the balance sheet, our CapEx investments, and the revised business models. Starting with the P&L. Revenue in Q2 2026 came to $460 million, the highest in the company's history, which works out to 11% growth quarter-over-quarter versus $414 million in Q1 2026, and 24% growth year-over-year versus $372 million in Q2 2025. Gross profit in Q2 2026 hit a record $138 million, equal to a 30% gross margin, up 72% from $80 million in Q2 2025. Operating profit in Q2 2026 was $90 million, which is 2.26 times the operating profit recorded in Q2 2025. Net profit in Q2 2026 reached $91 million, a 20% net margin, up 95%, or $44 million, from net profit of $47 million in Q2 2025. Earnings per share in Q2 2026 came in at $0.80 basic and $0.79 diluted, nearly twice the $0.49 basic and $0.41 diluted posted in Q2 2025. Turning to the balance sheet. As we announced previously, during Q1 2026 we received $290 million in prepayments from SiPho customers, mostly for 2027 capacity reservations; on the balance sheet at the close of June 2026, these customer prepayments appear as liabilities under short- and long-term customer advances, and they show up in the Q1 2026 and H1 2026 cash flow report as cash from operating activities. Our balance sheet remains strong, as the following indicators and financial ratios demonstrate. At the close of June 2026, our assets totaled $3.8 billion, made up mainly of $1.6 billion in net fixed assets, chiefly fab machinery, plus $2 billion of current assets. Our current ratio is still very strong at roughly 4.9x, and shareholders' equity hit a record $3.1 billion at the close of June 2026. Hedging. I'd now like to go over our currency hedging activities. On the Japanese yen: because most of TPSCo's revenue is denominated in yen, and the overwhelming majority of TPSCo's costs are in yen as well, we enjoy a natural hedge across the bulk of our Japanese business and operations. To offset part of the remaining yen exposure, we enter into zero-cost cylinder transactions to hedge against currency fluctuations. Even though the yen exchange rate versus the US dollar can move around, the effect on our margin is limited. On the Israeli shekel: we have no revenue in this currency, but some of our costs in Israel are denominated in shekel, so we likewise hedge a substantial share of that currency risk through zero-cost cylinder transactions to reduce this exposure. Although the shekel exchange rate against the U.S. The effect on our margin from fluctuations in the dollar tends to be contained, which is what we've observed across the last several quarters, a period when the shekel strengthened considerably versus the U.S. dollar. Let's turn to the CapEx investment plan. As we've said before, to keep up with rising SiPho and SiGe demand, we are carrying out a plan to invest $920 million aimed at expanding capacity and capability for SiGe and SiPho equipment at our eight-inch fabs located in Israel, Newport Beach, and Texas, plus our 12-inch Uozu fab in Japan; that investment remains on schedule when it comes to purchase orders issued, technology and process qualification, equipment arrivals, and ramp plans. Roughly 50% of that $920 million CapEx investment had been paid as of the second quarter of 2026 and appears in our cash flow from investing activities for the reporting period, with the other 50% anticipated to be paid across H2 2026 and full year 2027. Business model. As announced not long ago and presented by Russell earlier today, the company has revised its business model, which now factors in the planned investment in the Arai facilities, previously Fab 6, to convert it for 12-inch SiPho wafer manufacturing, which maximizes the company's 12-inch wafer output in Japan, with backing from METI, Japan's Ministry of Economics, Trade and Industry. This updated model rests on forward-looking operational business and financial assumptions, among them the assumption that every fab will run at 85% utilization following full installation and qualification of the $920 million in SiPho and SiPho CapEx investments; we keep executing to ramp up SiPho and SiGe capacity and capability at our Fabs 2, 3, 5, and 7, and following the planned repurposing of the Arai facilities and qualification of the new equipment tools slated for installation there. A further assumption concerns forward-looking wafer selling prices for existing and future products and flow. There are also forward-looking assumptions regarding cost and completion dates for the successful qualification of tools to be installed, process technologies to be qualified, and customer products to be qualified and ordered from the company. Under the updated model, we are targeting $3.6 billion in annual revenue, which comes in $760 million above the February model target, reflecting 27% revenue increase. $1.63 billion annual gross profit, producing a 45% gross margin versus 39% in the prior model. That annual gross profit sits $510 million above the prior February 2026 model, representing 67% incremental gross profit derived from the incremental $760 million revenue, reflecting our enhanced product mix. $1.38 billion of annual operating profit, producing a 38% operating margin versus 32% in the prior model. That annual operating profit sits $480 million above the prior February 2026 model, representing 63% incremental operating profit derived from the incremental $760 million revenue, again reflecting our enhanced product mix. Finally, $1.2 billion in annual net profit, producing a 33% net margin versus 26% in the prior model, and versus 20% in the second quarter of 2026. Compared with the earlier February 2026 projection, this yearly net profit comes in at $450 million greater, which works out to 59% incremental net profit generated by the extra $760 million in revenue, showing the benefit of our improved product mix; that wraps up what I had prepared; at this point I'd like to hand things back to the operator for your questions. Hello, Cody.
I appreciate you letting me ask. Hi, everyone. Nice work on yet another strong quarter and really solid execution. Well done, all. One thing I want clarified. A few fast questions. Regarding the SiPho bookings figure for 2027, that $1.3 billion — was it revised? Have you provided an updated figure for it?
No, that wasn't something I did. In the script I mentioned that the Q4 start rate, which would be completely reflected in Q2 2027 revenue, stood at three times the Q2 shipment.
That could amount to 680 times 3. That refers to capacity. Meanwhile, the $1.3 billion relates to prepayments committed by customers, under contracts that customers have committed to.
Even though the.
All right.
Right. Beyond what was mentioned, that added capacity already has parties committed to it. Has every bit of it been reserved? Not entirely, yet commitments exist for it. That's my take.
Would it be accurate to put it that way?
Pardon me. Please continue.
Pardon me.
I was about to mention that, as far as I can tell, nothing new was shared regarding the contracts and bookings, though an update was provided largely concerning what has begun or is going to begin in terms of shipments.
All right, that's great, thanks. Given the amount of visibility you've provided, you've managed to deliver record sequential and annual growth rates, on both a percentage and a dollar basis, and that has been quite impressive. With that degree of visibility, could you perhaps just handicap how likely it is that you can keep up this kind of growth rate, or even speed up from here?
So, we provided the 2028 financial model, and we said it's our expectation to hit those numbers, at a minimum, by run rate, nominally across the full year, in 2028. I believe you can work out what the growth rate is, based on a 2028 revenue level of $3.6 billion.
And then finally, Russell, perhaps you could give me some sense of the size of those Japanese projects — Track 1 and Track 2. What I'm after is figuring out the wafer volumes, or the revenue backing, that Japan will eventually provide. Less so Track 1, since I believe that's already clear in your fiscal 2028 model, where you're assuming that Track 1 build-out runs at full utilization. Is that the correct way to view it?
Correct — at 85% utilization.
How should we think about Track 2's scale?
We said Track 2 would boost 300-millimeter capacity by four times, and it's mainly meant for SiPho and SiGe. Specific figures weren't provided. One reason we held back on numbers is that we're in final negotiations — strong ones — not concerning pricing, only regarding the timing to finish the facility and complete its facilitization, plus talks with suppliers about the equipment itself. Budgets, all of it is approved, all of it is in place. Our focus is on funding everything through cash we generate internally, so it rests comfortably in our hands. Nothing, as we see it, is gated or leveraged by a fundraising event or anything like that. No dilution exists within the plans. At this moment we still lack the final schedule for tool installations and tool qualifications. Our goal is for all of it to be installed and functioning by Q4 2028, as stated in the script, which then delivers a seamless growth trajectory into 2029, having reached the full potential of the track one development in 2028. From there, our target would be to have that grow. The exact numbers of how much we would get from that we haven't said. Giving quarterly updates and financial models isn't necessarily our desire.
Sure.
What you should anticipate is that sometime during the early quarters, as we move into 2027, we'll be refreshing a long-term financial model. That timing hinges on how things ultimately shake out regarding the final completion of the factory and how many tools we're able to accommodate within that finished factory. Portions of it—including even parts of the building blueprint—haven't been locked down yet.
All right, that’s great. I appreciate the specifics. I want to make sure I understand—when you say track two’s capacity is quadrupled, 4x, does that refer only to the Uozu facility, or does that 4x figure also take Fab 6 into account?
Fab 6 is part of that 4x number. Anything we’d be adding step by step in Fab 6—which we’ll mainly use as an epicenter, plus for certain specialty packaging tools—yes, that incremental Fab 6 piece you can see in track one is factored into the 4x.
Alright, that's quite useful. Thanks, everyone.
Yeah, appreciate you taking my question. I've got a few. Russell, I'm hoping to get a clearer picture of how end market demand is shifting. When we think about the coming year and NPO starts to matter, does that open the door for you to raise the content within any particular transceiver?
As far as I grasp what you're asking, my answer would be no. What would essentially happen is that in those situations we'd be selling an NPO instead of a plugable, and quite often, in all likelihood, we'd be selling both. As for the content itself, no, at least based on how I'm interpreting your question.
Absolutely. What I'm attempting to do is get a clearer picture of whether synergy exists in this situation, particularly since I'm working off the belief that eventually the PIC itself is going to need SiGe and SiPho stacked together, and that's really where this question comes from. Should we end up with a structure that's more heterogeneous, a PIC-based one, my sense is that would work in your favor, and I simply want to check whether my thinking is on the right track.
My view is that your thinking is on the right track, and I'd say yes to that. In my opinion, the emphasis falls less on NPO and more on what comes in the next generation. That's the area where we discussed advanced packaging, and a substantial portion of the advanced packaging we are implementing involves die-to-wafer as well as wafer-to-wafer bonding.
All right, thanks. My guess is that this will also mean you need more SiGe capacity, and that's already included in phase I of the capacity expansion, correct?
Partially. For the most part, it's included in phase II.
All right, I'll keep it brief then.
The first phase has a very sharp focus — apologies — the first phase is aimed squarely and very intently at expanding SiPho, and as we've said, every bit of that growth we're pursuing in the first phase is already accounted for.
All right, excellent. Now for my second question, about where you manufacture. Since you're putting so much focus on Japan, and you've got this phase I and phase II, is it fair to think that most of your customers based in the U.S. would be served by plants in Japan, or is the U.S. another place where you'll eventually need to add capacity?
At both Newport Beach and San Antonio, we keep adding capacity, and we still intend to add more at each location. One reason we picked Japan is that it's geopolitically neutral. There's no problem with anyone being supplied from Japan. It's a great location for growth. There are several reasons to have picked Japan. The workforce we have there is genuinely remarkable. Much of our R&D in Japan came out of hires from Panasonic back when Panasonic was the crème de la crème, a top three Japanese company, and it was picking whoever it wanted, mostly from Tokyo University; our core R&D workforce remains exceptionally capable. On top of that, anyone anywhere who's spent time around Japanese manufacturing knows the quality is simply outstanding. It's embedded in the Japanese mentality. This is a side note, but we lived in Japan for a number of years while our youngest daughter was in kindergarten, and it's quite interesting. At just three and a half, four years of age, when a kindergarten student arrives at school, a plastic container holding a washcloth called an oshibori is part of their uniform, and the school day begins with taking that washcloth out and wiping down their desk, and the school day finishes with taking that washcloth out and wiping down their desk. Honoring the workplace matters enormously in Japan. That culture, when you begin at three and a half, four years of age, learning to honor the workplace, it never leaves you. I once made a film, which is kind of funny, just about the procedure of returning your tray and dishes after lunch. The activities in Japan, it's more or less an assembly line. There's absolutely no degradation of a quality mentality lunch, first when they're in the regular factory. When it comes to very high quality manufacturing, I think Japan is truly the best of the best. Those are several reasons for going there. The taxonomy we have of workforce is fantastic. Quality culture in Japan is fantastic. Education in Japan is amazingly good, and it's geopolitically neutral.
Understood. I appreciate the thorough explanation. One more brief question. At what point will RF mobile revenue level off? What you've said about preparing suggests that moving production to 300-millimeter is still ongoing. Will those adjustments to where you manufacture settle by the second half, or will this keep going through 2027?
Level off isn't the phrase I'd choose. If something levels off, that implies you've entered a phase without further incremental growth. As I think I mentioned, based on tape-out activity and the design work already secured, the biggest 300-millimeter RF SOI output we've ever achieved should come in Q2 and Q3 of next year. I wouldn't call it leveling off. It's really about reaching a stage of ongoing growth, and then what other initiatives let us expand from there. Yes, my expectation is that Fab 10, where our manufacturing happens, will be extremely full by the second and third quarter of next year.
Understood. Thanks a lot.
You're quite welcome. That was a good one.
Richard, hello there.
Excellent. Hey, Russell, how's it going?
Doing well, thanks.
Great, and congratulations on the fantastic results. Keep it up. My first question follows up on a few of Cody's questions, and it's about the eventual silicon photonics in SiGe 300-millimeter capacity once Track 2 is complete. You mentioned a 4x capacity increase, I believe at least a 4x increase. It feels like we need to grasp and describe how much of the capacity at that stage is 200 versus 300. Could you respond to that with numbers, or at least in qualitative terms, so we can think about it more concretely?
Honestly, I'm not certain I grasp the question. Would you mind restating it?
Okay. You mentioned that following Track 2, your capacity for 300 millimeter silicon photonics and SiGe is going to quadruple. To put a number on that, I believe we first have to grasp how capacity is divided between 200 and 300 at that moment, prior to the start of that expansion. In trying to work this out, I'm curious whether you could tell us what that capacity split looks like then, so we can attempt to compute it.
Yes. Let me just get my notebook quickly. Apologies. I'll return in a moment. Just a second. Apologies. Apologies, Richard. What I'd say is that, first of all, following — I'd say — 2027, mid 2028, every bit of our growth in SiGe and SiPho, and certainly in SiPho, is going to be at 300 millimeter. The 4x capacity expansion in Japan, the METI agreement, is truly centered on optics, and it concerns silicon photonics along with silicon germanium. At a minimum, our plan would be to add 20,000, 25,000 wafer per month of silicon photonics capacity. That figure can rise far, far beyond that; such is the country's ambition. Multiply that by 2.25 and you're looking at an extremely large quantity of wafer capacity relative to 200 millimeter. Even though I don't quite see why that ratio matters so much for you in modeling something, silicon photonics growth is very substantial, and the bulk of what comes from Track 1 — close to 100% — will be silicon photonics. You already have the delta figures for revenue between the earlier model and the current one. Once we move into the 300-millimeter facility, we haven't yet disclosed the exact split, but the largest share of all that 4x growth will be silicon photonics, and the second-largest share will be silicon germanium.
All right, thank you for that. Let me just say, Russell, I believe everybody is attempting to determine how the next revision to the business model will appear. My recollection is that you mentioned an update would come relatively early, or at some stage in the coming year, since all of us are trying to get that work done in advance. We'll be eager for additional specifics once you're in a position to provide them. Among my follow-up questions, one concerns NPO, and I would really like to understand how much of a meaningful contribution it will make to SiPho revenues in 2027. Could you describe the magnitude, whether in quantitative or qualitative terms, please?
My view is that it's going to be significant. What the precise % will be, I don't know at this point. My sense is that it won't land in the single digits, but rather in the tens when expressed as a percentage of our shipments, particularly during the second half of the year.
All right, excellent. That's quite useful. For my final question, I want to touch on advanced packaging. I believe there's some contribution from the Track 1 investments in Japan, though I suspect a larger portion sits in Track 2. Russell, I'd like to pose something with a longer horizon: when you consider the fullness of time, once all construction is finished and you factor in how the market will develop, what share of your silicon photonics revenues will come from packaging-related activity?
Great question. I'm not certain how much of that is strictly packaging-related as opposed to packaging-enabled. As things stand today, for the integrated laser — an indium phosphide laser — we'd anticipate several tens of millions in revenue next year from that integrated laser. That integrated laser involves a chipped wafer bonding process, which is packaging. Currently it's outsourced, though we intend to bring it in-house. It's not that we're being paid directly for packaging, and we aren't attempting to compete with packaging houses. What we're doing is adding more in-house capability so that we control the final outcome, and definitely gain far greater control over wafer start-to-ship timing instead of relying on a supplier, however capable that supplier might be. We aren't aiming to compete in packaging as such. Our view is that packaging serves as an enabler for expanding our core business with the highest quality and the fastest speed to production among any supplier out there. I hope that addresses your question, Richard. We won't break out a packaging revenue line. Our focus isn't on it as packaging revenue, but rather as something that enables our silicon photonics platform.
That's useful, and I appreciate it. Russell, I have nothing further. Thanks.
Thanks to you.
Yeah. I appreciate you taking these quick follow-ups, everyone. Russell, given all the capacity that's been added across the industry, from some of your competitors at their GlobalFoundries and—
Sorry, the beginning of that got cut off. My apologies. Would you mind beginning the question again?
Oh, no, that's totally fine. Given all the capacity being brought online across the sector by certain competitors of yours — GlobalFoundries, STMicro, Samsung have each put out announcements regarding planned additions to capacity. If you look ahead 12, 18, 24 months, could you perhaps just share your view on the supply-demand health of the industry, with all these various tranches entering the picture?
Supply is definitely going up. Right now we hold a clear advantage over anyone in that we're qualified with the lead customers globally, backed by very strong contracts running through 2028. At this point, with extra capacity coming online, there are strong interactions and discussions aimed at keeping contracts well beyond that. I don't really have too solid a sense of how much additional capacity will enter the market. Honestly, I haven't tracked that very closely. Our attention has truly been on how we keep sole share or majority share with our lead customers, and we're in very good shape there. The key thing is something I mentioned in the script, and that's speed. So long as we have programs underway for not only next generation, but generation plus two, and in many cases generation plus three, those programs always let you reach the market faster and stronger than anyone else. Related to that, what's very tangible from a business perspective, if we're engaged with a very large customer and we're dedicating substantial resources to next generation or multiple generation future developments, there are usually exclusivity agreements on both sides; we would let a lead customer get a head start, particularly for any module that's jointly developed, and we would ask for 100% market share. Again, for part of your question, I don't truly have a solid feel for how much capacity is additionally entering the market. What I believe we have a much better feel on is the market share we'll keep with our lead customers. We think that will remain extremely high.
Russell, I appreciate that. Just so I can wrap my head around it, is there some kind of benchmark, some sort of tiering, that you could direct us toward as a useful gauge, so we're able to watch how you keep leading the industry?
When you say benchmark, what do you mean exactly? Usually a benchmark refers to something technical that's been accomplished. What is it you're after? Are you after a way to gauge our share of the market?
Okay, no — I suppose what I'm really after is getting a sense of the industry chatter regarding the varying capabilities among the different competitors out there. I'm simply attempting to find out whether there's anything you could share that would let us grasp more clearly the places where Tower genuinely stands apart.
I hope this doesn't come across as flippant, but our differentiation lies in figure of merit. One of the most significant things anyone could examine is insertion loss. When you currently possess best-in-breed insertion loss, you genuinely assist the integrator, first, by removing the need to purchase pricier CW lasers for the sake of higher output, and second, through the capacity to cut down how many lasers go into the package. We put out a press release previously — it was a PR alongside InnoLight — regarding our insertion loss serving as an enabler for achieving the quantity of CW lasers required inside the package. My recollection is that it was a halving. I would need to revisit those PRs from a couple of years back. That really turns into the most substantial differentiator available to us: leading the industry on figure of merit. Turning to next-generation modulators, during the OFC conference we issued a joint PR together with Coherent covering our success in producing a 400G modulator in silicon. It's no secret that we — and possibly certain others — are engaged in work involving thin film lithium niobate; we're likewise pursuing indium phosphide for a modulator. The optimal approach one could take to preserving market share is asking how near you are to operating at the next generation and/or two generations ahead, and what your performance looks like in those domains. That is, if this addresses your question, Cody. I hope it does.
That answers my question, and it was quite useful. I appreciate it. One final item: you brought up InnoLight a moment ago. I realize your morning is full, so I'm not sure whether you've gotten the opportunity to look at certain stories in the press suggesting the administration might be considering restrictions on shipments of Chinese technology — optical technology, specifically — into the United States. I was simply curious whether you have any perspective on that.
Plenty of perspective, though not anything I'd be willing to share in a public setting.
Okay. Thanks a lot.
Morning. Appreciate it. I've only got two, just a pair of questions. To start, is your own output running into any scarcity or problems within the supply chain? I'm aware you dealt with the matter involving indium phosphide.
From a broad standpoint, no. Regarding starting materials and what we refer to as variable materials required for manufacturing, our situation is quite strong in every respect. Indium phosphide starting material had us somewhat constrained, and we think that matter has been handled quite well through the IQE contract. Apart from indium phosphide — where, amid a tough stretch, we pursued multiple avenues to boost substrate quantities — I believe that too has been sorted out. No, supply-wise we're sitting in excellent shape.
All right. For my final question, suppose we ran some kind of error analysis on your business model — where would you say the greatest variation in results might come from? Would that be expenses, timing, or the prices you project for your products? Where does the risk sit?
Yes, I believe I covered this during my prepared remarks — our model rests on a few key assumptions. The first involves the selling price for each wafer. The second concerns assumptions about costs, plus the installation and qualification timeline. The third is that we'll use 85% of the Fab, and those are the fundamental assumptions behind the model.
Understood. And out of those, which one has the widest spread of possible results — the one that deserves the most attention?
Typically that would be the price at which each wafer is sold, since a higher or lower volume brings a correspondingly higher or lower variable cost. When that price rises, the entire benefit flows straight through to the bottom line, and the opposite holds true as well. That's perhaps the usual case, though it is a broad generalization. As a rule, the selling price shows up entirely — 100% — in the margin.
Okay, we'll continue monitoring that, then. Thanks. That covers what I wanted to ask.
Thanks.
To begin with, just as I did when the call started, I'll close with the very same remark. I couldn't be more thrilled about the position we're in, the things we're pursuing, and what lies ahead for us. It was a pleasure to be able to walk you through all of it. One of the most thrilling aspects of working in a high-tech enterprise — and perhaps especially of serving in management or as CEO — is that in high tech, generally speaking, you're engaging and interacting with groups of individuals who, by any statistical measure, sit at the upper end of intelligence. That's a marvelous thing, a marvelous environment to operate in. When you pair that with individuals of exceptionally strong character, it turns into something fantastic, and the interaction is wonderful. I can honestly say that looking across business broadly, but at this moment specifically in silicon photonics, each and every one of our customers there, with truly no exceptions, are simply exceptional individuals that we work with. Every single interaction is beyond joyful. It's a partnership in which communication flows openly, you collaborate with one another, and from that you move forward into subsequent generations. The other day we had a really good call with a major customer concerning the fact that we have now announced — well, not announced, but concerning having very strong added 300-millimeter capability, and that sense of partnership, the manner in which we're expanding together, it's genuinely an incredible spot to be in. Among my very greatest pleasures in leading Tower is the chance I get each day within the organization to work alongside highly intelligent people of strong character, and to engage with customers who are likewise very smart and of high character, and to witness the two coming together in pursuit of creating new things. We've spoken about it many times, and it appeared in the script too, not only about Tower's expansion, but about the growth of Tower together with our customers; when you look at matters from that angle, from that perspective, it's a tremendous way to exist. Honestly, when we examine the financial model, when we examine reaching a 33% net profit, I'm simply delighted about the place we're in, the direction we're headed, and all of these opportunities sitting before us. What a truly extraordinarily wonderful adventure we're embarking on in Japan, building out a hub of optical excellence for the SiPho, for the silicon germanium. We're genuinely in an incredible position where everyone — not that this wasn't true previously, but people now arrive at work even more enthusiastic than they used to be. So much is taking place, and that's an incredible place to be. Having said that, my invitation goes out to anyone who is an investor: get in touch; we would be delighted to speak with you. We would be delighted to have as much interaction as possible, to help you gain a better grasp of what we're doing, where we stand, and to be able to share with you the strong results we're delivering at the company. With that said, we’re eager to connect with the investment community at the Jefferies conference coming up on August 25th and 26th in Chicago, as well as at the Benchmark StoneX Annual Conference in New York on September 10th. The opportunity to engage with investors is genuine, and the more clearly you grasp our story, the more clearly you see our current position and our direction, which we believe works out best for everyone. So with that, I’ll wrap up and simply express gratitude for your attention and your backing. Thank you. Bye bye.
That wraps up the conference call for today. We appreciate everyone for joining in. Everyone is free to disconnect at this point. Enjoy the rest of your day.
Read the full transcript via the API
You're viewing the first half of this call. Get the complete Tower Semiconductor Ltd. transcript - plus 132,000+ transcripts from 7,000+ companies, speaker segments and full-text search - through the EarningsAPI REST API or hosted MCP server.
Get an API key View API docs →For developers and AI pipelines
Programmatic access to Tower Semiconductor Ltd. earnings transcripts and 132,000+ others is available through the
EarningsAPI REST API and the hosted MCP server.
Quarterly plans from $162 - full transcripts, speaker segments, full-text search,
and the /api/v1/transcripts/recent polling endpoint for ETL pipelines.