Tower Semiconductor Ltd. (TSEM) Earnings Call Transcript & Summary
May 13, 2026
What were the key takeaways from Tower Semiconductor Ltd.'s May 13, 2026 earnings call?
In the first quarter of 2026, Tower Semiconductor Ltd. reported revenue of $414 million, reflecting a 15% year-over-year growth, and a net profit of $65 million, up 62% year-over-year. Management guided for the second quarter to reach $455 million, marking the highest revenue in the company's history, which indicates strong momentum and growth potential. The company also highlighted significant long-term contracts and customer prepayments totaling $1.3 billion for 2027, enhancing revenue visibility and confidence in sustained growth.
What topics did Tower Semiconductor Ltd. cover?
- Revenue Growth and Guidance: Tower Semiconductor achieved first quarter revenue of $414 million, a 15% increase year-over-year. Management anticipates second quarter revenue to be 'the highest revenue in the company's history' with a guidance of $455 million, plus or minus 5%.
- Profitability Improvement: The company reported a net profit of $65 million, yielding a net margin of 16%, up from 11% in the prior year. This reflects a strong operational performance and improved cost management.
- Long-Term Customer Commitments: Management stated that they have secured 'contractually representing $1.3 billion revenue in 2027' from photonics customers, which demonstrates strong demand and customer confidence in Tower's capabilities.
- Capacity Expansion in Japan: The transition to full ownership of the 300-millimeter factory in Japan is expected to enhance production capabilities, with plans to scale up to 4x current levels. This expansion is crucial for meeting growing customer demand.
- Silicon Photonics Demand: Silicon Photonics revenue grew 3x year-over-year, indicating robust demand in the market. Management expects continued leadership in this segment, with significant growth opportunities ahead.
What were Tower Semiconductor Ltd.'s May 13, 2026 results?
- Revenue: $414 million (vs $360 million est, +15% YoY)
- Net Profit: $65 million (vs $40 million YoY, +62%)
- Net Margin: 16% (vs 11% YoY)
- EPS (Diluted): $0.57 (vs $0.35 est, +63% YoY)
- Guidance for Q2 Revenue: $455 million (mid-range guidance, +22% YoY)
- CapEx Investment: $920 million (planned for SiPho and SiGe capacity expansion)
Tower Semiconductor's strong first quarter results and optimistic guidance suggest a robust growth trajectory, particularly in Silicon Photonics. The significant customer commitments and ongoing capacity expansions position the company well for future profitability. Investors should monitor the execution of expansion plans and the evolving competitive landscape in optical technologies.
Earnings Call Speaker Segments
Thanks. Hello, everyone, and I appreciate your participation today. This is Tower Semiconductor's conference call covering the financial results for the first quarter of 2026, and we're glad to have you with us. Joining us on this call are Mr. Russell Ellwanger, who serves as our Chief Executive Officer, and Mr. Oren Shirazi, our Chief Financial Officer. Prior to starting, we want to remind you that some remarks on today's call could be forward-looking and carry risks and uncertainties that might lead actual results to differ materially. Those risks are described in our SEC filings, Form 20-F and 6-K, along with documents filed with the Israel Securities Authority, and all of these can be found on our website. Tower undertakes no duty to revise any of those forward-looking statements. The results for our first quarter of 2026 have been prepared under U.S. GAAP. Certain data we present might contain non-GAAP financial measures as set out in SEC Regulation G. A reconciliation to GAAP numbers and complete explanations appear in today's press release and financial tables; as a reference, a supporting website can be accessed on our site and is incorporated into this webcast. Having said that, I'd like to hand the call over to our CEO, Mr. Russell Ellwanger. Russell?
Thanks, Noit. Hello, everyone. We're glad you could join us today. Our 2026 first quarter was strong and gives us a solid base for the significant expansion we expect over the course of this year. We maintained healthy financial performance even as we kept moving forward on our strategic goals. First-quarter 2026 revenue came to $414 million, representing a 15% increase compared with the same period a year earlier. Net profit for the quarter was $65 million, a 62% rise year-over-year, which works out to a 16% net margin — better than the 11% we recorded in the first quarter of 2025. Now looking forward, our guidance for Q2 2026 calls for the biggest revenue ever recorded by this company, with the midpoint sitting at $455 million, plus or minus 5%. That works out to a 22% gain versus Q2 2025 and 10% growth from the prior quarter. We're restating, without reservation, our goal of seeing both revenue and margin rise sequentially all through 2026. We continue to deepen our ties and collaboration with photonics clients by fulfilling long-term commitments from customers, which add up contractually to $1.3 billion in revenue for 2027, alongside considerably bigger contracts for 2028 that carry higher value, supported by roughly $290 million of prepayments we've already collected from our biggest SiPho customers. It speaks to how compelling our solutions are, and how much faith our customer partners have in what we can deliver as demand keeps climbing for next generation AI data center architectures. One key point: these reservations aren't the full picture of what these customers have said they want, and they aren't the ceiling on how much we intend to ship them either. They also leave out further wafer shipments going to a wider group — over 50 active SiPho customers we serve across a range of end market applications. Taken alongside our ongoing technology leadership and the strategic buildout of 300 mm as well as global manufacturing capacity, these commitments give us better revenue visibility and greater confidence in delivering sustained profitable growth. The restructuring agreement we announced not long ago in Japan involving TPSCo represents a major step forward for our long-term 300 mm strategy. Through the move to take complete ownership of Fab 7, our 300 mm plant located in Uozu, we're establishing a platform that's both more focused and more scalable, which will help us meet rising customer demand, especially for our differentiated optical photonics technologies. Having full ownership means we can grow and further develop a site that already runs several fully qualified high volume application flows and, crucially, is profitable at the volumes we're running today. Our 300 mm expansion, which is linked to the approval of numerous grants, has been set up to be strategic, operational, and capital efficient. Since we can tap into land right next door, our plan is to expand and eventually reach roughly quadruple what we run today, which should create a substantial growth driver over the long haul built on advanced, high-value technologies. Because we already hold customer qualifications and demand keeps rising, this strategy lets added capacity turn into sales and cash flow pretty much right away once new equipment comes online. That makes our 300 mm platforms both a major source of future expansion and a more durable profit contributor, strengthening the financial framework and the value we build over time. On top of that, we're signing a long-term supply deal with Nuvoton covering Fab 5 Tonami. That secures production continuity for our 200 mm client base, on conditions that work well for both sides. Turning now to how we performed in the first quarter of 2026, the year has gotten off to a very strong start, with silicon photonics leading the way on revenue growth of 3x year-over-year. Every one of our major technology offerings posted year-over-year growth: imagers rose 9%, RF SOI climbed 12%, power management increased 10%, and silicon germanium was up 24% year-over-year. Slide four has the reference for the Q1 revenue breakdown by technology. Now on RF infrastructure, last quarter was genuinely outstanding, both in how our team executed on aggressive capacity expansion and in reaching new breakthrough technology milestones. To start, we maintained a solid increase in shipments of 200 Gbps items across several clients, and at the same time continued to meet healthy need for our legacy offerings by making full use of additional capacity as it becomes available. Currently, we're engaged in a silicon photonics (SiPho) output increase at all of the following locations: Fab 2 in Migdal Haemek, Fab 3 in Newport Beach, Fab 9 in San Antonio, and Fab 7 in Uozu, Japan, which is a 300 mm facility. As part of this effort, during Q1 we reached a milestone with the first revenue-generating SiPho shipments from Fab 2 and Fab 7; notably, Fab 7 recorded a remarkable 95% yield on the initial SiPho wafers that departed the plant. We're still on schedule with our expansion plans. By the close of 2026, we expect SiPho capacity to reach five times what it was based on Q4 2025 wafer revenue shipments. Then, looking ahead to 2027, we think our main emphasis will shift toward expanding 300 mm capacity further at the Uozu factory, which we expect to fully own by then. Moving on, we've made several breakthroughs in next-generation technology alongside a number of our major customers. During this quarter, we shared news about an all-silicon 400 Gb per lane Mach-Zehnder modulator demonstration, done together with Coherent, a key partner and leader in the optical sector. Coherent is among the customers who have signed a long-term, high-volume contract with us. We've got some news about OpenLight: we just unveiled a 400 Gb per lane indium phosphide electro-absorption modulator that's heterogeneously integrated on the silicon PH18DA platform. We also moved significantly closer to high volume manufacturing for thin film lithium niobate, and we shared that we're working with Lightwave Logic and NLM Photonics to get organic polymers into high volume production, which will enable next generation compact modulators. Then, right before the Optical Fiber Conference, we told everyone about our partnerships with Salience Labs and Oriole Networks — together we'll manufacture advanced silicon photonics-based optical circuit switches, and both of these use the PH18DA platform with heterogeneous integrated indium phosphide optical amplifiers, delivering high bandwidth and ultra-low latency optical switch solutions for AI data center scaling. Finally—and this matters a great deal—Scintil Photonics, one of our partners, has made available something no one else has brought to market: heterogeneous integrated dense wavelength division multiplexer, or DWDM, laser sources, the first of their kind worldwide, built for near package optics and for CPO-based AI infrastructures. Through the close of this decade, pluggable optical transceivers should keep holding the leading position, according to the majority of market analysts. What we are in fact seeing is more high-density pluggable optics, referred to as XPO, an effort Arista is leading in order to extend how many generations of pluggables stay viable, and which Andy Bechtolsheim highlighted on his panel at the Optical Fiber Conference Executive Forum, alongside near package optics—and in the end co-package optics as well—which are now emerging and will coexist with pluggables over the next several years. So we are getting ready to embrace those technologies as well. During this year's OFC, Tower silicon photonics appeared in prominent XPO and near package optics demonstrations. Already, we're witnessing robust demand for NPO products in 2027. Because of this powerful customer momentum, we expect that Tower SiPho will keep its leading position in these novel optical form factors. Over the coming several years, NPO will likely grow in scale, and that will happen ahead of a significant CPO ramp among our main customers. We're directing considerable investment toward several CPO technologies — in particular, our internal 200 mm and 300 mm hybrid bonding that uses Through-Silicon Vias to join SiPho to electrical ICs seamlessly, laser sources built for both the more established architectures and DWDM architectures meant for CPO deployments, and more compact high-performance modulators created for CPO form factors where space is tight. Outside of SiPho, demand on our silicon germanium platform has been extraordinary, both for use in drivers and transimpedance amplifiers within optical transceivers and for active copper cables, an option that can act as an attractive replacement for optical approaches in short-distance scale-up architecture.
We're seeing a solid ramp right now for our RF silicon germanium technology, specifically for LNAs inside a Tier-1 mobile platform. On the silicon germanium side, we made an announcement not long ago about a partnership to build high-power silicon germanium beamforming ICs made in the U.S., intended for defense radar and satellite communication uses. These chips are fabricated at our U.S. sites, and the goal is to secure domestic supply chains while delivering superior performance for critical next generation defense systems. When our sites shift toward space — whether that's for data centers or global satellite connectivity — we believe our silicon image platform is well suited to support these growing applications. Turning to RF mobile: we're still in the process of shifting our RF SOI 200 mm technologies to 300 mm, which lets us benefit from finer line capabilities and other improvements that 300 mm makes possible, and at the same time we can free up that 200 mm capacity for SiPho and SiGe work that carries higher margins. Our Ron x Coff is dramatically better than what competitors offer, and combined with a lower layer count, that's driving solid design win momentum, putting our 300 mm RF SOI platform in a position for continued growth across the coming several years. Regarding power management, revenue has grown year over year for both our 200 mm and 300 mm BCD products. During the prior quarter, we shared news about the launch of Gen3, our newest power platform. For critical components running at more than 10 V, this platform delivers on resistance under 1.5 milliohm millimeter squared. Such a low on resistance makes possible strong power conversion efficiency in many different applications, and it places what we offer in a highly competitive spot versus competing foundry options. Our lead customers, leveraging this technology, have shown a 15% drop in power conversion losses — a notable result when set against even the most efficient alternatives out there. Among the end markets where revenue has risen for us are consumer mobile and automotive. Also, with the AI data center power delivery space moving toward an 800 V DC bus at the rack level, we anticipate a major growth opportunity in smart power stages and point of load converters built using our BCD offerings. And last, given the value we provide, pricing for our 200 mm BCD has gone up 13%. Image sensors. The CIS segments expanding most quickly are automotive, industrial, machine vision, and high-end video cameras. Within each of those areas, growth is concentrated at the high end, where high resolution, high dynamic range with low-light sensitivity, and global shutter technology are needed. Tower's global shutter technology, when paired with wafer-to-wafer hybrid bonding, delivers top-tier performance for low noise, high sensitivity, and high resolution. On top of that, we're working on an ultra-high density in-pixel capacitor to achieve best-in-class dynamic range, particularly for automotive applications. During the past quarter, we secured a second high-performance automotive product win. Notably, we have achieved full qualification with the next generation high-end video sensor at a leading high-end photography camera maker, and we're waiting on their product launch. Now for utilization. In the first quarter, Fab 2 utilization came in at roughly 60% as SiPho and SiGe qualifications proceed. Fab 3 operated at 80% utilization. That figure ended up somewhat suppressed since we added more SiPho and SiGe processes. We expect utilization along with output to rise once more during the second quarter. Utilization at Fab 5 reached 75%. Fab 7 is still fully utilized, well past the 85% utilization model we apply. Utilization for Fab 9 was 80%. Having said that, I'd now like to pass things along to our CFO, Mr. Oren Shirazi. Oren, go ahead.
Appreciate that, Russell. Morning, everybody. Our Q1 2026 financial results were released earlier today. What I'll do now is take you through the main takeaways from those numbers, along with the balance sheet and what we're putting into CapEx. Beginning on the P&L side: for the first quarter of 2026, revenue landed at $414 million, representing 15% growth compared with a year earlier and an increase over the $358 million recorded in the first quarter of 2025. When it came to gross profit in Q1 2026, we got to $111 million — that's a 52% jump relative to the first quarter of 2025. As for operating profit, it was $65 million, an increase of 96% year-over-year. On the income tax expense line in the P&L, you'll see $6.5 million, which works out to a 9% effective tax rate on an all-in basis. That's more favorable than what our model has baked in, because once Pillar Two rules kick in, we're estimating all-in tax rates north of 15%. The reason for that is a non-recurring income tax benefit we booked during the first quarter of 2026 related to TPSCo, our affiliate in Japan. Turning to net profit: for the first quarter of 2026, it came in at $65 million. That's an increase of 62%, or $25 million, compared with $40 million of net profit in the first quarter of 2025. The net margins behind those figures were 16%, versus 11% in the first quarter of 2025. For the first quarter of 2026, basic earnings per share reached $0.58, while diluted earnings per share were $0.57 — increases of 61% and 63%, respectively, compared with the same period a year earlier. As previously announced, we collected $290 million in prepayments from SiPho customers to reserve capacity for 2027; on the balance sheet at the close of March 2026, those prepayments are recorded as short-term and long-term customer advances, and within the Q1 2026 cash flow report they appear as cash from operating activities. Turning to the balance sheet, it continues to be solid, as the following indicators and financial ratios make clear. As of the end of March 2026, total assets stood at $3.7 billion, consisting primarily of $1.5 billion in net fixed assets — mostly fab machinery — along with $2 billion in current assets. Our current assets ratio is quite robust, sitting at roughly 5.6x. Meanwhile, shareholders' equity reached an all-time high of $3 billion as of the close of March 2026. Additional confirmation of our solid balance sheet and financial standing comes from Standard & Poor Maalot — a company wholly owned by S&P Global Rating — which, on May 5, 2026, completed its yearly rating assessment of our company, confirming the ilAA rating once more and raising the company's outlook from stable to positive. Let me now walk through what we've been doing on the hedging front. On the yen: TPS Co books the bulk of its sales in yen, and nearly all of its expenses are also yen-denominated, so for the greater part of our Japanese operations and business we enjoy a natural hedge. To address some of what remains of that yen exposure, we use zero-cost cylinder transactions as a way to hedge currency movements. Even if the yen's exchange rate versus the U.S. dollar swings, the effect on our margins stays small. Turning to the Israeli shekel: we bring in no revenue in that currency at all, yet since a portion of our Israel-based costs is denominated in shekels, we likewise hedge a substantial portion of that currency risk, employing zero-cost cylinder transactions to bring that exposure down. While fluctuations in the shekel-to-dollar rate do occur, their impact on our margins tends to stay contained—and that's precisely the pattern from recent quarters, when the Israeli currency appreciated quite a bit against the greenback. Shifting now to our capital expenditure roadmap: as we've shared over the past few quarters, in order to keep up with rising SiPho and SiGe demand, we're executing a $920 million program aimed at expanding both capacity and capability for SiGe and SiPho across our 8 in fabs located in Israel, Newport Beach, Texas, along with our 12 in Uozu fab in Japan. Regarding the purchase orders that have been issued, the qualification of both technology and processes, and the ramp plan, this investment remains on schedule. Of that $920 million in CapEx mentioned earlier, roughly 40% has already been disbursed and appears within our cash flow from investing activities for previously reported periods. The other 60%, meanwhile, should be disbursed across 2026 and 2027. Every bit of this CapEx — both the current and future investments — is completely captured in the model we shared back in February 2026. According to that model, our targets are $2.8 billion in annual revenue, $1.12 billion in annual gross profits, $900 million in annual operating profits, and $750 million in annual net profits. That wraps up my prepared remarks. I'll hand things back over to the operator now so we can open the line for your questions. Operator?
Yes, I appreciate you fielding my question. There are a few things I'd like to ask. Russell, when you step back and consider the broader picture along with the longer-term trends, could you somehow assist us in grasping how the possibilities for Tower in connection with silicon photonics and silicon germanium shift as we move away from a pluggable transceiver toward optical? Put differently, in what manner does your content evolve as this change is made from a conventional pluggable transceiver to optical?
As I said, moving ahead, when it comes to the content itself, I mean, lasers will of course still be required. We're pushing integrated lasers. We're pushing various material modulators to reach higher speeds. Also pushing, as I noted earlier, Through-Silicon Via so that some 3D packaging can be accomplished and the silicon photonics can be linked to the electrical ICs. When it comes to XPO, a lot more channels are simply being handled. When it comes to MPO, once again, a lot more channels are being handled, which means greater modulation.
Absolutely. The reason I raised this is that from where those of us on the investment side sit, we watch optical transceiver production changing, and various foundries are pursuing distinct approaches, so I simply wanted to get a clearer picture of whether Tower has emerged as a top-tier foundry and a partner for pluggable transceiver production. How does that shift once we get into optical, where other technologies come into play?
Like I said, our belief is that what we currently have in pluggables will carry over well into NPO and will also reach into CPO, though that is still a few years off. Looking at SiPhO ports on their own, it's quite fascinating. Starting in 2025, the number of SiPhO ports — and this comes from a LightCounting report — the number of SiPhO ports stood at 30 million, climbing by 2028 to 137 million SiPhO ports. Meanwhile, total ports rise from 90 million to 205. This reflects, you know, the data center, its build-up, build-out, and so on. The number of ports grows far, far larger. Pluggables aren't disappearing whatsoever. Pluggables will remain very robust at minimum through 2030. As we said in the script, we anticipate that the earliest item to ramp at a faster pace is near package optics, where we currently hold multiple design wins and noted that we have meaningful volume that should ship during 2026 and 2027. We also have, I'd say, a quite strong position in the XPO that Andy pointed to at the OFC, and as I noted, we had two solid demonstrators at OFC built on our SiPho. If you look once more, SiPho ports grow 4.5x between 2025 and 2028, and all of that port growth gets added between 2028 and 2030. I believe we are, by a wide margin, the leader in pluggables. I can't see any reason that would shift. Over the near term, the XPO and MPO form factors ought to transition smoothly, with Tower holding onto leadership.
Great, thanks for those details. Let me just ask one quick follow-up. I realize the investment community has its eye on silicon photonics, and I appreciate you laying out the revenue opportunities, particularly what's already under contract for 2027. Can we think in any way about the proportion between SiPho and silicon germanium? Silicon germanium also gets used in transceivers, after all. If your visibility into SiPho has been extended, would that in turn give you visibility into SiGe, and is there some ratio we could apply to get a handle on the SiGe opportunities you're targeting?
Yeah, that's right — the two go very much hand in hand. As port counts rise and the shift toward SiPho accelerates, then clearly SiPho expands. When port counts rise, demand for TIAs and drivers scales up in roughly the same proportion. What sets these two apart relates to how SiPho technology has progressed, plus the fact that while both carry very attractive margin requirements, the margin SiPho commands is greater than what SiGe delivers. Revenue expansion for SiGe trails the growth seen in SiPho revenue. In terms of unit volumes, the two move together almost step for step.
Understood, thanks. I'll rejoin the line.
Hello there, Cody.
Yeah. Appreciate it, everyone. Hi, folks. Thanks for fielding my questions, and congratulations on what was simply an outstanding quarter along with your guidance. Russell, perhaps I could receive a brief point of clarity regarding that $1.3 billion commitment. Pardon me. Does that cover wafers being delivered during fiscal 2027, or does it refer to wafers begun that would then push delivery into 2028? I'm only attempting to make my model accurate.
Wafers delivered in 2027. Regarding what I mentioned concerning a greater volume for 2028, that means wafers delivered in 2028. Once more, I believe it's extremely important to point out, even though I did state it in the script, that the $1.3 billion represents a contractual commitment. It isn't even with those customers from whom and with whom we hold that contractual commitment. It does not represent their complete volume demand. As for that $1.3 billion, if you examine it, where were we in 2025? I believe roughly $230 million. That's a two year span from $230 million to $1.3 billion. That $1.3 billion isn't what we are forecasting for SiPho in 2027.
Okay.
To put it plainly, yes — we're projecting a much bigger number, and that applies to the wafers we ship.
So that kind of leads into what I'd ask next. Given how much visibility you have, how much would you need before you'd feel at ease raising the long-term model you just raised?
Cody, that's a really good question. You referred to a long-term model — that was a model for 2028.
Right.
What I have in mind is more of a midterm model, I'd say. Our focus for 2027, as I mentioned, involves boosting 300 mm SiPho capacity — that's the strategic priority — and alongside it, naturally, comes 300 mm SiGe capacity too. At this point, no model factors that in. Visibility, in my view, is going to be on the shorter side. As for when a model gets revised upward to bigger figures, my expectation is that it happens over the coming quarters.
Alright, that's great, Russell. I appreciate you clearing that up and fixing the correction on—
That's a solid question, Cody. Thanks for asking.
Over the long haul compared with the medium term. I appreciate it. And then, if I could, a final item. How do you think about bringing indium phosphide lasers into an integrated setup? There has, I'm aware, been considerable worry across the industry regarding how dependable an integrated laser is as the shift to CPO happens, you know, when measured against the option of sourcing a laser externally. That, I understand, happens to be one of your strong points. Could you perhaps walk through what you've managed to accomplish on the reliability front?
We don't see any basis to claim that a laser which is integrated is less dependable, and also not any basis to claim it is more dependable, than one that is discrete. When it comes to reliability, what we mostly care about is simply the integration itself — getting the laser robustly integrated into the silicon photonics IC. So there isn't any possible drawback to combining those two, and my sense is that right now this is quite seamless. The work in that area, done alongside OpenLight and its OpenLight platform, is progressing very well. From where we sit, I'd say we hold a very optimistic view regarding the integrated laser, and on top of that regarding an indium phosphide integrated modulator aimed at 400 G. Both of those can be supported by our platforms, correct? Should any given customer prefer a form factor using lasers that are external, that's fine. Should they choose to proceed with us on the integrated laser, that's just as fine. So regarding whichever path they want to take, we're more or less agnostic — except for the reality that we are pressing quite hard on very advanced 400 G modulation. When you examine advanced modulators, indium phosphide delivers enormous form factor benefits.
Russell, if that's alright, given that you raised the topic, I suppose I'll slip in a final question here. Regarding the modulation approaches you referenced in the press release that followed your earnings release, you listed quite a broad range of modulation technologies that you're working on. Could you rank or give preference to any single one among them as the superior option for Tower across the medium to longer horizon?
No, my view is that we have equal capability across each of them. You know, thin film lithium niobate can be architected in several distinct manners. I won't claim it's easy, but it's possible to build a standalone lithium niobate modulator on its own wafer and place it alongside the PIC, or you can incorporate it as a chiplet on the PIC; both form factors are being pursued by us with various customers. When it comes to indium phosphide, clearly that one is, you know, integrated on the PIC. We lean somewhat neutral regarding which option a customer might choose. That said, once you reach many channels, indium phosphide genuinely offers an advantage in terms of form factor.
Okay, that's great. Thanks, Russell.
I appreciate that, Cody. Those were really solid questions. Thanks.
Thank you.
Hey, Richard.
Russell, hello there. How are you doing?
I'm doing very well, thanks.
Great. Sorry about the background noise. I’m about to get on a flight. If it gets too loud, I’ll try to join by phone afterward. I think I only have 2 main questions. The first concerns your capacity expansion in Japan. It’s sort of a 2-part question. First, how should we think about the possible revenue capacity scale for silicon photonics once you have at least the initial parts of the tranche—the tranches I’m sure you’re building there. Do you have any concerns about running short on silicon photonics capacity before that Japan expansion begins?
Great question. We are looking and hoping and believe that, very near term, we will get METI approval. Obviously, we’ve already started with contractors to complete planning, to get everything done. As soon as METI approval comes, we’ll most likely apply for permits; that’s probably about a 1.5-year kind of period from breaking ground to having a facility that can accept tools and begin ramping those tools. Where does that leave us? In the best case, that leaves us in the first half of 2028 for that capacity. When you say concerns—well, concern is an interesting word. Concerns can also be excitement, depending on how you look at it. The excitement right now is growing with an existing footprint, our silicon photonics capability in Fab 7, which is the current factory. We have approved capital spending for that for its first phase, which we have not announced publicly. I don’t want to discuss the figures at this point. We probably will within 2, 3 quarters, consistent with what Cody asked about updating a financial model. Then we have the ability to go further before the shell is finished by using an existing factory inside the TPSCo complex. It was one we announced a couple of years ago that we had closed, and that was the Arai factory, which Nuvoton Winbond owns. To place a certain number of tools and a very specific set of tools in that factory to grow capacity fairly quickly within an existing footprint, so to speak. Our plan is not to miss any upside, or not—our plan is certainly to satisfy customers’ demand within 2027 while developing, meaning building, the shell, which is substantially large. It would be an interim step. The good thing about the interim step is that any tools we purchase apply to the whole build. Even things we’d initially place in another factory would, over time, come back into the existing geography. Well, not the existing, but the existing plus built-out geography. Hopefully that answer wasn’t too obtuse. There’s a lot of explaining in it, but did that make sense, Richard?
That covers it. I'll go back over what you said and examine it in more depth, though it was quite thorough and useful. For my follow-up — and after asking this I'll mute myself, then we'll check what you come back with — regarding the news about your 400 G per lane silicon modulator for silicon photonics that you put out together with Coherent.
Yes.
I've got two things to ask. To start, is it possible to apply this tech, or something akin to it, when working with different clients? When you consider the 400 G per lane generation, what share of your business is going to come from silicon, from TFLN, from indium phosphide, or from some other modulator technology? Thanks.
Regarding your first question, the 400 G modulator we put out a press release about with Coherent definitely came from a Coherent design. Whatever we know, along with our know-how and our IP, can be used by any customer. The performance of that particular modulator stemmed from Coherent's design being paired with our platform. Hopefully that's clear — it isn't something we're able to hand over to anybody else. That IP belongs to Coherent. Since Coherent designed it around our platform, other customers who have or might potentially obtain the same capability as Coherent could create a design on our platform. As things turned out, Coherent was the one that did it, and Coherent was the one with the design capability to pull it off. That's the first point. Second, I believe what you're really asking is which technologies I expect to be adopted first. My own personal opinion — and it is just mine, since I'm definitely not a market analyst — is that lithium niobate will arrive with considerable strength for a single generation. My feeling is it won't endure across many generations. I expect the shift will be toward an indium phosphide.
Great. Thank you, Russell.
Thank you.
Hi. I wanted to ask if you might discuss gross margins a bit. They went up so much in the first quarter, which seems unexpected since revenues declined sequentially. Could you explain how the margins rose? Was it mostly product mix, higher prices? What drove that result?
Yes. Essentially, this aligns with the model we released during the previous quarter. In that model, you will notice we projected incremental revenue contributing at 59% to gross profits on top of a baseline, which stood at 20% a year earlier. As shown in this report, Q1 2025 gross profit was 20%; we assumed in our model that the incremental portion would arrive at 59%, and the fully built out model will get to 39%; this quarter, we have already hit 27%. As you mentioned, it's quite good that we are already up from the 20% to 27%. That represents the linear progression we anticipate toward 39% when we reach the 2.8.
So essentially, it comes down to moving more of the newer items that carry better margins.
Correct. That's right.
All right. All right. Let's circle back to the technology side. Is there anything that worries you, such as whether your capacity could run into trouble obtaining components or end up limited? I understand indium phosphide is difficult to source and is in short supply. Will that alter anything, or will customers simply move to a different option?
It's true that indium phosphide is currently tight. Where we sit, the end laser product isn't necessarily what's limiting us — the starting material is what's limiting us. No, I think we've got solid plans and solid supply chain efforts in place, so whatever bottlenecks may exist, we're addressing them. As these products scale up, and they ought to be scaling up on the indium phosphide integrated laser, it should be scaling this year, I believe we'll be well positioned to satisfy the demand.
All right. Thanks, that answers everything I had.
Yes. Hello, and I appreciate you fielding this. Russell, I've got a pair of items for you. First, I'm simply wondering how you'd size your current position in silicon photonics, and where you expect that position to sit a couple of years out, taking into account the longer-term contractual agreements you hold with your customers. My reason for asking is that GlobalFoundries is scaling up and TSMC is bringing their COUPE platforms up to speed. I'd just like to know where you believe your share could move between now and two years from now. And then I have one more thing.
What an intriguing thing to ask. Some others out there assert extremely large shares. To me, that seems entirely implausible. I won't put a number on share right now, yet I'd say we are clearly number one, and number one in silicon photonics today by a wide margin. Nothing suggests to me that this ought to shift.
Understood, Russell. I just want to ask a bit more about the tech angle for advanced packaging. Obviously, you've worked with hybrid bonding on CMOS image sensors. At what point do you expect that to carry over into silicon photonics? Is that a genuine edge near term, or is it more of a longer-term narrative?
I'd say it's a quite significant edge over the mid to long term. For this year, it isn't that important.
Got it. Thanks, Russell.
Thank you. Good question.
Yes, thanks. I've got a few quick follow-ups. RF mobile declined by nearly 36% from one quarter to the next. Clearly, that creates a low base, and for the rest of the year I'm anticipating growth. For 2026 compared with 2025, what's the right way to think about RF mobile? I have one more question after that.
Yeah. What I can say is that SOI did see exceptional growth in Q4 25 relative to earlier quarters. On a year-over-year basis, there was no reduction; it was simply a particular ramp during Q4 25.
Okay.
Russell will address 2026. Yeah.
Like I mentioned, we are stepping back from our 200 mm RF SOI, and our customers are doing the same—they're shifting toward newer design wins on 300 mm. The pipeline for design wins looks incredibly robust, and when I examine the forecast spanning multiple years, what I observe is that RF SOI overall is growing on 200 mm relative to 300 mm, yet the 200 mm business has nearly disappeared. Over the near term, not all of it has made the move to 300 mm just yet. You're right that it declined, and, well, when I consider the entire year—let me check to get this right—yes, I would anticipate the full year coming in lower than the prior year, including at 300 mm. My expectation is that during 2027 and 2028, we'll see record growth for RF SOI on 300 mm. A number of design wins have already been granted to us, though they relate to handset models launching for the first time in 2028. That's simply how this industry operates, and as a result, purchasing for those would need to begin in the third quarter of 2027.
Understood. All right. One more question. Looking at the broader industry, analog foundry, and excluding RF infrastructure, it appears that utilization across the industry is on the rise. A few of your competitors have mentioned higher prices, higher wafer prices, on a like-for-like basis. What I'd like to ask you, to get a clearer picture, is how we should separate out the mix of, clearly, RF infrastructure, which has a faster growth rate and better margin. Beyond RF infrastructure, should we likewise expect that some pricing power is accruing to you?
In my view, pricing power comes specifically from owning best-in-class platforms. When you offer something that lets a customer capture advantages by adopting your platform, you're able to command a premium for it. We're not the kind of company that takes pleasure in raising prices across the board simply because capacity is tight. Honestly, I believe that isn't a partnership approach. There are occasions when customers may end up on allocation due to demand being too strong. In terms of relationships and integrity, I don't think it's necessarily sound practice to lift prices merely because demand is constrained. Broadly speaking, our pricing gains come with new-generation platforms for new technologies. Each time we introduce a new platform, its starting price point is definitely above that of the platform currently being sold for that technology. I referenced earlier that in 200 mm power management, a 13% price increase was realized for, you know, starting, I believe, this quarter or across this year. That really came from assessing the platform's own value against possible price reductions granted over prior years, and it genuinely required a reset. It wasn't driven by seeing a capacity constraint and, you know, compelling a customer who has no alternative. It was essentially a matter of reassessing the worth of what we sell and what the pricing ought to be.
Excellent. That was quite informative. I appreciate it.
Hey there, everyone. I appreciate you fielding my follow-up questions. Only a handful of brief ones. Russell, perhaps from a thematic angle, could you discuss the shift toward CPO along with the industry's bigger integrated participant in the Taiwan Semi market — their CMOS integration on the SiPho side — plus their advantage as a single-source provider for a company such as NVIDIA over the long haul, and in what way does Tower go up against that over an extended period using your hybrid approach?
Yes. Regarding a portion of what you asked, you're essentially wanting me to act as a marketer for TSMC, and that truly isn't what I'm here to do. Seriously though, like I mentioned, NPO and XPO pluggable will remain the dominant source of demand, at minimum through the start of the coming decade, while CPO gains more momentum. When it comes to CPO, my view is that the significant advantage we bring comes from having PICs that deliver exceptional value due to how well the PIC performs. Theoretically speaking, nothing would stop TSMC from purchasing our PIC for their CoWoS provided our PIC significantly outperformed all alternatives. There's one item — actually, several items you brought up that are completely correct. I mean, with TSMC functioning as a one-stop shop, I doubt anyone can rival TSMC given what they accomplish in extreme deep digital content. That's simply not something in our arsenal; it isn't something we could challenge them on or attempt to pursue or consider. When it comes to offering modulation with three, five, and other materials that outperform what anybody else is capable of, that's precisely where our contribution would come in. That's among the major points I highlighted regarding our CPO roadmap — a genuinely intense emphasis on modulation.
Okay, that’s great. Thank you, Russell. And then, Oren, one last thing—could you give us your perspective on what to expect going forward for the tax rate, interest income, non-controlling interest lines, and any other miscellaneous income statement items?
On financing, that’s a separate income line, roughly $10 million in total, and my expectation is it stays flat. As for taxes, as I mentioned during my prepared comments, on a run-rate basis—meaning under a normal model—it ought to be no lower than 15% due to Pillar Two, which applies across every region, and could go as high as about 18%, since certain regions exceed 15%, for example the U.S. at 21%, Japan, and Italy. My assumption would be a tax range of 15%-18%.
What about that non-controlling interest line?
It ought to look more or less the way it does at present. What I mean is, during this quarter we saw a particular gain in that area, and I brought it up since it stems from the Japan-related income tax benefit, which is a non-recurring item. Once you take that out, it should resemble the prior period, so the figure was a fairly tiny one, correct? Our opinion isn't affected by it.
Cody, if I may, I'd like to make clear what I meant. When I say there's nothing stopping TSMC from purchasing our PICs, what I mean by that is that we turn into the reference design for the big integrators, and that would be one of the things they would use alongside TSMC.
Russell, could you say a bit more about that, regarding the connection you've built lately with NVIDIA at 1.6 T? Clearly, they've relied on Taiwan Semi throughout their past, yet going forward they're also working with you as a partner, right?
Yes. Cody, truthfully, I genuinely would like to be able to do so. I'm not at liberty to discuss particular customer programs beyond the press release we issued alongside NVIDIA. It came through pretty plainly that he described us as a partner in development.
Okay, that sounds good. Many thanks, everybody.
Yeah, thank you so much. We genuinely value the confidence and backing you keep giving us. My gratitude goes out to our teams—the Tower teams across the globe—whose commitment has enabled the advances we just described and the advances we anticipate achieving throughout the coming 12 months and beyond. That is genuinely thrilling. Achievement breeds achievement; the more wins we achieve alongside our customers, the more enthusiastic they become, and the more that energizes our own workforce, and it turns into a collaboration that is truly remarkable. I brought up Coherent as an example, yet we enjoy outstanding connections with numerous optical customers. I'd say with every one of them. Having that is genuinely thrilling, since when a customer feels truly pleased with you and how you perform, those exchanges turn into something highly motivating for next-generation work, and developments two generations out. I'm grateful to our teams, and I'm sincerely grateful to our customers. Regarding the equity stakeholders, I genuinely value the confidence and backing you keep giving, and we'd be delighted to see you at any of the various events we've lined up for the months ahead. On May 18th, we will take part in the 27th Annual Oppenheimer Israeli Conference in Tel Aviv. On May 27th, we will be at the 54th Annual TD Cowen Technology Media and Telecom Conference in New York. On May 28th, we will take part in the 23rd Annual Craig-Hallum Institutional Investor Conference in Minneapolis. On June 9th, we will be at the 2026 Mitsui Global Tech Conference in New York. Should you be able to attend any of those, we would truly enjoy getting together with you. Also, as ever, our investor relations team is quite willing to take calls and arrange video calls with anybody who wants additional updates and insight into what we consider a tremendously exciting endeavor and an exceptionally abundant roadmap. Thank you so very much.
That wraps up the conference call for today. Thanks for joining us. Everyone is free to disconnect at this point. Have a great 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.