Home / Transcripts / Tower Semiconductor Ltd. (TSEM) · February 11, 2026

Tower Semiconductor Ltd. (TSEM) Earnings Call Transcript

February 11, 2026

NASDAQ US Information Technology Semiconductors and Semiconductor Equipment earnings 62 min

Earnings Call Speaker Segments

Noit Levi-Karoubi executive
#1

Thanks. Hello, and I appreciate all of you being with us here. This is Tower Semiconductor's conference call covering the financial results for the fourth quarter and the entire 2025 year. Joining us are our Chief Executive Officer, Mr. Russell Ellwanger, and our Chief Financial Officer, Mr. Oren Shirazi. Prior to starting, I want to point out that some remarks on today's call might be forward-looking and could be affected by risks and uncertainties that would 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 Israeli 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 fourth quarter and full year 2025 have been put together following U.S. GAAP. A number of the items shown might contain non-GAAP financial measures as outlined by SEC Regulation G. Reconciliations to GAAP numbers and complete explanations appear in today's press release and financial tables; for your reference, a supporting slide deck can be accessed on our website and is part of this webcast. Having said that, I'll hand things over to our CEO, Mr. Russell Ellwanger. Russell?

Russell Ellwanger executive
#2

Thanks, Noit, and hello everyone. I appreciate you all being on the call with us today. It's a real pleasure to walk you through what we delivered in the fourth quarter and across the whole of 2025. On top of that, we're very excited to show you how those numbers have reshaped the financial targets we set for ourselves and pulled forward when we expect to reach them. The revised financial model we're about to walk through comes out of partnerships we already had in place with our lead customers — relationships that have deepened into supplier-customer technical alliances built on a great deal of trust. We closed Q4 of 2025 with $440 million in company revenue, which works out to 11% growth quarter-over-quarter and 14% growth year-over-year, so we hit the sequential quarterly growth goal we set at the start of the year. Beyond the top line, we also delivered bottom line growth all year long. Net profit in the fourth quarter came to $80 million, or an 18% net margin, compared with 11% in Q1 '25, 13% in Q2 '25, and 14% in Q3 — a sign that value-based growth is being powered by an enriched technology mix. Between Q1 and Q4 of 2025, revenue grew by $82 million, and $40 million of that dropped through to net profit — nearly 50%, 48.78% to be precise — and that's because of the high value of the incremental Photonics revenue. For the full year, revenue reached $1.566 billion, which is $130 million, or 9%, more than 2024 revenue. Now, to go through our 2025 revenue breakdown and talk about the key trends, please turn to Slides 5 and 6 as referenced. We grew year-over-year across our key technology platforms — power management, image sensors, and 300mm RFSOI — and on top of that, record achievements and unprecedented growth in our market-leading optical transceiver offerings, silicon germanium, and SiPho advanced platforms have pushed us into a favored and unique position, both fueling our growth for 2026 and giving us the ability to redefine our financial model, which I'll lay out at the end of my remarks. RF infrastructure posted a 75% revenue increase, 2025 over 2024, making it our fastest-growing application in '25, driven by hyperscalers rapidly adopting silicon photonics in 800G and 1.6T pluggable transceivers. Silicon germanium and silicon photonics revenues accounted for 27% of our corporate revenues, coming in at $421 million, up from $241 million, or 17%, in 2024. SiPho revenues by themselves were $228 million in 2025, up from $106 million in 2024. Looking specifically at the fourth quarter, RF infrastructure revenues made up 32% of corporate revenue, with SiPho reaching $95 million, or a $380 million annual run rate. That figure includes some non-wafer NRE to support future development for Gen+1 and Gen+2. As we highlighted in our recent announcement with NVIDIA, the insatiable demand for compute bandwidth in both scale-up and scale-out architectures, combined with Tower's exceptional ability to scale capacity flawlessly alongside our customer, has made 1.6 terabyte per second the fastest-growing silicon photonics node in the industry to date, with Tower by far the majority supplier of 1.6T silicon PICs. Our partnership with NVIDIA—just like every one of our direct module customers—highlights how dedicated we are to providing top-tier technology, along with the production flexibility needed to keep up with such an extraordinary demand trajectory. Beyond Fab 3 Newport Beach, during the last year we brought silicon photonics production up successfully at Fab 9 San Antonio, at Fab 7 in Uozu, Japan, and we remain on schedule to ship the first production—a very large SiPho ramp occurring in 2026—from Fab 2 Migdal Haemek as well. Because customer demand has come in even stronger than what we knew at the time of our previous quarterly release, we've raised our 2026 CapEx plan, and with numerous customers asking to sign capacity reservation agreements extending through 2028, this lets our customers pass firm commitments along to their own customers, having secured their supply. When it comes to next-generation 400-gigabit per lane, we keep making solid progress with heterogeneously integrated indium phosphide on silicon, plus other material systems. We're playing a key role, working alongside our lead customers to determine which material systems get selected, honing the flow, and therefore guaranteeing manufacturability readiness plus the ability to ramp immediately when the 3.2T market arrives. We likewise view co-packaged optics as a substantially incremental opportunity for us over the coming years, as optics gains adoption along with scale-up interconnects, as well as XPU to high-bandwidth memory interconnects that remain largely copper today. During Q4 '25, we announced that our mature 300 millimeter wafer bonding technology is being expanded to enable wafer to wafer integration of silicon photonics ICs and silicon germanium electrical ICs. We're also continuing to work with several customers on dense wavelength division multiplexing laser sources, which represent a critical component of many CPO implementations and can significantly expand our served optical market by now including the laser source. Past optical transceivers, our silicon photonics platform remains the technology of choice for physical AI applications, especially frequency modulated continuous wave LIDAR. Before CES, 2 of our FMCW LIDAR partners, AVA and LightIC, publicly announced they are collaborating with us to bring disruptive products to market. The demonstrated robustness of our silicon photonics platform, backed by many tens of thousands of high-yielding, high-quality wafers shipped so far, is letting silicon photonics capture growing share within the LIDAR market, opening up new automotive and robotics opportunities. Our silicon germanium platform posted strong growth year-over-year in 2025 of 43%, and it stays the optimal platform solution for low-power, low-latency, high-performance components like drivers, trans-impedance amplifiers for pluggables, LPOs, and active copper and active optical cables. Together with our silicon photonics production, our silicon germanium platform is now running at high volumes across Fab 3 Newport Beach, Fab 9 San Antonio, Fab 2 Migdal Haemek, and we've shipped 300 millimeter prototypes from Fab 7 in Uozu. RF mobile accounted for 23% of our 2025 corporate revenue and 24% of our Q4 25 revenues. 300mm RFSOI rose 5.5%, while RF mobile overall fell 15% year-over-year. That mostly stems from us teaming up early with our customer partners to deliberately cut back on lower margin controller products, shifting instead toward a richer optical and RF mix within the fabs, and it's also tied to the front-end module market moving away from 200 millimeter toward the greater digital content that's better handled at more advanced nodes on 300 millimeter. Our newest technology, the one we showed everyone last quarter, delivers a big gain in our <<inaudible>> versus what competitors have, along with fewer layers, which means more total value for every dollar a customer spends, and it keeps getting picked up enthusiastically. Lead customers have called it best-in-class and are getting ready to ramp into high volumes. Everywhere we look, design win momentum remains powerful, setting up our 300mm RFSOI platform for lasting secular growth. During 2025, we landed major wins — specifically 3 of the top 4 Tier 1 RF front-end module providers. One of them has started production, and all of them are planning a strong ramp in 2027 on the way to meaningful revenue volumes in 2028. Power Management climbed 20% year-over-year, showing solid year-over-year revenue growth across both 200 millimeter and 300 millimeter offerings, and it made up 16% of our 2025 corporate revenues and 15% of our Q4 '25 revenues.

Russell Ellwanger executive
#3

Now regarding 300 millimeter, that covers the ramp in production for the Tier 1 handset envelope tracker that we brought up earlier, and we expect to pick up additional share going forward. On the whole, our revenue growth in 300 millimeter power has significantly exceeded the expansion seen both in the power market and in the mobile handset sector, which shows how strong our offering is and how much share we're gaining in this important area. Sensors and Displays climbed 10% versus the same period last year, making up 16% of our 2025 corporate revenue and 15% of fourth quarter revenue. The machine vision sector has been robust and keeps demonstrating that strength, with new advanced offerings moving into production while our current products continue to win share. In addition, we expect our first ramp in the AR display segment—excuse me, in the AR display area—with our silicon backplane for OLED-on-silicon, which entered production in the quarter that just closed. We're watching this early adoption, and the wider market it represents, closely and with optimism, since it could bring meaningful value to Tower in the years ahead. Mixed-signal CMOS represented 7% of our 2025 corporate revenues, and Discrete accounted for 11%. Versus the previous year, those two numbers each fell by 18% and 14%, a trend that confirms our approach of growing through value and creates additional space for the higher-margin platform—the one with the highest margins—to take the place of these two application sets. Now on capacity expansion: in our November '25 earnings release, we stated that we would direct additional investment toward silicon photonics and silicon germanium growth, aiming to triple SiPho capacity relative to our targeted Q4 '25 silicon photonics actual shipments, and we said this would be online to begin silicon starts in the second half of 2026. Because demand keeps rising without letup, as of September 2023 we're here today to announce another $270 million in CapEx spending, which adds to the $650 million capacity expansion plan we already disclosed. Once that capacity is fully in place, the aim is for it to produce capacity growth exceeding 5x what we actually shipped monthly in wafers during the fourth quarter — that's silicon photonics wafer shipments — and you can compare that to the 3x figure we gave in our Q3 public release. Beyond that, over 70% of that total SiPho capacity is either already booked or is in the process of being booked through 2028, with customer prepayments firmly supporting it. Now to the fourth quarter, where utilization stood: Fab 2 operated at approximately 60% utilization, as we're currently in the final stages of qualifying silicon germanium and silicon photonics capacity across a variety of flows. Fab 3 remained exactly at our 85% model for full utilization, and we continue to add more capacity to enhance silicon photonics capability. At Fab 5, utilization reached 75%. Fab 7 was operating at complete capacity, and that sits well north of the 85% threshold that serves as our planning assumption. Turning to Fab 9, it was at 65% utilization, and right now it's working through the ramp stage for silicon photonics as well as silicon germanium. During September 2023, as we noted in the press release, Intel made clear it has no plans to fulfill its obligations under the Fab 11X agreement from that period. A mediation effort is currently in progress. Any volume that has moved, or is in the middle of moving, over to Fab 11X was initially qualified at Fab 7, our 300 millimeter facility located in Japan. Those customers are now receiving support routed to that plant in Japan. On the guidance front, for Q1 of 2026 we're targeting revenue at the midpoint of $412 million, give or take 5%, representing a 15% increase compared with the start of 2025; across 2026, our goal is for revenue and profitability to each climb quarter over quarter. Because of the robust corporate ecosystem we've established alongside deeply trusted alliances with customer partners, we're glad to present a refreshed financial model; this new version highlights our strategy of growth driven by value. Please turn to Slide 7. We'll begin with the assumptions. Beyond the $920 million in CapEx already made public, nothing else is required to arrive at this model — no additional CapEx, no clean room space, no further funds of any kind. In this model, we assume 85% utilization of the capacity Tower owns. Intel Fab 11X is left out of this model. We arrive at $2.84 billion in revenue, producing a 39.4% gross margin and a 31.7% operating margin — a drop of 7.7 points from gross to operating margin — and it shows a business that is extremely efficient, and if it's not the number one performer, it certainly ranks among the top performers in our industry. Such efficiency isn't apparent only in margin dollars. It also reflects the pace at which decisions get made and executed. Speed is a differentiator that can be maintained. Net profit reaches $750 million, equivalent to 26.4% net profit margins. All tool and customer qualification should be entirely wrapped up at some point during 2026. So, and above all, our objective is to hit this model in calendar year 2028. With that, I'd like to pass the call to our CFO, Oren Shirazi. Oren, go ahead.

Oren Shirazi executive
#4

Hi, everyone. Our results came out today covering the fourth quarter of 2025 and the entire year, and we also published the balance sheet and cash flow statements. I'll walk through the standout points in those figures, and I'll also discuss the main things we're putting money into on the CapEx side. After that, I'll lay out our revised target financial model, which brings higher revenue and stronger profit margins than the version we gave you earlier. Starting with the P&L: during every quarter of 2025 our top line went up, and the pace picked up in the second half, producing a record $440 million of revenue in Q4 2025 — up 14% versus a year earlier and 11% better than the prior quarter. Gross profit for Q4 2025 came to $118 million, which is $25 million more, or 26%, than the preceding quarter; operating profit was $71 million, a 40% quarter-over-quarter improvement. Net profit for Q4 2025 landed at $80 million, an increase of $26 million — 49% — over the $54 million of net profit we posted the quarter before. Per share, basic earnings were $0.71 and diluted $0.70, versus the $0.48 basic and $0.47 diluted reported in the prior quarter. One point deserves attention: inside the income tax expenses line of the P&L, Q4 2025 included a nonrecurring tax benefit, which lowered our all-in effective tax rate to 2%. Looking toward 2026 and beyond, because of Pillar 2 regulation, we expect the all-in tax effective rate won't drop under 15% at any of our manufacturing sites. For the full year 2025, revenue totaled $1.57 billion, a 9% gain over the $144 billion booked in 2024. Gross profit for '25 was $364 million and operating profit was $194 million — the comparable 2024 numbers were $339 million and $191 million. Net profit for 2025 came to $220 million, equal to $1.97 basic and $1.94 diluted per share, while net profit in 2024 was $208 million. Now, the balance sheet. Ours is still quite robust, as these indicators and ratios show. At December 2025's close, total assets were above $3 billion, made up mainly of $1.5 billion of fixed assets — largely fab equipment — plus $1.7 billion of current assets. As for why other long-term assets rose versus prior periods, the primary cause is the prepayment for extending the Newport Beach Fab lease — which we previously told you about and repaid in November 2025, and which GAAP requires us to hold as an asset. Our current assets ratio is at a very healthy roughly 6.5x, and shareholders' equity finished December 2025 at a record $2.9 billion. Moving to hedging, let me explain what we do regarding currency. When it comes to the Japanese yen, TPSCo typically issues invoices to its customers in yen, and nearly all of its costs are yen-denominated too, which means that for the overwhelming majority of our Japanese operations and business, we enjoy a natural hedge. To cover part of the remaining yen exposure, we use zero-cost cylinder transactions as a hedge against currency fluctuations. Consequently, although the yen-dollar exchange rate may fluctuate, the impact on our margin remains constrained. The identical reasoning holds for the Israeli shekel. When it comes to that currency, we bring in no revenue denominated in shekels, yet because certain costs we incur in Israel are shekel-denominated, we similarly hedge a large portion of that currency risk by way of zero-cost cylinder transactions to lower this exposure. Thus, even when the shekel-dollar rate shifts, the impact on our margins remains constrained. Let's now shift to our CapEx investment plan and the way it shapes our financial model. In today's announcement, we disclosed that we are directing an additional $270 million in cash toward equipment-related capacity and capability investments in order to satisfy rising SiPho and SiGe demand. That pushes our overall cash investment in CapEx to $920 million, a figure that incorporates the $650 million we previously told you about earlier in 2025. Through this $920 million CapEx investment, we will grow capacity at our 8-inch fabs in Israel, Newport Beach, and Texas, as well as at our 12-inch Uozu Fab in Japan. Capability CapEx linked to advanced development and high-end RF technology projects accounts for a meaningful share of this CapEx plan. Roughly 28% of the $920 million we are committing to CapEx has already been paid out, while the remaining 72% of that $920 million is expected to be paid across 2026 and 2027. Let's now shift to the financial model. Set against the model we presented more than 2 years ago, these investments ought to deliver greater revenue and improved incremental margins, and with them in place, we are providing a refreshed target financial model that signals considerably stronger targets for revenue, profitability and margin. Bear in mind that this projection rests on a broad set of forward-looking assumptions regarding operations, business, and finances. One such assumption is that once the $920 million in equipment tools we are funding are installed and qualified, all of our fabs will operate at 85% utilization. We also assume that a modest decrease in the average selling price of existing products and/or flows will be offset by introductions of new products and/or flows. A further assumption is that our cost estimates will not move much from their current levels. And lastly, be aware that this model excludes Fab 11X capacity, revenue, and margin, as well as any additional fabs and/or extra capacity that has not been secured, set up, or publicly disclosed as of now. This builds on the $1.44 billion recorded in 2024. Using the slide's model as our reference point, we're aiming for $2.84 billion in annual revenue—that's a $1.27 billion jump, or 81%, versus what we actually posted for full year 2025. Turning to gross profit, we're projecting $1.12 billion, which is over three times the figure we reported for 2025. That works out to a gross margin of roughly 40%, and when you stack it against FY 2025 actuals, the extra revenue generates 59% incremental gross profit. Moving to operations, the number comes to $900 million in annual operating profit. That equals 4.6 times the operating profit we actually delivered in FY '25. It also yields a 32% operating margin, and that margin reflects 55% incremental operating profit margins coming from the added revenue in that same model-versus-FY 2025 comparison. Finally, net profit lands at $750 million, over three times the full year 2025 net profit, and it represents a 26% net margin, as Russell noted, along with 42% incremental net profit margins created by the incremental revenue when comparing the model to FY 2025 actual results. To close things out, if you set this updated financial model beside the one we showed over two years ago, gross profit, operating profit, and net profit are each substantially bigger—rising 50% and 60% respectively against the earlier model—largely fueled by a stronger SiPho and SiGe mix together with the added value we deliver to our customers. That's everything I had ready, and I'll hand the call back to the operator now so we can start taking your questions. In 2024, that figure stood at $1.44 billion.

Mehdi Hosseini analyst
#5

I've got a few things I'd like to ask. Russell, let's dig into what you announced this past Thursday — the expanded work with NVIDIA. That press release mentioned a module, and I'd like to get a clearer picture of what that actually means. Are you going to be building a transceiver for NVIDIA, or is "module" really pointing to something wider — a sign of bigger services you'd be delivering to this customer? I also have a second question.

Russell Ellwanger executive
#6

No — what we contribute on the module side involves the output parameters coming from our photonics, or from the TIA, or from the drivers, or from the pluggable, and likewise for the copper or optical cable. What the partnership really speaks to is alignment and requirements, both directly and via our module customers, plus a shared grasp of supply requirements and supply shipment commitments.

Mehdi Hosseini analyst
#7

All right. Regarding that fivefold expansion in capacity for silicon photonics and silicon germanium — does it take into account the extra demand coming from NVIDIA and its partners?

Russell Ellwanger executive
#8

Yeah, that figure covers overall demand. Actually, I wouldn't put it quite that way — it isn't strictly about overall demand. It responds to demand, yes, but what we're really talking about is the capacity we're actually putting in place. So consider the $380 million run rate we mentioned for Q4; subtract from it a modest portion of NRE, the exact size of which we don't disclose — that's the silicon wafers we delivered during the fourth quarter. Measured in capacity, we intend to have 5x that quantity of silicon wafers by the fourth quarter of 2026.

Mehdi Hosseini analyst
#9

Understood. Alright. Then regarding the power business line of yours — could you also be in a position to assist potential clients when it comes to high voltage, particularly since the upcoming generation of AI server racks is going to need 800-volt DC?

Russell Ellwanger executive
#10

We've got numerous road map efforts underway. Right now, an 800-volt platform on an IC is not something we possess. In Fabs we certainly have 800-volt capabilities, though not on an IC. That said, IC capabilities at higher voltage, both with SOI and without it, are something we do have.

Tavy Rosner analyst
#11

I just want to add to what was asked about NVIDIA; to be clear, you're not sending products straight to NVIDIA — instead they go out via resellers, who then pass your technology along to that company.

Russell Ellwanger executive
#12

Yes, that's right. When it comes to the photonics side of things, we don't deliver that straight to NVIDIA. Regarding details about projects or work involving NVIDIA, anything beyond what the PR explicitly mentioned is something I'm not free to discuss. However, on the current photonics — silicon-based photonics ICs — those are entirely designed and shipped by way of other module makers or integrators.

Tavy Rosner analyst
#13

All right, got it. Now, regarding CPU — I mean, you mentioned that opportunity. If I remember correctly from the prior quarter, though perhaps this came up in a separate discussion, you both talked about how lasers and power connectors let you bring more value to the ecosystem, and also that you're pursuing some R&D on the CPU itself, possibly via third-party packaging, so that eventually you'd have your own end-to-end solution?

Russell Ellwanger executive
#14

Packaging the CPU directly — no, that's not it. We are definitely developing several CPO architectures, and the XPU certainly would be, or could be, integrated into the CPO. As for what we're specifically engaged in at this moment — well, that isn't entirely accurate either. Yes. I mean, we are definitely collaborating with XPU manufacturers on CPO strategies.

Tavy Rosner analyst
#15

Got it. That makes sense. Then here's my final question. Regarding the extra CapEx being brought online, I believe I remember you mentioning that everything would be operational by the close of 2026. Could it possibly happen earlier, given a number of variables — some of which might be outside your control? Or are you confident it definitely won't be up and running before this year ends?

Russell Ellwanger executive
#16

We'll be carrying out the capacity qualification ramp across the whole year. So the largest share of that $920 million ought to be easily live — meaning fully qualified — no later than the third quarter, with additional growth also occurring during the first and second quarters. A few of our latest orders include tools scheduled to arrive in the second quarter as well. But here's what we've said, and what I just repeated: we anticipate and aim for everything to be fully qualified by December so that customer starts can proceed. For all of it to be fully qualified by December, the tools essentially need to show up prior to the third quarter's end, and typically around the middle of that third quarter. So that's the thing to keep in mind — whether it's linear or not, tools will be distributed. A portion has already come in, and the majority of this $920 million will be delivered from now through the middle of the third quarter.

Cody Acree analyst
#17

Nice work on the consistent, solid gains. Russell, would you mind sharing a bit more detail regarding what you anticipate from your silicon photonics business in '26 and '27, particularly given the 70% commitment you've already mentioned around prepaid? Your visibility seems like it ought to be quite strong over the coming couple of years.

Russell Ellwanger executive
#18

Sure, absolutely. Demand exists. We're certainly quite conscious of that demand. At present — if what you're really asking about is the ramp profile, then that ramp profile comes down purely to operational execution right now. I'd say a bit of technical execution remains. Certain flows still need qualifying, whether that's San Antonio or Migdal Haemek — those aren't qualified yet, and they're part of the first order — actually not the first order, but only qualified over at Newport Beach, since that was the fab where essentially all of this development took place. So additional technical work remains to be completed, though for the most part the technical side sits behind us. Once everything is set to be qualified, several months remain for live testing so customers can qualify the flow on their own, if you follow my meaning. So in certain instances, it moves through HTOL and any other live tests that a customer demands as part of their own obligations toward their end customers, yet the majority of this amounts to operational execution alone. That's, I think, one factor behind why we feel so bullish and confident about where we stand and where we're headed with this model we just provided of the 2.8 — what was it, 2.84 and the $750 million net profit. When what you're aiming for and planning is to have everything online, for wafer starts in December. All right, suppose, worst case, you miss by 1 month, 2 months, 3 months, fine, perhaps, but it's there. So should it be — will you possess the complete start capability in December, we aim to have that, and I believe we will. Could it slip so that 1 or 2 tools aren't fully qualified for some reason. Clearly, plenty of equipment is arriving from suppliers. And while we're quite good and the suppliers are quite good at performing a final test at the supplier site, all the tools get disassembled and shipped. Through disassembly and shipment, something might break or go wrong that doesn't get spotted right away during what's called Tier 1, Tier 2 start-up at our site. So it's conceivable that, that might take slightly longer and 1 or 2, 3 tools could be delayed past the plan. It's equally conceivable that, for some reason, the supplier itself misses its initial target, and that does happen. Just as — not that tower would ever do such a thing, but occasionally wafer manufacturers fail to meet their commitments — just kidding there. The thing is, whether it lands in December or January, possibly February, I'm not certain; it could also be — look, we've already said publicly that we generally aim to have all of it operational by December. Folks who need certainty about commitments down the road will set targets where they figure there's some wiggle room. That means the number you set internally ought to be tougher than the number you tell Wall Street, correct? But what I'm really driving at, perhaps saying it in far too many words, is that regardless of whether full qualification of start capability arrives in December, November, January, or February, arrive it will. Demand exists, it's locked in, and it's going to get used. So the model gets achieved. Once you kick off all the starts, there's a stretch of time needed to ramp and qualify everything; after that, a certain number of months pass before shipping is possible and revenue comes in. That's why we're quite at ease discussing 2028 as the year we reach our model — the demand is right there. Customers have made commitments for that demand; to the degree that we never requested reservation fees from customers — they offered them voluntarily. They understand how valuable, particularly from tower, SiPho demand is, given that we are genuinely, by a wide margin, the leader in silicon PICs. So they're after the wafers, and it comes down to us on operational execution. Though I'd like to claim it's entirely within our control. Supplier performance also plays a role; fortunately they're solid suppliers. Our ties with them are strong. We put a lot of emphasis on maintaining robust customer relationships. And to keep a customer relationship healthy, you need the same approach with your suppliers, right? What goes around comes around, after all; so someone who won't deal fairly with a supplier can hardly expect fair treatment from a customer, and the reverse holds too. So, Cody, perhaps I've gone on — far too long. Still, our plans are extremely solid, extremely robust. Could something slip by a month or two in either direction? Naturally. But the plans exist. And should a month or two of slippage occur, it won't come from the majority of the capacity growth — it'll be 1 or 2 tools, or I don't know, a small number of tools, what people sometimes call a lemon tool. Not that it's truly a lemon; rather, some issue simply wasn't caught right away. That has the potential to push something back. Yet throughout this year, incremental capacity growth is absolutely going to happen.

Cody Acree analyst
#19

Could I pick up where we left off regarding your mobile segment? Is there any worry around the memory supply constraints that keep going, or the higher pricing that a number of others in this space have flagged, and what that might do to unit shipments in handsets?

Russell Ellwanger executive
#20

Cody, what I'd say is that whenever something sits in the market and you've got no influence or authority over it, worry is always present. So absolutely, that concern exists. We collaborate intently with our buyers to grasp where their stock sits; they, in turn, attempt to grasp where their own customers' stock sits. And to be as certain as we can that the target we've set for — our start plan covering this year is achievable. But do we — I would really like to claim there's nothing to worry about and that it can't touch us. That isn't true — there's — market forces are always at play that you'd never wish to be on the losing end of, so your best move is to plan as thoroughly as possible. And ideally to keep options available in case some capacity goes unused in the fab, so it can be swapped for another thing, which is where we mentioned deliberately developing certain lower-margin items to create space for higher-margin items. Those lower-margin items remain wanted, and it's always possible that if a hole opens up in the fab since demand you expected doesn't materialize, we can fill it with another product. That other product may not be ideal since its margin profile differs, yet it's feasible; so at minimum your fixed cost gets covered.

Richard Shannon analyst
#21

Tyler here, filling in for Richard. Apologies for the letdown. I've got a question. Regarding that model you provided and the 2028 timeline, I wanted to ask something. Does that figure represent a run rate for 2028? Or does it cover the entire year?

Russell Ellwanger executive
#22

Yeah. No, absolutely, we'll get there on a run rate basis, and our aim is to reach it for the full year. But what we've said is that it would be accomplished during that year. So our goal — and what I've communicated — is exactly that. It's entirely reasonable to think we'll achieve it by run rate. And ideally, we would very much like to achieve it for the full year. And that could happen.

Analyst analyst
#23

All right, that sounds good. Now, regarding silicon photonics — I realize you just said it could be used to fill in for other things, but looking at where we are now, given all the capital expenditure you're putting in, would that mean the fabs end up at complete utilization under that scenario?

Russell Ellwanger executive
#24

No, silicon photonics wouldn't get any of the factories to full photo utilization. However, silicon photonics isn't what I was referring to when I brought up backfilling. That particular question concerned RF mobile, driven by concern that the high-bandwidth memory producers are concentrating on data center instead of providing it to other markets. And absent that memory, there was a possibility of a downturn in the overall mobile integrator, since they wouldn't have the memory required for their phones. That's what the question was about. So what I meant was, if that scenario played out, that capacity can be repurposed.

Analyst analyst
#25

Understood. What I'm driving at, though, is whether the CapEx you're putting in now gets us to that 85% utilization.

Russell Ellwanger executive
#26

It does, as long as the remaining flows are utilized at the capacity levels we designated for them. So no — were it just silicon photonics, we wouldn't hit 85% utilization. But you also need to keep in mind something important here. Our focus is on silicon photonics, and our commitments tied to it, where I'd say RFSOI is largely interchangeable with power — there are certain layers that differ, but it's fairly interchangeable for power and fairly interchangeable for imaging. Silicon photonics operates under different ratios, yet it's highly interchangeable with silicon germanium.

Lisa Thompson analyst
#27

I've got a couple of questions about accounting, and they're directed at Oren. To start with, can you give us the precise figure for that one-time tax benefit we saw in Q4?

Oren Shirazi executive
#28

Roughly speaking, it comes down to what the model would show if the rate were 15%, 16%, or 17%. On pretax income of $81 million, you'd normally expect a tax expense somewhere in the 15% to 17% range — so somewhere around $12 million to $13 million. What we actually recorded was $1.5 million. That leaves a difference of roughly $10 million.

Lisa Thompson analyst
#29

All right, and could you walk through precisely what was received in exchange for that $105 million paid for extending the lease?

Oren Shirazi executive
#30

What we received was 3.5 more years on the Newport Beach facility lease. That was put out in a press release dated 13 November 2025. Rather than finishing up early in '27 as had been planned, it now runs through the close of 2030.

Lisa Thompson analyst
#31

All right, and that $105 million — was it paid as cash up front?

Oren Shirazi executive
#32

Correct, correct. Correct. And that figure sits inside Q4's cash flow from operations — that's why, just for this one time, it comes in $105 million below whatever any model shows. Mind you, we put out the announcement back in November. So there's nothing fresh here, no.

Lisa Thompson analyst
#33

Okay, okay, okay. Then I'm just wondering — with the U.S. depreciation rules shifting regarding what is eligible for write-off, has your model been affected in any way, or have your depreciation expectations for the future changed at all?

Oren Shirazi executive
#34

No. It doesn't affect us.

Lisa Thompson analyst
#35

No, that's not the case at all. Okay. Wonderful.

Russell Ellwanger executive
#36

I appreciate it greatly. With 2025, I finished my 20th year with Tower. Therefore, I figured I'd offer a somewhat broader perspective regarding what Tower represents, our direction, and our activities. During 2025, we adopted a corporate theme: bold growth, limitless impact, infinite reach. That theme resonates with me deeply, and I invested considerable reflection into it. It would genuinely be my tremendous honor if the path of my life proved deserving enough for those words to appear on my epitaph, alongside, naturally, loving, honorable, loyal, husband, father, grandfather, and friend. Yet if that appeared on my epitaph, wow, what a value-adding life I would have lived. Regarding bold growth, to me at least, it signifies remaining undaunted and building a legacy far more accretive than one's circumstances at birth. In the specific context of corporate leadership, that would mean growing the organization far, far past the state it was in upon one's arrival. As for limitless impact, that would signify that the person or corporate leader has succeeded in transferring knowledge and generating chances for staff, peers, and community, enabling one family to enjoy a rising growth trajectory far past what they would have otherwise experienced, what would have otherwise been. Infinite reach is quite a fascinating idea. I initially came across this phrase in David Deutsche's book, The Beginning of Infinity, and the meaning he proposed was that truth uncovered in any domain—if indeed a truth applies across all domains—and that is quite fascinating. When you consider learning, plenty of things genuinely cannot be taught, but instead have to be learned, and that learning arrives solely via doing. I pondered that considerably. It truly—a great many things are teachable, yet those things that can be conveyed are tools. What can be learned are principles and values, and it genuinely is acquired only via doing. I hold a very, very passionate conviction that work is the laboratory in which a person can and ought to learn and grow themselves across all capabilities, principles, and values required to become the person they wish to be. Anyone deserving of respect devotes the majority of their waking hours to work. What a pointless undertaking if that isn't where someone grows as a person. And I believed strongly that a good company has to permit financial and professional advancement, yet a great company permits the same while adding personal advancement. Years ago, earlier in my career, I had the wonderful fortune to reflect and thank Dr. Dan Medan, at that point the President of Applied Materials, and this is precisely what I wrote. When I arrived at Applied, I believe I was a good person. Thank you for building an environment that enabled me to become a better person. Tower wants to become that kind of high — that kind of company. Our priority is bringing aboard the most talented and enthusiastic individuals, and just as crucial, creating and caring for a setting in which talented, enthusiastic individuals can keep developing their skills, their zeal, and their goodness too. We recognize and promote a grasp of the idea that the greatest driver behind expanding skills and deepening enthusiasm is close partnership with our clients, along with the thrill that gets earned — genuinely earned — by taking part in one another's wins. There's a saying whose source nobody's sure about: suppose 2 individuals see eye to eye on every single thing, then one of them isn't needed. We cherish diversity. We cherish a range of viewpoints, though only when aimed at one shared aim, purpose, and set of deeds. Folks who act like organizational anarchists don't fare too well here at Tower. Still, regardless of what the view is or how varied it may be, when it aims at improving things, we hold it in high regard. I doubt I'm all that unusual compared to everybody else. It bothers me whenever someone takes a different position than mine, yet I genuinely prize it; that's a really powerful trait, and that's the culture we've got. So we've put in hard effort to be an organization that genuinely lets individuals develop. And once you let individuals develop, you wind up with a setting and a morale inside the company through which the company itself turns into a genuine masterpiece. Now, I've got no idea what pull any particular work of art or piece of music holds for anyone listening in, but I can tell you that I'm unable to pass a da Vinci without feeling its pull. Such is the power of a masterpiece. The very same holds for a company. When a company is filled with exceptionally enthusiastic individuals, and those individuals possess the loftiest character along with capability and knowledge, they act as a magnet drawing the client in, and that client wishes to remain alongside them. And that is precisely what opens the door to corporate expansion. That ranks among the factors enabling daring expansion, the kind that lets a company achieve boundless impact, and it rests upon the unbounded potential of people — the moment you shoulder major duties and assume complete ownership over them, you come to learn countless truths; and whatever holds true within one domain holds true across all domains. The very principles that let you thrive as a business leader also let you thrive as a father, a husband, a mother, a wife, a daughter, a friend who brings real value. So those are the matters Tower has genuinely devoted effort to and keeps devoting effort to. Former U.S. President Bill Clinton once offered a remark that, upon first hearing, comes across as quite pleasant: it goes, old age is when your memories outweigh your dreams. I'm no spring chicken, so these are the sorts of things I mull over. And initially, once more, it sounds quite pleasant, yet surely, possessing dreams does not in any manner characterize lively youth. So the claim might hold up in one direction: if what you remember exceeds what you dream, that indicates you're old. Yet simply having dreams doesn't prove you're young. A great many individuals, even when they're young, possess only dreams and lift not a finger to turn those dreams into reality. That's why I tacked something onto the quote and took that liberty. Being old is when your memories outweigh your dreams, along with the actions you take afterward to realize them. Tower is by no means a company that has aged. Our work draws on experience and knowledge that only come with age, yet it carries all the vibrance and excitement of use, and that mix can't be beaten — it genuinely drives customers to want to engage with your company. We presented this new financial model, and as I've said on several occasions and have been asked about too, hitting it is our goal, whether through run rate or across the full year, but reaching this model in 2028, which is fairly near term. The model covered all of the incremental margins, but essentially, it amounts to a revenue CAGR of 22%. That's a very attractive CAGR for our industry, a very attractive foundry CAGR. So a 3-year CAGR of 22%; however, it's a net profit CAGR of 50.5%, and achieving a 2.5% on top of the 22% CAGR — a 2.5% increase on the net profit CAGR — is truly remarkable, something rarely discussed since most companies don't have net profit CAGRs. But going from where we stand today to reaching this model comes to 50.5%; that is not an aged company. That is a company absolutely packed with youthful exuberance grounded in the capability of age, grounded in experience, grounded in having built up multiple years of strong, exceptionally strong customer relationships. So in closing, we came into 2026 with tremendous momentum toward bold growth, limitless impact and infinite reach. Thank you for joining us. Thank you for staying with us as we move toward achieving the $750 million net profit model. Thank you very much.

Operator operator
#37

That brings us to the end of today's conference call. We appreciate everyone joining us. You're welcome to disconnect your lines at this point. Thanks.

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.