Tower Semiconductor Ltd. (TSEM) Earnings Call Transcript & Summary

Ladies and gentlemen, thank you for standing by. Welcome to the Tower Semiconductor Conference Call. [Operator Instructions] As a reminder, this conference is being recorded September 5, 2023. Joining us today are Mr. Russell Ellwanger, Tower CEO; Mr. Oren Shirazi, CFO; Dr. Marco Racanelli, Senior Vice President and General Manager of our Analog Business Unit; and Dr. Avi Strum, Senior Vice President and General Manager of our Sensors and Display Business Unit. I would now like to turn the conference call over to Ms. Noit Levy, Senior Vice President of Investor Relations and Corporate Communications. Noit, please go ahead.

Thank you, and welcome to Tower Semiconductor Investor and Analyst Conference Call. Before we begin, I would like to remind you that some statements made during this call may be forward looking and are subject to uncertainties and risk factors that could cause actual results to be different from those currently expected. These uncertainties and risk factors are fully disclosed in our Forms 20-F and 6-K filed with the Securities and Exchange Commission as well as filings with the Israeli Securities Authority. They are also available on our website. Tower assumes no obligation to update any such forward-looking statements. Now I'd like to turn the call to our CEO, Mr. Russell Ellwanger. Russell, please go ahead.

Thank you, Noit. Welcome, everyone. Thank you for joining our call today. I'm very happy to reconnect with you. During the call, we will share the progresses that we've made and the milestones we've achieved in the last 18 months as well as our present plans. Our first open capital market interaction we have held since having signed a definitive agreement with Intel. Tower entered into that agreement being interested to accelerate its business growth through the much larger manufacturing scale that Intel offered as well as combining certain Intel high-end digital and [indiscernible] capabilities with Tower's leading analog solution. During these past 2 years of interaction with Intel, we have learned a lot, being able to benchmark our capabilities with a multi-decade semiconductor leader. However, unfortunately, as has been released, not all regulatory approvals were received, and both companies decided not to further extend the definitive agreement. Today, we jointly announced a large capacity corridor agreement with Intel, helping to fulfill one of our strategic goals. As per Pat Gelsinger's statement, in our previous mutual press release, we will continue to look for opportunities to work together in the future. This is Tower's intent as well. For example, to combine certain Intel high-end digital and packaging solutions, with Tower's leading analog solutions to create some potentially industry transformational capabilities. In the months since we signed the definitive agreement back in February 2022 and a lot of goodness has happened within Tower. We've continued to operate independently, maintaining our strong performance, executing strategically and tactically. -- Backed by a favorable market environment in 2022, the company set multiple financial records. At our last financial call in November 2021, we set an ambitious net profit margin target of 15% for 2022, a significant increase from the 10% we achieved in 2021. As has been released, we didn't just meet this target, we surpassed it. In the fourth quarter of 2022, we achieved a net profit margin of 20.7%. And for the entire year of 2022, we recorded a net margin of 16%. This accomplishment came against a backdrop of several quarters of sequential record growth and revenue. Since the fourth quarter of 2022 during the first half of 2023, the market has experienced headwinds that have affected our customers, intern us and as well our competitors. Nonetheless, despite a 16% year-over-year reduction in revenue during the first half of 2023, we maintained that 15% net profit margin model, strong evidence of the rich mix that we are manufacturing. In terms of guidance, we expect the third quarter to remain predominantly flat as compared to the second quarter. For updates, our revenue breakdown for 2022 and the first half of 2023, sensors and displays represented 17% from our corporate revenues in 2021, 18% in 2022 and 20% for the first half of 2023. Our RF mobile business, represented 20% -- sorry, 27% of corporate revenues in 2021, 22% in 2022 and 15% for the first half of 2023. Our RF infrastructure business represented 13% from our corporate revenues in 2021, 14% in 2022, 9% for the first half of 2023. Tower IP business represented 15% in 2021, 19% in 2022 and 27% for the first half of 2023. Our discrete business represented 15% in 2021, 16% in 2022 and 18% for the first half of 2023. Mixed-signal [ CMOS ] represented 11% in 2021, 8% in 2022 and 7% from the first half of 2023. For each of these periods, there's about 2% or 3% miscellaneous. Automotive, which we serve through several of the above-mentioned technologies, was 11% in 2021 and in 2022 and increased nicely to 19% for the first half of 2023. Providing second quarter utilization levels, this is second quarter of 2023, Fab 1 6-inch was approximately 70%, fab 2 8-inch was about 75%, fab 3 8-inch, about 40%, down predominantly due to weaknesses in data center. Fab 5 8-inch was about 70%; fab 7 12-inch was about 75%; fab 9 8-inch was about 65%. The current utilization levels is presented show that we have the capacity to support substantial revenue growth organically with our present footprint, which we will certainly see as the market recovers. As stated during the past 1.5 years, the significance of scale became increasingly evident to us. Our primary focus is continued growth while enhancing our manufacturing scale. This, in order to ensure we're able to support our customers' future demand, we actively pursue additional capacity growth. Since our last update, we expanded our manufacturing capabilities in a few of our existing facilities. At our 12-inch facility in Japan, we've increased our capacity by about 8%. At our 8-inch facility in Japan, we increased the capacity by 4%. And at our 8-inch facility in Migdal HaEmek, Israel, we increased the capacity by about 11%. For new capacity, in the Agrate 12-inch facility, we are on schedule with the capacity build while successfully completing typing for several products. This puts us in a strong position to commence production by the end of 2023 as planned and initially announced. We just announced the capacity corridor with Intel that will provide additional very high-volume 12-inch capacity with targeted full qualification in 2024. As we look to our future, we continue to focus on technology platforms, which differentiate and provide us leadership position. This aligns to the high-volume opportunities presented by existing and new customers while expanding our manufacturing scale. Dr. Racanelli, our Senior Vice President and GM of the Analog IC Business Unit and Dr. Strum, our Senior Vice President and General Manager of the Sensors and Display Business Unit, will discuss our differentiated technology leadership and customer partnerships as well as present market trends. Oren Shirazi, our CFO, will overview some financial highlights, and I'll then address a few summary remarks. Our financial strategies, combined with our operational excellence, position us very well for the future. We remain committed to delivering value to our stakeholders through ensuring sustainable and strategic growth for the company. Now I'd like to turn the time to Dr. Racanelli, Analog BU General Manager. Thank you, Marco.

Thank you, Russell. I will now review our analog business, which includes RF mobile infrastructure and power management segments. First, the RF mobile business is primarily composed of RF front-end products such as switches, antenna tuners and low-noise amplifiers built on our RF SOI technology. Over the past few years, this business grew significantly helped both by our strong technology and the adoption of 5G as each 5G handset requires significantly more RF silicon content than a 4G handset. Today, the handset market has been weakened by the combination of a post-COVID inventory correction, a weak Android consumer pull and somewhat of a plateauing of 5G adoption in developed nations. Looking forward, however, we anticipate inventories will normalize as we already believe to be at or near the bottom in the demand cycle. 5G adoption is expected to pick up in developing nations, especially India. And on the technology front, over the year, we have released PDKs for a breakthrough industry best RF SOI technology now prototyping with lead customers. The technology offers Ron-Coff figure of merit below 60 femtofarad, which up to just recently was thought to be only achievable through MEMS or other exotic technologies. This figure of merit enables our RF customers to deliver products with smaller die sizes and lower RF losses, which leads to improved reception and lower battery drain. The new technology is prototyping today in our 200-millimeter factory, and we will be releasing PDKs for a 300-millimeter version in the fourth quarter to further expand capacity available to our customers in this large market segment. Turning now to our RF infrastructure business. It is primarily composed of silicon germanium and silicon photonic components for optical transceivers, which are responsible for high-speed communications and data centers, artificial intelligence clusters and telecom networks. Our silicon germanium and silicon photonic technologies offer performance that has translated into high value for our customers and high margins for our business and thus growth in the space over the past years has helped us improve both top and bottom line metrics. This year, we have seen a pause in this market led primarily by the well-documented inventory correction at hyperscalers. This correction is resulting in low utilization in our Newport Beach factory, which largely serves this high-end [ RF ]. However, we have used this pause in the market to accelerate roadmap development in both advanced silicon germanium and silicon photonics. In silicon germanium, we have successfully prototyped our 86 or sixth generation platform, targeting the most demanding coherent applications while our fifth generation has ramped and continues to gather design wins and momentum in the 400- to 800-gigabit per second data center market. In addition to adoption in the traditional optical pluggable modules, these technologies are being adopted active optical cables and for retimer in active copper cables as well as in linear drive applications that are offering lower cost, lower power and lower latency alternatives to more expensive DSP-based solutions. Cost, latency and power consumption are becoming increasingly critical in many artificial intelligence and other applications, such demands, creating an expanded future market for these technologies. In silicon photonics, we have had a flurry of new technology and activity. First, we have added new low loss waveguide and coupling features to our production process enabling a record number of new design wins and tape in, in the first half of this year that cover optical applications in data centers and applications outside the data center in telecom, automotive LiDAR and optical computing. Next, we have been maturing our next-generation offering, which integrates indium phosphide lasers, along with silicon photonic components, resulting in the only foundry technology capable of delivering a single die optical front-end without the need for an external laser. We have design kits available to customers and have an announced partnership with OpenLight on this process, which helps our customers with IP and design services to speed adoption. Finally, over the past year, we have created a 300-millimeter version of our silicon photonic process to take advantage of scale and improved capabilities available in 300-millimeter and have released initial PDK to lead customers. As the hyperscaler market recovers from the current inventory correction, we, therefore, see not only a return of prior production volume but many new opportunities on these new technologies, which will more than double our served market and given our large market share in optical SiGe bring the potential to add significant high-margin revenue over the next several years in both silicon germanium and silicon photonics. Our power IC business serves a broad-based set of customers in automotive, consumer, computing and industrial areas with both 200-millimeter as well as 300-millimeter technologies. In the past few years, this business has grown significantly due primarily to our offering a very low Rdson figure of merit, which helps our customers achieve smaller die size, lower cost and improved power efficiency. In addition, we delivered unique high-voltage capability used in automotive EV battery management, which has benefited from strong market adoption. This year, the power market has seen mixed results although it has remained generally less impacted than the RF market. However, looking forward, in addition to market recovery, we are seeing unprecedented opportunity brought about by our recent focus on introducing more comprehensive mixed-signal content in our 300-millimeter BCD platform, such as improved low-voltage analog devices and more advanced digital libraries and memories. These enhancements are enabling us to effectively compete for more complex, larger die handset applications that can drive high volumes and represent a multibillion-dollar market we previously did not significantly participated. We have strong traction with initial tape-ins from new customers in this higher content platform and we'll be working to bring these to production over the next 18 months. Our investment in expanded 300-millimeter capacity at the Intel facility described in this morning's joint announcement will help provide the needed additional capacity to capitalize on these new opportunities. I will now turn the call over to Dr. Avi Strum, SVP and GM of our Sensors and Display Business. Avi?

Thank you, Marco. For the Sensors and Displays Business, in the past 2 years, we worked on multiple projects with immediate impact as well as long-term strategy to drive our revenue and market share growth in both the sensors and the displays markets. In the image sensors business, we continued to grow this year. And despite the decline in the overall worldwide semiconductor market, including image sensors, we saw an increase of 13% in the image sensor segment in the first half of 2023 compared with the first half of 2022. In parallel to this growth, which was to the most part, based on our previously developed technologies, many new products for factory automation, robotics and intelligent traffic systems market segments from a variety of leading customers using both our front [indiscernible] technologies on our 300-millimeter 65-nanometer platform has been prototyped and are now ramping to mass production. These products ensure our continuous growth in the coming years as well as the growth of our market share in these segments. In the past 2 years, we successfully developed new technologies that will allow us to expand our served market, such as backside illuminated stacked global shutter pixel for the industrial market, single-photon avalanche photodiode spectral range for time-of-flight 3D mainly for the automotive LiDAR and shortwave infrared technology for vision at harsh low-visibility environment also predominantly for the automotive. For the industrial machine vision market, we developed with a leading customer an industrial backside illuminated stacked 2.74 micron global shutter pixel for high-resolution stitched sensors and are in the final stages of the development of a 2.2 backside illuminated stacked global shutter pixel platform for the high-volume industrial sensor market. These 2 architectures would be the foundation of our next-generation platform for the industrial imaging market and allow us to grow our market share in the industrial market segment from about 15% to 25%. For the automotive market, we have developed with leading customers direct time-of-flight products for automotive LiDAR. Those are used for accurate distance measurement to provide 3D mapping of the environment used for collision avoidance and parking assist applications. Based on this platform, products for the commercial and industrial markets such as fast autofocus for cell phone cameras and autonomous robots were developed as well. These products are expected to ramp to production next year. We also successfully prototyped a commercial short wavelength in freight sensor with a partnering customer TriEye based on our germanium photodiode stacked wafer technology. This is a pioneering solution in the affordable commercial infrared market that is also targeted for the automotive advanced driver assistance systems, fast-growing market for aid in driving the low light and harsh fog or smoke conditions where visibility is extremely low. This will drive long-term and stable growth of our 200-millimeter well-established imaging platform. For the long-term, we are working now with our partners on providing organic sensors offering for the visible near infrared and shortwave infrared spectra. This will allow very small global shutter pixels with a very high quantum efficiency at wavelengths beyond what silicon can support. This technology is supposed to get into production in a couple of years as a complementary technology to our well-established stacked backside illuminated global shutter platform in the industrial, automotive and commercial markets. These technologies and partnerships with leading suppliers into the automotive sensor market will allow us to take a meaningful market share out of the current $700 million automotive sensor wafer market that is growing to $1 billion in 2027. For the medical market, we have seen substantial growth in the medical and dental [ stitched ] x-ray sensor market with our 200-millimeter platform and are developing now in addition to our existing 300-millimeter x-ray [ stitch ] platform and a lean flow on 300-millimeter wafers that will allow us to take a higher market share, especially in the more cost-sensitive segments. As our current market share is already higher than 50%, our way to increase our business is by increasing our served market. This linked flow provides 20% to 25% cost reduction and thus increases our served market as existing nonCMOS-based lower cost, lower performance technologies can now be replaced by higher performance cost-effective CMOS solution. In addition, we have developed for 1 of the market leaders a completely new direct x-ray photon detection platform for x-ray photon counting applications in the medical market. Photon counting is the next-generation technology for high-resolution CT machines and is being considered a real revolution in the medical imaging market. The expected volume is very high as 1 CT machine requires close to 1,000 wafers. These edge detectors are also used for particle physics in research institutes like CERN for example, that can also drive very high volume. Just for the CT market alone, according to our customers, hundreds of systems will convert into photon-counting technology. So we are talking here about hundreds of thousands of wafers for the totally new market. Moving to the display market. We have put a lot of focus in the past 2 years on the development of silicon backplane for micro OLED platform. The platform can support stitched and non-stitched displays for the AR and VR markets with voltage drive transistors and very low leakage required for this application and is very lean and cost effective. We are now completing a 10-volt platform on 300-millimeter and are working closely with leading customers in the VR display market. This 10-volt platform will enable much brighter pixels as OLED pixels can be stepped in series, providing 2 organic LEDs per pixel. The VR global market has been delayed in its ramp, but expected to drive very high wafer volume as the number of die per wafer is small. The number of [indiscernible] per wafer can run between 40% to 60%, depending on the display size and 2 displays per set are required so the opportunity is very large, several tens of thousands of 300-millimeter wafers. To summarize, we not only have a very strong position, especially in the industrial imaging and medical and dental x-ray markets evident by our growth despite of the semiconductor market slowdown, but a strong newly developed state-of-the-art technologies such as backside illuminated stacked global shutter [indiscernible] and [ SWIR ] pixels as well as OLED-on-silicon backplane technology that will allow us to not only grow our market share in our existing growing markets but also new fast-growing high-volume markets for us. We are very optimistic and confident about our future. With that, I would like to turn the call to our CFO, Oren Shirazi.

Thank you, Avi. Hello, everyone. I'd like to take a moment to discuss our financial position and margins following the strategic steps we have taken to ensure our company's financial stability and continued ramp-up of our capacity revenue and profitability. Our balance sheet is a clear reflection of our strong financial position and forward-looking approach. Our balance sheet assets as of the end of our last reported quarter, mainly Q2 of this year, totaled $2.5 billion includes $1 billion of fixed assets, mostly machinery and equipment, and $1.4 billion of current assets. Our cash, reflecting the multiple by which current assets are bigger than short-term liabilities, was 4.8x as of such date among the best in the industry. Another key highlight is our net cash position, which as of such date was approximately $700 million before the fees -- of the termination fees from Intel, which [ cash ] is the enabler for our investment strategies to drive increased manufacturing scale, including the one announced in today's press release. And in addition, provides liquidity to navigate any unfortunate challenges. Our strong financial position ensures that we can invest in opportunities that are aligned to up vision and promise significant returns, eg. Approximately $500 million of total aggregate cash is planned to be invested for equipment and other CapEx required to be installed in the 12-inch factory in Agrate, Italy, following the previously announced STMicro partnership agreement signed in 2021. We anticipate investing $300 million until the end of 2024, of which $100 million in the second half of this year and $200 million during 2024. Having invested already $100 million in each of the first half of 2023 and full year of 2022. We have [indiscernible] to invest an additional several hundreds of millions of dollars to continue to increase our capacity at Agrate, Italy. In addition, we expect our maintenance CapEx baseline level to remain as previously announced between $180 million and $200 million per annum. And the last addition to our CapEx plan is, as stated in today's press release, we will allocate an additional up to $300 million to buy equipment and other CapEx items to enable power customer roadmaps of production and manufacturing in Intel [indiscernible] in New Mexico. Now in relation to margins and efficiency, manufacturing scale and value-add are the drivers of gross margins, which in turn will drive all other margins. We have value-add products, as our scale now grow, so will our gross margins, which obviously is the starting point for all other margins. During recent years, as we have scaled our operations, our primary focus has been on improving efficiency. The drop from gross margin to operating margin sales as an indicator of our efficiency level. As you can -- while we've grown, we have consistently enhanced our operational efficiencies as follows: Our growth journey from annual revenue of $100 million per annum to $1.68 billion in 2022 is a story of strategic expansion and efficiency. In the early stages of this journey, we experienced a drop of more than 20 points from gross margin to operating margins. However, we managed to reduce this figure to about 10 points, among the best in the industry. We were able to continuously maintain this efficiency level in recent years despite the challenges and [ complexity ] that we face in such a ramp period, while we increased our capacity, revenue and profitability. And now I'd like to turn the call back to our CEO, Mr. Russell Ellwanger. Russell, please go ahead.

Thank you, Oren. And thank you, Avi and Marco as well. Over the past 1.5 years, our activities and learnings have been deeply underpinned with our corporate values. The value vectors of leadership, partnership, impact and innovation. These values overlap each other and then the best of what we must be to deliver the highest value. You cannot be a leader without partnership or without innovation. One cannot have sustainable partnership without close sharing and also innovating. I'm not speaking to just technology innovation, but as well innovations and business models. At times, the absolute best innovation is just simplification. In this, each and every employee can and does participate to simplify and streamline procedures and processes, making it easier to work with Tower. As we continue to drive all interactions, through these lenses, our footprint of impact increases externally and internally. The employee that goes home at day's end, fulfilled and encouraged through their days achievements is capable to perform wonders within the walls of their own homes and communities, which is accretive to the positive energy they begin the next workday with. To make these values into vectors of direction and magnitude, we strive for excellence in everything that we do. We evaluate each activity in terms of excellence. And we define excellence in 3 components, each being directly measurable, namely effective, efficient and of the highest quality. In purely financial terms, as Oren mentioned, effective is gross margin. Gross margin is a function of selling value-add products and doing so at scale. Our flow offerings have and continue to be of the highest value to our customers. To a good degree, from our multi-generation customer partnerships on roadmap developments. This is shown by both Marco and Avi has only grown stronger over time. It is clear to us that we must as well increase the scale of our manufacturing. We have taken many steps towards scale increase organically by adding capacity when feasible to existing factories over the years as well as inorganically with, for example, the acquisitions of TPSCo and in the past few years, the arrangements with ST at the Agrate facility in Italy and now with the large capacity corridor in Rio Rancho with Intel. And there are additional opportunities presently being considered for additional increases in scale. The financial metric of efficiency is well demonstrated by the drop in points from gross margin to operating margins. In this metric, being at or below a 10-point drop, Tower's among the best in the world. Reduced infrastructure enables firstly quick decision-making in itself a business differentiator. And secondly, it enables single-focused hence speedy execution. Our development model is such that all development or, if you will, our R&D is performed in the production facility that will be responsible for the customer ramp. All of our fab process and integration engineers are both production and development, enabling fast time to ramp and allowing a single mailbox for customers of the highest quality comes from in the first order, the quality of our employees. We performed a yearly ranking and rating in the company. Since 2009, when we implemented the ranking and rating system that we currently use, -- the annual retention of the top 20% of exempt employees has been above 97.5%, which would represent an average of 40 years of employment. Employees vote with their feet. Our employees, especially the top performing, love it here. They believe in the company, its values, its vision, mission and our aspirations. And considering knowledge, skills and attitude there is none better than the Tower workforce. All said and to summarize, the present customer traction and growth potential as the strongest we have ever had in the company history. We look forward to the next years of execution and achievements. Tomorrow, we are holding our Europe Technology Global Symposium in Amsterdam. We have over 100 customers that have registered for this event. Next week, on September 12, we will hold this event in Shanghai, China, for which we have over 300 customers registered for the event. In November, we should be holding the same event in the United States. We're looking forward to these many short-term interactions with customers to drive even stronger long-term relationships. I thank everybody for joining our call today. I would like now to turn the time over to our operator for the question-and-answer session. Thank you.

[Operator Instructions] The first question is from Cody Acree of Benchmark.

Welcome back. I'm not sure the best -- the most desired process to get back here, but welcome nonetheless. Maybe Russell, can you talk about business that you've been engaged in over the last 18 months as you work through the process with Intel. I'm sure that there are some customers that have been more comfortable to start process with you over the last -- over the acquisition period. So I guess the canceling of the acquisition pose any challenges to that business?

That's a very good question, Cody, and it's very nice to have contact with you again. I would say that the initial announcement of Intel acquiring us was very positive for Tower for who we are for what we're doing. The fact that Intel saw value to acquire us, the market cap that they assigned to the acquisition, it gave us a lot of visibility, maybe more visibility than we otherwise had and maybe warranted or not, it gave us more than we had had before. I could say probably through the fact of the acquisition, we became engaged in a series of opportunities that we might not have been engaged in otherwise, they were still being handled on an individual basis, meaning it was Tower interactions. And as stated, we maintained throughout the entire [ DA ] period as a fully separate company running Tower as Tower, and that was strongly respected by Intel. But there were opportunities that came about probably because of the acquisition. As Dr. Racanelli mentioned, there is a series of opportunities that the fulfillment of which will now be greatly enabled on the capacity side through the added capacity in the New Mexico facility. So I believe strongly that the trust between us and this, if you will, new set of customers would have maintained with or without the acquisition completing. And it's continuing to maintain and the acquisition did not complete, meaning the acquisition of Tower by Intel. But the capacity commitments that have been made will be able to -- or that had been foreseen will be able to be definitely fulfilled. And I think that, that was one of the very nice things about this capacity corridor agreement. However, even without the agreements, we do have substantial capacity that is at hand being built in Agrate. So maybe the specific answer is for one series of customers and opportunities, the announcement with Intel was maybe a catalyst for these opportunities. I don't see in the least bit that the deal not completing will be decretive towards the opportunities being fulfilled. And across the rest of the landscape, I don't think that it had too much of a plus or a minus on any of the other things that we went after. But there was one set of opportunities that I believe was catalyzed through the Intel announcement or through that series of thoughts through that announcement. Did that answer your question, Cody?

Yes. Maybe one for Oren. Can you just talk about the accounting expectations for the partnership with Intel?

Yes, sure. Actually, it's -- the CapEx that we said that we will buy up to $300 million of machinery and equipment, it will be like any other CapEx. I mean, it will be in the balance sheet. The amount that we will pay $300 million. And other than that, there is not anything special, like it's not -- we didn't form like a legal entity, a new legal entity or anything like that. So it will be just whatever we will sell to our customers will be revenue, whatever we will pay as the cost to manufacture it for Intel [indiscernible] will be in the cost of goods in the core.

I guess, Oren, how are you going to account for those fixed cost versus variable costs? Please [indiscernible].

No, there is a price -- I mean it's not any accounting of as if you acquire somebody or consolidated financial statement because it's not a legal entity. So we will sell to our customers. Whatever price we will get from the customers is revenue times the amount of volume. And whatever we will pay to Intel, we agreed in the contract on a pricing mechanism that we will pay to Intel, whether it's fixed or not, doesn't matter. Eventually, there is a price that we will pay, which will come against the cost of goods.

Okay. Can you help me there again as far as what are the negotiations with Intel? What does that entail? And what are your expectations for how favorable or an unfavorable past that might be to you?

I wouldn't want to speak on the favorable and beneficial, more to Tower or Intel. It's a confidential part of the commercial agreement, right?

Certainly, Tower would not have signed an agreement that we wouldn't see as favorable for Tower. And I don't think Intel will sign an agreement that they wouldn't see us favorable to Intel. So I think any time that a person goes into an agreement, the only way that it's sustainable is if both parties believe that they've done something that's good for their own company. And that's the very nature of partnership in any context. So the specifics of the contract, we cannot get into. I mean that's not something that has been disclosed. But I believe that for Tower and for our focus on increasing net profits over time, it's a good thing for our net profit.

Great. And then lastly, I guess, just on a higher level, Russell. So most of last year, you ran about four and a quarter -- $425 million in revenue, but that's always sort of falling off with the industry's challenges. I guess can you talk about the path back to that prior -- those prior peaks? What does that entail in your mind? What's a realistic logical walk to that? And does that assume market recoveries as well as just a ramp of business you already have in [ Albuquerque ].

To get back to the four and a quarter [Audio Gap] another $1.7 billion annualized run rate. It's slowly a function of market recovery. We don't need additional market share to get there, and we haven't lost market share. So I think it's just a question of market recovery. To get beyond those numbers is a question of the new initiatives that we've done over the past 4 or 5 years. And the realization of those either increases in served market or increases in share of market. But the $425 million quarterly run rate, I think, is solely a function of market recovery.

The next question is from Lisa Thompson of Zacks Investment Research.

Nice to have you back and have this firehose of information.

Thank you. I'm glad it's the firehose, not a...

Yes, it is, it is. I'm waiting for the transcript. So just to clarify a little bit relationship with Intel. Since this is a little bit different, does this mean that then the gross margin from that relationship will be lower and operating expenses will be lower compared to the average?

I really don't want to go into the specifics of the financials of the deal itself. It's -- we just signed the deal on what one can or say at this point is really just what's in the PR.

Okay. And when might revenue hit from that relationship?

We stated in the PR. I stated in the quote that we expect to be fully qualified at the end of 2024. So if we're fully qualified in 2024, our expectation is that we would start selling wafers, either coincidence or at least in the beginning of 2025.

Okay. And Italy, when does that happen...

We will start selling wafers in the fourth quarter of this year. That's my plan anyway. I mean I can't -- things can happen by plan. We have -- we're planning to fulfill purchase orders in the fourth quarter.

All right. Great. And then will that also be -- is that similar financially? Is your agreement with them? Or is it very different from Intel?

In the big picture, they're very similar. What we did with [ ST ] in Agrate is very similar to what we're doing with Intel in Albuquerque or Rio Rancho, a very similar deal.

The next question is from Richard Shannon of Craig-Hallum.

I think maybe a near-term question for me, maybe a couple kind of bigger picture. Wanted to ask a little bit about inventory correction, how you're seeing that? And I think your language, Russell, is seeing -- being at or near the bottom here. Maybe you can suggest when we might see that? Are you seeing orders -- potential orders to indicate when that could happen? And is that across the broad spot of your business or one or more areas that are lagging or leading that correction dynamic?

From an order basis, things appear to be pretty stable at the moment, but they're not stable on the high side, meaning we're not having decreases. I would say that up through the start of the second quarter, there were continual decreases, that is what was going on, but it looks stable at this point. There's always talk from customers about return and things rebounding, starting in the fourth quarter and then coming up in Q1, Q2. But candidly, we've not seen that in direct POs or forecasts. So how long will we be at the current level, I honestly couldn't say. But I don't see us going down further than where we're at now. That's in my interactions what I see on purchase orders and in forecast meetings. But I'll just move it over to Marco or Avi if they want to add any more color. Marco?

Yes, I think that's exactly right. And I think different segments of the market have seen different cycles, right? The mobile space dropped, let's say, sooner and relative to other segments. But all in all, the current -- we've kind of plateaued and continue to see kind of similar PO levels and forecast levels in the next little while. So that's the statement that we are -- it looks like we're at the [indiscernible] cycle is arising from that reality.

Avi, anything?

Yes. I totally agree. We don't see any reduction. We still don't see a very high growth that we expected to see in 2023. But talking to customers, they expect it to come in 2024. So we don't see a substantial increase in POs yet. But from talking to customers, they are very optimistic.

Okay. Fair enough. Russell, maybe you can talk to not just the Intel agreement from this morning, but maybe the STMicro. I may have missed some of Oren's comments, the quality of the audio here wasn't great. But maybe specifically with the Intel agreement here with the $300 million, how do we think about revenue capacity still there? And then also comparing that or maybe even quantifying what the opportunity for STMicro and Agrate looks like when you're fully built out with the commitments you made on CapEx there?

I really would have to go back and see what we've publicly announced about Agrate, the total capabilities that we have. The present phase that we've agreed to, we should reach the full capacity of the present phase in the fourth quarter of 2024 on capacity. And it's a substantial amount of incremental revenue. I don't know that I've given that number, but it is considerable, and it's 300-millimeter. In the case of the Rio Rancho, we gave a specific amount of photo layers that would be added. And that was 600,000 photo layers a month. So that's not necessarily so difficult to assign a revenue potential to. And we've not given an average price per layer. But at 600,000 layers, if you assume that, that would be sold at its full utilization, a sign what you think is 300-millimeter advanced platform would go for at 65-nanometer. And that's pretty easy then to calculate what you believe it would be. I don't want to give that number because I'd be giving a long-term forecast, which I'm not doing, but a sign what you think is a reasonable price per layer. And that's the max revenue capability of that capacity core present or at least that's the point that it's fully qualified.

Richard, just to answer your question, what -- the numbers I said before about Agrate where that the total CapEx for Agrate is expected to be $500 million, of which $100 million we already paid in 2022, $100 million we paid in H1 of this year. Additional $100 million should be paid by us in H2 this year and additional $200 million we should pay in 2024. So that's the spread of -- the schedule of the $500 million.

Okay. Appreciate that Oren. That audio is a lot better there. So I got that completely. Last question, and I'll jump out of line. I wasn't able to calculate the numbers. But Russell, you gave us some percentages of sales by product categories. And the interesting one here for me at least is on the Power IC side, where the percentage has increased dramatically since calendar '21. I'm not sure if I -- I haven't been able to calculate whether it's actual growth here, but certainly relative to your business, the increase is substantial. Maybe if you can talk to the drivers for that? Is that by new products, new voltage levels, new customers added? Maybe just kind of give us a broader idea of what's been going on there? It looks fairly dramatic relative to your revenue profile. That's all for me.

So thank you for the question and the comments. Indeed, it's not just that it's growing as a percentage of business. If you annualize the first half, it's growing year-over-year as well. And so obvious to the market, we're growing quite a bit in market share. And it deals with new platforms, very, very much with new platforms. We talked about it before -- while we were having the call, we had mentioned several new platforms that we released. One of them was this RESURF, 140-volt RESURF that allows you to have high voltage without the need for having an SOI substrate and that's gained very good traction. So that's a big growth driver. Marco had referred to in his script very strong activities within power management, battery management. That's been a very strong growth driver, if you will, the -- our automotive share increased by 70%. Now an increase of 70% is because it wasn't a very high percentage to begin with, but it went from 11 points to [indiscernible] point. So that's a very big growth. I mean there's many dollars that are involved in that, and that deals with new platforms, in particular, with battery management that are highly differentiated in the market. Going forward, Marco talked about the multiple additions that we've put into in order to enter into very strong mobile PMIC market and an increase in our served market, really in the order of $2 billion to $4 billion where we have very strong traction right now. So we've been doing quite a bit continually on enabling features to a platform that always was the best Rdson in the industry, but we're adding more and more features to it that makes them more applicable to expanded markets. So that's -- yes, if you go back and just reread some of our scripts in 2021 and 2020, many, many features were added. New voltage regimes are added to our power platform that are getting very big traction. And you're correct. When you have substantial increases in revenue during a market pullback, that's obvious that you're gaining share and you're gaining served market. Marco, if there's any color you want to add to that?

I think you've covered it. I think -- yes, also in the near-term growth, in addition to automotive, it's been the 65-nanometer platform. As Russell mentioned, we're adding features to now address new markets, but that also has grown in the last years or a couple of years significantly. And it's not a new platform or relatively new platform.

Where we still add features.

Right.

Okay. Good question. Thank you.

The next question is from Cody Acree of Benchmark.

A follow-up in here, guys. The -- I guess what I wanted to ask about is, have you done a dissection of your business by end application that would pertain to AI? I know you've got the data center business. Would this be primarily your optical transceiver business? Or would you lumpsum power management in that as a driver? And just curious if you had a look at how that is the driver of your business?

Predominantly, everything we sell is somewhat agnostic to AI or non-AI. It's very much needed in data center. The one thing that maybe is more of an enabler for AI would be CDRs because of latency. But that we have to see how that plays out and how what really happens in the market, meaning staying with some CDRs rather than DSPs. But if you look at, for example, everything we're doing with silicon photonics and really very big momentum, big customer pull, complete platforms with and without integrated laser, I would foresee and believe that SiPho is going to be a very, very big requirement for all of the activities with AI. But -- so -- but really, it's -- what we do is not necessarily devices between AI or non-AI. It's an enabler for data center period. Everything that we do with active copper cable, active optical cable, it all ties into the same capabilities, and it's really to enable speed at lower power. But Marco, anything that you would want to add there?

Yes. No. I think you mentioned power management. And I would say the same is true there. We do have some power management applications going into data centers. And as clusters -- AI clusters grow within data centers, they'll need power management. We're not doing anything specific to AI in that regard, but it will just provide additional market for us.

And then lastly, Russell, have you taken a look at what is your team looking at, how you might use AI in your own business as a driver of efficiency?

Yes, and yes.

Care to elaborate?

Effective operations deals a lot with AI. It deals with not just preventive, but predictive activities to prevent any type of excursion and to maximize output. So we have quite a bit of activities going on in that area with regards to operations. As far as our business itself and not internal to Tower, we have many direct strategic sessions focusing, be it with what we're doing with sensors, with EDGE AI, be it with what we're doing with data center to further understand and see how we can become more accretive towards enabling cluster AI, if you will. I mean our end customers are the AI -- the predominant AI champions in the world. So in both areas, we're looking at our business to always understand what can we do with our offerings to differentiate more full to enable AI and as well using what we can within the company to improve our operations.

And Russell, just one last thing for Oren. Again, with the cost of the wafers for the Intel agreement and back to STMicro as well, are those just going to be gross margin dollars, whereas as Intel is running the fab and carrying the operating costs in R&D or in the SG&A that would be allocated for those wafers? Or is it just quite a quick sales over on a gross margin dollar?

Yes. It's only -- like I mentioned before, it's only cost of goods because from our point of view, Intel is a manufacturer and all the payments that we will make to cover whatever will come against our cost of goods. Exactly the same, by the way, like with STMicro. In the end of the day, you just need to note that when we expand utilization in existing fabs that we have, we have the incremental model of 50%, 55%. But when we do inorganic growth, like STMicro, like when we acquired TPSCo, so you should not expect a 50%, 50% incremental model because it's inorganic growth.

I think maybe one very strong clarifier, that is important. Be it the Rio Rancho capacity agreement that we just signed with Intel or be it the ST agreement, the technology was, is and remains Tower technology. The road map remains a Tower road map. The customers or Tower customers, the customer interaction is a Tower customer interaction. So there the role of Tower in it is very strong as far as being the technology innovator and technology provider and partnering with someone that has a very strong, good manufacturing scale to enable growth in our manufacturing scale that we believe overall is accretive to our financials. And many partnerships should be beneficial to theirs as well.

And I guess, Russell, just to finish that thought, is there any minimum take-or-pay type capacity agreement? Or is -- I guess, what is Intel or STMicro holding out for their own capacity?

Well, if you look at the contract, the contract with Intel was for 600,000 photo layers. The Rio Rancho facility is a very big facility. So they have much capacity on top of the 600,000 photo layers. In the case of Agrate, we announced what percentage we would have of those factories. The rest totally belongs to ST -- of that factory, the rest is ST. Within the financial model itself are there commitments that we have on what we must buy or not buy, that's not public, and that's not something that I would get into at this point. But as stated, it's -- in both cases, I think that there are good agreements that Tower believes is the right agreement and that either Intel or ST believes is the right agreement for them.

The next question is from Natalia Winkler of Jefferies.

I wanted to follow up on Agrate. I think one of the questions I had was you mentioned that there was the first phase, right, and the CapEx effect, the $500 million is related to the first phase. Could you possibly speak about any potential phases beyond that? And how should we think about that from CapEx or any standpoint?

Yes. So like I said in my prepared remarks, I said we have an option to invest an additional several hundreds of millions of dollars. We continue to increase our capacity in Agrate, Italy. We still didn't stop or make a decision that we will stop, but we have this option. Now I'd like -- as Russell mentioned, we just this year are ending the process of qualification and starting to ramp production, will start to ramp production in the end of this year. So obviously, until we reach the full capacity utilization of what we are spending for $500 million, it's too early to make that decision. We don't need to make the decision currently.

Understood. That's very helpful. So the -- I guess, the phase 1 is sort of the $500 million and the phase 2 would be $700 million. So together, those 2 phases could be up to $1.2 billion. Is that fair?

We did not -- it's true that this is phase I, and it's true that we have an option for phase 2. But we never said or disclosed the amount that you say how much will cost the next phase because we still did not define it and so we don't have yet the [ true list ] and the exact [indiscernible]. So we never said what will be the cost of that phase.

Understood. Sorry, I think I misheard in my line. I apologize. And then, I guess, the second question is on maintenance CapEx. Like once Agrate becomes operational as you guys start seeing some revenues, how should we think about your maintenance CapEx on a quarterly basis? In the past you spoke about $45 million to $50 million. Is there going to be any structural change to that number once Agrate is ramping?

No, no. We believe that quote I gave $180 million to $200 million per annum will also capture any maintenance CapEx for that factory or to not be higher.

Understood. And then could you possibly help us with any updates on the CHIPS Act funding? Are you guys -- is it something you should get an application for? Is that something you're considering?

It's a very good question. We were seriously considering it at one point, independence of Intel acquiring us. At the time, we were in discussions with several states. We've held off on going forward, and I don't think that we've lost anything. And having done that, it's things haven't moved very quickly. We're reconsidering now if there's something or two that we might wish to do in the U.S. in addition to what we're doing that might itself makes sense to do with the CHIPS Act. And we'll potentially be filing something. Nothing that we've done as of yet.

And then, Russell, one last one for me. So as you guys are obviously signing some of the agreements with Intel on capacity, and it sounds like that's kind of a 60 units, fixed photo layers that you guys have there. Are you considering kind of maybe changing some of your customer relationships to include more of the long-term agreements, maybe something along the lines what some of your peers are doing?

It's a very good question. I would say that in any long-term agreements, you have to maintain some flexibility. Otherwise the long-term agreement winds up being a long-term losing a customer. But -- and people really do need to be very careful around it. Specific to the different businesses, you ask for different things, it depends on how much you're investing, specifically for an end customer's road map. And within that, you'll ask for some different terms. But are we looking specifically at changing the way that we do business? No. Are there some certain circumstances where we might want to have greater assurances over time? Yes. And there's many different forms that one can get greater assurances, multiple, multiple different models. And the point is, is to do something that's in the worst case, it's not punitive to a customer. And in the worst case for you, you don't lose money. So -- and there's multiple models that can be done after that drive that type of an agreement. But we're not fundamentally looking at changing the way we do business now.

This concludes the question-and-answer session. Mr. Ellwanger, would you like to make your concluding statement?

Yes. Again, I thank really, everybody for joining the call. We're excited about Tower. We're excited to be Tower. There's a beautiful statement. It's in a statue outside of the National Archives in Washington, D.C. And as I had walked by there, it hit me very, very strong. And it's something that's been very strong in my last few months and in particular, in light of the past 2 weeks, where we decided mutually to end the term of the definitive agreement. But the statement is the heritage of our past is the seed for the harvest of our future. And as I look back on Tower, at least since I've been here in 2005, what a heritage we've created. So much has happened in the company, so much growth, so much excitement, and it's quite rich. As you look at that, it really is a seed to build off of the plan to say everything we've done well. How do we do it now even better to increase the returns. Within a heritage or as well certain things that you don't want to do again that you'll avoid because of lessons that learned. But if you look at it, the heritage has been a very, very nice success heritage, having gone from the circa $90 million revenue in '05 to has been brought up in this call of $1.7 billion run rate that we had in 2022 with very, very nice financials across the board. I mean without any onetime events very, very nice net profits running up to close to 20% on a sustainable basis. So it's been an incredible heritage. And now it's a question of making that be the seed of the future harvest, and we're at about that. We're excited to reengage with the market. For those covering analysts that have been with us for a while, thank you, and thank you for being with us now, and we look forward to further discussions and help you develop your models. For investors that have been with us for a while, thank you for the trust. We look forward to continuing to work with you. And for those that are considering to invest in us now, we can promise journey. And we're -- I believe it's really true. The heritage of the past is indeed the seed for the harvest of the future. And our heritage has been a very, very good one. The harvest we expect will be a wonderful harvest. So those are my closing remarks. And thank you all very, very much. Bye-bye. Thank you.

Thank you. This concludes Tower Semiconductor Conference Call. Thank you for your participation. You may go ahead and disconnect.