Good morning, everyone. Welcome to GlobalFoundries's 2026 Investor Day. It's great to see so many of you here, and a warm welcome to those joining us virtually. We have an excellent lineup for you today of our excellent speakers, and I wanted to just go over a couple of things before we begin. The presentations will take approximately 2 hours long. We will have a 10-minute break after the third presentation, and we'll conclude with a 45-minute Q&A. It's been almost 4 years since our last in-person event, and a lot has changed. We're very excited to show you GlobalFoundries today. Do note that the slide deck and the replay will be made available after the event. Now, for some housekeeping. Note that today's discussion includes forward-looking statements. These statements are based on current expectations and involve risks and uncertainties that could cause actual results to differ materially.
Shown here and included in our materials is our safe harbor disclaimer. Note that we will be talking about non-IFRS metrics today. Before I hand it over to Tim Breen, our CEO, customers are very important to GF, and I'm glad and pleased to show you a customer testimonial video. Thank you, and welcome again.
Hello, everyone. I'm Charlie Kawwas, President of the Semiconductor Solutions Group at Broadcom. It's an honor to be here and to celebrate the strength of our partnership with GlobalFoundries. At Broadcom, we take pride in the fact that 99% of the world's data traffic flows through at least one of our chips. As AI increasingly disrupts our daily lives, our focus has been on open, scalable, and power-efficient foundational silicon for this once-in-a-lifetime transformation. Partners like GlobalFoundries have been critical to this journey. For decades, GlobalFoundries has been a trusted foundry partner to Broadcom, delivering innovation, reliability, and scale. GlobalFoundries' footprint with fabs in Asia, Europe, and the U.S. ensures solid execution and supply chain security. As more of our joint technologies move to their U.S. fabs, our strategic alignment only grows stronger.
Their leadership in RF-SOI, SiGe, and advanced packages enables us to push the boundaries of wireless and wired communications. We have partnered together with GlobalFoundries to be first to market with their most advanced technologies. Our relationship is a shared vision for the future. Together, we're advancing U.S.-based cellular connectivity, scaling AI technologies, and driving innovation that benefits our mutual customers. Thank you, and I wish you the very best.
A little over 8 years ago, I found myself in a blizzard in upstate N.Y. I had just joined the leadership team of GlobalFoundries, and as I walked the halls and factory floors of this company built on the incredible foundations of AMD, Chartered Semiconductor, and IBM Microelectronics, I came to appreciate the unique assets that we had. Over the years that followed, we clarified the company's strategy, honed our execution and quality, built a world-class team and a sustainable, profitable business model. The last 8 years have been incredibly fulfilling. I think for myself and for all of us at GF, they pale in comparison to our excitement about what is to come. AI represents a generational shift for the semiconductor industry, both in the cloud and in the physical world around us.
If that was the only story we would tell today about what's propelling our business forward, that would already be extraordinarily exciting. It's happening against a backdrop of a changing world, where supply security and the need for resilient semiconductor manufacturing matters more than ever. If I leave you with one thing today, it's that sense of opportunity. Opportunity that meets a company well-prepared to capitalize on it. Today, as Eric mentioned, we're gonna take you through four presentations. I'm gonna kick us off with our thesis on the three key mega trends driving the semiconductor industry. Greg will share a modicum of his boundless enthusiasm for our technology portfolio and our roadmap for the future. Mike will take us deeper into the end markets that we serve and showcase specific solutions we're delivering to our customers.
Sam will bring it home with a financial model that doesn't just demonstrate top-line growth, but performance improvement at the bottom line, creating and delivering value for our shareholders. We see three mega trends shaping the semiconductor industry. In the data center, the shift from generative to agentic AI to thousands and millions of agents operating autonomously at scale creates a surging demand for intelligence. When you add to it the imperative of sovereign AI, it's clear why data center CapEx numbers are only increasing. This year, the top four hyperscalers will spend more than $700 billion on data center capacity. That's more than 2x what they spent in 2025. With that build out comes new challenges. The challenge of connecting all of that compute together with massively parallel workloads, and the challenge of providing power to all of that capacity.
If you think the data center boom is big, we think the echo of physical AI in the real world will be even bigger. As AI enables new use cases, some of which we can predict, but many of which we can only imagine today, AI will unleash huge opportunities for our industry and create new requirements. Again, all of that is happening against a backdrop where the old rules don't apply, where supply chains are being reshaped by factors such as geopolitics, supply security, tariffs, and trade barriers. Let's go deeper into the data center. We're gonna talk about 2 trends that we're laser-focused on within GLOBALFOUNDRIES. AI workloads generate an incredible requirement for moving data, and they do it through a more and more heterogeneous compute architecture.
Just to give you a sense, an H100 GPU moves more data through its memory every single second than the equivalent of 1,000 human genomes. If you tore apart an AI data center rack, you'd see as many as seven different forms of compute. GPUs, CPUs, accelerators, switches, NPUs, DPUs, all working together to conduct complex training and inference workloads. That drives a necessity for a step change in networking, and it's accelerating that transition from copper to optical that we've talked about for a decade, but is finally upon us today. At the same time, power is one of the most incredible challenges for data center scale. A modern AI campus uses more power than the island of Manhattan. A rack in a data center in 2027 will use 50 times the power of the average rack powering the internet today.
This power challenge is huge. It requires innovations in systems and modules to provide that power more efficiently to those enormously power-consuming workloads. Let's go deeper into optical networking. It's clear from a number of performance metrics why optical is accelerating. Simply put, moving photons is easier than moving electrons, so it extends reach across networks. It's also easier to move more data through smaller spaces, increasing bandwidth density. It does all of that while consuming less energy per bit moved. Those metrics are all pretty compelling, but I think one metric on this page I want you to take away is the profound impact that optical networking has on compute efficiency. Today's CPUs and GPUs sit idle more than they should because they're waiting for data. Optical networking transforms the utilization of XPUs, enabling a step change in compute efficiency.
This is why in 2026, the whole industry aligned on the transition to optical no longer being a maybe, but being a now and actually being a yesterday. How fast can we go? There will be interesting discussions about pluggable near and co-packaged optics. At the heart of all of those systems are foundational technologies such as silicon photonics, which enable those systems to perform and conduct those complex workloads. With the industry increasingly galvanizing around standard approaches, we see this transition only accelerating. If we get into power, if you think of networking as the nervous system of the data center, then you should think of power as the blood. Today, with these racks moving from kilowatt to megawatt scale, a step change in power management is required.
With the conversion to high voltage DC requiring a very different power architecture from grid to chip, reducing conversion stages to as few as 2 stages, operating today at 400 volts, tomorrow at 800 volts. This requires much more dense power conversion. All of this works together to transform energy efficiency. If a 10% improvement of efficiency doesn't feel like a lot, think about that as a 40% reduction in losses. When power is the gating factor to data center scale and cost-efficient inference, 40% reduction in losses is a game changer. Again, this doesn't happen without semiconductor technology innovation. BCD for control, protection, mixed signal intelligence. Low voltage GaN for high frequency switching and high density power conversion. Together, creating a step function improvement in power management. In GF, we think about our competitiveness on 3 dimensions.
We think about our foundational technology leadership. We think about the design support and ecosystem we enable for our customers, and we think about our global footprint. In technology and optical networking, we have industry-leading portfolios in silicon photonics for that optical networking transition. We build on that differentiation with advanced packaging, 3D integration techniques, through silicon vias, broadband couplers, and more. We complement that with in-house design expertise that we've built organically and acquired with companies like InfiniLink. We enable a global ecosystem of EDA test fiber assembly partners, allowing our customers to reduce time to market. I think very importantly, and sometimes not well understood, is we serve that from a uniquely scaled global footprint. We manufacture silicon photonics 2.5 hours from here in upstate N.Y. in a modern fab well-equipped for scale.
With our campus in Singapore also ramping, providing our customers with supply chain optionality. We complement that with in-house advanced packaging and test, truly creating a one-stop shop. In data center power, we have industry-leading BCD and GaN processes today in the market with integrated drives and controllers, with integrated passive devices allowing us to shrink board area, allowing more efficient voltage regulation and power delivery to the chip. We enable our customers with advanced analog models and high voltage models, all of that allowing them to tune system performance, shorten development time, and reducing integration risk. Again, we serve this out of a global footprint in Germany, Singapore, and the U.S. Let's talk about physical AI. I think the way to think about physical AI from our perspective is thinking about technology waves, there's common factors that affect every technology wave.
What starts with an infrastructure build-out unleashes new applications, some of which we imagine, some of which we learn along the way. That leads to ultimately to a transformation of that technology vertical. What's also true about technology waves is each one builds on the last, and so they go faster. If you think back, some of us in the room can think back that far, to the launch of the commercial internet. It took about 30 years to get to the point of the globally connected cloud that's in use today. If we think to the rollout of cellular infrastructure and the enabling of billions of smartphones and further billions of smart devices, it took about 20 years to get to the point we're at today. Our conviction is that physical AI will happen even faster.
Within a decade, the boom we see today in the data center will be met with an echo in the physical world with tens of billions of devices unleashed and empowered with AI. What makes a physical AI device, and why is it different to something you might find in a climate-controlled data center? A physical AI device has to act in the real world. It has to deal with unpredictable environments. It has to be able to do 4 key things well, reliably every single time. It has to sense, to gather data using multimodal sensors, radar, lidar, time of flight, imaging. It has to be able to think, to process those data in real time with low power and low latency. No time to go back to the cloud to consult for a workload.
It has to be able to act, to control motors, drives, actuators, real world devices. It has to be able to communicate primarily in wireless form factors with its own system and with devices around it. As you can imagine, this drives a broad need for different semiconductors, and it lines up perfectly with GLOBALFOUNDRIES' portfolio. Even more so, the acquisitions we've made with MIPS and Synopsys ARC fit perfectly to the processor requirements that underpin these use cases. One of the best things about my job is the broad range of companies I get to interact with, some very large and some very small. Let me just give you a couple of vignettes into conversations that I've had just in the past weeks about some of these areas.
In transportation, last week, I was talking to one of the leaders in autonomous driving, and they were explaining to me their need for low-light imaging. Low-light imaging, one of the best ways to do that is what's called a SPAD sensor, a single photon avalanche diode sensor. That enables their autonomous systems to work in low light conditions, in poor weather, enabling them to see further and make better decisions, keeping those systems safer. Truly an inspiring use case. In the industrial world, we work with partners like Siemens, who dramatically are reimagining how a factory should work, empowering AI in all aspects of those processes, be it robotics, be it factory workers, enabling a step change in productivity.
On the consumer side, we've all seen the pictures of robots and humanoid robots in the world around us. I've been to some of these factories, I've been to some of these locations where they're developing and prototyping. By the way, we're using them in our own factories too. We're tasting truly, our own medicine and seeing how these devices can perform. Lastly, I think if I had to pick one area of physical AI that is closest to my heart, it's medical. When healthcare and electronics combine, incredible things become possible. Right here in New York City, I was having coffee with a technologist who was telling me about the device he's working on. He's building a sensor, looks a little bit like the CGM that you might see on someone's arm, that detects the precursors to a heart attack.
The second generation of this product will also connect to a delivery device that can deliver targeted medicine to reduce the likelihood of heart disease. 18 million people die every year from heart failure. This could address 30%-40% of those deaths. Again, physical AI isn't just a business opportunity, it's also something we have a responsibility to enable to improve those outcomes. From a market point of view, of course, we're excited about the growth. We've talked about this being an $18 billion market by 2030, but really this is just the tip of the iceberg of scaling AI into the physical world. Let's talk about the last mega trend, the need for resilient global semiconductor supply. This is clearly not the world that many of us grew up in.
Every day you read the news you're surprised by another development that puts strain on global supply chains. Where the operating principle used to be economic efficiency, today it's resiliency, geopolitics, national security. That's putting stress across the whole system. With rising uncertainty, this is leading to changes to policies at the country level, at the company level. It's no longer sufficient for people to wait and see what happens. People need to take action. This matters in every industry, but semiconductors it matters even more, and it matters for two key reasons. Firstly, we're beginning from a fragile starting point. 50% of essential semiconductor capacity is in China and Taiwan today, and that alone is enough of a risk to motivate action.
When you think about the role semiconductor played, and we got to see this live during the pandemic when factories were idle, when cars stood on the parking lots, when jobs were affected. The trillions of dollars of downstream economic activity enabled by semiconductors, the jobs that they support, and the critical applications, including defense, that they enable all mean that addressing this challenge is no longer a discussion. It's no longer on the risk register being monitored. It's being acted upon, again, at the country level and at the company level. GLOBALFOUNDRIES was born for this moment. We have the industry's most flexible global footprint. Not just fabs around the world, but over the past years we've qualified technologies in those fabs so a customer can design one time, tape our product out, manufacture it on three continents.
This creates incredible flexibility and optionality for our customers, but it also creates flexibility for us. We can respond to demand changes. We can react quickly. We can bring on capacity within our clean room space, adding in a capital efficient manner. Because those factories are incredibly important for the governments we work with every single day, the support levels we have from them for that build-out is the highest we've ever seen in our history. Since our fabs are not just automotive qualified but defense certified, we're able to build the most critical chips in all the geographies that matter. It's great having a global footprint. It only matters if customers are willing to use it. Last year when we announced our plans here in the U.S. to accelerate our investment journey, we had incredible support from the industry.
Not just from the fabless companies that you might expect, but strategic OEMs like SpaceX, Apple, General Motors. For them, what we're doing here in the U.S. is a critical part of them reducing reliance on China and Taiwan and reshoring those supply chains back to the U.S. We've announced 11 partnerships for reshoring. We had $3 billion of design wins in 2025 alone, driven by the reshoring trend. With more in the works, this momentum is only building. We take a step back and we look at the momentum across the broader business, we clearly see the leading indicators of our future growth. We had 500 design wins in 2025, more than a 50% increase over 2024. Perhaps more importantly, 95% of those wins were awarded to GF on a sole source basis.
That reflects the trust that our customers put in us and the fit to their portfolio of the technologies that we provide. When I think about our engagement, not just with our customers, but in many cases our customers' customers, and sometimes even our customers' customers' customers, we've never had this level of deep engagement across the value chain before in critical areas including industrial, automotive, mobile, hyperscalers, and the aerospace and defense industry. Again, clearly traction and momentum building for our business. If I net all of this out, hopefully it's clear why we're extraordinarily excited about what's to come. These three drivers of demand are extraordinarily powerful. As the data center build-out continues, as we hit new bottlenecks that require changes to networking and power, GlobalFoundries is ready.
As AI proliferates that physical world and those use cases, some of which we can only dream of today, the true statement of science fiction becoming science fact, again, we're ready with our portfolio of technologies, and we enrich that further with our IP. Lastly, we're ready for that reshoring trend that we see accelerating. The transformation we've undertaken over the last eight years has made us ready for this moment. We've built an incredible technology portfolio, which Greg will share with you. Not just a portfolio for today, a roadmap far into the future. We've enriched it with targeted acquisitions such as MIPS, ARC, AMF, InfiniLink, all adding specific capabilities to that roadmap.
We've built a global footprint, and we've readied it with technologies available in multiple geographies, and we prepared for growth with expansion plans that can hit the market quickly in a capital efficient way with deep government partnerships supporting those investments. Perhaps most of all, and where I spend most of my time, we've transformed the way we work with customers, not just building a broader set of solutions for them, complementing manufacturing with IP, software, and services, but also building trust. In the end, our business thrives on trust. Our customers trust us to succeed. In fact, they're counting on us to succeed. We hope you share our enthusiasm today, and we look forward to telling the rest of the story to you. For now, let me hand over to one of my customers in Japan.
Good morning. I'm excited to share how Renesas is advancing its strategy for intelligent connected systems and how our growing partnership with GlobalFoundries is helping us get there. The future will be shaped by smarter vehicles, more efficient infrastructure, and secure edge devices. Meeting these needs requires more than innovation alone. It calls for reliable execution, supply assurance, and technology platforms that scale globally. That's why we partnered with GlobalFoundries. GlobalFoundries' 22FDX+ platform is enabling us to deliver more compact, energy-efficient chips that power everything from next-generation automotive microcontrollers to connected medical devices. These solutions are designed for always-on applications where power savings and performance are critical. We are also working with GlobalFoundries on advancing power technologies, including BCD and Gallium Nitride. These platforms are essential for high-efficiency power conversion in battery systems and data centers, especially in demanding automotive and industrial environments where Renesas has deep expertise.
GlobalFoundries' global manufacturing footprint in New York, Vermont, Germany, and Singapore provides the geographic flexibility and supply chain resilience that complement Renesas its own. Together, we are giving customers worldwide the confidence that their systems will be supported and scalable for the long term. This partnership is not just about today's products and solutions. It's about enabling tomorrow's innovation. At Renesas, our purpose is to make our lives easier by complementing human capabilities. Together with GlobalFoundries, we are laying the foundation for the next wave of intelligent, secure, and sustainable technologies that will transform how people live and work. Thank you. We look forward to further strengthening our partnership with the GlobalFoundries team.
Good morning. I'm going to guess that you know most of the other speakers that are presenting today as they routinely engage the investor community, earnings calls, et cetera. Me, not so much. I spend most of my time interacting with my counterparts in our customer base, with our partner ecosystem, with universities and research consortia around the world. A bit of an introduction may be warranted. I've been 42 years in the semiconductor industry, specifically in technology and manufacturing. The more relevant point associated with that is the 17 years at GlobalFoundries. I joined 1 month after we spun out of Advanced Micro Devices in April of 2009. What that does is it makes me a bit of a corporate historian of sorts.
More, maybe more importantly, it gives me a position of authority in talking about the transformation of our corporation over that 17-year period of time. That gives me the two points that I want to convey to you today. First and foremost, as you heard from Tim Breen already, we have the strongest, deepest, broadest technology portfolio in the history of our company. Second, we're the only company that has all of these capabilities under one roof. In my presentation today, I want to talk about the proof points underpinning that breadth and depth of that technology portfolio. I'll start to touch on how our R&D investments are gonna translate into the financial returns that Mike and Sam Franklin are both gonna talk about. Let's dive into the technology portfolio and where we've been placing our investments.
I gave a keynote talk talking about an earlier version of this chart about four years ago, I said, "I think I have the coolest job in the industry." I look back whimsically now at the fact that I had no idea how much transformation we were gonna make in the four years that have passed since I made that talk, I'm as excited as you can possibly imagine. The best way for us to represent our technology platforms for our corporation is to map them into the 2 end markets that represent the growth vectors for our corporation. Those 2, as you've heard from Tim Breen already, are AI data center infrastructure and physical AI. In the data center infrastructure, it's as you see, 2 essential technologies, advanced packaging and power delivery.
In silicon photonics, it's pluggables and co-package optics, and I'll touch briefly on quantum computing. For physical AI, in addition to our ultra-low power CMOS platforms for FinFET and fully depleted SOI or FDX, with the integration of MIPS, it brings in the RISC-V core family, the software ecosystem that wraps around the products for those cores, and importantly, the introduction of custom silicon capabilities. Now, the underpinning of this is a broad array of other technologies that address automotive, IoT, aerospace and defense, things like our gallium nitride for fast-switching charging capabilities, the BCD platform, our historical strength in RFSOI and silicon germanium by CMOS, and more recently, additions of RF gallium nitride for fast switching in mobile handsets, as well as power amplifiers in cellular base stations. Finally, in feature rich, our workhorse 40-nanometer node with embedded non-volatile.
It supplies most of the high-end automotive manufacturers, and even the addition of a low-cost or cost-effective 22-nanometer bulk version of the technology. The key takeaway through this, as a consequence of the R&D investments that we have made and the acquisitions and integrations in particular of MIPS, we have created a technology portfolio that redefines GlobalFoundries not as a semiconductor process technology company that offers manufacturing services, but as a holistic technology partner. Let's dive into those, starting with the data center portion of it. We'll start with photonics. I've been going to the Optical Fiber Communication conference, or OFC, for many, many years. The news from this year's OFC conference was photonics has arrived. It's very different than the previous years, where it was how much further is copper interconnect going to be pushing out the ramp for silicon photonics. That wasn't the discussion.
If anything, it was a debate about co-package optics versus pluggables and which one's gonna win and how is that market gonna evolve. Of course, it's not just that it's arrived at an OFC conference. We see it in our financials as well. Disruptive technologies like silicon photonics take a lot of time to mature. It was a decision we made in the summer of 2015, believing that silicon photonics technology was ultimately gonna be required to supplant copper in the data center. The fact that we are now on the steep S curve, and we're in that rapid adoption phase is as a consequence of those investments we made more than a decade ago. Let's look at the portfolio that we have.
Today, in the pluggable space, it's 1.6 terabyte per second to 200 gigabit per lambda, going to 400 because we know that's the step to get to 3.2 terabit per second. We've already demonstrated that. We're working My GF Labs team is working on 800 gigabits per second. It's important to know in the silicon photonics space, many of the customers are not deeply experienced in knowing how to handle photonics, and they need help, in particular, in things like optical test and packaging. We have created an ecosystem and invested internally to support those activities as well. Our manufacturing footprint is an asset for us. You heard Tim talk about the geographic capabilities that we have. Through the acquisition of AMF in Singapore, we have 200 millimeter silicon photonics.
AMF has historically served long-haul telecom with coherent capability. In a minute, I'll come back to why that's going to wind up being a very important acquisition for GlobalFoundries. As Tim said earlier, our CLO photonics platform is north of here in our advanced 300 millimeter facility. Of course, to enable customers, you have to have the infrastructure. We have invested for many years in creating the right design kit capability that has both the optical and electrical capabilities that you need to design a electro-optic device there. In addition to that, we've created an OSAT ecosystem with the right test capabilities to enable our partners to manufacture at volume as part of their supply chain solution. There's no other company that has this capability for photonics technology and manufacturing scale to address the marketplace.
It's the proof point on why we are the clear leader in silicon photonics today. Data center architecture is a very important consideration. There are two news items that I'll highlight coming out of OFC this year. First, at GTC 2025, for the first time, they started talking about scale across. You know that scale up, of course, is the investments in the rack. Scale out is within the data center and scale across. Why is scale across an important consideration? As data centers begin consuming more than 500 MW or GW of power, the need to be able to disperse your data centers regionally is important. Historically, you've used connections between data centers for backhaul. This era now is connecting GPUs to GPUs because you're accessing the power grid separated by tens of km.
Scale across becomes an important consideration by creating large language models, GPUs-GPU connections 10s of kilometers apart. The fact that it uses coherent communication protocol is an important capability for us, something that came in through our AMF acquisition. You also see a list of a number of standards. There were three standards announced at OFC this year, which for a maturing industry is a very important consideration. I'll call out one of those, OCI, Optical Compute Interconnect, is a very important consideration, and I'll come back to that. The point of this chart, scale up, scale out, scale across, GF has silicon photonics platforms, optical test, advanced packaging, that covers all of those in addition to the manufacturing scale.
While co-packaged optics gets a lot of the news today, we know that pluggables is a very, very important market that will continue to grow for years to come. Our roadmap is tied to the industry cadence. That's the 1.6 terabyte per second today that is in volume manufacturing going to 3.2 terabytes per second. For GlobalFoundries, we distinguish ourselves in three ways with our pluggable solution. First and foremost is our ability to have a very low insertion loss fiber attach process. Our competitors have to do a MEMS-like technology. We have a very elegant solution that has the lowest loss insertion that can be done. Second, because this uses CWDM or coarse wave division multiplexing, you have to mux and demux the signals coming in.
We're capable of doing that without the use of heaters, which is a simplified design and reduces the overall power envelope for the pluggable solution. In addition to it, our advanced silicon germanium capabilities that are the TIAs and drivers couple very nicely into this overall solution with best-in-class performance. Now let's dive into co-package optics. Earlier this week, we hope that you have seen a press release where we announced our co-package optics optical module solution, which is SCALE, which stands for Silicon Photonics Co-Package Light Engine. It's what you see in the purple outline here. This is a optical module that has an electronic IC bonded to a photonic IC. The complex piece of this is the light in and light out, where you have to have waveguides that are coupling to a turning device, ultimately coupled to a fiber array.
Complex system there. I'll come back to this OCI, this optical circuit, optical compute interconnect solution. It's important for 3 reasons. First one is look at the list of the founding members that have come up with this multi-source agreement. You could not ask for a better list of companies that are the thought leaders in the architecture and the requirements in the data center. Number 2, it's a very forward-looking standard, and I'll go into some detail as to why that is important. Number 3, our SCALE solution is the industry's first OCI-capable solution. Let's get into a little bit of detail on why this is such a forward-looking standard. First of all, what it does is it goes to 8 wavelengths per fiber.
Practically what that means is many fewer fibers that need to be connected to your device because you can get bandwidth with 8 lambda or 8 different wavelengths on an individual fiber. That's a very important consideration on there. The way you get that is through use of both CWDM and DWDM. CWDM means that you've got 2 groups of wavelengths about 20 nanometers apart, 1310, 1330 nanometer wavelengths. Within there, you're doing DWDM separation of 4 wavelengths, each 2 nanometers apart. That uses DWDM. The fact that our SCALE solution is forwardly compatible to enable that OCI solution is not a mistake. Back when DARPA was investing, you know that this was early on, investing in creating DWDM capabilities with AMD and Ayar Labs, GlobalFoundries was their partner, and we established both 8 lambda and 16 lambda capabilities on there.
We feel like as a consequence of the demonstration of the technology we made years ago, it actually gave confidence to that list of founding members to establish the capabilities in this standard. We feel very, very good about our path forward. You have seen that the SCALE solution puts together a very complex engine on there. It takes an ecosystem to do that. We have invested aggressively to ensure that our customers and our internal manufacturing has all the necessary capabilities. If I just highlight one of these categories, right, this detachable fiber here that I just got done talking about. Our approach to this has been very systematic in that we worked with Teramont, Cinco, and Corning and designed that fiber coupler to be broadband capable.
Meaning that we already knew years ago when we started doing this work that DWDM doing 8, 16, 32 lambda was gonna be a really important capability. You can well imagine we're gonna continue to invest. In a few years, you'll see the same chart that will show advanced materials providers for the modulators needed for 800 gigabit and beyond. We've made a lot of investments to establish the capabilities that I've just shown you and the proof points for it. We also know that this is in a market with a very attractive compound annual growth rate. You see the SAM growing at 40%. Expectations that material volume in co-packaged optics will start in 2028.
We had a photonics and advanced packaging webinar a few months ago, and there we, one of the announcements we made is that we've pulled in the acceleration of our revenue profile at $1 billion from 2030 to 2028. Today, we believe that we can actually achieve $2 billion in revenue by 2030. I wanna stay inside the data center and talk about the second topic that Tim mentioned, which is power. Starting with the context of our leadership position before I get into the data center. We have had a long-standing leadership position in smart mobile devices, having shipped hundreds of millions of power amplifiers and haptic drivers.
I hadn't planned to mention the customer that we're doing that with, but they were very gracious in their earnings call last night when Cirrus Logic highlighted the strategic partnership that we have with them in being a really critical supplier to them for many, many power devices for smart mobile devices. Leadership position here has been established. Secondly, in automotive, we were fortunate enough to have worked with the leader in battery management system power management just as the electrification trend was taking off. As a consequence of that, they had very large market share in our 130 nanometer high voltage BCD technology, you know, was the technology that they've ramped with. Very important. We're now driving that investment profile for the data center, and of course, it's a very exciting one, and I'm gonna go into some detail.
Why, first of all, is power delivery such an important consideration in the data center? Data center racks, I'll take the Rubin Ultra rack. It'll operate at 600 kilowatts in a single rack. The racks aren't getting any bigger, you can imagine what that does for the power density. Let me walk through the 3 columns, top and bottom here, talking about why power delivery is such an essential technology. The first thing that happens is as a consequence of that power dissipation in the rack, most of the power going in gets converted to heat. You're operating in a very, very thermal sensitive environment. You heard Tim already talk about the fact you're going from 480 volt AC to 800 volt DC.
Most of that's to reduce the total amount of copperage for getting power into the rack. It also puts a new set of requirements on the voltage conversions, the step-downs that you have to do. At the same time, you're trying to go from maybe 5 or 6 step-downs to 3 or even 2. You get the point that there is a tremendous amount of effort you have to put in the design of your voltage regulators to do that. Of course, the key metric is how many watts you're putting in for the performance of the GPU, right? That GPU performance per watt is the key metric. Having said that, reliability becomes the second significant challenge. That rack at 600 kilowatts, when it's not running, it idles at 100 kilowatts.
It's still an enormous amount of power when you think about it, but it idles at 100 kilowatts. When it turns on full computation, it goes to 600 kilowatts in milliseconds. If you go do the math, 2,500 amps per microsecond. The current transients that these voltage regulators have to go through is tremendous. That shows up, first of all, as a ruggedness parameter. The ability to go through 2,500 amps per microsecond, switching this thing on, is an incredible feat of engineering. It also runs at a very high temperature on there. I mentioned earlier the automotive work that we have done. Those battery management systems have to be capable of 150 degrees Celsius operating temperature. This is in a very similar realm. Finally, highlighted already the need for very fast-switching devices.
As a consequence of going to the smaller power supplies, the form factor becomes a key consideration. You'll have to offer the designer high switching frequencies, but also the ability to have heterogeneous integration through silicon vias. What you want to have is your GPU and the voltage regulator in as close a proximity as possible. That requires engineering of the form factor as well. Let's dive into a specific configuration and talk about where GlobalFoundries, gallium nitride, and BCD solutions map into that data center. This happens to be a data center with a 3-step down, 800 volt going to 48 volt step down to a 6-volt intermediate bus conversion voltage level, and then ultimately the distribution down into a few hundred millivolts for distribution into the GPUs.
A Vera Rubin Ultra tray will operate at 10 kilowatts. Because you're delivering at less than 1 volt, you do that math, you're putting close to 13,000 amps into each tray. That's an enormous amount of current that you have to put in there. What that drives is the need for 50 to 100 voltage regulators per tray, each of them handling hundreds of amps. You get, first of all, the impression about the attractiveness of the market, but also the intense environment in which these devices have to do. I've given two examples. The third one, of course, is a step from 800 down to 48. For our 100 volt GaN solution, it is all about high reliability. We originally developed our GaN solution for fast charging in the automotive space, 650 volt GaN automotive qualified.
The applicability of that for a 100 volt solution, it's a direct translation of what we already know how to do for 650, translated into the 100 volt there. Voltage handling capability, the hard-switching ruggedness, those are all meeting the specifications. You see the key conversion efficiencies, 97% to 92% for the core VR, switching frequency on there. The step from 6 volts down to delivery can be done either with gallium nitride or BCD, and here I'm highlighting the key piece from our BCD portfolio. I mentioned earlier already operating at a 150 volt or a 150 degrees Celsius for automotive. These things operate around 125 to 135 degrees Celsius. We know how to do that. Our BCD platforms are qualified for it.
Being able to achieve better than 5 amps per square millimeter current density is a key feature of the design of the technology. Of course, automotive grade 1, I highlighted the role of reliability on the previous page, right. These voltage regulators are literally mounted right below the XPUs in the tray. This is not a chip that you just pop off. Being able to have automotive-grade capability at better than 1 part per million is a key enabler for the reliability that you have to have in the data center. Both our GaN platform and our BCD platforms offer the 3D heterogeneous integration capabilities and through-silicon vias that are necessary to have the implementation of them being very close coupled for power delivery directly to the GPU device in the tray.
From a market outlook, automotive, we do believe the electrification trend will continue. Gas prices certainly put some wind in those sails today. There's gonna be a trend towards higher charging efficiency and longer range. Automotive's gonna continue to be an important market. We see the data center every 2 years, there is a new form factor, right? If you looked at the architecture for the Vera Rubin, they're actually in the same rack, but they're actually standing them up on their edge, right? A complete every 2 years, there are significant changes in architecture on there that demand improvements in current handling capability, in reliability, and switching frequency on there.
We believe that this is going to be an incredibly attractive market, and it's a market in which GF has significant momentum, and we are targeting an incremental $1 billion of revenue in our power business by 2030. The final part of the data center, and it might be a reach to talk about quantum in the data center today, and your first question is probably, isn't it a little bit early to be talking about quantum? The answer is yes and no. The yes portion is today there are no at-scale utility quantum computing capabilities out there. We will look back at 2025 and 2026 as the inflection point where many quantum companies have established handfuls of qubits that demonstrate the viability of their quantum qubits. It moves from being a physics and research problem to a engineering and scaling challenge.
They're not gonna be coming to GlobalFoundries because, hey, we need a bunch of capacity to ramp up our massive data centers for quantum computing. The way that you get from handfuls of qubits today to 1,000, 100,000, and then 1 million, which is what people believe you have to have, around 1 million physical qubits to have utility-scale quantum computing with fault tolerance capability in there. The only way that happens is by leveraging the advanced semiconductor manufacturing technologies that go along with it. We're not strangers to the quantum world, of course. A longstanding, nearly a decade-long relationship with PsiQuantum doing photonic quantum computing. Have multiple partners in spin modality as well. Direct Quantum Motion is startups with which we work. We know what it takes to work with the quantum companies on there.
What has been a bit of a surprise starting last year and quite frankly accelerated this year is the number of companies that are now looking to scale from handfuls of qubits to be able to get to 1 million qubit, physical qubits in the 2030 timeframe. Here we saw a significant opportunity for us to establish some platforms. I won't go into all the details. There'll be a webinar at some point in your near-term future that we're gonna go into on quantum computing. We see 2 modality-agnostic platform opportunities. First and foremost, the Cryo-CMOS using our fully depleted. Won't go into the details of why that is a key technology. We see the same thing for 3D interconnect. Now, today, when you have an interposer, it runs at room temperature and elevated temperature in a data center.
In this case, they want to operate at millikelvin temperature and use superconducting materials like aluminum or niobium. The quantum modalities. No one knows which q- modalities are going to win. The folks that are leading today may not have scalability to 1 million qubits on there. Our approach is to not work on every modality, but the ones that we think have a high probability of success and work with the leading companies in each one of those modalities. That work is deeply in flight. A final comment perhaps on the business model. This is an NRE business for the foreseeable future, but don't write off the economics of this business. By the time that you are creating 1 million qubit module, you think there's value in a co-package optics module, a SCALE module.
Wait until you're putting a 1 million qubit module into a quantum computer. We do see, after the 2030 period of time, a significant commercial opportunity for it. I want to now transition away from the data center to the physical AI space. It's only started getting news in the last few years on there. The fact that the physical AI trend is starting to take off is perfect timing for GLOBALFOUNDRIES. It occurs at the same time we're redefining ourselves as a technology provider. Let me break that down. Today, data centers garner all of the headlines and all of the attention on there. We believe physical AI will transform our lives and societies far beyond the large language models that we're all interacting with today. It's the tens of billions of devices that you heard Tim talk about.
Think autonomous robots, think self-driving cars, AMRs, right? Autonomous mobile robots. Our factories are full of them moving our materials around through our facility on there. We see massive opportunities for this. Through the MIPS integration with the integration of ARC, we redefine GF from a semiconductor process technology company to a broad technology provider. We believe that this will create significant revenue growth opportunities for our corporation. Of course, MIPS is the first at-scale RISC-V company with a large suite of cores. With the integration of ARC, more than 300 customers and more than 40 IP and AI cores. Very significant capabilities there. The mission extends beyond the traditional MIPS business into custom silicon capabilities.
The combination, in physical AI, the combination of the MIPS real-time deterministic cores combined with our ultra-low-power CMOS technologies, FinFET and FDX technologies, creates significant synergy opportunities for us. By combining the MIPS engineers with deep knowledge of end market applications, physical AI figures of merit with semiconductor technologists under one roof, you have the virtuous cycle of knowledge at the system level, driving the figures of merit to semiconductor technologists who can optimize the silicon technology to produce better products for our customers. Let me give you a real-world example of that. MIPS has a, an active program, and it is for a motor actuator MCU for a humanoid robotic arm. It's a RISC-V core operating at more than 800 megahertz and has a gigabit MAC in it.
You go, "What on earth do you need a gigabit MAC for in a humanoid robotic arm?" You saw Tim earlier talk about sense, think, act, communicate. Pressure sensors, temperature sensors, current sensors, high resolution positioning and high resolution motion control. All of these real-time deterministic going through this chip. Of course, it's a humanoid robot, so it has a very high standard of its level of performance. As a consequence of the RISC-V core and the 22FDX platform being co-optimized for this particular situation, has to be very low power, has to have a small form factor on it. It has produced a competitive position that there's no other solution in the industry that can come close to touching it. We see significant synergy opportunities as a consequence of having that all under one roof.
By starting with the IP licensing and royalties, that has been the traditional MIPS business, accelerated by the software ecosystem that goes around that core developing products and into custom silicon, we see an opportunity to achieve a $1 billion run rate for this business by 2030. If you've been keeping track, I've now mentioned multi-billion dollar opportunities several time, and you can well imagine that Mike is gonna come connect that to the end markets in which this plays, and then Sam's gonna show how it fits into the financial model. Before I conclude my talk, I wanna take 1 minute to go beyond the current horizon and talk about our strategic focus in how we work with startups. Historically, we have done a lot of research and development with the research institutes that you see, you know, on the upper left-hand side here.
They're a wonderful source of technologies, and in some cases, they spin out startup companies. In the last few years, we have begun working with a number of venture funds, and in some cases, participating in those funds as limited partners. We get access to four things as a consequence of doing that. First and foremost, and you see that it covers all of the localities in which, you know, Singapore, Europe and U.S., we see the deal flow, right? You look at dozens of startups coming through the deal flow of each of these venture funds. You get a really good insights into what startups are coming, which ones are funded, how they're approaching it, what system architectures they're pursuing. Tremendous insights for us. That's number 1. Number 2, it's a source of customers for us.
Many of these companies are chip companies that need access to technologies. We have this incredible suite of technologies that we can offer them, and they become new customers for us. Number 3 is now some of these companies are operating in physical sciences, and they have demonstrated something at lab scale that need an industrialization partner, and it is a source of new technology capabilities that we bring into the corporation. Of course, the 4th one is either an investment vehicle opportunity for us or an M&A opportunity. The so what associated with that is we are plugged into a global network that has sources of both technology and commercial opportunities for us that are directly related to the growth vectors. You see the areas of interest for us, not surprisingly, the things I have said already.
It's in areas of advanced power, all forms of optical communications, optical circuit switching, neuromorphic computing, right? Spiking neural networks that we know are gonna be critical enablers for physical AI, because of the bursting nature of a lot of the edge AI capabilities. As you've heard, I have a little bit of passion around quantum computing these days. I wanna close by telling a story about why I joined GlobalFoundries 17 years ago. It actually goes back further than that to the early 2000s. I was leading Motorola's advanced CMOS R&D line, and we had a multi-year joint development program with Advanced Micro Devices.
Put you back in time, Motorola was shipping Power PCs to a company in Cupertino, you know, the early 2000s was the x86 wars, David and Goliath battle between AMD and Intel, of course. I came to admire the AMD team. We would innovate on things that are now old, but advanced nickel silicide capability, low K dielectrics. I came to admire them because they knew how to take those innovations, put them into products, and go compete with Intel seemingly overnight. Their ability to compete with better products because of differentiated process technologies. Fast-forward a few years, 2008 and 2009, started talking to AMD about joining this crazy vision of creating a foundry by spinning technology and manufacturing out of an IDM.
I said, "You know what? It's a crazy mission, but I know they know how to differentiate technologies. That is as good a starting point for competing in the industry as you could possibly happen. I look back now 17 years later. Of course, I joined the company, just a spoiler alert there. We have put dozens and dozens of competitive technologies into the differentiated process technology box. Also the transformation we've done, the integration of the MIPS team, advanced packaging services, and expansion into custom silicon technologies. I'm more excited about the outlook today than I was 17 years ago when I joined, and we're just getting started.
Good morning. It's a pleasure to speak with you today about Marvell's long-standing and valued partnership with GlobalFoundries. At Marvell, our strategy is focused on enabling accelerated infrastructure for the AI era. The scale required for these computing platforms is truly massive, and it starts with building the highest performance compute possible. Our approach to compute is built around custom silicon. We're developing custom compute engines for the world's largest hyperscalers. Marvell built its custom business through the acquisition of the custom ASIC group from GlobalFoundries. That team spent decades working together to build custom ASICs, and today they're working on some of the world's most advanced custom compute platforms for the world's largest data centers. That history reflects how foundational the relationship between Marvell and GF has been.
At its foundation, data center scale requires high-speed networking, where more specialty technologies such as silicon germanium and silicon photonics are essential to the connectivity stack. Marvell has a leading portfolio of products based on these technologies, addressing every layer of networking, including scale up, scale out, and scale across. GF is an important semiconductor manufacturing partner for Marvell, supporting key elements of our connectivity portfolio. In today's environment, geographic diversity and operational resilience are increasingly important to our customers. GF's global manufacturing footprint provides valuable flexibility and continuity as AI infrastructure expands globally. We're proud to partner with GF. We look forward to continuing to build on this collaboration in the years ahead. Thank you.
Okay. Good morning. Greg said I didn't need any introduction. I won't make one. I will say that I have been at GF, this is my seventh year, and with all due respect to Greg having the best job at GF, I think I actually have the best job at GF. I've done really nothing in those seven years but look at the end markets we serve and then lead the product teams that develop the products that try to meet those needs. I'm gonna do couple simple things with my section.
I'm gonna try to synthesize the mega trends that Tim went through with some of the technology that Greg talked about to help you understand where we focus, how we win, why we win. I'm gonna try to do it at a level that is more concrete than abstract. Try to make sure you walk away with a real understanding of how that all works. Right place to start with this presentation is where Tim left off, which is, AI is the new story in our markets, right? AI is having a either a direct or indirect effect on every market we serve.
Whether that's, you know, the AI boom specifically in the data center or the way it sort of ripples across and has impact across the other markets, a lot of profound change. I'm gonna try to provide four things. Context for the end markets, a little bit of history, a wider, a widened out view of everywhere we have opportunity, and then deep dive on a specific application so that you can understand sort of how the GF technology is matched to a revenue growth expectations for each of the end markets. Think of it as four chapters of the same story, and really building off of what Tim talked about, the sense, think, act, communicate.
What you'll see is that this is all pulling, these markets are all pulling towards the kind of technologies that GF focuses on: power efficiency, connected, RF, and analog. Okay. Before I do that, I'll cut to the chase. Our markets are growing very fast. Our servable opportunity is growing in a very significant way by 2030, up to $120 billion, as you'd expect. What's probably more interesting is on the, on the right side of the chart, you know, the four end markets we serve wind up to be pretty nicely distributed and ultimately large, and most of them growing quite fast. I'll take it from top to bottom. Not a real surprise. Data center, AI economics, big opportunities in the number of optical electrical connections that are out there.
Big opportunity, as Greg said, of not only getting power from the grid all the way down to the core or inside of these XBUs delivering thousands of amps to keep these things running, but also, and, you know, probably a buried lead, if you will, in the comms infrastructure side, I think we almost look past the fact that we're actually creating an entire new class of communication infrastructure. You know, if AI data center wasn't collecting all the ink, I think you'd be talking a lot more about satellite-based internet. I mean, that would have been the biggest story, but it's been overwhelmed a bit, but driving a ton of growth in CI&D. In auto, it's really just a continuation of what we've been seeing. It's this ICE to ACE transition.
You know, the internal combustion engine of the past becoming sort of an autonomous, connected, ever more electrified vehicle. Sometimes I don't think we appreciate it. We're in the early innings of a multi-decade transition. This is not losing steam. This is gonna go on for many, many years. In IoT, probably the best of all worlds. It's a market that has gotten bigger than anybody probably would have thought, and it's not done growing. It's not done growing for two really significant reasons. It's barely reaching some of the markets that it's serving. Yes, a lot of the consumer stuff has already happened, but it's really early days in industrial. It's super early days in medical. You've got more expansion of IoT happening, and then you've got two really great technology migrations happening.
The one that's been talked about for a couple of years is more AI moving to the edge. There's this notion of physical AI driving through the IoT market and things like Ultimately robotics, whether that shows up as industrial or consumer. Finally, maybe a little bit ironically, smart mobile devices, you know, actually becomes the smallest of our markets. It's really just the reality that the smartphone market, you know, peaked growth in 2019. While it continues to innovate and will continue to have opportunity because of the way AI impinges on what that device is intended to do, you know, it's probably not gonna grow a lot.
Cutting to the chase, the punchline is we'll outgrow the market in CI&D, outgrow the market in auto, outgrow the market in IoT, and we'll probably look to stay relatively flat in SMD. Let me get at it. Let's start with automotive. When I started in semis in the mid-eighties, there really wasn't much of a relationship between Silicon Valley and any of the epicenters of the automotive world, whether that be Detroit or Stuttgart or Tokyo. You know, the car was really just a mechanical device, and maybe over the years, a few electronic conveniences hung off the edge. Today is completely different. Semi content is fundamental to delivering not only the automotive, the vehicle itself, but the user experience. Body electronics, ADAS, zonal control, electrification, infotainment is a completely different world.
A great example of where the car is sensing its environment, acting autonomously, communicating continuously within and without the car. Each innovation, you know, has been moving more opportunity towards GF. It's not just more content, it's more of what GF makes, and that's a very powerful theme that I wanna hit on a few times in this discussion. To ground things, let's talk about what the opportunities are for GF in the car. Upper right side of the chart is a great place to start. All of the sensors that the car is being enveloped with are intended to enable the levels of ADAS that everybody is looking forward to. Every level of ADAS requires exponentially more sensor information to perform the job.
That sensing information, that torrent of data has to hit a very different compute infrastructure. This compute infrastructure never existed in the car. Whether you're talking about a zonal controller, you know, compute defined by area of the car, or a domain controller compute defined by the thing that it does, it's a complete redo of the compute and communicate architecture within the vehicle. Of course, that drives a very different word problem in terms of how you power all that compute. This is not everything in the car running off of a 12-volt battery stuck somewhere in the hood.
This is distributing power to a very complex compute and communication system, for the first time, more often than not, off of 1 battery source, 1 battery source that has to be managed, maintained, recharged, for which the propulsion of the car is competing with the essence of making the car more autonomous. Ironically, that actually gets you to the lower right part of the chart where all of this autonomy and self-driving and all the conveniences of modern automotive life actually create the opportunity to do more things like infotainment. You need to be entertained while you're in the car because the car is doing most of the work, but it's a complete redo. To me, this is truly sense, think, act, communicate in motion.
I promised I wouldn't stay at an abstract level. I wanna kind of give you a concrete example. Automotive radar. The number of radars in a car will continue to grow through the end of the decade. Our FDX technology is hands down, full stop, the best solution for automotive radar. Just quick gut check on what is radar anyway? What is that application? In automotive radar, you transmit a millimeter wave, high frequency signal out from the front of the car, the side of the car, the back of the car, wherever you like. That signal then runs into things. It runs into the trees and the pedestrians and the road ahead, and it is reflected back. To win in automotive radar, you need a few things, and FDX has them all.
First of all, you need the best receive sensitivity. That reflected image, the quality at which you receive that image is the single limiting factor in how your ADAS will work. If you can't capture that image on the inbound, you can never recover it. Super important, FDX is 75% more sensitive than the same solution done with bulk CMOS. Its partner in crime is the transmit side. On the transmit side, you need to transmit a very high fidelity signal at the highest possible output power. You can't have a signal that's integrity is diminished because you're trying to get output power, you can't lower the output power to try to make a better signal. You have to have both, that's what FDX delivers because that's what gives you range.
The higher power the signal, the further out the signal goes, so you can get those reflections. Then finally, to get resolution, these radars have to sit in a very specific form factor at a very specific power budget. The number of transmit and receive channels that you can fit in that budget define what kind of resolution you get. By resolution, I mean be able to do the discernment of if the object in front of you, is it a tree? Is it a traffic cone? Is it a child on a bike? Those 3 use cases really matter in terms of trying to make a decision. The quality of the radar is the only chance the ADAS system has to make the best decision with the most time. More range, more information, more granularity, better decision.
It's a pretty important application from a safety standpoint. Having said that means that our FDX platform has to be 150 degrees Celsius capable and meet all of the most stringent safety standards like ASIL D. There's a real-world specific example. I think that our experience in automotive is worth talking about. $100 million at the beginning of the decade, $1.4 billion at the midpoint of the decade. Growth 14x in five years with literally 0 help from number of vehicles, 100% content-driven. You know, short order, we expect to continue the growth. We're still early on in ICE to ACE. We think it's a low double-digit grower for us, and it's really based on simple math.
We're gonna triple the number of sensors to support all the autonomy. Even though we've been growing real fast, to date, we don't see any slowdown in the momentum. We actually had 1.5 times the design wins in automotive last year than we did the year before, so we think it's actually accelerating. Automotive is a tough market to break into. You have to have credibility. We've been in automotive for as long as we've been a company, and I think that's evidenced by, you know, we're the number 1 supplier to the number 1 automotive MCU company out there. We're super bullish on our prospects in automotive. Now moving to IoT. Some perspective on IoT. Amongst many other things I've done in my career, I worked at Broadcom.
15 years ago, I was in a super serious strategy meeting. You have to remember in 2011, this was in the peak smartphone craze, right? Smartphone in 2011 was a lot like AI data center in 2026. It was exploding. You know, people couldn't fathom that something could have ever been bigger than the PC or the feature phone or God forbid, actually have a growth in content in a phone. Remember at that time, feature phones were diving in terms of the content and the cost and the price. Smartphones went exactly the other way. I had the temerity to say, "Well, that's really nice, but, you know, we better get on with investing in IoT because obviously IoT is gonna be way bigger than smartphone." It was sort of like this room, crickets.
You know, it wasn't very popular at that point to talk about that. Honestly, what's the scoreboard say? I said, smartphones topped out in 2019. IoT, 10 billion installed base, 25 billion installed base, 40 billion installed base by the end of the decade. The new markets are still driving it. The new technology transformations, you got an installed base to upgrade now with AI and eventually physical AI. It's a really good story for somebody like us who focuses on low power, analog sensing, wireless connectivity, and let me show you why that is. I think this slide really captures the same idea, the sense, think, act, communicate, but in an IoT context. On the left side of the chart, let me back up.
Three things that make you an IoT device because if you don't have each of these three things, you're really not IoT. There's probably one fourth one coming. The first thing is you have to sense something. The left side of this chart is the sense side. You're hearing a wake word, you're capturing an image, you're sensing a change in temperature, vibration, humidity. I don't care what it is. Something in the analog world around the IoT device is doing something. Second thing to make you an IoT device is you have to be autonomous. You have to know what to do when you get that input. If the camera captures something, you have to know whether to wake up because you have to be smart enough to know if that was a thief or just the neighbor's dog running by, right?
The autonomy is the second, you know, have to have in IoT. The third is you have to communicate. The I still stands for internet. You have a plethora of wireless connectivity standards, whether you're connecting to the internet via cellular or Wi-Fi, or connecting to other devices via Zigbee. Whether you're connecting to the phone that you have via Bluetooth for provisioning, NFC for authentication. There's a ton of wireless connectivity if you are an IoT device. Finally, the fourth dimension that I think will come out is physicality. I think this move to robotics will build on the first three of these, and you'll see that in industrial and other humanoid applications as we go forward.
All of these require a massive power efficiency, massive connectivity, and what I'll talk about in the next page is embedded non-volatile memory. Why did I choose embedded non-volatile memory as the deep dive here? It's really because that's the heart of autonomy. I don't think people appreciate. We talk about eNVM as sort of like, "Well, you know, I don't even know why that matters." It matters because this is actually at the core of autonomy. What's important for autonomy? Three things. Security, latency, power. eNVM, break it up. It's embedded, and it's non-volatile. Why does it matter? Well, from a security standpoint, the code never leaves the device, right?
There's no external interface upon which the information, the password information, the who you are is inspectable by somebody with bad intent. It's what allows the device to be upgradable. To do an over-the-air upgrade, if you were to have a security problem, you need to have the non-volatile memory on the device so that you can upgrade it. You know that when you've upgraded it, you know, can store the old code, download the new code, know that it's good before you know, you reset the device. Finally, authenticatable, right? The other aspect of security is I know who that device is. How do I know that's the real thing? Do I know who the person is? Have I bound the person's identity to the device? Think about how you activate a smartphone, right?
You have your IMEI number. You have, you know, you the carrier authenticates you and the device to bind it. That's what embedded non-volatile memory creates on a security front. From a latency front, you don't have time to be woken up. You don't have time to download from DRAM into the device to wake up to hear that 1 of you came home and said, you know, Alexa, please play a song for me. Right? The latency is incredibly important and finally, power in every case. Memory interfaces cost pins, they cost power. Volatile memory requires power at all times to retain state. SRAM, just to stay, to know what it's doing, it has to be powered up. DRAM actually needs to be refreshed to keep its state.
None of this fits in with the duty cycle of an IoT device. IoT devices sleep most of the time. They have to wake up very fast, be consuming very low power, or else you're not going to be able to run them off battery. Very important, but at the same time, all eNVM is not created equal. There are tons of different attributes. The read and write speeds. Can I execute out of that memory, or do I have to have a buffer? The endurance. How many times can I rewrite the memory? The PPM, the number of bit failures I can tolerate. The bit size, the actual size of the technology. The error correction I have to wrap around it to make it work. The cost, whether that's the die size or the number of mask layers.
Or the mission profile. These go into different things. They have different temperature sensitivities. They have different immunity sensitivities. Bottom line for GF is I think we have the broadest eNVM portfolio, MRAM, RRAM, flash, across the widest array of technologies, FDX, FinFET. Even our power technologies have this. 40-nanometer, 28-nanometer, and that is the key to autonomy in an IoT device. Our expectations for IoT, very bullish. Whether it's because of these emerging markets, like think of the number of people with diabetes, think of the innovation now where you can have continuous glucose monitors available sort of everywhere. It's going to be like a 2.5x growth from where we are today by the end of the decade. I think Tim might have mentioned, you know, we had about 500 design wins last year.
More than 200 of them are in IoT. We've established it that we know that the design wins lead to the revenue. As the sort of physical AI gets going, you know, It'll probably take place in the industrial world. We're engaged with 7 of the top 8 IDMs. Mid-teens growth through the end of the decade. Smart mobile. Another really interesting one when I was thinking about how to describe this. I was sort of shocked when I looked back. It's been almost 20 years since the inception of the iPhone. The iPhone was announced in June of 2007. At the time, you know, there was a completely different list of players we were talking about.
Nokia was the king of the feature phone, and there's literally people we don't even think about much anymore. You know, BlackBerry, Siemens, HTC, they were all sort of taken by surprise by this big innovation. In fact, the famous CEO of Microsoft said, and I got this quoted here, "There's no chance that the iPhone is going to get any significant market share." He doubled down and said, "No chance" again. If you look at the scoreboard, there's only been 3 billion shipped. It's 15%+ of all the smartphones ever shipped. More than 40% of the volume, probably more than that of the pro-profits. For us, the most important thing is it wasn't just a new device. It actually was a completely new ecosystem. Never had such a high-volume device been so taxing to technology.
We got a ton of our DNA out of being successful in the smartphone business. Our RF, our analog, our sensing, our display. It's been a very powerful driver of R&D. While the market has probably matured and slowed, it's still got interesting dynamics within it, and it still represents an important market for GF in terms of how what we do there echoes through the rest of the end markets we serve. Again, just to set context, you know, exactly where you'd expect us to be successful. Same, same ideas. Wireless connectivity, the complex RF front end, a lot of the sensing, the imaging, the display, the audio, the haptics, the human machine interface. Very similar to what you saw in IoT.
Where did that IoT technology come from? It probably came somewhere from the smartphone, and all with a really complex power management scheme. Again, the significance of smartphone volume was it really tested. It was a point where the physics of what was possible met the scale of high volume on a really aggressive cadence. It's been very good for us. To diverge just briefly. The smartphone volumes, like I said, peaked in 19, but there's still some interesting activity going on in different form factors. Whether it's glasses, wearables, or hearables, important for a couple reasons. One, these could become very large markets. You'll see with in how we model it, we're not really expecting that between now and 2030, but they could become very large.
The more important thing is they almost always start by taking technology out of the smartphone to put into the glasses, to the wearable, to the hearable. There's a direct pipeline between those two things that we benefit from. Honestly, they actually drive an even more aggressive size and power requirement for us to meet. It's actually quite good for us. Back to the more specific examples and a great example of where you can have an end market that is not growing so fast, but something within the end market that grows very fast. A great example is wireless charging. Whereas smartphones aren't going to grow that much, the attach rate on wireless charging is growing quite nicely.
For GF, what do we use to get at wireless charging? Our BCD platform. What's BCD? BCD is bipolar, very high precision transistors, that's the B. CMOS, like CMOS logic, that's what gives it the brain, that's the C. LDMOS, which is the power transistor, which gives it the super high drive, right? BCD technology is a perfect fit for something like a wireless charger. It's the triple threat. You get all the power and analog control, the sensing to know sort of how are the batteries doing, and the eNVM to have it all in one device. The attributes are clear. Higher voltage, faster charging. Everybody likes faster charging, not slower charging. Lowest RDS(on), Greg mentioned. That means you have the lowest resistance, which you get the most efficient charge. You get the lowest thermals.
Finally, that all leads to the lowest system cost, best battery life cycle management. You know, people very commonly say, you know, "Hey, nobody's buying new smartphones because the features aren't compelling." Well, you know the reason you buy a new smartphone? It's 'cause the battery stops working right. Right? The way to stop the battery from going and getting fatigued is to have ever smarter charging capability, and that's what we deliver. By the way, BCD is not new to us. We're on our third generation. We have it from 5 volts up to 150, up to 175 degrees C for automotive. It's a technology that's very familiar. Cut to the chase on smart mobile. We're not counting on a lot of growth, but we're excited because there's, you know, probably 10x growth in the alternative form factors.
It'll come from the smartphone guys we're already all engaged with, you know, everything in here is wireless. You know, we are a wireless company at our heart, and we're deeply engaged with, you know, practically all of the top RF suppliers. Finally, final market, comms infrastructure and data center. It's almost you got to remember back to the time when Greg talked about when we came out of AMD. You know, a big chunk of our business was, you know, data center and PC. Most of that was legacy digital nodes that have since moved on to different nodes at different foundries. It was quite a headwind in our business for a while. The turnaround's been nothing short of a renaissance.
Now, you know, think about the things that Tim and Greg talked about. Silicon photonics for optical networking, silicon germanium for wired connectivity, BCD and GaN for power delivery. It's been a real turnaround, a real reborn market for GF. Greg and Tim talked about the two things that everybody talks about now, which is the photonics and the power delivery. I'll spend a little time on a more humble thing that people mostly underappreciate, the transimpedance amplifier. Now, that's not something you talk about at cocktail parties very often, I bet, but it's actually quite important. It's a, if you think about the application of a transimpedance amplifier, all those photons we keep talking about going over the light, eventually they have to hit a photodiode.
The photodiode then creates a very high frequency, very low current that is the representation electrically of the photons. The transimpedance amplifier then takes that very low current and amplifies it to a voltage. The technology that does that is our silicon germanium. Silicon germanium is not a new thing to GF. We actually can be the proud owners of that by virtue of our IBM acquisition. SiGe has been in our DNA for 40 years. All of those applications you hear about, like 200 gig per lane, 400 gig per lambda, all require you to not only convert the electrical to the optical, gets a lot of attention, but the optical back to the electrical. The best-in-class fT/fmax and GF has that.
You know, it's, it's sort of funny, you know, that RF is only wireless. It's true. All wireless applications are RF, but not all RF applications are wireless. There's a ton of RF going on in the data center, very good for GF. That was the D part of CI&D. Let me talk about CNI. Low Earth orbit satellite-based internet. What's the problem statement? You know how everybody's annoyed when their cell phone doesn't seem to get any signal? That's a tower that's, like, miles away. In this topology, the signal's coming from outer space. Imagine how small it is by the time it reaches Earth. The way you recover that signal is you have to have a very precision low noise amplifier.
In fact, you have to have an array of low noise amplifiers. Great example. We took the LNA technology that we had in the phone, we optimized it for this particular application, and it helps to generate an enormous market. What's the simple logic? You need the best LNA. The better your LNA performs, the less LNAs you need. The less LNAs you need, the smaller the box can be. The smaller the box can be, the cheaper the box can be, the faster you can grow subscribers. With two and a half billion subscribers out there underserved, it's pretty important for the growth of that market. Just for context setting, there's about 5 times the RF front-end content in every one of those ground terminals than in an average smartphone. Pretty cool.
Suffice it to say that leaves us with high expectations in the market, whether it's silicon photonics in our optical world, BCD and GaN in the power world, SiGe in the TIA and driver. This NSX technology for LNAs, you know, 3 times the satellite subscriber growth, 70% optical connections by the end of the network. We're form factor agnostic, but we're working with the most of the pluggable guys now as they move from pluggable to near package to co-packaged at whatever rate and pace. We say that's a 30+% grower for us, especially since the digital legacy or the legacy headwinds have abated. You know, don't believe me, believe our customers. I think the significance of the quotes on this chart, which I'm not going to read, are twofold.
It's interesting you have like Amazon LEO on there. Like I said, I think that would have been a bigger story if AI data center hadn't stolen all the headlines. More importantly, it's really an acknowledgement of GF, the technology provider, the guy growing up from, "Hey, you have to have great manufacturing services to get friendly quotes from customers." These quotes are all about the technology that we deliver. We are moving up towards a more complete supplier of technology. Final thought, yes, mega trends rushing through the 4 end markets we serve, exploding in a plethora of opportunity. You know, I've been doing this for 40 years, you know, I'm a rookie compared to Greg, I've been in every operating model. I was at an IDM.
I ran a couple startups that were fabless. I was at a mega-cap fabless guy, Broadcom. Been at Foundry now for seven years, and I've seen every one of these mega-trends, every one of these booms and echoes. What I would leave you with is just think about this possibility between AI and physical AI moving to the edge into IoT. Think about what we saw in auto. You had a 90 million unit market whose content exploded and drove 14x growth for us through the first half of the decade. Imagine what will happen if you have a multi-billion unit market like IoT with 40 billion units of installed base getting upgraded in a similar content way. I mean, it's just a massively different geometric progression of opportunity.
Just like Greg, I'm a little bit jealous of everybody who gets to start in semiconductors now because, you know, after 40 years of doing this, I actually think the cool part's just coming. You know, I just can't remember a time or company where I've been at where the markets, the technology, the geopolitical landscape have lined up any better. Very excited about the future. Thank you.
Decarbonization and digitalization are transforming industries at unprecedented speed. Artificial intelligence accelerates this development even further. We at Infineon see transformation as an opportunity. Together with our customers and partners, we turn innovation into tangible progress for society. Infineon and GlobalFoundries have worked together for more than a decade. Our partnership is built on the conviction that only innovation and consistent execution ensure sustainable success in the semiconductor industry. Together, we are delivering one of the most successful automotive microcontroller platforms into the market, manufactured across GlobalFoundries sites in Singapore and Germany. Our work with GlobalFoundries extends well beyond manufacturing capacity. We co-develop technologies. The processes we apply with GlobalFoundries are tailored to the requirements of automotive and security applications. This gives our products clear advantages in performance, ultimately benefiting our customers.
As vehicles become software-defined and intelligence shifts to the edge, innovation strength and speed are becoming even more important. Together, we are advancing next-generation automotive platforms, including our work on RISC-V based architectures. From a customer perspective, resilience matters just as much as innovation. GlobalFoundries' multi-site manufacturing footprint supports supply chain resilience and our ability to scale as we expect long-term growth in our core markets. Looking ahead, our partnership with GlobalFoundries will continue to evolve. We are constantly looking for opportunities to broaden the scope of our collaboration. With strong partners like GlobalFoundries, Infineon is well-positioned to capture the growth opportunities of our time.
Our final presentation of the day is from Sam Franklin, our CFO. Before we bring him on, we have yet another customer video to show you.
Good morning. It is a privilege to speak to you today about how NXP's unique hybrid manufacturing strategy enable us to navigate today's dynamic landscape and how our partnership with GlobalFoundries fits into our strategy. We are seeing sustained demand for secure and power-efficient SoCs across automotive, IoT, and mobile markets. At the same time, supply chain complexity and geopolitical uncertainty compels us to reassess the importance of manufacturing and sourcing. In this environment, foundry partnerships matter more than ever. GF is an important strategic foundry partner for NXP, and our relationship spans more than a decade. Together, we have delivered measurable business impact through consistent execution, innovation, and a shared commitment to customer success. NXP is leveraging GF's 22FDX platform to drive next-generation solutions across a range of end markets, including automotive and IoT.
This platform's power efficiency and performance enables our customers to build the next generation of connected and secure edge solutions. It delivers ultra-low power consumption, high performance, and system-level integration, bringing RF, analog, digital, non-volatile memory together on a single die. This platform fits well within NXP's ambition to be the leader at bringing intelligent systems to the edge. Additionally, GF's manufacturing presence in Germany and the U.S. helps support our goals of ensuring geographical supply resilience in our manufacturing base. GF's cross-qualified nodes spanning multiple sites align perfectly with NXP's hybrid manufacturing strategy, helping us mitigate risk and optimize cost. A key reason why we chose GF is their commitment to the highest automotive quality standards, ensuring NXP's products meet the rigorous demands of global vehicle manufacturers and deliver the reliability and safety our customers require. This partnership has also delivered commercial traction.
Together, we won business with major automotive and IoT customers, and we look forward to continued innovation and success. Ultimately, NXP's partnership with GF is foundational. It is about building the future. It is a key part of how NXP is delivering long-term value to our customers and shareholders. Thank you.
Good morning, thank you for joining us today as we articulate the next phase of GF's remarkable journey and the critical role that GF plays within the semiconductor ecosystem. We've covered a lot with you today. Tim set out the megatrends shaping our industry and the critical role GF plays within those megatrends. Mike ultimately set out how we think about the end markets and the investments we're making in those end markets to support our customers. Of course, Greg took us through the technology roadmap and how our differentiated solutions are serving those end markets. As GF CFO, my objectives of this section are to set out the financial context of our plan and lay out the framework for our projections, for how GF is positioned to capture growth and maximize value profitably and sustainably.
As many of you might know, I've been with GF for a little over eight years. First half of that in the capacity of the shareholder, the last four years as part of the management team. What gives me confidence in the plan that we've shared with you today is not just the strategic tailwinds that are driving our business and our industry, but the quality of the team, the team in the room with you here today, and the team that makes up our 13,000 employees around the world. As you've heard from Tim, Greg, and Mike, this is about what's changed structurally within our business and why we believe our financial model is stronger and more resilient than at any point in our history. Let me start with the core principles as we focus on the next five years of growth.
Everything that we do financially ladders up to three core principles. Firstly, diversified revenue and durable revenue. Reducing our exposure to single end market cyclicality. Driving meaningful margin expansion through mix, scale, the evolution of the services that we offer our customers. Ultimately, creating substantial shareholder value through disciplined capital allocation. You'll see these principles reinforced as I take you through this section. I'll start with our expectations for the opportunities across the end markets we serve and build upon where we're focused on driving our durable and diversified revenue growth. You heard a bit of this from Mike in his section as well, so I'll try and bring it all together.
We believe that we put together a set of growth expectations that are anchored across end markets with strong secular tailwinds, content expansion, and form factor evolution, all of which play into GF SAM today and going forward. This is a plan that's grounded in real customer momentum. You heard from the team today, we delivered over 500 design wins in 2025, 95% of those on a single-source basis. That momentum drives tapeouts, drives production revenue, gives us confidence in our plan.
I'll bring together what Mike set out in his individual components of his section, but I'll start with smart mobile devices, where the decisions that we've taken over the course of the last few years have created a model for us that means we're no longer reliant on the growth of smart mobile devices to drive towards the enterprise revenue CAGR that we believe is achievable. That's not to say that we don't see opportunities within smart mobile, and our design win momentum is consistently reinforcing the principle of form factor evolution. Those design wins take time to grow and to ramp, and so we think of those as more opportunities in the second half of our target model. Moving to automotive. As you know, this has been a real bright spot for GF. 14 times revenue growth over the course of the last 5 years.
There's two key reasons for that that Mike set out, which we believe continue going forward. The first, content expansion. We've seen a doubling of the content within the vehicle over the course of the last five years. The expectation is that grows 1.5x by the end of this decade. Content expansion matters. Notwithstanding some of the dynamics going on from a SAR perspective, I think you can see the proof in our numbers that we've out-bucked that trend. The second is about share gain, and one of the reasons we've shared with you today some of the customer testimonials is because it reinforces the partnerships that we have with those customers. Put all of that together, gives us confidence in our target to deliver low double-digit % growth. Next on to IoT. This is an exciting area for us.
We've talked at length today around the opportunities that come with the evolution of IoT applications, the growth of physical AI, smart connectivity, industrial automation, defense applications as well. Again, this isn't a plan which is based solely on future expectations. This is a plan which we see momentum today. We delivered 200 design wins in IoT in 2025. That was a 40% uplift on 2024. As we think about the growth over the next 5 years, we believe that a target growth rate of mid-teens is reasonable with that backdrop. Finally, Communications Infrastructure and Data Center. It's a critical engine of growth and profitability, and you're seeing that come through in our numbers today. This isn't just about data center, this is about satellite communications as well and the depth of the relationships that we're building with customers in that space.
We delivered 29% year-over-year revenue growth in Communications Infrastructure and Data Center in 2025. As you heard from our call earlier on this week, we expect to deliver high 30s% year-over-year growth in 2026. All said, we feel very comfortable with the target of 30+% growth that we've identified for this end market. We believe we put together an end market growth strategy and taken strategic decisions as a business that is underpinned by gains from a share perspective, by content growth across all of the end markets that we serve, and are foundational to all of the design wins that we've been winning over the last couple of years. You've heard from Greg and Mike where our technology platforms meet the market opportunity. The slide is really a convergence of the strategies being brought to life.
Multiple concurrent growth engines that serve all of the end markets. Greg took you through in detail the key accelerators we're driving the business towards, targeting a multi-billion-dollar revenue opportunity by 2030. I'll reinforce the points on silicon photonics. We've accelerated our targets there. Billion-dollar run rate by the exit of 2028. We're targeting $2 billion of revenue in 2030 as we target share gains in pluggables and the ramp in near and co-packaged optics. The demand for power is insatiable across all of the markets we serve. Greg laid out in significant detail today how important power applications and power delivery will be within a data center.
It's not just the data center, it's all of the end markets that we serve, and we believe that with those backdrops and those tailwinds, that we can target an incremental $1 billion of revenue from our power applications by 2030. Finally, custom silicon IP software. It's a big part of the thesis as to why we acquired MIPS at the back end of last year. It's a big part of the thesis as to why we're seeking to acquire the ARC IP business from Synopsys. We're scaling these technology services. These create an earlier inroad with customers. We can work with customers at a much earlier point in the design life cycle than we otherwise would be able to as a pure-play foundry.
We're also targeting custom silicon solutions as well, which take years to ramp, but again, gives us confidence in this target plan of delivering a billion-dollar run rate business by the time we get to 2030. All 3 of these businesses serve this broader objective, which I set out at the start of this section. Revenue diversification, margin fall-through, and operating leverage. We set out to build multiple engines of growth serving multiple end markets, and we believe that these decisions serve those objectives precisely. To that end, our M&A strategy continues to be highly targeted. We're creating new capabilities to support customer demands and targeting both data center AI and physical AI applications, as both Tim and Greg set out. The proof points for us are clear. We've closed 3 acquisitions in the space of the last year.
We have a 4th, which we expect to close in the next couple of months. AMF and InfiniLink are targeting those data center AI applications, while MIPS and ARC are targeting the opportunity that we see within physical AI. There's a 5th acquisition which is not on the page, 1 that we closed in 2024 called Tagore Technology, where we acquired their power business design capabilities, principally serving GaN power. You've heard about the importance of GaN power today in terms of the opportunities in the data center. Not only do these acquisitions strengthen our position within the physical AI and data center AI marketplace, but they are, in the case of both MIPS and AMF, driving revenue and margin contribution today, and we expect to scale that over the years ahead.
The strategic intent of these acquisitions deliberately target more capabilities, more customer engagements, and pulling more customer roadmaps into GF over the next 3 to 5 years. If you look at our revenue evolution from the decisions we've taken over the last 5 years, you can see that this becomes the foundation for how we think about the next 5 years. I'll look at some of those diversification proof points, we'll start with automotive. In 2020, automotive end market revenue was 3% of our end market revenue. By 2025, that was 23%. That's thanks to years of design innovation and design wins and momentum with customers. Over the same period, our revenue excluding smart mobile devices has expanded by 10 points as a percentage of manufacturing service revenue. This is a meaningful step forward towards the balanced portfolio that we're targeting.
More recently, as we've discussed, the very strong momentum that we're seeing in Communications Infrastructure and Data Center. That is the result of very deliberate remixing decisions that we took over the course of the last few years. It does not happen overnight. As I say, these proof points from the last five years establish what we believe to be the foundation for the next chapter. As those design wins drive the tape outs, drive the production ramps. By 2030, we're targeting revenue from manufacturing services across automotive, IoT, Communications Infrastructure, and Data Center to represent 75% of our manufacturing services revenue. That objective is to deliver a stronger, more diversified revenue mix, less exposed to singular cyclical dynamics. To put a finer point on it, we're deliberately prioritizing robust customer demand in high gross margin products with markets and applications supported by secular tailwinds.
These don't just accelerate revenue, they accelerate the mix shift that drives the margin expansion. You heard from us this week on our earnings call that we're making some changes to our revenue categorization, namely revenue from manufacturing services and revenue from technology services. With GF's shift towards becoming a holistic technology solutions partner to our customers, we believe that these categorizations better reflect the stacked value model that we're providing to our customers. Now, of course, manufacturing remains at GF's core, and that will continue to be the case both today and going forward. We see the opportunity to support that manufacturing revenue in two key areas. The first is manufacturing value add. As you heard from Greg, that's where we're growing our advanced packaging capabilities, driving opportunities in optical modules. We expect 2028 will be an important inflection point for when those opportunities start to ramp.
Secondly, technology services, which today principally comprises of revenue from masks, reticles, NRE. But as we integrate the MIPS business, the ARC business, as we scale those businesses, we expect to see an increasing contribution from IP licensing software. The great thing about these technology services is that they're expected to support gross margin expansion. They don't come with the same fixed cost intensity that you otherwise associate with manufacturing services. We believe it's the evolution of these two categories that fundamentally change GLOBALFOUNDRIES's revenue mix and earnings power. Let me now focus on our margin expansion objectives, what we've delivered, and what's ahead. Since our IPO in 2021, we've delivered structural margin improvement against the backdrop of relative cyclical headwinds across certain consumer-centric end markets that we serve. We delivered over 1,000 basis points of margin expansion in 4 years.
That expansion has followed three core principles: disciplined capacity investments, a maniacal focus on manufacturing productivity, and of course, as we discussed today, a consistent shift in the mix and the evolution of our mix. We've outperformed our peers on a relative margin expansion, and although we're acutely aware of the fact that we're still on the journey towards our medium and long-term margin objectives, this progress to date serves as the foundation for our margin objectives going forward. This isn't just a long-term expectation. The progress is visible today and as recently as our earnings that we announced earlier on this week. The performance is clear in 2 clear metrics. Revenue, up 3% year-over-year. Gross margin, up 510 basis points year-over-year.
This margin expansion is years in the making, and we believe the opportunity is still in the years ahead to continue driving that. Q1 is just a proof point. It shows that we're on the right track to delivering what we believe is possible with our margin expansion using the levers within our control. Over the last few years, we've reiterated our objective to be a dollar for dollar more profitable company through structural margin transformation. Look, we've got a long way to go, and implementing structural margin transformation doesn't happen overnight, but I'm pleased to report that we continue to make real progress towards achieving that. I've talked about Q1. What I didn't say about Q1 is that it's also the highest margin that we've delivered in a first quarter as a company.
We believe that as we go through the end of this year, we can exit 2026 with a 30% gross margin. More importantly, as we ramp critical technology corridors with a higher margin mix, as we integrate the acquisitions that we've made over the course of the last year, and as we expand our technology service offerings to our customers, we expect to achieve a 40% gross margin exiting 2028. Beyond 2028, that's where we view the next inflection in the business plan that we're projecting with you today. Co-packaged optics, advanced packaging, custom silicon, all targeted to ramp in that 2028 timeframe and drive a line of sight to a 45% gross margin over the longer-term horizon of our model.
The key mission we're seeking to achieve is to remove that reliance on single margin drivers and drive structural margin upside across mixed transformation, corridor optimization, technology services, and accretive M&A. That's why we have confidence in the trajectory. Of course, expanding our capabilities across manufacturing and technology services brings a need to efficiently scale our R&D investments as well. These R&D investments will go into driving differentiation across capability tooling, IP, custom silicon, advanced packaging, optical networking. Critically, we intend to scale this R&D where it differentiates us while maintaining SG&A discipline and AI productivity. As a result, we expect our OpEx as a % of revenue to be roughly in line with our historic trends, and over the duration of our model, we expect our OpEx to grow at roughly half the rate of the revenue CAGR targets that we're sharing with you today.
As we exit 2028, we believe that that will drive approximately 15 points of margin delta to gross margin, delivering approximately 25% operating margin. However, beyond that, as the R&D investments we're making begin to scale and drive the flywheel of incremental revenue, we expect our operating leverage to compound and support an operating margin of approximately 35% longer term. Moving to the culmination of our core financial mission, which is to raise the bar on shareholder value creation in the years ahead. Margins matter because they fuel cash, and cash enables choice. Our CapEx discipline continues to be a key area of focus for GF. In the last 5 years, our CapEx intensity has been approximately 22% of revenue. That's marginally higher than the prior model we shared. However, we've generated meaningful free cash flow during that timeframe.
The capital intensity that comes with being a leading foundry means that we have to plan and execute with discipline. Starting with the left side of this chart, you can see that our capital intensity in the early years was principally driven by the investments into capacity scale at our 7H facility in Singapore. That delivered approximately 400,000 wafers of incremental capacity, and we're scaling that business today across all of the technology services that we offer. In the years that followed, the focus was on capacity within the four walls. It was focused on demand-led customer signals. As we discussed at the beginning of this year, we reiterated earlier on this week, we do expect that 2026 is gonna represent an investment year for GF into capacity.
We're targeting about 15%-20% as CapEx as a percentage of revenue this year, but it's going into those fast growth, high demand corridors, SiGe, silicon photonics, FDX. Against this backdrop, we continue to believe that a long-term target of approximately 20% CapEx intensity through the cycle provides the right balance of CapEx discipline and demand-led growth. Moving from left to right, you can see how we've evolved from a little over 2 million wafers of capacity in 2020 to approximately 2.7 million wafers exiting 2025. These capacity additions have followed 4 core principles. The first is that we expand our capacity intelligently, focusing on existing clean room space first, and then ultimately over time, looking to modular expansion. We've got great examples of that happening right now in our Singapore and our Dresden facilities.
The second is that we invest in technology fungibility to reduce technology concentration. A great example of that, Malta, N.Y., originally built as a FinFET fab, where we've spent the last few years transferring technologies in to support our silicon photonics business, our 40 nanometer auto grade, our 22FDX platform. With that, we target delivering manufacturing scale and efficiencies to improve utilization and optionality. We're a high fixed cost business, so scale and optionality matter. Finally, we look to de-risk those investments with customer commitments, prepayments, and government support frameworks as well. To underscore this strategy, we intend to add capacity to meet customer demand in margin accretive corridors with meaningful stakeholder funding contributions.
The partnerships and government support frameworks that we have in the U.S., Germany, and Singapore are all great examples of where they're contributing meaningfully to both our CapEx and our R&D investments, where we expect to recover anywhere between 30% and 50% of eligible spending. Even though 2025 was a relatively low CapEx intensity year, we still recovered 20% of our CapEx in the year under these frameworks. We've invested in these government relationships for many years. We're proud to be such an integral part of the semiconductor ecosystem across all of the geographies that we operate in. In the last 12 months alone, we've successfully closed new partnerships with the Economic Development Board of Singapore, the German government, and of course, the U.S. CHIPS Office, all to support customer demand, which aligns with national onshoring priorities.
We expect more to come in the years ahead. What makes GF unique across all of our manufacturing sites is that it's not just one of our sites where we benefit from these frameworks. It's all of our sites. It lowers the capital intensity of our capacity objectives and supports our margin targets. It's the combination of government support frameworks, customer funding, customer demand signals that become the backbone of investing in our capacity and drives returns without compromising flexibility. To bring it all together, this really captures the culmination of both the qualitative and the quantitative metrics we've outlined as part of our strategy today. The track record since our IPO is clear. Notwithstanding relatively muted revenue growth over that period, we've expanded our margins, gross margins by 10 points, operating margins by 12 points. That's had a commensurate fall through to expansion of EPS.
The how we achieve that becomes the basis for the next phase of our growth, continuing to expand and remix our capacity, growing our technology services business, investing in critical growth corridors all the while to support customer demand. The right-hand side of this page sets out our target model in two distinct phases, exiting 2028 and longer term. As you've heard today, our business is evolving and our services are expanding, and we're targeting secular growth across the end markets we deliver. We believe that leads to approximately 10%-12% of revenue CAGR opportunity over the duration of our model.
The quality of the mix within this revenue and the structural margin drivers that we're continuing to implement all drive us towards the expected exit of 2028 with a 40% gross margin and targeting a 45% gross margin over the long term. As I've discussed, the resulting operating margin is expected to be in the neighborhood of 25% exiting 2028. As we get the flywheel of those R&D investments driving incremental demand, creating operating leverage, we believe that we can target a 35% long-term operating margin. All of that drives towards approximately $4 of EPS exiting 2028 and $6 of longer-term earnings power. All the while, we intend to do that with a disciplined CapEx framework, driving to approximately 20% of revenue to target approximately 10% free cash margin.
This strategy is centered on disciplined expansion, growth markets, margin-accretive corridors, and consistent free cash flow generation. To that end, free cash continues to be a key area of focus for us. We've delivered over $2.5 billion of free cash in the last 3 years. During that timeframe, we've also strengthened our balance sheet with approximately $5 billion of available liquidity. We've reduced our gross leverage to less than 1 times. We've returned $400 million to shareholders year to date 2026. We funded our capacity objectives in parallel, and we've invested in acquisitions that support our future growth and margin objectives. This has given us the strategic freedom to invest, acquire, and return capital, which I'll turn to now. Our capital allocation principles are simple and disciplined.
We reinvest in profitable growth vectors for our business, as you've heard from the team today. We pursue differentiated and value-accretive M&A that we believe will accelerate our objectives. Finally, we intend to return excess cash to shareholders through coordinated buyback and dividend frameworks. Today marks a milestone for GF. I'm very pleased to announce that GF's board of directors has approved the initiation of a quarterly cash dividend of $0.12 per share, payable on July 14th to our stockholders of record as of June 24th. Today has been about setting the opportunities that we believe position us well for growth while enhancing shareholder returns. To drive this objective, we're formalizing our capital allocation framework of returning up to 50% of free cash flow after investments to shareholders through a combination of buybacks and dividends.
As you've heard today, we believe we're at an inflection point in our business. We're maniacally focused on executing our strategy and supporting our customers where they need us. Tim set out his vision and the critical role that GF supports in the mega-trend shaping our world. How the AI proliferation into the physical world is only just getting started, how GF's geographical footprint matters within all of that. Greg took you through the how we get there, investing in our technology roadmap, optical networking, advanced packaging, power delivery. Mike set out how we win with our customers, how content growth across the end markets we serve play directly into our technology portfolio. Finally, I've set out what we believe this means for our revenue growth trajectory and earnings potential for GF over the next five years.
We believe we're at the most important point in our company's history with the strongest differentiation in both our manufacturing and technology services portfolio, clear margin expansion drivers, a line of sight to strong profitability, and a disciplined framework for delivering shareholder returns. We'd really like to thank you for your time today. We sincerely appreciate your support and your continued interest in GF. Thank you. I'll now invite Tim, Mike, Greg up onto the stage, and I'll turn it over to Eric for Q&A.
Thanks, Sam. Thank you for joining us today. I'm very excited to start a Q&A session. Just a couple of general reminders and suggestions. I will ask that you kindly limit your question just to one for our first round to ensure we get through everyone and allow them to have an opportunity to ask a question. We have two team members here, kindly raise your hand. I'll call on you, and we will get to you. Finally, I'm just excited to kick this off. Can we start with Jim here in the front?
Thanks for this. James Schneider, Goldman Sachs. Thanks for the presentation today. I was wondering if you'd maybe address the geographical footprint of capacity going forward. Clearly, there are a lot of specialty processes, which is your core, and some of them are really outgrowing, like silicon photonics, et cetera. Can you maybe kind of talk to sort of the in-wall and modular capacity plans, where geographically you expect them to land? Is that Singapore, Dresden, or Malta? Maybe just kind of talk about any color you can provide on where you'd expect geographically to add the most. Thank you.
Yeah. Thanks. Maybe I'll start with kind of, again, recap on how do we think about this. We seek to have as many of our platforms available in as many fabs as we can because we know that that creates that optionality for customers, and we can't predict how, you know, preferences will change in future. The more we can cross-qualify, the better because it also allows us flexibility. As demand changes, you wanna be able to adapt to a market that picks up in one area, but maybe is weaker in another. First principle is do as much as you can more, more broadly.
We're doing modular expansion already today in Germany. We're filling out the clean room space in Singapore. I'd say when you think about the bulk of the investments, the U.S. is where the demand supply balance is probably the most, let's say, different, right? Some of those comments we made earlier about how much customer demand there is for manufacturing here means that that will be the lion's share of our growth investments. It's also very much tailored to those technologies like photonics that we talked about. I'll let Sam comment about, you know, the mix in terms of financials, but I'd say overall more in the U.S. for the, you know, near term. Actually some really good accretion from expansion in Germany and Singapore.
No, I completely agree. I would say that, you know, if you were to pinpoint where a lot of our CapEx is going this year, as you'd expect, a significant majority is going into silicon photonics. That's filling out the existing opportunities that we have within Malta today. We also have pluggables capacity in Singapore. One of the really important reasons why we acquired AMF was because they'd reached the capacity of their space, and there's significant opportunity to ramp and scale them into our Singapore facility as well. That transfer is underway. We're targeting the end of that year, this year to get to that scale opportunity there. You know, I would say that there are other pockets of important CapEx being spent. FDX is one of them, but SiGe as well.
As we think about our Burlington, Vermont site, that's where a lot of CapEx into SiGe is going to. It reinforces that principle that Tim said, which is we're gonna optimize the existing four walls that we have before we go and start putting modular expansion on.
Let's go to Ross Seymore here in the second row.
Hi, Ross Seymore from Deutsche Bank. Another CapEx related question, I guess two parts. One for you, Sam, and then a bigger picture one. The first part is the 20% longer term capital intensity gross or net? The second question is it enough considering all of the growth opportunities, the geographic dynamics that are going on, geopolitics, et cetera, are you trying to get ahead of that? Does the 20% do that, or would you spend more if you could?
Yeah, maybe I'll start and then Tim Bowen, we dive in. We think of that as a net target. One of the reasons we think about that as a net target is because there are two really important dynamics that contribute towards that overall gross CapEx. The one is the customer prepayments that I touched on earlier. The second is the level of support that we get under government frameworks. We feel quite comfortable that a net target through the cycle of 20% does not inhibit us from continuing to be able to grow that CapEx, provided we're getting the right level of contribution coming through from customer prepayments and government support frameworks. That's kind of point one. The other point I would say is that, you know, we think about this as a through the cycle target.
One of the reasons I included within my presentation materials, the profile that we've had over the last five years is you can see, given the high CapEx intensity of our business, that there's often a degree of lumpiness to the timing of those CapEx. We may well have certain years whilst we're even on a net basis, we're above that 20% target. My expectation is that as we see those investments scale drive incremental operating free cash flow, we're still gonna be on track towards that 20%.
Maybe just to add, Ross, I think the, you know, the conversations we have with governments is, can you go faster? We have more help available if you need it. I think that part is, you know, not fully tapped, let's say, in how we think about it. On the customer side, there's a lot of interest in growing capacity. We're always gonna have to strike the balance, but when we have those partnerships together, it makes it a lot easier to add capacity. If you think about it, when you're adding capacity, you're adding tools, which means the productivity is from day one, right? You don't have to build out huge shells before you get there. That's how we're thinking about it.
Let's go to Krish.
Hi, thanks. It's Krish Sankar from TD Cowen. I had a 2-part question for Greg. On the SCALE platform, you spoke about integrating the EIC chip. Is that single-digit nanometer, and are you doing that in-house, or are you actually collaborating with a external foundry? The second part of it is the way to think about it, SiGe is for AEC and ACC and SiPhO for optical, or can you actually use SiGe for optical? Thanks.
On the optical module, the answer to is it single digit nanometer or not, it's an open ecosystem for us. In instances where a customer has an EIC coming from another source, we're happy to integrate that into it. Of course, we have our own internal capabilities, both as a consequence of the InfiniLink acquisition that brings the know-how to the EIC photonic IC design capabilities, and of course, our own digital platforms that we have. We're an open ecosystem. We're happy to take in silicon from wherever it comes.
The SiGe?
I didn't catch your question. Could you re-repeat the question on silicon germanium?
Is the silicon germanium mainly used for AEC for copper and SiPhO for optical, or can you use SiGe for optical?
Silicon germanium's role in the silicon photonics space is used as the drivers and the TIA for the transmit and receive end of it. It's, you know, as Mike mentioned in his comments, it's the high-performance RF, right? You're trying to drive modulators on there at, you know, gigahertz. As high a fT/fmax as you can possibly get from the devices. There are plenty of wired areas where silicon germanium fits in the portfolio as well. Specifically, the applications that I was highlighting in my talk were about the role that is the TIAs and drivers in the photonic solution.
Let's go for this side of the room. Charles, please.
Thanks. Charles Shi from Needham. I have a question on the gross margin target, exiting 2028, 40%, long-term 45%. Can you help us break down how much of that is, let's say, a product mix driven, and how much of that is efficiency driven? For one, I do think custom silicon, that's a new area you want to get to grow that business to $1 billion. That should by itself should be 40%, 50% gross margin. IP, MIPS, ARC by itself should be 90% gross margin. The base wafer foundry business, what's the long-term target for that part of the business, the gross margin? Obviously you're getting a lot of subsidies, 20% capital intensity.
If I do my math right, your gross capital intensity seems to be scaling to 30%. How much of the base wafer foundry gross margin is helped by the government subsidies there? Can you kinda help us break down all the pieces there? Want to really just nail down how much of the efficiency gain you are expecting over the next few years. Thanks.
Thanks, Charles. Yeah, there's a few pieces to unpack within that question. Maybe I'll start with how we think about bridging the gross margin from, call it, the roughly 30% that we expect to exit 2026 with the target 40% for the exit of 2028. Look, we've really hit on this a lot today. It probably comes as no surprise to you that mix is going to be a significant driver of that. You know, think of that in the neighborhood of 5 to 6 points of margin opportunity over the next three years. You know, we don't break out our end markets by margin, but it probably comes as no surprise to you that the composition of silicon photonics, silicon germanium within communications infrastructure and data center falls through at a very attractive margin to our overall corporate objectives.
That's not least the fact that we've also got healthy margin dynamics associated with the likes of IoT, with the likes of automotive as well. As I said in my prepared remarks, we're not reliant on smart mobile within this plan. That's sort of the manufacturing services side of the mix dynamics within that 5-6 points. The technology services revenue within that 5-6 points of mix also matters. You rightly touched on the fact that we do expect that revenue associated with MIPS to fall through at a very attractive gross margin. I would caution 1 point in terms of your direct compare to where some of the IP providers are, which is that a big part of this business plan with MIPS is custom silicon.
When you're in the early phases of new engagements for custom silicon, you typically have some cost associated with that as well. It doesn't fall through at quite the level that you can read across from IP companies, but it's comparable. You know, the second part of it is as we think about it, is productivity and manufacturing cost productivity as well. We've set ourselves internally some very ambitious targets, which we're on track to as it relates to how we think about our manufacturing costs and our cash costs per mask layer. We think that over this time period, it can deliver 1-2 points of margin growth as well.
You have a couple of points that we'd expect to achieve as we see, you know, opportunities come through from our scale efficiencies across all of our sites as well. Clearly there's utilization within that as well, although I've embedded some of the utilization dynamics within the end market mix that I outlined. Hopefully that gives you a little bit of color in terms of how we're thinking about the bridge to 40%. Clearly, depreciation, which I think is what you're getting at by reference to government support frameworks, you know, that actually as we think about an increased CapEx cycle, that probably is a point or so of headwind as we think about it over the next couple of years.
Nevertheless, as I said, we're going to invest in our capacity in a highly efficient manner, and if we're able to recover, you know, anywhere between sort of 30% and 50% of our CapEx intensity, then clearly that's helpful in terms of how we think about the offset to depreciation. That offset is not a single year, it's an offset which comes through over the amortized life of the investments.
We can do Natalia here.
Yeah, just wanted to follow up on CapEx. I think this slide presented the near-term expansion as more of a brownfield opportunity, right? Whereas longer term is greenfield. As we think about the 20% net CapEx or maybe kind of in the gross CapEx equivalent. How should we think about that through cycle from the standpoint of how capital intensive it is to deploy brownfield versus greenfield?
Maybe I'll just again frame it 'cause of when we say greenfield, it feels like we're building a completely new fab. Look, we've had this experience now in Singapore. We built our new fab in 2021. This is expansion of an existing fab campus. We have an incredible AMHS system. You can see a wafer transit between the entire campus. It means that the first tool that you install in that new capacity is already a productive tool, right? Wafers can go from the old fab to the new fab. You don't have that greenfield economic challenge where you're spending time to ramp a whole line, right? It has very good ROI and actually has really good capital efficiency as well. That 20% is a through the cycle number.
There will be years where we spend a bit more because we're gonna be investing, let's say, in a new shell. For the next few years at least, a lot of that expansion is coming within the existing four walls.
That's it. We have, Yes. Sorry.
Thank you. It's Mehdi Hosseini, Susquehanna. I have one follow and one question. Just going back to the gross margin topic. If I look at the past 12, 18 months, everywhere in the semi land has seen a price increase due to high utilization rate. Now we're beginning to see analog beginning to feel incrementally tighter, and I think that's gonna get even tighter into the second half. In that context, how much of the gross margin target that you have here embeds better pricing power? I have a follow-up.
Maybe I'll start specifically on the model and Tim, you can give a bit of color in terms of how we think about pricing more broadly. From a model perspective, we're not embedding significant price increases just associated with supply and demand dynamics. As we know within our industry, those supply and demand dynamics can change quickly. Really the gradual uptick that you'd see associated with pricing on an average enterprise level will be more a function of the evolution of that mix that I talked to. It's more a function of, you know, shipping higher ASP wafers with a higher margin structure to some of those end markets where we see the strong growth. Tim, you wanna talk about pricing more broadly?
I think the general framework we take is, you heard it from some of our customers, these are long-term partnerships. We don't wanna have, you know, surprises in either direction, right? I think in a lot of those end markets, that's what we've been able to achieve. Now, where that changes is when you're introducing new features, new capacity, or ramping new technologies get priced based on value, right? When you're unlocking data center performance with Photonics, it's a relatively small cost to unlock a huge amount of system-level performance. We're able, obviously, therefore, to capture more value with those new products. As Sam said, like for like pricing, very little in that long-term view, much more of that's coming from product mix growth.
Just a very quick follow-up on data center. What's the mix of the $1 billion incremental revenue by pluggable transceiver versus optical transceiver, and how does that mix change into 2030 when your incremental revenue doubles? Thank you.
Yeah. Maybe I'll start, and Greg can obviously add some technical color as we think about the ramp timings. You know, really the way we put our plan together over the period up until the exit of 2028 is that a lot of that revenue growth is principally coming from pluggables. That's demand that we see today. That's demand that customers continue to indicate. It's actually a broadening of the customer base that has come through with the acquisition of AMF as well. Think of our plan to the exit of 2028 as principally pluggables revenue growth-focused. It's the beyond 2028 inflection point where we think there's an opportunity to start ramping and scaling CPO, and maybe Greg, you wanna talk about that a little?
Yeah, happy to comment on it. By the way, I'll tell you, we've already seen the first few tapeouts taping out under that OCI MSA that I mentioned, right? They've immediately snapped to that as a configuration for the future on there, and it is a matter of then those going in and ramping up. Completely aligned with Sam. We're certainly preparing now for the capacity ramp that comes when you go to, you know, millions of optical modules per month. Some of the investments we're making today are in fact in preparation for that. We really see that in a significant ramp in the material way probably the second half of 2028.
With C.J., please.
Thank you. C.J. Muse, Cantor Fitzgerald. Thank you for hosting today. Great day, great day. Wanted to follow up and then ask another question, if possible. On the silicon photonics side, target model seems to apply about $250 million-$300 million of non-pluggable business exiting 2028. Curious, is your sense that is the full scale or is that selling kind of discrete picks? How do you kinda see the mix within that?
It's a great question 'cause you, as you think about that module, there's different components within it, and I think as we work with different customers, they're gonna take different parts of that solution. We mostly think of it as a full optical module, C.J., so that's most of it. We're still in the early days of ramp in 2028, as we've said, right? One of the reasons there's an inflection point to CPO, and I think the industry sees that too, is high level of attach rate to GPU, and you're talking about an integrated module, and as you said, not just the one wafer component of that story. By 2028, relatively small contribution of CPO, but as that ramps, yeah, it'll be more of that full module build that we include. Yeah.
Great. Then, I'd love to hear more on technology services. Is there a revenue number that or growth CAGR that you're targeting into the 2030 timeframe? You know, are there specific contributors that you wanna highlight? Lastly, as it relates to custom silicon, will that fall within the manufacturing services, or will there be contributions down to the technology side? Thank you.
Maybe I'll start as we think about it from revenue contribution point of view. Taking a little bit, a step back down memory lane, non-wafer revenue, right? Which shall forever now be consigned to the trash can. We are thinking about this purely as technology services revenue, nevertheless, that was principally comprised of mask reticles and NRE. That typically was in the range of 8%-12%. We were sometimes at the lower end of that range, but typically around the 10% mark. As we've evolved this into a broader technology services solution, and as we have already started to get opportunities come through from an IP licensing or software point of view, that's one of the reasons why you've seen that pick up as a percentage of revenue.
Two pieces of guidance we gave at the start of this year was that we were no longer looking at the lower end of that being 8%. We narrowed that range 10%-12%. We actually expect that for the full year 2026, we're gonna be closer to that 12% end of the range. We ended up actually having 13% in the first quarter. Now, as I think about this longer term, you can expect that as we continue to get more IP licensing, software-related business coming through with customer partnerships, it should provide some scale opportunity to increase that from call it 10%-12% now to 12%-14%, but that's sort of in the second half of our model.
You can kind of infer from that, CJ, as we think about the broader, sort of 10%-12% revenue CAGR we're putting on the enterprise. If you were seeing sort of roughly 12% of that being technology services in the next couple of years, you can kind of infer the contribution there that comes through under IP. The other part of your question, as we think about it from a custom silicon point of view, at the end of the day, if it involves manufacturing of wafers, it will sit within the manufacturing services. It was one of the reasons why we called out the manufacturing value add, because that's where we see the opportunity coming through from custom silicon, but principally in manufacturing services.
Maybe, you know, C.J. Muse, if you think about the life cycle of a custom design, that's something I think is quite interesting. You know, day 1 software. Why? Because you're enabling your customer with tools so they can do simulations to think about their product architecture. Pretty quickly NRE, right? As you work together on some optimization of that design. You then move, you know, perhaps IP licensing, if they're going to use your IP upfront. It moves into production. That moves into custom silicon within manufacturing. You still collect IP royalties if they're using your IP through that process, right? That revenue cycle will vary a lot by design. What's great about that is we can engage very early, and so you'll actually see an early leading indicator as software starts to grow because customers are using it to begin that design process.
Let's go to this side of the room for Chris. Chris?
Sorry.
Yeah.
Thanks. Chris Caso with Wolfe Research. I wanted to ask about the change in cash return and the dividend. And it's always been a cyclical business, so, you know, that's had some peaks and valleys in cash flow. I guess the question is: What's changed in your view going forward that gives you the confidence to, you know, both start returning some cash and to put the dividend in place?
Happy to take that, Chris. Look, I think it really ties with a lot of what you've seen today from when we last presented a model to you and we took the company public to where we believe we are as a company today and where we believe we're gonna be as a company over the next five years. That is a link to the maturation that we believe is achievable as a company. We also have confidence that where we've already invested in our capacity footprint and the opportunities we have around our existing capacity footprint, we can be efficient from a CapEx point of view as well.
With that increased CapEx, with the driving of incremental capacity to support the demand that we believe is certainly being signaled by our customers today, you know, we think that creates a healthy generation of free cash flow over the next 5 years in the 10+% range. The reason it's 10+ is because there will be some lumpiness associated with how we think about the CapEx itself. All said and done, you look across the peer group that we're in, as I'm sure you do regularly, they have established capital allocation frameworks. They have established principles between how they're doing that.
You know, this is an important day for GF as far as the maturation of the company, and we think that we've sized this dividend at a reasonable enough level such that we can continue with all of the objectives that we want to achieve as a business.
Go to Joe, please.
Great. Thank you. Joseph Moore from Morgan Stanley. You talked about the sort of importance of, you know, your geopolitical stability of the regions that you can serve. A couple of years ago, when there were automotive shortages, you had these kind of automotive OEMs starting to form direct relationships with you. You know, I guess the shortages aren't as intense these days, but do you still have those conversations? Do you still have that connectivity, and are you getting a prioritization that extends all the way to the OEM level in the automotive space?
Thank you, Joe. I mean, actually, one of the things I like to do in my role is spend time with the whole value chain that we serve. Actually, although the shortages haven't been as acute for automotive, maybe barring Nexperia and so on, the automotive interest in partnering has only increased. Last year, for example, we announced a partnership with Hyundai Motor Group, by the way, one of the largest players today, by the way, one of the largest importers into the U.S. today, so you can understand why they're thinking about supply security. I'd say that's actually intensified, if anything, in the years that have gone by. They like very much both the combination of the technology portfolio, but also the distributed manufacturing footprint.
Many of them have global markets, and they want that flexibility because they don't know if they're gonna see more growth in Europe or the U.S. or Asia and beyond. By the way, I would even include that to include Chinese OEMs as well, because for many of them, they wanna export as much as anybody else. They wanna attack those exciting markets for them. They love the idea of manufacturing some of that outside China to be able to have resilient supply chains that their customers and the governments that are gonna let them do it will also appreciate. I think it's only intensified, and it's probably broadened to other markets.
I do supply reviews with end customers, you know, two steps down the chain on a regular basis around what are you seeing, where are you investing, what do you think you'll need in 2, 3 years' time? It's super helpful for us from a planning point of view as well.
Vivek.
Thanks. Vivek Arya from Bank of America. Thank you so much for hosting such an informative analyst day. I had two questions. One is, there's only a handful of people making CPO platforms, and, you know, some of them are engaged with your competitors as well. I'm curious, as part of your planning process, are you assuming a certain market share? Are you assuming, you know, that you gain share? Just, you know, what the underlying assumptions are, because it's an important part of, you know, the growth from 2028. Tab, as I zoom out, you know, and I know we are talking about a very long term and, you know, shares and all these things can become a little vague over time.
If I look at your sales in 2025, right, they were about $7 billion or so, right? The market was $75 billion, so just under 10% market share. Then you're saying you're adding another $4 billion, so that gets you to $11 billion, and the market is $125 billion according to your You're kind of staying at that 10% share, right, with so much leadership technology. Are you not assuming much share gain, right? Like, how do you conceptually think about share gain versus the targets? Is this conservatism? Just what is the message in that trajectory?
Maybe I'll start with CPO, then we can talk more broadly about how we think about, you know, share gain. There are industry forecasts out there for overall CPO market size. It's probably more useful to use like 2035 because you're still seeing the ramp debate that's happening. 2035, $25 billion market is what, you know, we see written about today. There will be multiple players in the CPO market. We have very strong conviction. GF will be 1 of the top 2. You know, if we succeed, we'll be number 1. If we fail, well, I think we'll be number 2, that's because of the technology advantages that Greg built. I also think when we hear customers' views, they wanna have at least 2 solutions, right?
They're betting very big franchises on this. We would expect that they will actually have 2 solutions going. They need that combination that we bring in CPO, the photonics technology, the advanced packaging, the scale, by the way, the footprint that also matters to them. We're pretty confident that we'll be one of those 2. You'll decide which one, but you'll decide who the number 1 is in both cases. Yeah. Overall share gain, you wanna comment?
Yeah, sure, and we very deliberately set out on that chart earlier how we think about the relative market CAGRs and the CAGR opportunity that we see from a business perspective. There was one number that you may well have picked up on communications infrastructure and data center, where we said 30+%. We do think that there's an increasing opportunity there for us, and some of that ties to what Tim outlined from, you know, the optical engine, the CPO perspective. And really it's gonna be the rate and pace of that ramp in the second half of the current decade and how we think about it going into the next decade that I think provides upside to that opportunity as well.
I guess just to emphasize, you're not assuming share gains as part of your forecast because your SAM growth and your kind of overall sales growth ex services is about the same.
Yeah. I'd say at a macro level, yes. At a market level and application level, no. Right? Because that mix shift that Sam talked about is because we're being very intentional on where we're putting growth on the page and where we also think growth could be an upside to the plan. I think there'll be areas we'll deliberately focus on. For sure, optical communication being a big one. Areas like automotive net net probably will gain a little bit of share. Mobile, we saw us growing slightly lower than the SAM, so that's a bit of an offsetting factor.
Let's go with Matt, please.
Thanks. Matthew Bryson from Wedbush Securities. I have two follow-ups, if that's okay. Just going back to gross margins, you kinda laid out a path from 30 to 40. I guess my question is, in 2026, you're working off 29% gross margin end of last year. A lot of the trends you talked about that are benefiting the gross margin trajectory in place, why wouldn't we expect you to get some of that benefit this year?
Maybe a couple of points to answer that, Matt. One of those is frankly having the capacity there to be able to meet that demand. You know, as we talked about on our earnings call earlier on this week, like we're already seeing oversubscription in high-performance SiGe through to the end of the first half of 2027. The only reason that is through to the end of the first half of 2027 is not because the demand dissipates. It's because we have more capacity coming online to be able to support that as we get into the end of this year. Some of that is somewhat as a result of the capacity dynamics. You know, a big part of our job is to make sure that we're as closely as possible matching the capacity and the demand dynamics.
I think silicon photonics, we're doing a consistent job on that, and we feel quite good about the demand versus the capacity we have going through the course of this year. A big part of the 15%-20% step-up in CapEx intensity this year is to support that demand that we see growing through the course of next year and the year after and what our customers are telling us. You know, there's, you're absolutely right to say that those trends are there, and they're happening now. Our job is to make sure that we have the capacity there to be able to support that growth going into 2027 and beyond.
Thanks, Sam. I just wanted to follow up on Vivek's question. When you look at your competitors out there, they've kind of some of them have been linked with large customers. You look at ST, for instance, they're working with Amazon. You look at TSM, COUPE is being used by NVIDIA. Tower has been linked with a couple module makers. When we look forward, should we be expecting that one of the CSPs is gonna come out and say they're using GLOBALFOUNDRIES? Can you hit your targeted goals kind of being the second source if everyone's using, like you said, or like Tim said, 2 sources for, say, CPO?
I think maybe the two comments I'll make. One is, you know, Greg talked about the OCI, MSA and who backed that, right? That was a pretty good list of the players in the industry, and a lot of what they based that decision on was technologies that we've proven. You can imagine there's engagement with a pretty good portion of that group. I think it's because, again, they wanna see what the technology can truly do. They do want to have multiple options. I'm pretty confident that we'll have engagement pretty broadly across that group going forward. I don't think we need to have an incredible inflection point of new customer engagement beyond what we already have.
You are seeing, as you rightly said, the CSPs get more engaged in the space, so far more in shaping standards, right, than in being direct purchasers of these systems. I think they're getting much more active because they realize that they're architecting these racks with different compute stacks that they wanna control more of. For sure, they therefore will take the next step of thinking about optical networking, about power architectures, and so on. I think those relationships and conversations are pretty good. I think they're gonna move in maybe the latter part of the plan into more direct work as they start to own more of that architecture.
Thanks.
A follow-up from Natalia, and happy to open it up to any follow-ups as well.
I'll stay with one as well.
Thank you for the follow-up. The quick question I had two very short ones. One was, does the $1 billion in software and services revenue include any contribution from ARC, or would that be incremental once the deal closes?
That's included.
Included.
Yeah.
The second one, the AMF 200 millimeter capacity, would that be strategically helpful for you guys to continue having that, or would there be a way that you integrate that with your current footprint?
Maybe I'll comment on AMF. AMF, you know, was a company with incredible technology and good customer traction that grew out of a research lab basically in Singapore, and it was gated by future capacity growth. We already have some photonic processes in 200 millimeter in Singapore. In fact, we actually had a couple of processes that we were already helping them on. What we're doing right now is transferring those technologies to our, you know, world-scale campus in Singapore. No, it's gonna unlock, full growth in 200 millimeter in Singapore. We have a longer term roadmap for Singapore because there's a lot of opportunity to grow further as well.
Thank you.
Russ, please.
Thanks. One quick follow-up on that whole market share dynamic people are asking about in various markets. The one that was the most overt was that your home IoT is supposed to grow significantly faster than the TAM. I wondered maybe, Mike, if you wanted to go through some of the specific drivers. Is that share gains, technology specifics, partnerships? Kinda what gives the confidence in that mid-teens versus the 9% for the SAM?
Yeah, I think it's mostly share gain. You know, not to read too much into it, but if you even look at the customer testimonials we had and think of the customers that were in there talking about the attributes that we just talked about, like the way 22FDX is really suitable for connectivity. They're seeing it at their secure edge, you know. The combination of sort of subjective comments like that and more hard data points like the number of design wins that we're seeing, you know, it's not a perfect science, but when you start to rack them up, you say, "Okay, this is actually gonna accumulate faster than it has in the past." I think we will benefit from all of the above.
Some share gain, some mix. Some of the older technologies that were in some of the earlier IoT weren't quite as differentiated as the ones going forward. We'll get a better value prop out of that. If you add it all together, combined with those mega trends of physical AI and edge AI, you get a better outcome, or at least we're expecting a better outcome.
I think just you touched on it there, Mike, so I'll build on it, which is, you know, when you think about the growth of IoT, the opportunity within that, and you overlay physical AI, I think it's pretty well appreciated a lot of physical AI will-.
Ends up there, yeah.
manifest itself within IoT. Again, one of the reasons that we have acquired MIPS, one of the reasons why we have a, you know, good conviction in RISC-V architecture is that we believe that that will be a strong supporter of not just physical AI growth, but also IoT growth for the business over time.
Russ, not to be a third wheel piling on, but I kinda have to. The mix within IoT, there's a lot of different end markets that fall within that. What you saw perhaps in the past was more on the consumer side. What you see going forward is also a significant growth in industrial, aerospace, and defense. A lot of these applications that, you know, are really starting now to roll out. That's obviously gonna be, you know, working in our favor given the very important requirements those industries have.
Believe we have a follow-up from Charles.
Very quick. I think I look at the long-term CAGR of different end markets. The one, I mean, at least for me, the surprise to the upside is going back to the IoT question. Understand all the comments you made, design wins, and a share gain prospect. Wonder how you think about what's your assumption around the inflection point, let's say, for physical AI? I believe it may be a big part of your assumption there because I think that the industry may still trying to figure out when that thing's gonna take off. 2 to 7 years, there's a wide range. That's a forecast I have seen and can easily go out of your 2030 forecast range. If things gets a little bit delayed.
Wonder what's your thought on that physical AI, when the inflection point is coming? What's embedded in your assumption? Thank you.
Yeah. I'm happy to-
No, go ahead.
No, look, I think, you know, what we laid out is quite a broad range of use cases that define physical AI. We sort of said the definition would be those 4 key functions, right? Sensing, thinking, acting, and communicating. When you take that lens rather than the lens of when will we have humanoid robots, that's a good discussion to have. By the way, if you go to China, the answer is already.
That's true.
I think we're gonna see some accelerated growth there as well. You see already deployment of AI workloads in all sorts of edge devices, and I think those ramps are more on the earlier side of that than not. Again, it's about fit for the technology portfolio because sensing, something we've done for a long time, by the way, in mobile and auto, right? Thinking, distributed compute where low power like FDX really matters, right? Combined with RISC-V. Precision control of motors and actuators, BCD technology, very well placed for that. Then the whole wireless communication. You know, RF is, you know, very core in our DNA. Wi-Fi, Bluetooth, NFC, UWB, you know, a number of protocols these devices have to do.
Look, I think it's a good example where the demand is actually coming at the technologies that we have spent a lot of time developing. Look, I think we'll really very much outgrow that market.
I think, you know, if you to draw a comparison to what we've achieved in other end markets, right? Automotive, you know, the 14x growth that we delivered there is a big part, not just of content expansion, but share gain as well. We very much think about the design wins being an early leading indicator of where we expect future production revenue to be. Well, we had those early design wins in automotive in the 7 to 10 years ago timeframe. That became the leading indicator for the growth that we saw in automotive over the last 5 years. Again, the reason we called out that D win momentum is because we think it's a good leading indicator for the opportunities ahead.
We have a question in the back.
Hi, Ted Kang from Kylin Management here. I don't want to beat the dead horse here, just going back to the gross margin again, it's impressive, you know, projection you have there, could you break that out to different business segments? In the long-term target, I'm thinking 2030 number, right? By that time, what will be the gross margin for each business to achieve that 45% target? Also, second question to that is, when you're getting the gross margin expansion from 30%-45%, from 2026-2030, yet your OPM margin is going from 20% to only 35%. That's 15 percentage increase both sides. Why aren't you getting more leverage on the OP margin?
Maybe I'll just tackle the gross margin piece to begin with because we don't break out the gross margin by end market. There's a very important reason as to why we don't do that. When you think about the tooling layout within a particular fab, you don't have corridors of tools that are dedicated to a particular end market. As a high fixed cost business and the fungibility that we have across our tooling footprint, you know, it's not the right way to think about our business as end market margins. Now, we can point to where we see stronger margin fall through within those end markets, and Communications Infrastructure and Data Center is a great example of that. Big part of the 510 basis points of year-over-year.
It's no surprise that we saw such strong growth within CI&D as well. That's probably the. I won't reiterate the building blocks that I shared earlier in terms of more broader piece around mix dynamics, cost productivity, how we think about scale. To the second point on your question in terms of how we think about the operating margin delta, you're absolutely right. You know, there is a delta of, call it, 15 points in the early part of our model, 10 points later on, and it's very much tied to some of the business decisions we've taken around technology services. You know, they are, albeit strong gross margin contributors, they are also R&D intensive businesses. A lot of engineers coming across with the MIPS and the ARC acquisition as well.
We think as we invest in that R&D capability and resource over the next couple of years, it holds us in that sort of 15-point delta to the gross margin. In that post 2028 timeframe, that's when we start getting the inflection point of custom silicon advanced packaging, you know, optical engines, CPO. Like, that's how we think about the scale benefits that come through with revenue at that point that then creates operating leverage to narrow that gap to about 10 points at the end of our model.
We have one last one from Matt.
Thanks for the chance to ask another question. Greg, I know it's early days, but you want to take a stab at exactly how much content you might have in a million qubit system or what the TAM of that market might look like. I guess just to follow up on that, how has IonQ's acquisition of SkyWater kind of changed the dynamics in that quantum space?
Stay tuned for the webinar that we'll thing. Let me talk in some broad strokes. We've had dozens and dozens of customer engagements that gives us an early indication of it. First and foremost, Cryo-CMOS. While I highlighted the 22FDX that while designed for IoT and millimeter wave radar, it turns out it is a perfect cryogenic CMOS technology, and there's a bunch of technical details why. The area that we have been surprised about, you know, we've been working with spin qubit and Photonic Quantum companies already. The number of companies that have approached us about superconducting interconnect, and we stepped back and realized we actually are probably gonna see the heaviest traction on superconducting and cryogenic interposers. Interposers will take on a different context there because in some cases, it's wiring wafers.
The superconducting guys need, you know, they have large qubits, so they need large amounts of interconnect. I think that's the area that's probably gonna see the heaviest traction based on the customer engagements that we have. One of the disciplines you have to have, there are so many modalities that exist in the quantum space. I listed, you know, 5 of them, is not getting yourself overly diluted and trying to chase too many things that look like a really diverse MEMS business already. Bottom line for us, 2 of the three columns are platform solutions. FDX, we have discussions on our FinFET platform for Cryo-CMOS, our 130 nm complementary bipolar CMOS technologies. You know, the ion trap guys need high voltage capabilities, so things like 55 BCD.
I think we're gonna find a suite of digital platforms that are Cryo-CMOS, heavy traction on the interconnect capabilities, and then maybe more focused individual modalities. Won't comment on the IonQ one other than to say that there are a large number of companies internationally that are interested in finding their scaling partners on there, and we've had very heavy engagement.
I think a lot of people were looking for Greg's number the day that announcement came out.
Yeah. Great. Thank you very much. That concludes our Q&A session. As we advance the slides twice, I'd actually like to, again, thank you so much for your interest and support in GlobalFoundries, and I'd like to kick it back over to Tim for some final remarks.
Thanks, Dave. Thank you very much for being with us today. I hope you leave with even a portion of the enthusiasm we have for what's to come. You know, we thought we'd leave you with some takeaways. For us, it's very clear we're at the middle of some incredibly interesting mega trends. Those will play out over the years to come. That AI story in the data center, the physical AI story, really putting wind to our backs. Combine it with the footprint that's truly very special. That leads to multiple high margin businesses that we can ramp. We are already ramping. Everything that we're growing in the future, we're already growing faster than our future forecasts. We're broadening our mix in terms of technology. I'm really proud of the acquisitions.
These teams are bringing incredible skills into GF. I spend a lot of time with teams like MIPS and hope soon with ARC. They're bringing new capabilities, and they're challenging us in new ways. We're building where customers need us. I think you heard every customer talk about how important the footprint is and how important it is for their own supply story. Lastly, we're doing this all in a more holistic business model where we offer to them earlier and broader set of solutions. Obviously, at the heart of that is building trust for the long term. Thank you very much. For us, it's been great to spend time together. For me, the blizzard was certainly worth it. We hope you enjoy the rest of the day and enjoy your lunch. Thank you.
Thank you