Hello, ladies and gentlemen, and welcome to Nokia's Capital Markets Day 2025. It's great to see so many of you here in person, and welcome to those of you that are joining us for the webcast. I'm David Mulholland, Head of Investor Relations at Nokia. Before we get started, a quick disclaimer: during this event, we will be making forward-looking statements regarding our future business and financial performance, and these statements are predictions that involve risks and uncertainties.
Actual results may therefore differ materially from the results we currently expect. Factors that could cause such differences can be both external as well as internal operating factors. We've identified such risks in the Risk Factor section of our annual report on Form 20-F, which is available on our Investor Relations website.
Within today's presentation, references to growth rates will mostly be on a constant currency basis, and profit or margins will relate to Nokia's comparable reporting. The presentation will be published on Nokia's Investor Relations website following the conclusion of the event. In terms of the agenda for today, you'll get to hear from many of Nokia's leadership team.
Justin will start and outline Nokia's strategy and the next chapter in the company's evolution. Pallavi will explain the journey we're on towards AI-native networks, and Raghav, in his new role as Nokia's Chief Customer Officer, will discuss Nokia through the lens of our customers. We'll then move on to the business segment presentations, with David Heard outlining the growth opportunity we see in network infrastructure.
We'll then take a 30-minute break so we can all recover and get some refreshments, and then Justin and Patrik will discuss our mobile infrastructure business and how we're transforming to lead in AI-native networks. We'll then finish the presentation with Marco bringing everything together from a financial perspective. We'll then conclude with a Q&A session, and we expect the event to last until approximately 12:30 or 1:00 P.M. New York time. With that, let's get started.
The AI supercycle. It's something I touched on when I was announced in February, and it's something that I've talked about continuously since I started as CEO in April. When I say supercycle, I'm not talking about something that's just a single boom. I'm talking about a technology revolution, because a supercycle is a surge in innovation, multiple booms, and it's waves that build on each other.
Now, most of us in this room are probably old enough to remember the last one, the Internet supercycle. In fact, back in 1997, I was starting out my career as an engineer in the middle of a boom. I was building 2G mobile networks across the United States. We were just at the beginning of the Internet supercycle, and the dot-com boom was just about to reach its peak. However, there was another boom underway, the one in telephony. This was the boom that I was a part of at the time, because in 1997, while Netscape was making progress, the device capturing the world's attention and imagination was the mobile phone. Shortly thereafter, there was a device that everyone wanted to own, and you're probably familiar with it because it had a sound that was exactly like this.
Nokia, then a relatively unheard-of company from Finland that began in paper and rubber, rose to become one of the world's most recognized brands and most valuable companies. It was a market leader for over a decade, not just because of technology, but because of what it promised. Its promise was simple. It was about connecting people, and it was more than a tagline. It was an idea that shaped a company and shaped a generation. It was an idea that transformed how we thought about communication. As the world moved from 2G to 3G, we all became obsessed with it. We became obsessed with calling anyone, anytime, anywhere: friends, families, colleagues. We became obsessed with sending SMS messages, so much so that we forced ourselves to learn how to type on a 12-key keypad. Many of us became obsessed and attached to playing a game called Snake.
The reality was that boom ended, and the Internet supercycle, which was getting started in 1997, gave rise to multiple booms, including many that started after the dot-com boom, because the Internet supercycle was more than just about connecting people. It was about connecting information. That connection spread from the computers on our desks to the phones in our pockets. As the world moved from 3G to 4G, new leaders emerged as the market shifted from the Internet to the mobile Internet, to SaaS, and ultimately to public cloud. There is a lesson in that that is very clear for us. Sustained success is not about riding a single technology wave. It is actually about focusing innovation and investment, and investing R&D in particular, where future opportunities lie, where we create real impact and ultimately real value.
When I talk about the AI supercycle, I'm not just talking about what's happening today, the ubiquitous deployment of large language models and deployment of AI agents. I'm talking about a multi-wave structural transformation that will raise new opportunities, reshape industries, transform economies, and ultimately advance human progress. As intelligence moves beyond the data center into the physical world, it will transform how devices interact, how industries operate, and how people live and experience technology. For companies like Nokia, it's a moment of disruption and a tremendous moment of opportunity, because it's an opportunity when new leaders can emerge. In fact, when you think about it, let's talk about LLMs. Hundreds of millions of people are already using LLMs: ChatGPT, Grok, Gemini, Claude, Llama, Perplexity, and others. They're not using it just to gain knowledge.
They're using it to generate images, generate video, write code, explore scientific questions, automate workflows, and increasingly leverage intelligent agents to assist in their daily lives. That shift is already visible in the network. We're seeing far more intensity and actually far more variability in traffic patterns and entirely new demands on latency, reliability, security, and throughput. In fact, our researchers at Bell Labs have done analysis that shows that as inference workloads expand and token exchanges multiply, consumer AI traffic across mobile and fixed networks is set to grow at over 20% annually through the next decade, with enterprise and industrial AI traffic accelerating at nearly 50% annually, driven by more distributed and more real-time applications. The reality is, if you think about every generation of technology boom, connectivity has become more critical, bandwidth is more essential, and interruptions cannot be accepted.
This is only the beginning of the AI supercycle. What this means for Nokia as a trusted Western provider of connectivity is clear. The AI supercycle is an opportunity for us to lead again. To seize that opportunity, we also recognize that Nokia needs to evolve. That is exactly what we are doing. Today, the company that used to be known for connecting people is now the company known for connecting intelligence. Connecting intelligence is critical because it flows directly from our North Star as a company, which is advancing connectivity to secure a brighter world. The first wave of the AI supercycle is already reshaping demand across our industry. AI adoption is scaling faster than any technology in history, and that impact is already showing up in the network, because the network is not just carrying bits of traffic. It is increasingly carrying tokens, the currency of intelligence.
These tokens represent everything from the prompts and responses flowing through LLMs to real-time sensor data used by intelligent agents and applications. This is only the beginning when you think about autonomous vehicles, robotics, delivery drones, and more and more adoption of augmented and virtual reality applications. We are already seeing the beginnings of this transformation, because across the world, AI factories are being built at an unprecedented pace. Data center construction is accelerating. Optical transport requirements are increasing, and IP networks are being re-architected to move AI workloads. AI infrastructure is also evolving rapidly as compute and memory fabrics become bottlenecks, creating demand for the kind of high-performance optics and switching that Nokia is known for.
In fact, in optical and IP networking, we're already capturing this opportunity, and we're serving as the backbone of many of the world's AI factories, moving the data and the compute that power AI training and real-time inferencing. We enable hyperscale connectivity across continents and within the data center. As inference workloads expand beyond centralized clusters, our networks are enabling data and compute to move closer to the users and the devices that rely on them. It's as this phase accelerates that our opportunity in network infrastructure expands. Today, the NI market is forecasted to grow from $48 billion in 2025 to $60 billion in 2028. We see meaningful upside beyond that as AI workloads become more distributed, data center traffic becomes more optical, more coherent, and more performance-intensive.
The reality is today, this is a portfolio and a business that sits at the heart of the AI supercycle, and we're already capturing the early waves of this demand. However, the opportunity doesn't stop here. As AI moves into the real world, into physical AI, into industrial AI, into autonomous systems, robotics, industrial automation, augmented reality, digital twins, access networks, both mobile and fixed, will need to evolve just as dramatically as optical and IP networks are today. Ultimately, to deliver intelligence everywhere, connectivity will need to be ubiquitous, trusted, performant, and adaptable. It won't be just enough to deliver it the way that we have traditionally, deterministic and static. Connectivity will need to be optimized in real time. It will need to be able to adapt.
It will need to be able to support the demands of tokens with the right bandwidth, the right latency, and the right security for that specific token. That is why the next major opportunity for Nokia lies in AI-native mobility. While the mobile infrastructure market is stable today, we see long-term growth coming, and it will follow the rise of AI-enabled devices and the demand for these intelligent applications at the edge. I want to transition and talk a little bit about looking at this not from just a market view, but from the most important lens that we think about, our customers. Let me start with telecommunication providers. As we all know, telcos today have invested billions in spectrum, significant CapEx in access technologies and transport infrastructure, and in many cases, are looking to unlock more value from their networks.
They are already looking to us today to evolve their networks, to improve performance, and to identify new opportunities to bring new services, new sources of value to their customers. In fact, one great example of this is a customer of ours called AT&T. AT&T has a bold vision for fiber connectivity. They are aiming to double their subscriber base to 60 million homes by 2030. Our passive optical networking technology plays a critical role in enabling that expansion. We provide the next-generation fiber platforms that will upgrade and extend one of the largest fiber networks in the world. It is not just in fixed access. It is also in mobile. Turning to Europe, in Estonia, Elisa selected Nokia radio technology for their entire 5G network. That radio network has independently been assessed as one of the highest-performing networks in Europe.
Ultimately, all of these high-performance access networks cannot operate on islands. They require high-performance transport networks. What we do in Network Infrastructure in optical and IP networking provides a backbone that allows this traffic to be carried reliably, performantly, and at scale. That is what sets us apart, that we bring together the entire network for telcos: radio, core, and transport, helping them to solve our customers' most complex challenges and helping them to deliver more value, more capability, and more services to their customers. The second major segment for us is AI and cloud providers. These are the companies building the physical infrastructure of the AI economy. As we all know, they are investing at a pace the world has never seen. In fact, just to dimensionalize that, the largest hyperscalers are now investing more each quarter than the largest telcos invest in a year.
Increasingly, they're coming to Nokia to power their AI factories, co-engineering and co-developing some of the most complex networks we've ever seen. In fact, today, nine of the world's top 10 hyperscalers use our optical technology. They're investing and trusting our innovation, from ICE-X 400 G and 800 G pluggables to IP routing systems that handle petabit scale capacity. We don't just sell to them. We co-create. This was a key point and key value driver of our acquisition of Infinera, because Infinera deepened our presence in this segment. It brought us depth and experience in co-engineering with this critical customer base. Just as importantly, it brought a culture of speed and agility that has strengthened our roadmap and our relevance with these customers.
When you look at the order and revenue growth, we're already seeing, not even a year into this acquisition, it's clear proof that we're making progress. You'll hear more from David on that progress and the roadmap ahead. Our last major segment is mission-critical enterprise, and this includes defense. This is a segment where connectivity is a strategic asset and a critical differentiator. This isn't broad enterprise. This is a very specific group of enterprises where persistent connectivity for them is critical because they depend on it for their operations, and their operations simply cannot fail. Examples of this are public utilities, rail, transportation, public safety organizations. They need performant, reliable, trusted networks. Of course, as I mentioned, it includes defense. As I've talked about, this is an emerging adjacency for us.
We see it as an opportunity, though it's still quite nascent when you look at it in the context of our financial results. To capture that opportunity, we today announced that we will stand up a new business unit called Nokia Defense. While this business is still in incubation, it includes our acquisition of Phoenix and our existing portfolio of Nokia Federal Solutions. Its objective is simple. Its focus is on accelerating co-innovation and co-development with partners in the U.S. and Finland, across NATO and Five Eyes nations. When you think about these customer segments, they're all slightly different. One thing is consistent. They all share a common need for Nokia's core technologies, from optical to IP to fixed access to core and radio networks.
We're uniquely positioned for this moment and this opportunity because we are the only Western company with a complete portfolio to power the most critical networks in the world. However, technology is great. Passion is important, but it's ultimately not enough. Winning requires focus, execution, speed, and agility. That is why we've outlined five strategic priorities to guide our strategy. Let me walk you through them. First, number one, accelerate growth in AI and cloud. We've been talking about this quite a bit since I started. What we're doing is simple. We're aligning our portfolio to the infrastructure that powers the AI supercycle, from data centers to the intelligent edge. We're capturing the demand that we see in the market today. Second, we want to lead the era of connectivity with AI-native networks and 6G, because we're pioneering trusted, secure AI-powered networks that make connectivity invisible and intelligent.
Our partnership with NVIDIA is a signpost of our strategic intent in this space. Third, I've talked about this a little bit already, but let me be explicit. We want to grow by co-innovating with our customers and partners. The reality is, for us, for where we are today, the products we have in market, and the future we're building, many of our best innovations don't come just within our own labs. They actually come from co-development and co-engineering with our customers. While we're doing well, I want to see us lean in here more, driving co-creation that improves performance for customers and strengthens our own competitiveness, giving us more sustainable differentiation in the market. Fourth, this is all about focus, deploying capital where we can differentiate. This starts with R&D.
We will invest where we see a technology advantage for Nokia, where it's clear, where we can deliver it at scale. Everywhere else where we cannot deliver that differentiation, we're going to partner with the best in the industry to accelerate time to value. There's a great example of this in our core, our core networks, core software, where we focused on real software differentiation. I'll talk about that a little bit more later. In that, what's not as obvious is we made a very important strategic decision. We recognized that building our own cloud stack was going to be something where we could not add differentiated value, and it was something that would slow customer adoption of our solutions. We made a bold decision and partnered with Red Hat instead of building our own stack.
When you look now at what we've done, we've accelerated our footprint across multiple public cloud platforms. Finally, most importantly, taking these steps to ensure we're unlocking sustainable returns. This comes down to a few principles. Number one, we want to drive AI-enabled productivity. Number two, we're empowering Team Nokia. Number three, we want to make sure that everything we do is associated with delivering durable value for our shareholders. Our strategy is clear. It's focused. It positions us to lead. Ultimately, now it comes down to one thing: execution. Just so I'm clear, since I started, we've already actually begun to execute against these priorities. One of the key steps was making Nokia a more outside-in organization.
What I mean by that is aligning us more closely with how our customers buy, how the markets are evolving, and making sure we're deploying capital against where we see the greatest opportunities. Now we're taking the next step, simplifying and streamlining our organization, because we want to be able to accelerate innovation, unlock operating leverage, and move faster in the markets that we're in. Starting January 1, we're going to move from four business groups to two business segments. Network Infrastructure is the first. Network Infrastructure is our Optical Networks, our IP Networks, and Fixed Networks portfolio. If you look at the past 12 months, Network Infrastructure generated approximately $7.8 billion in revenue and at a 10% operating margin. It also delivered a 43% gross margin. NI will continue to be led by David, and you'll hear from David a little bit later this morning.
What you're going to hear from him is that we view NI as our near-term growth engine because we have clear technology leadership and strong customer momentum. Optical and IP are winning categories for us, and our leadership in fixed networks positions us well as global fiber investment continues to scale. The reality is this segment is already sitting at the center of the AI supercycle, and it's where we're already capturing the early waves of AI-driven demand. Our second segment will be mobile infrastructure. This brings together our core software, our radio networks, and technology standards. When you look at mobile infrastructure over the past 12 months, it's an approximately EUR 11.6 billion business with 48% gross margin and 13% operating profit. We're in the process of seeking a permanent leader for MI. In the interim, it will be led by me.
MI is where we're focused on improving execution, product performance, and ultimately returns. We have some strong foundations to build on in this business. Our leadership position in core software and in automation is already enabling us to grow at above-market rates, and the robust portfolio we have of IP and technology standards provides durable profit and cash flow. We also recognize that while we've improved product competitiveness in radio networks, we still need to deliver stronger financial returns. We also fundamentally believe that while this market is flat today, this is where the next wave of innovation and opportunity will come from. Because ultimately, as I touched on earlier, as AI extends into the physical world, we see a tremendous opportunity for the need for AI-native networks.
This will be not only with 5G today, but ultimately the foundation of what we believe will be 6G. We have a unique opportunity to lead the industry with what we're doing in AI-RAN to make sure that we can execute on this opportunity and capture it and lead again. This is important because this is a fundamental tenet of this new structure. This structure gives us the clarity and the accountability that we need to not only address the near-term challenges, but to position the business for long-term sustainable growth and value. More importantly, it aligns our business with how the customers buy, ultimately how our customers see us and where their demand is shifting. It also gives us flexibility across both businesses to allocate capital and talent in the areas where we can create the most value and unlock operating leverage.
This is one step. The other thing we need to continue to do is to pursue active portfolio management. We have announced a set of steps today around taking exactly those steps, allocating capital away from areas that do not fit our strategy or do not fit our business. What we are announcing today is that we are creating a portfolio business segment. This segment will include four businesses: fixed wireless access, customer premise equipment, site implementation and outside plant services, enterprise campus edge solutions, and microwave radio. I want to be clear. These are good businesses. When we looked at these businesses in terms of the market opportunity and most importantly, against the five strategic priorities that I outlined, they are simply not the right fit with our strategy.
Over the past 12 months, these businesses generated approximately EUR 900 million in net sales with a gross margin of 22% and an operating loss of about EUR 100 million. We have set out a plan to improve execution in these businesses and to define the right path forward for each of them during 2026. Our priority is to ensure continuity for our employees and for our customers. As we all know, this is an important step, but organizational structure is ultimately only a step in the process of execution. Execution ultimately comes down to having the right people. That is exactly what we have been building since I started: a new leadership team to take Nokia forward. This team is a combination of existing Nokia leaders and new talent from joining outside.
In fact, when you look at the existing leaders, what you'll notice is almost every single one of them has had some kind of role expansion since I started. In terms of new members, last week we announced Kristen Pressner will join as our new Chief People Officer. She's joining us from being the CPO at Roche Diagnostics, and she'll start on May 1, 2026. Her charter will be to develop our culture and talent to ensure that we're prepared to meet the challenges ahead. As I said, not only have we transformed some of the existing members' roles, we've brought in a number of new members to the team since I started. The key thing here is when you look at this organization and you look at the GLT, they're all aligned to the leadership structure for the new Nokia we're building.
I'm excited about this because this is a team that's rich in experience, diverse in background, and united in mindset and purpose. I'm really thrilled you're going to get to hear from a number of them today: Pallavi, Raghav, David, Patrik, and Marco. The last thing, of course, in strategy and execution is the structure team is ultimately culture. This is something that we've been hard at work at. We've been starting to evolve Nokia's culture. We still have a lot of work to do here, but we have a new emerging culture, and it has a straightforward approach and a simple name. We call it Team Nokia. As I said, it's a simple name, and it's actually a pretty fundamental concept.
It's based on everyone having clear roles and accountability, empowerment to go execute them, and the importance of while we each play our position, we all come together to ensure the shared success of the company. That's called winning together. The reality is, and we're already seeing benefits of this, because when we operate as one team, we move faster. We're more agile. We make better decisions, and we deliver better outcomes. When we take accountability and lean in to help each other succeed, we create results that show up for our customers, our employees, and our shareholders. A recent example of this was a customer win, a win with Vodafone. We won a significant market share expansion in their 5G standalone network in the U.K.
The reality was, when you go back and look at this deal, this was not about one segment or one team winning a deal. It was about core and radio networks teams coming together to provide an integrated solution. That is how we deliver for our customers and ultimately how we win together. It is what the Team Nokia we are building today looks like. Nokia changed the world once by connecting people. We are at a point in time where we have the opportunity to change it again by connecting intelligence. This is the new chapter of Nokia: focused, differentiated, trusted, and already putting points on the board in the AI era. Marco will give you more detail, but I want to also summarize briefly how it shows up in our results.
First, we're using two key indicators, revenue growth and operating margin, to demonstrate how we're unlocking the opportunity for AI while we unlock operating leverage in Network Infrastructure. Second, we're establishing a clear baseline to position Mobile Infrastructure for improved returns while we invest in the long-term opportunity we see in the company. Third, we're driving company-wide efficiency, running a leaner center, unlocking corporate operating leverage across the business, and ensuring strong, consistent cash generation for our shareholders. All of that underpins our long-term value creation. It's why we believe that this is both a compelling near-term investment opportunity and a platform for long-term sustainable shareholder value. Fundamentally, what we offer today is an opportunity to play in the AI supercycle in the near term and will deliver mid-term profit expansion targeting double-digit operating profit growth through 2028. We're not just stopping there.
We're investing in areas that are going to continue to position Nokia for sustainable growth. In doing so, we're going to be incredibly disciplined. We'll be disciplined about our approach to portfolio management and to capital allocation and continuing to drive productivity improvements. This is fundamentally a unique and compelling investment opportunity. I'm excited for you to hear more about it from the rest of the team that's here with me today. With that, I'd like to introduce Pallavi, our Chief Technology and AI Officer.
Thank you, Justin. It is so nice to be here in front of all of you as part of my very first capital market day with Nokia. Now, as a technologist, I have a strong eye for spotting when the next wave of disruption is going to hit, ride that wave, and then be the disruptor. I was born doing networking. A decade back, I moved to go back and build the compute for the AI infrastructure. Started with the cloud, then to the edge, and then to the data center. Now, as I have been building the AI infrastructure, I can now see that the next wave of disruption is coming, and this time it is going to disrupt the networks. Fundamentally, if you look at it, the AI supercycle, it has put an exponential growth on the demands of these networks.
Now, whether it is latency, whether it is capacity, whether it's bandwidth, whether it's reliability, and these metrics, you know, they've always been vital, but today they are increasingly becoming critical in scale and in urgency because this demands a new kind of network: networks that connect intelligence and networks that become smarter by using that intelligence. Nokia is the only Western company which has a comprehensive portfolio spanning radio, core, access, transport, IP, all the way up to the cloud, uniquely positioned to shape and deliver the networks of today and tomorrow. This is why I'm so excited to have joined Nokia as the Chief Technology and AI Officer. Now, let me start by sharing my journey.
I told you about riding these waves, so how I've gone about and gone through those waves, the lessons that each of the waves have taught to this industry, and how we are going to use those lessons as we go about and make a leapfrog in building these AI-native networks. Now, Justin was talking about the internet supercycle. That's where I started my career, building networks for the early internet. At that point of time, networks were all about speeds and feeds. This industry was disrupted with software-defined networking. I was a core part of the team that went about and defined how the control plane and the data plane should be separated, and then a core architect to move networks from CLIs to APIs, having authored many, many RFCs. This is where programmability became the norm. Next was cloud.
Cloud gave users scalability, elasticity, flexibility. To leapfrog cloud's adoption, we introduced a new currency, and that was about trust and security. That just made the clouds fly. As I was building the cloud, I could see that there was a tiny little thing called machine learning, which was starting to pick up. You could see intelligent weather predictions coming in. You could see some critical drug discoveries being possible because of the technology. I decided to ride that wave by building supercomputers for high-performance compute workloads. Remember, at that point of time, AI was just yet another high-performance compute workload. As we could see, the next wave of disruption was just around the corner, and that was with model training and inferencing. That is when, by building the world's first exaflop supercomputer, we moved this industry from petaflops to exaflops.
All of this innovation actually made vision-based inferencing happen. It started first at the edge. It then moved over to the data center and finally moved over to the cloud. As I started to see more and more AI move over to the cloud, I moved over to that side to build the AI-based cloud infrastructure that has enabled the next wave of disruption, which is generative AI. Now, suddenly you could see that model sizes started to explode from millions of parameters to trillions of parameters. ChatGPT, I think it started with about 117 million parameters, and today it is close to about 2 trillion parameters. Today we have entered the world of agentic AI. This is where intelligent agents interact, and there is massive machine-to-machine chatter. If you see, we have had massive technology disruptions in such a short span of time.
The key to this transformation was actually fueled by open ecosystems: open data, open framework, open models, because openness fosters innovation and collaboration. Now, throughout this AI supercycle, networks have been foundational, and networks have also gone about and changed. You know, whether you're talking about scaling up for machine learning, whether we are talking about scaling out for generative AI, or we are talking about scaling across for agentic AI. Now there is another disruption waiting to happen in this AI supercycle, and I can actually see it coming because AI is now on the cusp of the next wave of transformation, which is physical AI. This is where the boundaries of the physical and the digital worlds are going to get blurred. Now, think about it. This is the autonomous vehicles, drones, AR/VR glasses, intelligent factories, healthcare.
This is the world where critical essential services will demand that the network should always be on, where sub-split-second decisions need to be made, and every millisecond matters. With physical AI, we will have robots standing hand-to-hand with humans with heavy machinery. Now just imagine, just imagine that these robots, you know, the safety and the motion control loops that they need, they need decisions to be taken in the matter of microseconds. One network slip and the robot can actually miss a safety stop. In these environments, it's about safety. Now, let's step back and look at, you know, this whole transformation that I'm talking of. GenAI was three years back. 2025 is the year of agentic AI. Now I can clearly see that physical AI is knocking at our doors.
This evolution of AI will exponentially change the dimensions of the network themselves. Now, whether it is bandwidth, whether it is capacity, whether it is latency, whether it is reliability, all of these KPIs are changing with AI. Now, let me hit on a few examples and talk about how AI is changing traffic, and then what does it mean for each of the KPIs that I just spoke about.
Hi, this is Pallavi. I just joined. Hello, can you guys hear me? I'm going to turn my video on. Can you guys see me now? Is the connection working?
Now, as you see in this video, I started with a voice call. Can you hear me? I moved over to a video call. Can you see me? As I moved from voice to video, you could see that the bandwidth increased, the bandwidth demand increased. Still, the traffic was very predictable. The other important point to see here is that the downlink traffic is higher than the uplink traffic. Now, this is the world of yesterday, as I like to call it. Let's look at the world of today.
Hey, where am I? All right, I see those huge digital screens and all the bright, bustling ads around you. It looks like you're right in the middle of Times Square in New York City.
Now, as you see, the traffic is bursty. The traffic is unpredictable. What you also see in here is that the uplink traffic is higher than the downlink traffic. In fact, my teams and Bell Labs are predicting that in the new world, today the downlink to uplink ratio is 12:1, and it's going to move to 4:1. We are seeing more and more AI-native traffic originating from mobile devices: 48% of ChatGPT, 61% of Gemini's AI traffic is originating from mobile devices. Now, but wait, what we just saw was just one AR experience on one device. Tomorrow, the connected landscape will look radically different. You know, AR glasses are going to be common, as common as smartphones. We will be using them for navigation, for shopping, for translation. 50% of autonomous vehicles will be driving by themselves. Tens of thousands of robots will be delivering parcels.
AI agents will be there in every store for pricing, for merchandising, you know, for better customer engagement. You know, think about all the city-edge AI factories that will be orchestrating traffic, you know, energy, logistics, all in real time. Now, let's just start doing the numbers. Multiply that across cities. Multiply that across countries, across the world. The cumulative effect is staggering. We are talking of billions of devices generating AI-native traffic, streaming video, telemetry, immersive experiences, all happening simultaneously. Now, to power this future, networks must also scale across three critical dimensions. We are talking about reliability. We are talking about bandwidth. We are talking about latency. Let's start with reliability. Today's networks, they deliver five nines. Now, what does five nines mean? It means that, you know, the networks, there could be minutes of downtime in a year. We were talking about robots on a factory.
When robots on a factory floor are working hand-in-hand with humans, the downtime has to shrink to seconds. We are now starting to talk about six nines, six nines reliability, because one missed safety signal can cause an accident. Now, let's look at bandwidth. You know, we talked about, we showed you an AR/VR example. We have so much of AI inferencing that is happening. All of this, because of this, the bandwidth has already jumped 7x to 400 gigs per port. Tomorrow's network, they need terabit-class capacity to feed these giant AI models and real-time video streams. Now, think about all the AR glasses that are going to render 3D environments, autonomous vehicles, all of them streaming sensor data at the same time. Now let's look at the third metric, which is latency.
Today, latency is about 10 ms-20 ms, and that works well for video calls. But imagine, once again, the robots in a factory floor. When they are making safety decisions, we are talking of sub-millisecond latency. You know, when you look at these numbers, you might feel like, okay, you know, these are some small jumps that we are talking about. As a technologist, having done this in my life, I can tell you that in order to make these leaps, it is a lot of hard work and a lot of research and development that goes in order to make it happen. Because when the industry moved from three nines to five nines, it took us a lot of work. Now we are pushing even further to get these networks ready for AI-native traffic.
You know, we spoke about bandwidth, how we have to increase bandwidth. We spoke about how we need to reduce latency. One of the very fundamental things that we also have to do is you cannot increase the power envelope. In fact, we at Nokia, we are constantly pursuing to go about and innovate to decrease the power per bit, not in a linear fashion, but in an exponential fashion. I spoke about all of that. You know, there is also, we are also starting to hit some fundamental laws of physics. Spectrum is finite. Energy is finite. The question really is, is there a smarter way? Is there a different way to now build these networks? I feel that this is the opportunity that is in front of us. I joined Nokia about two months back.
When I joined Nokia, I already knew that Nokia had these end-to-end assets across the network. In the last two months that I've been here, I have been digging very deep across our portfolio. I have discovered that, you know, we have some incredible work going on. We have a clear roadmap. You know, we have a clear roadmap that spans across chipsets, across software, and all our assets. I feel that, you know, all of this is going to enable us to lead through the next wave of building these truly AI-native networks. Now, let's start with RAN. In RAN, our strategic asset is our software. This is the AnyRAN portfolio. In fact, this was the very first thing that I did after I joined.
You know, we went about and forged and expanded our partnership with NVIDIA by bringing in NVIDIA's R-Pro GPUs into our radio networks. Now, what this enables us to do is this enables us to bring in AI-based innovations into the basebands. Now, with this, we are bringing in GPU support to our proven field-hardened software on the AI-RAN platform, which now enables a seamless software-defined evolution. Let's hit onto the core. This is where the team has actually done some fundamental, really rock-solid work, as Justin was talking about, in making it fully cloud-native, which means that this software has no hardware dependencies. Now this team is working on making it more AI-native because, you know, in the world of AI applications, in the world of AR devices, in the world of robots, they are not going to speak 3GPP.
We are evolving the core beyond the traditional boundaries, making it programmable, making it agentic, making it service-aware, making it ready for AI-native functions, functions like real-time translation, you know, fraud detection, and much more. In IP routing, our portfolio is built on unique silicon. This is a silicon that we call FP5. On top of it, we have some robust software which has hardened protocols built, APIs, you know, the programmability that I was talking about. In fixed access, we are pushing the performance frontier. We recently launched the world's first 50 G PON. This is where we are getting ready for ultra-fast broadband for enterprises, for campuses, and for AI-heavy edge sites. In IP networks, we deliver the fabric that ties everything together. Our new switches have actually doubled the performance at 1.6 Tb per second. Now, optical.
This is where our differential comes in by owning the full stack. Now, this is a market which is rapidly evolving. This is where our innovation in material sciences is absolutely essential. This is the work that David and his team are doing. Because it is work like this that helps us achieve capacities like 1.2 Tb per second, you know, capacity like 1.6 Tb per second. It's only possible through this level of innovation. Now, our 800 G coherent pluggables, these give operators and hyperscalers the headroom that the AI era demands. Across all these domains, on top of this, we have our software, which is becoming more and more intelligent. Our EDA platform actually brings in AI ops into operations, cutting the downtime by up to 96%. Altiplano's new AI models and automation tools are making networks more reliable and easier to operate.
Now, like I shared, the KPIs that the networks of the future demand is very hard. That is why myself, as the Chief Technology and AI Officer, I'm hyper-focused on taking all of Nokia's differentiating assets and doing the hard thing, which is harnessing the rich data, the deep insights, and the domain expertise that we have built over years and bringing AI into the protocols and the algorithms that make these networks happen. In short, what I'm talking about is that the networks that have powered AI are now going to become more intelligent by using AI. Building this intelligence does not happen by chance. There are two foundational elements that you need to go about and bring in intelligence. It's data and it's domain expertise. By the way, this is exactly where Nokia stands apart.
With more than 30+ years of experience in digital networks, I like to say that, you know, from voice to packets over TCP/IP, we own the network from radio to cloud. Now, what I mean by this is, you know, when a voice call evolves, this is like sound waves. It gets translated to analog signals, to digital frames, to RTP packets over TCP/IP. It traverses the radio. It traverses the transport. It traverses the IP switching domains before reaching the cloud edge. This is where, with assets spanning radio, core, access, transport, data center, cloud, we manage the full protocol stack. I like to call it from waveform to workload. Now, this means that we have unmatched visibility across the network. We have unique, hard-to-replicate data that gives us a training and inferencing edge to go about and create these AI models for the networks.
From a domain knowledge perspective, this is where decades of experience working closely with operators around the world means that, you know, we have a very deep understanding of the full network. We understand how our customers go about and design. We understand how our customers go about and install their networks to how they go about and optimize these networks. Now, let me give you a couple of examples to show how we will be bringing in this intelligence and why it matters. I spoke about the strategic advantage in our portfolio with our AnyRAN software. Now, let's talk about beamforming as an example. This is where the partnership that we did with NVIDIA, which is the AI-RAN partnership, also comes in. What is beamforming? Beamforming is a technique which is actually very, very foundational for modern high-performing, you know, high-frequency networks.
Essentially, what beamforming does is that it allows you to focus radio signals like spotlights, spotlights that focus users rather than broadcasting everywhere. Now, for beamforming to work, it has to continuously go about and track the network and figure out where the best beam placement should be. Now, today, the way you do it is because these algorithms were written, you know, some 30 years back. The way you actually go about and do it is using brute force scanning. So the network goes about and sends hundreds of probes, measures each one of them, and then selects the one which has the best signal. Now, this works, but it's slow. It's bandwidth-hungry. It's compute-intensive. Now, what if instead of brute force scanning, the network could predict where the optimal beam should be?
That's exactly what our researchers at Bell Labs have done by bringing in AI-driven Bayesian optimization. What we've done is we've made beamforming now into a learning problem. Now, what happens is every time the network sends a probe, we go about, the AI model goes about and updates the signal landscape, which means it updates the model. It starts predicting where the best beam placement should be. Now, instead of scanning hundreds of probes, hundreds of beams, the network just goes about and sends few probes and can still land within 1 decibel of optimal precision. The results are straightforward. You know, we're talking of faster beam alignment, which means better user experience. You now need less spectrum overhead, which means less CapEx. You need less compute, which means less energy, which means less OpEx.
Now, when I was talking about our portfolio, on top of our portfolio, we are also building a unifying intelligent layer that we call autonomous networks. This is a layer that is going to make the network simple to manage. This is where we are bringing in intent-based self-management, you know, self-optimization, self-healing, self-security. To show you what it looks like, let me actually take an example. Now, Justin spoke about how Nokia today powers the data centers that connect, that interconnect, you know, nine of the ten top hyperscalers. Now, what these hyperscalers really want is they want reliable optical backbone that connects these massive AI GPU, AI factories. This is where we are bringing in a layer of software applications called WaveSuite. It has three big things.
First, it brings in deterministic AI to accurately model the performance of the optical network and then do end-to-end optimization. Second, it uses generative AI to explain the trade-offs to the human operator who's going about and configuring and optimizing the network. Third, with its intimate connection with the optical signal processors, it can monitor the extremely sensitive optical signal properties and can predict non-trusted activities before they affect any service. Now, you know, we are starting to talk about networks that just do not react. These are networks that are now anticipating. In fact, in a live trial we do in the UAE, we saw that by using this, the optimized performance modeling, it cuts the planning time in half, and it actually improved the design efficacy by 30%. Now, as I wrap up, I will reiterate, our pursuit is simple.
We will use every ounce of intelligence available to us, our assets, our data, our domain expertise, and our partnership, our deep partnership with telcos and hyperscalers to build networks for the new era of AI-native traffic, networks that just don't carry intelligence, but networks that continuously learn, adapt, protect, and improve. As we build these networks, we will partner with the best-in-breed partners. You know, Justin was talking about how we did the partnership with Red Hat. We will partner with folks in silicon, with folks in software, with folks in platforms to help our customers unlock their next wave of growth. Now, talking about customers, I'm going to hand it over to Raghav. Thank you.
Thank you, Pallavi. That was very inspiring. It's fantastic to be here. Good morning to all of you, by the way. Just listening to what Pallavi and Justin have just shared with us, there are some very powerful themes that actually start to emerge in this era of what are you calling is the AI supercycle. The pace of technology change is unlike anything that we've seen before. I think you guys already know this. Silicon is moving at a pace that is actually every two years, you're getting a five-fold increase in just the speeds of the GPUs. Actually, if you put that into a stack, that's about 30x in terms of improvements that you're getting. As we just heard from Pallavi as well as from Justin, AI is scaling at a pretty alarming rate. I can tell you that whatever we've predicted there in terms of growth rates, I'm sure it'll outgrow that as well.
What this is really accelerating is learning and decision-making at unprecedented scale. All of this is also causing a lot of growth in the data traffic, but not just data, but as Justin talked about, was in the area of tokens, giving rise to new digital currencies. Simultaneously, we're also seeing this emergence of the physical and the digital world that Pallavi talked about. There are new types of devices coming out from Meta, Microsoft, other players. There are robots, drones, and sensors and virtual platforms, and they're exploding in numbers, creating new immersive experiences that give rise to an entirely new set of services. Amid this transformation, the fundamental challenge that we all have is how do we keep up with this change?
The challenge that we have with the customers is that how do they future-proof these networks and data centers to scale with the massive but unpredictable data growth? That's the biggest challenge they're facing. The key thing for the telco customers is that you have to be able to deliver this at the improved power per bit and cost per bit, building automation into these increasingly complex systems while all maintaining five nines and also addressing the cybersecurity risks that we are actually facing out there. Now, about the change, it's not so much about the change, it's more about how we actually adapt to that change, how our customers do it, how do we do it, and also fast enough. Those who can master this transformation are actually going to be the ones that actually create tremendous value.
This was the opportunity that actually got me excited personally. I had to just sit down with Justin and say, "Listen, this is an amazing moment in time in our industry." I wanted to be at the center of it. I knew the customers, and I wanted the role to be very close to the customers. That is why I took on the responsibility for being the Chief Customer Officer for the company. We all travel. I have been traveling extensively, meeting our customers. Actually, if you want to send me a Christmas present this year, you can direct it to seat 16A on United Airlines out of San Francisco. Feel free to do that. What is really interesting is that when I meet them, the customers are extremely optimistic, but there is a level of anxiety in them as well.
What they're really looking for is that in this journey, in this change, how do they latch on to trusted partners that actually can guide them through this journey of this enormous change that the world is witnessing? There is a second fundamental challenge that they're also facing, and that is skills. How do they upgrade the skills? Where do they get new talent to adapt to this change at this speed? Now, what is interesting is that as I talk to the customers, they are actually looking at redefining the engagement model between customers and suppliers and vendors. What they also have to do is they recognize that they have to partner with a much, much broader ecosystem out there. The traditional way of acquiring technology was to set up your requirements in an RFP, and we used to respond to those RFPs.
With the innovation cycles that we are in today, that is something that's just not quite going to work. The fastest clock speed that matches the velocity of technology itself is what we've got to implement. The other area that has also become increasingly obvious is that these transactional relationships need to turn into strategic relationships and this notion of customer intimacy. I think we heard it from Justin, we heard it from Pallavi, and this powerful notion of co-creation. What that means is it's not just about selling products, but it's actually our engineers sitting down with their technology teams and actually shaping the future, aligning the roadmaps, working side by side to deliver value and outcomes. That is the journey, and that is the way, the mantra of working together going forward.
Now, what is interesting is, as I talk to these customers, we have a very, very unique position. Customers appreciate the relevance Nokia brings to them. You say, "Why? What are those reasons?" We talked about the broad portfolio. Pallavi went through it. We have a portfolio that cuts across the entire gamut of the network and also within the data centers themselves. That is something that is really appreciated by our customers. The second, our heritage in delivering mission-critical networks over the past multiple decades that we've been doing. That is not easy to do. The investments that we are making that we've talked about in cloud and AI, and they see this supercycle as a very key part of what they have to play in, and they're happy to see that. We've talked about this, and you'll hear this again and again.
It's this co-creation learning mindset, which really helps them modernize networks and data centers while ensuring two fundamental things: trust and security at scale. Now, what they are really intrigued about is the vision. If you know the game of ice hockey, the vision really is where the puck is going to be and the journey to that position. It is not so much of where the puck is today. They really, truly appreciate that. Now, this slide is something that Justin already showed a little bit earlier, but this is just a place as a reminder of the customer segments that we serve. The market opportunity in the CapEx is pretty large, as you can see in the segments we serve.
What I'm going to try and do over the next few minutes is to really unpack these customer segments to show you how we unlock value and actually give you examples of customers of how we are innovating with them. Let's get on this journey for the telecommunication providers. The telcos have been a foundation of a business that we've built Nokia on. You can see a whole group of logos that are out there. These are some of our strategic customers that we do business with. Here are some interesting facts that I'd like to share with you. If you look at the 5G standalone network in itself, 70% of the world's 5G networks as standalone networks actually include the Nokia core platform. Let's look at fiber. 70% of the fiber broadband connections in North America run on Nokia solutions. Here's another interesting statistic.
Over 1 million base stations, which actually power 15 of the world's 20 fastest networks, run Nokia solutions. That's pretty cool. Now, what the AI supercycle really enables is, while it creates opportunities, let's be real, there are challenges in this industry. The customers are facing issues around capital returns, monetization, RFU slowdown. The opportunity that the AI supercycle now presents with this explosion in traffic brings opportunity to monetize just not bandwidth, but things like AI tokens and other types of services that will develop on top of the network. That's where the excitement is. That's where the opportunity is. Now, this requires that you've got to transition these networks to support this large explosion of traffic.
This is where our customers are actually moving down a curve to drive fiber modernization, mobile densification, and implementing cloud-native 5G standalone core capabilities and augmenting satellite as well to be able to provide coverage on an end-to-end basis. That is where the opportunity is. Now, the most innovative telcos are leading with value. It is very, very important that that is the metric that they are setting. They are adding new offerings like security, APIs, dedicated bandwidth, and laying the foundation of AI and networks to the path of 6G. We heard this from Pallavi. All provisionings in their networks for the changes that she described require the impact of uplink, inferencing, and AI at the edge and enabling autonomous automation, to name a few that they have to embark as a journey on.
Now, it's time to give you some real concrete examples where I actually had the luxury and the privilege to have personally spent some time on. Let's look at T-Mobile in the United States. It's one of the industry's most innovative operators. What does partnership mean to them? It means, in practice, where they work with us, NVIDIA and Dell, to lead an AI-RAN. What did we do here? We combined radio expertise with accelerated computing to create networks that learn and optimize in real time using artificial intelligence. Now, I'm very, very pleased to announce that last week, we actually achieved the first live RF call that was established in an outdoor lab system using a commercial device. This is a powerful first and a great proof point of innovation in the industry that we accomplished with T-Mobile.
Now, Justin spoke about AT&T, another truly innovative operator. A year ago, we signed a five-year deal to deploy the next-generation fiber access technology to support one of the largest fiber networks in the world. This includes a range of next-generation PON technologies from 10 G to 100 G, giving AT&T the choice and flexibility to optimize its network to specific business needs. We certainly appreciate AT&T's business, and we look forward to continuing to work with them in new areas. Now, let's jump across the world and get into India. Bharti Airtel is one of the leading operators there, and they needed automation at enormous scale. I was personally involved in it. When we talk enormous, we're talking about a country that has the highest population in the world. It is pretty large.
Now, what we did with them is we actually sat down with their technical engineers and our technical engineers and actually co-created an appliance-based packet core innovation to manage traffic at the edge. This is really, really cool innovation, which is now actually deployed by more than 100 operators worldwide, reducing cost by 40% in global deployments. This is exactly what a clear demonstration of when we do something impactful with leading telcos, the rest of the world adopts. It is very, very important that we co-create with the leading players in the world. Let's jump across to Europe. Deutsche Telekom is a leading operator there. We actually co-developed a transition with them from the legacy to our most advanced radio platforms. What did they want? They wanted flexibility. They wanted interoperable networks. We earned their trust to truly open-standard-based suppliers of radio networks.
That was also pretty cool. Staying in Europe, I still recall these early discussions that we had with Telefonica Germany. Boy, they had a desire to really embrace the cloud and do something very, very different to bring cloud-native functions like core onto AWS, moving policy engine onto Google. This is another striking example of pioneering industry-first approaches. Now, telcos are something we continue to serve, but there are also subsegments of wholesalers that are growing faster than the telecom providers. One example is Zayo. We won IP there, and we displaced a competitor. We already have a long-standing relationship supporting their optical networking infrastructure with Zayo for already a while. This is continuing to land and expand with our customers.
In conclusion, for telcos, whether it's AI-RAN or fiber builds in the United States, or it's automation at scale in India, or openness in Europe, telcos tell me they choose Nokia because we are the only Western company that provides a broad market-leading portfolio that encompasses radio, core, fixed, fiber, transport with complete network automation and secure. Moving on, let's move on to one of the most exciting spaces of AI and cloud players. This is where we're seeing, obviously, a lot of growth that we all know about. I was with Cloud and Network Services, and this is where I had personally the opportunity to lead the charge in forging strong ecosystem partnerships with leading cloud players like Google, AWS, Microsoft to transform and cloudify network functions to be running on hyper-scale platforms.
Now, the scope of that relationship has now expanded, accelerated by our Infinera acquisition, which now puts us in a very, very strong position where nine of the world's top 10 AI and cloud companies actually use Nokia solutions. As you can also see, there are hundreds more that actually continue to be our customers in this space. This is a truly exciting space for us. Actually, our success is just not limited to hyperscalers. We're rapidly expanding into emerging and innovators such as CoreWeave, NScale, and these are really new providers scaling fast in the AI economy. We also see opportunities emerging in the intersection of cloud and telecom, particularly in managed optical fiber networks that actually link data centers across the fast-growing regions like the Middle East and India. Let me give you a few examples here where we are winning.
In optical networks, two major hyperscalers have chosen Nokia's 800 G coherent pluggables and next-generation line systems. We co-designed these with our customers to actually deliver 800 G over hundreds of kilometers with the lowest power and cost per bit in the industry. If we look at another key differentiator that we provided here was our U.S.-based fab and packaging facilities. This provides supply chain assurance and resilience, which is very critical in this environment. Let's move on to IP. We secured wins here for our high-performance switching solutions with multiple hyperscalers and newer cloud players. Our superspine platform here is the lowest power design in the industry, built for scale and future expansion. These customers are really true partners in every sense of the word. We collaborate on multi-year roadmaps, tools, automation, and services that will shape the next phase of the AI supercycle.
Let's go on and look on to the last segment and look at the mission-critical enterprises, which also includes defense, an equally exciting area of growth. Here is where we're actually deploying 5G networks with the same reliability and performance and security that we actually deliver to the telcos. These are purpose-built for agencies and industries in public safety, utilities, transportation, medical care, and other industries as well. For example, in the Middle East, we're actually co-innovating with leading partners on secure private networks. This is for critical infrastructure like utilities, combining telco-grade reliability with unmatched cybersecurity expertise because these networks are critical networks that have to be secured. A new growth area is in defense, where the bar actually is a lot, lot higher. Here, trusted connectivity and defense-grade performance are simply non-negotiable.
Another reason why this sector is important for us, if you look back in time, some of the best innovations, such as the internet, actually came out in defense, the DARPANET. This is an important area where we can co-innovate and co-create new and emerging technologies. We've already delivered tactical wireless solutions in the US Marine Corps. Together with Telia and the Finnish Defense Forces, we successfully completed and conducted the world's first seamless 5G standalone slice handover between multiple countries in the live network. We are proud and honored to serve the armed forces. In short, across defense and industry, Nokia is building trusted networks that actually protect, connect, and enable the world's most vital infrastructure. The future of the AI economy runs on these networks, and we are actually engineering it together.
I want to close with actually sharing with you three core principles that will actually guide us how we engage with our customers. I got to tell you, top of the list is put the customer first. This is extremely, extremely important. What does that mean? That means that every action we take must start with how it delivers value to our customers, ensuring that they sit in every interaction at the center of everything we do and the decisions we make. This is something I will take personally as a responsibility as Chief Customer Officer. I know these customers well, and I will make sure that we actually deliver on our promise and actually a lot more. Number two, co-create. We've heard this word, and I think you'll continue to hear this word for the future. Our co-creation learning mindset is what truly, truly differentiates us.
As I shared with you earlier, when we team closely with our customers, we just move faster together. When we get it right with the lead customer, others follow, turning innovation into scale. Quite frankly, with EUR 4.5 billion invested annually in R&D, we have the muscle to solve real-world problems and deliver measurable business outcomes. Finally, I do not think I could get off the stage. My CFO would not allow me to do this is to make sure we deliver profitable growth together. This is really, really important. Here, we have to have laser focus on the markets and customers where we can jointly innovate, create value, and drive success on both sides. The operating model that Justin outlined really reflects these principles in action.
Now, we've engaged with customers as multiple business units in the past, but now the path forward is to actually make sure that we come together unified to the face of the customer as Team Nokia. By aligning around all of our business groups, the value is really, truly realized because when we focus on solution and needs of the customer and remove the complexity of business groups inside Nokia, that's when you win. That has to be our focus. Now, we are also going to expand our client executive program for strategic accounts. This is to make sure that we make it easier for our customers to access our best talent, leverage our R&D, and see a faster time to value. We live in the most exciting times in this industry that actually I've known across telco, AI cloud, mission-critical industries, and defense.
By putting customer first, co-creating for the future, and delivering profitable growth, we bring the actual AI supercycle to life. This is just not a concept, but a reality that we are building together with our customers every day. This is what really energizes me about the journey that we're on. I want to thank you, and I want to also make sure I pass it on to my friend who will take you deeper into giving you insights into network infrastructure. Thank you very much.
Great job. Thanks, Raghav. Thank you. Thanks, Raghav. I appreciate it. It's a super exciting time. The power of Nokia and its brand and that reach with that customer set, as well as that ability to co-collaborate and leverage the assets that the company has and the forward-looking work of Bell Labs as an ex-competitor to Nokia, boy, it feels good to have that on my side today. For those of you who do not know me, and it's good to see some familiar faces, my name is David Heard. I came into Nokia, back to Nokia, actually, through the acquisition of Infinera. I was the CEO at Infinera. Prior to that, a long history in the telecommunications world. Justin mentioned that he spent his early career in 2G wireless. I started in 1G wireless with Bell Labs. I know it's hard to believe that I'm the older one. When it comes to the network, I've been in the access network.
I actually left Bell Labs running the fixed network, so I have a very good idea of what's happened there. I'm super proud of the spot that we're in, developed switches and routers out in the field, and obviously, the recent optical experience. Bringing all these assets together is super exciting. It's also super interesting that being the ex-CEO of a company that got bought, usually at this point in time, I'm a cost synergy. I'm not up on stage talking about the future. I'm super happy that Justin has provided me the opportunity to show you that I'm a revenue synergy for the company going forward. I want to talk a little bit about why am I here, what excites me about the future. I've seen the move from analog cellular to digital cellular, from circuit to packet, moving into a cloudified network.
Those were all great growth trends. There were ebbs and flows. This supercycle that we're talking about is something very, very different. It is something that when we see the power of what this is putting together, Pallavi talked about that upspeed. I want you to lock in your head what happened with Uplink in that video call moving to an interactive AI call. When you look at network infrastructure, I don't just want you to think about the short-term prospects of growth, wonderful wins we're having, and great market position we have, but what happens in an AI inferencing model when that traffic shifts from 12 : 1 to 4: 1? When you look at kind of the four key fundamental areas, one, there is no lack of growth opportunity here. I'm not going to spend lots of time on this.
This is a EUR 60 billion TAM, SAM, excuse me, that we're going after. It is plenty of market. This is not, I've been times in my career where I'm around searching for market opportunities. This is not that time. There's plenty of market opportunity in front of us. Number two, when I'm looking at things that excite me, great growth opportunity is wonderful, but where are we starting from? This core network infrastructure fabric that Nokia has is number one or number two in every single segment that it's in. Look, I've been a smaller player in an industry. I've been number seven or number eight or number six or number five. That is a very, very difficult position to try to wander into the wild world of AI. I feel very good about those number one and number two positions.
I'm going to talk a bit more about that. The other thing, the other perspective, kind of the compare and contrast from being in a smaller company is having the financial ability as well as the technical ability to execute. When I look at our research and development investment, and Marco, it is investment, it is not spend, of $1.6 billion. I compare that to some of our competitors in the optical space, maybe in the access space. This is a multiple of what they're doing. Not only do I feel good about the number, but the people behind it and the roadmap competitiveness that this lays out. Not only do we have a terrific roadmap today that's winning, as Justin talked about earlier, but we're laying the foundational roadmaps that win even beyond this strategic period.
Lastly, and I want to, this is a careful listening moment, I think, for this audience, but it's great to have a big market. It's great to be number one or number two. It's wonderful to have the bulk of investment to be able to leap ahead of the competition. You want to make sure there's momentum in what you're doing. Momentum is a wonderful winning elixir. What you'll see is year-to-date in orders in that hyperscale space, we have now brought in $1.5 billion of orders. That's through six wins with hyperscalers, two in optical systems, two in IP, and two in a brand new billion-dollar-plus market for us in pluggables. We are beginning to scale that. Again, that's a year-to-date number that if you look at both Infinera and Nokia together and I a year ago, that is a 3x increase.
If you look at where Nokia was in part of the strategic acquisition, that's more like a 6x-7x increase. It's early days, but I like what I'm seeing in terms of the momentum in this business. It's about having the right market opportunity. It's the largest I've seen, the right team, the right technology to execute. Those customer engagements that Raghav talked about. Let me kind of break down network infrastructure for a second. It's made up of three foundational pieces. If you think about the fixed network as where all the residential and enterprise comes in, we are the clear number one in that position. Usually, when I'm at one of these conferences, I'm saying, "Well, we're number one or number two except for Huawei." In this case, we are number one in the world in fixed networks.
That is an incredibly strategic position. When you think about it in the future, it is not a super high-growth market today. When we look at what inferencing does to that piece of the network, it is an incredibly strategic position to be in. Coming in from that kind of catcher's mit of the network into the fabric of the network where the application meets the network, that is our IP Networks for routing and switching. That historically was a $2.6 billion market. We are number one in edge routing there. Again, I will tell you, being number one in edge routing, routing is much tougher than switching. We are now applying those assets, as I demonstrated in the last slide, into the hyperscale space. Optical Networks, again, about a $3.1 billion. These are all trailing 12-month figures where we're number one or number two in every segment.
Feel good saying that because I did start at number eight in the market, and that was no fun. Climbing up now, we actually are in a wonderful position to take advantage of the AI supercycle. Raghav did a very nice job talking about the market segmentation. I'm only going to make a couple of key points here. Certainly, nine out of the top 10 hyperscalers are there, but you're starting to see business wins and scaling those business wins. 1,500 of the top service providers, it's wonderful. We can get into places that I never could get into before. We have the wonderful assets of all of Network Infrastructure and all of Nokia to be able to pull together that value for our client base.
When you look at the mission-critical segment, this is a great segment I'm going to talk about where people need trusted technologies. Quite frankly, in many cases, they want to buy from somebody in network infrastructure that has all of the pieces. We are uniquely positioned to be able to do that. When you look at our growth rates and what I'm going to talk about in terms of growth rates, I want you to remember a couple of things. One, if you look at 2024 results of the companies, most of our business, 80% of our business was with those telco players. Again, they're going through lots of transition, but they are not growing at a super fast pace. You are seeing the hyperscalers that I'll talk about growing at a faster pace, but traditionally, 80% of our business was with those telcos.
By 2028, that number will be just under 60%. You are going to see a traversing of new growth. It is weighted average math. We are going to continue to grow with mission-critical, actually ahead of the growth rate, as well as service provider. The sheer growth, as Justin talked about, of the hyperscalers and cloud majors around the world, the people building the AI factories is just tremendous. Again, pulling forth a ton of growth. All of these segments are looking for the same value proposition. They want to drive down power per bit, cost per bit. They want to make sure that they can scale and scale with some level of elasticity, as well as drive a huge amount of agility and resiliency in their network. As Pallavi so eloquently said, the move from five nines to six nines, it is not easy.
When you get both inside a data center, there is no room for failure as we go. That's just a quick view of the overall Network Infrastructure business. Look, there's no doubt AI is growing in terms of traffic. One thing some people don't understand is just how important that fiber rollout is and how early it is. This is typically where I say, "My doctor says fiber's good for me, but this is getting ridiculous." And you kind of give the corny laugh. Two billion kilometers of fiber will be deployed over the next five years. You haven't seen anything yet. As the number one player in fiber, again, at the access, the person pulling that in in the network and then optical connecting it, that's a good position to be in. Infrastructure spend, again, $3 trillion-$4 trillion, there's no lack of market opportunity.
I was with one of the hyperscalers a couple of weeks ago. We were in a meeting. We were talking about CapEx. I kind of chuckled, said, "Where are you at in your CapEx deployment?" They talk about quarterly CapEx deployment. They were talking about, "Well, we got to deploy still this. Here's what we need. We're spending $28 billion a quarter." If you think about that in the old days, when I used, if you look at Verizon and AT&T and others, they were about a $20 billion CapEx for the entire year. This is an opportunity where speed matters. I love the size that we have in network infrastructure in Nokia. The whole point is making that move at a completely new clock speed. You make that move like a startup.
Boy, that's exactly what our hyperscale customers, our AI cloud customers, and even our telcos and mission-critical customers want to see. Back to understanding the growth rate, I mentioned that 80% of our historical business was with the telecommunication service providers. Again, the telcos growing at 1%-2%. Enterprise and mission-critical is about 7%-9%. We intend to grow ahead of that. Where you're going to see a lot of the growth is in the pink going forward. That 16% growth, remember for Nokia, this was a new market segment. It was part of the thesis of purchasing Infinera. A lot of people ask, "I can't be in Nokia Bell Labs.
I can't be in a technology company without showing a nice network diagram here. Traditionally, what I would talk about is the wide area network, which I've already talked about with you, which is, hey, how access connects into the edge of the network and metro, how that connects into a long-haul network with cables traversing this planet, 10,000 km underneath the sea. What's really changed for Nokia, the real change in this exposure to the AI market that you need to be aware of is what is happening inside these data center builds. They're not just data center builds. These are AI factories people are building. We play in the front end of the business. That is the traditional data center, which is still a significant portion of growth. That is the leaf spine top of rack architecture.
We have products that traverse in the space, again, using open merchant silicon on the switching side. I'll talk about two of those design wins that I talked about. One was in the front end. One was in the back end, which is where all the GPUs are clustered. All three of our products are applicable in the front end, including fixed. Fixed has an out-of-band solution to be able to manage the data center environment for resiliency purposes. In the scale-out and scale-across networks in the back end, again, we have products that go across this. I'll give you a case example of a win here. We have an IP switching scaling across networks, as well as the pluggable wins that we have are really for that interconnect into these AI gigafactories. I'll give you more of the technology details going forward.
Again, when I talk about that EUR 1.5 billion of orders to date, right, through that Q3 period, that's a big portion of where that spend is going. When I talk about that recent order momentum, again, to give you a little bit more detail, I gave you the detail on the six wins on the left side here in terms of order intake. This is all orders year-to-date 2025. Optical networks, it's growing quite nicely. I think fortifying the investment thesis for Infinera: 40% optical growth year-over-year, year-to-date. In the 800 G coherent pluggables, these are shipping. These little babies are shipping out there. I know you can, I'll pass this around, Simon. You'll be able to take a look at it. These little babies are shipping. We've won, again, two contracts for these that, again, are ramping or in the period of ramping.
I would expect to see that as revenue, more material revenue as we get into 2026 and continuing to grow and beyond. Lastly, again, our switching products, rock solid out in the wide area network, a little later getting inside the data center. Boy, is the quality of what we're doing, as well as the resiliency and the open nature of what we're doing, low power of what we're doing mattering. It's early innings, but year-over-year order pattern, 150% growth. I'm going to very quickly try to step through these. Can't tell I get a little excited about what I do. In the optical space, if you talk about what people buy, this also explains the growth rates. Last year, 100% of what we did was optical systems.
When you build optical systems, you are talking about selling in terms of the optical engines that drive them, tens of thousands, twenties of thousands, and thirties of thousands of units. This business, again, was 100% of where we went, where we've been. These applications, you'll always hear metro applications, long-haul, subsea where we're a leader, wholesalers or carriers' carrier. That is where, again, these optical systems are deployed. You interconnect data centers in between. You use systems in many cases to be able to do so. The two new businesses for us, one in the pink here on pluggables, that is a new business for us. When you are selling pluggables, both with hyperscalers, the annual needs are typically hundreds of thousands. That is an exponential gain from systems. In the component space, there is the ability for us to be able to sell components.
Now, this is very early. We're looking at designs. It's not in our revenue for 2026, but we are, because of our foundational technology, building small components, right, that power networks that lower the power inside of data centers in between GPUs. I know you can really see this, right? There is a chip in my hand to lower power 80% inside the data centers. When you're doing these, the chips that drive these that we build in our own fab, we're talking about fives to tens of millions of units annually. That is a whole new scale. The good news is we here at Nokia have our own vertical integration. We have our own fab.
Two and a half years ago, even before Nokia bought Infinera, we had the forethought to say, "We're going to need more capacity here." For those of you that are following it, there's a lot of people very interested in indium phosphide laser capacity. When you talk about 5 million- 10 million units interconnecting GPUs driving power down, that drives some pretty incredible demand. Two and a half years ago, we have one fab in Sunnyvale, California. We embarked on building out a second fab in San Jose to increase our capacity by 25x. Great news. These fabs, relax, Marco, relax. These are not billion-dollar fabs, two billion-dollar fabs, a couple of hundred million-dollar fabs. In addition, we had some U.S. CHIPS Act support. Really, really great to see the support of the U.S. government.
We then are able to package those photonically integrated circuits, which, by the way, we invented the photonically integrated circuit. We package them in transmit-receive optical assemblies. This is packaging, very intricate. This is an 800 G package done in Allentown, Pennsylvania. Very, very strategic. The last time we had a supply chain pinch, we were able to gain share. If they're, again, building out the right capacity and having that, super important. Lastly, on the DSP team, the DSPs are what drive kind of the Moore's Law of the optical network. Pallavi talked about having dollar per bit, power per bit. That's what we do. I think our biggest problem before is we had too small of a DSP team, so we could only spin new DSPs every so often.
The typical design cycle for a telco in the past would have been a new DSP every four years or so, five years would be fine. With the hyperscalers, they're looking for new DSPs for both pluggables as well as systems every two years. The great news about putting together the two largest coherent DSP teams in the world, and they're not just the largest, they're the best, that will continue to drive Moore's Law going forward for us. Again, to Justin's term, moving us at a very different clock speed. Very exciting. Back to the growth rate. You understand how I think about the weighted average math. Optical systems traditionally are growing at anywhere from 3% - 4%.
You're going to see our big growth both in the optical systems that we've now closed with hyperscalers because they're growing faster than the telcos, as well as in this coherent pluggable space. Again, full disclosure, we're in some qualification and testing, design testing. We've made test chips for this intra-data center space. Again, brand new SAM. This is not contemplated in our forward numbers. We have the capability. This is part of the beauty of Nokia coming together with Infinera. They have the resources that we've been able to apply to this brand new field. A nice market, great strong position, the right assets to win. This is my favorite part of the presentation. I've never done this in my career. I've been bought, and I get to grade the acquisition of Infinera. On the left were the things that Nokia put up when they bought us.
They said, "Hey, we want to create optical networks powerhouse. We don't want any disruption to customers. We don't want a lot of overlap." What have we found? We put together a unified roadmap. The customer response has been outstanding. Overall, how have the results been? I haven't seen disynergies. We've actually delivered more than we expected. How do you see that? The second piece kind of helps you there, which Nokia wanted to increase its presence in North America. Order presence there is, again, up 40% year-over-year in terms of orders. Increasing our presence in AI and cloud, again, I won't repeat this over. Good early. That number might seem big. To me, it's still small. There's plenty of forward opportunity. We are not getting yet our fair share, but those design wins are there.
As we begin to ramp and move on in our roadmap from 800 G to 1.6T to 3.2T, that is plenty of green space for us to be able to grow. Lastly, everybody wants to hear about the synergies. We are ahead of schedule, on track, and ahead of schedule to be able to deliver those synergies. We have plenty of investment areas, again, to be able to grow back to that revenue synergy rolling forward. I will prove that I am a revenue synergy. In terms of customer use cases, I will not go through this because I think I have gone through it nicely. I am not going to grade myself that nicely. I have gone through it. You can tell me whether it is nicely or not. Those pluggables were co-developed with two hyperscalers. They wanted to have a standard way to do probabilistic constellation shaping.
They wanted to be able to have certain software parameters on their pluggable. We deliver these pluggables can go kind of up to 1,700 km, so they can be used for all different applications depending on distance. Again, each win you do here is a couple of hundred million annually. The ability for us to get into components, this is not a win. This is a design example of using our component for 1.6 Tb inside the data center to lower the cost, to lower the power of the GPUs. Each deal like that could be, again, hundreds of millions annually. What Marco really likes about this is that when we're selling that, we're selling just the photonically integrated circuit.
In terms of inventories and ease, this is a great business, all accretive to the margins for the company, which helps us with that 300 basis points-700 basis points expansion in network infrastructure. In IP, I'm so excited because the next thing you plug a pluggable into is IP. On the access edge, this is a company that has the aggregation switching, the BNGs, number one, and cell site routing that, again, this platform is rock solid. Like in the optical domain where we have our own silicon, or in that case, indium phosphide, we have our own chipsets here too. Again, the FP series, the FPCX that helps power a huge amount of applications required as you get closer to the edge. Again, they tend to be very, very quality sensitive, very, very sensitive on the feature sets required. We then bring that into the IP edge and core.
These are the 7700, the 7200 series products. We have developed rock solid software and SROS that our customers just love. In addition, we have the network automation, the NSP to be able to make that happen. That is where we're bringing in AI tools to help people build their networks and be able to spin up digital twins to be able to implement with ease, with high reliability out in the network. What we've been able to do is translate that. We've taken our SROS software, we've opened it up, developed it on SR Linux, and made that applicable for the data center play, right? The most fresh software stack out in the industry, built on, again, decades of rock solid experience out in the network. We offer that as an open platform. We've also implemented our hardware with Sonic in an open environment.
When you get into the hyperscalers, they like to choose. Our goal is to build the world's best network elements that can be put together as a solution. The remainder of the platform down here, again, this tends to be there are some elements towards the edge where we're using merchant silicon as well. We use the right silicon, as Pallavi said, for the right scenario. We're not going to spend lots of money spinning up our own silicon if there's a merchant approach that makes economic sense. You put this all together, we have applications for DC gateway, front-end switching, and back-end switching. We also have events-driven automation for AI operations for this portion of the business as well. Super important as you get some of these neoclouds that are building out massive AI factories and need the ability to configure, might not have the technical expertise.
They can use normal language to be able to configure, troubleshoot a network. In this case, just as an example, a couple of weeks ago, last week, actually, the 7220 was announced. That is our 102.4 Tb solution, 1.6T, Tomahawk 6 platform, as well as I'm going to talk about our 18E design win that you're talking about a half a petabit of capability. Very similarly, when I look at the growth rates for this business, again, traditional IP routing and services, that is a 2-3% market. Look, it's a fundamentally sound market. There's areas we're going to continue to grow there. We're going to grow, again, back to that mission-critical segment that's growing 7%-9%. That has been a nice segment for us with healthcare and utilities.
You're going to see the growth, a big piece of the growth in the pink here, where we've again had design wins. Our goal here is you go land a couple of design wins, you land and expand. We did that this year. Our plan is we've aligned our roadmap. We've seen a nice pull-through impact with the teams from both Infinera, Nokia Optical, now working together with the IP team, and by the way, with the FIX team, to be able to bring real solutions inside the data center. From that weighted average math of the growth rate, that's what you're going to see. That scale across opportunity I talked about that we won, this is hundreds of millions of dollars of actual implementation as you win annually. This is the scale across, which is our 18E, 7250.
This is a half a petabit with 576 ports of 800 G connectivity. Who's got the pluggable? Yeah, go get another 575, and that completes a solution for this particular hyperscaler. On the neocloud side, some of these are a bit smaller, but again, a bit smaller being $50 million a year. This is where we can use our entire platform. See, I talk my own ear off. That's when you know you got to stop. The 7220, 7250, all of these platforms, they're using that plus EDA for these fast buildouts. Two case examples here of wins that we're seeing out in the marketplace. Lightspan, again, is somebody I competed against a long time ago, and they beat me. They were number one out in the space. They are still a strong number one out in the market.
They are implementing the OLTs and the ONTs out in the market to power the residential market, the enterprise market. You heard references from Justin and Raghav talking about players like AT&T growing from 30 million subscribers to 60. That's a big deal. There's plenty of white space open. You see we've covered 600 million subs, seven out of the top, seven out of ten in the U.S. are ours. This is a space we know very, very well and will be very, very important for inferencing rolling forward. We also have the automation to be able to make this work in the home. Has anybody ever had a problem in your home with your internet? 60% of the time, that problem is Wi-Fi. It has nothing to do with the fixed solution.
By us deploying Corteca, which is an in-home software, we're able to avert 60% of those calls. We're able to make call times go down by 50% and improve the net promoter score of that service experience by 20 points. This is just starting business for us, but it's a way to continue to add value to our clients and margin to the network. Managing the OLTs out there, super important for us. We're able to do digital twinning. Actually, think about it. It's a passive optical network. How do you test active gear? We're able to do that through a digital twin through Altiplano. There are a few new applications. I've been in the job for about five months. The team is really working hard off that strategic position, looking at optical LAN. We've announced our first product there very early.
The Otaban solution for data centers came by us pulling together with optical and IP and listening to our hyperscale customers. Again, a great business, a business that kind of the first thing we're doing is continuing our leadership position, especially on that OLT side and ONT side with fiber. The second thing is, again, we're driving the margins up through software and the rationalization that Justin talked about. One comment, I talked about my corny joke of fiber being good for you. Look, 50% of the networks are not even covered yet with fiber, so we are not in the late innings of the ballgame here. With coax, we're less than 10% of the way there in conversion. With optical LAN, the game hasn't started. I'm still getting a beer trying to find my seat.
There are lots of opportunities ahead, but right now, we're in this business as a very stable business. It's always been profitable. We've got work to do to be able to drive that next level of innovation, feel good with the early insights. This is all very wonderful. Big markets, great product categories, early wins, but having a clear plan. Justin talked about the importance of execution in driving that. We've got a clear plan for networks. It is going to be to drive that systems growth with both cloud and mission-critical. You'll see us ramp those pluggables, not just 800 G, but then the next generation of 1.6T. We will enter and scale the components market. It's still early to call the ball on when that will take an impact in our financials, right? I'm sure we're going to keep everybody plugged in as to that.
In IP, it's driving growth in that data center switching, land and expand, rinse and repeat. Land and expand, rinse and repeat. Accelerating that mission-critical business. That is really a great segment for us. The big thing is continuing to take routing share in the telcos as they position themselves for inferencing and AI. In fixed networks, it's about continuing to extend. This is a great team, right? In fixed networks, they've done an unbelievable job. This move to XGS-PON, they've got their own chipset as well, Quillion, that drives six nines of availability. They're ahead of the curve here. That is a chipset that if you want to upgrade 10 G to 25 G, I don't have to dispatch trucks to go do that. I can fast provision that given we own our own chipset, which is also effective for our margins in the OLT.
We are driving our own chipsets in some of the ONTs platforms. We're driving the automation. It's early days, but that will provide accretive value to us to expand our margins. Lastly, I mentioned there's a couple of really, really cool areas that you can leverage off this business, but it is very early days. With that, Marco will go a little bit more into that. Hopefully, I've explained that six to eight points of overall growth includes, again, a large degree of telco, a large degree of we're carrying over the historical systems, a much larger embedded base, fixed that's growing a bit slower. If you pull out fixed and just look at optical and IP, that growth rate is 10%-12%. You're going to see us driving, again, 300 basis points-700 basis points of margin expansion through the initiatives I've talked about.
Why am I excited? It is not just because I have six cups of coffee. We are well positioned in the market. Feels great. We have a clear strategy to win. There are opportunities to open up billion-dollar markets. Who does not want to do that? We have the scale and the vertical integration that, I do not know, seven years ago might not have seemed as important, but seems extremely relevant. As you want to increase your exposure to the AI supercycle, it is a great place to start. We have the business momentum with our strategic customers to make that happen. I hope I did not wear you out. Hope I gave you a little bit of a snippet for what I see in network infrastructure. I appreciate your time and look forward to your insightful questions coming up. With that, I'm going to bring it back over to my buddy David Mulholland.
Thanks, David. Thanks, David. Just one clarification. Everywhere David said dollars, he meant euros. We will not take a 30-minute break, so we'll see you back here to restart at about 10:40. For those in the room, you're welcome to try and catch up with David's six coffees outside, but we'll intend to restart at 10:40. Thank you. Welcome back, everyone. I hope you managed to get a drink or a coffee over the break, maybe not as many as David. We're now ready to resume the event, so let me hand over to Justin to talk about our mobile infrastructure business.
All right, thanks, David. Hopefully you got a sense of why we're so excited about NI. I also just want to touch on this point that David talked about. He's absolutely a revenue synergy. More than that, he's breathed a ton of energy into our NI team. I talked a little bit about the culture of bringing Infinera into our Network Infrastructure business, and there have been great cultural synergies and integration. The leadership energy and the pace that that team is moving at now is just so much faster and much more aligned to the market, exactly what we need at Nokia. Let me turn now and talk a little bit about mobile infrastructure. You know, I just touched on this point about making us faster, simpler, and more aligned to the market. That is exactly what mobile infrastructure will do. What I want to touch on on mobile infrastructure is the opportunity we see.
I want to break it down in terms of what it is, why it matters, and how we're positioning to win and create long-term value. As I touched on in my earlier comments, mobile infrastructure on January 1 will bring three connected areas. They are connected as our customers see them and as the market defines them. We'll bring in our core software business from Cloud and Network Services, our baseband and radio systems from Mobile Networks, what we call radio networks, and our technology standards business, formerly Nokia Tech. Bringing them together gives us one single integrated platform focused on mobile connectivity, from core to radio to standards. It's a platform with greater scale, better leverage, and now we'll have much sharper focus. I want to be clear, this isn't an internal reorganization exercise. This isn't a financial engineering exercise.
It's about simplicity, aligning our portfolio with how our customers see us today and with our vision for the future of connectivity. Let me break down a little bit of this for you. Core software leads in autonomous and programmable networks. Today, we're giving telcos a cloud-native core with zero-touch automation and API-based monetization. Over the past five years, this business has outgrown the market. It's become the clear leader in voice core and in subscriber data management while gaining market share in packet core. It generates about EUR 2.5 billion in sales with 49% gross margins. In radio networks, we provide the radios, the baseband, and the associated software, as well as service that powers the networks behind every smartphone and connected device. On a trailing 12-month basis, this business generated about EUR 7.5 billion in sales with 36% gross margin.
In technology standards, we manage one of the industry's strongest patent portfolios and most experienced and talented licensing teams. Our portfolio exceeds 26,000 patent families. You will hear a little more about this from Patrik in a few minutes. As you look at it, not only does it have a massive patent base, it has incredible duration. 70% of these patents have more than 10 years of life remaining. As we think about standards, it is easy to just focus on the IP licensing stream. It is also important to recognize that it is an enabler for standardization. As we invest in 6G standardization, this is a portfolio that will continue to grow and continue to extend its duration and value. Financially, this is a business that has an annual contracted run rate of EUR 1.4 billion and delivers an operating profit of EUR 1.1 billion.
This is a foundation of durable profit and cash flow, not just for MI, but also for Nokia. As we think about the market, this is a market that we play in that's largely stable. I think, as many of you know that cover this industry, it's largely cyclical. Telco investments have been steady. They're not expanding. As Raghav detailed out in his explanation about the customer base, in many returns for our customers, their returns are under pressure. We have a unique position in this market. Today, Nokia is one of only two scale Western vendors capable of delivering across the portfolio, across core software, across radio networks, and with a robust standards portfolio. In North America, Europe, India, and parts of Asia-Pacific, we're positioned to capture incremental share as governments and telcos prioritize trusted vendors and seek innovation partners.
At the same time, in markets that are open to both Western and Eastern suppliers, we're being disciplined. This is a shift. We're focused on winning share only with customers where innovation and value are recognized. I want to be clear, we have some great customers in these markets, and we'll continue to invest and grow with them. We're not going to chase volume for volume's sake. We need to be delivering an acceptable return on this business. This is a clear balance. It's a balance between growth and profitability. It reflects the discipline that we have with mobile infrastructure for managing it for sustainable returns. This is ultimately the foundation for reshaping MI for long-term value creation. Because when we look ahead, we see a very, very different market.
If you think about today's market, today's market is centered on what is largely consumer-based connectivity. Tomorrow's will be very different. Tomorrow's market will be defined by AI-enabled industrial and mission-critical applications that will come to depend even more on intelligent, secure, and programmable networks. As the AI supercycle accelerates, networks are evolving. We're no longer just connecting people and information. We're now connecting intelligence. As you think about this, if you look at what's happening today, today, almost every device is largely treated like a mobile phone. My phone is a phone. My watch is treated like a phone. My tablet is a phone. My connected vehicle is a phone, largely through pre- and postpaid subscriptions and largely providing a consistent revenue stream for our operators. In the future, it's going to be very different.
In the future, when we have machines, sensors, and distributed systems connected, the shift will be around trust, security, and performance. That is really where we're headed. Because we recognize that the networks of the future and ultimately our customer's success in the future won't depend just on delivering consistent revenue streams and dealing with homogenous devices. They're going to need to deal with heterogeneous services. That is a very different world. That is where we are positioning Nokia to lead. We're going to be leading where we can differentiate, but ultimately by delivering the innovation that we anticipate is needed for the transformation of AI-native networks. I think this really gets down to a different business model.
This is really critical because largely when you think about the model, the model David talked about starting in 1G, I intercepted us in 2G, but the model really hasn't changed much. If you think about the model and the journey we've been on in this industry, our customers are largely monetizing their value through ARPU, pre- and postpaid subscriptions. Our business is largely centered around hardware, a little bit of software, and a heavy services content. That has been the business. We still pack a lot of value into hardware today. Where we're headed is a very different world. In order to deliver in this AI-native 6G future, first of all, starting with our customers, the telcos, they're going to need to deliver different revenue streams. Revenue streams will not just be about subscription. They'll also be about tokens.
Because different platforms, different physical AI devices will require different services. As that happens, their networks will also need to evolve. What that means as a service provider is we're going to have to change how we innovate. Ultimately, that innovation can't depend on the pace of hardware upgrades. We need to be far more software-centric. This is a simple but critical fact in how the industry needs to evolve and where we see the opportunity for Nokia to lead it. In order to move more value to software, we have to commoditize hardware. Hardware needs to become simpler, and it cannot be the source of where all innovation resides. Rather, we need to leverage software and AI applications that can optimize and learn in real time and can be deployed at a far more dynamic pace.
If you think about this for us, this is a unique opportunity. This is an opportunity not only to innovate, but to deliver a higher quality business, one that's software-driven, one that delivers higher gross margins, allows us to deliver a faster pace of innovation, and ultimately deliver an increased share of recurring revenue. This will take a bit of time because, as we all know, this is not an industry that transitions overnight. It is a direction the industry has to go. We've already started to build it. By the way, if you just think back, I'll go back to the story of the Internet Supercycle. Think about the dot-com bubble before and after. In the late 1990s, the kind of hardware in IT houses, specifically around compute, were vertically integrated stacks: Sun Microsystems, Digital Equipment Corporation. In fact, they were the early winners.
You look at their revenue and growth profile. When you get into the 2000s, we moved into industry-standard servers, general-purpose silicon. We moved into areas like virtualization, optimizing the use of our hardware so we could build more and more applications, leverage more of the hardware utilization. Over time, we moved into containers and microservices. That journey is played out over multiple different technologies. It is quite predictable. In fact, think about the storage industry as another analogy. It is an industry that went from appliances to virtualized software with a lot of general-purpose build hardware underneath it. This is the journey that we are on. This is why we recognize and we have already started to build our platforms towards this. In fact, I have to give a lot of credit to the team in core software and ultimately to Raghav.
Because under his leadership, we began this transformation five years ago in core. As many of you know who have been in this industry, core used to be big iron. When David and I were talking about our time in the early days of the industry, switches, base station controllers, these were big iron systems. Raghav and his team recognized we needed to shift. We needed to shift to being cloud-native. We moved from the appliance-based systems and platforms like network function NFV or network function virtualization to committing to fully cloud-native software. The reality as we stand here five years later is this transition is largely complete. It is paying off. We actually radically simplified our voice core. It became the number one platform globally, as I said earlier. Not just that. We migrated our packet core to a true cloud-native architecture.
We have gained four points of market share. You might look at it and say, "Gosh, voice is a legacy application. Where is the innovation?" We are seeing that already. You heard a little bit of that from Pallavi and from Raghav in their talks. We are already seeing innovation in this area. We have demonstrations of immersive audio. Think of the experience you get today. Sometimes if you are listening to, you are watching a video at home. We can actually deliver immersive audio for a phone call, for a video call, leveraging some of this technology. Real-time translation, which if you are an American like me living in Finland, is probably the best solution to learning Finnish. Because Duolingo, while I am working on it, is not helping me move fast enough. Real-time translation, another solution. The ability to identify and block robocalls.
There's a whole bunch of services that will have value and tremendous duration for voice that we could not have innovated in if we weren't built in a cloud-native stack. Beyond that, packet core is really a step towards this AI-native future. Actually, we've leaned in. We've embraced multi-cloud openness, multi-vendor interoperability. I touched on the Red Hat example in my earlier comments. This is giving us deployment freedom and no supplier lock-in. Raghav touched on some of the examples of where we're lined up with cloud providers. Effectively, what we're seeing is the industry shifting. Our customers are unifying IT and OT environments through this cloud-native stack. When you look at our leadership, they were on the prior slide. The results are clear. We've won over 125 5G standalone contracts and over 25 core-as-a-service customers running on Google Cloud, AWS, and Microsoft Azure.
This business has grown well above the market rate. Operating margins have more than doubled since 2022. We're not stopping there. In fact, we see a future where the network fabric needs to become truly autonomous. Because delivering the services in the future cannot rely on human intervention. That means it needs to be powered by agentic AI and be exposed through open APIs that allow intent-based operations to be deployed at the speed of the devices needing them, and new monetization models to emerge, giving operators the opportunity to participate in this value stream. Actually, if you look at the core market, what you see is largely a flat market in core networks, but in these areas of autonomous networks and APIs, growth.
One of the areas that we've been growing in is a place that we recognized early on, our API network as a code platform. This was a place we invested with a principle around open source. Open source is a principle of the technology industry. We all know that open source enables more developers to get access to tools, to build capabilities, and create value in their networks. We're seeing that already. We have more than 35 customers adopted globally. Ultimately, the other message here is differentiation matters more than scale. It's very clear. Our differentiation is our software stack. Our stack gives us a competitive edge. It's zero-touch. It's intent-based. It's critical to emphasize it's not something we built just in a lab. It's something we've co-created with our lead customers. It's AI-native, and it's 6G ready. It supports ecosystems beyond 3GPP.
This open programmability, the APIs I talked about earlier, allow telcos to expose and monetize network capabilities across industries. The architecture extends across the network from core to edge, providing a unified fabric for automation, observability, and intelligence. As we look ahead to what's coming, AI-native networks and 6G, we're accelerating our development, evolving this platform for agentic capabilities, extending it to the RAN and to the cloud so we can deliver the performance, security, and reliability that intelligent tokens will demand from the device itself all the way back to the host. Let me turn and talk a little bit about radio networks. This is a place where we've had challenges. First of all, it's very clear that this is a business that has not delivered acceptable returns. It's ours. We own it. We know we need to fix it.
It starts with portfolio focus and ultimately emphasizing technology differentiation. What I want to make sure I emphasize here is we've actually made quite a bit of progress. Over the last five years, the team that's worked on this, led by Tommi Uitto, has made great progress. They're focused on core product competitiveness and, as importantly, quality and stability. When you look at our results, our latest baseband platforms deliver up to 90% lower energy consumption compared to the previous generation. Our Habrok Massive MIMO radios reduce power by 30%-40%. As critically, we've been investing in our RAN software stack. We recognized that software and hardware need to become distinct in this future. We've been investing heavily on our AnyRAN software.
The reality today is that our AnyRAN software can run both on our native hardware, AirScale, and on cloud-native stacks that run on common off-the-shelf servers called COT servers called Cloud RAN. These are the servers you can buy from partners like Dell Technologies, who participated in our announcement a few weeks ago. We've also embedded machine learning across the portfolio. We talk about AI a lot. In fact, this is something that we worked on for a number of years. We started with machine learning because we had the technology in the purpose-built silicon. We continued to invest in enabling that technology. Machine learning actually delivers some of those performance results that I talked about.
For example, machine learning enables 10% higher downlink throughput, 30% higher cell utilization, and 90% faster issue detection and repair, meaning that our customers' networks are more available for their customers. This was an early application of AI, as I mentioned, and Nokia was a leader. We have not stopped there. We talked a lot about open and core software. We have also been committed to a foundational principle of open interfaces in the radio network. In fact, we have been a leader in what the industry calls open RAN or ORAN. Customers like NTT Docomo and Deutsche Telekom have recognized our leadership in open, high-performance RAN. In fact, Deutsche Telekom recently commented publicly on the successful trials we have done with Fujitsu around an open RAN solution in their networks. Ultimately, while open is important, performance is still what matters and still the key differentiator.
We've deepened key partnerships in this space. We've extended our collaboration with T-Mobile U.S., which we announced earlier this year. We've added new wins such as the one I touched on at Vodafone 3 in the U.K. Earlier this week, we just announced a deal with TIM in Italy to deliver their 5G network. In fact, when you look at the 20 of the world's fastest 5G networks, Nokia is a RAN supplier for 15 of those 20. We have a strong and solid base to build from. In that context, we're building on that base in a prudent manner that recognizes the reality of the market. Because while we see that the near-term forecast is flat, we do believe that the transition to AI-native 6G networks presents a tremendous long-term opportunity for growth. Why?
Because if you just look at the market forecasts, they're showing a 14% CAGR over the next 10 years. Remember earlier, I said our own Nokia Bell Labs research had identified that that traffic would probably grow at 20% a year. We believe it's actually higher than this. We recognize that even that will drive investment. Just think for a minute that for all of the work that's been done in 5G, we're only 35% penetrated in the market globally. That represents a significant opportunity even in this flat market. What this does is create structural tailwinds for investment. It also underpins the focus of our roadmap, focusing not on being everywhere, but innovating where we can deliver differentiation, moving from 5G advanced to AI-RAN to AI-native 6G. That's the opportunity we're zeroed in on.
This is where it's critical to understand a little bit more about our partnership with NVIDIA. NVIDIA's capabilities are central to our shared vision of AI-native 6G. As we announced last month, we talked a lot about the partnership around bringing our AnyRAN software to combine it with NVIDIA's accelerated computing stack. I just touched on earlier that we've been investing in the portability of the AnyRAN software. Those investments have positioned us to be able to deliver this capability. That's why if you attended GTC a few weeks ago when we made this announcement, you could actually see our AnyRAN software running on a Grace Hopper platform. You hear that T-Mobile U.S. has already made a successful call using this stack. What you may or may not understand about NVIDIA's stack is that NVIDIA's stack is completely portable.
Once you build on top of the CUDA platform, I'm not just locked into a Grace Hopper platform. I can deploy that over any one of their GPUs, to L4S, to their graphics cards, even to Jetson the robotics platform. This is really important because it means that all of the development we've done and everything we're continuing to do will port seamlessly to NVIDIA's Aerial RAN Computer Pro platform, or Arc Pro, that they announced a few weeks ago. What's even as important is getting onto this platform means that you're not locked in and dependent merely on hardware upgrades for enhancements. For those of you that understand their stack, CUDA continues to unlock more performance. If you look at the performance, for example, of Grace Hopper or Grace Blackwell when they were released versus what they're doing now in the field, they continue to improve.
This is just in traditional AI training or inferencing applications because the software continues to unlock value in the hardware. That is going to be very similar for us in the RAN. The ability to unlock value through their software will be one way that we unlock value even after a customer has purchased the hardware. The second thing is we will continue to be able to build value on top. Pallavi talked about this a little bit in her comments, but the ability for us to use models in this AI-native stack to continue to learn and improve on the software. What that fundamentally means is there is no longer this lock between the hardware that is running the baseband software and the software that runs on top of it. I am no longer dependent on making a hardware upgrade to get better features and better performance.
This is a tremendous shift for our industry and one that we are excited to be leading. The last thing I'll say about this is there's a lot of talk, and I understand because of the history of AI at the edge, the potential for edge inferencing, mobile edge compute. There's been a lot of talk about how you monetize the GPU. That's not our focus with NVIDIA to start. We absolutely believe there's value in those applications. We've been co-innovating and co-developing with them and other ecosystem partners, exploring those, actually linking those back to our Network as Code platform that I touched on in core software as we bring this together. The fundamental principle of this hardware platform is to make sure that the hardware card actually functions as a pure baseband processor.
We mean that in terms of performance per watt, which is essential for our customers because if the performance and the power efficiency isn't there, it doesn't matter how compelling the application is. Not just at a performance level, but a performance per watt level. The other thing is around the ROI because we recognize the investment needs to be something that can be available to them on hardware. This is the shift. Finally, the last part of the announcement we made that's incredibly significant is this Arc Pro hardware won't just go into new devices. It will also be an upgrade option for our Nokia AirScale baseband platform. The reason that's important for customers is many customers have told us the investments they have in AirScale, they would like to see through the AI-native transition and ultimately into early 6G.
What that means is that the million units of AirScale that we have deployed globally now have an option to being AI-native platforms, which means that our existing customers can start deploying AI-native networks as soon as we move into production in 2027. This is why we had T-Mobile join us in the announcement, and they continue to collaborate and innovate with us as the clear performance and innovation leader in the U.S., and why we've also had interest from SoftBank and from IOH in Indonesia because they recognize the potential and the power of this for their networks and ultimately for their customers. Ultimately, what this means for us is we're moving away from a legacy hardware model to one that is built much more around software and building a software-driven business model. What may not be apparent is that this also frees capital.
This means we can shift investment into software and ultimately deliver differentiation and value where it matters. This is the shift from proprietary hardware to general-purpose hardware. In that context, we have a set of priorities for this business that are very clear. Number one, we want sharper commercial focus. We've been chasing business at times. Sometimes our service offerings have not necessarily yielded returns for our customers. We're going to be very disciplined around share capture. Ultimately, we're going to compete where we can innovate with our customers, deliver value to them that they recognize, and capture value by enabling them to do more for their customers. We're also reallocating capital with discipline, allocating capital to building a different model, one that is much more software-driven in value capture, focusing on where we can generate higher returns and reducing exposure where returns aren't acceptable.
In parallel to all of this and consistent with what you heard from David and what you'll hear from Marco shortly, is we're focused on driving better operating leverage. As we grow in this business, we'll continue to get scale efficiencies. That means simplifying how we work, consolidating overlaps, and embedding AI-enabled productivity across the portfolio. This, over time, means we'll have a more focused, more disciplined, and more profitable portfolio with higher margins, stronger recurring revenue profile, and sustainable returns. Now, those of you that know and have been through transitions to recurring revenue models, you know that it takes a bit of time to see those returns. That is why you see the strategic KPIs as being very balanced here. We're being disciplined around gross margin. That's our North Star for value capture. We're being disciplined around the base of operating profit.
Obviously, many of you in here know how to do math. If you take the $1.1 billion that I talked about in IP licensing and you think about the $2.5 billion at 15 percentage points of operating margin, we have a pretty strong base to start from here. The reason we have such a strong base is actually technology standards. I think what gets lost a little bit in our technology standards business is how durable the revenue stream and the profit stream are because of the essential patents that we have. As I said earlier, it's not just about the revenue and profit stream. It's about enabling interoperability. It's about enabling any device anywhere to connect to every network securely and efficiently.
This is actually really important because as you think about AI expanding and taking on greater significance, greater significance economically, greater significance geopolitically, this standard foundation becomes even more important. It enables innovation to scale globally. It reinforces Western technology leadership, Western technology leadership in connectivity, Western technology leadership in networks, and ultimately Western technology leadership in AI. In closing, mobile infrastructure is a business we're transforming. It's built on unique and differentiated assets. We've positioned it for stability in the near term and ultimately growth in the long term. We believe it has the potential to deliver solid returns today and outsized returns as the industry moves into the AI-native and 6G era. Simply put, our focus right now is not to build a bigger business in the short term. It's to build a better, more valuable business for the long term.
With that, I'd like to have a little bit more time so you can understand with greater depth the value of technology standards and why we think it's such a unique asset. I'd like to ask Patrik to come up and talk to you a little bit more about what we do in tech standards. Patrik.
Thank you, Justin. I've been with Nokia for 20 years and with the Nokia Technologies Unit since it started operating. I can't remember when we would have been in a stronger position than we are today. I'm going to talk about our industry-leading activities and achievements in standards, patents, and licensing. The work is crucial for Nokia's future technology leadership and also financial performance. I have four topics. First, the fact that our licensing business is based on Nokia's foundational technologies.
Second, our smartphone renewals ensure predictable long-term cash flow for our investors. Third, we are diversifying our revenue pools with success in our expansion areas. Fourth, we're investing in our portfolio to future-proof our business. Nokia's patent licensing business is based on the virtuous cycle of IPR. We conduct groundbreaking research in solving the most complex technical problems with fundamental inventions. We then contribute these inventions to open standards so other companies can build on our inventions without making the original investment. The agreed compensation for the innovation are the patent royalties we receive from our licensing activities of our standard essential patents. These royalties then fund our research of future key technologies, which we then again license, and so the cycle continues. Looking at our business, we have strong execution and momentum in all of our licensing programs.
In February last year, we completed our smartphone renewal cycle, locking in billions of EUR in contracted revenue for years to come. This required closing seven major smartphone deals in 13 months. Since completing the renewals, we have focused on the remaining addressable smartphone market. We signed an agreement with Transsion, the market leader in Africa, at the end of last year and have signed additional deals this year. As a result, we now have virtually all of the global smartphone market licensed for our cellular technologies, something I believe no other patent licensor has at the moment. Turning to our expansion programs, automotive, consumer electronics, IoT devices, and multimedia services, we are ahead of our peers in all of these programs. In automotive, we have 4G and 5G licenses with almost every Western car company.
Many of these agreements were secured via the Avanci Automotive Licensing Pool, of which Nokia is a leading member. In addition, we signed the industry-first bilateral agreements with Chinese automakers last year. This year, we have made further progress and now have four Chinese automakers under license. In automotive, our Wi-Fi technologies are also relevant, as these enable local connectivity inside the car. We now have bilateral license agreements with seven major automakers covering the use of Wi-Fi technologies in their vehicles. The most recent was signed last week with Mercedes. In consumer electronics, we primarily license the use of our video and Wi-Fi technologies in devices such as tablets, laptops, and connected TVs. This is also an important pillar of our expansion areas, and we have market-leading coverage here too.
Over the past year, we have signed agreements covering the use of our technologies in HP's, Amazon's, Samsung's, Casio's, and GoPro's products, just to name a few. We also have strong momentum in IoT devices. This is a fragmented market where we earlier simplified our go-to-market approach to vertical by vertical. This ensures we focus on one vertical before moving on to the next one. In IoT, point-of-sale payment terminals have been our first focus area. We have agreements with all major Western vendors, and last year, we signed an industry-first agreement with a Chinese point-of-sale vendor. Since then, we have signed agreements with three more Chinese vendors. Again, we're ahead of the rest of the market here. Going forward, we will increase our total addressable market in IoT through controlled expansion into new verticals. Multimedia services is, of course, also a meaningful opportunity.
Nokia's inventors have contributed to the development of all market-adapted video codecs. Fast-forwarding or rewinding a video by scrolling through it while simultaneously displaying the current scene is just one example of an invention by Nokia that many of us are likely using. Our work in this area has secured patents, but it is also recognized otherwise. For example, we just received our sixth technology engineering Emmy a few weeks ago. We are leading our peers in unlocking the licensing market. We signed our first agreement with a streaming company in the summer of last year, and since then, we have signed five more deals. The most recent was signed last month with Starz, a U.S. pay-TV company. The successful smartphone renewals and the momentum we have established in our expansion areas means we are well-positioned for near-term stability and longer-term growth.
We have more than 800 million of annual contracted recurring revenue locked in each year all the way through 2030. The current annual revenue run rate from our expansion areas is over EUR 200 million. As you can see from the chart, there are further opportunities across all of the programs. Turning to our patent portfolio, Nokia's patent license business is, of course, based upon our industry-leading portfolio. The strong portfolio is why we have been able to renew most of our agreements without litigation. It is also why we have been able to sign industry-first agreements in automotive, IoT, and in video streaming. We now have over 26,000 patent families in our portfolio, up from around 20,000 five years ago, and well over 7,000 families declared as essential to 5G.
When it comes to the longevity of the patents, as Justin said, we are in a strong position. The vast majority of the current portfolio has more than 10 years of life left. While we do follow quantity, we're, of course, obsessed with quality. One example of the strength of our patents is the multiple court rulings in our favor. Here, our patents have been found to be valid and infringed by independent courts. For each new generation of standardized technology, there is a time-bound golden window when the standard is defined. We're currently in the golden window for 6G, for Wi-Fi, and for video compression codecs. We're making investments in all of these technologies and have updated our accounting for 6G patenting to better align the costs with the 6G revenues. We're also making strategic acquisitions to strengthen our portfolio in certain areas.
As you know, we pay patent office annual fees for each of our patents. When the size of our portfolio increases, these costs also rise. To offset this, we routinely trim patents that no longer generate value. We also take pride in operational excellence and automation to streamline our activities where we're possible to manage cost. In summary, we have a strong record in maximizing value from patents and a strong foundation to continue to do so going forward. We currently have contracted recurring revenue of EUR 1.4 billion. Some agreements will, of course, come up for renewal over time, but we have over EUR 800 million of annual contracted recurring revenue until 2030, giving us predictable long-term cash flow. With the world-leading team that Justin also referred to, we would expect to renew the agreements that are up for renewal between now and then.
We also have strong momentum in our expansion areas thanks to a series of industry-first agreements and with plenty of opportunities ahead. This strengthens our revenue mix. To future-proof our business and further strengthen our industry-leading portfolio, we are investing in next-generation technologies, 6G video codecs, and Wi-Fi. I personally feel there has never been a better time to be part of our technology standards unit or to be part of Nokia. Thank you. With that, over to Marco.
Thank you. Hello from my side as well. It is good to see so many familiar faces here. I have been with the company the past five years, and I must say that I am extremely excited to see where we are today. What is extremely exciting is the opportunities that we see. Perhaps you wonder what is different now. Just giving you a few examples. First of all, clock speed is totally different. We have market opportunities. We are exposed to new segments and technologies as well. AI and cloud is a good example of this. We have much better product competitiveness today, and we are disciplined in investments with focus on growth and margin expansion. We have been very clear now with what is core, what is non-core. We are driving M&A and strategic partnerships to create value. We are clear on operating leverage plans.
Today, I will go through the following areas in my presentation. I start with the long-term targets and our financial framework with KPIs. I will go to balance sheet position, capital discipline, and capital allocation. I summarize in the end how we are driving value throughout the strategy. We have set a clear direction for the business going forward to create long-term shareholder value. There are three principles here. One is delivering profit expansion, position Nokia for long-term growth, and maintaining a disciplined approach to capital allocation. Today, we are introducing a new operating profit target to show our ambition over the next three years. We now target operating profit of between EUR 2.7 billion and EUR 3.2 billion by 2028. This is an increase from the EUR 2 billion we have had the past 12 months.
This means a double-digit CAGR on the operating profit side. To improve our financial performance and deliver sustainable returns for our shareholders, we will grow and drive operating leverage in Network Infrastructure, grow profit in Mobile Infrastructure, and at the group level, drive efficiency and capital discipline. We are also introducing a set of strategic KPIs that illustrate how we translate the strategic principles into financial focus areas. Let me now walk through each KPI and explain the underlying initiatives to reach our targets. As Justin explained this morning, our priority in NI is to accelerate our growth. I do not think that any of you missed how excited David was about the opportunities in NI. The first strategic KPI in NI is 6%-8% net sales growth through 2028.
We target to grow slightly faster than the addressable market in IP and optical, but we'll take a disciplined approach to fixed networks, focusing on higher value segments. With the actions we are taking, we expect to have higher growth opportunities beyond 2028, thanks to the better product mix. The second strategic KPI for NI is improving its operating margin to 13%-17%. There are four main drivers that will support the operating margin expansion. The first is capturing the synergies from the Infinera acquisition. You heard David also said that we are well on track, growing the higher margin products and the disciplined approach on fixed networks business mix. The fourth, operating leverage, means that our OpEx growth is lower than the net sales growth.
This margin expansion might be limited during the first half of 2026, as we will be ramping up new products. Over time, we are confident there is meaningful room to expand our profitability. Turning now to mobile networks. The focus in mobile infrastructure. More focus in MI will be increasing our value capture going forward. Our first KPI here is to increase our gross margin from the current 48% to 48%-50% levels by 2028. There are three elements that will support this. The first one is that we will benefit from our cloud-native platform in core software and expect to have higher growth than the market with improving gross margins. The second, we will take disciplined actions and approach the radio networks, focus on profitability over top-line growth.
We will continue to strengthen our technology position and differentiation to drive better gross margins. The third one is that we see a stable profit contribution from our licensing business in the coming years, just like you heard from Patrik recently. The second KPI is that we gradually expand our operating profit in MI from the base of $1.5 billion, which we delivered over the past 12 months. In addition to the levers that improve gross margin, we will have a disciplined mindset towards investments. We will invest to accelerate our AI-RAN competitiveness, but also focus on efficiency and maintain a similar OpEx base. We are not providing an explicit guidance and target for the net sales, as we, for example, make decisions for the future about the investments in AI-RAN.
This could result in a more focused and more profitable player as we move from a hardware-centric to a more software-focused business in radio networks. This will be a gradual transition with a largely stable market outlook, but we expect a similar trend for the net sales in the coming years for MI. We are also taking steps to make our corporate center operations more efficient. We started at a cost level of EUR 370 million and reduced that now to about EUR 350 million. We will save about EUR 50 million in 2026. By 2028, we target corporate center costs at about EUR 150 million. To accelerate this, we are allocating about EUR 120 million of costs in the segments effective January 1, as these are largely operating-related costs.
However, I want to be clear here that we are tasking the segments to reduce the G&A expenses by this amount and deliver an additional $30 million reduction at the group common by 2028. As you can see on this slide, we are now targeting $1.2 billion in gross cost savings between 2023 and 2026. This is an increase from our earlier target of $1 billion, reflecting our commitment to accelerate the efficiency and value creation across the group. The cost restructuring or the cost for the restructuring program is planned to be about $1.2 billion, as we've said earlier, that usually the cost is the same as the annual savings. To date, we have already achieved about $800 million in savings, demonstrating a strong momentum and discipline in our approach. These savings are being delivered in part through a reduction of our workforce.
At the start of the program, we had 84,000 employees, and now we target to have or we target to reduce that by 14,000 in total, of which 9,000 have already been reduced by September this year. That additional 5,000 headcount reduction, half of that already has been communicated. Just a note here that these data points exclude now Infinera and ASN divestment, as these were unrelated to this program. To improve the operating leverage is one of my absolute key focus areas. Now, when I've taken over the IT organization, we can drive further efficiencies by AI implementations and process simplifications. We are also implementing a continuous improvement culture in the company throughout the whole group. In summary, we are moving decisively to reshape our cost base, ensuring Nokia is leaner, more agile, and better positioned to capture the value.
We will continue to update on the progress on our cost savings programs. In recent years, we have seen some volatility in conversion of operating profit into cash, which has been mainly from three reasons. First of all, we had some significant project ramp-ups that consumed and then released working capital, and the India ramp-up is a very good example here. The second, that we had supply disruptions, as you remember, in 2021, 2022. Of course, the third one is that we have some timing differences related to prepayments in our licensing business. Occasionally, we get questions from you guys and others about our use of some off-balance sheet items, like sale of residuals. I want to be very clear here now that within Nokia, we do operate with a very modest level of sale of residuals.
Currently, the level is the lowest since we acquired Alcatel Lucent. It is mainly used to mitigate risks, like FX risk, country risk, and credit risks. Additionally, we aim to recover the cost from our customers. We disclose also the net cost for Nokia in our annual report. The impact to the balance sheet, where the cost is not being covered by the customers, is currently about EUR 500 million. We expect to keep that level, considering the geographical mix that we have in our sales. Going forward, we expect to move to a more stable framework for cash generation in the business. We now target free cash flow conversion from comparable operating profit in the range of 65%-75%. There can be some variations year- by- year due to the customer payment dynamics.
In the next couple of years, we will have higher CapEx as we invest in additional Fab capacity in optical, which we see as a significant value-creating asset for the company. We're investing over $100 million in the optical Fab capacity, as David mentioned earlier in the presentations. The main deltas that we have between operating profit and free cash flow are CapEx and tax items. Restructuring costs will fall after this program that we currently have. What comes to our balance sheet, we have a very strong balance sheet. We had, by the end of quarter three, about $3 billion as a net cash.
I would say that the two items that I want to bring up that will impact our net cash position, the first one is that, as you know, we have exercised the call option in our Chinese joint venture, Nokia Shanghai Bell, and acquiring the remaining 50% of the shares. We have estimated that value in a balance sheet to be around $500,000,000. We expect the deal to conclude in the next quarters or so. This will negatively impact our net cash position with the same amount. Why we're doing this? This deal will give us much better ability to simplify our operations and operational structure in this region. The second item is that we have now received the funds from the NVIDIA investments. Going over to the portfolio businesses.
This was based on the detailed strategic review that we decided to move the number of businesses into new unit. The review focused on our market positions, the margin profile of each of the units, and their strategic fit. These are good businesses, just like Justin mentioned earlier, but these are not deemed to be core to Nokia. Thus, we are not the right owner. Over the past 12 months, these businesses have had net sales of about EUR 0.9 billion and an operating loss of EUR 0.1 billion. We are assessing the best path for each of these businesses and owners as well and expect that we will have a solution by the end of 2026. When we complete this transition, it means that we will be much more focused on our core businesses.
What comes to our capital allocation priorities, those remain unchanged. The first priority is organic investments in R&D and other investments where we can create value, just that Fab investment that I mentioned earlier. The second is that we remain open to acquisitions that could accelerate our strategy execution. Here, we are looking at both minority investments, like we did with NScale, and potential bolt-on acquisitions. The main thing is that in all M&A, it is extremely important that we remain disciplined around the strategic fit and financial rationale. All M&A activities must create clear shareholder value, just like Infinera we believe is a good example of that. The third is dividend. We target recurring, stable, and over time growing dividend that takes into account the financial position and the business outlook.
Finally, the fourth one is that if we deem that we have excess cash, we continue to consider share buybacks as an approach to return that cash to shareholders. Justin talked about these five strategic priorities earlier. The strategic KPIs that I've talked through now are the financial outcomes that we expect from these. Let me leave you with three key points here. The first one is that we are focused on delivering growth and operating leverage in network infrastructure. The second, that we will focus on increasing our value capture in our mobile infrastructure business, underpinned by our technology standards and core software, while we transform the mobile networks. The third one, we will continue to execute with discipline and actively manage our capital going forward. Ultimately, we believe this will all lead to a double-digit operating profit expansion for the business through 2028. I hope you will join us on this journey. Now I will welcome David back on stage so we can start the Q&A.
Thanks, Marco. Thank you to all of the previous presenters that we had. We will now move to the Q&A session. During the Q&A, I would kindly ask that you limit yourself initially to one question and a brief follow-up. If you have further ups, we can, of course, come back around to you after we've gone around a bit of the room. We will today be taking questions only from the room. If you'd like to ask a question, once everybody's back on stage, please do raise your hand, and one of my colleagues will make their way around to you with a microphone.
When you get the microphone, please introduce your name, the company you work for, and then state your questions. With that, let me welcome the presenters back onto the stage, and we'll get started. There you go. Oh, thank you. I think we'll take our first question from Simon.
Thank you very much, Simon Leopold with Raymond James. I'll ask my two together, so we don't have to go back and forth. The first one's, I think, a broad topic that really didn't get addressed today. That's in the past, we've talked about these opportunities of Huawei swaps. Now, there's been more recent news in Europe about potentially pushing out some of the high-risk vendors. It wasn't addressed. I understand we want to be conservative, but I'd like to hear your thoughts about that particular opportunity, how you're thinking about it, how you'd size it.
The next question, probably more oriented towards David Heard, is we've heard more about this newer architecture of scale across, which I think some people may confuse with data center interconnect, which has existed for many, many years. Could you discuss how you see these opportunities in scale across? Is this factored into the forecast? How do you size it? What are you thinking in that particular application? Thank you.
Yep. I'll take the first one. David, I'll let you have fun with scale across. First of all, as you look at it, we put up a geo map, Simon, for a reason, which is we think the markets, it's pretty clear if you look at the direction of travel, that the markets are going to continue to bifurcate.
There's going to be markets that only permit Western vendors, markets that strongly prefer Eastern vendors, and markets that have a clear strategy around having both. If you look at Europe, I mean, right now, Core Networks has largely executed. Europe has already implemented the 5G toolbox for Core Networks. Look, we're optimistic that the commission is making progress on this. The reality is, for those of you that know the history in the industry, this has been talked about for a long time, and very little has been implemented.
If you think about the TAM, without giving you a hard number, if you were to look at the opportunity today, I think you'd take something around the EUR 2 billion-EUR 2.5 billion of opportunity right now if there was a radical, if there was a more aggressive replacement, an accelerated replacement in the market, just based on Huawei's share and the size of the market. That's a little bit how we think about it. The key thing for me is it's really easy for us to execute on the capacity if the demand comes in, but I'm not going to run the business expecting it to happen. Now, do we talk about it with the European Commission and governments, of course?
Our advice to them also is, look, this is something you need, if you commit to, you've got to actually go fund, you've got to go fund. When we have this conversation, if you're going to go enforce the EU toolbox, it's got to come with capital. Because if I think about it from my customer's perspective, there's no justification to upgrade a network in terms of their investor base without some kind of capital return. Unless that capital is offset, why would I have an incentive to replace it? That's how we think about it. We think it's an, obviously, we believe it's important. I also think it's important more broadly, which is just the principle of this is Europe. The two major players are European companies.
It's kind of incredible that from a balance of trade perspective, Europe is allowing this dynamic to persist. I think the U.S., the U.K., India have all shown a very clear direction. By the way, the U.S., the U.K., and India are all showing investment in networks and performance in networks. Because ultimately, a race to the bottom is not a value-creating experience for anybody. It's not better, I know people get excited because maybe the ARPU of the customers is lower, but that's not better for operators. You don't get better service in those markets when you look at it as a customer. Those are the places that aren't as deployed for 5G.
If you think about the economic impact of what's coming, you're not going to be able to invest to get the kind of innovation and development, and ultimately economic development and GDP growth per capita that you'll get if you invest in connectivity as a core asset. I think there's a lot of opportunity. I'm optimistic, but I'm also pragmatic. When we see that opportunity, we'll be very aggressive about pursuing it, but we're not going to run the business anticipating something that we're not in direct control of today. David, scale up, scale out, scale across. That's something that used to be close to my heart, but I'll let you talk about it.
Go ahead. Right. Now, Simon, yeah, certainly Metro DCI, over 50% of the investments are still going on in that front-end data center today. That business, and by the way, when I look kind of at the early success we had in hyperscale, that was it. It was DCI interconnect at, again, those high data rates carrying all that traffic. That will continue. What we see growing at, let's say, that exponential rate is, again, this scale across, just the sheer magnitude, as I mentioned, of the port, both the overall switch capacity required and then the port densities that people are looking to go connect. That is a sweet spot for us that, again, by landing it and expanding, I'm seeing a, call it larger average deal size than what we saw in the DCI interconnect.
Again, just not to bore you with speeds and feeds, but again, in the case we gave with the 18E, you're talking about full chassis, full seven-foot racks with half a petabit, and again, 576 ports that require thermally efficient 800 G pluggables traversing. That market has a long way to go. I mean, it's early innings, so feel great about that. As I mentioned, when you look at Broadcom's release plan for Tomahawk 6, and when you look inside the data center, port densities, right, the switching has not yet moved to 1.6 yet. There is still a lot of 800 G to go, as well as that move to 1.6 first in the data center, followed down below in the back end.
I know I probably didn't give you the SAM split up that you would like, but what I would tell you is that deal size is a multiple of the average DCI deal size. I would say the thing to add that I'm not going to say excited because all you guys keep saying I'm excited. What I will say is what I am feeling good about is, look, we get the switch and the pluggable. We have both. We have the solution to be able to do that. You can do some unique things. We are very open. Everything we do is open. When we do do it together, there's some power management things we can do. Like in the wide area network, we're doing that, where you can take power down 19% or 21% on a per application basis. That's a big deal.
Yeah. Maybe what I'm excited about, since I must have stolen your coffee at the break, what I'm excited about in this is look at the pace of innovation, right? I think that's the, for me, the thing that's compelling about scale across is, okay, AI factories, we've got a power constraint. How are we going to solve that? We're now going to interconnect data centers, right? This is the pace at which this has changed. Because we were talking about 10 KW, 15 KW, 50KW kilowatt racks, like those were a big thing. Now we're at 400 and even more. Now we're talking about the footprint being constrained, so we're solving AI factories. I think both Pallavi talked about this from a technology perspective. David's done a great job of breaking it out in terms of a business opportunity for us.
That pace of innovation, that's what's coming. That is also, when you think about the broader opportunity for us, why the agility, the clock speed, the shift in the company. I know you guys would love for us to probably lean in a little more on the opportunity, but I'm much more focused on how we're acting internally. That is why what I see from the team in terms of responsiveness, coming out from very far behind to catching up on 800 G to launching to shipping, I mean, that's the kind of pace that we're moving at to being in a position to deliver 1.6 to meet the market. I think that's the change in clock speed that we're all talking about and why we are excited about the opportunity, even if we're measured on what we're showing. Yeah.
One thing, I'll add a third question for you to ask me, and then I'll answer it for you. No, the other impact that has just on optical systems that I didn't cover in my rapid talk is that we can connect these things, again, via direct pluggable, direct coherent pluggable into these big switches, or we are still seeing, right, the traditional optical systems growth happening. We're not fighting a religious war here. We're on both sides. To give you just a little bit of an insight into what's happening, these line systems used to be, right, single fiber pair, right, single rail. You're seeing now, because of the power requirements, and if you look out in the huts where you're amplifying in the wide area network, you're seeing demands for things that are multi-rail line systems, where you're now looking to put 128 pairs, 160 pairs.
Pallavi covered this. Traditionally, in a hut, you're talking about 3 KW. You can't go out in some of these locations and add more power, right? That's all, again, nice opportunity in systems, switching, and in this pluggable demand. Sorry to ask your third question, Simon.
All right. We'll take our next question from Terrence.
Clearly an advantage to sitting in the front of the room. Those lights are interrogating. Yeah. Yeah. Thanks very much. It's Terrence Hill from Morgan Stanley. Just another question on optical and IP, and exploring the target that you have for revenue growth of 10%-12%, which you're clearly very excited about. However, your largest peer is talking about 17% growth in optical. I just wondered if you can do a bit of compare and contrast and talk about if there are parts of the market where you think you could be better represented in it.
You got it. Yeah. Yeah. You see, I twitched when you did that. Yeah. No, I mean, I was from a day where 10%-12% was pretty good at these things. Yeah, look, I think we had a later start as a company into the hyperscale space. As I said, the EUR 1.5 billion, euros, not dollars, of orders that we brought in, in terms of the total spend, it's a great start, but that isn't where we need the destination to be.
When you compare me with maybe that particular competitor that you're talking about, a couple of things to note, they do have a larger exposure into the hyperscalers, number one. Number two, they were earlier, although not vertically integrated, into the pluggable space, including 400 G. We're really intersecting at 800 G. Number three, when you look at where a lot of their business comes from, it's where either Infinera nor Nokia had any real big presence in optical. The major carriers here in the United States, in terms of you can think of their names, right, that was a very large when you look at a percent of their revenue. Now, the beautiful thing about what's happening in the network, again, I'm talking my ear off, is these are insertion points where we're focused right now.
As we get those wins, when I looked at that carrier space growing to one to two, if you're not in those large carriers, it's very, very hard to, again, grow ahead of that. That's all green space for us. That is not space we've been in, and we intend to get there. Did that answer your question?
Thanks.
Go to Felix.
Hi, Felix Henriksson, Nordea. Thanks for the presentation and for taking the question. Another NI question from me as well. You've presented the 6%-8% revenue growth target today. To me, it doesn't sound like it sort of factors in much tailwind from items such as AI inferencing and fixed, the optical components business, or the NVIDIA partnership. Is that a fair assessment? Yes. What do you consider as the biggest upside risks to your targets?
Yes is the answer to the first piece. Look, I think in the rest of the business, again, I think Justin mentioned it, for us, it's execution. The opportunity's there. We have the critical platforms. We've gotten through typically acquisitions, no matter how much somebody paints them as lovely, we're hitting the objective, but it's not easy. That's hard work, right? We have the focus in the roadmaps to go intersect. To me, the number one risk is execution.
I think the other thing too, and just to call it for what it is, because obviously if we do the math, we're pretty conservative on Fixed Networks. I think we're very optimistic on the business. The reality of what we inherited, David and I inherited in that was a portfolio that was focused on chasing what I would say is low-calorie revenue.
That is why we made the decision we did in fixed wireless access CPE. Again, it is a good business, but it does not fit our profile. It is not a place where we can differentiate. We can add value. We are going to be more measured on that side of the business. To be frank, I think the business has not consistently focused on where it is differentiated. You look at a couple of customers. I mean, certainly we talked about AT&T today. In earnings last quarter, I talked about Frontier, two U.S. customers where our technology on the OLT and the systems and the associated service side, where when we deploy, we are delivering tremendous value. I think that for us is getting the business refocused. You have a little bit of that capacity that we are building into the profile.
Yeah.
That's the math, I think, that you're probably trying to, you're seeing a little disconnected.
Yeah. I would also add that in the past, I think that business was held to a bottom line operating income very strictly, where optical didn't have the scale and wasn't contributing. There was a little bit of taking from pockets with different areas, taking a different tilt at innovation. Across the three businesses, we're investing. As Justin mentioned, whether it's AT&T moving from 30 million subscribers to 60 million, that's pretty profound. On top of that, Japan is going through the entire, if you look at what's happening standards-wise, there's a move from EPON to XGS-PON, where we're the leader. There are pockets, but again, as Justin said, we're focusing that business on growth and on margin expansion and on preparing for the inferencing.
To your point, I think it just would be way too early to put anything down that says, "Okay, here's the financial metrics." I think we have enough to deliver in front of us, and execution is the key.
Sandeep.
Thank you. Sandeep Deshpande, JP Morgan. Two questions, if I may. Firstly, on the mobile infrastructure business, I mean, you've talked about this transition to a software-based architecture going forward. I mean, in terms of the customer base, you've talked about 15 of the 20, I think, global telcos you are in, two of the biggest U.S. telcos you are not in. I mean, is that going to change over the next three years? Clearly not in the three-year horizon you've given on the estimates, but is that a target, or is that not a target?
Because that could change the operating margin of the business very substantially if you were able to get into one of those big telcos as such. Then secondly, back again to network infrastructure. In switching, I mean, Nokia has been behind the curve, but now gaining share. What is it that is making Nokia gain share in that business? I mean, there have been existing players in that market. Is it your software? Is it your silicon? Is it something else? Maybe trying to understand what is it going to make Nokia going to be a bigger player in that market? Clearly optical, you're already there, but in switching, that is the question.
Okay. On the first one on Winbacks, look, I mean, obviously if we had news to announce, we would announce it. I think right now our focus is on building the best portfolio that we believe is going to enable differentiation for the customers that value it. Look, they're important customers for us. By the way, if you think about those two telcos where we don't have presence in the radio networks today, we've got presence across most of the rest of the portfolio. They're important customers for us. Right now, our focus is delivering for the customers that are partnering with us, innovating, and you see that obviously with the announcement on AI-RAN, and making sure we're pivoting the business for the future. I think there's tremendous opportunity for us ahead. I think Simon touched on one, obviously, with his question earlier in Europe. I think there's the potential with high-risk vendors for a growth vector. Obviously, this is another place where there's potential.
What I can just say is, if you think about the capacity constraints to be able to deliver hardware manufacturing, obviously software resources to be able to go deliver capabilities, those are areas where I feel really good when if and when the opportunities arise, that we'll be ready to support those for customers. Again, we can't run the business expecting those to happen. That's the way I would think about it. Look, on IP networking, I think the first answer to your question is David , but I'll let David explain what he's doing about IP switching.
Again, I don't want to get into a religious war on this, but routing is really, really difficult. I'm not saying switching is easy. I think the first instances, if you think about hyperscalers and their first applications and workloads, is the workloads are now becoming way bigger and way more complicated. The networks they operate are now becoming way bigger and way more complicated. They need diversity in their supplier base. I think, one, we were not there early enough. Like listening, co-collaborating, this is not do a binder of a spec and then go, as Raghav alluded to in the presentation, things have changed. This is put a bunch of engineers in a room and start working heavy, heavy on both network workload all the way down to the switch. I think there was an absolute need first for the rock-solid software I mentioned.
Because we were able to pivot SROS into SRLinux with some heavy listening, that rock-solid nature, people can take in the team that does that is just awesome. You can take an alpha version of our software, and we found our customers like, "Holy hell, this is like we can't take an alpha version of our competitor's software." I think the bar is raising in terms of complexity and reliability on the software side. Obviously, on the hardware side, they need rock-solid hardware. We have a history not just of doing our own FP and FPCX chipsets, but on the edge of the network where you're using some of Broadcom's open chipsets and some others in the industry, we know how to integrate that. That's what's given us the opportunity. I mentioned to you that 18E, when you're talking about half a Pb, there's not many people that can do that. In the speed we did it, it's probably the thing I'm proudest of.
Yeah. I mean, I think just to add to that, when I first came into the company, it was very apparent to me this just wasn't a focus. I mean, by the way, it's hard to criticize Nokia. If you look at the three players that were delivering telco-class routing and two of the players that were doing enterprise networking, they largely missed the transition to cloud, right? That's why Arista has been so wildly successful. Nokia was one of those, just being very blunt about it. It wasn't considered a strategic priority. I think what we've reprioritized now is SRLinux as a critical OS.
That's a differentiator, as David said. The discipline of being time to market with Broadcom Silicon, which is very important for the industry, and we know that. It's just getting focused on this as, "Hey, this is a market where the left edge of technology is getting developed." By the way, that doesn't mean we're not going to focus on telcos. We're very focused on that core customer base, as David highlighted. You have to recognize the market transition. The reality in switching is that started in cloud, and we missed that. Now we're getting back at it. We think because of the pace of change in AI factories and what's happening within the data center, there's opportunities for us. That's being reinforced by customers. There's a lot of strong competitors. There's a lot of incumbency. That is also why, again, we're being measured about the progress that we're making.
Yeah. I'd say if you look like maybe even as soon as five or six years ago, if you missed a cycle with a telco or a Mish Critical, you were out. You were in the penalty box for six years. The great thing about the hyperscalers, cloud majors, and these folks building these AI factories is price, performance, cost per bit, power per bit, reliability. Look, there's insertion points all the time.
I'll also add automation because especially in our switching portfolio, David was talking about SRLinux. What comes on top of it is our EDA tooling. Now, especially when you're looking at switching, these are your leaf, spine switches. Configurations can actually become very complex. With the EDA tooling that we bring in, we're getting very positive feedback from customers.
Yeah. Okay.
We'll take our next question from Sami.
Thanks. Sami Sarkamies, Danske Bank Markets. I have a two-part question related to recent NVIDIA cooperation. First part goes to David, who was excited about a number of things but did not really mention NVIDIA in the presentation. How do you see the opportunity for network infrastructure? Are you able to size that opportunity together with NVIDIA? The second part to Justin, if we think about the implications to mobile networks, which will be included in mobile infrastructure, how do you see the financial impact from NVIDIA cooperation? If we think about 2026, 2028 timeframe, you probably need to hike investments, but we will not yet see much revenues from those new products.
Yeah. You want to take a minute? I mean, I think we've said most of what we're going to say on NI, but I don't know if you want to—there's something else you want to— Yeah.
I mean, thank you. You've uncovered a weak point in my presentation. The hats off. I didn't mention NVIDIA. I didn't mention NScale. I didn't mention Supermicro. There were three things I should have mentioned. Just only so much time and my mouth gets running, and my brain doesn't catch up. Partnering is key to what we do going forward. I think all of those are wonderful partnering opportunities. I see that impacting our roadmap in optical and IP going further and actually potentially even in FN. It is too early to unpack that. I will say I'm excited about the opportunities those provide, but would be reckless to put anything out there in the near term.
Yeah. To your question on AI-RAN, I think we've been pretty clear on the revenue forecast point, proof of concepts in 2026, early deployments in 2027, broad commercial deployment in 2028. One thing that maybe was not obvious in the presentation today is we're going to reallocate capital. That means stopping doing certain things and investing in places where we see value. I think as you think about radio networks, because remember, Mobile Networks was a slightly different portfolio. It included microwave. In radio networks, we see opportunities to reallocate capital. One of those things that we're investing in is AI-RAN because we see it as very strategic. Also, we're not starting from zero. We've actually done a ton of the work to cloudify the NERAN stack, to do the initial trials with NVIDIA to validate this technology worked well before the announcement.
The development work for us is incremental, and we'll trade that off against lower priority opportunities. The reality, again, in this business is when you look at the business, there's a lot of product investment that hasn't generated return. Ultimately, if you guys do the math on the return on investment in what was previously Mobile Networks, and even if you can click into radio networks, what you'll see is that the business hasn't generated its cost of capital. We're going to be very disciplined about making investments that we believe can generate excess returns. This is one of those things.
By the way, it turns out when you look at our biggest customers and the customers that we support and serve, and those customers we do not have that we see as incremental opportunities, they largely give us the feedback and the roadmap on where we want to invest. Core, I think we did a better job over the last few years of being able to deliver against those requests, and the results speak for themselves. The reality in radio networks is we are executing that pivot aggressively now, and we have been for the last couple of quarters. It is going to take a little bit of time to show up in its results. I think for the prior question on the U.S. operators, you do not just flip a switch overnight. It takes time to deploy into these networks and grow.
I fundamentally believe where the market is headed, if you listen to Pallavi, we talk about this a lot, the things that are happening in the market, the devices that are being deployed, robotics, autonomous vehicles. I mean, I'll get back to this. I use this simple example for folks. Living in Helsinki, you've got delivery robots today. We're going to have autonomous vehicles because Finland is both connected and advanced. Autonomous vehicles, AR, VR glasses, all these things are coming. What you're going to need is a performant network because the reality is, while these devices are intelligent and autonomous, they can't be disconnected. There hasn't been a point where we've gone through technology where we determined we need less bandwidth and less connectivity. When you think about these devices, why in the world would you allow them to run on an untrusted network?
Untrusted radios, untrusted core networks, untrusted transport. Simply not going to happen for many markets. The third thing is, when you have all of that, you need a network that performs because those devices can't be sitting there waiting for an instruction, as Pallavi articulated in her presentation. That change will happen. I think we're in a position here, much like we were in core five years ago, to be a leader and drive a model shift and an innovation shift in the i ndustry.
I'll shake things up and go to the back. We'll go to Rob.
Yeah. Just a quick question for David on the Indium phosphide fab. I think when you announced the Infinera deal, that fab was quite lowly loaded, maybe only 35% loaded. Now you've announced a major expansion quite recently. Could you just talk about how fast that ramp is looking and how the utilization sits today? The second question would just be around for both sides of the business, just around potential supply constraints, whether it's memory or passives, or there seems to be quite emerging tightness in a lot of key component areas. I'd love to get some clarity on that. Thank you.
Yeah. I'll do a job of bridging the last question. The one thing I did not mention is when I was at Infinera, while we've done some things with players like NVIDIA, the power of Nokia opened up a major partner door, which, when you get inside that door and look at some of the quantities and the need, it makes me feel good about filling the utilization of the fab.
The great news is two and a half years we started that fab. It will qualify at the end of this year, and we will be actually producing. We already have one existing fab that, let's just say, the utilization is going up. I will just tell you that this components business, it was in green in my presentation, and I said, "Don't put it in any number because it's way too early." I believe it was 283%, some crazy growth number. Look, the core capabilities in that fab and building high-power lasers, building arrays for co-packaged optics, that's where we came from. We invented the photonically integrated circuit. Our first pick of real like the ICE 3 pick was a, if you think about it, it was a 12 by 100 arrayed solution.
If you think about what's taking power down, it's moving electrical lanes into an optical array. While the religious wars in optical of what you hear will go, "Well, co-packaged optics or linear pluggable optics," or, "You know what? We've got both," and it's all off that same fab. That will qualify. Again, when I said 25 x capacity, it's 25 x capacity. Some of those applications that we're talking about have much better yields. I expect even a better view of that. It's way too early. I'm encouraged by now the doors open at places like NVIDIA. Again, we're not going to probably report our fab utilization. Marco doesn't want me to do that, nor would I.
No. I mean, I think the one thing to say is just a nod, and I'll let you comment, is we're going to be prudent on capital allocation. If we're investing in fab capacity, it's because we see demand.
Yeah.
Absolutely. No, absolutely. Extreme discipline on securing that we get the returns. In this field and this fab, we are very confident that this is a good investment. Also that it's a North American, it's a U.S.-based fab, which is an extremely big deal here now.
I'll take our next question from Artem.
Artem Beletski, from SEB, thank you for all the presentations today. I would like to ask on your number one strategic priority, which is really to accelerate growth when it comes to AI and cloud. This business represented 6% of revenues in Q3. Where could it be by, let's say, 2028? You're talking about addressable market growth of 18%. David spoke about quite nice growth in terms of orders. Could you maybe comment on this topic?
Yeah. As we said, Q3, 14% of NI, 6% of group. I think NI is probably the better metric because the reality is that's where the exposure is. If you do the math on kind of what we shared, we think it could double. We think the business could be—we think the non-Telco business could be up to 40% of our overall NI mix. That includes mission critical. That's just at the rate that we're seeing the growth. I mean, obviously, if we see the market accelerate faster, we execute better than what we're assuming, obviously, that penetration becomes higher. I think our planning assumptions, when you look at it, are that Telco is quite predictable in this period, right? That's largely fixed.
The real variable here is how much. Periodically, we'll give you an update on this number on the mix in NI to have a sense of how we're doing. We'll do our best to give you a view of where we are. I think the other piece here, and David unpacked a lot of this detail in the presentation, is the business model is evolving, right? If you think about—you know, he touched on this in his talk—that we've started in systems. We're now going to pluggables. He just touched on the fact that we could be in components if there's an opportunity. In an opportunity, as we go into linear packaged or co-packaged optics. The business model will change. The discipline for us is making sure that in each of those cases, we're delivering differentiation, we're generating return.
For me, that's the other thing that we're really looking at internally with Marco and David is making sure we're lined up to capture those opportunities, but giving ourselves the flexibility to make sure we have the right business model. Because the miss here would be to be too attached to pluggables if components are an opportunity, or too attached to systems if pluggables are an opportunity, right? We need to follow the market, leveraging the unique technology we have. That's what we'll continue to do. Like I said, we'll update you as we make progress.
Thank you.
We'll take the next question from—I can't—Paulo, you can pick someone from down there. Maybe Andrew. I can't quite make our faces with the lights.
Thank you very much. The mystery asker is Grant Lenehan, Appledore Research. For everyone's knowledge, I'm the square peg in this round hole. I'm an industry analyst, not a financial analyst. I want to start with what's been going on in the industry. If you look at the telecom service providers and you look at how their spend every year is broken up, actual network capital is about 12%-17% of the total. The rest is operations. It's people. It's controllable expense. We also see service complexity rising, volumes going up, technological complexity, configuration complexity rising. You've made a real argument that you want to bring together businesses to provide differentiated capabilities and better technology.
Yet what I didn't hear much about, though kudos to David for talking about Altiplano and NSP and all of those things, and your talk about AI and the RAN, I didn't talk about how you were going to bring all of the software together to address that other 85% of the money and help your customers transform their business to make you successful. I may be a bit of a software nerd, but I'm dying to hear about that either now or another time.
Maybe what we'll do is let Raghav and Pallavi talk a little bit about that. That was really what we touched on in the core networks space in the MI overview. Raghav, maybe you can talk a little bit about what you're seeing and then really what we've been doing in this space because this is a problem we recognize. Then Pallavi, maybe you can add.
Yeah.
Yeah. I think when you go and speak to the customers themselves, I mean, you think about the things they talk about these days. These networks are getting extremely complex, and you just can't run them just by humans by themselves. You have 2G, 3G, 4G, 5G. These networks have become extremely complex from that perspective. Automation is one of the key things that the customers are now asking for. Automation is a very big word. There are industry definitions of that by the TM Forum that you go from the L2 plus that the industry's at to L4 and then eventually into L5. This is going to be a journey, but this is very clearly front and center of every customer in the telecommunications company that they're looking for.
This is where we are building the autonomous network fabric that will really allow the journey from where they are today that as you get into 6G, you're going to have these autonomous networks that are self-healing, self-provisioning, self-managed networks with a little bit of human augmentation for critical activities. This is a very key area of our investment cycle in terms of what we're investing in. This is very key to come. The other area that's very, very specifically that they're looking for is making sure that the networks are secure because as you disaggregate getting into cloud-native capabilities, this disaggregation allows a great amount of automation, but it also opens up the fabric where you can have many more cyber attacks.
In this whole autonomous network piece, security is a very, very key part of how to make sure that you drive that into the fabric itself and into the architecture as you go forward. That is another area we continue to invest heavily in. In this particular area, we are also bringing the generative AI capabilities to be able to make sure that we can act as a co-pilot when these security incidents actually happen. This is very, very key. I think you raise a great point. I think Justin covered it a little bit on the autonomous piece as well as Pallavi, but I will let her add more comments to it.
I think two vectors to look at this. I mean, Raghav, you covered everything from an automation point of view. When we are looking at autonomous networks, we are looking into not just how customers are managing their networks, but right from the time when they start planning their networks. How do we go about and enable them to optimize the planning process? How do they go about and design? How do they go about and deploy? When they have actually gone about and deployed the day-to-day, I like to call it as a day minus one, day zero, day one, and then the day end. Those elements are what we will actually—you are very focused in looking into how we bring in the autonomous network layer. The second point that I will say is every customer, every operator is in a very different stage of automation. This is what Raghav was talking about, the TM Forum having definition of L2, L3, L4.
The way we are building up this automation is through APIs. Depending on which stage the customer is in their journey of automation, they can plug in into what we are providing.
Maybe one last comment I'll make on this, and then just to move on so we leave time for a couple more. The last comment I'll make on this is in MI. If you think about that business, core and radio, we were investing in these things in parallel. We brought them together. Raghav led some of the—or led this effort with a couple of key customers to start bringing them together because we recognize, and our customers recognize, that we can't just build an autonomous layer for one part of the network. It needs to be across the entire network.
Now, I think that will extend into fixed because, obviously, as you think about transport and backhaul, there's opportunities. We see that linkage. We're going to be very customer-driven in this space because the risk here is we start building capabilities that our customers don't want. The key thing here is building it with a fabric and a platform, as Raghav touched on, thinking end-to-end from the device, obviously, to the network edge, the access network edge, which we manage, and then all the way back through to the host. That view for us is really where we're taking this, particularly in MI. It means bringing these resources together. That's already really started ahead of the announcement because it's just logical.
You can't deliver a network slice or intent-based networking in the core if it doesn't actually go and deploy to the radio. Ultimately, you need to be—some of the things Pallavi was talking about with dynamic traffic patterns—you need to be reallocating the RF in the area to make sure that you have the bandwidth available to deliver the experience. If it's an encrypted service, making sure that's following as the device is moving. There is a lot more complexity here. The last thing I'll say here is this is fundamentally a multi-vendor environment. We recognize that the work we're doing in this space is multi-vendor. That's why APIs are important. It's also why extensibility matters because you can't assume that I'm going to have one technology provider in 100% of the network. You need to design for diversity.
It's Tim Savageaux at Northland. Since you guys reported, which seems like a long time ago, given all that's happened, we've seen a lot of indications across, especially the supply chain and optical, of a pretty dramatic uptick in demand, kind of a step function in real time. A lot inside the data center, but also a lot outside. David, given your comments on the order strength, I wonder if you can provide any color on how that's continuing into Q4 or a general sense of the momentum you may be seeing there.
Yeah. You want to talk about orders or supply chain?
Orders. Orders. Okay. You want to talk about orders? Yeah. Obviously, I don't want to provide any update to any guidance, and we don't guide orders. Let me refer back to something I said, Tim, and I know you know this, in terms of in the traditional days, if you're shipping subsea, maybe it's 10,000 transponders a year, right? I think what's being realized now is if you're getting into these AI gigafactories and you look at kind of Medcav's law, the number of just ports you have to connect. When I talk about moving from 10,000 in subsea to 5 million in a GPU ecosystem interconnecting GPUs, people are getting nervous. They're looking for high-power lasers down the roadmap for CPO. What I would tell you is people are looking to secure design partners for the future. What I love about that is it's not a binder and an RFP process, and then I wait.
Because we can kind of see things coming as we get design wins, then we can start to say, "Great, here's the business in front of you." Yeah, look, I've seen pieces written, heard things from clients that say, "We're worried about the tightness in the optical component market around the ability to get enough lasers to be able to make what's going to be required in this next cycle of build."
Just to follow up briefly, on the MI side, it looks like the plan or the aspiration is to grow from $1.5 billion, but the guidance seems kind of flat. That's one of several areas that seem conservative, whether it's service provider down and network infrastructure, fixed down, and I know you're being more selective there. Am I looking at that right, and to what extent are we looking at a baseline outlook here versus an aspirational outlook? That's it for me.
Yeah. Look, I think clearly we've got to focus on where we see opportunities in the market. You're making a lot of changes. If you think about the changes from the time that I came in, this is a pretty dramatic shift for the company. The other thing we have to acknowledge is historically, this is not a company that has delivered predictable results. A lot of one-offs in earnings, good and bad, and a lot of inconsistency.
While I'm pleased with the start of the operational rigor and the forward visibility that we're seeing, and I'm pleased with where we're going, I think we have a much better handle on the business as a team than we did when I started. I also recognize that we're a couple of quarters into this journey, and these things don't change overnight. I think we're being prudent in the guidance based on what we see. David, he talked to you about why we've tempered the growth forecast because we're starting with such a heavy mix of telcos and NNI. We also recognize we're going to go through some portfolio transition in MI, particularly around radio networks. We want to be disciplined. Obviously, we feel like this is prudent. We think it's a great platform from an investment perspective.
I'll talk a little bit about that at the end. We also think it's prudent. What I want to see, and I've talked about this a little bit in earnings and certainly in other forums, I want to see us be predictable and stable in terms of execution. That, for me, is important because I think that's an important part of investing in us and seeing us as a valuable option of where to put your capital.
I think we'll go to Andrew.
Thank you for taking the question. Andrew Gardiner from Citi. Another one on fixed. If I look at how you described the addressable market growing, yes, less than switching and optical, but still it was a high single-digit growth number. If I sort of back out the fixed assumption within your guidance between the 6%-8% for NI and the 10%-12% for optical and IP, you get to sort of down mid-single digits. So there's a big delta when you compound that over three years. You've talked about sort of pruning and being a bit more selective, but it seems a much bigger step in terms of the seeding of market, the choice to seed market share there. Just how bad was the problem within fixed in terms of taking low-margin business previously that you've got to make such a big step in terms of seeding market share in that division? Thank you.
Yeah. Look, I think we have very healthy market share in OLT. As David said, right, fixed was kind of used as the way to find marginal profit to offset some of the other businesses and sometimes to find top line. We are going to be disciplined about it. We have a very strong portfolio, but we also need that business to go through a portfolio transition. A great example is just look at the impact of the business. If you look at, and you will see it in the numbers, if you look at fixed when you back out FWA and SIOP, you see some of that. I think we are giving ourselves a balanced view to be able to go grow in the core business. Ultimately, I think this is a very good market for us. We are a clear leader. We are investing in this space. We have differentiated technology in Quillion.
However, some of the revenue there we've had is what I would call low-calorie revenue, and we just need to be disciplined about it.
I think we'll take our last question from Emil.
Hi, Emil Immonen from DMB Carnegie. Thanks for taking my question. Maybe one last on mobile infrastructure. Could you elaborate maybe the long-term thinking? Because I think Pallavi described pretty well that kind of the network requirements are changing. Is the target of having this almost flat in the next three years just because we're in the middle of the 5G transition? Then the growth comes when we go to 6G.
Do you want to make a comment at least on that?
Yeah. The way we are seeing it, and especially the AI-RAN portfolio that we have, it is going to be 5G ready. For us, 6G is just a software upgrade. The way we look into it is some of our leading operators are the ones who are actually going to take up the AI-RAN. That is where we are actually co-creating with them. The rest of the industry is actually going to go about and latch on because as you go about and develop this, and we were talking about it, this is where we start creating value for the operators. Those values can come in into simple examples of node slicing, for example, right? How you can take care of different customers, different segments in a different way, how you can actually bring in all these better user experiences, whether it is in terms of connected cars, AR, VRs, and all of that stuff.
Yeah. I'd add two things, Emil, just in terms of thinking about the business. You have two things happening at once. One, there's parts of the portfolio, parts of the services portfolio, product portfolio where the businesses don't generate a return, and we're just putting simple discipline around the business, right? That's what we touched on. That puts a headwind on growth, right? That's one element. The other, if you think about what we're doing, we're basically going through a hardware-centric transition to a software-driven business model. I touched on this a little bit in my comments, but storage is probably a good analogy to think about as another technology segment that went through this. Or core networks is obviously a very close one for us. As you think about that, what you have is the hardware commoditizes. You see top-line pressure on the hardware because you're not going to—you're not trying to pack value into the hardware.
The software goes from being either embedded in the hardware or a term license, depending on the model, the customer and the model, to being a subscription recurring revenue model. That means that it's recognized radically. Those things put some pressure on the top line of the business and a little bit on profit, obviously, because you're making that transition. That is why we're being prudent and disciplined. That is why when I said the comment I made before, if you just think about the core business and Patrik's business, you have a very stable profit base, and that's what allows us to make this transition. Ultimately, again, if you look at this industry and you look at the two big Western players after all of the consolidations, this hasn't been the most attractive investment.
I believe we're going to go through an interesting cycle. Now is the time to change the model. When I look at what's happened with core, I think the opportunity is to do something very similar in radio. Also, everything Pallavi has shared technically, the market will need it. Because if we just deliver the same radio networks with faster bandwidth, this won't actually work for the market that we're in. It won't work to bring industrial AI and physical AI into the world. I think it's a—I'm sure it's challenging to kind of see the logic and the math, but we actually feel like we've got a lot of opportunity here. As I said, we've gotten started on this for some time. Now it's just about focusing and executing and seeing that trajectory.
The last thing I'll say is remember that the early deployments of AI-RAN are a hardware upgrade, right? A million units of AirScale, a million units of AirScale deployed in the market. That install base is the first place we anticipate upgrades. T-Mobile talked about that in the announcement we did with NVIDIA. That's not a huge—the big revenue sweeps come from the radio. If we're upgrading to 6G or, sorry, 5G AI-native to 6G with just an upgrade to the baseband hardware, a sled in the baseband hardware, that's also a lower revenue performance. Most of the top line really comes when you deploy a large radio network.
It's a little bit to unpack, and we'll look to continue to talk to you about this and explain it as clearly as we can as we go through it because obviously, it's important. No one asked questions of Patrik, but I think hopefully you got a good picture today of the fact that he's got a portfolio and a revenue stream that's incredibly stable. We're investing for the future. That gives us a strong base. When I think about the comps in his market, he's a significantly larger and compelling business. I think that positions us well with the cash flow and the stable profit to make this transition. Long term. Yeah.
Always important to have an ATM. Longevity is extremely important. I hope that Patrik could convince you as well that this is not some business that is just in a good level now and then it's going down. This is something we continuously continue investing to new standards, new innovations to secure that we create that cash flow and profits.
Right. Cool.
Thank you, Justin, and all the presenters. Thank you all for joining us. I think I'll hand back to Justin just for a couple of concluding remarks.
Yeah. No, good. Look, first of all, thank you for taking the time to be with us today. As I touched on, and there was a question in the room, I thought it was a good one, so I'll re-answer it again, which is all of this opportunity, fairly conservative forecast. It's a balanced forecast. Again, Nokia is a company that has not consistently executed.
In some cases, we haven't consistently delivered a compelling return on investment. That's a part of the journey of what we're changing. What I'll also share is this is more than just the GLT you saw up here today. This is the set of leaders. In fact, over half of our employees contributed their feedback and input to our vision and strategy and where they want to see us differentiate.
Over those half, they all talked about the core of being an innovative company, partnering better with our customers, and empowering our team to lead by connecting intelligence. I think we have a clear North Star for the company now. We have five key strategic priorities that are informing our decisions. I think we have a reasonable platform which will deliver both top-line growth and ultimately midterm profit expansion, targeting double-digit operating profit CAGR through 2028.
In addition, I think this time period over these three years sets us up to being a far more profitable and far more compelling growth story down the road as the markets we're in both continue to grow in the case of NI and return to growth in the case of MI. Thank you, and we'll look forward to seeing you soon.