Nokia Oyj (NOKIA) Earnings Call Transcript

Hello, ladies and gentlemen, and welcome to our Nokia Mobile Networks Progress Update Event. I'm David Mulholland, Head of Nokia Investor Relations. And today in Espoo with me is Pekka Lundmark, our President and CEO; along with Tommi Uitto, our President of Mobile Networks. Before we get started, a quick disclaimer. During this call, 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 have identified such risks in the Risk Factors section of our Annual Report on Form 20-F, which is available on our Investor Relations website. In terms of the agenda for today, Pekka will provide a brief introduction on his perspective of the progress we've made this year, before Tommi then provides you with a presentation to explain the progress we've made in technology leadership in Mobile Networks, how we see our addressable market evolving and then some thoughts on the trends we see related to O-RAN and Cloud RAN and how they might impact our business. We expect the presentation section will last for around 45 minutes and then we'll move to a Q&A section. For those looking for it, the presentation can be found on our Investor Relations website. With that, over to you, Pekka.

Thank you, David, and thank you all for joining us today. Before we talk about Mobile Networks, I want to say a few words about how we want to communicate with you going forward. As we said at our Capital Markets Day in March, we are taking steps to further improve our investor communications. We want to give you the information you need as clearly and as simply as possible. The key steps we have taken are summarized on the slide. We have streamlined our financial reporting, we have become more transparent about the underlying progress in our business performance, and we have been providing a balanced outlook commentary and brought greater visibility to our ESG achievements. But we also appreciate that sometimes there can be too much information to fully digest in a single event. So while we will continue to hold Capital Market Days, they will probably not become annual events. Instead, we feel it is better to give regular progress updates focused on a specific business group, where you can hear directly from the business group presidents. After all our business groups are fully accountable for their performance, so it makes sense for them to explain their own journey. We want to increase predictability. So you can expect to hear at least once a year from each business group. These events will focus on our strategy, product and technology and will not be updates on our financial outlook. Today, we start with Mobile Networks, which makes sense because of how much progress they have made this year. But before I hand over to Tommi, let me remind you of 2 things. And the first one is our reset agenda. Against all of the metrics we had for this key focus area, I believe we have made meaningful progress, as you can see on this slide. We have taken significant steps to strengthen our technology leadership. Just to name a couple of product highlights. In June, we launched our next-generation AirScale 5G product family, which Tommi will talk about in a moment. And in September, we launched our next-generation FP5 routing silicon. Our new operating model and divisional structure have now been firmly embedded, and I'm really pleased with how quickly the team has started to demonstrate the benefits of increased accountability. We are continuing to optimize our cost base and ensure we make the right investments to drive technology leadership in our target areas. We have also adopted a new purpose and started the process of renewing our culture, and we have streamlined our leadership team structure to further increase accountability and agility. In short, 2021 has been the year of the reset, and I'm proud of the progress we have made despite the ongoing and industry-wide supply challenges. We have created a strong foundation for ourselves as we move towards the accelerate stage of our 3-phased plan. The second thing I want to comment on quickly is that while today is not meant to be an update on financials, I am really pleased with the financial progress we have made this year. You can see the key figures on screen, these are of course familiar figures. In the first 9 months of this year, we have delivered 6% constant currency growth year-to-date, expanded our comparable gross margin by 270 basis points and significantly strengthened our net cash position. I'd like to remind you that these numbers have had some one-off benefits, which combined with cost inflation and supply chain uncertainty, may limit our margin expansion potential in 2022. That said, I believe we are firmly on-track to take Nokia to the next phase of our plan. Mobile Networks has made great progress in key focus areas this year and that is testament to the entire team's hard work and commitment. With that, let me hand over to Tommi to give you more details.

Thank you, Pekka, and thank you all for joining us today. So my name is indeed Tommi Uitto, I led the product part of Nokia's Mobile Networks in 2019 and 2020. And as of 1st of January this year, I have been leading the complete Mobile Networks business of products, attached services and network management. Today I want to take you on the journey that we have been on in Mobile Networks this year to help you understand how we see the future opportunities ahead of us. I will touch on some of the proof points of how we have executed this year, before we look at how we're working to secure product, technology and services leadership, how we see our addressable market evolving, and the next iterations of mobile technology in 5G Advanced and preparation for 6G. Finally, I will also address our positioning on Open RAN and Cloud RAN, which are topics we understand raise some questions amongst investors, but which we believe can be very much an opportunity for us at Nokia. If you cast your mind back to my presentation at our Capital Markets Day in March, we committed to 4 action items for this year, and I believe we have executed very well on those. The first commitment was to secure our product portfolio competitiveness, in particular, in 5G. I will touch on this in more detail shortly, but with our AirScale portfolio launch in June, together with the 5G network performance improvements, I believe we have delivered on that and closed the gap to competition. In some areas, we even set the benchmark in the industry and got ahead of competition, for instance, with the industry lightest high-power 400-megahertz bandwidth massive MIMO with 32TRX; capacity and connectivity of our new SoC-based multi-radio baseband and some features like the 3-component carrier aggregation and Voice over New Radio, or VoNR. The second commitment was to improve our 5G momentum and deal-making with CSPs and private wireless. Enabled by the new competitive portfolio and roadmaps, we have now stabilized our 4G to 5G conversion rate and I am optimistic we have opportunities that it could start to improve again. In private wireless, we have grown from 260 customers at the start of the year to 380 in Q3. And even if most of them include 4G radio, already more than 70 of those cases include 5G radio. And then we continue to invest in our O-RAN capabilities. The platforms that we launched in June are all Open RAN-ready. We are working on the open interfaces and the cloud platforms. We have had our full-fledged Cloud RAN with both virtualized CU and virtualized DU with hardware acceleration in trials already with our first contracted customers. Finally, on our cost base, we have continued to invest heavily in R&D to ensure we achieve technology leadership in 5G and beyond. Despite of that investment, the improvements we have made in the cost competitiveness of our products and services, together with efficiency improvements in SG&A, have put us on a path to deliver a much better operating margin this year than we initially forecasted. Let's then get into some details of the strength and merits of our renewed RAN product portfolio. It's actually a very exciting time for Mobile Networks and I feel that we have now all bases loaded to secure 5G technology leadership. We have launched our new SoC-based AirScale radios, including the industry's lightest high power 400 megahertz 32TRX massive MIMO. These radios, of course, contribute to our 50% reduction in power consumption of massive MIMO radios from 2019 to 2023. Our AirScale baseband is now the industry benchmark for connectivity, capacity and flexibility. It also comes with significantly improved energy efficiency, reducing the baseband's power consumption by up to 75% compared to the previous baseline. This also contributes to our commitment to half the total base station power consumption by 2023. We're on-track to power our full portfolio with the latest ReefShark system on chips or SoCs by end of '22. And finally, this year we brought together our software to a common development trunk, which means updates to system software from 2G to 5G in one single release, bringing our customers the speed, operational efficiency and quality that they need. The customer feedback has been humbling. For example, Johan Wibergh, Group Technology Officer at Vodafone, just recently acknowledged Nokia to be truly back in 5G. Similarly, Telia Group CEO, Allison Kirkby told media that Nokia has fully caught up competition in 5G. I will take you through a closer look at the portfolio. But before this, let's see a short video. [Presentation]

Let me explain why our new SoC-based multi-radio baseband has become the benchmark in our industry. Multi-radio means of course that we now tick the box of having the same hardware for different radio access technologies like 4G and 5G. In a baseband platform, there are 6 competitive factors that are above everything else. Data throughput capacity, in other words, how many gigabit is going to be pushed through the system per second. Simultaneously connected subscriber connectivity. In other words, how many subscribers can be connected simultaneously through the basement system. Maximum cell connectivity. In other words, how many radio cells can be connected to the system. And then power consumption, in other words, the power consumption of the entire system or power consumption per gigabit per second, per subscriber, per cell. Scalability of the capacity. You see, when traffic volume grows, the so-called Layer 1 and Layer 2 processing need grows faster than the so-called Layer 3 and transport processing need. This is why we keep those functions on separate boards. You have 1 or 2 common boards for Layer 3 and transport, and then you can add up to 6 Layer 1, Layer 2 boards as you need. If you have all the functions, Layer 1, Layer 2, Layer 3 and transport on the same monolithic board or pizza box, you end up buying and paying for Layer 3 and transport capacity that you don't need or use. This architecture is also more future proof, because there are empty slots in the chassis and you can keep adding new boards as you go, rather than taking the risk of having to swap the whole pizza box if there are new features in 3GPP that would require new processors. We know this all too well from the past. Our Flexi Base Station architecture from almost 2 decades ago had that pizza box architecture and it just became clumsy in network evolution. We have packaged the latest baseband SoCs in a form factor of a single board too, but this product is rather positioned as an entry-level product for low capacity or rural base station sites where high capacity would not be needed. Finally, there is plenty of headroom in our new SoCs to make them able to support many features introduced in new software releases, new features requiring further computing capacity. Here, amongst all these competitive factors, we lead in almost all of them, and where we don't, we are on par. Let's move on to the radio units. We have a complete portfolio to serve any capacity and coverage needs of operators and enterprise, anything from our unique 3 sector RF modules to 4TRX single sector radio heads, to 8TRX radio heads, to 32TRX and 64TRX massive MIMO with beam forming, to integrated passive/active antennas where we are the market leader. These platforms are versatile in enabling us to spin any frequency variant with minimum effort and customization. We design for manufacturing, design for logistics and design for serviceability in mind. Let's look at the massive MIMO radio platform. There the key competitiveness factors are volume, i.e., size, weight, occupied bandwidth, instantaneous bandwidth and power efficiency. Size and weight are important because of cost of installation, the cost of the supporting side structures and of course visual impact. The bandwidths have to be wide enough to support operator spectrum holdings today, but also the spectrum they may get in the future. Wide bandwidths are also important for operators will combine their spectrum assets to share a network or who may do so in the future. Power efficiency is basically the ratio of output power and input power. The nature of wireless technology is that you do need output power for capacity and coverage. That's just the laws of physics, but what is key is the ratio between input and output power, and that's the way to reduce power consumption for a given required level of output power. Our new fully SoC-based AirScale massive MIMO platform features industry-widest bandwidths, industry lightest high-power 32TRX massive MIMO products. We of course, have also a new sleek 64TRX product in this platform as well, but we have focused optimizing size and weight more for the 32TRX, because that configuration enjoys and will enjoy the largest volumes in the world. Let's talk about the heart and brain of base stations, the SoCs, the silicon. Since the beginning of 2019, we have tripled the size of our SoC design team. We diversified the SoC partner base and we now work with Broadcom, Intel and Marvell in alphabetical order. Thanks to this, we now have all bases loaded with SoCs, let me explain. We can simplify a bit and say that there are at least 3 distinct computing functions, each requiring specific type of processors in a base station, if you are to fully optimize performance, cost and scalability. As discussed earlier, Layer 1 processing, Layer 2, Layer 3 processing in baseband, digital front-end or DFE also known as digital pre-distortion in radio units. In massive MIMO radios, there is also the so-called Layer 1 low and beamforming as the fourth function. And for reasons that will become evident shortly, I also show eCPRI here as a computing function. eCPRI is the 25 gigabit per second interface between baseband and radio units. Nokia entered the 5G era without any SoC capable of 5G Layer 1, Layer 1 low and beamforming and eCPRI. We had to resort to FPGAs for those functions. Back in '19/'20, our main challenge was to get out of using such expensive and power-hungry FPGAs as fast as possible. And by the way, FPGAs are also difficult to program for these use cases, especially when you have to optimize the use of the FPGA. So the first job of the expanded SoC workforce and our associate partners was to introduce SoCs for these functions with the fastest possible time to market. And they are now all available and shipping in volumes. We have now had the next generation of ReefShark SoCs in development for quite some time. These chips will be 5-nanometer silicon technology, which will be the best technology available at that time, because at that time 3-nanometer technology is not yet mature. And that's how you know you will be in the lead in terms of the use of semiconductor technology. We will continue to push aggressively this boundary as semiconductor technology develops further. The way we typically work here in SoC space is that we develop some of the IP blocks for the chips, the partner develops some of the IP and then the partner packages our IP, their IP and third-party IPs in an SoC design and takes it to foundry. There has been some discussion in our industry about access to the latest semiconductor technologies. We believe that being on path to 5-nanometer is critical to achieve the required capacity, connectivity, performance, size, weight, power consumption, et cetera. You can offset some of that deficiency with software and good algorithms if you don't have the latest silicon, but not all of it. This may over time create an advantage to those suppliers who have access to latest silicon technologies. Where we are now is the result of hard and focused execution by my renewed R&D team that I am so proud about, and our strategic priority to build technology leadership by improving our portfolio competitiveness over the past 3 years. I am extremely proud about how far we have come. We have done well to improve our R&D productivity, quality, predictability and overall capacity and output. We made significant progress in increasing our R&D capacity and output during '19, '20, '21. Our 5G R&D headcount was increased by approximately 60% from the beginning of 2019 until the end of this year and at the same time, our 5G software feature output in terms of so-called harvested hours and number of features normalized for feature size, increased by 130% during the same time period, outpacing the increase in our headcount and demonstrating improved efficiency and productivity. In other words, the feature output improvement is a combination of increased R&D headcount and improved R&D efficiency. At the same time -- and actually it's really part of the same phenomenon -- we have improved in all operational and quality KPIs, whether internal or external. For example, our feature build accuracy has increased from 60% to 90%. Definition of done or DoD compliance has shot up from 30% to almost 100%, and our total fault inflow reduced by 30% between second quarter of '19 and third quarter of '21. Then a few words on services, which are an important driver for our competitiveness. Many of our customers buy services attached to the products, such as network planning and optimization and deployment services, and of course, all of our customers naturally buy technical support services for the products. The more competitive our attached services are, the better the overall solution. The more profitable our attached services are when customers buy services attached to products, the better the overall profitability. What this also means is that we have embarked on shaping the mix of our business as we want to become less heavy on lower-margin deployment business, especially turnkey. With continued focus on digitizing service delivery, we can enhance speed and quality of deployments, maximize network quality, while reducing TCO for our customers and improving our margins. A couple of hot examples are explained on the slide in more detail with our work on driverless acceptance, site design automation and intelligent issue resolution with Nokia Digital Assistant with machine learning techniques. Our increased R&D capacity has enabled us to not only introduce new hardware platforms but accelerate feature development. Some features impact network performance, such as throughput or coverage, while other features enable new services such as Voice over NR or VoNR. Let's see how we have been doing this year in downlink throughput. In other words, the data speed from network to a device in 5G. The data in these graphs was crowdsourced by Tutela Technologies from one of the South Korea networks that has 5G radio from 3 different suppliers in different areas of South Korea. Crowdsourced data is meaningful in that there are millions of data points gathered by apps in iOS and Android devices in the background of using the apps. And it is therefore much more representative of the real customer experience than drive tests, where tricks can be played with deep packet inspection and policy control. The upper graph on the left is showing the evolution of average downlink data speed in the 3 suppliers' radio networks of this operator this year. You can see that downlink speed in the network supplied by Nokia, the blue curve, jumped during the summertime. That's when we upgraded to our release 21A. The bars at the bottom on the left show the average data speeds per city or region. This shows that the Nokia network again with the blue stack is present in important cities, such as Busan, with lots of traffic, making it more challenging to achieve high speeds. If you then look at the user experience, specifically more than just the basic KPIs, look at the graph on the right. The data on user experience in video conferencing and cloud gaming shows that the Nokia supplied area delivers very good quality, more of the dark green portion of the bars, and has also the lower share of the bad experience samples in the data. We take pride in optimizing network performance, not just for the basic KPIs, but the real customer experience per service. South Korea was the first country to launch 5G networks in the world on April 3, 2019 and we supply all 3 networks there in the whole market of one of our key competitors. Let's look at how we are doing in the rest of the world since then. Today, Nokia is a radio supplier in approximately 40% of all launched 5G networks and we supply half of all countries that have live 5G. Further, we have 59 5G standalone deals with 9 of those networks live. Since the beginning of 2019, we have won 30 completely new radio customers that we did not have prior to 2019. And in the same time period, we managed to increase our radio market share in 23 existing accounts, where we were already supplying. Since the beginning of 2019, to-date, we have won approximately half of all the geopolitical opportunities that have risen. All of this has helped us offset some of the market share loss that we experienced in North America and of course in China, though the latter was less detrimental in terms of margins. As reported alongside our Q3 earnings, our 4G to 5G conversion rate, excluding China, has therefore stabilized at around 90%, and we still have opportunity to grow it. More than anything, I believe this speaks both for the strong relationship we have with our customers and the renewed trust they have in our 5G technology. So a lot of what has been now discussed was focused on the CSP space, but one of the other success stories we've had this year is the continued growth of our private wireless business, which has expanded to now include over 380 customers, of which over 70 are on 5G radio. We have highlighted in the past the verticals we target in different industries. But today, I wanted to give you 2 concrete examples of customers that we are working on in this space. Lufthansa Technik has deployed a Nokia 5G private wireless solution to enable virtual engine parts inspection. Their technicians use 5G enabled mobile devices to allow customers to remotely attempt engine parts inspections. Entering pilot stage just before the pandemic, the solution quickly demonstrated its value and quickly moved from trial to business-critical infrastructure of Lufthansa operations. We have another brief video to explain more. [Presentation]

Isn't that fantastic? We have also just recently announced our work with Agnico-Eagle in Finland, who is using a 5G stand-alone network to bring connectivity, both on the surface and up to -- or should I say down to 1 kilometer below ground for its people, sensors, devices and vehicles. In fact, some of the machinery will be autonomous and remotely controlled over 5G. The solution will both improve operational efficiency and increase the level of safety for the teams working at the mine. We are still in the early stages of what we believe will be a significant opportunity in 5G private wireless over the next decade and we recognize that much more work is still needing to be done to prove the whole business, but we have the solutions today and we are working with customers, channel partners, distributors and CSP partners to prove the business case exists. I would like to finish this section of the presentation, before we start to look more into the future of our industry, with a brief revisit of what we said in March and how we have fared, including our financials in Mobile Networks. Pekka mentioned that these events are not intended to be focused on financials, but we still want to give you a summary picture of how we see it in my scope. We have clearly executed ahead of the schedule and much faster in 2021 than we expected in March, and I'm pleased to say that this has really been down to focusing on those factors that we highlighted on this slide at the CMD. As we strengthen our product competitiveness, maintain and build scale, manage the cost base, we do see a positive trajectory for our margins. Pekka has previously highlighted that we face some uncertainties related to our supply chain, both in terms of component availability and cost inflation as we looked into 2022. The challenging situation makes it difficult at this point to comment on the short term, but I want to make our ambition for the business long term clear. Whilst we presented a target at our CMD for 5% to 8% operating margin comparable by '23, our ambition continues to expand beyond that. Like I said, back in March, in this business, one should make north of 10% operating margin. And as we execute on our strategy and specifically on those focus areas of scale, competitiveness, technology leadership and efficiency in our cost base, we have a clear ambition to take this business group to a double-digit comparable operating margin. Let's now start to look forward. I would like to give you a picture of how we see our addressable market evolving over the next decade or so. We currently forecasted between '21 and '24 our addressable market for Mobile Networks will grow at a 2% CAGR, excluding China. We have been questioned about the sustainability of the current demand cycle, and whether '21 might be the peak, and whilst growth will likely slow down a bit, we do not see a peak yet and particularly as private wireless starts to become more material in volumes, we believe there is good scope for our market to grow. Even when we do see the 5G-related peak and CSPs building 5G for mobile broadband connectivity in wide area, we believe that the industry spending is then more stable for a few years as private wireless ramps up and fills the gap between that point with the CSPs and when 6G investments start to pick up towards the end of the current decade. I would also remind you of one other commitment Nokia made at the CMD, which was to grow faster than the market in our outlook for '23, and we believe we are putting Mobile Networks on that same trajectory relative to its addressable market as we push towards technology leadership. Preparing for future opportunity starts now. In 5G, beyond building the basic coverage and capacity for mobile broadband connectivity, our key focus areas include the following: industrial 5G use cases, private wireless networks, IoT solutions, sold through our CSP customers, through channel partners and in some cases direct; more advanced applications and services utilizing 4G, 5G slicing, AI and ML and new IoT capabilities; improved network efficiency with automation and energy efficiency improvements and network architecture evolution utilizing Cloud RAN and partly O-RAN. With 5G Advanced introduced for the second part of the 5G decade and cycle, the following. Actually, this one I will discuss in more detail on the next slide, but it will enable new usage areas, while also improving user experience and different services availability. Towards the end of the decade, we will then start introducing 6G. It is still early stages on the 6G journey, but we envision it to bring massively more capacity, new spectrum band such as sub-terahertz and very low FDD bands and a new golden band from 7 gigahertz up, requiring more sophisticated massive MIMO, adaptive AI-based air interface or AIAI and more use of machine learning techniques. But when the time of the 6G comes, we should not assume that we can all start from a clean slate. Our customers will want to ensure seamless evolution of network architectures, 5G, 6G products, user equipment, platforms, chipsets and software. It's going to be an exciting evolution from 5G to 6G. But first, there is still a lot to gain from basic 5G and 5G Advanced. So let's take a look at 5G Advanced. Here are some of its planned characteristics. Extending 5G coverage and connectivity with link budget, beamforming improvement, sidelink for vehicles and RedCap for wearables and industrial sensors. We actually use the core RedCap EMTC or extreme machine type communication in the past. Expanding 5G functionality with even 10-centimeter accuracy in positioning, timing synchronization and all sorts of IoT improvements. Enhanced 5G user experience for humans with edge computing and mobility improvements for use cases like cloud gaming and extended reality. Actually, extended reality can be expected to become a killer app in the 5G Advanced era. B&S Intelligence predicts that there will be over 100 million XR users by '25 and 1 billion by 2030 enabled by 5G, 5G Advanced. And finally, 5G Advanced is also planned to boost network operability, network optimization and network efficiency, including energy efficiency, for that matter, by leveraging data analytics, machine learning, centralized baseband unit resiliency, slicing enhancement and more. For the final section of my presentation, I want to address more directly a key question we get often from investors, which is what does O-RAN and Cloud RAN mean for Nokia's business? So let's start with some background. Open RAN Alliance or O-RAN Alliance was launched in June of 2018 by merging xRAN Forum with C-RAN Alliance. The Alliance also works in close collaboration with the telecom infrastructure project or TIP, whose mission is to deploy end to end disaggregated telecom infrastructure in varying environments. There are currently more than 290 contributors in the alliance, 30 of which are operators and many of them are my largest customers. The alliance is focused on enabling multi-vendor combinations through open interfaces that it believes could bring lower TCO along with increased programmability. Nokia has been and is the leading contributor in O-RAN Alliance. We are contributing to almost all O-RAN working groups and we are co-leading 2 of them; the group working on RAN Intelligent Controller, which is a new innovative network function, and the group working on this open front-haul interface between radio and baseband. It was Nokia who contributed this open eCPRI 7-2 interface between radio and baseband, or more specifically radio unit, i.e. the RU and distributed unit or DU. Here's a picture of the O-RAN architecture shown here vertically with the network functions and open interfaces, some of which have been defined in 3GPP and some in O-RAN. Generally speaking, O-RAN outlines an architecture with different implementation options, and nothing is mandatory, which adds to the variety. RAN Intelligent Controller or RIC is the only new network function with many potential use cases of AI and ML in network optimization, for example, a sort of API to the radio network. Here baseband has been split into 2 parts. You have the centralized unit or CU for non-real time-sensitive Layer 2 and Layer 3 processing. You have the distributed unit or DU for Layer 1 and real-time sensitive Layer 2 processing. And then there is radio unit for transmitting and receiving radio signal and converting it from and to digital signal. The open interface that is of most interest to CSPs is the open front-haul between RU and DU. Note that so far, I have not said anything about virtualization or Cloud RAN. O-RAN is really about splitting the base station, about the horizontal disaggregation even if shown here as a vertical stack in this picture. But as such, O-RAN doesn't take a stance on separation of hardware and software. In practice of course, companies making O-RAN compliance CU and DU software make it for hardware, using commercial processors as opposed to custom silicon, and that then makes sense to try and make use of benefits of cloud computing in CU and DU, those benefits which we know from Core Networks and IT functions. Remember that one of the objectives of CSPs is to lower the entry barrier for new suppliers who can then focus only on RUs, DUs, CUs, only on software or hardware. In other words, it can be said that O-RAN is about horizontal disaggregation of the base station, whereas Cloud RAN is about vertical disaggregation and separation of the baseband hardware and software. And actually, you can make O-RAN-compliant products, DUs and CUs with purpose-built hardware and you can make Cloud RAN, which is not O-RAN compliant. They, of course, often go together hand in hand. In this picture, the upper most example shows an O-RAN system with purpose-built hardware where the CU and DU are in the same baseband unit. The picture here is an AirScale baseband chassis to illustrate this example. Here in this picture, the O-RAN is connected to a cloudified core network, so you have the cloud symbol. The second picture from the top shows an architecture option where the DU and CU are both virtualized and in different locations with an open interface between them, the F1 interface. This shows a cloud symbol for virtualized CU because that function is easier to cloudify than DU and it's going to be even co-located in edge clouds or regional edge clouds with, for example, CDN, content delivery network, or gateway part of the packet core. The vDU here is shown with a picture of Nokia airframe open edge server -- but it could be any server -- a cloud server optimized for far edge cloud deployments. It is using commercial processors as opposed to custom silicon, like Nokia ReefShark. This virtualized DU reside at the cell site in a CSP network with the RU or it can be at most 20 kilometers away from the cell site, if we are to make use of the low latency capability of 5G. But it's not really a data center in the classical way of how people think about data centers. It's a small cloud server. And of course, the radio units then are -- well, they are radio, radio transceivers and they cannot be virtualized. Let's stop here for a second. What is interesting is that the market has now started to understand that general purpose processor, CPUs, are very expensive and power-hungry in Layer 1 processing in a DU. So-called hardware acceleration is needed. So CSPs have now understood and accepted largely that hardware acceleration with add-in cards is needed in cloud servers for Cloud RAN to make any sense. Otherwise, it's just too expensive and power hungry. And there are now several chip suppliers who have announced that they will make commercial O-RAN compatible chips with hardware acceleration for Layer 1. Interestingly, the big 3 webscale companies have had hardware acceleration for various workloads in their Public Service Clouds for quite some time. What is also interesting that even if one uses hardware acceleration, a full-fledged edge cloud or far edge cloud server has higher product cost and consumes more power than purpose-built hardware today. We can expect that this difference may diminish over time and of course, the overall benefits of cloud computing may eventually outweigh any remaining cost deficiency and power consumption deficiency of a far-edge cloud server. But in any event, this deficiency, together with the transport network capabilities and limitations is one of the factors that is currently slowing down the adoption of Cloud RAN. The third picture from the top shows what can be an enterprise use case where the virtualized DU resides in an on-premises or on-prem server with other workloads in the private wireless network at an industrial campus, for example. In other words, we believe that most of the so-called far edge cloud sites will actually be on-prem in case of enterprise or at cell sites in case of CSB networks. Once again, why are operators, some technology providers and even some governments after O-RAN? One reason is the programmability, especially the programmability of the RAN Intelligent Controller, especially the industry is into introducing open APIs to the architecture. Another reason is stimulation of innovation by allowing different players to focus on where they can add most value and potentially complement each other's capabilities. There is also the diversification of supplier base. Yet another reason is that operators believe that network prices would come down because of more competition or because of cloud hardware. Further, by having the option of using different suppliers for RUs, and then DUs, or CUs, operators can avoid costly swaps if a baseband or DU/CU solution must be changed for whatever reason, but by keeping and re-using the existing RUs. And some governments have their geopolitical reasons for supporting O-RAN in the supplier industry. At the same time, there are several engineering challenges to be solved. If base stations are disaggregated, then someone must re-aggregate them for features, for performance for security, for operational concepts and lifecycle management. And this goes beyond traditional interoperability or 3GPP network elements. The feature sets, the roadmaps, the different implementation options in the spec, the operability concepts, they have to be aligned adequately between the suppliers of RU, DU and CU. And this of course, has an operational cost still in the foreseeable future. And when things go wrong and from time to time they will go wrong, CSPs would want to have clear accountability for the performance of the radio network with one party. O-RAN complaint networks will have to coexist with non-O-RAN compliant parts of the network, just like Cloud RAN will have to coexist with purpose-built hardware at least during the time when they are being built. And then there is the topic of the cost and power consumption. The O-RAN challenger suppliers are using hardware platforms that use commercial chips and even FPGAs because the technology is still nascent and specifications are maturing, and the result is a hardware platform that has a cost and power consumption that is multiple of that of purpose-built hardware with custom silicon. So O-RAN and Cloud RAN is not necessarily cheaper, at least not any time soon. As said, the difference can be expected to diminish over time and then eventually, whatever remaining cost and power consumption deficiency may eventually be more than offset by cloud computing benefits in terms of automation, orchestration, scalability, capacity elasticity. Why then is Nokia so active in both O-RAN and Cloud RAN? First, one can look at this through the lens of innovators dilemma. If there will be RAN disaggregation then experience and history from business world tells us that it's better to be part of defining it -- be on the inside track, even if it means disrupting yourself. So we can now prepare our SoCs, our hardware platforms, our software platforms for Open RAN and Cloud RAN and prepare ourselves for benefiting from the disruption or minimizing any damage that it could cause. Second, there are the geopolitical market dynamics at play here. Some governments have outright banned some suppliers, while other governments are taking more subtle measures and while some operators are worried about these things and the security of supply. Many of these operators are even -- and even governments are mandating O-RAN as a requirement to replace some specific suppliers and also to enable and secure supplier diversity and choice. This is where we can win more than we would lose. You see there are many large radio networks in the world with 2 suppliers but without Nokia today supply. If operators are considering changing one of those 2 suppliers, the race will be typically between those who embrace O-RAN and not one of the incumbents. By being a leader in O-RAN among established suppliers, we can win in such replacement cases and we can also compete effectively against O-RAN challengers because we have better radio performance, better algorithms, which is important to maximize spectral efficiency to make the most of what is typically the CSP's most expensive and valuable asset. in other words, the spectrum. We have better robustness, better security, richer feature set, and so on. Third, there is the aspect of innovation. We do truly believe that RIC as a new network function can add a lot of intelligence and innovation to the ecosystem. Further, we already have fantastic multi-vendor references of complementary portfolios between Nokia and another supplier, where we have replaced one of our key competitors at large stadiums and arenas by connecting our baseband system to another supplier's special antenna systems, systems that we don't have in our portfolio and we shouldn't even have. These cases have been a true win-win-win for the CSP, Nokia and the antenna system supplier. When you add all these things up, we believe we have more to win than lose. So O-RAN is a big trend in our business. I referred to the 30 operators already. We are already engaged with more than 50 of our customers in this topic. And in general, and I said, if we play this right, we can win more than we would lose. So how fast do people then believe O-RAN and Cloud RAN would grow their share? For one, Dell'Oro says that the 10% threshold would be reached around '24, '25 for Open RAN and 25% threshold some 2 to 3 years later. Before '24, we would agree with Dell'Oro that O-RAN sales would not reach a meaningful level of scale. Note however, that it is possible that some CSPs will require O-RAN compliance in the contracts and O-RAN compliant products, but still buy both RF and baseband from the same supplier. So it's an option for the future. It can also be about performance. We expect multi-vendor O-RAN to proliferate first in lesser areas with 4TX radio whereas CSPs are typically expected to use the same supplier's massive MIMO radio and baseband in dense urban areas for maximum performance. One way to try to understand these opportunities and threats associated with O-RAN are to consider a few scenarios. In these 2-by-2 matrix, you can see 4 scenarios based on whether O-RAN technology maturity and performance will be low or high, whether operator demand for O-RAN will be low or high. If O-RAN technology does not mature fast enough or perform well enough, we can continue to lead the market with complete verticalized base stations. If O-RAN technology does mature fast enough and thus perform well or even faster than anybody thought, we can win more from the established competitor suppliers by embracing O-RAN more than the others, so that we -- and more than we would occasionally lose to any new O-RAN challengers. And you can draw similar conclusions with Cloud RAN as an adjacent topic and partly overlapping topic with Open RAN. And then to finish, there are 3 points I would really like you to take away from my presentation on Mobile Networks progress. Thank you for bearing with me. It's been a little bit techy today. First, we have delivered on the objectives we set for ourselves for '21 and the results are coming in even faster than we expected. We believe we have now completed the reset stage and we can now look forward to accelerating scaling up and helping Nokia's success in the future with our customers. We have increased our R&D investment, funded that partly by reducing SG&A. We have closed the 5G gap to competition and in many aspects, we have surpassed competition. We have what it takes to lead the market in 5G, 5G Advanced, private wireless networks, Open RAN, Cloud RAN and toward 6G. We continue to see robust demand for our addressable market. We forecast a 2% CAGR for our Mobile Networks addressable market through '24. And many of you have been concerned about Open RAN and Cloud RAN and what they might mean for our industry, but I hope I have been able to provide you with some reassurance that we know what to do and we can win more than we would occasionally lose. We are actively and deeply engaged in developing O-RAN and Cloud RAN. We understand the challenges they face, the opportunities they bring, and we believe we have the strategy that will serve us best. Thank you for your attention, and now back to David for the Q&A section.

So we'll now move to the Q&A session. And as mentioned previously, this event is focused on our products, technology and strategy, rather than a financial update. We would ask you kindly to focus your questions in these areas [Operator Instructions]. Thank you Tommi and Pekka. We will now move on to the Q&A. As mentioned previously, this event is focused on our products, technology and strategy, rather than a financial update. We would ask therefore that you would focus your questions today on Mobile Networks, considering we have Tommi with us. [Operator Instructions]

[Operator Instructions] I will now turn the call back over to Mr. David Mulholland.

Thank you, operator. So I think the first question is coming from Simon Leopold of Raymond James.

David, appreciate you folks doing this today. This is a good level of detail and helpful. I wanted to go into a little bit more on the Open RAN trend, particularly in light of the letter that European operators published a couple of weeks ago. I guess I'd like to get an understanding of the implications for Nokia. It's not clear what the operators are asking the European community to do and whether or not this means a potential for some funding directly from governments for Nokia. How you see this possibly playing out in the EU?

Yes. So it is indeed true that as O-RAN, like I said, the disaggregation would require that somebody re-aggregates for performance and features and operability concepts and so on, and that of course, does create some additional cost and workload, and I believe that operators have approached the European Union and asked for some funding for different types of labs and R&D work required for this type of O-RAN re-aggregation and making the different bits and pieces work together. In some cases, it is possible for suppliers also to apply for such funding, and we are of course looking at how to adapt to that opportunity.

Our next question comes from Aleksander Peterc from Societe Generale.

Could you maybe expand a little bit on your enterprise strategy in Mobile Networks? First of all, we've seen a bit of a slowdown in overall enterprise sales at Nokia. Is that also true in Mobile Networks? And how do you -- if you could just shed some color on that, explain that? Then as a side question, do you think your current initiatives with private wireless networks are the right routes to additional revenue streams in this area or would you also consider M&A, seems to be preferred by Ericsson at the moment? And then just finally, a very quick follow-up. Do you have any initial thoughts on what AWS announced yesterday with a private 5G initiative, is that going to be a new disruptive or new player in this area? Any thoughts on that?

Tommi, maybe I take that question, because the overall enterprise strategy is of course a group-wide question. And also, it's important to understand that when we look at our private wireless systems, that is the responsibility for the system solution is actually within the CNS Cloud and Network Services division, but of course, the Mobile Networks division provides the radio subsystem for that. Well, first of all, enterprise is 7% of Nokia sales. We expect that part to grow faster going forward than the CSP part, because obviously that is a fast growth market. We did have some slowdown in our own sales development in Q3, but we also did say the private wireless is actually growing double-digit. So we continue to be bullish about that market opportunity and this is actually validated by some of these latest announcements, including the one from AWS. There are 14 million industrial composition in this world and we did recently a study through Bell Labs, which came to a conclusion that 70% of world's large enterprises would invest in enterprise wireless in the next 5 years. Now this is a market that will have a different dynamics compared to the traditional CSP market. CSP certainly will be interested in this market, and we have a platform that we call service provider as a partner. But in addition to that, we are working with multiple partners to create an ecosystem for all of this. We are working with web scalers. We are working with other platform companies. We are working with the traditional distributors. We are working with systems integrators. The opportunity is big, it's a fast-moving market, and your question if I understood it correctly, that have we done enough so far? I don't think anyone can say that because it is such a fast-growing opportunity. We are doubling down on this, we have just some time ago decided to significantly increase our R&D investment in this and also go-to market investment in this and we are not announcing any specific M&A strategies, but definitely this is a segment where we are not ruling out any possibilities going forward.

Anything to add, Tommi?

No nothing much to it.

And we'll take our next question from Dominik Olszewski from Morgan Stanley.

Yes. Just one really on R&D for Tommi. In the slides, you mentioned a 60% increase in R&D headcount directed towards 5G. How should we expect that to develop towards 2023 and specifically talking about the opportunities of Advanced 5G and potentially the 4-quadrant matrix around O-RAN, because it requires some sort of flexibility? So how are you seeing the R&D headcount evolving there?

Yes. So as the headcount estimation for the future would automatically imply certain OpEx spend, I can't really comment that we will comment our future outlook then after the Q4 results. As to the O-RAN work, the O-RAN work is really, now that the chips are already -- the SoCs are O-RAN compliant, O-RAN work is really about software work and that's part of our plan.

We'll take the next question from Frank Maao from DNB.

Very helpful with all these details and also more longer-term thoughts that you're sharing. And I would like to ask a little bit about the next generation, 5-nanometer process node chipset that you've announced. First of all, do you expect this to ramp in the first part of 2023? Or how fast would you expect that to ramp compared to the current generation, which is of course taking quite a bit longer time? And then also, how important do you see that your customers see this type of technology road map with access to the most advanced cutting edge process nodes from the foundries, how important they see this release to have a credible tech roadmap compared to perhaps some of the competitors that for some reasons geopolitical or otherwise, don't necessarily have access to the same technology? Is that something that your customers is starting to appreciate the importance of having that type of credible technology roadmap? That's my main question. And if you could also potentially answer to some extent whether or not that next generation chipset would offer you opportunities to define out other materials of the building materials to compensate for rising chip prices, such as cooling or casing or designing out all the components to compensate for that? That would be very interesting.

Yes. Thank you. Yes, I should not go into the details of the exact availability dates of any of our processors. What I can say is that when we move to 5-nanometer technology, this industry is -- or these base station functions are probably not the first use case, you would expect some machine learning use cases to be really the first user of that for obvious reasons. But we are then riding that wave. And we believe that when we have 5-nanometer technology in our base station products, well, we will be in the lead and at that time nobody has 3-nanometer technology available. I should not really comment on behalf of my competitors. But generally speaking, experts do say that that access to 7-nanometer, 5-nanometer, then 3-nanometer as opposed to some of the previous process nodes, does have an impact on product cost, size, power consumption, weight because of power consumption. So you need a bigger heat sink and potentially some of the performance, because it could be that some of the algorithms can only be run -- they can be so heavy that they can best be run by latest silicon. And the third question, I actually have to now say I missed. What was the third one?

Frank, do you want to repeat the question?

Yes, it was just really about whether or not your customers really appreciate this type of credible tech road map compared to perhaps the ones that they are committed to that don't have the same kind of credibility road map?

Yes, I shouldn't speculate on what the customers believe or not. You're better off asking them.

We'll move to the next question from Francois Bouvignies from UBS.

I echo what my peers said about their usefulness of the presentation. I have 2 quick ones on the Open RAN, if I may. The first one is, you mentioned the requirements of some of your operators or customers around the compatibility of Open RAN, or Open RAN ready. So I wanted to ask you if you can give more color on that. I mean how much today you think it is in terms of requirement of your customers and more specifically how it works exactly, when you have an Open RAN-ready base station radio, how does it work? You can just enable the feature at some point, how easy it is? What do you need to do when you enable it, would be very helpful? And the second one is on the integration. You mentioned integration as an important hurdle maybe for Open RAN. So what's the situation today on integration? Do you see a solution in terms of possible players being able to do that? Is Nokia interested in doing that? Just some color on the integration hurdle and how you see things going forward?

Yes. Thank you. Excellent question and as well, like the previous one. So first is this O-RAN compatibility. So what it means in practice is that if I start from the SoC, the system-on-chip. There are certain things that you have to take into account there to make sure that they are O-RAN compatible and that's basically coming from the O-RAN specification. And then the same thing for the hardware platform, so that then the overall hardware platform can then implement or can support the open interface. The open interfaces are simply software. And then this readiness means that even if an operator would not mix and match different suppliers, RU, DU or CU, they could do so, so that they may buy RU, DU, CU from one supplier, but they would have the optionality to later change one of these and connect to the existing one. We have now made O-RAN compliance in the SoCs and the hardware platforms default in all of our system. So then what it means is that our customers can then software upgrade to this open interfaces when we have them all available. There is more than the open front-hall of course. Some customers may not need it immediately. They will take whatever standard eCPRI interface or the old eCPRI interface or they may actually even want to implement the open interface even if all the different elements come from Nokia. But the software upgradability is really the key thing there. Then to your question on the integrate…

Is it differentiation?

Pardon?

Sorry, it's just a follow-up. Is it a differentiation versus your competition, this feature that you offer?

Yes, it is a differentiator, especially in case a supplier has hardware platforms that are not prepared for eCPRI 7-2, because not all platforms in the market support eCPRI 7-2. And then to your question on integration, indeed, there is some work, especially this year and in the next few years, there is still some work in the integration to be done. One of the challenges in Open RAN today is that the number of so-called profiles has proliferated and almost exploded and a profile of means a configuration between -- or certain combination or permutation of certain type of radio head with certain type of baseband. So it is not yet so simple that you could just buy 2 different things and they would immediately work. So some integration work is required just for the basic interoperability and getting the performance right. And many operators are looking at us because of our system experience and our insights and our technical capabilities and they're asking if we could be the integrator. Then there are some operators who are thinking about doing it by themselves and even selling this type of integration service to the market.

We'll take our next question from Sandeep Deshpande from JPMorgan.