Nokia Oyj (NOKIA) Earnings Call Transcript

Hello, ladies and gentlemen, and thank you for joining us both in person here in London and those of you that are joining us virtually on the webcast. I'm David Mulholland, Head of Investor Relations here at Nokia, and I'm delighted to have Tommi Uitto, our President of our Mobile Networks business for today's progress update, and we'll be focusing obviously on our strategy, technology and the journey we see for the business, as we have done in all the events so far. 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 Factors section of our annual report on Form 20-F, which is available on our Investor Relations website. In terms of the structure for today's event, I'll shortly hand over to Tommi for his presentation, then we'll have a short Q&A session. And when that's over, those attending in person are welcome to join us for a brief drinks reception on the other side of the building, but you can follow us over. You will find the material for today's presentation on our website after the event. But with that, let me hand over to Tommi. Thank you very much.

Thank you. Thank you, David, and, hey, great to be here with you today. Great to see some real people. And for those of you online, I have to tell you that you're missing something. We have a beautiful sunny, crispy wintry day here in London. So we're busting some meets here. Every time I come to London to meet with the -- with British Telecom or O2 or Vodafone or Hutchison. It's always sunshine. So I don't know what all that talk about the British weather is. Anyway, so great to be here. And so, this mobile networks update is about building sustainable leadership in wireless. And so, what I'm going to be sharing with you is how we're going to further strengthen our leadership in wireless and how do we do that so that we can create value to shareholders on a sustainable basis. And this presentation will have 4 parts to it. So first, I will describe the dynamics of our addressable market, and especially the radio access networks and [ attached ] services market. Then I will explain why we believe that we are so well positioned for growth as to where we are today. And then I will go into further detail of what are those future opportunities that we see where we can grow faster than the market and create sustainable shareholder value. And then we'll quantify that as much as we can today without going into too much of guidance for the future for -- with some implications on our financial journey. So with that, let's take a look at the market. So this may be a familiar picture to some of you from the previous presentation that I made on the same topic. There are actually many points we can make on this chart. But this chart is showing the addressable market of radio access networks plus the [ attached ] services, and in constant currency from '21. So it's almost the entire addressable market of mobile networks. I also have the microwave radio link plus [ attached ] services market and some custom solutions, but this is the vast majority. And so, you see the evolution of the market value over the years. And if you take a look at one of the data points, we see that there would be a growth of 2% CAGR from '22 to '25, during that 3-year time frame. So modest growth in the overall market. This market captures both the mobility voice, mobile data of the operators as well as then the private wireless networks, which are built either by the -- or provided by -- either by the operators or then build provided by other players such as ourselves selling directly to the enterprise or through some other partners. It's interesting to see that, that 2% is sort of ironed out over the 3 years because you can see that there's a sort of -- there's a little blip there. If you would compare our view on what we think is going to be the size of the market in '23, it's some 10 percentage points. 10% higher than our view was about 1 year ago for year '23. So the market value in '23 seems to be higher than we still thought 1 year ago. So that would then suggest that there's growth in the market now from '22 to '23, and then we would come back a bit, about 2% CAGR over the 3 years' time. But what is also interesting that this hasn't changed from the last time, that within that 2% -- and then, if you start to extrapolate to what happens in the future, we see very significant growth in the addressable market in private wireless networks. So 28% CAGR from '21 to '30, so for the 10 years' time frame, and that's, of course, a very significant part of the market. What has not changed since the last time when we talked about this is that we see an extended peak in 5G. First thing I want to say that we don't think 5G market has peaked yet. It's true that there are some markets where 5G population coverage is good, take South Korea, take Finland, take maybe United States. But in most parts of the world, 5G coverage is actually really poor. Even in Europe, we have something like 50% population coverage, and in most parts of the world, significantly less. Some 75% of the 4G base stations don't yet -- have not yet been upgraded for 5G around the world. So there's still quite some growth there. Last year, you have those 2 sort of overlapping curves with a little time shift. You have this basic voice mobility and mobile data by the operators, and then you have the private wireless. And when you sum those 2 up, then actually, you have an extended peak in the market. So if we then get into the second part of why do we believe that we are so well positioned to capture growth going forward. We have said that we will be growing faster than the market. So let's just see first, where we come from, where we are today and what we're going to do next. So last year was the year of reset at Nokia. And for mobile networks, that was sort of natural continuation because we had already been running our mobile networks turnaround program in 2019, 2020 and first half of '21 as well. Originally, I thought that the turnaround would take only 2 years to implement in terms of the operation. But it took 2.5 years, but it got done. And then -- so -- but it got completed in '21. Then, when Pekka Lundmark and Marco Wiren came on board, they helped us design this new operating model, which then became effective from the beginning of '21, which has certainly made it a lot better and easier to manage this business. We moved away from this confusing matrix organization with blurred accountabilities and trying to manage the P&L somehow in the customer dimension and product dimension, and it was unclear as to who is accountable. So now we have the divisional model with the 4 business groups where they are almost like companies inside of company, and they are alone responsible for the P&L. And there was something else that was particularly -- or that was specific to mobile networks, which was that, we also started to operate the product business together with the attached services business and, of course, the network management part. Because that's really how our customers see the network and how they build the network and how they buy the network. So finally, we organized ourselves for mobile networks according to how customers do this, and that has made the business management a lot easier. With this improved P&L responsibility and accountability, we also got better visibility into the fixed cost, into SG&A, and that enabled us to challenge, to streamline, to save cost in SG&A. And in particular, in G&A, I mean, we didn't really reduce sales, customer-facing people, and we certainly didn't reduce marketing, but in the G&A space, especially. And I think we've said it, but it's remarkable that, when we increased our R&D OpEx spend in mobile networks in '21, half of that was funded from savings from G&A -- half of it. And then, despite of all of this, we have been able to do well with the operating margin, as we will see shortly. We -- also, last year, we launched some new product platforms, and this was particularly important because we had to introduce better performing, more cost-efficient, more power-efficient platforms based on system-on-chip technology. We had been relying on FPGA technology for many key functions in a base station. So now we were able to introduce some great new platforms that were very well received in the marketplace. Now we are in the accelerate phase, which is '22 and '23, and, of course, '22 almost done. But where we are is, we have created a high capacity and predictable R&D machine, and I will go into some further detail about that. And that machine is going to continue to improve. We are also going to introduce, and we have introduced this year also some new platforms, both on the baseband side and in the radio products, some new dual-band, triple band radios, and we will continue to do so in '23. Our market share has -- now in radio networks, has stabilized. For many years, we were losing market share. First, we stabilized it, and now we are on the upswing. So I will also show you some interesting data points on the market share growth and what we expect to happen. We have been making investments in supply chain resilience, partly, of course, because of the learnings of the semiconductor crisis and supply constraints, but also partly because of the geopolitical situation and how that might be evolving. And then, after this accelerate phase of 2 years, we talk about the scale phase after that, where, we believe, that we have what it takes to be the trusted partner for critical networks, both for CSPs and enterprise, as a supplier who has access to leading silicon technologies, be it 5-nanometers, be it 3-nanometers or like. Already today, we enjoy this position and this reputation amongst our customers, that we are a trusted partner, somebody who operates on -- ethically well, somebody who has security designed into the products and into our operations. And that is, of course, particularly important when we go for the private wireless networks that are going to be a sort of nerve system for not just humanity, but for businesses. So that is an obvious competitive edge that we're going to have. I will be talking more about this advanced silicon and why that is creating certain opportunity for us to grow faster than the market, because not everybody will have access to leading silicon. There are these growth opportunities that we have in the big picture in the long term over the coming years. Some of them may not be immediate, but they are there. Now the first one is enterprise. So as said, the private wireless network market, which is the biggest part of my enterprise business now and in the future, that is something that -- yes, growing by 28% CAGR, and there's really no reason why Nokia would not have a higher market share in private wireless networks than we have in the operator space, because in the operator space, our market share today, even if we stabilized it, stopped the decline, and we have grown -- we have started to grow. The reality is that our market share with the CSPs today is not reflecting the competitiveness of our products today and tomorrow, because it takes some time to recover some of the losses from the past years in the CSP space. But in the enterprise space, that's a nascent market. It's a new market, and there's really no reason why we shouldn't have a better market share there. Another reason for that would be that the enterprise customers, they want solutions. They want something that works. They are not in the business of making science out of slicing and dicing network elements into the smallest possible cube and then running reverse e-auctions to get them for the cheapest price for -- wherever. I mean, the enterprise customers want solutions that work. There's the geopolitical opportunity because, already today, we can say that -- and we can still say -- this has been said before, but we've done the recalculation that we have won approximately 50%, approximately half of all the value that operators have moved away from the so-called HRVs or high-risk vendors over the last years. And that has -- the decisions have already been made in places like Canada, United States, U.K., Australia, New Zealand, Japan, some European countries, some more publicly, some more discretely. And there are more of such opportunities for a couple of reasons. One is, of course, the overall geopolitical environment. And then the other one is what some people expect to happen to the competitiveness of those suppliers who may not have access to leading silicon or chip design tools in the future. Cloud RAN is an opportunity for us because, even if cloud RAN's technology has picked up maybe slower than some people would have thought, it is creating certain disruptions at an inflection point where we believe that we have what it takes to do better than the rest in this transition. And then in 5G Advanced, well, we now have the R&D capacity predictability, operational quality to make it big. And 5G Advanced will be a bigger thing in the 5G decade than LTE Advanced was in the 4G era. There are 7 particular future solution areas I'm going to describe, and I'll explain why Nokia will be the leader in 5G Advanced. So let's first take a look at -- in some more detail about what we have done to R&D. And -- so on the left-hand side, you can see the R&D OpEx expense, 12-month rolling going from -- about EUR 1,800 million to EUR 2,200 million on a 12-month rolling basis. So we have increased R&D spend. And you may recall that my President and CEO, Pekka Lundmark, said that we will do -- we will invest whatever it takes to restore technology leadership. And certainly, when Pekka came on board, he -- of course, he scrutinized our turnaround plan and what we want to make out of it. But once he understood it, both thumbs up and go ahead, and we continued to invest even more in R&D. But more than the spend itself, what is interesting is what has happened to the capacity in our R&D. So if you look at 5G R&D capacity, from end of '18 to end of '21, so in 3 years' time, we increased the head count by 60%, but we have also concluded that the overall capacity grew by 130%. So we have an average of 70% productivity improvement in R&D because of a lot of things that we did in the MN turnaround 2019, '21. And the best part is that it continues. So this year, I have another 11% improvement coming from productivity, even if we haven't increased our head count. So my R&D teams and my R&D leaders, they have this hunger and this ambition to take things even further. Of course, the steps are now smaller, it's more incremental and sort of continuous improvement rather than painting with the big brush, but they have excellent ideas that they still have going forward. So I have commitments from my R&D for further productivity gains in 2023 and '24. Knowing more or less based upon competitor intelligence, what our competition is doing, we have a reason to believe that, even if our product has already become very competitive, it will become, relatively speaking, more competitive relative to other suppliers in the next couple of years. One piece of the whole puzzle is, of course, the chips -- the system on chip. And there are basically 4 distinct -- 3 or 4 depending on what type of radio product it is, but there are 3 or 4 distinct computing functions in the base station where you need to have custom silicon, if you are to make the highest performance, lowest cost -- product cost and lowest power consumption product possible. So you have those 4 functions. And you may recall that we indeed had FPGAs in the beginning of 5G. But where we are today in '22 is that we have ReefShark SoCs across the board. And these were sort of time-to-market SoCs that we just had to develop as fast as we could with whatever we could make available. And some of them are 10-nanometer, some of them are 14-nanometer, some even 28-nanometer technology. By the way, you should always be specific, is it Intel 10-nanometer or TSMC 7, or the process nodes are not totally comparable. But you get my point, these were time-to-market SoCs at that time. Now the chips that I have in development today, where we -- in some cases, we have, of course, the samples already -- the engineering samples working, and then testing continues and products are introduced in '23. We have all bases loaded for all 4 computing functions. So layer 1, the physical layer in baseband, layer 2 + also in the baseband, RF digital front-end for the radio units, which, together with the analog front end is doing the analog to digital, digital to analog conversion, so that the thing can be processed in baseband. And then, in Massive MIMO radios, you have layer 1 low end Beamforming. So you have 3D digital Beamforming which requires also specific computing. So I have all bases loaded now with chips that are coming to the market in products, and mostly during '23 and partly in '24, but the same applies. I mean, it's not like there's always 2-year cadence across the board, but typically ASICs come with a 18-month, 24-month cadence. And these are predominantly 5-nanometer technology. At the time when 5-nanometer technology is mature, it's cost efficient, somebody else already figured out the early issues in some bigger industries and where they had a bigger need for 5 at this -- at that time, and at the time when 3-nanometer is not yet technically or commercially making sense, technically not mature and certainly financially too expensive. And then, we have started to already make the decisions and started to make the specs for the future next-gen SoCs in ReefShark, which predominantly are 3-nanometer technology. So you can see where we come from, from FPGA to whatever SoC to leading edge, bleeding edge SoC with 5-nanometer and then into 3-nanometer. And this, of course, has a big impact on product performance, product cost, power consumption, size, weight, wind load, this type of characteristics. So let's move on to South Korea and our network performances. So South Korea is a great place to make comparisons because it is the world's most advanced 5G market, very wide population coverage, very loaded networks, very competitive market. 75% of traffic -- mobile traffic already over 5G. Operators have ARPU increase with 5G subscribers compared to 4G. They've done a lot of things, right? Not the only place, but a great place. The reason why we like to compare -- and this is showing the downlink speeds and the uplink speeds in between 4 different suppliers from summer of '20 to -- or Northern Hemisphere, summer of 20' to summer of '22, is that there are 3 networks. We supply all 3 networks in some areas. There's another supplier who supplies all 3 networks, and then, there's one who supplies 2 networks and then the fourth one supplies one network. So that's why you can have 4 bars here representing the downlink speed and uplink speed of different suppliers. We all supply dense urban areas, which is the toughest. You have lots of traffic, lots of subscribers in those areas. And what this tells you is that, back in the summer of '20, probably no surprise to anyone, you really cannot say that Nokia would have been a leader in downlink speed. And in uplink speed, it doesn't come across that clearly here because all numbers are so low. But in the uplink speed, we were even 50% behind competition in the summer of '20, which was a big problem for us at the time. But where we have come now -- and you can see all suppliers typically improving because there's new features, new functionalities -- despite the traffic growth, we have been able to improve the throughput. But where we are -- where we were in the summer of '22 in July and after that, we have continued to improve. We already have taken the lead in downlink speed, and we have come on par with the uplink speed. So with this particular spectrum holdings that they have in South Korea across the 3 operators, and looking at just the 5G part, so not the combined 4G plus 5G throughput, we have been able to get to 1.1 gigabit per second downlink speeds in -- on average in these measurements. We've seen a 60% increase in the downlink speed over a 2-year time and 60% improvement in uplink in just 1 year time. And, of course, the key features to make that happen, if you like, technology would be 256 QAM modulation as well as sounding reference signal Beamforming in uplink. So those would be the key features to get on par in uplink. And of course, if we have been coming from behind -- and of course, others have not been standing still -- if we can keep up the pace and if others don't pick up pace, you can imagine what's going to happen next. Now if we look at the product portfolio, we can look at it like at a high level like this in radio. We launched those new product platforms in '21, so new Massive MIMO radio platforms, new baseband platforms. Starting from the Massive MIMO, which was particularly important for us because we were able to introduce SoC-based Massive MIMO products, very well received by the market, and we optimized the product for the 32 TRX use case because that is enjoying the biggest volumes in the world outside of China. Even in China, they have actually now switched from 64 to 32 to help some suppliers preserve chips. But 32 TRX is the biggest volume. We have the industry lightest product in the wide band with 400 megahertz bandwidth, high-power product. And that's important because, in radio, power, output power means capacity and coverage. So it is particularly important. And it doesn't stop there. So obviously, as I said, with the new chips, we can make that even better. In the base -- and what is important here when you think of these products is, in the radio products, the RF bandwidth instantaneous bandwidth, occupied bandwidth product volume in leaders, weight in kilograms, power efficiency, how much of the input power you can turn into output power, and of course, from our perspective, product cost and serviceability as well. Now if you move to the baseband, the layer 1, layer 2, layer 3 transport processing, what is important there is cell connectivity, how many cells you can support, subscriber connectivity, throughput, how many gigabits per second you can push through scalability of the product, headroom, product cost, power consumption and so on. And we believe that we have the market-leading product in baseband. Global data was analyzing our services capabilities. And one thing that we're particularly proud about is that, they concluded that the network planning and optimization, we are the market leader. What happens with these radio networks, even if it's sort of super sophisticated technology, but the real world out there is quite complex, and it's very difficult actually to make extremely high-performance radio network. So you need to be doing the radio planning properly and the radio optimization properly. And operators can do that themselves, and then there are some third parties who can do NPO, network planning and optimization. But we have proven it time and again then, that, when Nokia's engineers in our services organization, go and optimize the network with our digital tools, with our algorithms, with our digitalized tooling, we can get the best results in the market. And then the fourth example we pick here is our SON, the self-optimizing network solution. So that is -- the brand name is EdenNet SON. That has been found many times by many analyst firms as the market-leading SON solution, and also multi-vendor SON solution. So you can use this SON solution for optimizing the performance as well as, of course, the power consumption of the radio. And it's interesting that there are some die hard customers of some of our competitors who are using our multi-vendor SON to manage the radio of my competitors. So it tells you something about the competitiveness of that product. Energy efficiency has been -- it has always been important. Power consumption has been important, but now even more so. So our operator customers, in particular, they are really grappling with this topic of how to improve the energy efficiency. And what we are doing is, we are pulling all possible levers. You can say there's 4, 5 different levers that we can pull to reduce power consumption of the radio networks. This is important because in the mobile network, some 80% of the power is consumed by the base station sites and 80% by the radio unit. So it's clear that this is an area where we have to invest, and that's what we are doing. So first, we introduced new SoCs, as I said, when we move from 14 to 10 to 7 to 5 to 3 nanometers. That helps improve power efficiency. Then there's the rest of the hardware platform. So for instance, in radio, you would have improved power amplifier efficiency with better digital predistortion algorithms where we are investing a lot of PhD work. Another example would be our microwave radio link product, which is not today the main topic here. But the big European operator group has told us that your microwave radio link is clearly the most power-efficient, energy efficient in the market. That's hardware. Then we have software features, and this is linear software or deterministic software where a human being decides what features are activated and whatnot. So you have a lot of features that you can use to reduce power consumption. So you can be switching sales off on particular layers of different frequency layers. You can be doing micro DTX or micro discontinuous transmit, which means that at the millisecond level, you don't transmit anything if there's nothing to be transmitted. Kind of self-evident, but that's not how these systems used to work. And Massive MIMO or MIMO [ muting ] where you can -- for instance, one 64 TRX Massive MIMO has a 32 TRX when you just don't need the capacity. Or you can even go to deep sleep modes where the base station is truly in deep sleep, the one where the human being should get at least 1 hour per night. Some of us get even 2. And then, you wake up the base station only when there is some traffic. So very important features. But this is linear and deterministic. And there's something even better, which is AI and ML. So in our EdenNet SON, we use machine learning, for instance, where the machine is learning based upon previous weeks and previous days and previous hours as to when is the best time to start shutting down some particular parts of the base station, without impacting overly negatively customer experience. It's just fantastic solutions, and there's more. You can do that either with the EdenNet SON, or in the Open RAN world, you're going to use RAN intelligent controller, which is a new network function on -- sort of on top of the network. And then you have the rest at the sort of site level. So with our site solutions, we can make base station sites where it's 0 footprint. Everything is up in the mass. You don't really need -- everything is outdoor. You don't need a cabinet that requires cooling, for instance. Or if you have baseband hotels where you concentrate a lot of baseband units into the same location, you can use liquid cooling so that you don't need air conditioning which consumes power. Or if you have to use cabinets, then we have some all-in-one optimized cabinets that are optimizing the energy efficiency at the site level. And then there are some cool new stuff, which is like how to use battery backup systems together with artificial intelligence or machine learning. So for instance, we can run the base station on the battery when electricity is expensive, and then we can load the battery when electricity is cheap. Sounds very simple. Some of us do that nowadays at our homes. And -- but it hasn't yet been done much in this type of networks, but increasingly, it will be. So in other words, we have to pull all possible levers to make that happen. And then, now I think we have the video. Because you have listened to me, but there's a better way to do this, which is to ask or hear from one of my biggest customers in the world. [Presentation]

So that was Neville Ray, President of Technology at T-Mobile, U.S. We're super happy to work with Neville and his team. They drive us very hard, I'll tell you that. They are technically very capable and competent, and we would probably not be as good as we are today without them driving us so hard. But there's more examples. So here's a couple of examples from the recent highlights that we have made public in terms of the technology leadership. One feature area in 5G, where we decided to take the lead, is in carrier aggregation where you combine carriers from different bands or even the same bands into one carrier, for ever higher data speeds for the users. And with British Telecom here in the U.K., we were the first one to show this -- first one in Europe to show this 4 component carrier for CCA in the 5G stand-alone network, as one example. Or similar, we do in [indiscernible] so -- and then MediaTek as a chipset supplier -- one of the 2 leading chipset suppliers. We combined 300 megahertz worth of spectrum for a carrier to get world record speed in 5G, 1.5 gigs. Telia, Finland, has nicely combined 2 interesting technologies in 5G fixed wireless access as an alternative to, say fiber to the home, but with slicing, because obviously, you want to protect the user experience of the fixed wireless access users and the mobility service users, and make sure that it works properly. So they were the first one to do that with 5G stand-alone. Then the Google case is also fantastic. I just met with Google earlier this week, and just a great partnership we have with them. combining the best of both companies' technology assets. So people talk about slicing, yes, and you can talk about slicing, or like segregating a logical slice of a network without making a separate network. You can talk about that, that radio network level, or network level, radio transport core, which is something Nokia can do because we have the transport products as well, but you can even extend slicing into the device. So with Google, of course, having Android, we extended slicing into the device so that we were able to optimize user experience with some particular applications like YouTube over an Android device. Then we get into the area where we need rocket scientists like [ Josep ] from my team, to get into further detail in the Q&A, that would be AST SpaceMobile. So AST SpaceMobile is a company from the U.S. who, we believe, has probably the best chances of succeeding, making a very good commercial business success out of non-terrestrial satellite network for mobile devices. So they just recently launched the first satellite [indiscernible] nicely. They use our technology -- our base station technology with their satellite technology to create this service where they can extend coverage of operators today into previously uncovered areas in such a way that their customers can use standard devices instead of using some clunky separate satellite devices. So just a fantastic collaboration we have. And that's going to be also one of the areas of 5G Advanced where we will see some new developments in NTN, non-terrestrial networks. And then, looking a little bit further to the future than 5G Advanced, the 6G, Nokia was chosen as the leader for Europe's 6G flagship project on research called Hexa-X. And we're particularly proud about that in terms of what we do in research, and then the university collaboration and other technology institute, the collaboration we have in leading Europe's charge in 6G. Now as I said, our market share in radio has stabilized and is on the upswing, and I will show you some market share-related chart later. But you may recall that somewhere there in 2019, we started to report publicly a so-called 4G to 5G conversion rate, which was important for us to demonstrate how well do we manage to win the 5G business when our 4G customers are making their 5G decisions. So that if you have 100%, then it means that, basically, all of your 4G customers choose you in 5G. It can also happen that you lose some customers in the conversion, or that you win some new customers, and these are netted against each other. Now during the early 2019, first half of 2020, when we were still quite weak in competitiveness, we lost some customers. I mean, we lost share in radio in some -- a little bit more than 20 accounts, either partially or completely. We started then also winning new customers. And then what happened was, when all things netted and when we had some particular headwinds in the U.S. market, the conversion rate dropped to about 90%. And then, all things considered, it stayed -- it sort of stabilized at about 90% in the subsequent quarters, and we stopped reporting it at the end of last year. But now just for the interest in it that -- what is it today, I mean, we are now missing a couple of quarters from in between Q1, Q2, and this is now the October figure. Taking into account the new customers that we have win and increase share with some existing customers, we have actually now concluded that the conversion rate is up by 20 percentage points, from that 90% now to some 110%, excluding Mainland China, of course, because in China, I have like 1% market share, 2% at best. It is very difficult for any non-Shenzhen headquartered supplier to have any meaningful business in China. If you look at the biggest customer wins, so like I said, so we lost -- let's be complete here. We lost some share in some 20 -- a little bit more than 20 accounts, but we have won completely new customers in 38 cases in the CSP space. Of course, many more in private pilots, but the big CSPs are like these 38 completely new radio customers. And the biggest ones would probably be Reliance Jio in India, the biggest network in the world outside of Mainland China, 300,000 base stations, and we are going to be a very significant supplier to Jio network. They previously had -- in 4G, they had another supplier, and we managed to now grab a very significant part of the 5G business. And there's no swap involved here because it's 5G stand-alone from the outset. There's no need to swap the underlying supplier. AT&T Mexico, we also took share from another supplier, not the high-risk vendor, by the way. So we took share from another one. CBN is a greenfield network in China. We only won 4%, but it's 19,216 base stations, to be precise. We also won -- so that's a greenfield. Then Proximus and Orange in Belgium. That's a high-risk vendor, so 100%. Tele2, Telenor in Sweden, also high-risk vendors in network mobility joint venture, just like Proximus and Orange and the [ M ] WING's joint venture. And then, we increased our share in T-Mobile, Orange, Poland. They have their Networks joint venture. So that's also a competitor swap. Then we increased our share with KDDI and SoftBank in Japan, and in Bell in Canada, and we won back TELUS in Canada. So this would probably -- and British Telecom, of course, is a case where we were already supplying some 1/3. Now we supply some 2/3 of all the networks. So these are probably the biggest ones among the 38, plus 28 new customers or increased market share customers. And as said -- I forgot to say it, but, indeed, most of these decisions, almost all the decisions taken against us where you would be swapped, they were taking in 2009, before that even, of course, but 2019, before -- during my time, 2019 and the first half of '20. And there's always this time shift that, when there's a decision like that against you, you start losing sales immediately. When you win something, there's always a time lag from the decision to the deployments to the first revenue recognition, like my CFO, Wiren, would be able to eloquently explain. So then the supply chain resilience, where we have been working very hard to improve the resilience. And we have shared with you earlier that we had some supply constraints that were holding us back earlier in the year. Now in third quarter of '22, we shipped the biggest volumes in my business since the start of pandemic, and of course, since the beginning of the semiconductor crisis which hit us in early '21 -- hit everybody in early '21. And our sort of worst moment was at the end of second quarter when we had the biggest amount of backlog that we couldn't deliver on time, so it's really late deliveries, if you will. And at the end of third quarter, that had reduced by more than 1/3, some 34%, from my memory, and it is further going down now towards the year-end. We have been able to improve our inventory health so that -- we are not living completely hand to mouth with some components and then having too much of other components. So there's a better balance in the componentry. Supply remains tight, but that's, to a great extent, also due to the strong demand for our products, the growth that we see in particular in '23, for instance, in India, where -- I forgot to say, it's actually very important to remember that, I spoke about Jio, but we also won our 45% of Bharti Airtel who has the world's second largest network outside of Mainland China. So they have 240,000 base stations, and we kept our 45% market share in this transition. So also a very, very important deal. And these 2 are now raising, if you look at what they have publicly said about how many base stations they are deploying, when are they pan-India. It's an arms race, and of course, we are in a very significant part of it. Now -- so there are also some signs in the market -- in the component market that it's easing for a couple of particular reasons, more capacity coming online in the semiconductor industry, silicon wafers or substrates in particular, but also some demand decreasing like in consumer electronics or personal electronics. Car demand probably is also coming down, but that doesn't really help because the EVs and hybrids are consuming so many chips that it doesn't really give you any relief. We have been building resilience. So we have been, as I said, normalizing, help making the inventory levels and the inventory quality healthier, and we have been mitigating the risk of different geopolitical scenarios of what might happen. We have been building strategic buffer -- component buffers. But we have also been regionalizing not just the component purchase where we can, but also we've been regionalizing manufacturing and distribution so that we would always have at least 2 live dual sources or manufacturing facilities and distribution centers for any volume product so that, if something bad happens in one part of the world, then we can cope with the other one. And then, we've been improving the product design so that there's less -- there are fewer components that would have a single supplier, and that there are rather several suppliers for the same or the different type of component. So we have multi-source components as well as then better reuse across different platforms that we have in RF and baseband. And this work continues. This is important because we need to meet this growing demand for our products for '23, but also for the future because, remember what I said about the big picture fundamentals going for us, and not just going for us, we're making them go for us. So I spoke about the geopolitical opportunity, the cloud -- the private wireless networks, 5G Advanced and so on. So we need to prepare for that in our supply chain. So if we then look at the future opportunities for this, it is still fair to say that 5G is still at early stage of the entire life cycle. So we are basically 4 years into the decade, you could say, even 3 years because 2019 was a kind of [ hasty ] start. 5G should have been launched only in 2020, but there's many reasons why it was done like this. But if you just look at what's happening, it is still mostly EMBB. It is just evolved mobile broadband, some fixed wireless access, which is somewhat corresponding. It is still early deployments, a mass rollout in many markets, but coverage is really poor still in most parts of the world. It has been distributed RAN with some centralized RAN with baseband hotels. That's where we are today. But increasingly, there's the 5G enterprise opportunity. And this is very interesting that, out of the more than 500 customers we already have, more than 100 are using 5G technology in radio, even if there is no URLLC or this RedCap functionality. I'll just explain that shortly. So it's just plain vanilla 5G, and yet they wanted to have 5G for private wireless instead of 4G. I think that's remarkable. So we can imagine what will happen when we have URLLC and RedCap capable devices. So you have the enterprise growth in the -- with the private wireless, and then you have 5G Advanced during the second half of the decade, and I'll go to some detail on that. In terms of the architecture, you will start seeing some more Edge clouds with some further decentralization of platforms that are server-based, and hopefully [ cloud native ]. But indeed, URLLC -- so, it's interesting. It was actually standardized, already in release [ '16 ] -- in 2020, but the reality is that there are no devices yet available in the marketplace to support this type of ultra-reliable, low-latency communication wireless IoT. And that would be important for, say, wireless robotics, where it's really important to have a robust and secure connection and be able to have very low latency so that you can accurately locate and position a moving object like a wireless robot. So there are no devices for that yet, but they will come now in '23, and then increasingly in '24, which tells you that there's a lot of further potential in the 5G-based private wireless networks. And then RedCap, which is a different feature set for wireless IoT, that will be standardized in release '18 and then come to the market for the second half of the decade. That is for connecting a very large number of relatively cheap wireless IoT devices that consume very little power, so sensors and that sort of things. And -- so that will create further opportunity in the private wireless. There is a shift from 4G to -- 4G, 5G nonstand-alone to stand-alone 5G. That is happening now, and that, of course, then enables operators to make different use of the spectrum assets as well as introduce some new 5G specific services, like say voice over new radio. And then we will start seeing some 5G cloud RAN and hybrid RAN deployments. It is not possible to go overnight into Cloud RAN. In reality, purpose-built RAN and Cloud RAN will be coexisting, which has some important implications, and I'll talk more about that. And then we -- in 2029, we can expect the first [ 60 of our ] launches. I personally would hope that it would happen only in 2030, but we know how this works. Everybody wants to be the first, and then, there's a [ ways ] when you balance the time-to-market and the risk and all that. But we can, of course, see -- expect to see trialing proof of concept well before that, but '29 as a commercial 60. But plenty of work, plenty of mileage still with 5G technology -- 5G Advanced technology, because it's still basically just evolved mobile broadband, some fixed wireless access, some 100 on our side. We are the biggest private 5G network supplier in the world. Some use in the private wireless, but there's a lot more to be done. Here, you see the segments -- the industrial segments where we see private wireless networks potential. So you have -- we have already sold to more than 500 customers. Last time I checked it was 515. And like I said, 100 of them are a bit more with 5G. You would have power to electricity companies, water companies who would want to have their private wireless networks for coverage and security reasons, for instance, oil and gas, oil rigs. You can count how many oil rigs and multiply that. So you get that segment. Then there's mining, which means mining companies for their operations as well as mining equipment makers who want to use private wireless networks. That was one of our earliest segments where we managed to sell these things. Then you have the transportation segment where you would have railway operators, railway companies. You would have airport operators, airlines like Lufthansa, you would have airport -- sorry, maritime or seaport operators and cargo handling equipment like cargo tech type of companies. So plenty of opportunity there, again, for certain performance reasons, security reasons, coverage reasons. You would have manufacturing, wireless robots, for instance -- manufacturing, warehousing logistics. And then you have governments, public authorities and so on. You would have public safety, fire, police, first responders, ambulance. Also, government and defense, very interesting segment for us, as well as then some smart city use cases. I'm not going to go to detail of this, but this is a good example of utility case. So you can basically look at the private wireless networks in 2 different solution areas. So you have wide area networks which are covering wide areas, and then you have campus networks which are relatively small for very local coverage. This would be an example of a wide area network. So we won this project, 450connect in Germany. Germany-wide, 450 megahertz private LP network for all electricity and water companies. Beautiful. You would have a campus network underground by Posiva in Finland, built by Telia, and then digitize as a system integrator partner. So this is about building a secure private wireless network underground. Obviously, there's no other coverage there. And for security reasons and other reasons, they want to have it this way. So the Campus network underground using Telia spectrum. Let's move on to a very complicated topic, which is -- there's a lot of hype about this and a lot of misconceptions about this. And this is the 2 forms of RAN disaggregation, open RAN and Cloud RAN. So as a friendly reminder and a quick reminder, what are we talking about? You can say that Open RAN is about disaggregating a base station horizontally into 3 different functions. You have the radio unit, then you have some of the baseband in distributed unit, and then you have some of the baseband in centralized unit, or CU. There's also a fourth element of function, which is the RAN intelligent controller, which I spoke about earlier. It's not in the picture. And then this whole thing would be connected to Cloud Core, which is not part of Open RAN. But of course, there's open interface. There has always been open interface between core and radio, unless we go back to CDMA days when you had to have the same supplier for core and radio. And -- so that's Open RAN. But then you can think of Cloud RAN being vertical disaggregation. So you're separating hardware from software, and software, you are even splitting into different layers. Like you have the Container as a Service layer and then you have the -- you can have the application layer, like DU and CU function in this case -- DU and CU, and then you can have the orchestration. And -- so these are the different dimensions of disaggregation. The reason why these both have been picking up slower than expected is the following. Open RAN specifications are not quite yet complete and robust enough. They are not detailed enough. There are too many implementation options. It is not a plug and play today. You cannot go to the shop and buy somebody's RU and combine it with somebody's DU like [ Lego ] box. It takes system integration effort still. Another reason would be that -- and in that area -- actually, security is one area where the specifications still have to be made more complete. Another reason is that some of the new suppliers for whom the entry barrier was sort of reduced by splitting the base station into 3 parts, their progress has been disappointingly slow in terms of their feature set, their network performance in the field, as well as their product cost and power consumption, because they have to use this type of server hardware, which I'm going to explain shortly. Now we are committed to Open RAN, have always been. We have contributed more to the Open RAN specifications than any other company. We chair or co-chair 3 of the 10 working groups. And it's just that, whatever we do with Open RAN, we, of course, have to be clear in our strategy that, by opening the interfaces, we will need to win more from those suppliers who don't embrace Open RAN, or who are otherwise not selectable. Then we would lose to some newcomers. Otherwise, why would we do this, right? So that's the fundamental of our strategy. On the server side or Cloud RAN side, one of the reasons why Cloud RAN has not proliferated more, or faster, is that the server hardware is still very expensive in terms of product cost, hardware cost per sale, and it's also power hungry, power consumption per sale compared to purpose built. So purpose-built hardware still wins by a wide margin. Another reason would be that, people have not been able to yet make use of, or reap the benefits of cloud computing, as you would expect. I mean, the benefit of cloud computing here should be network operations efficiencies gained from orchestrating several different workloads, including the DU and CU. But if you don't do that, then what is the benefit of Cloud RAN. If you don't have that, then there is no benefit. If somebody knows a benefit, they should tell me. So we think that we understand the potential here. We have the capabilities to make this transition in such a way that operators get the best price and performance, the supplier diversity, the best of this both -- dimensions of this aggregation. And I'd just say that, in the Cloud RAN dimension specifically, there are 4 cornerstones that we have to address to make Cloud RAN successful as a business, as a technology, as an industry. So first of all, we do have to make sure that the server hardware is cost competitive in cost per cell and power consumption per cell, which it is not today and not in the foreseeable future. Number 2, as I said, Cloud RAN cannot be built overnight, and then you throw away the old. And that means they have to go exist. But the operator customers, they want to have service continuity. When they move from the purpose-built area to the Cloud RAN area or vice versa, they expect to have the same services, the same applications available to them and the same network performance. So you need to have proper feature parity and performance parity from day 1. And then, as I said, really the benefit of cloud computing in this sense, in Cloud RAN, is that you would get network operations efficiencies from orchestrating several workloads with the same orchestrator. And of course, to do that, you need automation and you need to have the DU and CU software as cloud native. And cloud native is not something which is a yes or no, binary, 0 or 1. There are -- depending on who you listen to, like 4 plus 10 cloud nativeness characteristics and attributes. And then the question is, in this scale of shades of gray, how cloud native is it? And most of the Cloud RAN solutions today in the market, the software -- the DU and CU software is simply not cloud native. It is not fulfilling the requirements of cloud nativeness in most cases. Number 4, of course, to make this truly a sort of ecosystem, then these different bits and pieces, both in the vertical layer as well as the horizontal dimension, would have to be interoperable and work well together. So if we look at our Cloud RAN strategy, I mean, we certainly want to make use of cloud computing in radio networks. And that is -- again, it applies to the DU and CU function because the radio unit is exactly the same, and most of the value is actually in the radio, with Massive MIMO in particular. So you have -- the radio is the same between Cloud RAN and purpose built. And -- so here, the CU and DU function, we are decomposing into micro services that can be run in containers. Then, as a cash layer for Container as a Service, in our blueprint solution we have Red Hat, because we have to have a blueprint solution. But we are accurately aware that some of our customers would have different preference. So they might want to have VMware or [ Wind River ]. okay? So we will make our solution compatible with that. But we also want to be sure that our Cloud RAN is compatible with the public or private service clouds of the hyperscalers, because they have their own cash layer. If you look at the hardware for DU CU -- the server hardware, it is okay to use x86 computing, general purpose computing for layer 2, layer 3 and transport function. It's not quite as good as purpose-built, but it's okay. And we have our own product, but we are not religious or jealous about it because that's really not the name of the game here, optimized for Cloud RAN use case. But we work with Dell and HPE to make sure that our Cloud RAN works, our DU and CU function works well with -- in their servers, If our operator customers want to build Cloud RAN using servers from HP or Dell. And the same applies to the hyperscalers. So we are working with them also to make sure our Cloud RAN, DU and CU, work well in their data centers if our customers want to move DU and CU workload into their private cloud or public cloud. And one point we want to make here is that, this layer 1 computing in particular, it is so challenging that general purpose processes don't do a good job and you need hardware acceleration. And this was something that was very evident to the hyperscalers and the server makers from the outset because, for them, it's every day. They know that you need to have hardware acceleration for specific workloads. They have a lot of smart mix in their data centers. And in any event, computing is going to a direction of workload optimized custom computing. But we have our Smart NIC card for layer 1 computing that can be inserted into any of these services. So then, 5G Advanced -- looks like I'm running a little bit late today. So 7 different feature functionality areas. We are -- the [ repertory ] 3GPP standard with the XR-related improvements as well as the uncrewed aerial vehicle solution area. But you see a lot of different improvements here. Like I said, this is going to be a big topic in the 5G era. And this just happens to coincide nicely where, when these features and functionalities have to be introduced, the old -- oldest 5G base stations start to be out of their life cycle. So there's a modernization round. We get to sell 5G base stations, again, both to our customers as well as to our competitors' customers in case we can use this inflection point. One such reason, of course, is the energy efficiency, because the oldest 5G base stations in the market, or in the networks, are not really that power efficient. So let's move to the last section, so implications for our financial journey. This is looking at the past 2 years. We have made significant financial progress in mobile networks in the last 2 years. We are very proud about that. If you look at the sales development, then, yes, you all know that what were our difficulties in the past and why we were not growing, and there were quarters when we were even declining. No need to go there. Now we have been growing in the last 2 quarters. And actually, in the product business, we have been growing in 3 quarters, and we are gaining market share. We're growing faster than the market, and we expect to grow going forward. Our gross margins have been improving for a couple of particular -- specific reasons, product cost reduction, services cost reduction, better product and service mix, especially becoming lighter on the low-margin deployment services business. So it's more geared towards NPO and technical support and the product. And then, we have also had some help from the regional mix. If you look at our '22 gross margins on a 4-quarter rolling basis, we're actually quite proud about that, despite of this inflationary environment, we've been -- and component cost increases, that we've been able to deliver that type of margins on a 4-quarter rolling basis. And then the last one is the operating margin where it has been hovering sort of there in the 9% on a 4-quarter rolling basis -- 9% area, despite the fact that we have increased significantly our R&D OpEx, as you could see from one of the charts that I showed. So that's what we've done in the past 2 years. Looking then into '23. So we know that many of you have had some concerns about how our business will develop in '23, especially because some regional impacts or regional mix effects. Already, it was said in Q3 that would impact in Q4, but in '23 as well. And you saw in our Q3 commentary that, at the group level, we believe that our market will grow, addressable market will grow in '23, and we will grow faster than the market. That's what we said. So I'm not saying anything new. We have had good success in India with Bharti, with Jio. It's a very big market in 5G, a key driver for the growth. But of course, we understand that you have had some concerns over then the margin impact of this regional shift. So we wanted to show you this type of structure as to how we're looking at this '23 today. So today we cannot give you any specific guidance that will happen with the fourth quarter results, or Q4 results. But just to help you understand how we are seeing our '23 developing. So if you take -- we still have one period, a little bit more than that to go. So we first need to see how '22 is panning out and where we are landing. But then, as directional vectors, yes, we expect a certain negative component of vector from this regional mix of the U.S. market being a little bit softer, and then the India market taking a bigger weight in the mix. But then we have the positives. We have supply chain normalization, which means less expedite fees, lower freight costs, as well as less open market purchases where you have to pay huge premium to get some components from other than your supplier where people are abusing the situation. So we will have less of that. We will have product cost improvement. We are now still in the middle of the annual price negotiations with the component suppliers in the semiconductor space. And of course, with the situation easing, with supply easing, we can yield some benefits from there. Plus, we have some product design improvements, new platforms, smaller improvements that we can do in product costs as well as in our services cost, thanks to the end-to-end digitalized tool chain. And then we will have the scale impact. Our OpEx doesn't really move because of -- in any material way, based upon the recent new customer wins. And that means that then, we can throw this OpEx mass over a bigger mass of sales margin and then get the positive impact on operating profit and on operating margin. And then there are some other smaller elements. And that's why we believe that this is the right way to look at our '23, and we now work with our customers to better understand what they're going to invest in, what other CapEx plans in '23, where are these component price negotiations landing in the next couple of weeks so that we can then give you better visibility as part of our Q4 results. And then we can look at a little bit longer term beyond '23. So first of all, like I said, we have stopped losing market share. We are on the upswing. So Dell'Oro would tell you that we have been growing our market share in the recent quarters in 4G plus 5G, excluding Mainland China. I also remember seeing in the Dell'Oro report that we -- in 5G specifically, we have taken market share now in 5 quarters -- consecutive quarters year-on-year. So we're on the upswing. And then, for the long term, how do we -- why do we believe that we can improve profitability with this? So the scale. So there are still some 4G customers we will win in converting to 5G. We have new customers we can still win because of the reasons I explained. The enterprise segment is there. The product is more competitive. It will become more competitive relative to other suppliers. There's further improvements, including also product cost and services cost. And then we have this 5G Advanced opportunity ahead of us. The product or life cycle maturity. We know from previous cycles that when you move from the coverage build out more into the capacity build out, margins will improve because you have more software, you have more capacity upgrades as opposed to cabinets and frames and this type of basic infra. And we believe we can maintain efficient cost base. I don't think we have yet maxed out on the benefit of the new operating model. So we can gain further improvements in efficiency -- in cost efficiency. So with that, I will have the last slide and the key takeaways. So like said, 5G market, from our perspective, has not peaked. It will have an extended peak. Like we said last time, we maintain that view, and we are well positioned to grow faster than that market, and we gave you the growth figures. We have clear plans on how we will do that, how we will capitalize on these opportunities in enterprise, in this geopolitical environment with Cloud RAN, with 5G Advanced in particular. We have made significant financial progress, and we believe that the scale effects, the volume effects can, in the short term, in '23, offset some of the weaker regional mix, but we'll provide more detailed guidance on that later. So thank you for listening this far. And I think we are now ready for some questions and comments and jokes and insults.

Thank you, Tommi. Thanks, everyone, for your attention during the presentation. We will now move to Q&A. And as a reminder, these events are intended -- more focused on our products, technology and strategy, and we've talked a little bit about financials as well. But obviously, please keep questions focused on our mobile networks business. [Operator Instructions] But with that, we'll take our first question from Andrew.

Just first one on the 5G sort of peak plateau point that you're making. I'm just wondering what kind of visibility you're getting today from your operator customers into their plans, certainly for next year, perhaps a little bit longer than that? I mean, I suppose first, for next year, are you getting better visibility than you normally would because of the component constraints, because of the lack of supply? Is that helping you in terms of the visibility? And then, perhaps a little bit longer term, this idea of a sort of longer plateau. What are you seeing in those early markets like South Korea in terms of usage levels and operators needing to come back to add capacity or not, as the case may be? And how does that compare to your expectations?

Yes. I wouldn't say that the visibility into our customers' plans for next year is any worse or better than it normally would be. They are all going through their budgeting cycles now. And we're, of course, asking a lot of questions about what is your spend next year. And we have some good understanding of what it probably is. Part of the reason is that, of course, during the semiconductor crisis, we have asked our customers to give us proper 12 to 18-month forecast -- at least 12-month forecast. So we've had that and they [ orders ] that. And so, I wouldn't say it's any different. One thing that we always look at in this business is in the big picture, what is the coverage development by country per network, how well is it evolving. And then, of course, with experience and insight, we can then also triangulate from that angle that how do we think operators will be deploying these networks. The supply side doesn't really impact. I don't see any of our customers making any particular comments on this way or that way because of supply capability. So supply situation is easing enough to not make people worried about that in terms of what they will be spending. It would be more about than the discussion of, will there be a recession, how big and where and how deep and how long and that type of things.

We'll take our next question from Sandeep.

If you see your slide on the South Korean performance as such really, which shows your position in terms of your product at this point, I mean you lost a very major U.S. customer in the past couple of years. I mean, does that position now help you at some of those big losses that you've had in Germany or in the United States, et cetera, really?

Yes. In Germany, it's a very long time ago since we lost any customers. So I know that some of our competition likes to remind you about how we lost the customer like 10 years ago. I mean, that's a very long time ago. So I wouldn't really go back to Germany. And in the U.S., of course, the situation is that we have stabilized our situation. So we supply on a continuous basis. We supply a lot of 5G to 2 of the big 3. And the third one, we are still an important supplier to them. They have a lot of product in their network. And we continue to work with them to see if there's opportunity for us to do more business together. This overall competitiveness, of course -- or the improvement in competitiveness means that we are then in a better position to win back some of what we may have lost in the past. And you see some of that. I mean, I mentioned TELUS in Canada. That would be a win back case, for instance. And we do see that there would be further opportunities, thanks to the improved competitiveness.

I mean, what I'm trying to say is that, as you said that you have some existing -- 75% you mentioned, of 4G base stations haven't been upgraded across the world. I mean -- I don't know what that figure is in the United States, and particularly with that customer and whether you could actually keep some of that footprint that you had?

Yes, I can't really speak on behalf of any of the operators there. So I'll just say that we work very closely with them. They are our strategic customer in many technologies, in many parts of the network. And of course, we always -- periodically, we discuss radio networks with them as well.

We'll take our next question from Peter.

You spoke about the reduction in the production cost this year. And obviously, we know that the new SOCs has had a significant impact. You say you have a new generation coming next year. How important will that be in time to have a meaningful impact on cost in '23? Or will that primarily be a '24 thing? And do you have any early thoughts on how the new restrictions on the chip side may impact the industry and yourselves?

Yes. The ReefShark chart was meant to be sort of the overall logic and descriptive of how this works, that you have basically, like 18 to 24-month cadence with these different ASICs. Sometimes you can design some of them in parallel. So it's not like all chips came this year and then next year, all chips come again. So they come a little bit [ interleaved ], if you will. But the new chips, they have an important impact on the product performance, the product cost, power consumption, size, weight. In case of radio, it's important, not in the baseband. And -- but we would not give you particular details about how big the impact is. But just generally speaking, it is an important thing.

The next question from Felix.

I want to ask about the rising energy prices in Europe. So I'm curious to hear in your conversation with European operators, what do you think that will happen to their CapEx plans heading towards next year? Will they sort of push forward CapEx to protect your cash flows? Or to the sort of situation even accelerate demand for certain software and hardware applications that could drive improved power efficiency?

Yes. I have not really heard of any significant discussions of shifting CapEx out because of the increasing energy prices. Operators know quite well that, for them to be competitive, they will have to have proper coverage and capacity, including with 5G. But then there is a very significant increase in the interest, and actually the pressure on our suppliers as well to reduce the power consumption of our products. And it can be pretty far reaching as well, in that some operators are even discussing shutting down 5G during the night time because 4G would be enough. Because in 5G, if you are to use Massive MIMO for those very high data rates, then Massive MIMO radios [ will ] consume power -- more power than traditional radios. Of course, they are more power-efficient per gigabyte transferred over the year. But if there's not so many -- if you don't really need the speed, then 4G would be enough. So I haven't really seen or heard a particular discussion of CapEx shift because of increasing energy costs. They are coping with that in other ways.

Sami?

I have a question regarding upcoming Indian rollouts. Can you promise that you'll make good money in India? I think a lot of investors are concerned that, as it has been a difficult market in the past, you might be faced with cost overruns. So I mean, how are you protected? I think about, for example, the agreements, the rollout plans are very aggressive and there is a risk for cost overrun, and you may not even know all the component prices yet.

Yes. Excellent question. But actually, it's a very exciting market, and I'm personally fascinated about the work that we're doing with both Bharti Airtel and Reliance Jio. And then, there will hopefully be a third one soon because you know that we -- Vodafone Idea Limited, they also acquired 5G spectrum. They have not yet started the rollout, but -- they haven't announced the contract get. But that, by the way, is an other operator where we have 45% market share in 4G. And it's very natural. Of course, we have to compete for that business, but it would be very natural for them to work with Nokia. In India -- there were many parts in your question, I will answer them all. Our services delivery organization is just absolutely fantastic in India. I mean, they do a super job in terms of the project ramp-up capacity of project management, subcontractors for the deployment, network optimization. We have always done very well in India in terms of the services delivery. If we have had project start-up problems around the world and in new projects, they have certainly not been in India. We can do -- from my memory now, I think we can deploy 12,000 base stations a month in India at the peak. So it is a very significant capacity that we have there. And then, in terms of the -- of course, we have been planning this for quite some time. It was always -- we didn't take it for granted, but we, of course, expected that we would become a supplier to Bharti Airtel with whom we have worked 20 years in 2G, 3G, 4G and now 5G. And -- but we also started to work on the Reliance Jio opportunity very, very early, and we made all the proper preparations to make sure we have then the capacity. And both projects are very nicely ramping up. I'm sitting in the steering teams, and we're looking at all the deliveries, subscribers, the work, this and that, and it's progressing very well. In terms of the financials, it is true that India is probably lesser gross margin quality market than some others. However, these were sound business decisions. They are improving our operating profit. They are improving our operating margin. There is no swapping involved. You don't have to modernize anything. You don't have to order swap or modernize your own base station, which always is something where customers try to get a discount. We don't have to swap our competition and give any swap discount. So Reliance Jio, for instance, it's a stand-alone. They can keep the underlying 4G network. So I'm very happy about how the India 5G battle is -- has progressed and turned out for us. Of course, it's a raise every day, but very good progress we have there.

We'll go to Artem.