Guy Daniels, TelecomTV (00:24):
Hello, you're watching the Future of RAN Summit and we have almost reached the end of this year's event. I'm Guy Daniels and it's time now for our closing summary programme where we wrap up the major findings and answer a few more of your questions. We are live, so if you still want to get in touch and ask a RAN-related question, then go ahead and use the Q&A form on the TelecomTV website, but you do need to be quick as we only have about 30 minutes or so remaining. Well, I'm delighted to say that joining me live for our final show of the summit are Neil McRae, Chief Technology and Information Officer at CityFibre, Francis Haysom, Principal Analyst for Appledore Research, and James Pearce, Editor at TelecomTV. Hello and a welcome back to Neil and Francis, who were on our earlier panels, and hello to James who has joined us for this special closing session.

(01:31):
So we started the summit with a panel that looked at the emergence of AI RAN, a hot topic this year obviously, followed by a discussion on RAN architecture for 6G, and we finished with a panel on how to modernise legacy RAN. So can we now try and bring this together? And I'd like to ask you how you see the RAN evolving over the next four or five years. Neil, if I could come to you first, what do you see as the future of the RAN?

Neil McRae, CityFibre (02:06):
Yeah, so look, I think the topics that we've covered this week are all kind of interlinked. And I think in RAN now, in my mind, it's more of a game of inches rather than silver bullets. So all of the things that we've talked about this week, including not forgetting some of the basics, are really what's going to help RAN evolve. I think we still have some big TCO questions, especially when it comes to AI RAN, but even still with just pure cloud RAN where in some geographies, the TCO, just to move to a cloud-based RAN is still quite a challenge. You have to believe that we'll solve that over the next few years and we'll see much more cloud-based architectures, but if I was to put it into five or six points, I think the first one is virtualising the infrastructure, getting the most out of the infrastructure in terms of RAN deployment and RAN utilisation, extending that into managing small cells.

(03:14):
I really believe that the next couple of years are going to be big on small cells as we get more and more demand in locations where the macro isn't really well suited. I think the second thing is, so we've kind of had the, some might say the Open RAN ideology, we've now got to, well, we're being pragmatic about what's really going to make a difference in that space where practically all the big vendors have migrated to leverage Open RAN. I don't think it's really brought us all the benefits that we hoped for, but I think we might see some evolution on that, but ultimately a pragmatic approach to that.

(04:01):
And this is the big question, is AI RAN going to be affordable enough or drive enough value that we see operators deploy at scale? We've got some operators doing trials maybe in three or four hundred base stations or slightly more, but I don't think we've really seen anyone commit to a network-wide deployment. And I think there's a lot more work to do to really show what the benefits of that are, both in the RAN, but also is there a way to monetise that AI GPU investment in other ways. And then finally, spectrum and spectrum evolution, I still feel today that from a spectrum point of view, there's much more to extract in spectrum in terms of how we use it, how we manage it, how we deploy it, and then actually the geography of the network. How do we optimise for the geography of the network?

(05:01):
Cities have changed, locations have changed, many things have changed, but we kind of still have the geography from 20 years ago. I really believe that's another massive opportunity that many network operators aren't looking at today. So those are kind of the three, four, or five things that I think we need to look at. And then finally, making sure what you've got works as well as it should. It still amazes me quite often how a base station might not be performing as well as it needs to. I think there's no excuse for that in today's network and we really need to move on from that. And then of course sunsetting all the legacy. We've got to be much more aggressive on that in my view and ensuring we take customers on the journey, of course, as we do that.

Guy Daniels, TelecomTV (05:52):
Great points there, Neil. Thanks very much. So moving forward there, evolving, but as you said, there's still some big questions we need answers for. James, I'm going to come across to you next. As part of TelecomTV, you're covering news on a daily basis for us. So what are your thoughts on what you're seeing in developments in the RAN?

James Pearce, TelecomTV (06:13):
I think what's interesting is a lot of the noise has been around AI RAN and we saw that at MWC. There were so many announcements. And what I'm really interested in, Neil kind of touched upon this, is AI RAN starting to steal the thunder from Open RAN and how do those two elements fit together? The industry has been really focused on Open RAN in the last few years, and then AI RAN's kind of came in and NVIDIA making a lot of noise there. And it raised a lot of questions as well about the role of the GPU and where that sits as opposed to having a CPU in the network. There's a lot of tasks that we get told by vendors that might need a GPU when actually probably doesn't at the moment. And I think just touching on something else that Neil said that I think is really important is spectrum and that's always been a big challenge for the industry and it's not one that's going away.

(07:03):
So how can the evolution of the RAN and all these new technologies that we're being told about allow the operators to get the most out of the spectrum because it's a limited resource. It's a challenging resource. And even though the sunsetting of legacy networks has opened up some bandwidths and we've had some come available from TV streams and things like that, there's a lot of new spectrum possibilities that still always, always we need more. So how do we go about getting that and how does the network and especially the RAN make sure that we get the most out of what we've got? So I think those are some of the really interesting developments that we're going to see over the next few years. And it's just what I think I'm interested in is seeing how the industry deals with those challenges.

Guy Daniels, TelecomTV (07:54):
Thanks very much, James. Yes, operators always say they need more spectrum since the dawn of time. They never have enough spectrum. Francis, let's come across to you. The RAN is evolving. We're trying to get the most out of what we've got. We've got prospects to look forward to. What's your

Francis Haysom, Appledore Research (08:13):
Take? I think the most important thing to recognise is that we kind of hit a bit of a crunch point, particularly with 5G. The reality is we haven't quite got those use cases that were promised or the new monetisation use cases that we had. So I think a lot of the emphasis, particularly in the new AI RAN area, is likely to be in that kind of getting more from what you have already invested. I'm seeing AI RAN, AI for RAN as being one of those key areas. How do you specialise capability AI inference in AI for the RAN to make it better in indoor applications, make it better, easier to manage in indoor applications, make it easier to manage, for example, for neutral hosts, etc. Because at the moment we still really have a RAN that is very much, as Neil said, it's based on 20 years ago, it was kind of macro cellular basis, and we've always had that difficulty in getting into those new niches.

(09:22):
So the opportunity in AI for RAN is, I think, one of the big opportunities. The question is, will we take it? We're very used to having a kind of one size fits all, fits all operators, standardised, etc. Are we prepared to look at those innovation niches in there? I do hope so, but that remains open. I will just make one comment about the other two areas of AI RAN, which is AI on RAN and AI in RAN. Both of those are actually really AI on RAN is really kind of an edge 2.0 proposition. We've got to get the economics right for that, for telco for it to make sense. Otherwise, we're going to go through exactly the same problems we had with the previous MEC initiatives. And similarly, AI in RAN is really a data centre, putting computing and network together, do they make one plus one equals three?

(10:26):
And again, historically, we failed to do that. So I hope those two initiatives can build, but we've got to look very much better at the economics of it and the stickiness for a telco versus somebody else doing it.

Guy Daniels, TelecomTV (10:40):
Thanks very much, Francis. I tell you, we might have got better at that, but we haven't got better at naming some of our technologies and solutions and approaches. I mean, AI for RAN and RAN, on RAN, utterly confusing. I hope someone grasps this and really sorts it out. We could all do with a refresh there. But until that happens, we've got some additional viewer questions I'd like to put to you. Here's the first one we've got in about an hour or so ago. Francis, if I can come to you on this one, this question asks, how mature are the associated RIC platforms, AI schedulers, and data pipelines, and what gaps still prevent large scale deployment? I

Francis Haysom, Appledore Research (11:23):
Think it's important with RIC to actually distinguish between two RIC. There are two RICs or they certainly are. There's the near real-time RIC. Now, that's really about applications controlling the controller of the RAN. I think that story has kind of passed. Yes, I think there will be innovation in that area, but it will be very much tied to the actual vendor that is delivering the RAN equipment. I think as far as the non-real-time RIC, there's huge opportunities there, but they're really in the kind of ... Initially, they're really kind of super SON applications that are optimising the whole network and ideally using applications to better manage and change things in the network on an ongoing basis rather than on a weekly or monthly basis. I think there's big opportunities there. Again, though, I think the way to characterise this is that one of the challenges with Open RAN has always been this kind of, do I want innovation or do I want insurance?

(12:38):
And innovation, I can get lots of vendors coming in, lots of our apps, for example, working in my area, but how do I ensure that they work together? How do I ensure if something goes wrong with them, who is to blame? Again, we have to acknowledge a lot of the industry is actually quite conservative. It just wants a RAN to work. And therefore, how do you balance that kind of a company like an Ericsson or a Huawei saying, "I will own the whole thing. I will insure you. I will come in if it's a problem and own the problem," versus lots of innovation and bringing that together. And I think that's really, the RIC platforms are there, but it's your ability to verify, ensure, validate that all these things work together that I think is the biggest problem in terms of the implementation here.

Guy Daniels, TelecomTV (13:34):
That is a great point, Francis. Thanks so much for making that one. Yeah, we sometimes overlook that and forget that. Thanks for answering that one. Another viewer question we've received, let me read this one out. AI and ML have been used in the RAN for many years, so what makes this new AI RAN so different? The last panel spoke about the huge increase in compute that will be available to operators in the RAN. Is that really likely and how can operators use this? Well, Neil, I think it was you who mentioned or brought up this subject in that third panel. So I'm going to come across to you. Can you handle this question for us?

Neil McRae, CityFibre (14:14):
So look, I think what's different about it is kind of, and it's almost linked to the previous question as well to some extent, which is how can we bring more real-time analysis and AI power to the RAN in real time? So how do we adapt a radio network from when it's commuter time to when it's night-time, to when it's a stadium event, to when it's indoors or outdoors. And I think that the difference in compute, I believe, will make that journey a lot more easier. Is it a fully fledged 100% certain thing yet? No, I don't think it is. I think with some of the things that are possible with the RIC and with AI RAN potentially, I think you could see a scenario where you can adjust the network dependent on certain scenarios. We've got a long way to go on that though. I think anyone expecting a sudden boost of capability in the next 18 to 24 months probably will be disappointed.

(15:23):
But then there's the other capabilities of what applications can we use that use the network to connect, but can also use that kind of GPU and AI capability to create new use cases or new solutions for today's problems. And I think for me, that's where I'm interested in terms of what the application builders potentially could use this for. Things like real-time digital twins, managing traffic or managing crowds or managing, even the journey that you have through a shopping mall to ensure that every footstep you take is a valuable one. Using this type of technology to run a logistics centre from a health and safety point of view and having the power to analyse and manage that. So one of the big winners in 5G, and sadly, there's probably not so many we would like to call out is transportation and particularly ports where we've built 5G, both public and private networks and some hybrids, and we've used some compute power to help them run it.

(16:35):
I believe that we'll get more and more out of that, but I think still a question is, is it enough, back to Francis's point about the economics, is it going to be enough to justify all this investment or is there a way that we can do this investment in a more staged approach where perhaps we don't put GPUs at every single base station, but we maybe put them in certain locations that are close to the radio network. I can't help but think as we get more compute power in our devices and not just handsets, but other types of devices, industrial devices, and we get more compute in the network, then working together to ensure that the connectivity is as good as it needs to be for their application, for me, feels like the right thing. I've been saying for over 10 years, the network is too complicated for humans to run it at a real-time view.

(17:29):
And this is where my hope is, is that AI comes in and allows us to keep pushing the quality and the capability of the network in a way that we as humans today, we're just not able to do anymore. But I think it's early days and I think there's some risk in that, but I also think there's some opportunities and some of the different use cases that we might be able to bring to bear or at the very least, being able to ensure that we're getting the maximum out of all the assets that we've deployed in the radio network.

Guy Daniels, TelecomTV (18:01):
Thank you very much, Neil, for those insights. I will come to James, but Francis, I'm going to come across to you next though for your input on this one.

Francis Haysom, Appledore Research (18:10):
I think the question actually summarises it very well. We've been doing AI a long time. We've been doing it in the RAN at the edge of the RAN. I think the really only difference that a GPU has, yes, it's a newer technology and I'm sure Jensen wants to sell us one of them, but I think the important thing is to recognise it's a programmable device that is programmable by others. Typically, the device that has been in the RAN up till now is vendor specific. We can build AI into that vendor specific, etc, but equally the advantage of a GPU is it's an open platform that we can build other applications in, but it comes back to the original thing about insurance versus innovation is does the CSP want that kind of innovation in close to the RAN? Does it want that kind of inference to be driving the RAN in that area, in which case the GPU is probably a better platform because it is not a proprietary ... It's proprietary, but it's an open platform that you can program things on versus the closed platform tied to the existing network equipment vendor.

(19:20):
But if all you're going to do is build on the basis of stack, it's another form of silicon on which a RAN is being built by a proprietary vendor.

Guy Daniels, TelecomTV (19:33):
Yeah. Thanks very much, Francis. We were talking about the programmable RAN years and years ago. These things take time to come through, aren't they? James, I promised to come to you as well. So with all this extra compute available or potentially available at the RAN, are operators able to use this? Are they likely to use it? What are your thoughts? What are you seeing?

James Pearce, TelecomTV (19:53):
I mean, it's really interesting because, and we've kind of touched on it, the AI, the grand story of AI is actually that the vendors and the people who are working on AI telling us that this is something brand new and actually we've been doing it for years. It's the great myth, but there are new things and new opportunities that are presented. And I think one of the ... Kind of coming back to what I said in the first answer around spectrum, I think particularly here, spectral efficiency is something where we're already seeing tests that are interesting. Some of the tests in poor coverage areas, I think we covered something on TelecomTV that found that there was up to 20% efficiencies shown by using AI RAN in low and poor coverage areas, which isn't inconsequential at all. And if you start to think about that and then the ability to pick areas of the network and select certain bandwidths and to use sensing to really make the network a lot smarter, then it opens up the door to all of the applications that Neil was talking about.

(21:01):
And I think as somebody who's written about enterprises for a lot of years, that's where it gets interesting and the monetisation options become available because if you have the next generation use cases that are going to connect to it, so autonomous vehicles or robots or things like that, that are going to need low latency networks and you start to think about networks that can actually be programmable, actually be really smart, much smarter than they've been previously and AI will play a role in that. Then we suddenly start seeing a picture emerge, which makes a lot more sense and maybe can generate money for the operators, which is always the big question because these things are going to cost a lot of money and a GPU is not a cheap thing to invest in, as we all know. So if there's a reason NVIDIA are being valued so highly at the moment, and if we don't find a way of getting an ROI on that, then that's going to be one of the big challenges as to whether this will all work out or not.

Guy Daniels, TelecomTV (22:02):
Certainly will, James. Thank you very much for those insights. That's great. I just want to go back. Francis, I'm going to come back to you. You just quipped there earlier on the answer about Jensen wants to sell us all GPUs. Absolutely he does. We've got a question that's just come in. It came in since we started the show here. And the question says, look, if you go to run your RAN on NVIDIA GPUs, aren't you just back to a lock-in supply chain again? Are we locking ourselves into yet another supply chain, a vendor supply chain? Have you got any thoughts on that one from your evidence?

Francis Haysom, Appledore Research (22:41):
To some extent you are. I think it depends on what the ecosystem you're building on top of that silicon, etc. To some extent, it does depend how much you've disaggregated the underlying capability. So there may be alternatives in terms of the silicon, if you've at least some degree, decoupled the hardware from what the application that's running on it. I come back to the fact that whilst the GPU is from NVIDIA or from wherever else is proprietary, it is at least a programmable thing that a wider set of developers and application is available to you. So you do have that wider spectrum. But again, coming back to innovation versus insurance is, are you willing to take that view of opening up your ecosystem that can run on that GPU, which place you haven't tied yourself, you're not into a proprietary platform. But if you have made it that way, then yeah, it's proprietary silicon to all intents and purposes because of what you've put on top of it.

Guy Daniels, TelecomTV (23:52):
Great. Francis, thanks very much for answering that question for us. Much appreciated. Okay, thanks everyone. Before we get onto another question, we do have a couple more. Let's take a look at the audience poll that we have been conducting during the summit. And as usual, the format is one question, seven answer choices, and you can select as many as you think are appropriate. So we asked you which areas you think are most important for operators when evolving their RAN over the next five years. And there we go. This is what you've told us. These are the votes in so far. No surprise that AI RAN is right out in front there. Also, strong interest in improving energy efficiency, but at the back of the pack is Open RAN, quite interesting. Oh, and also preparing for 6G. Make of that what you will. We're going to keep this poll open for the rest of the week.

(24:58):
So if you haven't yet taken part, then please do so. You'll find the poll on the TelecomTV website.

(25:09):
So that was our viewer poll, which we run every single year. Perhaps no surprise that AI RAN topped the poll there. And it's also probably no surprise that quite a few of the questions we've had from our audience are AI related. And Francis, I just asked you that one there about the lock-in supply chain, but we've had others. I'm just going to read another one out here. Well, I'm going to merge a couple really, because a couple of questions are asking about traditionally operators have been quite conservative about sharing their RAN data, but when you look to the AI hunger for data to train RAN models and the use of AI to compare different RAN configurations, are we opening the door to security issues, possible copyright issues? And we've had all this with the training phase of some of these foundational models over the past couple of years.

(26:05):
As we start implementing AI in our RAN, does this open the door for lots of other complications? Neil, am I able to come to you just for some general thoughts here as to what this about all might mean for the future of the RAN?

Neil McRae, CityFibre (26:20):
Yeah. I mean, I think that's ... So first of all, security risk, I don't think this introduces any big new security concerns that we're not already day by day working on ... Since we moved from telegraph to the internet to IP networking, we've always made them secure and resilient. And actually that's what telcos do that no one else can do, frankly. That doesn't worry me, but of course we can't be complacent. I think in terms of some of the other challenges around data, I mean, I've worked in practically every operator on the planet here in the UK, and I would say none of us are great at managing our data. So I think if there is value in that and we can translate that and work with companies to help us manage that data better to improve the network, then I see little risk in kind of copyright or anything like that.

(27:22):
Sure, there's some privacy data that the radio knows where you are, etc. We need to be careful about that. And again, I think telecommunication operators are the kind of geniuses of doing that. We're heavily trusted in the global community, and I think we need to manage the privacy concerns and ensure that they are taken care of like we do with practically every other service. But if we can somehow leverage that data ... And I've worked in many operations where we've tried to monetise the data, and I don't think anyone's really been successful at it. So maybe we can't monetise it for cash, maybe we can monetise it for experience or for cost out to save us money. And I do think that there is an opportunity there. And I think that's where AI comes in. I mean, we made, I think, an unfair comparison to the ML and SON days.

(28:22):
I mean, if you ever looked at the SON platform, it was heavily rigid rules. There was very little flexibility in it, and I think to compare that to today's AI is like comparing horses and cars, frankly. So in my mind, I think that data is valuable to the operator and improving the operator service. And I think there's many ways that we can leverage that data, both in real time, but also as leveraging more and more digital twins, which I do see many operators doing. And I think there's some value in that in terms of capacity planning, planning your capital investment effectively, reusing components from one place that's not so busy to another. I think there's an opportunity in all of that, but we've got to be able to curate and run and manage that data. And it's not something I think most operators, and I've worked with practically all the top operators across the world now, and I would say none of us are brilliant at that.

(29:25):
I think there's more opportunity to come.

Guy Daniels, TelecomTV (29:27):
Great. Thanks very much, Neil. Appreciate those comments. And Francis, thoughts on opening up the RAN to AI and privacy and security and copyright?

Francis Haysom, Appledore Research (29:38):
I think Neil's already really covered that we can do this. I think the only point I would add to Neil's is the clearer we can be about what is the business model. And it doesn't mean that telco makes money because it gives its data. A lot of these can be in terms of better RANs worldwide. How can a mobile operator in the UK, for example, benefit from information about similar equipment based in Japan in different scenarios? Can we see patterns? Can we see specialisations? At the moment, we're very limited. We're very siloed and we are very, very protective of the data, seeing that as our crown jewels. We need to flip the model and say, actually, there's a benefit to us actually understanding, I don't know, what is the worldwide performance of Ericsson equipment versus Nokia equipment across the globe? If we can see those things, we can see the patterns, the improvements, where it works, where it's less efficient.

(30:45):
That's a benefit to every telco globe-wide. And that doesn't necessarily have huge privacy applications. It has the belief that sharing the data actually gives us better results. But it comes back to what is our business model? What are we trading here and what are we hoping to get back?

Guy Daniels, TelecomTV (31:04):
Thanks very much for those comments, Francis. We probably have just about time for one final question. We've got three minutes or so to go. So let's have a quick question here. We've spent today looking at the future of RAN and the many opportunities for telcos, but what about the challenges? Can we identify some quick challenges, some danger signs to watch out for, hype to ignore, dead ends to avoid? James, I'll come to you first. What can you tell us about some of the challenges ahead?

James Pearce, TelecomTV (31:37):
I mean, I think we touched upon one of the biggest ones, and it's one that we've written about a few times on TelecomTV recently. It's that challenge around where to use GPUs and where to use CPUs and how to mitigate those costs. I think Neil made a great point about the fact that we may end up with networks where we have a GPU on some specific targeted sites and then stick to a CPU on other ones because the fact is that they're really expensive. And if telcos don't need anything else, it's massive investments. Because it's an investment-led industry, they don't need an extra cost that isn't actually going to bring an ROI. And I think where we're kind of sitting there, it's not just the cost of the actual chip as well. It's also the cost of the energy to run them. And we're really having this kind of pushed on us as an industry.

(32:28):
Francis mentioned NVIDIA and Jensen trying to sell to everyone, but I think as we kind of look at that, if we step back from it a little bit, we've got to work out exactly what the benefits of AI RAN would be because that seems to be where the industry's going. And the poll that you mentioned shows that that's where the interest is at the moment. So if that's where we're going, what are the benefits? How do we get the most money out of it? And make sure that it's not just another rabbit hole for us to throw money at as an industry which we don't get anything back from. So I'm interested to see where that goes, but that's a concern at the moment, I think for a lot of our operator partners and friends.

Guy Daniels, TelecomTV (33:11):
Thanks very much, James. And Neil, can I quickly come to you as well? Any last thought about what we've got to avoid or watch out for?

Neil McRae, CityFibre (33:22):
Yeah. I mean, I think it's back to Francis' point, which is really about getting the business models right and ensuring that we ... We're very good in telecoms at inventing science projects that we get everyone behind and we have a signing ceremony and everyone's like, wearing a logo, but it doesn't really come to much. And I can't help but think that we need an approach that's more about kind of failing by doing rather than talking about stuff and then hoping that it works, which is kind of the telco way. If you've ever been to a kind of KubeCon, the way that that kind of community works is by trying stuff and trying it and iterating and iterating it. And I kind of think that's the approach that we need to take as we move forward in the space. As I said at the start, I don't think there's any silver bullets in the radio or in mobile networks at the moment.

(34:25):
So all the inches count. And if you've got more people trying to develop across a wider kind of ecosystem, then I think you'll get a lot more inches and you'll get a lot more benefits. So let's not try and find the next kind of poster child that we're all talking about. It was NFV, Open RAN, now it's AI RAN. Meanwhile, there's a lot of other things that we could be optimising and working on that kind of gets left behind. And if you think about, we've been talking about GPUs and AI. One of the biggest parts of network growth over the last few years has been all around data centres. And I think from a telco perspective, we haven't really banked much value of that when, and perhaps we should have done. So how do we keep an open mind to everything that's going on and how do we really take a much more incremental try and make it better and iterate and iterate and iterate such that we're consistently learning, we're consistently trying.

(35:29):
And it doesn't feel like we have to wait for the lowest common denominator for everyone to agree forward that we've just done too much of that over the last 10 years in our industry.

Guy Daniels, TelecomTV (35:40):
Yeah, absolutely, Neil. Thank you very much indeed. This is the way. That's what we've got to remember. This is the way. Francis, quick final comment from you before we close the show.

Francis Haysom, Appledore Research (35:52):
Yeah, I'd really raise a couple here. I think the biggest issue is going to be the challenge is monetisation. We have to admit that we have failed to truly get the monetisation for 5G and that sort of is like a sort of damage hanging over the industry at the moment in terms of where we go next. I think with this important thing of how do we overcome that? Well, I know Guy, you and I sort of corresponded about the recent 6G new use cases, but the challenge with those use cases is they're really kind of ... Telco is the answer, what's the problem type analysis. It's telco seeing itself as the solver of other industries' problems. And I think if we were to change this monetisation problem or challenge that I say, it's going to be telco imagining themselves as part of another industry's solution and being much more circumspect about what is the real issue that they're trying to solve.

(36:59):
I'll give you one very clear example. We get very fixated as telcos about latency. Low latency is going to deliver us everything. Most industries don't have a latency problem or it's a problem that they can solve by doing other things. It's for us to actually understand, okay, latency is an opportunity that we can provide, maybe AI inference at the edge, but why somebody will buy that versus doing it themselves or doing it on the cloud is for a lot of other reasons that are not to do with the technology of networks. It's to do with the overall business value to the other enterprise or the other consumer, for example. So if we can stop seeing ourselves as the sun and start seeing themselves as orbiting different enterprise business needs and business cases, I think we'll do a lot better here.

Guy Daniels, TelecomTV (37:56):
Yeah. Thank you very much, Francis. It's so often, isn't it? It's what is the problem we are trying to solve for our customers? Let's focus on that and work backwards. Well, we must leave it there. That's all the time we have. Thank you all so much for taking part in our programme today. And that brings to a close the Future of RAN Summit for another year. Thanks to all of our guests and sponsors and to all of you for watching and taking part. There is no summit next month because we have our in-person DSP Leaders World Forum event, and that's being held in Windsor, just outside London, but we shall be live streaming both days for those of you unable to attend. And then in June, we are back for the Network APIs and Agents Summit, looking out for more details very soon. Don't forget that you will find links to all the panels and programmes from the Future of RAN Summit on the dedicated page on TelecomTV.

(39:01):
On-demand versions will be available from next week. For now though, thank you so much for watching and goodbye.