Guy Daniels, TelecomTV (00:00:10):
Welcome to the Great Telco Debate 2025. Good to see you all. Come on in and take a seat. You all, you're taking seats. Great. Hope you've all got some teas and coffees. I'm ready to go on this hour. I believe our 12th Great Telco Debate. Yes indeed. I can't believe it is 12 already. Very good to see you all. I know many of you were here with us yesterday for our first Digital Sovereignty Forum event, so thanks for coming back today for the great Telco debate and also we are joined today by our online viewers, so welcome to all of you. So agenda wise, let's have a look at what we've got for you. We have got five debates as usual. Each of these debates is an hour long, the format's an hour long. We've got breaks in between, networking breaks in between and lunch at one o'clock.

(00:01:09):
So we're looking obviously with the AI infrastructure opportunity to start with. We're moving on to the network edge and then APIs. And then after lunch I'll be joined on stage by my colleague Ray Le Maistre, who'll be giving you a brief update on our DSP leaders coverage and some of the research we have planned for the coming year. And then we'll follow that up with two more debates, autonomous networks and the business case for 5G standalone. Right. Those are our debates, but of course we do have motions. It's a unique aspect and feature of the Great Telco Debate is we have motions for you to consider and then to vote on and to ensure we have a heated debate. As always, we are joined by Chris Lewis and Graham Wilde to my right who will be arguing for and against each of the motion. So welcome back both of you.

Graham Wilde, Three Group Solutions (00:02:03):
Thank you very much going,

Guy Daniels, TelecomTV (00:02:04):
Looking forward to another year of debates.

Chris Lewis, Lewis Insight (00:02:09):
Lovely to be here, and a dozen - dozen debates. Can you believe it?

Guy Daniels, TelecomTV (00:02:10):
Very impressed. Absolutely spectacular. Well, thanks very much. We'll hear more from Chris and Graham as we go through the day. Let me give you some of the rules for the debate because we are going to be joined by expert witnesses on each of our debates and they will each give a three minute witness statement, an expert witness statement to come to the podium and deliver that three minute statement. Three minutes is strictly enforced because at the back of the room here, we've got a screen and it's a giant countdown clock and it will turn bright red. Here we go. Thank you very much. Three minutes. You see it's already primed, ready to go. So please do not exceed the three minutes. Please, please, please, please. You don't want to see what happens if you do. Now these statements you're going to hear are going to be on a range of topics, but the idea is to give you a range of thoughts and ideas and views that will sort of broaden out the debate.

(00:03:04):
And then once our experts have delivered that, we'll reveal our motion and we'll hear from Chris and Graham. And this motion's going to be closely linked with the theme of each debate and then we'll have a q and a your chance to ask questions. We really don't want to hear from you 30 minute q and a. So get your questions ready, we'll come to you as many of you as we can and then we're going to vote. So for those of you new to the great telco debate, you'll notice these little plastic paddles on the table. This is how you vote. It's not digital, it is analog and it works supremely well. So long as I can count and I'm fair and honest, which I usually am not. Now boo, if you decide that you agree with either Chris or Graham and you are for the motion, you point the green screen towards me. If you think the motions load are rubbish and you are against it, then point the red side towards me and we'll tot them up and we'll see if the motions are passed or rejected. And if those of you watching online, the motions are on the website so you can also vote and we'll have a look at those in about a week's time or so. So you can participate too, right? That's enough. Let's get on with our first debate, shall we? Oh, no, no. There's an interjection. This intervention

Chris Lewis, Lewis Insight (00:04:17):
Already. Yeah, I would like to point out that whether we speak for or against was decided by the great intellectual presence of Alexa. So it was a virtual coin toss done by Alexa. It was witnessed by my wife and Ali. So if anyone wants to question the validity of it, so just in the way that Graham and I speak in for against that is completely random as to whether we believe in it or not. That's the challenge we face.

Guy Daniels, TelecomTV (00:04:41):
Yeah, yeah. Chris always says that. Yeah, I had to worry with Ley earlier and he told me a completely different story. He does, right? We're going to start our first debate, the AI infrastructure opportunity. So please could we have our first set of expert witnesses on the stage please? Andy, you're four or four. You might be against, I dunno, but you're number four anyway. Right? Well thanks very much everyone. So who's going to start us off? Well actually to start us off, we have a last minute substitution. So Warren Beak, who is the vice president technology office at Wind River is going to deliver our first expert witness statement. So Warren, thank you very much for doing this, and away you go.

Warren Bayek, Wind River (00:05:40):
Thank you Guy and thank you for inviting me here and glad to be here at the last second. So over the next five years, AI will transform every industry on the planet. As Jensen Huang puts it, AI is becoming physical. That means intelligence is no longer centralized only in data centers, it's moving outward out toward factories, hospitals, autonomous systems and vehicles, port cities, basically all kinds of devices that humans interact with. The real value isn't just in training these massive models, it's deploying them everywhere to interact with humans. And that requires an infrastructure fabric that reaches out into the real world. That fabric is the network, the network that this industry built.

(00:06:33):
But here's the challenge. If telcos don't define their role in this coming AI economy, the hyperscalers and the AI giants and software giants will define our role and their future vision of telcos puts us squarely in the bandwidth utility class, the dreaded dumb pipe that others monetize. But telcos can choose a different path. We can become active participants in this coming AI infrastructure stat. What does that mean? We can provide sovereign, secure ultra low latency compute at the edge where AI actually interacts with the real world. We can expose network intelligence through APIs and webhooks QOS, location mobility, jitter intelligence, radio energy awareness. We can monetize connectivity as a service, not by selling gigabytes but by selling outcomes. Industrial automation improvement, computer vision, autonomous systems, robotics, real-time analytics, all the kinds of things I talked about earlier. And we can partner with and not just host the AI ecosystem.

(00:07:48):
Cloud providers can train models, they're really good at that, but telcos can deploy them reliably globally, locally when networks become AI powered and when AI depends on the network, telcos shift from cost center to value creation center from vendor driven to ecosystem leading from connectivity provider to AI infrastructure operator, AI will transform every industry. Telcos can either help power that transformation or be disrupted by it. The only real question is whether we're paying attention early enough to be an economic player in this new landscape or not. The next trillion dollar opportunity is built on intelligence at the edge. Telcos can either continue to fight to squeeze every nickel and diamond shrinking CapEx and opex or claim their share of this coming trillion dollar opportunity.

Guy Daniels, TelecomTV (00:08:54):
Look at that Guy, look at that. Perfect. Three minutes Warren, thank you very much indeed. Excellent applause for Warren. Thank you very much. Manish, we'd like to take to the lecture. Our next speaker is Manish Singh, who is CTO telecom system business at Dell Technologies.

Manish Singh, Dell Technologies (00:09:09):
Alright, thank you. Good morning everyone. Pleasure to be back here. When we talk about ai, I think lemme just start by a simple fact. We all know AI will be everywhere and AI requires a new compute paradigm. It's going to be on the device, your AI pc, on the edge in the core, all the way into the cloud. And there are some constraining factors when we think about that continuum. On one hand, you're limited by power, battery life, et cetera, and heat cooling. And that changes as you go across the continuum. You've got very limited power on your device, limited capabilities on the edge, and the capabilities grow as you go more into the core and into the cloud where liquid cooling and other paradigms start to come in. So power cooling are the constraints and then the workloads which will have AI models embedded in them.

(00:10:13):
And these workloads have different sizes requiring different computes. And where do these workloads land? What's the counterbalancing force? The industry loves to talk about latency. My view is latency is not the counterbalancing force. The counterbalancing force is going to be data gravity. Bringing AI closer to the data source is what's going to make where and define where do we actually land the workload. So that's number one. Number two, what are newer things coming up? And I give you two to think about. You've got XR coming up, which is glasses and other form factors, which requires contextual AI and you've got physical ai, robots, connected vehicles, et cetera. All of these will demand a lot more context going up to the contextual ai, multimodal ai, which also has impact on the network for higher uplink, et cetera. But more importantly, bringing that contextual AI closer to these devices, that's another opportunity.

(00:11:20):
So how do the telcos address this? And I think there are two things to think about. Number one, telcos spend a lot of money, billions of dollars on their CapEx, but these systems have primarily been proprietary ASIC base. We need to transform that to more cloud native open systems and really by doing that we create the opportunity to land other workloads on that infrastructure and since only 20 seconds are left, the second thing I will say is telcos need to change their go-to market motion. Telcos are optimized to sell arpu, oh sell sims and collect arpu. And we need to change that because in the B2C context, to unlock these new B2B use cases, you need to understand the use case, the data and the value creation for that enterprise. And that's different than the current go-to-market motions. Thank you very much,

Guy Daniels, TelecomTV (00:12:20):
Manish. Thank you very much indeed. Round applause. Thank you very much, Manice. Great. And our next speaker is Neil McCray. Neil, please take the letter. Neil is chief network strategist at HPE Networking.

Neil McRae, HPE Networking (00:12:31):
Good morning everybody. I'm going to start with something that telcos rarely start with the customer. My two colleagues have only proven me right coming before that. So as telcos, what are we doing? We're using AI in our call centers, we're using AI in our logistics centers, we're using AI and network design. Telcos are using AI everywhere to improve their business. Why not take that to market? Telcos run the biggest customer contact centers. They run the biggest infrastructures, they run the biggest logistical challenges in terms of running networks all over the world and all over the country. I believe that telcos can take those solutions and market them to customers, but that requires a big change in infrastructure, the core infrastructure of the telco. For too long we have been running that by having a thousand people sat at desks and screens. The time to move to fully autonomous networks is now not because we want to have less people in the network, but because we want to repurpose those amazing network engineer experts to be at the front of our business serving our customers and helping them deploy ai.

(00:13:42):
Now I know some of you have got an NFV steering board at the end of the day, I know some of you are panicking trying to get this year's IT digital transformation finished before the end of the year. But lemme tell you, none of those things are important when it comes to getting on board with AI and being able to position that with your customers. And the great thing about it is when I talk to customers, they would welcome our engagement. They trust us more than the hyperscalers. They believe we understand their business at a local level more than the hyperscalers. So their choice is simple and my colleagues have put this really well either go for the low bit low value solutions or engage with customers in helping them be the business that they can be. That is the simple choice. And if we allow over the top players to come in and take this opportunity from us, the future for us will be very bleak.

(00:14:36):
We have done a phenomenal job, I've said this before, of connecting this device. We can use AI to connect more devices, health and safety in logistics centers, education and training in schools. All of these solutions are powered by the network and then there's AI connectivity services themselves. Today in Barcelona we're demonstrating how you can use the network to separate some of the AI functions and take the cost and reduce the cost of AI and not be dependent upon the LLMs. That in itself is also an opportunity for telcos to sell on-ramp AI services, connectivity services, the thing that we are brilliant at to customers whilst also upskilling and up capability, if that's even the word, probably not bit early in the morning, but up demanding and up powering all of those opportunities that our customers as telcos and their customers need and most importantly want big up for Neil.

Guy Daniels, TelecomTV (00:15:37):
Thank you Neil. Thank you. Our next expert witness is Andy Linham. Andy, please take to the podium. Andy is principal strategy manager, Vodafone group. Thanks guy.

Andy Linham, Vodafone Group (00:15:48):
So I'm an out and out network nerd, right? I have spent my entire career designing networks, building networks, fixing networks and breaking networks. What it means is that when I think about AI infrastructure, I don't think about GPUs so much. So yes, we have to build these massive compute classes and yes, we have to run the inferencing and the training. I think more about how we get to that training workloads, right? So if it's an inferencing workload, how do I get my user across to their application? If it's a training workload, how do I get my data to the gpu? And then within that GPU cluster, how do I get all of those nodes talking really, really quickly to enable that kind of real time parallel processing that AI needs? The thing that I love the most, right? The thing that really makes me smile is that at least half of those training workloads are that proper bleeding edge capability. At least half of it is running across ultra ethernet right now. Ultra ethernet is basically ethernet with maybe a couple more apologies, a couple more bandwidth speeds that we support. Does anybody here know how old the ethernet protocol is? Anyone care to guess?

Chris Lewis, Lewis Insight (00:16:51):
86?

Andy Linham, Vodafone Group (00:16:52):
Anybody as nerdy as me,

Chris Lewis, Lewis Insight (00:16:53):
40 years old,

Andy Linham, Vodafone Group (00:16:55):
1973.

Chris Lewis, Lewis Insight (00:16:57):
No,

Andy Linham, Vodafone Group (00:16:57):
Right? 52 years old. He is older than even me, right? And this protocol is supporting the latest and greatest, most bleeding edge capabilities that we can build. I absolutely love that about networking. The other thing, and bear with me here, ethernet also reminds me of my wedding day, which sounds like a leap, right? And you're thinking what a boring wording day do you have out here? It was quite good, right? I enjoyed it. But fundamentally, I remember the morning of my wedding day. I must have looked really nervous. I must have looked probably terrified a little bit overall my best man wanders up and so he met a university doing technology degrees and he comes up to me, he says, Andy, he says, what's an ethernet? And my right hyper brain kicks in. Well obviously it's a layer two pros code that we use for switching packets across or switching frames across the network. And he looks at me and he goes, handy, handy. Two, an ethernet is what we use to catch the bunny. Thank you. I'm here a week,

Guy Daniels, TelecomTV (00:17:55):
Andy, what a wedding. Thanks very much. Our next speaker is Fran here, and Fran is Vice President Global Telecommunications by Red Hat.

Fran Heeran, Red Hat (00:18:04):
Thank you Guy. Good morning everybody. My first great telco debate, so great to be here. Typically I categorize, I've spent a long time in Telco, mostly on the vendor side, a little bit on the operator side and probably quite crudely. And typically I categorize all of our conversations with our customers in one of two categories. Help me make money and help me save money. Everything else goes from there. AI is no different. If we categorize it across those two areas on the help me make money, the discussions are ramping up around can telcos credibly monetize ai? And if so, why? So what differentiators do we have as a telco, right? One is the edge potentially, right? Unique edge locations out as far as the radio, we hear about ai ran AI for ran AI with ran second differentiator, we talked about it yesterday. Is sovereignty.

(00:18:53):
So in country location and presence, can I offer unique service using the edge, using my sovereign presence in country to offer something that is maybe not unique but differentiated enough to be competitive? And thirdly, do I do it myself? So do I or can I afford to build AI infrastructure on my own given the costs? Or will we see telcos getting together to offer national in-country AI services? So open questions there, but certainly we're seeing globally that conversation ramping up. The majority of discussion in AI though is on the help me save money. So using AI internally in the operators for operational efficiency, customer experience, usually breaking down across the three typical categories inside of Telco, your ran your edge, I kind of grouped them together, the core and also on the it. On the O-S-S-B-S-S, interestingly, many people when they hear the word AI, automatically assume I need GPUs.

(00:19:51):
We need expensive hardware that is becoming less and less true. We're seeing a split happening between the requirements around the training of these models, right? The large language models, the small models, large action models now as well that does require heavy infrastructure. GPU intensive, do I build it myself? Do I partner open? Question. The inferencing, however, is changing significantly. So the cost and the resources required given the optimizations we're seeing and the algorithms is improving week by week. So we're seeing essentially AI becoming a set of questions, which is do I have the right infrastructure for the right application in the right location? Is my data pure and complete enough to be practical? We're hearing stories from customers around trying chatbots with less than stellar success because of bad data. And then automation, which is everyone's favorite topic, very general using AI for automation. I would argue and based on what we're seeing, it's less today about the technology and more about the trust. Do I trust the machines? Do I trust the software to run my network end to end? It will happen over time, maybe not today. And above all, does it pass the bullshit test, which is typical for telco. I think it's important we discuss the use cases and we ensure their practical use cases. They have a practical business model and differentiated enough as a telco to be successful. Thank you

Guy Daniels, TelecomTV (00:21:12):
Fran. Thank you much indeed. Thank you. And our final expert. Oh dear o. Dear, dear, dear, what have we did? Mark. Mark Matt, please introduce Mark Gilmore. Mark Gilmore is the Chief technology Officer at Connectivity Europe, believe it or not.

Mark Gilmour, ConnectiviTree (Europe) (00:21:30):
What do you mean?

(00:21:32):
Right? I'm going to request a little bit of audience participation, but don't worry, I'm not going to ask you to do anything outrageous. Could you raise your hand or your paddle as a test if you have children aged between about 20 and 30. This will determine just how successful this next part goes. Okay, well, alright. If you have children around that age, you may remember one of my son's favorite children's TV characters, Bob the builder. Now for royalty infringement reasons we can't play his theme tune, but does anybody remember his catchphrase? Join in, if you remember it went something like this. Bob the builder. Can we fix it? Bob the builder?

Chris Lewis, Lewis Insight (00:22:18):
Yes

Mark Gilmour, ConnectiviTree (Europe) (00:22:19):
We can. Fantastic guy that went better than I thought it would. Had we been allowed as part of these three minute expert statements to use visual aids, I would have shown you a picture of my 18 month old son dressed up as Bob hard hat tool belt and baby grow. But the great rules of the great telco debate have spared his blushes. So for visual context

(00:22:44):
I shall wear his signature hard hat. Now, when I first looked at the topic of this panel, I couldn't help but think of Bob the highly respected and recommended builder from my son's childhood. And why? While because we call ourselves telecom operators, but at the heart of it, we're actually network builders. So considering the AI infrastructure opportunity, telcos are network builders and should be at the heart of it. Now I shall mention three areas where I believe there's a real but perhaps unglamorous opportunity for telcos and one where I think we should be cautious. The three are AI factories and their connectivity and their build. These data centers or these AI factories or big data centers which are being located based on power and cooling more than any other factor. This creates an opportunity for telcos to get involved in providing high capacity, resilient, secure connectivity.

(00:23:38):
It's not sexy but it is critical. Now, sovereign ai, we discussed that at length yesterday, but by nature operators are sovereign. So we can step into the frame of this and use that to build data centers but to also to get involved with partners. Look at the news this week, Deutsche Telecom amongst others. And the last is collaboration and the enablement of AI related services. Now this is an opportunity to create that environment that fosters innovation in the AI space that my colleagues have spoken about, but it's a difficult areas for telcos to engage with given our track record with, sorry, Dean, OTT players. Now the one I believe that telcos should be cautious of is the edge AI at the edge. We've spoke about that yesterday. We'll return to that now return it to Bob. One of the most memorable episodes for me was when Bob was tasked with building a new data center. Okay, that isn't quite true. You can't believe everything you read on the internet. The closest he got was building farmer pickles barn in series one, episode 13. But what did frequently occur in the 250 episodes that aired over that 16 year period would he would rally his team under the careful supervision of Wendy. Can we build it? Yes we can. There we go.

Guy Daniels, TelecomTV (00:25:00):
Round applause for Mark. Mark was given an extension there. He was dressed up and talking of extensions. I've got one half finished, if you wouldn't mind popping round, finishing it off. Good. I dunno what to say. Thank you all to all of our expert witnesses. Terrific set of views there, right? We've heard from our six guests, what we need now is a motion and we're going to display our first motion. Our first motion is, let's read the first

Graham Wilde, Three Group Solutions (00:25:35):
Motion. Let's read the first motion. The motion is telcos must become AI infrastructure players to benefit from the AI economy.

Guy Daniels, TelecomTV (00:25:42):
Well done Graham and Graham, you are going to be advocating

Graham Wilde, Three Group Solutions (00:25:46):
For Yeah, after the virtual toing costs. I'm speaking for this motion and Chris is speaking against it. So let me stand up. Oh ladies and gentlemen, I urge you when the vote comes to point, the green bit of your paddle towards guy and I want to tell you a little story. I want to take you back to the year 1974 in the Philippines and there was a man there called hero, or he was a Japanese soldier who had fought in the second World war and he had stayed in the jungles of the Philippines for 29 years believing that the war was still ongoing. He was found eventually by a Japanese adventurer who befriended him, but he wouldn't come out of the jungle until his commanding officer was brought from Japan. His commanding officer is now an old man running a bookshop. By the way, he was brought from Japan, taken the jungle in the Philippines to persuade Mr.

(00:26:46):
Oda to surrender. Mr. Oda came out of the jungle, still wearing his uniform, still carrying his rifle. Now why am I telling you this story? Telcos, you could be in the jungle. You could be in the jungle for the next 29 years eating rats, okay, you could be there. Mr. Aoda missed some key things while he was in the jungle. He missed the invention of television. He had missed people putting a man on the moon. He missed the invention of ethernet one year before he came out, right? He was very excited about that. You have heard from our speakers the many ways in which telcos can participate, particularly at the edge, particularly in terms of taking the expertise, as Neil said, putting it in call centers, offering that as those services to customers. You would be very, very foolish to pass this up. So I urge you, ladies and gentlemen to vote for the motion. Thank you very much, Graham. Well

Chris Lewis, Lewis Insight (00:27:48):
Argued. So are you seriously, he had the same clothes on 29 years. He did.

Graham Wilde, Three Group Solutions (00:27:53):
Yes he did.

Chris Lewis, Lewis Insight (00:27:54):
I think you're talking, you're hallucinating Graham, along with your proposition. Ladies and gentlemen, as Graham will tell you from his psychology degree, the red side of the paddle will make you angry. But much of the discussion around AI makes me very angry in relation to the telco side of the discussion as an industry, as Neil quite rightly says, we've been building, building, building. We do a great job providing that connectivity. Over the last 40 years that I've been an analyst looking at the industry, every time a new wave of interest comes along, we believe as an entry we should be embracing it. We should be taking control of it. Everybody should play by our rules. The AI explosion that we've seen in the last few years, but it's been going on for longer than that, really takes out of the center of comfort, the comfort zone from the telcos.

(00:28:47):
So to suggest that we should become AI infrastructure players is looking for an extension in the way that the investment we put in place, we know from the return on investment across the industry, even in the US with ARPU over $60, but all the way down to India with an ARPU of $2. We have a massive array of country markets, a massive array of requirements we need to keep building if we're going to get anywhere near connecting. All of those things that Neil touched upon and the quality that we need to support it, we need to focus on the connectivity side of it. Now in many respects when we think about AI and one of the analogies that we often use is think of a sporting analogy in the Premier league of the future digital services market environment. Does AI promote or relegate the telco part of that structure?

(00:29:41):
In many ways it pushes it farther down the scale. The power required both physical power, energy, but also the software power required actually is going to be developed by other parties. Ladies and gentlemen, I recommend to you, you vote against this. This will be another black hole of investment. It will be a future hallucination. It takes away from what is required, which is focus on delivering the part of the equation, the part of the economy, which does underpin what happens elsewhere, but it needs to be focused. Please do not get distracted by the AI bubble. Wow.

Guy Daniels, TelecomTV (00:30:20):
Chris, thank you very much. Lovely. Chris and Graham there. Thank you very much. We'll revisit this motion in about 30 minutes time or so. A reminder to our online audience, please go ahead and vote on the telecom TV website and we'll reveal the online results after the event. Right? We are going to start our discussion with our guests now on the topic of the AI infrastructure opportunity. Before I come to my guests, I'm going to go around the room and just see if there's any questions because we've had 30 minutes worth of expert witness statements and I am sure there are going to be some questions from the room. Let us go to, let's go in the middle of the room first. Let's go to Beth in the middle of the room there first and then we'll make Hannah walk backwards and forwards all afternoon.

Beth Cohen, Luth Computer (00:31:05):
Okay, thanks for making me not walk. So is it all black and white? Do we have to invest completely in the AI full AI stack, including the GPUs and everything? Or can the telcos cherry pick what is the most effective part of the AI infrastructure?

Guy Daniels, TelecomTV (00:31:29):
Beth, thank you very much. Great question. Who would like to address this? Is this an all or nothing question? Andy, do you want to come in?

Andy Linham, Vodafone Group (00:31:37):
Yeah, I mean I think going all in is one way to go, but as telcos, we don't typically have billions of dollars worth of CapEx to go and invest in GPUs. I think what we need to do is to play to our strengths. So we have huge amounts of physical locations, we have massive customer bases. We have the ability to do things that frankly the hyperscale at the moment. So I think that ability to pick and choose the right areas to invest in is absolutely the way we need to go. Part of that is partnering to, so finding the right people to fill in the rest of the stack I think is where you're going to find telcos getting the most value for their resources and the effort.

Guy Daniels, TelecomTV (00:32:13):
Great. Andy, thanks so much Warren, your come next.

Warren Bayek, Wind River (00:32:16):
Yeah, I agree generically with what you said and I think telcos should consider this sort of as if we are going to become digital landlords, if you will. I'm a little bit of a financial nerd. Let's face it. There has never in the history of humankind been as much money pouring into the people running the AI revolution. It's not telcos, right? Those people have tons of money to spend and they want to do something with it. So the telco collaboration and becoming sort of a digital landlord model I think makes a lot of sense. Them bringing all that money into our infrastructure to prevent us from going down that ROI rat hole that we've been in so many times in the past.

Guy Daniels, TelecomTV (00:33:02):
Thanks Warren. Anyone who's got a Fran?

Fran Heeran, Red Hat (00:33:04):
I think, well the answer is no, it's not all or nothing. I think you can start by renting out the space. So telcos have this valuable real estate at the radio on the far edge. You could just lease out that, let somebody else build the infrastructure. The question is how far up the stack do you want to go Put in basic compute, do you want to do the applications yourself, build your own model? I think it'll be a partnership model. And then the question for each telco will be where do I want to play and how far do I go up the stack? Thanks Brandon. Mark.

Mark Gilmour, ConnectiviTree (Europe) (00:33:29):
Yeah, I agree with that statement, but Warren's point is interesting about the amount of money that's pouring in because it does beg the question about ROI in the long term because if all that money is coming in, somebody's got to make money out of it at the end of it and if it just becomes poured in, then it becomes a bubble and it explodes. And so I think that's why telcos can't go all in actually, because even with this, there is a risk involved, a financial risk. I agree with my other panelists really that it's going to be right sized for you. You as a telco, what fits your approach, your piece. Because not all telcos are mobile operators. That's the thing that I will bring out. We tend to forget that these days. And so the telco industry, we play in different spaces within that. And so there's different parts to play in this AI infrastructure play, but it is something where telcos have to get involved at some point. We can't just ignore it.

Guy Daniels, TelecomTV (00:34:44):
Thanks very much Mark Neil.

Neil McRae, HPE Networking (00:34:46):
Yeah, so I think one of the advantages that telecommunication operator have got right now is something that we always wish that we had more scale actually with the scale that telecom operators have got. They don't have to spend tons of money and they can build as they grow. And all the vendors in this space have got platforms that allow you to do that. And also many of these platforms are also supporting other telco workloads. So I think the question's spot on Beth, I think it's where do you start, where do you start, where you're going to be successful, generate momentum and then push out? And definitely I think we're scared by we see these headlines, 20 billion, 30 billion, 40 billion, and immediately we step back and say, wow, that's a lot of CapEx. It is a lot of CapEx, but you've got to think of your end game and what you're trying to serve and to serve.

(00:35:42):
If I was to think of this market here in the UK to serve that, you're not talking about anywhere near that certain number. And also if you pick the industries you want to support carefully and optimize for those industries even less. And so I think my point, logistics you could do a lot with in the logistics space and actually not have to spend a huge amount of money. Fran, spot on. It's a partnership model for sure. And I think in that scenario, bringing the right experts together is actually what's going to deliver this. I reflect on 5G and I know we've got a topic about this later on in 2016 when we were talking about 5G, I said, unless we get the partnership model right with knowledge experts, communication experts and customer experts, 5G will go nowhere. Look what's happened. So we have to change that and we have to be brave about it. And my fear is the do nothing is seen as low risk. Actually I think the do nothing is the highest risk a telco could take today.

Guy Daniels, TelecomTV (00:36:44):
Interesting. Thank you very much. Neil. Manish, do you want to come in? Yeah,

Manish Singh, Dell Technologies (00:36:49):
I would say the answer is yes and no. The yes part, lemme start with that. Lemme just start with the yes part first. The yes all in part that I think is a must do is to move away from the legacy way of building the networks from legacy infrastructure to start to move towards more cloud native AI native architectures. And must I say this? Please take note of the disruptions that are happening in the semiconductor industries where large part of the capacity of that industry is getting sucked into the AI world, whether it's CPUs, GPUs or memory implication being a six and proprietary systems are going to become even more expensive and difficult to bill and the CapEx dollars that telco spend are going to become even harder and will go lesser distance given what's happening in the semiconductor because of fa. So the yes part is please go all in and start thinking about transforming your infrastructure. The no part is I do not believe that they should go all in to address mass use cases. Rather identify where you are differentiated sovereignty and edge are clear friends for the telcos forces that really put tailwinds behind it. And that's why I said the go-to market motion here will have to change. This is not a sell sim collect ARPU game. This is about understanding the customer, the use case, the data and where can we unlock the value.

Guy Daniels, TelecomTV (00:38:32):
Thanks very much Manishh. Great question and great responses from all our guests there for the opening one, we've got a lot of hands going up. So Dean, we're going to come to Dean next with the big orange box.

Dean Bubley, Disruptive Analysis (00:38:44):
Thanks. So where and what exactly is this desirable property? It's not cell towers. They've got no power spare and frankly mobile's irrelevant to AI pretty much anyway. It's not edge data centers in an old phone exchange that has maybe got a few kilowatts spare and doesn't have that many diverse fibers into it. So you're not running critical workloads, maybe some aggregation points. It's not a streetside cabinet for a fixed operator. So where is this? And frankly a lot of cases they've outsourced it to tower companies or a rooftop of this building or something else and they're not trying to become an AI edge. So where is this desirable property and what's it connected to and how much power and cooling does it have?

Guy Daniels, TelecomTV (00:39:33):
Dean, thank you. Who'd like to start us off and Neil?

Neil McRae, HPE Networking (00:39:38):
Yeah, I mean look, I think it's a fair challenge, Dean, that telephone exchanges and central offices are, they've been around since as long as ethernet, right? But I think if you look at, I know most of the UK networks reasonably well, there are key sites that could easily do AI that are pushing big chunks of data and power already. And actually AI would help them optimize those sites both for cooling because these aren't like big square buildings where your yield's really easy. These are really, we buildings with staircases in the middle of them, Victorian buildings, et cetera, et cetera. AI can actually help you plan the workflow, the airflow, and I go back to my scale point. If you're building a patchwork quote across the UK and you build that well and you use the tools that Manish is talking about, about a new cloud-based architecture, you can easily build that out and scale out is I think a word you're going to hear a lot over the next few weeks where actually we've ran out of a capacity here, we need to put something somewhere else and we need to connect the two together.

(00:40:48):
So I strongly believe that there is an opportunity to do that. There is the power, there is the space and most importantly, that meets the demand of where these customers are and gives you the latency benefits and all the other benefits. Is it 18,000 base stations in the uk? Hell no, right. It's definitely not that. Is it 150 core network sites in the uk? Maybe

Guy Daniels, TelecomTV (00:41:15):
Neil, thanks Warren and then we'll come across to you. Frank Warren.

Warren Bayek, Wind River (00:41:18):
I think it's a great question, but I think it depends a lot on what you're talking about when you say ai. Well, I agree that the power, the cooling, that's really important. Once you get into complex models, as someone mentioned, I think maybe even Manish, there's various levels of what AI is designed to do and what we're going to make it do in the telco world and in the inferencing space, it's actually already happening at the far edge. If you've created a virtual environment out there as it's been stated by other panelists, that's an absolute essential. That is the all in Beth, that we cannot do this without virtualizing the edge states period. It will never work with legacy systems. Some of the North American carriers, Verizon in particular have gone a long way to virtualizing their entire network. They're already running or capable of running workloads out there, just inferencing.

(00:42:12):
Now let's not get complicated with these large LLM models. We're not going to train out there, but we can and already are running some inferencing out there. Some of the virtual ramp providers are using it to optimize their radio efficiency, their energy usage, their spectrum usage. So it's already happening. If we can virtualize the entire edge, we can push some version of this AI infrastructure out to the far edge. Now there will be other plays and I agree with other panelists who have said there's a lot of room in the telco world to use AI to do operational efficiencies, to do O-S-S-B-S-S improvements to do changing the billing models of critical part of that. But there's a lot of different opportunities in telco for ai. It's not a one size fits all that AI is this box. It varies where you want to put it, where the workloads are going to sit and how the data is going to be ingested and what data we're going to use. But I fundamentally disagree, respectfully with your comment that we will not have AI at the edge. I not only believe it, I've seen it happening and it will happen in dramatic fashion in the next few years. I'm convinced of it.

Guy Daniels, TelecomTV (00:43:20):
Great. Warren, thanks very much. And Fran probably,

Fran Heeran, Red Hat (00:43:22):
Yeah, no, I mean it's a fair question Dean and I think, but just to build on what Neil was saying, it's also important as well to look ahead. If we look at six G, God forbid, ring the bell, someone said six G before six G mentioned. I was hoping to be the first to do that, but six G, the fundamental structure and nature of the edge will change in six G. You'll see more applications from the core, especially moving towards the edge on the radio side, maybe not every radio site, but the six G specs are seeing a change in how it's structured. So you'll see more general purpose compute, I think a different edge structure. So yes, today it's a fair question. I think it will evolve maybe not completely in every site, but I think it will change in six G. Even 5G advanced will drive some of that as well.

Guy Daniels, TelecomTV (00:44:07):
Thank you. Mark, do you want to come in?

Mark Gilmour, ConnectiviTree (Europe) (00:44:09):
Yeah, I'm not wholly convinced by the AI at the Edge, telco edge, yes on the far edge devices, things like that. See that I'm not entirely convinced yet on the cell site, but I also don't think that's the only play when we talk about telco AI infrastructure in this, look at the Deutsche Telecon news, we discussed a little bit yesterday about other examples. There's an opportunity there to build new sovereign based AI factories, data centers, whatever you want to call them. And I think that, but that comes back to the partnership angle that we talked about when operators are not doing that on their own. So I think there is space there and the thing is, is telcos bring, we kind of bring sovereign compliance, something that we inherently do as an operator because we're regulatory based, we are based on a particular governance, whether that's uk, whether that's Europe, whether that's US Asia, it's just how we operate and that is something that a hyperscaler doesn't do. And so that's where I think we bring operators bring value into the discussion for that point. We talked a little bit about it yesterday. I can't see too much of a difference between the edge story that we had 10 years ago and the edge story that we're doing now. I don't see too much difference right at this very moment, but perhaps there is something future coming, but my years in the industry now tell me sometimes that doesn't come

Guy Daniels, TelecomTV (00:46:05):
Up. Okay. Okay, well we'll see Mark. Thank you. We're going to hear from Andy as well.

Andy Linham, Vodafone Group (00:46:09):
I think the last point I mentioned is just you need to think about the geography of where you're operating. So the uk, as much as we think with this massive global power, it's a small country fundamentally. So if I put a couple of nodes north and south, one in the middle, I can get really high performance from just about everywhere. If I look at the United States or India or China, I get a lot more nodes. In Europe in particular, unless you're going to Germany and France, you can get away with a couple of nodes, a handful of nodes per country to get really good performance. So we don't actually need that higher volume of sites. And I think if you look at the end of church, look at the physical fixed points of presence that we've got, there's enough capacity there, there's enough space there, there's enough power there to give us the ability to run the in workloads in particular at really good performance. Andy, thank you very much.

Guy Daniels, TelecomTV (00:46:54):
Manish, did you want to add?

Manish Singh, Dell Technologies (00:46:55):
Yeah, a couple of quick things. Number one, the debate is not way to train, way to infer just to start with, training is going to happen in deep cloud centralized locations as somebody who's been in the middle of building some very large training clusters. Yes, I mean these clusters require directly liquid cooling. There's complete different facilities that you need including power. The question is where to infer and when it comes to inferencing, I think somebody mentioned it's not only just GPUs, I mean inferencing actually will happen on CPUs, GPUs, and it's a continuum, as I said, including the devices exactly as you said, there's going to be inferencing even on the devices. So inferencing, there's a continuum that's required, which brings us all the way I argue to the RAN base station and to the telco edge. Not to forget that AI in the RAN itself has tremendous potential to unlock things from beam forming to multi-user MIMO pairing and things that traditional computer has struggled to give us the unlock, which we expect and we are already looking at even things like link adaptation and all AI in the RAN itself will demand compute facilities in there.

(00:48:17):
And so where I conclude this is we are going to see the workloads in the network themselves evolve, and if we can unlock the infrastructure from the workloads, we can make the network workload smarter with AI and create pathways for inferencing workloads, not all, but the ones that truly require to be brought to that edge for data gravity or sovereignty.

Guy Daniels, TelecomTV (00:48:43):
Manish, thank you very much. We're going to go for another audience question. We had one at the front, so let's go to Francis at the front.

Francis Haysom, Appledore Research (00:48:51):
Thank you Francis Haysom from Appledore Research. My question is really how and where are we going to see the investment by telcos in this opportunity? Some background, apple door research last year did an analysis of 20 major telcos globally and looked at their performance or financial performance over the last 10 years. And in sort of if it was a hospital drama, the patients dead basically. And we also compared it with say the hyperscalers and even some utility providers, Ashley. And basically if you compare it with the hyperscalers, the hyperscalers are investing CapEx and they're investing CapEx not just to build things, they're investing to change operational. How in a flatlining industry are we going to get the investment? That actually one allows us to build this thing but also allows us to compete operationally for that one because I think that's the big opportunity. We talk a lot about the opportunity. We very rarely talk about what the investment is needed to actually achieve it.

Guy Daniels, TelecomTV (00:49:59):
Okay, thanks so much, Francis, who's got the crash card? Mark.

Mark Gilmour, ConnectiviTree (Europe) (00:50:02):
So I'll tell you where I'd like to see the investment go beyond the operational efficiency things that we've spoken about in the past. I'd love to see it go towards customer engagement, customer improvement, and reflecting the customer needs kind of to where Neil was pointing in his address. That's why I'd like to see it start. I'd like to see telcos engaged in participating in the service side of that and stepping up the stat. I think on the infrastructure side, which is the point of the panel, then I think that will come in joint investment into factories, into connectivity for those. And that's kind of an operating model that we kind of know we can draw from consortium based applications that the industry is good at doing already where it needs large scale investment. But I'd love to see the drive towards focused areas, whether it be in ports, whether it be in, I did point a note.

(00:51:27):
Oh yeah, energy. There's an interesting one. I made a note of this. Things like grid forecasting and things like that, some investment into those sort of areas. And even on the general, one of the great things I do in my role is I get to look at some innovative things that are coming back. We don't innovate particularly well as an industry, but we can enable innovation in that. And I think the partnership in there was talking with a Silicon Valley based startup that is looking at orchestrating inference, which I find fascinating so that you as an operator, I can offer to the customer inference across a whole host of LLMs rather than just a specific one. Fascinating.

Guy Daniels, TelecomTV (00:52:16):
Thanks Mark. Investment. How and where Any other, Neil, go

Neil McRae, HPE Networking (00:52:21):
Ahead. I mean I think you've just demonstrated the reason why telcos have to do something different. We've been blowing money on radios and stuff for years and what's happened is the business is not great. I mean 10 years ago, 20 12, 20 13 when 4G the wall was a very different place. We've now got to a wall where the value proposition that telcos have is a challenge in one. It is kind of somewhat fortunate that the next, the last two big upgrades, so all the fiber that's being built right now, all the 5G that's being built right now has happened at a time where it's affordable, fast forward 10 years from now, we be able to afford those upgrades. So I kind of feel that without the network there is no ai and for me that means that there has to be a play for this industry to participate in that.

(00:53:19):
Is it some of the robots delivering packages around the street, something like that? Yes. Is it flying jet fighter planes? Probably not. That's probably not where to focus, but I can't believe that in the widest range of where the world is, there is not a space for telcos to drive more value. One area I believe, and just to refocus is security. I mean, some of the breaches that we've had over the last year have been frankly outrageous. And it's because the complexity of technology, the complexity of the network is beyond what we as humans can cope with. So we need this technology to help us do the basic things we need to do. We as telcos, we as vendors and our customers who are running their businesses absolutely knew it. I think that's the space where telcos could absolutely operate in and that they could build the infrastructure for it. And again, I don't think we're talking about billions of pounds or dollars here. I think there's a scalable way that you can grow that and build the value at the same time. Is it easy? No. If it was, we'd done it right, but this goes back to my point I made earlier. The do nothing scenario is higher risk than do something at this point. That's my strong belief. Thanks

Manish Singh, Dell Technologies (00:54:45):
Manishh. Very short point of view on this one, where does the investment come from? The investment in my view comes from the existing CapEx investments. And that's why I keep going back to that point of we got to think about how do we take that CapEx investment that we are doing in the network and make it more cloud native, more AI native where I can run my network workloads, but then create space for bringing in AI workloads. Thanks Manishh. Andy?

Andy Linham, Vodafone Group (00:55:16):
I think all I would say is that as much as yes, there's money to be invested, there's also the investment in terms of effort, in terms of resources, in terms of power and space, and that's where we do have an abundance of talent and capabilities. I think if we can pivot towards investing that kind of existing capacity and capabilities that we've already got and reuse our people, reuse our space, reuse our power, and maybe let somebody else spend the CapEx and work in partnership, I think that's where we find the most return on investment, frankly. Thanks. And Warren?

Warren Bayek, Wind River (00:55:44):
Yeah, I think Neil said a really important line, which is without the network ai, especially physical ai, which Jensen is really pushing now can't happen. So while I agree that you don't need to spend quadrillions of dollars out there, I think the important thing is bringing in these collaborations with the people who will benefit from it. The reason that telco is in low single digit growth with their CapEx is not not spending money, it's because we're not recognizing the other side of it, right? It's either in or out and we've really focused on the out. We're really focused, squeezing every penny out of operational efficiency, which is great and we have to keep doing it. But the folks who are doing the 40, 50, 60% on that over the last 10 years, as you well know, they're spending gobs of money, but they're also realizing gobs of revenue.

(00:56:38):
So I think the collaboration and we've built this incredible real estate venue, if you will, that no one else has the serenity of the data, all of that telcos already have that. So we need to offer that to the folks who are going to really make money in this space and say, look, we're coming in, we're giving you something of value, let's not give it away this time. So let's collaborate with them and let that money be invested in the infrastructure play. The telcos will do our part in that, but we do need to bring in other investment through collaboration with the deep, deeply invested pockets of the large players in this space who are going to make a ton of money and telcos have to change the paradigm if we're going to partake in that. Thank you Warren

Guy Daniels, TelecomTV (00:57:25):
And Fran, did you want to add to this investment question?

Fran Heeran, Red Hat (00:57:27):
Yeah, I mean just to build on that, I suppose the other topic is national interest, right? Does this become a government investment topic? We've seen it with broadband before. I think we're still missing the use case though, right? So there's kind of a leap of faith to be taken on. What will the edge use case be? Look at where 5G is going with the next speculations. It's very device focused, right? So will those devices drive an actual use case that will need the edge TBD? So it could be private investment, private partnership, or even government investment as well.

Guy Daniels, TelecomTV (00:57:57):
Thanks so much Fran. I'm aware that we've only got about four minutes left of this.

Fran Heeran, Red Hat (00:58:03):
Oh boy.

Guy Daniels, TelecomTV (00:58:03):
James, can I come to you for this one? I'm sorry if everyone else I'm going to miss out here.

James Crawshaw. Omdia (00:58:08):
Mine's very quick. So you're probably able to squeeze in another one guy. In fact, I've got two questions that are for you. Unfortunately they're slightly awkward. So the first one is I couldn't quite understand why is it schedule when it's school and scheme? Surely the Americans have got this right? And secondly, how can we have this debate when we don't have a representative on the panel who is a hyperscaler?

Guy Daniels, TelecomTV (00:58:34):
Okay, so do have another question in the room?

James Crawshaw. Omdia (00:58:37):
Alright, so I don't expect an answer, but just two

Guy Daniels, TelecomTV (00:58:40):
Observations, no answers for me on that one. James,

James Crawshaw. Omdia (00:58:42):
Do you mind checking this out?

Guy Daniels, TelecomTV (00:58:43):
Okay, look at this. The time says three o'clock now and three minutes to go and it's already clocking down. I'm going to go at the very back though, if you don't mind. I'm going to go to Andrew at the back there. We don't throw the patch. Not anymore. Health and safety. We can't afford the insurance.

Andrew Collinson, Connective Insight (00:58:57):
Hello. A very pro infrastructure panel. What is your view on the time to return of investment of an infrastructure in this? Because we've talked a lot about how attractive it is from a technical point of view, or at least in theory, how long I've heard people talking about 17 years, that kind of time for infrastructure and why infrastructure. Clearly there's a need in ai. So what's the time to return and why must telcos get apart from the excitement of the yellow jacket? Why must telcos get into building infrastructure for this?

Guy Daniels, TelecomTV (00:59:38):
Right? Brilliant. Two minutes to go. As you can see from the clock. So anyone tackle the timescales first? Anyone to just check?

Neil McRae, HPE Networking (00:59:44):
I would challenge the 17 years because if you think about the mobile cycle, it's not 17 years, but what

Andrew Collinson, Connective Insight (00:59:51):
Is

Neil McRae, HPE Networking (00:59:53):
It? So in this scenario, I think it's someone like security. You could easily, I think, build a case within a five year cycle. I think that's absolutely doable because Is that infrastructure? It's infrastructure, yes. It's partly infrastructure. It's partly services. It's partly capability. Some of these things telcos already got 'em they just need to leverage them in terms of the It's mix isn't it? Isn't not infrastructure, stop it. Not pure infrastructure. But you can't just like, Hey, I'm going to chuck A of servers here and boom, we are making money. That's just not how it works. Right, exactly. You've got to build it. You've got to extend the service. You've got to build as managed, so you've got to change the model in terms of how you engage. And we see that from customers, particularly in security actually. So I think that it's absolutely doable.

(01:00:37):
Again, it kind of feels like Brewster's millions, we've got to spend 30 million in 30 days. No, we don't. Right? What we've got to do is build the infrastructure. We need to start this journey off. And actually to start off, you might not need to build it. Many of us have got platforms where you can rent the infrastructure to start with, get the ball rolling and then I invest as you want. And actually back in the early telco days at co, that's exactly what we did. We put in a lease line from the incumbent and when we had enough customers, we migrate our own fiber to it. So I think there's a lot of different models that you could make this work, but it isn't just infrastructure is one part of it. It's how do we turn that into actually a service enabled solution with partners that is valuable to customers. And I genuinely think security is one area and it is everything in security from risk management all the way to perimeter management.

Guy Daniels, TelecomTV (01:01:29):
Neil, thanks very much. Mark, you've got 10 seconds.

Mark Gilmour, ConnectiviTree (Europe) (01:01:33):
If we're building brand new, then that's a model that actually as a telco we understand pretty good in terms of the ROI ON connectivity and things like that. I think we can manage on that side of thing and that's where the partnership piece looks. Look at what is happening in these big partnership sovereign builds.

Guy Daniels, TelecomTV (01:01:50):
Great. Mark, thank you very much. Thank you everyone. I will answer and answer questions in the break, but thank you very much. I'm sorry we didn't get to all your view questions, but we are out of time for our q and a. However we've had our debate. We do now need to remind ourselves of the motion and to see what you think of the motion. Here's the reminder. The motion was telcos must become AI infrastructure players to benefit from the AI economy Graham was voting for and Chris was voting against. So please pick up your panels and vote now. Oh, work that one out. Oh, I need some assistance from, I need some. What do you reckon here? This. This is the two complex. This is quite split. This isn't it? Split decision. This is a really, you got some greens in here? What's that gt? How many people with a red? It's green. The greens have it. I can't believe it. The greens. Thank you very much everybody. Thank you very much. Perfect time for short break. We are back in 20 minutes. Please avail yourselves of refreshments in the room next door. Thank you. We'll see you soon. Thank you. And a round of applause to our guests.