To embed our video on your website copy and paste the code below:
<iframe src="https://www.youtube.com/embed/sdvYGWgOF4A?modestbranding=1&rel=0" width="970" height="546" frameborder="0" scrolling="auto" allowfullscreen></iframe>
Guy Daniels, TelecomTV (00:11):
Hello, you're watching TelecomTV and this is the spotlight on 5G After Show. I'm Guy Daniels, and today we are going to wrap up our coverage of the MWC 2026 event. What were the highlights, takeaways and talking points, surprises, good and bad. And what have we learned about the state of the industry and its direction for the next year? Well, I'm delighted to say that joining me on the program to share their expert findings with us are Steve Daigle, who is global head of Telco Systems Engineers at HPE. Nico Marziliano, who is regional VP Telco Sales for Wind River. Rob McManus, who is principal product marketing manager with Red Hat, John Abraham, who is principal analyst at Appledore Research, and James Pearce, editor of our very own telecom TV. Hello, everyone. It's really good to see you. Thanks so much for taking part today.
(01:11):
Hope you are all now fully recovered from the exertions of MWC. Now, our editorial director, Ray La Maistre, unveiled his four C's takeaways last week, connectivity, collaboration, context, and celerity. Don't ask. So let's hear what your key takeaways were and got to start, of course, with AI. It dominated the show and it was part of so many of the conversations that we had. And I know you all have views on this, but Steve, I'll come to you first. What has been the impact of AI on the network?
Steve Daigle, HPE (01:50):
Hey, thanks, Guy. We had a really good show and it's always a good show, but it was our first time with HPE and we had some significant announcements around some of our products. But I wanted to add a C to your list there to raise a list about conversations. We had conversations around AI adoption, but really around the networks to support the growth and the new types of WAN traffic that we're seeing with AI adoption, especially the importance of latency capacity and dedicated AI paths. And I tell you, latency is just becoming more of a conversation that we're having with customers than it was before. It's like the broadband measurements and the speed tests of the past. It's more and more about latency. I met with a couple of European carriers that have latency improvement projects that are underway. And the LEOs, that's a real important part of their expansion as well, which aligns well or aligns with what we've seen around AI specific SLAs.
(02:58):
These conversations reinforced our view that the network needs to evolve for these AI workloads and to being more deterministic. We had a few conversations around customers looking, again, at optical, at waves, and then back even IPMPLS is dedicated capacity and not shared capacity. Reinforcing, again, some of the results of a study that we did last year. So yeah, I mean, what we saw is a lot of what we thought before. It's customers needing the dedicated bandwidth, a low latency and high bandwidth and bandwidth that can scale on demand. So we saw a lot of reinforcements to some of the beliefs that we had in the past.
Guy Daniels, TelecomTV (03:46):
That's good to hear, Steve. And conversations. I can't believe Ray missed that. That should have been number one on his list, conversations. We desperately needed conversations. Nico, let's come to you next. I'm keen to ask you, is AI native now becoming regarded as the norm? Is it becoming a new operating model for telcos?
Nico Marziliano, Wind River (04:07):
Yes, it is. And thanks, first off, for inviting and for the good question. The AI now has been the late motiva in the show. Everybody has to have at least one AI slogan somewhere, but I think we are moving from a buzzword from what was supposed to be a nice interesting new topics, new toys for the industry into a practical adoption. So when you mention the topic of being the new operating system of the network, yes, indeed. So not only the people has been talking about AI infrastructure, but also client and partners, they have been reaching out and talking about how do we put on the ground an AI infrastructure that will serve the next version of our networks deployment. And that's why it's linked also to the other topic, which is edge AI. So when we say about edge AI, edge is because it's the new frontier, so is the new border of the networks.
(05:11):
In many cases we talk in the past and the far edge, that becomes the intelligence going to be so distributed. In fact, in Windrill, we talk a lot about what we say intelligent system and intelligent edge because the more the intelligence will be distributed, the more we'll reach an edge of networks domains as well as technology. And one of the things, for example, we demonstrated in our booth was the physical AI. So we had a couple of showcases where we demonstrated how an operator, a partners, a client that can monetize using AI. And probably the last frontiers into the AI space is exactly the physical AI. It used to be a buzzword. As I said, it used to be an ideal topic. Now it's become reality. So we are moving from the pure intelligence at the core, at the cloud level, down to the physical system in every single device that can be a drone, can be a robots harm, can be a car, it can be any device that will require intelligent action at the very far edge.
(06:34):
That is exactly what we have been listening to the show. And I think that we did in our booth, but walking around, we see a lot of robots and cars and drones and everything that was possibly connected into an intelligent edge.
Guy Daniels, TelecomTV (06:50):
Great. Thanks very much, Nico. And I think that term physical AI is going to stick. Rob, let's come to you next. Again, on the AI topic, which was so prevalent to MWC, from your discussions and what you saw MWC, are you now seeing your telco customers look to operational AI across their businesses?
Rob McManus, REd Hat (07:13):
Yeah. Thanks, Guy. Great to be here. It's a question that was asked many times during the show. A lot of service providers, telecommunication service providers have been experimenting. I think it's fair to say lots of POCs, proof of concepts. So it's a very different sort of ballgame when you need to put this into production. So definitely lots of conversations around operationalizing this at scale. We had a very clear focus on sovereign AI and how to build those capabilities because these capabilities are really necessary to capture these high value market opportunities. So if service providers can actually get their hands around this, they can simplify and scale. So to do that, they really need an integrated platform to make that happen with associated tooling because without that, they can't reduce the operational burden. Another thing that we spoke at length about was inference. So service providers are looking to train smaller models using their own data.
(08:25):
How do they get that out there? How do they set up the endpoint? And again, there's a couple of really interesting open source projects. One's called VLLM, and that tends to be focused on inferencing on a single server, but the one that is sort of at the forefront of a lot of people's minds, but it's quite nascent is something called LLMD. So that takes a lot of the work from VLLM and uses it at scale. And that's particularly interesting for service providers because of their very distributed networks. So LLMD allows you to inference a bunch of models in different places, but it has an intelligent sort of routing capability higher up. So that allows you to sort of direct AI specific queries to certain pockets of resource. Now with AI, there's two main sort of phases. There's something called a prefill phase, which takes the prompt in, and there's something called a decode phase, which then obviously returns the results.
(09:25):
Now, GPUs, as we probably all appreciate, is a very finite resource, very expensive resource. So if you can set this resource up very optimally and have this intelligent routing layer that LLMD provides, you can very intelligently and optimally direct the AI workloads to the correct place. So a lot of conversation around how to do that efficiently.
(09:50):
Service providers really want to own their own AI strategy. I think it's fair to say many have been down the hyperscaler route because that's the easy button. You can do a lot of things with hyperscalers and that will prevail, I'm sure. But again, touching back to the sovereign AI point that I made a few moments ago, they really are looking to own their AI strategy. And what that really means is they want to run all sorts of different types of AI, predictive, generative, and as I mentioned a few moments ago, training models securely on their own data. And then last but not least, agentic AI, again, was a very popular topic. How do they build agents? How do they build intelligent, autonomous, multi-step workflows, and how do they access the plethora of data they have in their networks? Because we know, again, this is a big challenge.
(10:43):
So to summarize, I think those were the key conversations, but a key element of this is this common platform capability that you need because without that, you can't deliver operational consistency, you can't do it at scale over a distributed architecture. And the other benefit that the common platform brings, it allows you to deploy different types of workloads. And AI at the end of the day is just a different type of workload. So the platform perspective is very important here.
Guy Daniels, TelecomTV (11:16):
Thanks very much, Rob. It's such a fantastic time to be in telecoms, and yet the learning never stops, does it? I'm learning more and more every day in this business. John, I want to move on to you because we should never forget the customer and the commercial objectives of telcos. So what's your take on the AI developments at MWC and how they underpin the commercial needs and business models of telcos?
John Abraham, Appledore Research (11:46):
Sure. Hi. Anyway, Guy, great to be back. So from a customer facing perspective, I think the biggest trend that really stood out for me was the emphasis on the underpinnings of AI. If you go back the last few years, I mean, first beginning with generative and then onto AgentTech, we have had a lot of emphasis on the applications, what AI can do for you. And the way we see it at Appledore is that we see it as a three-tier pyramid at a high level with applications at the top, the ontology or the semantics in the middle and the data fabric at the bottom. And by and large, if you go back the last two or three years, the emphasis was on the application layer, what generator or then agentic AI can do for you. And this time around, I felt that the focus has shifted downward on the underpinnings.
(12:45):
The data fabric, we all know that telcos have a problem with legacy silos, and that becomes a particularly difficult challenge now because the effectiveness of AI is closely tied to its ability to understand context. And if data is split across, say, 10, 20, or 30 different systems, how exactly are we going to make that happen? So there is far more greater emphasis on bringing that together, different strategies being applied or explored at this phase, but I don't think it's going to be one path to fix all that disparate data that we have across multiple systems. I believe operators are likely to pursue different paths to bring that data together. On the ontology side, again, this is quite a unique and distinct challenge given the volume of data that we have. How do you map them? How do you try and build a semantic backbone across a multi-vendor environment, some of which actually go back to three decades.
(14:00):
And ontology, one of the challenges we have there is that we don't have a consensus on how to define it. So we are close to getting into situations where we need to have different vendor solutions talk to each other that requires a common framework and we don't yet know how to define ontology. So that's going to be something that remains in focus because the ability of the AI application at the top is critically tied to the ontology, the strength of that, semantics, and the ability to bring all the different desegregated data together. The other point I will call out is the growing interest in outcome-based pricing models. I mean, strictly speaking, this is not a new thing. We have seen different forms of SLA-based pricing models in telco in the past, but that used to be defined mostly from a functional aspect like uptime or things that can be linked to technology or can be easily quantified.
(15:11):
But now that we are looking at AI and its potential impact on the bottom line and its impact on improving customer experience or the NPS codes and so on, how do you exactly quantify it? And those are questions ahead of us, I think, for this particular business model to mature. We still don't have a clarity on the definition. And then one part of the challenge we have there is operators are very concerned about any sort of KPEx that gets shift to OPEX for obvious reasons. So this has to be navigated carefully, but we are probably in the early days of this conversation about how to define what an ideal outcome is and how can vendors get paid for the services they offer linked to those outcomes. So watch this space. I think there's a lot more conversations coming up on these topics.
Guy Daniels, TelecomTV (16:10):
Oh yes, absolutely. I'm sure there are, John. Thanks very much for your insights there. And as you say, it's early days on a number of these issues and we will follow them with great interest. And James, last but not least, you were there daily reporting from the show floor, MWC followers on telecom TV. What did you notice about AI at the event? And in particular, how telcos are reacting? Are they moving fast enough? What are your takeaways?
James Pearce, TelecomTV (16:37):
Thanks, Guy. I think there's a really interesting area around AI and the RAN, and I think especially how best to leverage it. And the guys have touched upon it a little bit because there's still a really steep learning curve for a lot of the operators. The developments that we've seen in AI over the last few years have been rapid compared to a lot of previous technological developments. We've gone from that kind of initial AI boom and when ChatGPT launched through to the Copilot model and then into the Agentic era, and now we're kind of looking forward at potential autonomous networks and then also robots on networks and things like that. So there's big questions as to how telcos keep up with that. And I think it's really forced a change in the industry's approach where the operators and the vendors as well are all having to work together to be able to bring a rapidity to that lifecycle, product life cycle.
(17:39):
And it also ties into the customer experience as well, because their customers want products faster. They want to be able to use AI tools faster. So the telcos are having to work with their partners to develop them a lot quicker. I mean, Ray in his piece used the term celerity, which I still think might well be a made up turn, but I think it does kind of sum up that feeling that everything is moving a lot faster and AI is playing a big role in it. But I think also it's bringing interesting new people into the conversations. We touched upon how important the role of the GPU is, but also how expensive they are and some of the challenges that they cause. But there's still big questions as to where the GPU's going to sit and where it's best in the network and how they get the most out of it.
(18:27):
So if you're an NVIDIA, you're kind of saying, "We want it to be everywhere." Of course you are. You're the one selling the GPUs. But other vendors that I had conversations with were kind of less keen or said, "We don't need a GPU in everything." And the telcos kind of feel the same as well. They're looking at it and trying to get an assessment as to where they'll benefit most. So it's a really interesting time, but it's definitely speeding things up in a way that I don't think we've seen before.
Guy Daniels, TelecomTV (18:55):
No, it's good to hear. James, thanks very much. And I do like what we say there about bringing new people and sectors into the conversations, exactly what telcos need. They need that engagement. Right. Well, that was AI, which does lead us very nicely into automation and autonomous networking. Rob, I'm going to come back to you first. Did you see any evidence that telcos are operating autonomously to deliver OPEX reductions and efficiencies?
Rob McManus, REd Hat (19:25):
Yes, but they're still doing it primarily in silos. So that trend hasn't changed. But I think before I get into sort of talking about where we're going with autonomous networks, it's really predicated on higher level business drivers. So there's three of them, and these drivers haven't actually changed in a number of years. So ultimately, service providers want to drive new business and revenue. They want to optimize cost and mitigate risk. And then thirdly, they need to evolve their network. So automation touches all three of those. Primarily, from our perspective, it's the cost optimization one that's most sort of important right now. So during the show, we're seeing now a lot of shift towards automating intelligently and operating autonomously. So what do we mean by that? So as I said at the beginning, so a lot of service providers have done automation projects in different places within the organization, typically disconnected, but they're now looking at it more holistically.
(20:35):
They know they can use automation to streamline operations, and if they can do that across the network, this delivers very measurable OPEX reductions. So that shift then is automation that embraces AI, and I'm sure that doesn't come as any surprise, but to do that, you need the right components to build what we call autonomous intelligent networks. So how we define that then at Red Hat is that an intelligent autonomous network is a network that follows recognized frameworks and also contains domain specific automation. So that will be at the lower layers. Some examples, the radio access network, the core network, for example, those are the domain specific layers of automation, but they need to connect up into higher levels like business and service layers. And again, that touches back to frameworks. There are frameworks out there that sort of deems as good practice to do that.
(21:38):
Also, the use of cloud native methodologies for what we call day zero, one, two operations. And some examples of those are something called zero touch provisioning, which as the name implies, you press a button and then it configures both the infrastructure and the workloads. Another one is infrastructure and policy as code. And also, we see a lot of telecommunication service providers now looking at GitOps. So this notion of a single source of truth, which again, ties back into infrastructure as code. If it's a single source of truth, that sort of confidence that you can do different configuration changes carefully. And obviously that touches back to the mitigating risk point that I mentioned. And then thirdly, the ability to enforce security. So with these networks, again, typically because they're very distributed, there's a lot more in terms of what we call attack vectors now. So the ability to enforce security through automation is very important, certainly with these environments, and they need to do that to make sure their networks are both reliable and are obviously in compliance because failure to do that, as we've seen many times in the media, is catastrophic.
(22:53):
So those were the conversations we had around automation, but to summarize it, it's about connecting the automation to the AI, and that brings in this intelligence to it.
Guy Daniels, TelecomTV (23:05):
Great. Thanks very much, Rob, and break out of those silos. Steve, what about you? What was your take on autonomous networking at MWC? Any examples you can give us? Especially one of the areas that we are really focused on at the moment is use of agents in networks. Anything you saw there?
Steve Daigle, HPE (23:25):
I did, and autonomous was actually on almost every one of the customer conversations. Some of my customers are deploying AI powered digital twins. So it's the combination of AI and automation, as we said previous speakers said, especially digital twins, especially for the control plane. But something that I saw that was interesting was the partnering of the hyperscalers with our service providers for their automation. And I think that was mentioned earlier also. Google had a few announcements at MWC on their network digital twin. It's with the immediate root cause analysis and they claim predict failure propagation. And a few of my customers are announced to be working with GCP DT as one is work, they've actually demonstrated their autonomous fault remediation, Vodafone using AI agents for the real-time service degradation prediction. So I think customers are now in production, moving from pilot to production. They're deploying these AI agents to configure their networks with intent.
(24:39):
I met with one who plans to be ZTP, zero touch provisioning completely by 2030, and then fully self-driving after that. So I really think, Guy, we're moving from experimenting with AI to actually running some of at least a portion of the networks in AI and very domain specific at the same time.
(25:02):
A couple of things that we, announcements that we made around AI co-polyfire routing director, but the idea around those products is to expose the WAN automation capabilities to an LLM based system for natural language interaction, what if scenarios and autonomous operations? There's the Agentic that you were mentioning earlier. And then another thing I found was, and this isn't a revelation, but customers are in really different points along the whole continuum of self-driving, but some are going beyond remediation to provisioning for adaptive growth. And I think that's further off into the future, but it calls back to a few conversations that we've already had here or examples that we've had here where you can actually use the network to adapt itself to where you can maximize the GPU utilization. And that's one of the ultimate goals of a network that provides itself like a grid for AI consumption.
(26:10):
We talked again about the domains, and I heard that quite a few times at the show. I think that we have to master the single domain autonomy first as we grow and on our autonomous journey, so to speak, you get to level three autonomy within a domain, and then you build that cross-domain visibility with human interaction and orchestration. And then finally, we see that you can close the loop on cross-domain decisions. So that's kind of the spectrum of the continuum or the journey that we see. And in the conversations that I had, Guy, there were a few customers that were definitely in different places along that continuum.
Guy Daniels, TelecomTV (26:56):
Thanks very much, Steve. It's really good to hear the move towards production. And it's secretly good to hear about the level of ambition from some of these telcos. Let me just move on briefly to Nico. I'm going to come across to you because I don't think yet we've touched yet on AI RAN, and that certainly we picked AI run up as one of the major talking points at this show. Obviously, the jury is still out to a large extent, but what was your takeaway on the viability of AI RAN?
Nico Marziliano, Wind River (27:28):
Yeah, very good question. The topic of AIRN came across, and let me also bridge the two questions, the previous one about autonomous and AI networks and the AI RAN. In the recent past that we have been talking a lot about moving from a traditional RAN to VRAN and to ORAN. This day, everybody talk about AI RAN, which is a natural evolution of the cloudification of the networks moving, as I said before, the intelligence to the far, far edge of the networks. And the radio station are definitely from a network perspective, probably the far edge you can reach. So AI came to touch the last miles into the network side. That's brought a lot of implication back to what mentioned by my colleagues before about how do you orchestrate and manage in autonomous way those devices and those components like radio station, like edge site, core site, application.
(28:40):
AI is a kind of holy grail. It's a promising to bring the level four of the autonomous networks that everybody's looking at. In the recent past, many has been in a good way implementing automated networks function. I think that has been demonstrated and is in place in many operator, the possibility by domain to automate things, making networks becoming autonomous, so taking independent decision AI is perceived to be the possibility to bring that to life. Is it going to be true? It's going to be for real? Again, my colleagues before articulated very well from the difference between application, the ontology layers and the physical layers, as well as the automation layers mentioned about having the LLMD capability or to make inference in the proper way so that the different devices talk inside the domain or infra domain in the same way And having the same data source available, that is still to be solved, but a lot of steps has been done.
(30:08):
So AIRAN, back to the original topic, it's full of expectation. Talking with client, everybody's looking at AI RAN to be delivering what has been promised probably from Eron Oran perspective. So the virtualization and the cloudification of RAN, bringing the final automated, autonomous networks delivery with the possibility to upsell having additional workload and component at the very far edge. That will bring still enormous level of topic to be solved. Before it was mentioned, the GPU utilization is going to be GPU at the RAM side. I have seen partners presenting GPU in the box. Others not believing the GPU will be really down to every single site by the cost. If I ever take maybe anticipating future question or next question from you, is one of the interesting topic I see as a late motive in the various conversation was the cost of the silicon and the challenge of GPUs as well as RAM availability, which is now, in my opinion, becoming a little bit of a critical point in the boost of the AI explosion, which for me is really is a natural trend, is happening, probably maybe just impacted by this potential cost factor of GPUs and RAM availability.
Guy Daniels, TelecomTV (31:51):
So those were our takeaways from the show. What about anything else? Any surprises, anything you heard or saw that you weren't expecting? James, let me come to you first. What else caught your attention to MWC?
James Pearce, TelecomTV (32:06):
I think it was a really interesting year for non-terrestrial networks. And it feels a bit old-fashioned in a way talking about it, because if you went back 10 years or so, you'd go to see ... They'd all have stands at the conference, but you'd go around those stands and the satellite companies would be a little bit quiet. Perhaps I'm being a bit unfair. Somebody might come back to us and say, "No, no, we used to have loads of visitors." But the fact is there's definitely a rise in the conversation, the interest. And a lot of it's being driven by two factors, I think. The first is just clearly the economic factors. So because of Elon Musk, because of SpaceX and the ability to send rockets back into space, it's easier to put satellites up there. More satellites means that you can cover a wider area and that makes things much better.
(32:57):
For the economics, it brings down the cost, the wholesale cost for both broadband. And the other area that's really, really interesting is directed device, which has had a bit of a comeback. And I think you can see that in the way a lot of the operators are partnering with the likes of Starlink. VMO2 in the UK launched their service recently. O2 now offers satellite services. We've seen it with T-Mobile and Deutsche Telecom, but also Vodafone have invested a lot of money in their partnership with AST Space Mobile. And they announced the new company, the European Venture, just ahead of the conference. And they announced a load of partners who have taken them up orange among a bunch of really big name operators who are going to use their services for direct to device. And I think what's really interesting is you're kind of looking at a market going, well, satellite wasn't that relevant a few years ago, but now it's got a two-pronged attack.
(33:57):
There's real interest in the broadband side, but also in that director device. And there was some research that kind of came out before the show that really said that the end users are actually interested in it as well. Customers who had previously maybe been used to having hot spots, whether it's in their home or living in the middle of the Welsh Mountains or somewhere like that, they were used to not getting connectivity. They now see an opportunity to get it and they really, really want that. Whether people will be able to be willing to pay extra for it remains to be seen, there's still question marks around it, but I think it's fair to say if you'll forgive the pun, the satellite really took off this year.
Guy Daniels, TelecomTV (34:41):
I'm not sure we can forget that pun, James. It was a terrible pun, but thanks very much. Absolutely. There was a lot more NTN and satellite interest at the show this year. Rob, what about you? What surprised you this year at MWC?
Rob McManus, REd Hat (34:56):
The lack of appetite to talk about, dare I say it, 6G. So we just didn't really go into any details on 6G, which was somewhat surprising from having been involved with Mobile World Congress for far too many years. If there's another G on the horizon, typically it dominates the conversation. Just wasn't the case this time. So from the conversations I personally had, I think there's still wide acknowledgement of the slow uptake of 5G services and also deployments. We know as a fact that a lot of networks are still on what we call non-standalone, which essentially is a 4G network with 5G radios plugged into it. And those work fine. They give increased speeds and operators are monetizing from them. There's still no killer application. There are 5G standalone networks, of course, in various parts of the world. And we also heard when we started to sort of ask and put feelers out about 5G advanced, limited interest in that, a lot of operators really just still struggling to monetize these huge investments they've made a long time, well, a number of years ago now, and it's just not happened.
(36:17):
So as James said a few moments ago, the focus in the sort of area of 5G was non-terrestrial networks. Again, we heard that in terms of maybe seen as a competitive differentiator for coverage perspective. And in the very few conversations, we did touch on 6G, I should say, very much needs to be a software upgrade we heard over and over again. I think these huge investment cycles that we saw on 5G aren't going to surface, in my opinion, for 6G. So very much it needs to be software. And once we get there, it also needs to be AI native rather than infusing AI into different things. So again, came as a bit of a surprise, but I think it was interesting to see that when we touched on the topic of 6G, it was a consistent, yeah, let's not go too much into the weeds there at the moment.
Guy Daniels, TelecomTV (37:17):
Yeah, absolutely. Right there, Rob. And it's really refreshing. We know the research is going on. We know the study work's going on. We know we're a year away from standards even being written for this, but it's like AI has stolen the news cycle. So the marketing side, we can focus on AI. We don't need to talk about 6G. Nice and refreshing. I'm going to come to Nico first and also James. So we just want to pick up some points on that. Nico.
Nico Marziliano, Wind River (37:40):
Yes, thank you. I tend to agree that 6G was a bit surprise, but at the same time, it's not a surprise. I remember in one of your telco debate we had a year ago, there was this conversation about CG, 7G, 5G, and people start saying, "Nico, don't talk about Gs. I think we still need to materialize what the 5G was promising." And until we solve it, don't talk about Gs anymore. So yes, it was a surprise, but it was definitely, I would say it was an industry feedback that it's not time to talk necessarily about Gs. Of course, we know that the next six will bring the further acceleration of cloudification and probably autonomy and AI, but the people prefer to talk about the real AI. And in fact, what I was mentioning before was I was surprised to see a lot of physical AI.
(38:47):
I mean, we, as a company, we presented a couple of cases, we bet on it, but we were not 100% sure that all the others were looking to this kind of next level of AI application. And we find surprising a lot of booth and partners and players presenting their own view of what the physical AI looks like. So definitely that was interesting. I have a comment on the non-terrestrial networks. I think that there is a very interesting angle that has been brought here. It's going to be maybe the next last frontiers where operator, yes, Vodafone, Timore, but many others are experimenting. I know that, for example, recently VMO2 in the UK, they connected with a satellite service. And so satellites will be clearly one of the topic where operator are investing just to provide connectivity or to provide AI distributor as some of the big players are try to move and transfer their data center and AI on the space, so basically shipping on the moon, but that will be definitely the next frontier I'm looking for.
(40:05):
So yes, that was another interesting angle. Thanks for reminding.
Guy Daniels, TelecomTV (40:10):
Thanks, Nico. I mean, I'm still concerned about the uplink power budget from terminals to NTNs, but let the researchers work that one out. James, I promise to come back to you.
James Pearce, TelecomTV (40:21):
Yeah, I think Rob made a really great point about 6G and the fact that you could walk around the halls and you'd barely see a sniff of it. And prepared, if you remember when it was the lead up to 5G, you had politicians across the world talking about it was getting really, really excited. It was in the mainstream news, not just the industry news, but now there's a certain reluctance around it. And I actually think it's down to AI. I think AI has caused the problem because essentially AI has fallen in the middle of a life cycle between two generations. It's fallen between 5G, which wasn't designed specifically for AI and 6G, which is going to have to be AI native, but 6G is still some ways away, as you touched upon, the standards aren't going to be with us for a few years. And there's actually a worry within the industry, within some of the operators that there's going to be a kind of too much of an acceleration to try to drive to 6G.
(41:17):
As Rob said, they haven't seen the benefits of their investment in 5G. A lot of them, they've not made the money back from it. So they don't want to go into that cycle of plowing loads of money into new 6G infrastructure. But at the same time, they've got this issue where if you remember in the buildup to the launch of 5G, there was a lot of push to try and be the first to get to 5G. And there was a lot of marketing spin around some of the launches, certainly from some of the US operators and so on as to where they'd take 4G, but it wouldn't be and rebrand it 4G plus and rebrand it as 5G or ... There was a lot going on, which was kind of slightly, I wouldn't say misleading, but made it look like 5G was here when it wasn't in.
(42:14):
The 5G standalone certainly was some years off. With the Olympics in the US in 2028, that's obviously a big wart mark. And there's always a desire across some spectrums to have a big announcement and have something big technologically. And LA have already announced that they're going to have flying cars and stuff like that in operation. That's one of the big plans for it. You need good connectivity to be able to do that. So you can see what's going to happen. You can see the potential that there'll be politicians pushing for 6G to launch and it won't be ready. The industry won't be ready. The operators certainly won't be ready. And I've heard this from several people high up at several operators when we're having conversations over TAPAS in Barcelona and they're saying, "We don't want to talk about 6G at the moment. We're not there." But at the same time, there'll be pressure elsewhere to get onto 6G sooner rather than later.
(43:14):
And I can see there being a real potential inflection point around the 2028 Olympics.
Guy Daniels, TelecomTV (43:20):
Yeah, James, you're absolutely right. There's also the spectrum angle in the US to consider there, which is a bit different from elsewhere, but I guarantee ... Oh, and it was misleading that those early marketing moves towards 5G, which weren't 5G, that really was misleading and it was proven. We'll see the same with 6G. I guarantee it. Someone can't resist it and we'll have operators coming out with these services before pre-standard services, I'm sure of it. Right. We've just got about time for one more take or question. I'd just like to ask you if anybody's got any feeling about the industry post MWC, how we're feeling more confident, less confident, and how is this going to translate into the next 12 months? What might we see? Steve, can I pick on you? I'm sure you've got some views on where this might be going, and then we'll come across to John as well.
Steve Daigle, HPE (44:13):
Sure. What I thought I saw, I'm a positive guy, so I'm always going to have the positive takeaways from what we saw. But around the adoption of AI, but not only that, the virtualizing of the RAN made a lot of good progress. I saw some things like T-Mobile's announcement about their Magenta AI call assistant, a partnership they have with a company called 11 Labs, where you can get on a call and say, "Hey, Magenta," and they'll do live translation of 50 languages while you're on the call and do contextual assistance. These kind of things are pretty, I mean, it's pretty exciting. I know you can also do it on your phone using an app, but not using an app and using the actual core network, using it in real time and have AI applied there, that it's kind of exciting to me things that we saw that are very, I think, good for the future.
(45:14):
Then there are a few other things like that partnership with the hyperscalers between the SPs, they've done that dance before and it hasn't quite worked out, but it seemed to me as though there's more of that going on around AI and the autonomous conversations. So it'll be interesting to see how this plays out within a year between some of the cool things we saw, lots of robots and some of the other things that we should probably look at a little bit more closely. The 6G conversation was a great one. But yeah, I mean, I'm looking forward to next year, but I'm ready for there to be more than 360 days between now and then to recover.
Guy Daniels, TelecomTV (45:57):
Yeah, nicely said, Steve. Absolutely. Yeah. And yeah, we're all positive. I'm sure all positive. John, are you positive?
John Abraham, Appledore Research (46:05):
I am, Guy. I'm quite optimistic. Remember, we are in unfamiliar territory in terms of how quickly a new technology has been adopted by telecom operators. It was generally considered to be quite conservative in terms of experimenting with new technologies. So yet we have seen a lot of value additions already. And as I mentioned earlier, one of the key distinctions between previous years and this year has been that there is a less focus on chasing the shiniest use case like we have seen in the past couple of years and far more emphasis on how do we actually get to value and what does it take to get there? Because there's a growing awareness there. It's one thing to demo an application, another thing to actually set it up within a complex legacy oriented environment. And we have already seen a lot of, let's say, niche or compartmentalized deployment say, such as within the ... And I'm talking about it mainly from a customer facing perspective because this has been the most popular landing spot for AI, at least starting from the initial years, given that a lot of this was inspired by what other enterprises were doing.
(47:17):
So we are seeing a lot of traction in terms of value addition, say, within the call center domain, improving the efficiency of the agent by having a couple of demos talked about how the screen changes based on what the AI understand the customer is talking about. So the agent doesn't have to go and click on multiple options. Automatically, the screen changes or the ability to monetize different types of advanced use cases, either because now the time to market is faster because a lot of these things are highly personalized and ready to go. Even the ability to do upsell, just because the AI provides a script. So we have seen a lot of this being deployed within compartmentalized use cases. And I think more and more, the reason for my optimism is that we are seeing far more emphasis on the data and the ontology layers that can bring all these disparate pieces together, because I believe that's the only way you can scale it, especially across a complex environment.
(48:25):
So I'm quite optimistic based on what I've seen in terms of operator, how close they are. There's still some ways to go, but at least I feel they are on the right path.
Guy Daniels, TelecomTV (48:35):
That's a really good way to end this show. John, thanks very much for optimism and thank you all. We have to leave it there for now. We are out of time, but it's been a great conversation, a lot of great takeaways, a lot of interesting insights from the event. Thank you very much for taking part in the program today. You can find all of our coverage from MWC 2026 on our website. Just go to the spotlight on 5G section, news, interviews, panels, and more. Well, that's all for now until next year and MWC 2027. From all of us here, thank you for watching and goodbye.
Hello, you're watching TelecomTV and this is the spotlight on 5G After Show. I'm Guy Daniels, and today we are going to wrap up our coverage of the MWC 2026 event. What were the highlights, takeaways and talking points, surprises, good and bad. And what have we learned about the state of the industry and its direction for the next year? Well, I'm delighted to say that joining me on the program to share their expert findings with us are Steve Daigle, who is global head of Telco Systems Engineers at HPE. Nico Marziliano, who is regional VP Telco Sales for Wind River. Rob McManus, who is principal product marketing manager with Red Hat, John Abraham, who is principal analyst at Appledore Research, and James Pearce, editor of our very own telecom TV. Hello, everyone. It's really good to see you. Thanks so much for taking part today.
(01:11):
Hope you are all now fully recovered from the exertions of MWC. Now, our editorial director, Ray La Maistre, unveiled his four C's takeaways last week, connectivity, collaboration, context, and celerity. Don't ask. So let's hear what your key takeaways were and got to start, of course, with AI. It dominated the show and it was part of so many of the conversations that we had. And I know you all have views on this, but Steve, I'll come to you first. What has been the impact of AI on the network?
Steve Daigle, HPE (01:50):
Hey, thanks, Guy. We had a really good show and it's always a good show, but it was our first time with HPE and we had some significant announcements around some of our products. But I wanted to add a C to your list there to raise a list about conversations. We had conversations around AI adoption, but really around the networks to support the growth and the new types of WAN traffic that we're seeing with AI adoption, especially the importance of latency capacity and dedicated AI paths. And I tell you, latency is just becoming more of a conversation that we're having with customers than it was before. It's like the broadband measurements and the speed tests of the past. It's more and more about latency. I met with a couple of European carriers that have latency improvement projects that are underway. And the LEOs, that's a real important part of their expansion as well, which aligns well or aligns with what we've seen around AI specific SLAs.
(02:58):
These conversations reinforced our view that the network needs to evolve for these AI workloads and to being more deterministic. We had a few conversations around customers looking, again, at optical, at waves, and then back even IPMPLS is dedicated capacity and not shared capacity. Reinforcing, again, some of the results of a study that we did last year. So yeah, I mean, what we saw is a lot of what we thought before. It's customers needing the dedicated bandwidth, a low latency and high bandwidth and bandwidth that can scale on demand. So we saw a lot of reinforcements to some of the beliefs that we had in the past.
Guy Daniels, TelecomTV (03:46):
That's good to hear, Steve. And conversations. I can't believe Ray missed that. That should have been number one on his list, conversations. We desperately needed conversations. Nico, let's come to you next. I'm keen to ask you, is AI native now becoming regarded as the norm? Is it becoming a new operating model for telcos?
Nico Marziliano, Wind River (04:07):
Yes, it is. And thanks, first off, for inviting and for the good question. The AI now has been the late motiva in the show. Everybody has to have at least one AI slogan somewhere, but I think we are moving from a buzzword from what was supposed to be a nice interesting new topics, new toys for the industry into a practical adoption. So when you mention the topic of being the new operating system of the network, yes, indeed. So not only the people has been talking about AI infrastructure, but also client and partners, they have been reaching out and talking about how do we put on the ground an AI infrastructure that will serve the next version of our networks deployment. And that's why it's linked also to the other topic, which is edge AI. So when we say about edge AI, edge is because it's the new frontier, so is the new border of the networks.
(05:11):
In many cases we talk in the past and the far edge, that becomes the intelligence going to be so distributed. In fact, in Windrill, we talk a lot about what we say intelligent system and intelligent edge because the more the intelligence will be distributed, the more we'll reach an edge of networks domains as well as technology. And one of the things, for example, we demonstrated in our booth was the physical AI. So we had a couple of showcases where we demonstrated how an operator, a partners, a client that can monetize using AI. And probably the last frontiers into the AI space is exactly the physical AI. It used to be a buzzword. As I said, it used to be an ideal topic. Now it's become reality. So we are moving from the pure intelligence at the core, at the cloud level, down to the physical system in every single device that can be a drone, can be a robots harm, can be a car, it can be any device that will require intelligent action at the very far edge.
(06:34):
That is exactly what we have been listening to the show. And I think that we did in our booth, but walking around, we see a lot of robots and cars and drones and everything that was possibly connected into an intelligent edge.
Guy Daniels, TelecomTV (06:50):
Great. Thanks very much, Nico. And I think that term physical AI is going to stick. Rob, let's come to you next. Again, on the AI topic, which was so prevalent to MWC, from your discussions and what you saw MWC, are you now seeing your telco customers look to operational AI across their businesses?
Rob McManus, REd Hat (07:13):
Yeah. Thanks, Guy. Great to be here. It's a question that was asked many times during the show. A lot of service providers, telecommunication service providers have been experimenting. I think it's fair to say lots of POCs, proof of concepts. So it's a very different sort of ballgame when you need to put this into production. So definitely lots of conversations around operationalizing this at scale. We had a very clear focus on sovereign AI and how to build those capabilities because these capabilities are really necessary to capture these high value market opportunities. So if service providers can actually get their hands around this, they can simplify and scale. So to do that, they really need an integrated platform to make that happen with associated tooling because without that, they can't reduce the operational burden. Another thing that we spoke at length about was inference. So service providers are looking to train smaller models using their own data.
(08:25):
How do they get that out there? How do they set up the endpoint? And again, there's a couple of really interesting open source projects. One's called VLLM, and that tends to be focused on inferencing on a single server, but the one that is sort of at the forefront of a lot of people's minds, but it's quite nascent is something called LLMD. So that takes a lot of the work from VLLM and uses it at scale. And that's particularly interesting for service providers because of their very distributed networks. So LLMD allows you to inference a bunch of models in different places, but it has an intelligent sort of routing capability higher up. So that allows you to sort of direct AI specific queries to certain pockets of resource. Now with AI, there's two main sort of phases. There's something called a prefill phase, which takes the prompt in, and there's something called a decode phase, which then obviously returns the results.
(09:25):
Now, GPUs, as we probably all appreciate, is a very finite resource, very expensive resource. So if you can set this resource up very optimally and have this intelligent routing layer that LLMD provides, you can very intelligently and optimally direct the AI workloads to the correct place. So a lot of conversation around how to do that efficiently.
(09:50):
Service providers really want to own their own AI strategy. I think it's fair to say many have been down the hyperscaler route because that's the easy button. You can do a lot of things with hyperscalers and that will prevail, I'm sure. But again, touching back to the sovereign AI point that I made a few moments ago, they really are looking to own their AI strategy. And what that really means is they want to run all sorts of different types of AI, predictive, generative, and as I mentioned a few moments ago, training models securely on their own data. And then last but not least, agentic AI, again, was a very popular topic. How do they build agents? How do they build intelligent, autonomous, multi-step workflows, and how do they access the plethora of data they have in their networks? Because we know, again, this is a big challenge.
(10:43):
So to summarize, I think those were the key conversations, but a key element of this is this common platform capability that you need because without that, you can't deliver operational consistency, you can't do it at scale over a distributed architecture. And the other benefit that the common platform brings, it allows you to deploy different types of workloads. And AI at the end of the day is just a different type of workload. So the platform perspective is very important here.
Guy Daniels, TelecomTV (11:16):
Thanks very much, Rob. It's such a fantastic time to be in telecoms, and yet the learning never stops, does it? I'm learning more and more every day in this business. John, I want to move on to you because we should never forget the customer and the commercial objectives of telcos. So what's your take on the AI developments at MWC and how they underpin the commercial needs and business models of telcos?
John Abraham, Appledore Research (11:46):
Sure. Hi. Anyway, Guy, great to be back. So from a customer facing perspective, I think the biggest trend that really stood out for me was the emphasis on the underpinnings of AI. If you go back the last few years, I mean, first beginning with generative and then onto AgentTech, we have had a lot of emphasis on the applications, what AI can do for you. And the way we see it at Appledore is that we see it as a three-tier pyramid at a high level with applications at the top, the ontology or the semantics in the middle and the data fabric at the bottom. And by and large, if you go back the last two or three years, the emphasis was on the application layer, what generator or then agentic AI can do for you. And this time around, I felt that the focus has shifted downward on the underpinnings.
(12:45):
The data fabric, we all know that telcos have a problem with legacy silos, and that becomes a particularly difficult challenge now because the effectiveness of AI is closely tied to its ability to understand context. And if data is split across, say, 10, 20, or 30 different systems, how exactly are we going to make that happen? So there is far more greater emphasis on bringing that together, different strategies being applied or explored at this phase, but I don't think it's going to be one path to fix all that disparate data that we have across multiple systems. I believe operators are likely to pursue different paths to bring that data together. On the ontology side, again, this is quite a unique and distinct challenge given the volume of data that we have. How do you map them? How do you try and build a semantic backbone across a multi-vendor environment, some of which actually go back to three decades.
(14:00):
And ontology, one of the challenges we have there is that we don't have a consensus on how to define it. So we are close to getting into situations where we need to have different vendor solutions talk to each other that requires a common framework and we don't yet know how to define ontology. So that's going to be something that remains in focus because the ability of the AI application at the top is critically tied to the ontology, the strength of that, semantics, and the ability to bring all the different desegregated data together. The other point I will call out is the growing interest in outcome-based pricing models. I mean, strictly speaking, this is not a new thing. We have seen different forms of SLA-based pricing models in telco in the past, but that used to be defined mostly from a functional aspect like uptime or things that can be linked to technology or can be easily quantified.
(15:11):
But now that we are looking at AI and its potential impact on the bottom line and its impact on improving customer experience or the NPS codes and so on, how do you exactly quantify it? And those are questions ahead of us, I think, for this particular business model to mature. We still don't have a clarity on the definition. And then one part of the challenge we have there is operators are very concerned about any sort of KPEx that gets shift to OPEX for obvious reasons. So this has to be navigated carefully, but we are probably in the early days of this conversation about how to define what an ideal outcome is and how can vendors get paid for the services they offer linked to those outcomes. So watch this space. I think there's a lot more conversations coming up on these topics.
Guy Daniels, TelecomTV (16:10):
Oh yes, absolutely. I'm sure there are, John. Thanks very much for your insights there. And as you say, it's early days on a number of these issues and we will follow them with great interest. And James, last but not least, you were there daily reporting from the show floor, MWC followers on telecom TV. What did you notice about AI at the event? And in particular, how telcos are reacting? Are they moving fast enough? What are your takeaways?
James Pearce, TelecomTV (16:37):
Thanks, Guy. I think there's a really interesting area around AI and the RAN, and I think especially how best to leverage it. And the guys have touched upon it a little bit because there's still a really steep learning curve for a lot of the operators. The developments that we've seen in AI over the last few years have been rapid compared to a lot of previous technological developments. We've gone from that kind of initial AI boom and when ChatGPT launched through to the Copilot model and then into the Agentic era, and now we're kind of looking forward at potential autonomous networks and then also robots on networks and things like that. So there's big questions as to how telcos keep up with that. And I think it's really forced a change in the industry's approach where the operators and the vendors as well are all having to work together to be able to bring a rapidity to that lifecycle, product life cycle.
(17:39):
And it also ties into the customer experience as well, because their customers want products faster. They want to be able to use AI tools faster. So the telcos are having to work with their partners to develop them a lot quicker. I mean, Ray in his piece used the term celerity, which I still think might well be a made up turn, but I think it does kind of sum up that feeling that everything is moving a lot faster and AI is playing a big role in it. But I think also it's bringing interesting new people into the conversations. We touched upon how important the role of the GPU is, but also how expensive they are and some of the challenges that they cause. But there's still big questions as to where the GPU's going to sit and where it's best in the network and how they get the most out of it.
(18:27):
So if you're an NVIDIA, you're kind of saying, "We want it to be everywhere." Of course you are. You're the one selling the GPUs. But other vendors that I had conversations with were kind of less keen or said, "We don't need a GPU in everything." And the telcos kind of feel the same as well. They're looking at it and trying to get an assessment as to where they'll benefit most. So it's a really interesting time, but it's definitely speeding things up in a way that I don't think we've seen before.
Guy Daniels, TelecomTV (18:55):
No, it's good to hear. James, thanks very much. And I do like what we say there about bringing new people and sectors into the conversations, exactly what telcos need. They need that engagement. Right. Well, that was AI, which does lead us very nicely into automation and autonomous networking. Rob, I'm going to come back to you first. Did you see any evidence that telcos are operating autonomously to deliver OPEX reductions and efficiencies?
Rob McManus, REd Hat (19:25):
Yes, but they're still doing it primarily in silos. So that trend hasn't changed. But I think before I get into sort of talking about where we're going with autonomous networks, it's really predicated on higher level business drivers. So there's three of them, and these drivers haven't actually changed in a number of years. So ultimately, service providers want to drive new business and revenue. They want to optimize cost and mitigate risk. And then thirdly, they need to evolve their network. So automation touches all three of those. Primarily, from our perspective, it's the cost optimization one that's most sort of important right now. So during the show, we're seeing now a lot of shift towards automating intelligently and operating autonomously. So what do we mean by that? So as I said at the beginning, so a lot of service providers have done automation projects in different places within the organization, typically disconnected, but they're now looking at it more holistically.
(20:35):
They know they can use automation to streamline operations, and if they can do that across the network, this delivers very measurable OPEX reductions. So that shift then is automation that embraces AI, and I'm sure that doesn't come as any surprise, but to do that, you need the right components to build what we call autonomous intelligent networks. So how we define that then at Red Hat is that an intelligent autonomous network is a network that follows recognized frameworks and also contains domain specific automation. So that will be at the lower layers. Some examples, the radio access network, the core network, for example, those are the domain specific layers of automation, but they need to connect up into higher levels like business and service layers. And again, that touches back to frameworks. There are frameworks out there that sort of deems as good practice to do that.
(21:38):
Also, the use of cloud native methodologies for what we call day zero, one, two operations. And some examples of those are something called zero touch provisioning, which as the name implies, you press a button and then it configures both the infrastructure and the workloads. Another one is infrastructure and policy as code. And also, we see a lot of telecommunication service providers now looking at GitOps. So this notion of a single source of truth, which again, ties back into infrastructure as code. If it's a single source of truth, that sort of confidence that you can do different configuration changes carefully. And obviously that touches back to the mitigating risk point that I mentioned. And then thirdly, the ability to enforce security. So with these networks, again, typically because they're very distributed, there's a lot more in terms of what we call attack vectors now. So the ability to enforce security through automation is very important, certainly with these environments, and they need to do that to make sure their networks are both reliable and are obviously in compliance because failure to do that, as we've seen many times in the media, is catastrophic.
(22:53):
So those were the conversations we had around automation, but to summarize it, it's about connecting the automation to the AI, and that brings in this intelligence to it.
Guy Daniels, TelecomTV (23:05):
Great. Thanks very much, Rob, and break out of those silos. Steve, what about you? What was your take on autonomous networking at MWC? Any examples you can give us? Especially one of the areas that we are really focused on at the moment is use of agents in networks. Anything you saw there?
Steve Daigle, HPE (23:25):
I did, and autonomous was actually on almost every one of the customer conversations. Some of my customers are deploying AI powered digital twins. So it's the combination of AI and automation, as we said previous speakers said, especially digital twins, especially for the control plane. But something that I saw that was interesting was the partnering of the hyperscalers with our service providers for their automation. And I think that was mentioned earlier also. Google had a few announcements at MWC on their network digital twin. It's with the immediate root cause analysis and they claim predict failure propagation. And a few of my customers are announced to be working with GCP DT as one is work, they've actually demonstrated their autonomous fault remediation, Vodafone using AI agents for the real-time service degradation prediction. So I think customers are now in production, moving from pilot to production. They're deploying these AI agents to configure their networks with intent.
(24:39):
I met with one who plans to be ZTP, zero touch provisioning completely by 2030, and then fully self-driving after that. So I really think, Guy, we're moving from experimenting with AI to actually running some of at least a portion of the networks in AI and very domain specific at the same time.
(25:02):
A couple of things that we, announcements that we made around AI co-polyfire routing director, but the idea around those products is to expose the WAN automation capabilities to an LLM based system for natural language interaction, what if scenarios and autonomous operations? There's the Agentic that you were mentioning earlier. And then another thing I found was, and this isn't a revelation, but customers are in really different points along the whole continuum of self-driving, but some are going beyond remediation to provisioning for adaptive growth. And I think that's further off into the future, but it calls back to a few conversations that we've already had here or examples that we've had here where you can actually use the network to adapt itself to where you can maximize the GPU utilization. And that's one of the ultimate goals of a network that provides itself like a grid for AI consumption.
(26:10):
We talked again about the domains, and I heard that quite a few times at the show. I think that we have to master the single domain autonomy first as we grow and on our autonomous journey, so to speak, you get to level three autonomy within a domain, and then you build that cross-domain visibility with human interaction and orchestration. And then finally, we see that you can close the loop on cross-domain decisions. So that's kind of the spectrum of the continuum or the journey that we see. And in the conversations that I had, Guy, there were a few customers that were definitely in different places along that continuum.
Guy Daniels, TelecomTV (26:56):
Thanks very much, Steve. It's really good to hear the move towards production. And it's secretly good to hear about the level of ambition from some of these telcos. Let me just move on briefly to Nico. I'm going to come across to you because I don't think yet we've touched yet on AI RAN, and that certainly we picked AI run up as one of the major talking points at this show. Obviously, the jury is still out to a large extent, but what was your takeaway on the viability of AI RAN?
Nico Marziliano, Wind River (27:28):
Yeah, very good question. The topic of AIRN came across, and let me also bridge the two questions, the previous one about autonomous and AI networks and the AI RAN. In the recent past that we have been talking a lot about moving from a traditional RAN to VRAN and to ORAN. This day, everybody talk about AI RAN, which is a natural evolution of the cloudification of the networks moving, as I said before, the intelligence to the far, far edge of the networks. And the radio station are definitely from a network perspective, probably the far edge you can reach. So AI came to touch the last miles into the network side. That's brought a lot of implication back to what mentioned by my colleagues before about how do you orchestrate and manage in autonomous way those devices and those components like radio station, like edge site, core site, application.
(28:40):
AI is a kind of holy grail. It's a promising to bring the level four of the autonomous networks that everybody's looking at. In the recent past, many has been in a good way implementing automated networks function. I think that has been demonstrated and is in place in many operator, the possibility by domain to automate things, making networks becoming autonomous, so taking independent decision AI is perceived to be the possibility to bring that to life. Is it going to be true? It's going to be for real? Again, my colleagues before articulated very well from the difference between application, the ontology layers and the physical layers, as well as the automation layers mentioned about having the LLMD capability or to make inference in the proper way so that the different devices talk inside the domain or infra domain in the same way And having the same data source available, that is still to be solved, but a lot of steps has been done.
(30:08):
So AIRAN, back to the original topic, it's full of expectation. Talking with client, everybody's looking at AI RAN to be delivering what has been promised probably from Eron Oran perspective. So the virtualization and the cloudification of RAN, bringing the final automated, autonomous networks delivery with the possibility to upsell having additional workload and component at the very far edge. That will bring still enormous level of topic to be solved. Before it was mentioned, the GPU utilization is going to be GPU at the RAM side. I have seen partners presenting GPU in the box. Others not believing the GPU will be really down to every single site by the cost. If I ever take maybe anticipating future question or next question from you, is one of the interesting topic I see as a late motive in the various conversation was the cost of the silicon and the challenge of GPUs as well as RAM availability, which is now, in my opinion, becoming a little bit of a critical point in the boost of the AI explosion, which for me is really is a natural trend, is happening, probably maybe just impacted by this potential cost factor of GPUs and RAM availability.
Guy Daniels, TelecomTV (31:51):
So those were our takeaways from the show. What about anything else? Any surprises, anything you heard or saw that you weren't expecting? James, let me come to you first. What else caught your attention to MWC?
James Pearce, TelecomTV (32:06):
I think it was a really interesting year for non-terrestrial networks. And it feels a bit old-fashioned in a way talking about it, because if you went back 10 years or so, you'd go to see ... They'd all have stands at the conference, but you'd go around those stands and the satellite companies would be a little bit quiet. Perhaps I'm being a bit unfair. Somebody might come back to us and say, "No, no, we used to have loads of visitors." But the fact is there's definitely a rise in the conversation, the interest. And a lot of it's being driven by two factors, I think. The first is just clearly the economic factors. So because of Elon Musk, because of SpaceX and the ability to send rockets back into space, it's easier to put satellites up there. More satellites means that you can cover a wider area and that makes things much better.
(32:57):
For the economics, it brings down the cost, the wholesale cost for both broadband. And the other area that's really, really interesting is directed device, which has had a bit of a comeback. And I think you can see that in the way a lot of the operators are partnering with the likes of Starlink. VMO2 in the UK launched their service recently. O2 now offers satellite services. We've seen it with T-Mobile and Deutsche Telecom, but also Vodafone have invested a lot of money in their partnership with AST Space Mobile. And they announced the new company, the European Venture, just ahead of the conference. And they announced a load of partners who have taken them up orange among a bunch of really big name operators who are going to use their services for direct to device. And I think what's really interesting is you're kind of looking at a market going, well, satellite wasn't that relevant a few years ago, but now it's got a two-pronged attack.
(33:57):
There's real interest in the broadband side, but also in that director device. And there was some research that kind of came out before the show that really said that the end users are actually interested in it as well. Customers who had previously maybe been used to having hot spots, whether it's in their home or living in the middle of the Welsh Mountains or somewhere like that, they were used to not getting connectivity. They now see an opportunity to get it and they really, really want that. Whether people will be able to be willing to pay extra for it remains to be seen, there's still question marks around it, but I think it's fair to say if you'll forgive the pun, the satellite really took off this year.
Guy Daniels, TelecomTV (34:41):
I'm not sure we can forget that pun, James. It was a terrible pun, but thanks very much. Absolutely. There was a lot more NTN and satellite interest at the show this year. Rob, what about you? What surprised you this year at MWC?
Rob McManus, REd Hat (34:56):
The lack of appetite to talk about, dare I say it, 6G. So we just didn't really go into any details on 6G, which was somewhat surprising from having been involved with Mobile World Congress for far too many years. If there's another G on the horizon, typically it dominates the conversation. Just wasn't the case this time. So from the conversations I personally had, I think there's still wide acknowledgement of the slow uptake of 5G services and also deployments. We know as a fact that a lot of networks are still on what we call non-standalone, which essentially is a 4G network with 5G radios plugged into it. And those work fine. They give increased speeds and operators are monetizing from them. There's still no killer application. There are 5G standalone networks, of course, in various parts of the world. And we also heard when we started to sort of ask and put feelers out about 5G advanced, limited interest in that, a lot of operators really just still struggling to monetize these huge investments they've made a long time, well, a number of years ago now, and it's just not happened.
(36:17):
So as James said a few moments ago, the focus in the sort of area of 5G was non-terrestrial networks. Again, we heard that in terms of maybe seen as a competitive differentiator for coverage perspective. And in the very few conversations, we did touch on 6G, I should say, very much needs to be a software upgrade we heard over and over again. I think these huge investment cycles that we saw on 5G aren't going to surface, in my opinion, for 6G. So very much it needs to be software. And once we get there, it also needs to be AI native rather than infusing AI into different things. So again, came as a bit of a surprise, but I think it was interesting to see that when we touched on the topic of 6G, it was a consistent, yeah, let's not go too much into the weeds there at the moment.
Guy Daniels, TelecomTV (37:17):
Yeah, absolutely. Right there, Rob. And it's really refreshing. We know the research is going on. We know the study work's going on. We know we're a year away from standards even being written for this, but it's like AI has stolen the news cycle. So the marketing side, we can focus on AI. We don't need to talk about 6G. Nice and refreshing. I'm going to come to Nico first and also James. So we just want to pick up some points on that. Nico.
Nico Marziliano, Wind River (37:40):
Yes, thank you. I tend to agree that 6G was a bit surprise, but at the same time, it's not a surprise. I remember in one of your telco debate we had a year ago, there was this conversation about CG, 7G, 5G, and people start saying, "Nico, don't talk about Gs. I think we still need to materialize what the 5G was promising." And until we solve it, don't talk about Gs anymore. So yes, it was a surprise, but it was definitely, I would say it was an industry feedback that it's not time to talk necessarily about Gs. Of course, we know that the next six will bring the further acceleration of cloudification and probably autonomy and AI, but the people prefer to talk about the real AI. And in fact, what I was mentioning before was I was surprised to see a lot of physical AI.
(38:47):
I mean, we, as a company, we presented a couple of cases, we bet on it, but we were not 100% sure that all the others were looking to this kind of next level of AI application. And we find surprising a lot of booth and partners and players presenting their own view of what the physical AI looks like. So definitely that was interesting. I have a comment on the non-terrestrial networks. I think that there is a very interesting angle that has been brought here. It's going to be maybe the next last frontiers where operator, yes, Vodafone, Timore, but many others are experimenting. I know that, for example, recently VMO2 in the UK, they connected with a satellite service. And so satellites will be clearly one of the topic where operator are investing just to provide connectivity or to provide AI distributor as some of the big players are try to move and transfer their data center and AI on the space, so basically shipping on the moon, but that will be definitely the next frontier I'm looking for.
(40:05):
So yes, that was another interesting angle. Thanks for reminding.
Guy Daniels, TelecomTV (40:10):
Thanks, Nico. I mean, I'm still concerned about the uplink power budget from terminals to NTNs, but let the researchers work that one out. James, I promise to come back to you.
James Pearce, TelecomTV (40:21):
Yeah, I think Rob made a really great point about 6G and the fact that you could walk around the halls and you'd barely see a sniff of it. And prepared, if you remember when it was the lead up to 5G, you had politicians across the world talking about it was getting really, really excited. It was in the mainstream news, not just the industry news, but now there's a certain reluctance around it. And I actually think it's down to AI. I think AI has caused the problem because essentially AI has fallen in the middle of a life cycle between two generations. It's fallen between 5G, which wasn't designed specifically for AI and 6G, which is going to have to be AI native, but 6G is still some ways away, as you touched upon, the standards aren't going to be with us for a few years. And there's actually a worry within the industry, within some of the operators that there's going to be a kind of too much of an acceleration to try to drive to 6G.
(41:17):
As Rob said, they haven't seen the benefits of their investment in 5G. A lot of them, they've not made the money back from it. So they don't want to go into that cycle of plowing loads of money into new 6G infrastructure. But at the same time, they've got this issue where if you remember in the buildup to the launch of 5G, there was a lot of push to try and be the first to get to 5G. And there was a lot of marketing spin around some of the launches, certainly from some of the US operators and so on as to where they'd take 4G, but it wouldn't be and rebrand it 4G plus and rebrand it as 5G or ... There was a lot going on, which was kind of slightly, I wouldn't say misleading, but made it look like 5G was here when it wasn't in.
(42:14):
The 5G standalone certainly was some years off. With the Olympics in the US in 2028, that's obviously a big wart mark. And there's always a desire across some spectrums to have a big announcement and have something big technologically. And LA have already announced that they're going to have flying cars and stuff like that in operation. That's one of the big plans for it. You need good connectivity to be able to do that. So you can see what's going to happen. You can see the potential that there'll be politicians pushing for 6G to launch and it won't be ready. The industry won't be ready. The operators certainly won't be ready. And I've heard this from several people high up at several operators when we're having conversations over TAPAS in Barcelona and they're saying, "We don't want to talk about 6G at the moment. We're not there." But at the same time, there'll be pressure elsewhere to get onto 6G sooner rather than later.
(43:14):
And I can see there being a real potential inflection point around the 2028 Olympics.
Guy Daniels, TelecomTV (43:20):
Yeah, James, you're absolutely right. There's also the spectrum angle in the US to consider there, which is a bit different from elsewhere, but I guarantee ... Oh, and it was misleading that those early marketing moves towards 5G, which weren't 5G, that really was misleading and it was proven. We'll see the same with 6G. I guarantee it. Someone can't resist it and we'll have operators coming out with these services before pre-standard services, I'm sure of it. Right. We've just got about time for one more take or question. I'd just like to ask you if anybody's got any feeling about the industry post MWC, how we're feeling more confident, less confident, and how is this going to translate into the next 12 months? What might we see? Steve, can I pick on you? I'm sure you've got some views on where this might be going, and then we'll come across to John as well.
Steve Daigle, HPE (44:13):
Sure. What I thought I saw, I'm a positive guy, so I'm always going to have the positive takeaways from what we saw. But around the adoption of AI, but not only that, the virtualizing of the RAN made a lot of good progress. I saw some things like T-Mobile's announcement about their Magenta AI call assistant, a partnership they have with a company called 11 Labs, where you can get on a call and say, "Hey, Magenta," and they'll do live translation of 50 languages while you're on the call and do contextual assistance. These kind of things are pretty, I mean, it's pretty exciting. I know you can also do it on your phone using an app, but not using an app and using the actual core network, using it in real time and have AI applied there, that it's kind of exciting to me things that we saw that are very, I think, good for the future.
(45:14):
Then there are a few other things like that partnership with the hyperscalers between the SPs, they've done that dance before and it hasn't quite worked out, but it seemed to me as though there's more of that going on around AI and the autonomous conversations. So it'll be interesting to see how this plays out within a year between some of the cool things we saw, lots of robots and some of the other things that we should probably look at a little bit more closely. The 6G conversation was a great one. But yeah, I mean, I'm looking forward to next year, but I'm ready for there to be more than 360 days between now and then to recover.
Guy Daniels, TelecomTV (45:57):
Yeah, nicely said, Steve. Absolutely. Yeah. And yeah, we're all positive. I'm sure all positive. John, are you positive?
John Abraham, Appledore Research (46:05):
I am, Guy. I'm quite optimistic. Remember, we are in unfamiliar territory in terms of how quickly a new technology has been adopted by telecom operators. It was generally considered to be quite conservative in terms of experimenting with new technologies. So yet we have seen a lot of value additions already. And as I mentioned earlier, one of the key distinctions between previous years and this year has been that there is a less focus on chasing the shiniest use case like we have seen in the past couple of years and far more emphasis on how do we actually get to value and what does it take to get there? Because there's a growing awareness there. It's one thing to demo an application, another thing to actually set it up within a complex legacy oriented environment. And we have already seen a lot of, let's say, niche or compartmentalized deployment say, such as within the ... And I'm talking about it mainly from a customer facing perspective because this has been the most popular landing spot for AI, at least starting from the initial years, given that a lot of this was inspired by what other enterprises were doing.
(47:17):
So we are seeing a lot of traction in terms of value addition, say, within the call center domain, improving the efficiency of the agent by having a couple of demos talked about how the screen changes based on what the AI understand the customer is talking about. So the agent doesn't have to go and click on multiple options. Automatically, the screen changes or the ability to monetize different types of advanced use cases, either because now the time to market is faster because a lot of these things are highly personalized and ready to go. Even the ability to do upsell, just because the AI provides a script. So we have seen a lot of this being deployed within compartmentalized use cases. And I think more and more, the reason for my optimism is that we are seeing far more emphasis on the data and the ontology layers that can bring all these disparate pieces together, because I believe that's the only way you can scale it, especially across a complex environment.
(48:25):
So I'm quite optimistic based on what I've seen in terms of operator, how close they are. There's still some ways to go, but at least I feel they are on the right path.
Guy Daniels, TelecomTV (48:35):
That's a really good way to end this show. John, thanks very much for optimism and thank you all. We have to leave it there for now. We are out of time, but it's been a great conversation, a lot of great takeaways, a lot of interesting insights from the event. Thank you very much for taking part in the program today. You can find all of our coverage from MWC 2026 on our website. Just go to the spotlight on 5G section, news, interviews, panels, and more. Well, that's all for now until next year and MWC 2027. From all of us here, thank you for watching and goodbye.
Please note that video transcripts are provided for reference only – content may vary from the published video or contain inaccuracies.
Review Show Panel Discussion
TelecomTV’s post-MWC discussion show returns. Guests from HPE, Wind River, Red Hat and Appledore share their thoughts and insights from MWC26. AI dominated conversations at the event, from the impact of AI on the network and the move towards AI native as the new operating system, to developments in AI-RAN and the implications of sovereign AI capabilities. Autonomous networking also emerged as a major theme, along with the role of automation to streamline ops and deliver measurable operating expenditure reductions. Partnership models and the emergence of viable non-terrestrial network (NTN) services also featured, alongside the monetisation of services and outcome-based pricing models.
Featuring:
- Steve Daigle, Global Head of Telco Systems Engineers, Hewlett Packard Enterprise
- Nico Marziliano, Regional VP, Telco Sales, Wind River
- Rob McManus, Principal Product Marketing Manager, Red Hat
- John Abraham, Principal Analyst, Appledore Research
- James Pearce, Editor, TelecomTV
Recorded March 2026
Email Newsletters
Sign up to receive TelecomTV's top news and videos, plus exclusive subscriber-only content direct to your inbox.
