The Unthinkable MWC 2026 Recap (with Dean Bubley and Andrew Collinson)

neXt Curve had the privilege to have Dean Bubley of Disruptive Analysis, and Andrew Collinson of Connective Insight join us to have an "Unthinkable" recap of Mobile World Congress 2026, which took place in Barcelona, Spain, last week. It's a sequel to their chat with Leonard Lee of neXt Curve in London a couple of days before Ericsson's pre-MWC analyst event that covered many of the key themes that shaped the MWC 2026 agenda.

In this episode, Leonard, Dean, and Andrew give their "unthinkable" takes on the top topics of Mobile World Congress 2026. 

➡️ Andrew's "Unthinkable" impression of MWC 2026.
➡️ Dean's "Unthinkable" impression of MWC 2026.
➡️ Making sense of "kinetic tokens" and physical AI.
➡️ Is the nature of the network changing with 6G?
➡️ Making sense of ISAC (Integrated Sensing and Communications).
➡️ Making a case for better 6G use cases.
➡️ Thoughts on OCUDU and GSMA's Open Telco AI initiative.
➡️ Who benefits from all the AI buzzwords?
➡️ Andrew's proposal for healthy thinking!
➡️ Key themes in mobile/telco in 2026...

Hit Leonard, Andrew, and Dean up on LinkedIn and take part in their industry and tech insights.

Check out Dean and his research at Disruptive Analysis at www.deanbubley.com.
Check out Andrew Collinson at Connective Insight at www.connectiveinsight.com.

Please subscribe to our podcast which will be featured on the neXt Curve YouTube Channel. Check out the audio version on BuzzSprout or find us on your favorite Podcast platform.

Also, subscribe to the neXt Curve research portal at www.next-curve.com and our Substack (https://substack.com/@nextcurve) for the tech and industry insights that matter.

NOTE: The transcript is AI-generated and will contain errors.

DISCLAIMER: This podcast is for informational purposes only.

Transcript

Andrew Collinson 0:06

Next

Dean Bubley 0:06

curve.

Leonard Lee 0:09

Welcome everyone to the next Curve, rethink podcast. And, as you all know, we are here to break down the latest tech and industry events and happenings in the world of mobile and telecom and many other things into the insights that matter. I'm Leonard Lee, executive Analyst at ncur, and I'm joined by two very special guests. Dean bubbly of disruptive analysis and Andrew Collinson. Of connective insights and these two gentlemen are going to help me unpack the stuff that mattered at MWC 2026 and we'll cover impressions of the, you know, what are the key themes of the event, highlights that impressed us and our prognostications for what we think will carry the industry in 2026. So, gentlemen, welcome. Nice meeting you with man in the metaverse. Is Metaverse still a thing, Dean?

Dean Bubley 1:14

was it ever a thing?

Leonard Lee 1:17

Good question. It's great to have both of you on. I'm really looking forward to the discussion. I'm sure that the audience is going to be absolutely thrilled, but of course, everyone wants to know Dean. How did things look, from the outside, like from the outside of the free bureau ground beer?

Dean Bubley 1:37

Yeah, I spent the week, not at NWC, I was more what I call the sort of, oh, the Dean WC or, or NWC Fringe. And so, I spent most of it either having coffee with people or watching. The, various, keynotes from the comfort of a cafe or a hotel room on one occasion is actually sitting on a park bench by the beach, watching the streaming on the streaming channel. So, thanks to the Mobile World Live guys for that. But no, it was interesting for me because. I've done this for a couple of years, and you sit outside it a bit. You watch you got a little bit more luxury time to, watch some of the, the things that go on stage. And you meet people when they're not dashing around with 15 minutes slots with their hair on fire. So I feel like I've got a decent read of what was going on.

Leonard Lee 2:24

Oh, okay. Cool. Hey, before we get started, please remember to share and, react, comment on this episode. Also subscribe here on YouTube and on Buzzsprout. listen to us on your favorite podcast platform, opinions and Statements by my guests or their own, and don't reflect mine or those of next curve. We're doing this. To provide a, open forum for discussion and debate on all things mobile and telecom, tech and industry. And this is for informational purposes only, but also we have Andrew Collinson of, connective insight. So, yeah, why don't we get things. Picked off impression since you were actually in the FEA with me. Right. And we bumped into each other several times, serendipitously. Right. Why don't we start off with you, what were your impressions of the event at a high level, key themes?

Andrew Collinson 3:23

Sure. Thanks Leonard. First of all, I'd spent the first day mainly in the keynotes and, meeting a couple of people there, which was interesting. And then the rest of the time I was in a couple of summits. I was in the security summit, and then I went into the Open Gateway summit, and then I bumped around looking at, meeting people and looking at the stands and so on. So I kind of got quite, and I was also outside the fair, occasionally even met Dean Bubbly. I had a fairly broad experience of the whole thing and what I felt most of all, or if you like, my superficial first, impression was it wasn't one of those years where there was one big topic. Now we can talk a lot about ai and obviously AI was on the menu. But it wasn't like a year ago or two years ago where AI was this sort of. it was really your headline, but nothing else. It was, it, AI had been kind of absorbed into what everybody thinks is normal, which I think is a reflection of modern, you know, executive life particularly, you know, if you're in a white collar profession, you're using it a lot and you're seeing it now. It's not mature as a, an enterprise grade. Solution for many places, but it just felt it had come off the hype burner and I didn't really think there was anything else to, to, to fill that. In the end. It came to the view that a lot of what we saw was about trust. because clearly we live in a world of geopolitical, very fast changing and quite disrupted. Political forces and that's destroying trust at a number of levels and that it sadly creates an opportunity to be some, if you can be a party that is helping, deliver trusted interactions, then that's really one of the great opportunities. And I think it's quite interesting'cause you look at operators and you think, there's a lot of bemoaning operators not being tremendously innovative, but they don't really invest a huge amount of their r and d. You are in r and d, it's 1% or so, so they can't really be expected to be producing new technologies. And what are they're carriers, they're operators. They operate things. And I think that gives them the opportunity in some places to operate things in a way that delivers trust to their clients. So I illustrated a few examples of it. One of them is sovereignty. And I know Dean and I doing some work on that for unthinkable. Another is the whole kind of API stuff, the network, API, stuff about identity and CI swap and so on. It's another form of trust. there was some quite interesting demos of things that were call screening and so I've also been doing a lot of working in AI voice, so there's, there was a lot of interesting stuff about AI voice, which is really exploding. I think the other commentary I would give, which came out through some of my discussions around the stuff I wrote about it was, the other way you could look at it is you could say Six G's not ready. AI is not quite hitting the road with all the benefits. And probably satellite's the only other thing that's around, which is real big and shiny. Yeah. So satellite got a bit of a kind of airing on the stage. So I think there was a bit of contextual business that there's not one big, new technological way rushing through the industry that everybody's trying to get on top of, and the geopolitical stuff behind it was really creating that kind of, not exactly a vacuum, but this opportunity around trust in different donors. So that's what I felt about it. Anyway, that was, you asked, and that's what I thought.

Leonard Lee 6:49

Yeah. and you wrote that piece, last week. Was it last week?

Andrew Collinson 6:53

The beginning of this week?

Leonard Lee 6:54

value needs

Andrew Collinson 6:55

service?

Leonard Lee 6:56

Really well done. So folks should check it out. it gives you a really great perspective on, trust and the, actually the, you wanna call it the unfortunate opportunity that operators have, with, maybe positioning themselves. As, trust providers or enablers.

Dean Bubley 7:15

I would say in terms of things I noticed and the discussions I had and the things I, first thing is there wasn't much mobile. At Mobile World Congress, I think certainly listening to the keynotes, I heard much more about Fiber subsea satellite than I did about 5G.

Leonard Lee 7:34

Mm-hmm.

Dean Bubley 7:34

the at and TCEO was talking about saying, well, basically he sees telecoms as a dense grid of fiber with different accesses hanging off the end of it. Mm-hmm. And that was an interesting framing. The other thing is there was, I agree with Andrew, there was a fair bit about, voice communications and I've been talking about a renaissance of voice for some time. and I used to run workshops 15 years ago called The Future of Voice. And it seems like the future is about now. and then there was definitely on the other side of trust and geopolitics, there's quite a lot about, there's, a theme I, I. Picked up last year. Just the sign signs of that really kicked off this year was the convergence between telecoms and defense. and I think that's something that's gonna, we're gonna see more of for, for obvious reasons. and then there's this sort of, physical ai. And, I know you noticed a post I wrote on that the other day.

Leonard Lee 8:26

Yeah. That was, what, entitled Tyranny of Consensus. Right. So you tackled two buzzwords of the conference. six G and, Physical ai, right?

Dean Bubley 8:37

Yeah. There's this, the physical AI is this term, it's not bad. It's basically, it's like IO OT meets ai. It's sort of smart things, which aren't smart phones. So it's robots, it's autonomous vehicles. It's ar, vr, it's, industrial automation with ai. pixie dust sprinkled over it. It's things which exist in the real world and use, sensing an AI to, decide what to do. Yeah. And whether that's an autonomous vehicle, whether it's a drone, whether it's a humanoid robot, you can see that there's an appeal for lots of different stakeholder groups, whether it's the. The radio people, the device vendors, the cloud players, the AI model providers, even politicians, core network providers. Everyone has an angle on it and they quite like it because it's all the stuff, which is, that never happened with 5G. It's bundling it all together and giving it like a sci-fi. Sheen with a picture of a drone and a robot, to compete with the people selling rockets and nuclear powered data centers and all the other sci-fi stuff that's going on and say, Hey, yeah, and see there's things like this idea of kinetic tokens. People were talk, talking about, I think T-Mobile us, which as far as I can see is slicing meats. Ultra reliable, low latency, repackaged and rebranded in a way that makes it look vaguely billable and perhaps regulators won't stand on it.

Andrew Collinson 10:11

It's all got a bit of a whiff of sulfur about it to me. It, it kind of like you look at it and you listen to it and you think. What, yeah, what you actually talking about? I say robots, I think. Okay. I I understand your drones. Yeah. Oh, fair enough. But physical ai, yeah, it's, yeah,

Leonard Lee 10:26

I mean, yeah.

Andrew Collinson 10:27

There's be a German word for this, isn't there? The words that are smashed together into sort of meaninglessness in some sense. But, sorry, I shouldn't be so skeptical, but

Leonard Lee 10:36

No, no. I don't think you're being skeptical. It's a branding twist. or it's marketing. twist. A lot of the things that they're packing into physical AI are things that have already existed for a really long time. Hmm. It's nothing new. whether it's the, the Association of Digital Twins or iot or, sensor technology. These are all things that we're very familiar with. It's just, I think it boils down. That's

Andrew Collinson 11:02

But I thought your point, Dean, about the kind of six G and physical AI was a really well made point. You know,

Dean Bubley 11:07

it,

Andrew Collinson 11:08

you've, it can produce an accidental outcome. That's actually not what tremendously desirable.

Dean Bubley 11:15

But it's also, it's, yeah, as Lauren says, all this stuff has been around, so it's not like drones and robots are waiting around for six G to rescue them from their lack of connectivity. and this is the typical thing the mobile industry does time and time again, whereas it positions itself as the critical single enabler of all this cool new stuff of, you know, we will be, contributing to. Trillions of dollars of GDP because no robot or drone would be able to work without us. And yeah, I sometimes stereotype it is a bit like the person who turns up at a party at three o'clock in the morning with a cheap bottle of wine going, Hey, I'm here now. The party can get started and everyone looks at them and go, who the hell are you? and I have a feeling that six G is trying to gate crash existing parties. No, absolutely. Six G will add. To, it could be vehicles, it could be drones, but that's like incremental and it will have to integrate and fit within those existing systems. Right? It's not, there's gonna be stuff which is six G Native and oh, we've, you know, thankfully you've arrived. We've been waiting till 2032. Yeah.

Leonard Lee 12:19

But

Dean Bubley 12:20

that's not happening.

Leonard Lee 12:21

Yeah. And I think that is actually a better perspective for the industry to whole than what they've traditionally done with like getting ahead of themselves with these use cases. the thing is. This transition, and I know that you hate the term cyber physical. I have no idea why Dean. This guy hates all kinds of words. It's like really weird. But here's the thing. The infrastructure, the network is changing whether we like it or not. The way that they're defining six G. you know, the integrated sensing and, communications, but also the high precision positioning. There are all these new functions that are coming out that are being, repurposed. Actually. These are functions that the network and the infrastructure already that it uses in, delivering. The air interface. It's just now you're looking at new features that are capitalizing. On these capabilities, right? These, physical capabilities that now, end users can tap into, right? Developers can eventually tap into that enable, it, it's actually not mobile computing. It's, positioning the infrastructure as, kind of like the sensor, right? So E ISAC for instance, it basically. Converts the network into a radar. But the capabilities that the, these new features introduce, I don't think should be so trivialized because let's say for instance, smart cities holistically, right? It can introduce a new capability for, let's say, creating spatial mappings at a higher sample rate. Than what we have today. Right. There's an opportunity to think, beyond just connectivity and, consider and contemplate applications that haven't been possible before.

Dean Bubley 14:22

I agree. And I think that ISAC is a particularly interesting one. But I'm not sure that actually the eye. Makes as much sense. And for example, do we even taking on board what you just said for smart cities, does every country on the planet need three competing radar networks?

Leonard Lee 14:41

Yeah, I saw you write about that.

Dean Bubley 14:44

I'm actually meeting with, someone from a re regulator tomorrow to specifically talk about what the regulatory surface around sensing is in terms of spectrum, in terms of privacy. Even just. The fact, yeah, the competition. Yeah. You know you are right. It's interesting, but is it something that belongs in a public mobile network or all of the public mobile networks, or is it something that's a bit more like some of the specialist networks we see for defense or rail or utilities and so, you know. Yes, maybe you should have a shared city radar network, firstly, because it means you can put the infrastructure in the places that need it. Secondly, that you've got governance that because there's gonna be national security implications, there's gonna be a completely different regulatory surface. There might be different physical sighting for it. So, absolutely. And you might wanna have sensor fusion. So you combine the wireless radar function, frankly, with cameras.

Leonard Lee 15:42

Yeah,

Dean Bubley 15:43

so ideally you have a network of cameras and a network of, radar, capabilities and probably pollution sensors and motion sensors and a bunch of other stuff. And you have some clever sensor fusion layer which blends it all together. And is that something that belongs in six G or is that the sensing function and maybe you plug in wifi sensing for the indoors and you have. The set, the fiber sensing to, to look at underground and a bunch of other stuff. Right. so yeah, completely agree. That sensing is interesting. I'm just not sure it belongs in a, in a public six G network.

Leonard Lee 16:17

I think it, it's incumbent on the industry to. offer as an alternative, as you cited in your tyranny of consensus piece, right? It's, it's an option. I think there is some uniqueness to it. It's just all a matter of the industry. Just thinking through the details, like what you're suggesting. They're articulating how it's a viable alternative for developers, for solution architects in, either augmenting existing applications or, or innovating in new frontiers. Right. and I guess that's the point that I am making because, wifi is doing the same thing, right? There's all kinds of new sensing features, but yeah.

Dean Bubley 17:00

You get sensing in subsea cables even. you're sensing everywhere throughout the entire network.

Andrew Collinson 17:05

The challenge is for sensing as well in the sense that, you know it, it potentially has quite a lot of privacy and security. Absolutely. Implications and inter inference is perhaps the wrong one, but certainly privacy and security and it feels to me like. I'll tell you what it feels like. I'm not sure that it's, but you know those diagrams you always get when you have a new network evolution, which has a triangle and a lot of circles and things like that on it. Yeah. Like you used to get clouds in the old days. It's that kind of diagram and it's one of those things that always pops up on the new network diagram, and I think it's a bit like, what should we put in here? Oh, let's chuck sensing it, it feels a bit. Like light on consideration. I'm not saying it's a bad idea, I'm just saying it doesn't seem very thought through from what I've seen. But maybe I'm wrong, Dean. You are deeper in it than I'm, but that's what I feel about it.

Dean Bubley 17:57

There's a lot of interesting stuff. I met up with the, the Etsy, standards group that's looking at ISAC and sensing, and there's particularly a lot of interest around. drone detection is one that everyone talks about. And then also, vehicle to vehicle, vehicle to X. and I think both of those are very interesting, but again, drone detection, you don't, well firstly, you probably prioritize certain areas, so airports, frankly, data centers, other places which are vulnerable. If you look at what's happened over the last week in the Gulf. The places you have to worry about drones are the sort of strategic national infrastructure locations, which includes compute facilities and energy facilities and things which are not traditionally the areas which are telco strongholds and certainly if I look for the European lens, they're the place that if you had ISAC plus the AI to interpret it, you suddenly put yourself in the category of sort of high risk AI use cases and you are. Regulated to the max. Whereas if you're doing it to find someone's lost dog or is there a parking space or something like that? Yeah, it is a much lower bar. But I think we need to think what's the, what's it for? And, do you end up then exposing the regulators to yet more regulation? And I think that no one's even started looking at what, ISAC regulation looks like.

Andrew Collinson 19:15

Just take,

Dean Bubley 19:16

you could argue that drone

Andrew Collinson 19:17

protection. Version. Let's say you got a network running right now and you can detect drones. What are you gonna do with that information? How's it gonna be useful when you're normally detecting drones, it is attached to a radar and a rocket at the other end of it, or some sort of defense mechanism or an alert to the, area in which, you know, if it's an airport or something like that. I cannot see a public network in that role at the moment, or at least without significant consideration of The pros and cons and benefits and whys and where for what? If there's a gap in coverage and it doesn't see it, are you liable? What if, what if it's too slow? What if you miss? Yeah. really you're gonna go up track. and where else are you gonna use drone detection? It could be useful to say if someone's overlying.

Leonard Lee 20:02

Yeah.

Andrew Collinson 20:02

I think that you could, maybe you could, that might be more realistic. I don't know.

Leonard Lee 20:06

Yeah. And this is where I think, the use cases that are at the center of explaining a technology sometimes can do a disservice for the technology. I, I think it actually has some pretty good potential in, pa, active, spatial twinning. So if there's a particular space and you want to create a quote unquote spatial twin or digital twin of a space, it helps you spatially contextualize the area, and then you can use features like. a precision, positioning, to track things that are actually connected to the network versus trying to, detect a drone flying around that isn't even registered with the network. there are these other applications that consider, it's one element that extends the possibility of, let's say, a broader. Or more expensive application that may be more valuable than just what Esec does. Mm-hmm. Functionally within the context of that solution.

Dean Bubley 21:14

Also gotta coexist with the camera sensing and lidar and uwb at short range and all the rest of it. Yes. yeah. So it is part of that. Yeah. I, a few years ago I went to a sensor exhibition or a trade show and I was just blown away by. How many different sorts of sensor there are. Oh yeah. From depth perception to motion, to chemical sensors to density, to, you name it. Every possible physical characteristic, there's five solutions for it. And I think that this the, as well as the tyranny of consensus, the other phrase I used was assumption of uniqueness. And the risk is that the mobile industry thinks that we have the only way of doing. Sensing when in fact they fit into an existing really complicated sensing landscape. And this could well be an, a valuable add-on, but there's also limits to it. And, you, there'll be other things where you use UWB or you use, frankly, a$20 camera and some, cheap image recognition.

Leonard Lee 22:17

Yeah.

Andrew Collinson 22:18

I think the points about like the cases that are sensible are really good. But the problem is a lot of the cases that are being put out there are not really good. and like everybody in the world has got limited attention. So if you come to me with a sensing case and I've got five minutes to think about it, and you come to me with a silly one, I'm just gonna think you're talking rubbish. And I don't mean me, I'm talking, I'm the regulator. I'm your customer, I'm your investor. If you're talking. Nonsense use cases, you are expecting people to apply a level of imagination and knowledge that they probably don't have. And it's, I sometimes think the industry lacks or has never heard the story about the boy with no, the emperor's No new clothes. Yeah. Where they just haven't spent enough time arguing about it, which is something, as spent a lot of time doing going, that's Robert, why doesn't, why does that make sense? And it's. I suppose in a way it's quite good for us because it helps us make a living, but it can be quite frustrating sometimes that people just haven't, put it through the washer a bit more, you know?

Leonard Lee 23:16

this is one of the reasons why, in my observation, Industrial 5G, if you wanna call it that? Mm-hmm. Private networks and stuff like that has been such a challenging thing. there's. Idea of uniqueness as you were alluding to, right? It's the only thing that can deliver U-R-L-L-C, which it just happened to be, really expensive and challenging to realize in an industrial setting, much less out in in a macro network, right? assuming that. Your technology is unique, which it's not. If you go into the industrial, iot, there are all kinds of protocols out there, different, lot of different technologies that are being used for wireless. but also the singular focus on connectivity. Has also undermined, the positioning of 5G in industrial and enterprise, scenarios, right. And context. So I think that it's really undermined the. progress that people had anticipated with, 5G private networks, especially for enterprise and, industrial. So, these are all great points, but, the thing is the deficit at the moment is that exercise that you're alluding to, Andrew, is that debate and then, somehow, advancing the thinking, beyond just generic and, shooting from a hip type of concept. Hmm. That undersells the technology, right? And its potential to, enable, innovation and improve.

Andrew Collinson 24:51

I'll tell you one area I've seen, an interesting thing and you know, if, forgive me Dean, I'm gonna say something nice about the GS MA. they have a program called Fusion, which is kind of the sister program to the Open Gateway program. So they realized that what they needed to do was get closer to enterprise demand. So they've got a program called Fusion where they're going out and talking to enterprises and finding out what they actually need and want, and turning those into statements of requirement. And I think that's good. I think that's a step forward. I've looked at some of the cases. I'm just about to do a little more research in APIs, so I will be in that. I can see Dean smiling, so I imagine he's got something to,

Dean Bubley 25:27

well, I, I, something else I posted you. Yesterday, which is, if you're gonna ask enterprise some enterprises what they want, it's, we want our own spectrum to build our own networks.

Andrew Collinson 25:36

Yeah. Yeah.

Dean Bubley 25:37

We don't wanna buy it from you. and that's perhaps one of the answers that, that, that will be quite difficult because, I'm a big believer that there should be an, essentially an enterprise spectrum users association who ha speaks with a similar weight to GSMA. To regulators and ITU and others, but at the moment it's quite difficult to get everyone from Miami Airport to Jaguar Land Rover to the BBC to the US military in a room saying, actually, we quite like to have a couple hundred megahertz for our private networks. rather, yeah. Which are with this specification and these power levels and so on. Because frankly, we think that we can do it better. And I think that's, a difficult one. No, I, I take your point. And to be fair, it is good to see that GSA is paying attention to enterprise, whether it's on the API side of things or whether it's on connectivity. I'm a slicing skeptic, but, certainly it makes sense and they brought together some communities of interest on, from memory, automotive and a couple of others.

Andrew Collinson 26:38

Manufacturing.

Dean Bubley 26:39

No, they exist,

Andrew Collinson 26:39

but,

Dean Bubley 26:40

but 5G A-C-I-O-I mean, is that just because people got bored with five G-A-C-I-A, which is the, in industrial automation group and 5G aa, which is the automotive. Right. So they seem to repeat the two verticals where this is already existing trade association. So maybe they're just trying to po poach members. I don't know.

Leonard Lee 26:57

Yeah, easier, I dunno. I dunno. Yeah, I think it might be things just didn't ground themselves in, the previous effort. So the looking Yeah.

Dean Bubley 27:09

Next evolution, another five G-A-C-I-A. People were slightly irritated that they'd gone all through all this work of trying to design what they wanted from 5G. which included the famous one, millisecond latency. They can't actually buy it'cause no one's selling it.

Leonard Lee 27:24

Yeah. And then, uh, CV to X, that, that was. Challenged as well, you know?

Dean Bubley 27:29

The industry in the past have got history of getting together and saying, we want this.

Leonard Lee 27:34

Right, right.

Dean Bubley 27:34

And then not being able to buy it.

Leonard Lee 27:37

Although CV two X has some interesting new defense applications, that could be very valuable. But see, that's one of the things that yeah. Yeah. You, you made a great observation last year or even previously. it was actually at, MWC Las Vegas about three years ago, three, four years ago, where the defense industry. It took a interest in, 5G, so that, that is now playing out even more prominently, this year.

Dean Bubley 28:08

kudu and things like that.

Leonard Lee 28:10

Kudu and the what was the other thing? The Open Telco AI initiative by GSMA. Any thoughts on that?

Dean Bubley 28:17

No, I haven't looked at that one yet.

Leonard Lee 28:20

Oh, really? Okay.

Andrew Collinson 28:22

What was the open. Because there's been several of those sort of things. Long

Leonard Lee 28:28

look

Andrew Collinson 28:31

launched about like a couple of years ago with a group of seven operators.'cause it was, they were talking about

Dean Bubley 28:38

TSKT and a few, I think it's like, there's people talking about like a telco language model as an i I think the other thing is the word ontology and there's another phrase which is like ontology, which I can't remember which crops up quite a lot.

Leonard Lee 28:51

graph,

Dean Bubley 28:54

it's, it's sort of all, basically all the stuff you feed into rag so that gen AI actually understands what reever Hmm, sorry,

Leonard Lee 29:01

retrievers.

Dean Bubley 29:03

Maybe,

Leonard Lee 29:04

vector, vector

Dean Bubley 29:05

No, no, no. This, it's like, it's essentially, it's a library of telecom's terminology of all you, you basically, you load in all the standards, you load in all the, descriptions of how the bits of networks and OSS and BS fit together, so that when you ask gen AI a question, it doesn't go off on complete tangent. You anchor it in, telecom language. Yeah, and so if you're talking about, I don't know whether it's about power or if it's about customers or whatever it looks through a telecom lens.'cause it's got a telecom, essentially dictionary, ah, built into the, the gen ai. So it doesn't go off in too many hallucinogenic tangents.

Leonard Lee 29:41

Yeah. the only thing I can think of is Laura, which is a way of extending a foundation model to be more domain specific. It's a methodology and post training. But yeah, no, I think it's really interesting, because there's so many efforts to try to make these models domain specific. But then, the field is becoming very diverse. And the way that you're applying, AI is, is actually, vendors. Specific, it is a competitive advantage, right? For instance, between Nutcracker and Amdocs, the way that they might train L-M-L-L-M to be optimized for their particular. Application portfolio will be different. And there can be competitive advantages. having an LLM that's trained on your system or your application ontology, right? Which could be very different from a competitor. So I think people go overboard with open and shared. Sometimes don't appreciate, the benefits of competition and differentiation, you know?

Andrew Collinson 30:50

Mm-hmm. Yes.

Leonard Lee 30:51

And then some of these open efforts become futile. They become anti-competitive in a way. I don't know.

Dean Bubley 30:57

I think that coming back to sort of things like acu, I think where that makes sense is there are, particularly in defense, I can see why there is a desire to have. In the US a platform for sort of sovereign radio applications. Maybe you wanna plug in a special waveform for something, that makes sense to me to have the sort of subset of O ran or something adjacent to it. That's relatively constrained, that has, obviously a sort of more national or friendly, nations spin to it. Yeah. So that, that I can see it is the same with the uc on the spectrum management side and stuff that is being done by the National Spectrum Consortium in the us and. Others in, in, in other parts of the world. the, that does make sense to me is to have a certain level of openness. And I think if I look at what's going on with AI native in six G, a lot of the Standardization work is on things like model lifecycle management and data management and sort of, yeah, signaling between different parts of it. It's not necessarily standardizing the model, it's making sure that some of the hooks and the scaffolding and the frameworks are there. And I, I can see the logic for that.

Leonard Lee 32:13

Yeah. But even that, I don't know why you need to standardize any of that stuff. there may be best practices. like agile, right? You just adopt whatever makes sense. but standardizing it, to me it doesn't make a lot of sense.

Andrew Collinson 32:28

there are areas. I mean, I'm not sure that I've fully understood this, so I'm a danger of leaping into deep hole with both feet here. But

Leonard Lee 32:34

no, just

Andrew Collinson 32:35

One of the things we talked about with, Dean and I ran an unthinkable lab about MCP and Agen AI and APIs.

Leonard Lee 32:42

Yeah.

Andrew Collinson 32:43

And one of the things, for example, Philippe, who's Philip, who's the v VP of Tech at Orange, he was very, interested in having open standards on things like, MCP, interactions with Agen so that the, if you run an agen query, you get the same thing from different places. So the format of the query that you runs should be within a certain. Within certain bounds and the output should be in certain bounds. So I think that And why would you want that for operator's point of view? Because if you want a standard which is interoperable, then you need that level of compatibility, I think.

Dean Bubley 33:24

And I also think that there's an nagle there, which is, if it's something is standardized, then you can get the test vendors to support it. And so I think that a lot of the telecoms industry. Won't. Deploy stuff unless there is a box from Yeah. Keysight or Viavi or whoever it is, right? That flashes a green light and prints out rapport metaphorically. And I think that in order to have, a testing suite, you need to know what to test to. But if you want to, and yeah, I know in the UK for O Ran we, there's a security testing lab that's government funded to make sure that essentially it's trustworthy. Yeah. So you need to be able to have things to test to, and then attest the ecosystem to tick the box and flash the green light.

Leonard Lee 34:10

Yeah. But here's the big problem. agents are. Very difficult to test testing is no one's cracked a code on that. And also what you described, Andrew, is addressed by something that already exists, which is the API. So this is the problem with what we're seeing with AgTech frameworks and AgTech AI and MCP is at the end of the day, what businesses want, what probably consumers want is something more attuned to, programmatically configured, agent, which basically takes us back to RPA and workflow. Management. That is ironically where we're going, but what I think many industries are trying to do at the moment because there's so much pressure to adopt. AI is shoehorning all these buzz words, right? to make AI fit. But then when you look at the requirement sets and how industries are reacting to what is being presented in front of them. they're realizing, oh, we do have to step this back to. The old, reliable stuff. And, I'm not trying to sound like a Luddite, it's an, it's a true observation.'cause when you look at the ISVs, the likes of ServiceNow, Salesforce, you name them, they're embedding logic and scripting into their agentic frameworks. And guess what? At the end of the day, this just all looks like RPA and the old school stuff because you're taking away so much agency from the agent. It's no longer fuzzy. It is being programmed

Dean Bubley 35:52

on the other hand

Leonard Lee 35:53

or the best, for a de deterministic outcome or,

Dean Bubley 35:56

but otherwise you end up with a telecom equivalent of OpenCL and you, you basically sort of give it a task or an intent and it says, make my video run better. And like, well, I can buy some QOS, I can down rate the resolution. I can, send letters to all the electors to get them to vote in someone who doesn't like their neutrality. I can, knock everyone else off the network by hacking them. So it is like you need to have some guardrails'cause otherwise you can

Leonard Lee 36:22

be a very dangerous person.

Dean Bubley 36:27

Yeah. He's a very

Andrew Collinson 36:28

dangerous person. And you, you, you should know this by now.

Leonard Lee 36:32

Yeah.

Dean Bubley 36:32

But quite the imagination. But that, but that's what happens if you have like open-ended Oh, an intent. I'm like, well, yeah, just how far do you want to go? What, what, what's your limits on that?

Leonard Lee 36:43

Well, yeah. And then, there's, there it, I think it is becoming increasingly known that these models have alignment issues, right? these are like all these fundamental. Challenges you talked about limitations, Dean, that impact the efficacy and viability of anything that, any of these permutations of a gen, you know, gen AI above it, including the physical stuff. Even though, you know, like what we said, physical's been around, I think if we talk about physical AI in a better way, maybe unpack it to reveal all the stuff that's familiar, then we can figure out how all the new stuff. Can actually, make things meaningful, for the industry.

Dean Bubley 37:26

I think that's right. Oh, I, well, the question is, you gotta define is what is the industry and is the industry at that point? Seriously, is it the robotics industry or is it the mobile industry, or is it the networks industry or the cloud industry or the, application side of this or the enterprises and everyone, and this is where I came from, Tyra Consensus, is that everyone's using this nice, fuzzy term. They all mean different things by it, but it's nice. It's a nice buzzword to hang all there. Diverse motivations on

Leonard Lee 37:55

Yeah, but I think if you unpack it, then you can hold people accountable to their marketing intent.

Dean Bubley 38:04

Hold people accountable to their marketing. That would be in advance. We probably put all of us out with Job as well.

Leonard Lee 38:10

Oh, no. That, that's what keeps you, keeps you busy, Dean, come on. And you, Andrew. It's all good stuff. Are you okay there, Andrew, or are you just thinking?

Andrew Collinson 38:23

No, I was thinking, I'm just, I got, was a bit distracted there because I was thinking that earlier today. I was having a chat with, same guy, Philippe and saying. I was thinking we ought to have a campaign for healthy thinking. Like you've got a campaign for healthy eating. Just because there are thousands of ais available doesn't mean you should just use them in the same ways that you can buy millions of hamburgers and different things. And so, sorry, I was just distracted'cause we went off on an MCP rabbit hole and I was thinking Yeah, pay for healthy thinking, you know, down.

Leonard Lee 38:59

So why don't we do this? Let's, let's start to wrap this up. Any thoughts on what are some of the, what are those key themes that are gonna carry the industry through 2026? that we'll define what we'll be talking about at MWC 2027. And of course, Dean being like. Not there.

Dean Bubley 39:25

Satellite voice, fiber, sovereignty.

Andrew Collinson 39:32

Yeah, I think voice is a big deal. It is been deeply unsexy for a number of years and, uh, I've been on, in a combined and separate path looking a lot of this.

Leonard Lee 39:43

Yeah.

Andrew Collinson 39:43

I've just almost finished a report on it. it's just taken me ages'cause it's just changing so fast. There's so much going on in it It's classic telco, it's been on sexy for 15, 20 years. And of course a lot of people have neglected it, but the clever folk in the other companies, the innovative folk haven't. So you've got a lot of really interesting companies doing amazing stuff in AI voice, and, it does two things. First of all, you've got these services, new services coming to boot, so you've seen the first ones, which call AI and every call idea where you just call in your AI to help you translate or take notes or something. That's fantastic. A lot of the stuff I've been looking at is more on the enterprise side, where you integrate it into your enterprise operation. What's really interesting, I think about that is not just the automation prospects of it, because that pretty, who wants to put people outta jobs? Some people do, but not me. But there are some quite good benefits of it, in the sense of where it's being used to help companies do things better.

Leonard Lee 40:41

Yeah.

Andrew Collinson 40:41

But where I think it's really interesting is it turns every call into a proper source of information. So if you imagine you get all of your calls in one place and you can properly interrogate them and interrogate'em in the context of other corporate information, instead of a few lines from your CRM that you've typed in because you know you are in a hurry and you said I spoke to Leonard. It was great. We talked about MWC full stock. Goodbye. You've got the whole conversation in there. If you can run your AI on that and the other conversations you've got and the other things like that, you can do so much more. You can identify real time problems. You can identify trends in operations that you weren't available, that you weren't aware of, even. and you can start to understand different needs of customers because you can see across essentially a vastly different landscape of information. So I'm a big, I'm a big fan of the whole AI voice thing I think is, found in the sense I think is really interesting. Where it goes and what it does, we shall see. But that's in its own right. Interesting. So I picking up on Dean's theme. Yeah, satellite's gonna be interesting. It was fascinating to see starlink and SpaceX talking about launching, I can't remember. They were talking about launching 50 satellites with each spaceship when they, the Starship, when they get to launch it, and they're gonna launch 1200 of these things next year in a year. And I think the thing that's interesting about that from my perspective is not the kind of like the shiny tech aspect. It was just, I sat listening to these people and I thought, wow, I'm not sure that I understand what the motivation of people at Elon Musk is doing what he's doing, but you look at the scale of the technological ambition. And you, you feel this is a bit lacking in, on the other side of the room. Now we talked about a lack of r and d and we, they're very different types of company, but you look at it and you think, my god, the pace of change and the pace of investment in some of this stuff is immense and we're learning very quickly that you can't. It's hard to forecast very hard, far ahead in a lot of this stuff because each thing interacts with the other thing and changes it. So we've got geopolitical change going on. We've got a technological change going on, probably some, how can I put it? Sociological, socioeconomic change going on. Yeah. Because of ai, not just the technical stuff, and it's all. Oh, a bloody great melting pot. And you know, it, it's very hard to call it even a year out now. I think maybe, you know, I'm sure we could write down a list of keywords.

Leonard Lee 43:10

Yeah.

Andrew Collinson 43:11

But exactly what we would, what's gonna be said is hard to, to judge I think at this point. But the topics are probably right.

Leonard Lee 43:16

for me, I think trust, so. Off of the thing that you wrote about that you thought was, important? definitely. I think given the, and we talked about this at the bar, at World's End, right. and Dean you mentioned as well, there's forces that are driving the sovereignty. Discussion in, places that we haven't seen before. And I think work counter to what has benefited, the mobile industry, for last few decades. And so I think the thing that the industry is going to have to reckon with. This year, given a lot of the geopolitical volatility and, conflict is how does that impact the future of the industry? And then I think also, when we look at forward to six G, how is it going to influence what six G is going to be about and maybe even 5G advanced, right? because, those who, follow the evolution of the standards and how they're, are cognizant of how the standards are arrived at, it's a tyranny of consensus, right? Dean? Some will claim as a tyranny of consensus uh, and I don't think a lot of folks recognize that, especially policy makers, who might have a naive, notion about how these standards are arrived. At, globally. So that's probably gonna be one of the biggest issues. And then for the operators, I think they're gonna be grappling with, modernization and hopefully they'll re be revisiting, the, readiness of 5G. I've been talking about this for quite some time now. We're at a different place with 5G. So is it a time to look beyond the initial disappointment. Then, maybe reconsider what the prospects are for 5G given we have more devices that support, 5G. And also, we're, what is that release? 19? So there's a lot of good features that are being untapped. What is that value proposition now at this phase? And, Hopefully the industry for its benefit will reflect on that and maybe, focus on things that matter a bit more than, humanoid robots.

Dean Bubley 45:50

Yeah. Yeah. Things like indoor connectivity. Actually one last thing that I picked up there was a presentation from a futurist I really called Aza, who's a general futurist who speaks really well. He's got a great podcast, but he was saying that he's currently on, I think he said he's using a hundred million tokens a day for his AI agent workforce. And the issue is, and I had a bit of a comment, a comment thread with him on my LinkedIn is how much of that goes over the access network, whether it's fixed to home or office or over the mobile network? And it's gonna be small. And this is a big problem with AI and mobile industry, telecom industry is it's all inside the cloud. And I noticed, at and t and making a big investment announcement and a lot of that is around long haul fiber for data center interconnect.'cause the amount of the AI stuff, which, whether it's for robots or just LLMs that goes over any sort of access network to us, is a tiny, tiny fraction of the total that's going on in internally. It's a bit like the Hitchhikers Guide to the Galaxy, where there's this sort of, the earth is actually this computer that's run for millions of years. And it comes up with the answer to the A, the great question of life, the universe and everything, which is of course 42. In other words, there were two bytes of information on the access network after a million years of compute. And I have a feeling we, we might be heading that direction.

Leonard Lee 47:07

Interesting. That's nice insight to, uh, cap off the episode. Gentlemen, thank you so much. It's always fun hanging out with you guys and, Dean, I look forward to hanging out with you next week. Hopefully, we'll, you help me find some good places to. Enjoy here in San Diego. And, by the way, if you don't know, I live in San Diego, and Andrew, hopefully one of these days you'll be, visiting San Diego and you can help me find good places to eat here as well.

Andrew Collinson 47:38

I look forward to that. And of course you remember we, our agreement, we have to practice so that we're ready. Ready for the next break? Yes.

Leonard Lee 47:45

Yes. And Dean, you'll be singing your AI generated lyrics on stage. No excuses. hey everyone, thanks for tuning in. I hope you enjoyed this, conversation. I know it's been, uh. Great learning experience, for me having these debates with, Dean and Andrew, if you don't already follow'em, you should. And by the way, Dean, congratulations on 30,000 followers on LinkedIn. That's a great accomplishment. Kudos. Thank

Dean Bubley 48:16

you. I've got a long way to catch up with you though,

Leonard Lee 48:17

Yeah. you'll eventually get there. No worries, I have no doubt. And, so follow him and, and disruptive, analysis at www dot disruptive Dean. Right.

Dean Bubley 48:31

I, Dean Bubbly is my website, to be honest. Just LinkedIn's easier link. LinkedIn's easier to be honest.

Leonard Lee 48:37

And then, help him get to 180,000 followers. And of course, the wonderful Andrew Collinson of Connective Insight. Follow him at www dot connective insight. That's insight. Dot com, right?

Andrew Collinson 48:51

Just LinkedIn. I'd say the same. And also, I need your help more than Dean does.

Leonard Lee 48:56

Okay. Yeah, so

Andrew Collinson 49:01

I haven't been on this trail for as long. And, so I'm very grateful to you both for allowing me into your hallowed world of conversations. And no, I genuine, genuinely enjoy it. And thank you so much, Leonard, for

Leonard Lee 49:13

no. And make sure if you want an awesome guitar, by the way, hit up Andrew. Amazing luthier. And that's coming from me. Amazing. Astonishing. So kudos to you. And, also, please remember to subscribe to our podcast, which will be featured on Next Curve YouTube channel. Check out the audio version on Buzzsprout or find us on your favorite podcast platform. Also subscribe to the next curve research portal@www.next-curve.com, as well as, our substack for the tech and industry insights that matter. And we'll see you next time, gentlemen. Thank you so much.

Andrew Collinson 49:52

Thank you. Thank you. Cheers, deal. Cheers. Enjoy your trip guys.

Leonard Lee 49:56

Thanks.

Andrew Collinson 49:57

Bye.