Where is Telco going in 2026? (with Sebastian Barros)

In this episode, Leonard Lee, Executive Analyst at neXt Curve, and special guest, Sebastian Barros, managing director at Circle, former-Ericsson, and former-Google telco leader, tackle some of the top questions of the telco industry for 2026.

• Where is Telco going in 2026?
• What is 6G and what should it be about?
• Why will GenAI matter in telco and how? What about AI?

Sebastian is one of the bright thought leaders in the telco industry who has and continues to generously shared his insights and analysis at an amazingly prolific level. It was a pleasure to have him on the reThink Podcast to have an unscripted, honest discussion of the state and direction of the telco industry faced with AI hype, hope, prospect for disruption, and opportunity. How much of each will we see in 2026 and beyond, and when?

Leonard and Sebastian unpack the following:
💥 What will the role of the GPU be in the RAN?
💥 Is Open RAN dead, or does it have some life left in it?
💥 What is AI-RAN, and what is its value proposition?
💥 The reinvention of the telco and the telco business model - Telco Neocloud
💥 Addressing the telco industry stagnancy and 5G monetization gap
💥 Was telco and ecosystem readiness the thing that made 5G disappoint?
💥 Will 6G need solution-oriented thinking to be successful?
💥 Does the industry know what innovating on 5G looks like? What about 6G?
💥 How will the volatile geopolitical picture and outlook impact 6G?
💥 Will 6G be able to sustain 5G's greatest success?
💥 5G happiness is all about timing!
💥 Massive AI investments in telco will need to show ROI this year!
💥 Agentic AI will be the big focus and the big AI exploration of the year!
💥 2026 will be the year of the great AI unlearning!

Hit Leonard and Sebastian up on LinkedIn and take part in their industry and tech insights.

Check out Sebastian on LinkedIn and his Substack at https://sebastianbarros.substack.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

Sebastian Barros: 0:06

Next curve.

Leonard Lee: 0:10

Uh, welcome to this next Curve Rethink podcast episode where we break down the latest tech and industry events and happenings into the insights that matter. And I'm Leonard Lee, executive Analyst at Next Curve. And I'm here with a very special guest, Sebastian Barro, or is it Barros?

Sebastian Barros: 0:28

Barros. Barros,

Leonard Lee: 0:30

yes.

Sebastian Barros: 0:31

You have a good association.

Leonard Lee: 0:33

Yeah. I took, Spanish in high school, junior high, and. High school and all I know is Don.

Sebastian Barros: 0:45

That's actually quite handy.

Leonard Lee: 0:47

Yeah, yeah, yeah. It's worked for me, on several, very important urgent occasions. But, hey, Sebastian is a former Erickson, former Google, and then former Erickson again.

Sebastian Barros: 1:03

Yeah.

Leonard Lee: 1:03

And a provocateur, and a thought leader at the same time. And of course in all things telco and ai, right? You were like one of the few persons that I know who's handled a real human brain, not of a live person, mind you.

Sebastian Barros: 1:21

It was alive was, I think it was pretty. well, of course a disease patient.

Leonard Lee: 1:28

Yeah, yeah, yeah. Hopefully not a live one.

Sebastian Barros: 1:31

Yeah. A few recordings in, university, in the medicine department.

Leonard Lee: 1:36

Yeah.

Sebastian Barros: 1:36

And they allowed me to record not only the whole autopsy, right? I mean,

Leonard Lee: 1:40

yeah.

Sebastian Barros: 1:40

I so hold the brain. And so it was a quite big shocking and humbling experience, to be honest.

Leonard Lee: 1:50

Well, that was a really, impressive moment for you that impressed everybody. it goes to show the depth that you're going with this topic of ai, right? Because I think a lot of folks are enamored by what surfaces at the glass, but there's so much stuff underneath, and it's not just about GPUs or infrastructure. Ai, broadly speaking, is about so much, and it's great to see you, doing these explorations because. it's important for folks like you who have a technical background who have worked with technologies and surfaced them to customers as solutions to, understand, what the architecture, the solution architecture is for ai, right? And it apparently starts with a brain, a fresh brain. Yeah. So, we're gonna cover a simple agenda here, right? number one, I just wanna pick your brain, uh, no pun intended on, your thoughts on where Telco was going in 2026, which should be a fun one. I know that you, present a lot of points of view on that topic, especially, lately. And then the question everybody is asking, what is six G and what should it be about? Not what has it been documented to be about, but what should it be about? And then, touch on the topic of, ai, right? So what is it about, and then why will gen AI in particular matter in telco, given what we've learned, right? Because, one of the things I say is that there's been a lot of learning and there's an urgent need to unlearn a lot of things that were, learned that were not right, but which is

Sebastian Barros: 3:41

probably the most difficult part. I agree.

Leonard Lee: 3:43

and that's where you need to filter, right? And so before we get started, remember to like, share and comment on this episode and subscribe to the next. Curve Rethink podcast here on YouTube and Buzz Sprout. Take us on the road and on your jog on your favorite podcast platform. This podcast is for informational purposes only. Statements by my guests are their own and don't reflect that of next curve. We just wanna provide a forum for debate and constructive conversation on all things tech and on the industry. And with that, hey, I have to say this is so cool to finally, meet you virtually, and I'm looking forward to seeing you in Barcelona. I know you're gonna be there, but it's great to have you on the podcast and I'm really looking forward to this chat. So let's start. Where is Telco going in 2026? And you already have a lot of thoughts, right? By the way, you gotta check out a substack. This guy is like prolific, right? What you got like top 10 telecom startup investment strategies.

Sebastian Barros: 4:46

You know, it's funny. One person unsubscribed from my substack because he told me it looks, and I love your stuff, but it's too much. They cannot go with so many things. I mean,

Leonard Lee: 4:58

it's good though. I like this one. AI agents are not employees calling them. that is a technical bull crap. I totally agree with that. And you have, resuscitating open ran. Does it need a doctor

Sebastian Barros: 5:12

or a priest?

Leonard Lee: 5:14

Standalone is out. still waiting for revenue, right. Um. Me, uh, me, which I take as, um, 5G standalone is not like out the door. It's out right there for everyone to adopt, but still waiting or is being deployed. But where is the revenue kind of question, right? I mean, these are all important questions. And so, Hey man, the floor is yours. Where is Telco going in 2026? Tell us.

Sebastian Barros: 5:42

Yeah, I think 2026 is gonna be a super interesting year for the telecom industry and the main reasons because there's so many uncertainties. Yeah. I think this year will bring some lights, in terms of those uncertainties. I mean, one thing I feel it's one of the big things at least I want to know, is what is gonna happen with GPU in run, with opia announcing quite a big announcement actually. Besides 1 billion. Also saying, look, I gonna commit my RAN roadmap, which by the way is like 30, 35% of their total revenues with Nvidia, and I gonna bet that inference will be needed at the access level.

Leonard Lee: 6:30

Mm-hmm.

Sebastian Barros: 6:30

And they have in RAN or GPU in ran GBU program. Yeah. I'm very excited about that. the telco community has not been so excited about that. and here you have a clash of two school of thoughts, right? The deterministic guys, which I will put Ericsson on that side saying, look, I mean, that's overkill. Why you need to put a BPU in Iran? I think, yeah, we can do it with accelerators. It's good enough. Yeah. Then you have a new breed of companies saying, no, look guys, there are many things that we can solve from a signaling processing, and an RF that can be done better with neural networks. So I think that's gonna be a big definition for me because it change the way we think of networks. Coming back to what Jason one say is like, data centers are becoming ai factories, right? Intelligent factories is something that telecomm networks also should follow, or is only something that will happen at the regional data center level. So,

Leonard Lee: 7:32

yeah,

Sebastian Barros: 7:32

I think that's, that's a hot topic for me. Yeah. I think we should see some lights this year. Not deployments, but pilots and things that will give us some lights.

Leonard Lee: 7:41

Yeah. Yeah.

Sebastian Barros: 7:42

The second one, I think you mentioned open run, Uhhuh. Nailing the coughing or we gonna see, a revival story that will, impress us all. If you see the exit from NAC, the exits from ma from this business.

Leonard Lee: 7:59

Yeah.

Sebastian Barros: 8:00

Ericsson being the largest open run player, which is almost like a irony.

Leonard Lee: 8:06

Yeah.

Sebastian Barros: 8:06

Being followed by Nokia as the second largest.

Leonard Lee: 8:08

I know.

Sebastian Barros: 8:09

And probably the industry after this is a funny thing, right? After open run, which you could say it started somewhere around 2017, let's say, until now, there is more consolidation that diversity of vendors. Yeah. More Huawei, Ericsson, no. And Zoom than ever before. So

Leonard Lee: 8:32

yeah.

Sebastian Barros: 8:33

the open run folks will need to think a bit different in terms of what is their value proposition.

Leonard Lee: 8:39

Yeah. Yeah.

Sebastian Barros: 8:41

I don't know what you think there, but it's a question mark for me. I don't know if they will survive.

Leonard Lee: 8:46

Yeah. I think it'll still be around, the missed opportunity. that whole consolidation you're talking about was like the whole open ran squid game that I coined like four years ago. And, you know, it was sort of a cautionary type of thing that I came up with just to tell them, look, if you're gonna scale, you can't think open, you can, the technology can be open. The way that you build out the ecosystem, you have to create pockets of scale and they fail to do that. Right. Nobody wanted to own anything, you know what I'm saying? They thought they wanted to just. Owned the software layer. Nobody wanted to really own the stack, and that was the biggest problem in my mind.

Sebastian Barros: 9:26

I agree.

Leonard Lee: 9:28

and I talked to a lot of the big guys that were playing in the early innings of the open ran, um. You know, uh, race, if you wanna call it that. I hate that term, but anyways, I'll say it, but yeah, nobody wanted to own anything. You have to own it, right? You have to own it.'cause that's what the customer wants. And then we went to that phase of integration. Remember systems integration. Nobody thought, Hey, well, you know, wow, you have to integrate this stuff and then you have to have certification, attestation, all this, all the stuff that was typical of, building, a full, ran solution, right? These are things that were at the very early phase, not considered, but yeah. And now the real que the question now is going back to your comment about GPU, what is AI ran really about and what is that value proposition? And, you know, I'm curious, what do you think that has to happen there in order for AI ran to really be a substantial and meaningful. Because right now there's a lot of squishy stuff being put out there. Hey, you can do AI on Ran and then in Ran and for ran and blah, blah, blah. But, my view is it's still a, a conceptual exploration more than, a really well thought out, solutioning exercise.

Sebastian Barros: 10:53

I would put it as a, it's a very long short bet because two things needs to happen. Right. Let the first one, there needs to be demand. I was talking with, some of them media thoughts and, okay. When they build data centers with a massive amount of dpu, they already see the demand. so of course, and then that gives you a guideline, and at least some view that in the future there will be more demand for inferior. Yeah. Not so much of training. Training I think is more of a one time off, but now what we see is the massive demand on inference. Yeah. Now, the first thing that will need to happen, I think that's why this year is very interesting. We'll need to see some cases of enterprises or even AI models or hyperscalers that say, Hmm, I cannot do all the influencing at the global data center. I cannot do all inference also at the device level. So I need something in between that demand, that need. We haven't seen it yet.

Leonard Lee: 11:59

Yeah.

Sebastian Barros: 12:00

We understand the vision and it is a bold vision, especially if you hear Gene one and there's some logic to it saying, well, if the move the world is becoming intelligence everywhere, probably you will need inference closer to the user and the devices and the machine now that, that is something to be proved. Then the second question, which is also more down to earth, is can GPUs because it will be expensive and they will consume a lot of power, right? So you will not be paying a hundred thousand for a site. You will be paying probably 200,000 for a site. Yeah. the second thing that they need to prove, which is very concrete, is, okay, Ericson can do a lot of improvement on the RF interface. signaling processing, being forming and so forth by just using accelerators, right? Now, if you put GPU Run, can you improve those metrics? Can you make now a spectrum efficiencies of 30%? If you can show something like that, then the operator could say, okay, I'm spending several billion users in a spectrum.

Leonard Lee: 13:08

Yeah. To me it just sounds like that whole performance debate that was happening even with Open ran, because you did have these, everybody coming out with accelerator cards, accelerators. so this is not a new conversation. It's GPUs coming in, but my view is this, Sebastian, their value proposition has been from the beginning general purpose in that you can leverage the, the same accelerator, in fact the same infrastructure, okay, to handle almost every aspect, if not the entire, stack of, I ran compute needs all the way down to L one, L two, processing everything, right? and then on top of that, have the general purpose flexibility to support all these inference applications or business applications on top. And when you think about. How things have turned out, through every single one of these, cycles of whether it's crypto, GPUs were using crypto mining. they all moved to asic. They all went to application specific, right? And so building that case, a substantial case for the general purpose and, accelerator and accelerated architecture, I think that's really what has to happen. I think it's, my view is this, it's a tall order because there's non GPU, accelerators that are better for inference. And you, I don't think you're gonna be doing a lot of large scale model training, across a ran or you a network.

Sebastian Barros: 14:54

No, will be mainly in experiencing, because of course, I mean why you will use a distributed architecture to train, right? you will usually, I mean, most likely, train, where you can concentrate compute power. So I think

Leonard Lee: 15:08

Yeah. Yeah.

Sebastian Barros: 15:08

Is evolve. to be honest, it's a very bold bet from Noke. so we would need to see, find out if they can show some something this year.

Leonard Lee: 15:19

Yeah. I mean that's, yeah, that is a big, it's a big change. It's,

Sebastian Barros: 15:24

no, I agree. Agree. and it's not, it's for me, it's, I put it in the question mark. I mean, it would be great if we can have some sort of, Experiencing, at that level. And, we can integrate it with larger hyperscalers. But, there's a lot of things that need to happen. And also sometimes the things are not technology, right? We come from an industry Yeah. Is extremely adverse to risk. they are really pressured by shareholders on dividends and to be very cautious on the investment, right? they're not open AI or any of these, models that they can just go and burn money, And see what happens.

Leonard Lee: 16:03

So,

Sebastian Barros: 16:04

yeah, I think we'll need to wait, but this is one of the cool things to look out, this year.

Leonard Lee: 16:10

Yeah. Well, I do wanna remind folks that. telcos used to have a lot of money, enough to buy Time Warner, right? That was a lot of money. A lot of money that didn't go anywhere, because, um. Yeah. Uh, you know, a telco losing its identity, I think is problematic. So I wanna, pivot to the topic of, business models, and then maybe reinvention at Telco.'cause I know that you share a lot of points of view on that. I, I don't know if you wanna call it, telco needing to become a TechCo. I'm not a big fan of that. but, in light of what we're seeing here with this huge interest going and probably going into, I know that I was an interest last year even, probably even more so going into MWC 2026 in Barcelona, is this whole neo cloud adjacency play. what are your thoughts there? Because you have shared, Points of view on that. Maybe you can share the finer bullet points.

Sebastian Barros: 17:20

Look, AI data centers, run by telcos, or built in partnership by telcos is now the new big thing, right? And we have some big announcements like, they talked telecom with MB and also with SAP.

Leonard Lee: 17:33

Mm-hmm.

Sebastian Barros: 17:34

we have seen also SKT in Korea, launching, several initiatives, one include. So I'm looking forward to see what are the, but launching a sovereign, data center. we have been there before as well. Do you remember 2010. Telco said, look, guys, the biggest thing now is gonna be building our own data centers. We have the Fiverr, we know how to operate data centers because we have core networks, and now the business is gonna be co-location. We gonna co-locate whatever you need and we gonna offer these to enterprise. That thing didn't really work so well. if you look at, now by 20, by 20, 15, most of these telcos already sold for peanuts, any data center investment, because it was really bad. And what happened there? It became commoditized. A commoditized business. Pure real estate. There was real differentiator for a telco to do, right? So now the question is for me, okay. At this, telcos repeating the same mistake. And I think not really. I think there are some few differences. Of course, they could fall in the same mistake of, trying to compete in combat. But I have one rule, right? One rule is a following. I do believe that inference is the next big thing. It's not moving beats, but probably moving intelligence and moving intelligence mean you need to move inference. Now, inference, in many cases, especially in the world we are living, which is completely, polarized, will have some limitations to move anywhere. From regulatory, political views, security or latency and performance. So I think that's a place is not the next. Future of Telco, but it's a place where telcos can create some, so I think for instance, what telco, this one is Smart Europe is all about. they have very strict rules. even American companies cannot really freely move, workloads from one place to another. there is a lot of conscious about, security, aspects. So I think in a way, Doch Telecom is saying, well, this makes sense for me, right? I'm the local owner. I comply to all the regulations. Nobody knows better regulations and the local networks and me. So I think in those kind of scenarios makes a lot of sense.

Leonard Lee: 20:04

Yeah, I think as long as the operator understands the nature of the business of. Neo cloud and the economics of it, and that, when you lose sight of that and you're just looking at, the cycle of spending, and making bets on demand, the thing is the demand has to be monetizable. What we're seeing is non monetizable demand, in other words, the stuff is just being given away largely for free or compensated at a much lower level than the capital spending that's going into the infrastructure that supports or caters to that demand. That is that bubble dynamic that everyone, is debating at the moment. and so as long as the operator understands that and is able to proactively manage the economics of Neo Cloud. A neo cloud business, then I think they can survive. If they don't, the, it's like oil and water. The telco business does not match what's happening with Neo Cloud, right? The economics are so different it's gonna be interesting for me to see what, Deutsche Telecom, how they handle it, you know what I'm saying? there'll be a huge test case. you mentioned SK and, kt, the Korean, telcos taking on that, deep dive into the neo cloud, diversification, right? And then you see a lot of, the Chinese operators doing sort of the same thing, but they've been like CSPs, a blend. So they might have a better feel for what it takes to successfully pursue a, quote CSP adjacency and make that work with telco economics than anyone else. Quite frankly, what are your thoughts there?

Sebastian Barros: 22:02

If you look back, the mistakes of the past was that they build this amazing colocation capabilities, super capital intensive, and then workloads become standardized super quickly.

Leonard Lee: 22:13

Yeah. Yeah.

Sebastian Barros: 22:14

And there was no real business and margin was super compressed on data centers.

Leonard Lee: 22:20

Yeah.

Sebastian Barros: 22:20

and there was a complete organizational mismatch between. Telcos and data center. So that was one big issue.

Leonard Lee: 22:28

that's good insight. Yeah.

Sebastian Barros: 22:30

Today I think if they learn from the past, there are certain opportunities coming from. So I think that's one thing. Maybe the experiencing cannot move, of course, for a telco, which is a local owner of the network can be a good Thing. And there are some few things that I like. For instance, STC data are very interesting. partnership with Humane Human is trying, oh, you could say the open AI of Middle East, they're investing a lot. One of the challenges for Humane is that they need to accommodate more workloads in, data centers. PC was super fast to say, Hey, my friends, I already have a center three, which is their data center brand. I can accommodate for you, one, gigawatt of AI data center capacity. And also this will be so, and I think that that is a smart move because essentially STC is just taking captive workloads. Workloads so they can run, right? But again, if they think that they can go and compete head to head with Equinix, I think that will be, version two. Mistakes. The,

Leonard Lee: 23:43

the, the cloud. The cloud, telco cloud, the real telco cloud on 1.0. Right. So, hey, let's talk about everyone's favorite topic or question. six G. What should it be about? Tell us, Sebastian,

Sebastian Barros: 24:04

I think

Leonard Lee: 24:04

you, you have no lack

Sebastian Barros: 24:06

of would be the CEO of, I don't know, one of the big vendors.

Leonard Lee: 24:11

Yeah.

Sebastian Barros: 24:12

the problem with six G is that 5G didn't do a great deal of, promoting, these G standards. Right? Yeah. I mean there was a lot of disillusionment.

Leonard Lee: 24:24

Yeah.

Sebastian Barros: 24:24

Some of that, is understandable. I think, yeah, the mistake, made, especially if you remember 2017, right? when all the marketing machines start moving. it was crazy. I will not say the name of this person, but if you Google, 5G as, equals electricity, there was one CEO saying that 5G will be as important as electricity. So that was the height level that the market was expecting.

Leonard Lee: 24:53

Oh, aren't they saying that about AI right now?

Sebastian Barros: 24:57

I would say it was even Wilder. I remember 2017 remote surgery, flying cars. and you will ask me. for a remote surgery. Okay. The connectivity is important, but what else? What about the roles that will be doing the surgery? No problem. So give us a massive hype then. Because of all these geopolitical, stuff, the standard was divided in two. So you go initially 4G with asteroids, that was 5G NSA, and then everyone start complaining. This is around 2022 A, we're not getting any revenues. A you need to wait, you need to upgrade to 5G SA, because that's where the real funds start. Now most operators are either operated or in the process and still the revenues are not yet to be seen.

Leonard Lee: 25:48

Yeah. So

Sebastian Barros: 25:50

that's the backdrop. Now what's the problem, right? That I think the industry need, a new. Architecture, they need to rethink networks because they're changing rapidly. But nobody, I think even if you ask in the vendor community is really sure how to reposition. and we are running out. We are agree. If you think about it, this is, we are in 5G advance. 5G advance has released 1819, and then 2020 will come in 2027. Now we are commercially release 19, but 2020 supposedly is the link to six G.

Leonard Lee: 26:31

Yeah.

Sebastian Barros: 26:31

So kind of the natural evolution of the releases. And we are not there, we are not really understanding what we need. Now from my perspective, my humble perspective, why, what do we need? So there's few logics that we need to understand. the first one is that although traffic is not growing at the same rate still when you look 2030, with the most conservative projections, we will increase the total traffic in mobile networks by 2.6 to three three times.

Leonard Lee: 27:02

Yeah.

Sebastian Barros: 27:03

So they were running 108, 180 Excel lights. By 2031, we're gonna be running almost 500 exabytes.

Leonard Lee: 27:11

Yeah.

Sebastian Barros: 27:12

So that massive load needs to be managed. Yeah. And just doing, feature releases will not help. We need to rethink, the radio layers. the second part, which I think is important is that I don't expect an AI data tsunami on the access network. here. I might fair with, yeah, I don't expect that. I don't see, but I expect that the traffic composition will change dramatically.

Leonard Lee: 27:41

Yeah.

Sebastian Barros: 27:41

there will be more burst of information coming. there will be, more signaling coming, especially from wearables, uplink workloads, which is something we never, we, we never networks.

Leonard Lee: 27:57

Yeah. I wanna interject and point you to this, work that Mike Thelander did recently, looking at smart glasses and your comment about signaling as well as, maybe marginal impact that he saw with, uplink. probably the most interesting finding was that, one regarding signaling where, It was more prolonged than he had anticipated. if a application was active with the meta smart glasses versus just, running application with the smartphone. So I just wanted to highlight that and say hi to Mike.

Sebastian Barros: 28:34

is a great, report. I read it and I think it gives some lights to where the traffic will change in its composition. I don't think more like a tsunami, but probably we'll see, more uplinking, more signaling and many other things now.

Leonard Lee: 28:46

my point to him was it's still early. You don't have a lot of these devices to really, Be able to witness any kind of impact on, traffic today. And then there's a lot of stuff happening on device that, requires different thinking. It is not gonna be so much about the network.

Sebastian Barros: 29:04

one thing that we really don't know because I was also thinking in this case, the problem with all the wearables we use today is that they don't behave as the endpoint. They don't have really the 5G model.

Leonard Lee: 29:16

yeah. Right.

Sebastian Barros: 29:17

So we don't really know, for instance, if economics work, let's say Mike Zuckerberg decide to say, look, screw this Bluetooth thing. I want to put a 5G model in the glasses itself. That will be a very different scenario to the scenario we have today. Because what happened today is that since the, meta RayBan communicates to the smartphone, a lot of the functions, even AI functions, inference, is done by the device. So what is left to the network is not so much. So I agree there are many unknowns. then the other part that we need to think about is that with three times, more load, we need to think, that energy is a massive constraint we have seen in 5G. So in my view, we need, a new standard. but that standard might be needed for different reasons Now. The problem here is a following, right? I think this is the big elephant in the room, and I don't think anyone has a, a good question is telcos are not really growing. They're stagnant. They have been stagnant for more than a decade. the traffic keep growing. of course AI is coming to save them, in a way automating, different things. But, six G, I think, and, at least from what I see now, has good propositions to improve the rf, improve energy, bring AI to radio. So it's all about, you could say, opex and CapEx proposition and

Leonard Lee: 30:51

yeah,

Sebastian Barros: 30:51

course A, but I need something to make money, right? I don't see that yet. and that's I think where we,

Leonard Lee: 30:58

the problem there is that, number one. The mobile application is missing, whether, it's catalyzed by a or it comes about by a new device type, right? Whether it's any of these proposals for an existing, an emerging forms of wearables and alternatives to the smartphone or evolution of applications for the smartphone itself. I think the smartphone, people want the smartphone to disappear by kind of like smartphones. I don't really have problems with them, you know, but, there's this narrative that somehow it, it's going to disappear and not, but it actually continues to have a very important function. But those applications. Are missing. And a lot of that has to do with the maturity of the network, maturity of the operators themselves to have a channel of monetization, but more importantly, the supporting services or enabling services that those are not there. and so your earlier comment about 5G being kind of like a letdown, I would argue NSA made sense because the availability of endpoint devices that can support anything remotely close to, 5G, like, release 16 didn't exist. Right? Very small numbers, on the premium end. And, SA is probably landing where it makes sense, but we might be anticipating benefit. Too soon. because I, I, you know, I I, I'd love your reaction to this. I think that we have been so disappointed as an industry about 5G. Nobody has put a lot of energy into thinking about what is innovating on top of 5G.

Sebastian Barros: 32:53

Oh, okay. You, I think you hit the nail there. So my, my, my view is a follow it. and this is maybe because we're so used to think on the standards releases and features.

Leonard Lee: 33:03

Yeah.

Sebastian Barros: 33:04

Mm-hmm. The thing is that we tend to think that. A technical feature equals monetization.

Leonard Lee: 33:11

Yeah.

Sebastian Barros: 33:12

And honestly, that was true. when I joined Telco, that was true. SMS was a technical feature and you can make money, voice, you can sell it and data the same right now until that point, it makes sense, right? You have a new technical capability, you sell it fine, but then, we have a huge amount of technical features.

Leonard Lee: 33:37

Yeah.

Sebastian Barros: 33:37

I think what we are lacking of is. What do we do with them? Yeah. to some extent camera APIs try to answer that, saying, okay, at least expose it to someone. I think we haven't dig deeper there saying, okay, I'll give you an example. yesterday I was watching this, video from Ericsson on, mobile positioning. Yeah. So mobile positioning. Now with the 5G advanced, you can do outdoor positioning, And also inside indoor you can do precision, positioning. Fantastic. And they have like a nice little bit video with a robot, everything. Then you say, okay, these guys got a super cool feature. That could be used in industrial purposes, but have we thought about all the things we need to do to make money out of it? Ecosystem, partners, enterprises, device support. Usually the answer is not, We leave the feature, we say we have the feature, but we don't think too much on, Hey, what to do on top. Right?

Leonard Lee: 34:34

Yeah, yeah. Yeah. in my interactions with, operators, they're not really good at figuring this stuff out. I know that there's a lot of great collaborations like industry and consumer collaboration, going on. But the telcos themselves struggle with how to innovate. I think the industry in general, you brought up network APIs. People really struggle figuring out what the hell a API is. And now you throw MCP into the mix, it gets even more confusing. And, and that's the problem. And I think one of the things that is missing is that, solution oriented thinking, versus product oriented thinking like what you were saying before, right? You have a technical specification or a feature coming out of, the whole, 3G PP pipeline, and that would easily translate into a monetizable service. Right? Now you have to think about layers, right? SMS is just like, it's a communication. API somehow got squished into network, API, which I think is a different thing. And, now in order to get to those higher order applications and services, that support that you have to compose something that's very complex, right? It's not enabled down at the feature level, right? The technical level.

Sebastian Barros: 36:06

You need some orchestration layer. I don't know how to put it, but we obviously are missing, some sort of a layer between what the vendor produce because Ericsson, Okia and Huawei, after the standardization comes with so cool features. I was reading the latest, release 19, and there's so many cool feature even with 18.

Leonard Lee: 36:28

Yeah. Now

Sebastian Barros: 36:29

they will only leave until there, right? Yeah. Now then you go to the telco and the telco is selling seems and selling, enterprise solution, but they're not in a position to take those features and build all the product management you need, all the way to it. So this kind of a gap, I think this, joint venture, from Ericsson Aona, API, is trying to bridge that gap, at least for network APIs

Leonard Lee: 36:53

at a certain level, at a very, actually a very low level,

Sebastian Barros: 36:57

in a very primitive way, I would say. Because the problem I see, I mean, if you look back, right? I mean if you take for instance, Jason Hu I love the guy because he's really smart. He was saying, look, my biggest differentiator here is not the EPU people tend to believe that it's EPU, but it's not really EPU the real differentiator for me. The platform and that platform is Cuda. And of course you have many different, services, but on top of Cuda you have 2 million developers.

Leonard Lee: 37:30

So

Sebastian Barros: 37:31

way he has built this level of traction. he builds the machine and The E Ps. But on top of that, he has this orchestration layer where you have one or 2 million developers and they create whatever you want communications do, a space and, new materials, right? So this is the kind of aspects, we are missing. and I think I've been very bold to say that, maybe we need someone like DSMA to take a more active role to build developer community. It is really disappointing to see that the largest industry in the world with 9 billion Yeah. Doesn't have a developer community. I mean, It's weird. I think we are still the boys. It's a mess kind of mentality.

Leonard Lee: 38:15

and my observation because, I don't just cover telco, I cover a lot of stuff. I think you're absolutely correct. it's missing, and I think the industry is still struggling with this whole topic of, that next level of innovation. Right. What does innovation really look like? And, I don't blame the industry. it's a really tough, if it was so easy, then somebody would've figured it out by now. Right. But just constant. training of, mutual confusion. But I know that sounds really horrible, but, you know, you and I, we've been witnessing this stuff for years, so, we are trying to help. We're not just trying to be like, right, I know that you're trying to help, you're trying to c clarify stuffing, you know, get the ball moving for the industry because it is a great industry. I wanna get final take on a six G. I know we're going kind of long, but I don't care. I love talking to you and this is really cool. you mentioned the geopolitics it looks like a couple years ago or three years ago, even Five years ago we were talking about bifurcation of, the ecosystem, the mobile wireless ecosystem. Right? Now it looks more than bifurcation. It looks like fragmentation. and that's completely counter to the dynamic that we saw that brought everything together for 5G. how do you think that's gonna impact, almost sounds like a loaded question, but I would love to get your

Sebastian Barros: 39:42

perspective. Oh, no, no. Actually is one of the trends we need to see in 2000, 26, when, Trump administration started with the bans in 2017, right. With the, I think that was the first Huawei ban and then it was the semiconductor, It was a more soft stand via diplomacy to say, look, we prefer to you, if you remember, there was this clean networks, concept. Yeah. and some few countries speak up, but some other countries say, no, no, no. I prefer to use mind Huawei as a vendor. So it's not really so dramatic. now I think we see a different type of, us push, and I think this one is more dramatic. it will incorporate not only diplomacy, but even sanctions for countries who are not aligning to their, tech stack. Right? Yeah. And that also critical, infrastructure, like daily communications. So that's one part. Now China used to be kind of patient about it. They were not responding so hard, but I think now they realize that, this is a real geopolitical tech war. So they have responded. if you look at, Huawei, sorry. No. Ken Ericson shares in the largest market for telecommunication has dropped to, single low single digit from 15%. They user have, that's several million users, several million base stations less.

Leonard Lee: 41:11

Yeah.

Sebastian Barros: 41:12

and now Europe is coming again and saying, look, we are gonna revive the huawe and ban. we have Latin America also with a strong push now from the US administration to also go for, what they call, allies, network. So. I think that will definitely affect the standard. And we are at a big, big risk that, this 5G six G standard, maybe never see the lights as in the same shape as a 5G. Yeah, 5G. So, for me it's not good because the reason why we can have, I mean, we can complain about the industry and, but the real reason why we have 9 billion subscribers, phones that work anywhere in the world, almost magic, yeah. With anyone, Price per G gv that is as low as in India, I think is 10 cents. The ratio for that is because we have been able to remove all our differences as fellows, world citizens, and come together to build standards. Now we go back to, the CDMA era, right?

Leonard Lee: 42:21

so let's back into your earlier statement that, 5G disappointed. Why don't we recognize that as probably the greatest success, right? Is that with 5G, we had a single standard. Globally and what a lot of folks don't recognize, and I think a lot of policy makers don't recognize this either, actually, most people don't recognize 3G PP. The standards come together through a consensus process, and it's a constellation of global bodies that agree upon that standard to this consensus process. It's conducive to, a global, agreement. I think this is gonna be probably the most important topic going into MWC this year. If it isn't, then it's a huge miss. but I think it, it's,

Sebastian Barros: 43:20

no, I agree.

Leonard Lee: 43:20

Probably the biggest topic and it's this will define what six G can be, right? At the end

Sebastian Barros: 43:27

of the day expectations. And I tell you, I, I've been very, glad and I think he has always been a super, strong voice of Boria. eCom has been raising this for years, I think the first time Polio raised this issue was actually in the first, Trump administration that, look, any geopolitical divide here will only affect. the end user. So, so I think it's important and I can tell you, coming back to three bp, it's easy to criticize, right? And I think sometimes people might misunderstand me, but imagine the task of someone sitting in 2000, 10 when the 5G started and trying to imagine the future on how networks will behave in 2030. That's kind of the engineering path. Yeah. I don't even know where I gonna be in one week. And, and these guys, they need to think 20 years ahead, 15 years ahead. So building a standard is, is really, one of the most amazing, engineering, task in humankind. And we, I think we do a pretty good job there. And I can tell you, I think coming back to the 5G thing, this is what really. I really struggled. Leonard 5G as a product is not a bad product. I was reading about India, right? India Indian 5G. Well by the way, 5G is working somewhere and it's in India. Indians are super 5G, so at least we have one country, I don't know, 280 or something. Now, one thing I was reading with which I was shocked is that in India from LT to 5G, you get 19 times the speed. Of course, geo was as smart as forget about 5G and AI go 5G as a directly and they have scales to have very cheap devices. Yeah, 19 times now the price is the lowest in the world.

Leonard Lee: 45:23

Yeah.

Sebastian Barros: 45:23

So there's two other industry in the world that will give you something that is 19 times better in a way. And

Leonard Lee: 45:32

that is pretty incredible. Yeah,

Sebastian Barros: 45:34

that's pretty great now. The only problem we have is that we don't have a clue how to monetize that amazing improvements. Right. we don't know.

Leonard Lee: 45:42

going back to the earlier discussion we had about the readiness 5G readiness and timing of monetization or even scaling, that is, I think the leapfrog dynamic. they weren't first, they came in about the middle. a lot of, folks in the industry focus on the, network. But the thing is think about what, Qualcomm and others have done on the device and the modem, and bringing 5G features and capabilities to the lower tier. Smartphones. And so you have this diffusion of capability, and so that makes it ripe for India to take advantage of that technology diffusion at that point and penetration to do what they've done. going back to my argument about NSA had its place for the earlier adopters, because, the cloud transition was also the cloud native transition. Very, very difficult. We all know about that, right? It is huge architectural shift and lift from the legacy architectures of, the leading, telcos. Yeah, that's not generative ai. but, let's talk about generative AI and why it matters and then what about ai? What are your thoughts there? So we're going into 2026 six. I know that we've already talked about, the neo clouds, but in regards to generative AI and how it's going to influence the industry this year, what are some of those key thoughts that you're taking in and sharing with your audience?

Sebastian Barros: 47:24

Well, first I expect that anything in Mobile world Congress will have the AI attack.

Leonard Lee: 47:31

No doubt.

Sebastian Barros: 47:32

Even the coffee was gonna be AI coffee. I think of course, we are at the top of the hype. and sometimes, as a former boss used to say, humans are very good at overestimating the present, but underestimating the future. So we're probably still in a very big hype in terms of what GEI and AI can do. and I think this year, and I think this is important part, we will need to start seeing, the ROI of many of the projects that are running today. And here's where we'll lead to listen carefully to the CFOs. after spending a lot of money in different projects to see if the money's coming, to ai. I think I have two views, right? First of all, AI and ML has been used in this industry for almost 60 years. I mean, actually it's interesting. One of the first industries to adopt AI and neural networks, this was like one layer neural network, I think was telecom industry. Under Bell Labs. they were using, like super simple neural networks for signaling. So we have a longest history as a telecom industry using artificial intelligence and machine learning. And now, of course, 2017, the transformer revolution happens. Realize that if you put attention to certain words, you can create whatever you want, right? And then on top of that, on parallel, you have all the deficient models that also develop, but now you have these generative ai. I think the impact for the industry, is threefold. first I think, when it comes to coding, there is a massive, massive opportunity, And it's, it is well proven. There you have the hype, right? In my view, it will help, developers to build fast. You listen to he say, my developers don't code anymore. and then you have Max saying, look, 50% of my code is already I, so, but definitely that's a big part. and why is it's big because that will accelerate their and d especially from vendors, but also will help, telcos to develop applications much faster. So I think that's one bad, one big thing. The second part, which is more general to the industry, and I gladly saw that it is a lot, is doing something with, I think with Microsoft is a, now that you have an AI assistant, we can debate how intelligence, that assistant, you can definitely use it to improve your productivity. So I think, if I remember, it's a lot, right? they were training 30,000, employees on using copilot. The same with Tel was doing something similar, also with the partnership with Accenture and Microsoft. And I think that is a smart, quick win, right? Even if you write an email, if you can write faster or if you have a meeting and you can take notes, and persistent notes and send the many meetings, you're saving minutes of time. So I think that's a pro quick win productivity tool that I think all companies are adopting. the third one. Which I think is interesting is essentially the use of Geneve ai, for improving, the customer experience. And I think we have seen many cases, Where it can be a really solid tool, for companies to approach. Now what I think is the big question for now, and I think AI is one thing, but is agent ai, right? So now we are gonna make these. LM based models to also be connected with, actions.

Leonard Lee: 51:13

Yeah.

Sebastian Barros: 51:13

And that's where the fund will start, because now, one thing is that you as a model to check your emails and do a summary and other things that you ask the model to look into those emails, find which are relevant, and of course reply back and even take actions, in different other systems. So for me, Gentech AI is where we need to start looking into. there are so many pilots on automating network operations with Gentech ai, customer service sales. So I think that's where I will put my attention and if there's one case, I will put all of attention is the one that it ISAD is doing with IBM. They have sign, actually an AI agent framework using what's on x. and they're planning to roll out agents in edges a lot as not employees, but as kind of a Second layer of workers.

Leonard Lee: 52:07

Yeah. Well, I like what you said in one of your articles recently is that treating agents, like employees is

Sebastian Barros: 52:16

this is the height that pissed me off, because I can imagine the McKinsey CO is a very smart person. the statement sounds mis ingenious. I mean, agents are still based on LLM models, LLM models, many, many issues that we cannot solve because they are unsolvable. One big problem is hallucination. and the hallucinations are pretty high, like 15.

Leonard Lee: 52:40

they're much higher than people think. And even with reasoning models, I had some practitioners tell me that it's under, more practically under order of, anywhere from 40 to 50%. That's pretty bad. Even though the people are going up on stage saying 90% accuracy, it's much lower than that.

Sebastian Barros: 53:00

so then you can imagine, Lorna, that I connect because the agent will be based on that model and now will have that agent, for instance, checking fraud.

Leonard Lee: 53:10

So

Sebastian Barros: 53:11

imagine a bank whereby the agent miss 20% of the frauds.

Leonard Lee: 53:16

It's really bad. Yeah. Yeah. Well, that's the thing. The term AI has been misappropriated to mean generative AI or to include generative ai when there's all this other work that has been done. This is the reason why I wanted to distinguish between what, how will chain AI matter versus ai, the, like you said, telco has been working with ml, and a AI broadly for decades, and some of the stuff that works, which took a lot of effort, right? it finally has a degree of reliability where exhibits value that's being either confused or overshadowed by generative AI narrative. And so that, that's one of the concerns and I mentioned before. So my whole view on what's gonna happen for the industry in 2026 is what I mentioned earlier, unlearning. There's going to be a mass scramble to unlearn, and then there's going to be a need for filters. Folks like you who can kind of see through all the hype and help the industry land on something that's workable. And because most of the things that are floating out there are not workable. they're gonna be wasted investments, based on, misguided narratives. So,

Sebastian Barros: 54:35

I think it's also one of the things you learn in these domain, right, is, even the most expert AI and ml developers, they tell you it's trial and error. I mean, it's really, almost a science field that is mixed with art. So, I think if I were in any telco shoes, I will be experimenting like crazy. I will definitely make sure that all my employees. Understand ai, understand how to use it. I think that's, a no brainer. it's like the internet era, right? when we move from fax machines to

Leonard Lee: 55:14

Yeah.

Sebastian Barros: 55:14

To emails. You

Leonard Lee: 55:16

don't go there,

Sebastian Barros: 55:19

train people to use an email. I think that's a no brainer. and then the second thing that we have as an industry, and maybe this is politically incorrect to say, is that we are in an industry where our CEOs are not very technical. if you compare to tech industries, I mean you have, I don't know, from El Mass to Jason Juan to Max Berg, to ela. I mean this, like our developer level CEOs. Our CEOs are not coming from that ground. They're coming mainly from CFO ground. so we will need. Them to understand ai. And, and I think you made a super good point, like basic understanding. Like, hey, look gene AI is just one piece of a much bigger picture. So when someone talks to you about gene ai, understand that's one context. So CEOs will need to make a big effort to also learn and unlearn what they know about ai.

Leonard Lee: 56:16

And the unlearning part, there's a lot of unlearning that has to happen. But, hey, with that, man, this was really fun and it's an absolute pleasure to hang out with you. we can, this is gonna be the longest podcast I ever recorded. We're having too much podcast.

Sebastian Barros: 56:31

Make sure you have a big, coffee.

Leonard Lee: 56:34

Yeah. But, why don't you share with the audience what you're up to now, and how they get in touch with you. obviously Substack people hit'em up on Substack big time.

Sebastian Barros: 56:44

I mean, as, as you know, learn, I love to interact ideas. If you want to reach me I'm always available on LinkedIn. If you really want to go with more provocative, thoughts and ideas, we can meet in Substack. I have my substack that is actually growing really, really fast. and I will be in Barcelona. So happy to meet you all with tapas.

Leonard Lee: 57:07

Yeah.

Sebastian Barros: 57:08

And be your guide as I can speak Spanish so I can be your guide. So, yeah, looking forward to everyone and, and yeah, I love what you're doing, Leonard. It is really important for us to build the telco community, where we can talk about the real thing, right? I think this is super important for us. So I think really appreciate your effort. I know you put a lot of effort and you are everywhere also, so you have a super busy agenda. So really appreciate the effort for the industry.

Leonard Lee: 57:35

Yeah. No, I appreciate that and I appreciate you. And so, again, thank you so much. And everyone, please subscribe to our podcast, which will be featured on the next Curb YouTube channel. check out the audio version on buzzsprout. Find us on your favorite podcast. Platform. Also subscribe to the next curve research portal at www dot next curve.com and our substack for the tech and industry insights that matter. Sebastian, it's been an absolute pleasure and we'll have to do this again, and I will see you in Barcelona, right. We'll make an appointment for sure.

Sebastian Barros: 58:13

Fantastic. Thank you very much everyone. Thank you, rna. Take care. Thank.