5G Americas Analyst Forum 2024 Recap (with Earl Lum and Prakash Sangam)

The 5G Americas Analyst Forum 2024 took place in Dallas. Leonard Lee of neXt Curve, Earl Lum of EJL Wireless Research, and Prakash Sangam of Tantra Analyst were there. This year's event highlighted the need for 5G standalone deployment to enable advanced features. The event also focused on the complexity of Latin America’s wireless industry and the importance of affordable 5G handsets for the region. 

Despite the hype surrounding AI, discussions were more realistic about its practical applications and the need for right-sizing vendor operations. In true 5G Americas fashion, we couldn't get enough of the expansive agenda and top-notch engagement with the leading operators and tech firms in the telco industry in the Americas.

Earl, Prakash, & Leonard discuss the following topic that mattered from 5G Americas Analyst Forum 2024:

• Impressions of 5G Americas Analyst Forum 2024 (2:40)
• 5G Americas by the numbers (4:42)
• The Latin American 5G story (6:48)
• 5G Standalone, the key to the 5G promises (7:43)
• The GenAI hype in telco is cooling and AI is practical again (11:20)
• FWA is the evolving killer 5G app if there was one (14:10)
• Innovations in the RAN. Anything new? (18:29)
• AI appropriation in telco (24:42)
• Scaling out a 5% error rate on your network (28:26)
• Automation versus AI - they are not the same (32:39)
• Network APIs in the work in progress  (37:30)
• The panel predicts the themes of 5G Americas Analyst Forum 2025 (40:10)

Connect with Earl Lum at www.ejlwireless.com and Prakash Sangam at www.tantraanalyst.com. Hit us up on LinkedIn and take part in our industry and tech insights. 

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

Also, subscribe to the neXt Curve research portal at www.next-curve.com for the tech and industry insights that matter.

Chapters

• 2:40 Impressions of 5G Americas Analyst Forum 2024
• 4:42 5G Americas by the numbers
• 6:48 The Latin American 5G story
• 7:43 5G Standalone, the key to the 5G promises
• 11:20 The GenAI hype in telco is cooling and AI is practical again
• 14:10 FWA is the evolving killer 5G app if there was one
• 18:29 Innovations in the RAN. Anything new?
• 24:42 AI appropriation in telco
• 28:26 Scaling out a 5% error rate on your network
• 32:39 Automation versus AI - they are not the same
• 37:30 Network APIs in the work in progress
• 40:10 The panel predicts the themes of 5G Americas Analyst Forum 2025

Transcript

Earl Lum: 0:06

Next curve.

Leonard Lee: 0:09

Welcome everyone to this next curve rethinking podcast episode, where we break down the latest tech and industry events, things, you know, into the insights that matter. And I'm Leonard Lee, executive analyst at next curve. And in this episode, we will be recapping 5g America's 2024 events. It's it was the analyst. Forum, which took place in Dallas, Texas on October the 24th. And I'm joined by two of the leading analysts covering mobile wireless and telco infrastructure, Prakash Sangam of Tantra Analyst and Earl Leung of EGL Wireless Research. Gentlemen, how goes it? Doing well, let me hear. Don't talk at once. Uh, yeah. Hey, before we get, and you know what, I'm just assuming both of you are doing fine. That's what I think I heard. But before we get started, remember to like share and comment on this episode and subscribe to the rethinking podcast here on YouTube and on buzzsprout. Take us on the road and on your job and everything else. Listen to us on your favorite podcast platform. And so gentlemen, I would like to confirm that both of you are doing very fine. Is this correct? Correct. Prash? No, no. I said this. Hear a thumb, Hey, it was great hanging out with both of you. At 5G Americas and this year event was really another great one. I have to tell you I think this is the best analyst event that is put on during the year for pretty much Any industry. I think it's amazing what chris pearson and his team have done over the years and the advocacy that they've fostered for the telco or Months Wireless industry in particular in the Americas. And it's not just the United States. It's all of the Americas, right? So Canada and all of the Americas below our southern border. It's a huge mandate and it's a great, I don't know. What do you guys think? I freaking love it.

Earl Lum: 2:40

I thought it was great. And I thought that the second keynote was. Great that because it focused on that Southern part of the Americas that we normally don't talk about. Jose Ortega's presentation, I thought went really well and really highlighted the problems in the region and what they're trying to address and. The extensive number of people and regulatory officials and agencies and everything that he has to coordinate by himself. Essentially, or what seemed like by himself, and that was just an amazing. Update to really understand what he was doing and what they were doing as an organization and. I guess I had wrote on a feedback form. I don't see enough about Latin America and whatever. So I guess they read it and and that was great. So I really thought that it was nice to have that focus. On south of the border.

Leonard Lee: 3:35

Yeah, what about you? Gosh, what do you think about in general? In general, I

Prakash Sangam: 3:40

agree with you. This is 1 event where, because there is no media, everybody operators, mainly. Tend to open up a little and talk freely rather than just repeating their corporate messaging and corporate messaging guide, if you will. And then I agree with and you on Latin America focus. Although it's been part of what Fudgy America has been doing for a long time. Thank you. But this year there was a specific focus on the complexity. We tend to think, okay, maybe, there is Brazil and few other countries, right? Mexico, Brazil, and few other countries, what? At least I don't cover Latin America that much, but that's my view of it. But when you, And open their apps, you see it's so many countries, so many regulatory agencies and so much of complexity. So I think put it very well on the complexity part of it and all the things that Fudge America does for its operators and the vendors, pushing Yeah. Wireless boundaries there.

Leonard Lee: 4:42

Yeah. And speaking of Chris Pearson, he wa he was not able to make it this year. Viet Nen, who is the VP of I believe, isn't he the VP of Marketing Correct. Patients. He really stepped in and yeah. And really huge shoes. He moderated and he actually led the chair. The board meeting that preceded the analyst event, and then he also emceed the whole event with Jan Walker and I thought they did a great job. Obviously it's weird. To have a 5g america's without chris and hope to see him next year filling his own shoes but yeah, I think this year was interesting, you know speaking of The keynotes you had Yal Elba, who is the his title escapes me, but he, VP of sw P at t? Yeah. Yeah. At and t. He's a board member. He stepped in and did the keynote and. I guess this year's theme was the dawn of 5g advance and technology readiness and he brought up some interesting points, I know that the they always love to start off with the global Count which this year now they're touting is now over 2 billion 5g connections globally with two about 250 or so in north america which is great, but I think it, seems like the handset vendors, the handset guys have done a great job of propagating, you know, the 5G readiness and capability. We need 5G advanced and the 5G SA transition to really take hold, I think is really the point that everyone at the conference was trying to make, or at least, the vendors and the operators were trying to come up.

Earl Lum: 6:48

For North America, I think if you remember what Jose said, having a sub 100 5G phone for the Latin and Caribbean markets was a key enabler, because it's still mostly poverty down there, and they're not going to pay 400 or 500 for a 5G handset. So that is certainly holding back the sub subscriptions down there, I think, at least from a handset side, right?

Leonard Lee: 7:13

Yeah. And it's also holding back the value proposition, the value proposition of 5G is really anchored on this transition toward 5G S. A. There's a lot of talk about 5G advanced, but most operators out there haven't even made the transition and broad deployment of standalone. Standalone. And that just gets you release, six types features. Do you know what I'm saying? It sets you up for 5G events.

Prakash Sangam: 7:43

Yeah, so anything more than broadband, if you want to do it with 5G network, you need standalone. Most of the things that you need is 5G advanced. You need standalone. I know it's very clear in everybody's mind that standalone is needed. However, what is not clear is what it will, what will. enable operators to think and invest in going, network wide with 5g advanced. Sorry, with 5g SA. So what they have now is, you might have noticed The kinds of applications and the next level of 5G everybody's talking about. It was the same as before last year. So same application, same verticals, which means the thing has moved on that front, right? There was a lot of optimism last year because we had just come off of a good 5G broadband launch and deployment, and then there's a lot of enthusiasm and optimism of this next phase of 5G, right? But in last year, I don't think a lot of it has materialized. We are still thinking about, oh, these are the possible use cases, verticals, and so on. A lot of groundwork has been happening, but actual deployment, I think, is still very much lacking. And a lot of it is because of 5G standalone. And in my view, it's a catch 22 situation. Cases that, that need SA are localized for specific locations, industries, verticals, and so on. But for an operator to do a network wide deployment, You need to be sure that it will be used across the network, right? If it's only a few pockets, we'll use it, but they have to deploy it across the network. It's not going to work out from a ROI perspective. So I think there is a push and pull on both sides. We'll see when that converges and make it happen. That's one of the questions I asked during the The panel after the keynote is red cab is one of those features because that might need the kind of network wide coverage. Would that be a catalyst in pushing operators towards 5GSA? However, I should note that T Mobile has much wider network of SA compared to other two operators, right? Right,

Leonard Lee: 10:10

but yeah, but they're rolling out the deployment, right? Yeah. It's not that there, they, you flip the switch and all of a sudden everything is a standal, right? Yeah. But yeah, and last year there was, I think there it's not that I think there was a lot of excitement about private networks. I thought it was a lot louder, lighter this year. I don't know why I got that perception that conversation wasn't as robust. And yeah, oddly iot started to peek out a little bit more this year because of what you mentioned, the red cap, what, quite honestly, I thought the way that folks were talking about red cap in one of the sessions that I was in, it was not that robust. I thought it was a little too, let's say non differentiated. The positioning, but anyways, we'll talk about, our impressions on the sessions actually right now I know that we all try to go to different sessions and rare occasion. We were in the same session, but which ones were the ones that you guys are focused on this year? And, Which ones are the ones that you thought were really key this year?

Prakash Sangam: 11:20

I can start. Of course, 5G, right? Sorry, AI, right? Not 5G, AI. Any of the AI roundtables I went to, there was not even a sitting room. They had to bring an extra chair to sit, right? I think me and you and a couple of them, Leonard, there was no sitting room at all. So the ones I went are AI related specifically on operations and use cases and and private networks. I wanted to find out private enterprise networks on how it is, I have the same feeling. I think the positive part is. Last year, there was a lot of, it is not a lot of hype, but the realization has started that it is not another huge, step up increase in terms of market size like we did cellular network, it's another round of kind of a public cellular network kind of deployment. The realization is started, but this year it has sink in. So the discussions were more realistic on what's working, where it should be focused and so on. And then when you look across the vendors, they're more focused. They're employee base supporting this kind of has shrunk as I've seen and and they're not basically right sizing, if you will, in terms of operations and the expectation that they have on we can talk about AI right now, but I think it's dealt as a separate topic, right? What about you, Earl?

Earl Lum: 12:52

I did the open ran development session. A, I was covered open ran cool. F. W. A. was another 1 and then advances in 5g radio technologies on the antennas because I love the tennis and then ran architecture innovation. And again. For those listening or watching, there was way more than five that we had to pick from and it's always that case, which is annoying in that I want to go to all of them, but you just can't. Yeah, I agree. So we can't go back and listen on what was said at any every 1 of these, close

Leonard Lee: 13:32

rooms. And the worst thing is that each of those sessions, because they'll double up on the sessions. They'll have 2 sessions for, let's say, a certain topic and each 1 of them. Each of those discussions is unique.

Earl Lum: 13:45

Exactly. There's always a great lineup, but then you have to again, sacrifice. And make concessions on which are the top five that you really want to go to.

Leonard Lee: 13:58

Yeah, you gotta divide and conquer like we did. Yeah, no. Yeah, so what were some of your key takes? Especially, can you share some perspective on FWA? I didn't attend that one. Prakash, I don't know if you did. No, I did not do that.

Earl Lum: 14:10

Yeah, the FWA was, I think, The most important thing that everyone should understand is I don't want as a mobile operator, 100 percent of my subscribers to be FWA. That would kill my network. I don't ever want that to happen. The question is what is that breaking point? What is that line that I want in terms of my mix? And then again, if we look at T Mobile as an example, they deploy FWA where they have spectrum they're not using. Makes absolute sense. You're not using it for mobile. Why not use it for FWA? But in areas where everything's already congested with mobile, you really don't want to use for FWA or you don't want, you need to be judicious. And how do I allocate my network resources now to a pipe that's being sucked dry by all of these subscribers that are paying way less than my mobile subscription monthly rates? And how do I balance all of that? And what spectrum is the best spectrum to use for that? And at what point do I run out of. And 41 for Timo and do I start looking at millimeter wave or some of these other pieces of spectrum to get that 12, 000, 000 number or whatever that Sievert said on their last right?

Leonard Lee: 15:26

Did they talk about energy at all? The cost of serving?

Earl Lum: 15:30

No, it was more about the, that session was about business FWA versus consumer FWA, and I think everyone was in agreement that you want the business FWA because of the stickiness of that subscriber compared to the potential churn you're going to get on the regular consumer side, and they're willing to pay more for whatever other added value services and such like that. I think it was a nice mix of the people that were the subject matter experts in that room. One of them was from Ericsson Cradlepoint, the other one was from Cisco, to talk about what are the differences, obviously, between a business account for FWA and what you garner in revenues and the ROI for doing a truck roll there versus a consumer account.

Leonard Lee: 16:16

Yeah that's really insightful stuff. That's really insightful stuff.

Prakash Sangam: 16:20

And also, in enterprises, they can actually differentiate further and tear the market, right? With the slicing and other things, they can provide SLA for the ones that are needed and so on. Yeah. But at the same time, I think if you listen to the the T Mobile investor day, was it, that they're looking at fixed wireless as the major revenue generator for their 5G network in next few years. All right, I know they started the network slicing for emergency services, first responders and so on. Their forecast on how much more revenue they could bring from the next phase of 5G, I think is limited and they're being cautious there. Yeah I think the,

Leonard Lee: 17:12

the idea that, you're going to have FW in every market Kari he mentioned or was it him or someone mentioned a T Mobile has split up the market into like over 400 little hexagons that have the same availability of FWA in each of those hexagons, in those areas and The economics of all of this is really going to have a huge impact on number 1, where they're going to make those services available, but also just the revenue opportunity. And it's augmenting. I don't think it's core. I never thought it was core, but we have to consider that these are net new opportunities for wireless operators, right? And even. Cable codes that are coming in with their wireless strategy, be able to get to the last mile using FWA. So yeah, I actually, this is a very super interesting modality of connectivity. And it, that's really insightful stuff that you guys are sharing right now. It's

Prakash Sangam: 18:14

the, it's the only new revenue earning and new subscriber bringing service 5G I've seen so far. So yeah.

Earl Lum: 18:22

And it's what wall street. Is focused on which is why they're focused on it because they want their stocks to go up.

Leonard Lee: 18:29

Okay. Hey what about innovations in the ran architecture you mentioned are all open ran any any other types of innovations or Takeaways that you got this year that were different from what we heard last year

Earl Lum: 18:45

In my session that I went to for the Open RAN, because there was two, the focus was trying to understand this new D2 interface between DUs where you could do carrier aggregation. So it was another spec that got thrown into the pot that they're going to have to figure out and the question is, how does that delay anything that's already in the works today? Because you just added another thing that you have to go and make sure you're compliant to. As part of the overall specifications, and I think in terms of the market, there wasn't anything new that was really being discussed in terms of on the radio side or on the Rick side or anything else. It was again, just this new interface that has popped up now that you have to comply to potentially. And when. I'm really not sure on the timing of that. And is it critical today to get it deployed? Probably not. But it's just another layering on of all the new specifications that have to be there. To get everything going on open RAN. So it just seems like we're stalled right now, given the global events that are happening around open RAN and what everyone's feelings are. So maybe next year will be the year where we decide, is it going to happen at all, or is it just continuing them to work its way through and muddle through whatever it's trying to do?

Leonard Lee: 20:10

I didn't see much either this year other than some evolved thinking around SMO and the RIC. So taking more of a top down approach, even though we're starting to hear more about distributed distributed apps or D apps. But that feels more like everyone trying to figure out, okay, how do we keep The momentum going here, but to your point, I don't necessarily see that the movement is humping in on actual problems that are going to help the operators accelerate the modernization of their network. And which is, I think, really important. Like what we were talking about earlier this transition toward the 5G SA, deployments is really more important than open. Right. I think open continue continues to have to sell as relevancy. And this is what we've been talking about for a long time, especially you and I Earl about open RAN and what, this what's turning out to be a pretty broken mandate. More I hear about the open RAN, the more I think that it's actually curtailing innovation. Because a lot of these things are just stuck, right? These ideas are stuck, they're not moving anywhere.

Earl Lum: 21:31

Right, or I have to spend more resources figuring out what is this thing that I have to be compliant to and I'm not working on some other problem that I should be working on that's more tactical for today.

Leonard Lee: 21:44

And I do hear, I did hear some attempts to interweave AI into the conversation, especially with the, with a lot of talk about AI RAN, right? And So now let's talk about AI for a moment. Maybe that, I don't know, that AI also didn't seem to hit as hard this year as it did last year. Yeah, last year it was nuts.

Prakash Sangam: 22:10

So everybody has a view of AI, Ryan, but they're not so sure what it is. So it's basically infusing AI in different parts of the ecosystem, parts of the network, right? That's the general idea. And I attended a couple of sessions. I think there was a robust discussion on, You know what AI means for the network. So is it a gen AI that is we are talking about? First time I hear people when you talk generally people asking are you talking about regular AI? Is it gen AI? So that to me shows some maturity and people thinking about it's not just throwing AI word and talking of things. Talking about it, using it. And then some specific use cases where not necessarily gen AI, but the AI has been used. AI is not new, right? Regular AI for telecoms. So a son has been around for a long time. And then using that in the networks, using that for the wireless link optimization as part of 5g advanced leads 18 and above, there are some good I know robot discussions around it that I was in and is there, when I asked the same question last time. Okay. So all of this needs lot of AI models who will bring these models will own the IP, who will own the market share? Who will optimize manage them? They seem to be, some clarity on it that it's not going to be just one provider bringing all the I and I P owning the I P. So it's bringing AI models specific for use cases, maybe coming from different parts of the ecosystem. And that's when, the discussion of SMU, SMO, for example, becomes important that you basically organize at that level for the operational part. And then on the radio link side itself, there's still a lot of work yet to be done on making sure this cross learning happens between network on the device on their model side. So I think There is some concrete work being done on the AI side rather than just being hype, and there are some promising use cases which will solve problems for challenges for the operators, especially in terms of operations both the network operations as well as customer operations and so on. Yeah, it's at a very high level.

Leonard Lee: 24:42

Yeah, I think there's at the moment some significant. AI appropriation happening where the generative AI community is now labeling every, the before, last I would say previous, actually previous 12 months, mm-Hmm. introduction of generative ai. Everyone was talking about generative ai. Yeah. These folks are appropriating AI and then throwing everything that. People assume it's in this AI bucket into this extended gen AI definition, but what they're actually talking about is gen AI. And so we saw this in a number of the AI conversations, but you're right. We started to ask them, what are you really being here? Are you talking about generative AI? Are you talking about all the other good stuff that, some of it works, but there's some stuff that actually works really well. And, they're really talking about generative AI, but here's the thing that last year my, one of my key takes was it's experimental at best this year, it's experimental at best. 1 of the things that really concerns me are these people going around saying they've solved hallucinations. They've solved all these things, and I'm telling you, I don't hear that from researchers. I don't hear that from people who are actually technologists and data scientists who actually know the technology. They don't say that. So there still needs to be this reconciliation, you know what I'm saying between what the researchers and the scientists and the people who actually know this stuff in terms of a technology are saying versus what some vendors are saying, and maybe even some operators. Right? And so we, my take is we have not moved that far. And there's no one presented a case where they were able to deliver some kind of generative AI magic.

Prakash Sangam: 26:39

Yeah, Jenny, I think, the realization is you don't have to wait for Jenny. I make AI useful for operators, right? The deployments of the uses is more on the non Jenny, if you will, in operations. And so I asked this simple question. Sadly, in the panel I was in On the round table, there was no operator, right? I did not get the response from the horse's mouth, but I got some feedback from the providers of AI to this. If you consider operator chatbot on their website should be the lowest hanging fruit for using generative AI, right? And still, It is if you and I keep on checking when I checked last time, not to, a few days ago, still most of those those chatbots are static chatbots. You ask any little bit difficult, complex question. They connect you to the So the operator or the answers are totally gibberish, don't make any sense, just static. And then, many people over there, many, the subject matter expert, cited hallucination as could be one of the reasons why operators are not, switched on Gen AI on those chatbots because, It's the primary interface for them with the customers and they don't want to mislead or do anything that might jeopardize their relationship by providing wrong information that they have to go correct later and so on. And then many of them cited hallucination being 1 of the reasons why they may not be using Genia and the chatbots yet. So there is some realization of that sort. So sadly. I didn't have any operator SME on that panel.

Leonard Lee: 28:26

Yeah, but I did have an interesting analyst interaction where one of our friends said something about, well, you know, it is about 85% effective. So that's pretty good. I said, well, scaling out at AI speed and velocity, a 50 percent error rate can be catastrophic. So actually not being able to do that. It gives an operator time to make sure that kind of error that's, let's say, human driven to be corrected. The fact that you can't scale out 15 percent automated highly scalable Disruption of your business is actually a good thing. Right? And I don't think we've arrived at that yet because we're so impressed with these numbers. Like, oh, 95 percent accuracy. It's like, well, that 5 percent scales out very, very fast and the last velocity of that impact and scale up is also very, very fast. And massive and get to scale very quickly. And so these are things that, I find absent in the generative AI discourse in particular, not the. AI, this, the broader AI discourse around ML, because they've already gotten beaten up several times, right? So they're a little bit better. I don't think the generative AI narrative has been kicked upside the head. Yeah.

Prakash Sangam: 29:53

It just won a comment title then after the We have to

Leonard Lee: 29:57

let Earl say something.

Prakash Sangam: 29:58

Yeah. So just one, one comment. So I think the accuracy the the importance of accuracy or the impact of accuracy is a little bit different. If you're doing anything, which is directly focusing with the consumers and your uncertain customers, then you've got to be extremely accurate because any hallucination. will have a catastrophic impact. But if we are using it for internal for process improvement, I think 85 percent would be if you're if the improvement in terms of, it is improving your cost structure, reducing the cost when it's only 85 percent accurate. I will take it if it is not directly touching the customers directly. Right. So if it is Saving me, 50 percent of more of cost, but it's only 85 percent accurate. I think it'll be a good trade off to make in my view.

Leonard Lee: 30:50

I don't know. I think that somebody needs to do it's going to be a case by case and people need to do the math. How much does the human Intervention and how much in human cost is required to deal with in reverse a scaled out.

Prakash Sangam: 31:09

But if you're reducing the effort no reducing the human enrollment by 50, 50%, but it's only 85% accurate, I think it's a good tradeoff and also. Yeah

Leonard Lee: 31:20

the only reason why I say that is, I have a consulting background. I've seen bad code do that, put into production. Bad processes create a tremendous amount of manual work. Yeah. So the entire value proposition that you went in with. With this theoretical automation, you have to put quality in there and that and it has to be deterministic and intentional. That's the only reason why I'm pushing back a little bit on what you're saying. But hey, if there are, that's not to say that there aren't. Scenarios where what you're just suggesting play out. It's just that it's not going to happen everywhere. It has to be

Prakash Sangam: 32:02

And you have to be selective the reason i'm saying is that 15 percent, you know Will take much more effort than getting to 85 percent, right? So if you want to clear, 15 percent gap the effort you needed to get to the 85 percent and with x You Then to gather to close that 15 percent gap will be much more than X in my view, the amount of effort needed to make it 99. 9999 percent accurate. Okay, sorry. No,

Leonard Lee: 32:32

no, now we need to let say something because this whole time he's been thinking about saying something really, he's thinking how stupid and I'm

Earl Lum: 32:39

also trying to make sure that I don't forget. What I was thinking is fleeting. I actually was in a session where there was an operator who was realistic. And he said, we don't use the term AI. We use the term automation, automate things. And we want to automate it with machine learning. And then I asked, well, do you beat people every time they say the word AI as a negative reinforcement or something? But I think that operator is realistic. And I think that's where, at least in the RAN. I don't need a 90 billion parameter model to tell me that at 2 a. m. in the morning, I should turn off the electricity to a bunch of my sites to save money. Right? So the question then becomes what's how big are these models needed to be for channel estimation or some of these other things? Am I doing it in real time? Or I'm just learning from it? A trillion parameters I've already gathered that I just need to sort through. And I'm using AI to just sort through all of this stuff to figure out, is there a trend here? Do I need to have it done in real time? Probably not. As long as I know what that trend is, I can adjust the parameters of the network. And those are the easy things, at least from a RAN side. If you're talking about from customer service and everything else, sure. That, that has to happen, but if you went to someone in manufacturing and said, I can give you an 85 percent yield, would you be happy with it? I'd say probably not because that 15 percent of yield loss is going to kill me in whatever I'm manufacturing. I don't know if that correlates to, you can get an 85 percent improvement on something. But I think in the Rand side, we don't need to have a GPU stuck at every site. Doing whatever the channel estimation and stuff is going to be happening data that everyone can gather from already about what's happening in the environment to where they don't need to do this crazy stuff. But it's not exciting to say it's not Jenny. I. Because that's what the industry wants to promote today.

Leonard Lee: 34:47

Yeah, and it's interesting that you say that because there were a lot of scratch people scratching their heads saying, what are we going to do with all this additional AI compute capacity that's also handling L1 processing. It's like we just need an accelerator as a simple thing. It's like, and then what do you do with the rest of this stuff? And so nobody had an answer for what you do with the rest of the stuff, other than a very hypothetical you can make revenue. It was like, well no, this is cost, right? I, I think monetization happens layers above. You drop a, a Grace Hopper or an NDL 32. I, at the the base of your site. I mean,

Earl Lum: 35:30

I don't think that's ever gonna happen. I don't think that is a good use of resources. I think the OEMs have already figured out how to use AI in a way where they can generate a very simplistic and efficient source code to embed into whatever they're running into their equipment at the edge there. And they're not gonna need to be learning about all this stuff at the edge in real time. I think the Timo thing about I want censoring of every cell and every parameter and every cell at the edge of the network isn't realistic and is not a good use of resources.

Prakash Sangam: 36:08

So, I actually on the sideline, talk to a senior executive of an operator. Which I'll not name. He was pessimistic on the edge, a lot of edge use cases that you load up the basis of the GPU and bring a lot of use cases. He said, one, it is very, constricted environment. Anything, even when you want to add an ounce of per weight is there. You think 10 times unless there is a immediate usage, you will not add and for a very long time, just ran is going to be the only workload at the base station on the edge and server. It was very pessimistic on, putting a GPU on every on the cell, for sure. And there was also a panel attended, a roundtable attended on edge, not AI, but edge applications and edge market as such, network edge. I think how we approach the market, who will do what? I mean there are a lot of discussions still on use cases. We can do that. We could do this, but how we do it more importantly, what is that go to market strategy? I don't think there is good alignment across operators and the vendors and how to address it. So it was still in a discussion. It is still in a discussion stage in my view. Yeah,

Leonard Lee: 37:25

that's a good observation. No, it's just lots of great stuff coming out of

Prakash Sangam: 37:30

this. So one thing I think more important thing we did not discuss, which, you know, one of the roundtable, I can probably hear there on that as well on the API stuff. Right. Yeah. So there's a lot of. I mean, on the recent announcement of JV between Ericsson and, a bunch of operators on standardizing API. And obviously, I know there are a lot of, there are more questions than answers, but I think it is a good robust discussion. I think that. This has not been tried before API, but at least there is a realization across the board of undersigned operators that, we have beaten the 5G use cases horse to death. Nothing is coming out. We got to do something. And. APIs could be one option. So there is a lot of interest to make it work. But as rightly pointed out, I forgot who pointed out in the discussion that there is not many things that are API you can get from API right now. Actually, you could expose quite a lot, some of the proprietary API that I'm in a company that I work with, we can expose a lot of network things through the API. Till now, operators have been very apprehensive on how much they open up. Maybe they are a little bit more open now. They were very concerned on how much they could open up the network through APIs. Consortium, see that many people, many operators are doing it together. Maybe there'll be a little bit more open now and less restrictive. But the point is that not many things can be. Opened up the through API right now. Yeah. Yeah, the network so you can provide device Network and device situation in terms of whether it's in good courage bad coverage area that kind of is the only thing that is exposed Right now.

Leonard Lee: 39:30

Yeah, one of the interesting things I had a chat about one of the operators was that exactly what you're talking about, this initial limited number of Camara compliant API APIs, right? And he says, no, we internally have tons of them. There's a lot more to expose. Yeah, a lot more to expose. Should we expose them? And then there's the security and liability stuff that needs to be resolved through the sharing aspect of all of this, right? Or, sharing directly with a quote unquote developer or going through aggregators or brokers or exchanges or what have you. Yeah, that's a great point. Final words. What do you think we're going to be talking about next year?

Earl Lum: 40:18

I think, I think, I

Prakash Sangam: 40:20

like to finish. Yeah, I talked too much. Sorry about that.

Earl Lum: 40:25

No, I think we should have a shout out to Verizon who refuses to join the 5G Americas. And doesn't show up every year. Just join what's holding you back. If you're still upset about CDMA, that was a long time ago. So let's just have them join because we're missing a very big input into the America's market from that operator next year. Probably the same stuff, maybe more of whatever this wonderful six G thing is, or not six G is an incremental partial. Evolution of the network as some of the people out there were saying is it maybe shouldn't be another G, right? It's just. Another revision of the network and it's, rev 20, 25, rev 2026, whatever we call.

Prakash Sangam: 41:16

Yeah, sure. I'm not so sure that this not calling 60 will go with. I'm really pessimistic. What I'm thinking is some operator, bring in one, feature of even 5G advanced or whatever. See know, release 1920 and say, Hey, we are six G, and use it as a marketing and then if one calls him six G, everybody else has to follow. Otherwise they'll be, thought they're behind. Right? And the six nomenclature is too sexy if you will. to not use it by somebody, so we'll see how that goes. And then what will be next year? I think it will be more a realization of more proof points of actual non genii AI working in the networks and how it is solving problems and challenges, and maybe a clear view on. Where we stand with Gen AI. Did it solve anything? Can it solve more or all the investment? Are they worthwhile? And I think there will be more realization of Gen AI if you will scope of Gen AI, I would say and then how we are using the non Gen AI in the network, which I think there's a lot of scope. By the way, I have and also there are issues that was discussed on Collecting the data in the network itself for making gen I work is a challenge, right? So some of them pointed that out. I forgot to mention that. Hopefully those will be solved and probably, more 5G advanced and we'll see without tasted. What having 5G standalone network might be and what it can bring and so on. And then API, I think, I hope will be further along the discussion because people thinking it is crunch time. We need to do something about it. Yeah,

Leonard Lee: 43:03

yeah by the time it gets to me, everyone's already made all these are all great stuff. Insights for the listeners to contemplate a

Earl Lum: 43:13

follow up letter is where is our spectrum where is our next year? When we're back, we're going to look at what spectrum do we have what is the FCC's policy on this bold move by T Mobile on millimeter wave, and, what else do we have to bring to the table in terms of spectrum when we come back? Or is there anything?

Leonard Lee: 43:37

Yeah, maybe, hey, maybe we will see the first conversation next year about how do we make the most out of the spectrum we already have, but more importantly, how do we help operators and get to a place where they can do that? And especially by making the transition from, toward the standalone and doing it in a way that that addresses a lot of the concerns that are constantly brought up about making that transition from cost to, quality of service in some cases. Some of our analyst friends have, they, they've switched off their 5G, right? And helping consumers, but also other stakeholders understand clearly and without hype what the value proposition of 5G is once you switch over to a 5G advance. And then, setting a much more clear criteria for operators. In general, and I think this really falls on the vendors shoulders to have, give them some real concrete, non hype driven, non BS recommendations on, how to make this transition. In a way that is valuable and that's probably gonna be the toughest thing, right? And that's a tall order, but we'll see, I think we've been hoping for that for several 5G Americas, but with that. Yeah let's wrap up here. But I just wanna say you guys are awesome. Thanks for jumping on and sharing your perspective. It always ends up being a great debate and this is a debate and I'm glad that Prakash and I were able to provide a little bit of drama to our But it's always good stuff, right? And it's not fun agreeing with everything everybody says boring. It is boring when that happens. So let's, I want to highlight to our audience, these 2 gentlemen, reach out to them. There's a lot of unpublished, unmentioned insights that you can tap into through their research, but more importantly, through engagement with talk to your analysts and our EGL wireless. Earl, I respect them tremendously. Prakash as well. Both of these gentlemen really know their stuff and that's why time and time again they're on my podcast because I learned from them they enhanced my point of view and You know You can benefit from them in the same capacity. So gentlemen, thank you so much Really quickly tell our audience how they can get in touch with you and then Let's put a wrap on this recap 5G Americas Analyst Summit 2024. Ernie, go ahead.

Earl Lum: 46:24

All right. Well, you can find me on LinkedIn. The website is ejlwireless. com. There it is. Nice and simple.

Prakash Sangam: 46:31

Yeah, so you can find all of my stuff on my website tantranalyst. com and on LinkedIn. A quick plug here. I'm releasing a report on role of AI in wireless s next week or so, and there's a webinar as well that you can sign on the website.

Leonard Lee: 46:47

Alright,

Prakash Sangam: 46:48

for the opportunity a

Leonard Lee: 46:49

awesome and thanks for everyone, for to everyone for listening in. We really appreciate your listenership. Please subscribe to our podcast, which is featured on the Next Curve YouTube channel. Check out the audio version on bus bra or find us on your favorite podcast platform. Also subscribe to the next curve research portal at www. next curve. com for the tech and industry insights that matter. We will see you next time. Thank you, gentlemen. Thanks, Larry.