5G Factor: AI Rules at 5G Americas; T-Mobile, Verizon, Nokia Make Key Moves – EP76

Hello and welcome everyone to the 5G Factor. I'm Ron Westpa, research director here at the Futuring Group, and I'm joined here today by my distinguished colleague, Tom Hollingsworth, the networking nerd and event lead at Tech Field Day here at the Futuring Group. In fact, I believe we've are coming off a successful string of tech field days, and we'll touch on that in a moment.

And, but first, Tom, welcome back to the 5G Factor. Once again, I trust, uh, you've been bearing up since our last episode pretty decently. Yeah, I have, uh, spooky season is almost over, but I'm on the lookout for all of those things that are still hiding out there.

Um, you know, it's, uh, a world of security breaches and all kinds of other stuff, so I'm kind of excited to be talking about 5G and some cool things for once. Right on. And I think nobody's gonna be surprised if we inject AI into our conversation about 5G.

And before we dive right into, uh, the 5G Factor, I just wanna let folks know, speaking of Tech Meal Day, there's an upcoming Networking Field Day event on November 6th and seventh with a host of innovative maverick companies, at least I believe so, including Arista Meter, I Potential Pass Solutions, elicit and, uh, vs. And Tom, do you have anything to add to the upcoming networking field Day A we'll be participating in early November? I bet you you're gonna hear some topics around ai.

A couple of these companies are using it in their platforms, and a couple of them are using their platform to help enable it. com and, uh, set your calendars because we have the presentation schedule lined up. You might even hear the, uh, dulcet tones of Ron's voice in the audience.

Well, thanks for the shout out, Tom. I'm confident we'll be hearing your voice as you know, the event leads. So this is all, I think kismet, and that's also I think a great segue into, you know, diving into 5G Matters.

After all, it's all about the ecosystem, and they're all interrelated. And recently I was at the annual 5G Americas Analyst event, and lo and behold, AI was the hottest topic. Nobody's surprised.

But what I think is gonna be interesting is what's going on in terms of what the operators are thinking about potentially in terms of how they can play an integral role in the evolution of AI and hopefully monetizing it to customers out there, both consumers and naturally, uh, enterprises. And, uh, to start off, T-Mobile is a member of 5G Americas and was a participant there. And, uh, they also certainly s spotlighted the fact that they're using AI technology as well as billions of data points to determine exactly how to upgrade and expand its network.

And basically it's an effort that could be described as customer driven coverage, and it's a strategy that T-Mobile has put into action after developing it for more than a year. So already, uh, we're al almost about two years into this, you know, whole advent of the AI era as it could be aptly described, I believe. And so we're seeing, you know, T-Mobile as a player, as an operator specifically, that is implementing this to really determine how their network is going to be built.

That is, they're taking new AI capabilities and just making it better. After all, many of these companies have been using AI to some degree, at least machine learning. Uh, and now we're, we're seeing is I think, a taking off of these capabilities.

And to dive down a bit more, T-Mobile is correlating these data points with business data and with real customer outcomes. So this is really about the bottom line. Now, this is connecting, okay, here's ai, the technology and the capabilities, and how is it going to do just that, improve business outcomes as well as customer experience.

In other words, we're seeing that the customer is integral to how they're going to build their network. And how's that coming about? Well, T-Mobile is assigning a customer lifetime value, or CLV to a grid across the country.

And now that's more than 4 million hexagons as they describe it. That has been created across the country here in the us. And these 165 meter wide hexagons, T-Mobile, is using them to assign those values relative to the competition to allow to know exactly where the company could build to well satisfy the customers naturally, but also to expand its presence when Mindshare, when market share.

Of course now T-Mobile has tens of thousands of future projects that get ranked based on some practical issues that we're all familiar with or certainly on in the operator space, zoning permitting. And a lot of that's based on outputs from its AI driven algorithm model now. And again, it's about this customer driven coverage that I believe can enable them to stay ahead of the demand curve at the very least, let alone maintain a competitive edge.

Also, T-Mobile's not just using AI on the algorithmic side, but also on the Gen AI side. And what I thought has been, um, somewhat underappreciated is that, um, uh, across all of these AI announcements by T-Mobile, but also by all the operators, is the fact that they are engaged in aji AI allowance, uh, alliance with open AI that, you know, basically combines, you know, its ex expertise in cultivating customer relationships alongside open AI's AI technology, both the knowledge as well as the research and development experts to basically custom build an innovative intent driven AI decisioning platform. And for now they are calling it intent cx or intent customer experience.

Now, I think this is significant because with the secure access to T-Mobile data and the ability to comprehend customer intent and sentiment in real time, and this capability will be available starting in 2025, and TED CX will, I think have the ability to apply meaningful understanding and knowledge of the customer to everyday interactions now. So it's not just about, you know, customer support, it's about the entire experience whenever you have to use the T-Mobile network or interact with T-Mobile at any level potentially. And that I think is something that is going to be a difference maker.

T-Mobile's proactive approach to ai, I think has been a bit more creative. And as a result, I think it can allow to maintain its competitive edge against major rivals at and t and Verizon at least he be judged by the most recent Q3 results that have come out. And with that, Tom, what do you see that is this, you know, making a difference That's, you know, even potentially groundbreaking in terms of how operators can use AI well, uh, to make a big difference.

It's the data that's the part that I don't think they're really getting yet. And T-Mobile finally figured it out. They have access to all the data they could ever need.

They have all these handsets that are out there sending back telemetry and helping them understand coverage patterns and growth patterns and travel patterns. And up till now, they really haven't been doing anything with it. Because the problem is, is it's a massive amount of data and what am I supposed to do with it?

Well, you're supposed to sift through it. That's what we've always done with analytics and having an AI platform that can go out there and can say, okay, we've noticed these kinds of uses patterns, or we've noticed that when these things happen, this happens, that is hugely valuable. But another thing that, that they've done that i, I really wish other people would do is they've broken it down into those hexagons.

They're not taking these wide coverage areas of a tower and saying, oh, you know, we've got coverage over here and we've got coverage over here. If you're an enterprise wifi person, you know what this feels like? Oh, well, we're just gonna put an access point right here, and as long as everything's green, we're good.

Right? No, no, we're not. Because at the edges of those cells is when weirdness starts happening, right?

And even if it looks like you've got good coverage, what happens when your handset's trapped going back and forth? How do I improve that, um, capability? How do I tell my handset?

You need to start being more selective about sticking to a tower and maybe dropping when your noise floor hits at a higher threshold. And that's what AI will do, is it will analyze all of these data points in these areas, you know, whether they're directly underneath the tower or at the fringe, and then they'll push those recommendations back to T-Mobile and it'll say, okay, for these areas, you need to change your thresholds for these other areas. It looks like you're getting a big growth pattern in this direction, and you need to be able to build more towers in this area in order to cover all those potential users.

And when you think about the way that these companies market their networks, you know, I, I remember a time when having a cell phone was like, oh, well, you get coverage along this interstate 'cause that's where our towers are, but as soon as you get off of there, good luck because we don't have really have any coverage now. People expect to have coverage everywhere. It is an odd situation when you don't, and for T-Mobile to basically kind of have a chance to move into that number two spot, or at least contend to be in that number two spot, they've got to have better coverage and happier customers.

But going along with that, a lot of the ways that companies have historically used this data is they've relied on their customers to tell them when something's not working right. That doesn't always work. I mean, one thing is people tend to bias us in the wrong direction.

If they're having a bad day or they're having a particularly bad incident, they're gonna give it the lowest possible rating, even though the call completed and you could kind of hear what was going on the other side. Likewise, they tend to over bias to being more permissive when they really shouldn't be. It's like, oh, well, I heard like robot voice once, but how bad of a deal can that be?

So by allowing the handsets to kind of give them backup data on this and being able to pull, uh, quantitative analysis off of it, what they're really allowing is an unbiased opinion. And the, the consumer may not understand how valuable that is, but they'll understand how valuable it can be later on when T-Mobile puts the right assets in place to make their experience a whole lot better and the customers didn't have to do anything to get there. Yeah, I think that is succinctly put.

I I couldn't agree more, Tom. In fact, I think it's showing that T-Mobile is being smart about that, uh, how to really leverage AI and quite simply make themselves a more intelligent carrier. And, uh, we didn't even touch on the recently formed AI ran Alliance along with Nvidia naturally as well as Ericsson in Nokia.

But the idea is the same principles. How can we even further optimize RAN capabilities using AI at, you know, the cell sites themselves or at where the rans are distributed and so forth, and combining AI computing with RAN optimization. And so state two, that has not actually been something that has a tangible timeline like we touched on with these other AI initiatives.

However, I think it's just indicating this is the smart thinking that needs to take place for the operators to really, I know, recreate how they can be much more meaningful to not only their customers, but across society. I think that's a great point that now people expect connectivity, at least mobile connectivity, and that's something that is on par with, you know, utilities and water and so forth. And with that, I think it's a great way to now touch on now what are the operators going to do to really take advantage of not just making their networks better, the customer experience better, but how can they potentially monetize serving up ai?

So let's shift a little bit and let's look at what Verizon came out just within the last, uh, couple of days. And what we saw is that the Verizon, uh, executives said that they're planning to profit from the sale of computing infrastructure for AI operations. But this is just, you know, committing to the principle of it that, uh, they haven't come out with details yet, but I think we can bet that they are leaning toward coming out those details in, uh, 25 at least, or at least they really need to.

And so what Verizon has done is it indicated it's getting a lot of good orders from, you know, the hyperscalers or it's dark fiber or, uh, lip fiber, and it's going to, you know, simply keep growing. Now, what Verizon's believing is that more than that, it's not just about the fiber, but also its assets in terms of power space and cooling resources, which is and really high demand combined with, you know, having the resources in the right, uh, places to really, uh, assist with things like latency and other key considerations when it comes to AI workload performance or AI experience, uh, optimization. And I think this is, uh, also important because we're seeing the hyperscalers are already reaching these critical power consumption and energy demand levels that they're investing in modular nuclear energy that they're investing in, you know, re uh, uh, firing up nuclear plants, three mile island, we saw that.

So this is something that I think is also going to impact the operators in terms of how they can also get into this AI game. Now, what I think is important though is that they're not going to compete directly against the hyperscalers when it comes to heavy duty, large language model, uh, training at, you know, the, uh, major data centers. And that is, you know, where the GPU clusters are heavily concentrated right now, pulling up all this energy demand, and also, well quite simply enabling that these AI capabilities will be in place for where the next shoe will very likely drop.

And that is on the AI inferencing side, that is enabling the resources at the edge throughout the edge to take advantage of the large language models and other models that have been trained to allow customers or organizations to get intelligent AI well interactions or what we saw just with, with the T-Mobile example. And so what I think is going to be interesting is that this is kind of the back to the future scenario for the operators, where the business case and monetization of AI inferencing can provide the warrant for the service providers to go ahead and take advantage of all these edge resources that not even the cloud providers have that is, you know, central offices and other, you know, um, hubs where they have equipment already that can potentially host computing at, uh, much closer to the customers. And that I think, uh, can drive offerings such as AI as a service, which was, you know, specifically invos with the ai, uh, ran alliance, uh, debut.

Now, telecom infrastructure can play a role in the future of ai, I believe, because as consumers and businesses use the technology, it's going to require driving AI processing a great deal outside of those massive data centers and quite simply close to the user, including, including those devices that, you know, are always, uh, accessible to, you know, the folks out there using a mobile network. Now, however, the idea of operators selling computing resources is not new. We've been here before.

In fact, we've seen that operators including Verizon, exited that model when they sold their data center businesses to companies such as Equinix about a decade ago. But what's different here is that they can be, I think, more intelligent or smart about how they distribute the AI computing resources. It doesn't have to necessarily be newly built data centers, but using ran locations and using other parts, other, uh, already existing resources to make this happen.

And Tom, well, I know that's a lot, but from your perspective, do you think that operators are up to it? Can they return to the future and start selling a high computing resources across their vastly distributed edge resources? Yes, not only do I think they can, I think they should, and they may be the only ones that ca are capable of doing it right now, there was a stat that came outta the Open Compute Summit last week that said that data center power budgets are expected to triple in the next five years.

Where are we gonna get all that power? You mentioned restarting three mile Island, there's, uh, research being done to small scale nuclear reactors and the kinds of things that can give us, you know, dozens or even hundreds of megawatts of power in order to run these things. But companies like Verizon and others have resources that are already positioned out there.

And like you said, central office, uh, that's kind of what that whole edge computing thing was all about, was pre positioning these resources close to the edge where they can get good response times. And I think that one of the things that Verizon really wants to focus on here is that they have more compute resources closer to the people that need them. And like you said, they're not gonna be running these gigantic Blackwell water cooled systems that are consuming dozens of kilowatts of power with every cycle.

They're gonna be doing things that are a little bit more focused. And that's really where Verizon's data comes in handy, is they know what their customers want, they know what their customers are using. So rather than trying to boil the ocean and create new algorithms and new LLMs that allow their customers to come up with these grand new ideas, they can literally just offer the things that their customers would've normally wanted in the first place.

Things like voice transcription, right? How many times have we seen the value of being able to turn on closed captioning on something? Well, what if you could do that live?

Well, that's not something I necessarily wanna do at a central location because every added millisecond of time creates lag in that call. So if it's something that I can do on the edge where the, you know, the, the pop is the one doing all the heavy lifting before it gets sent on through the network, that's a huge benefit for me. And, and those, that's just one example of the things that Verizon's wanting to do, because building out these massive, like Equinix style data centers is gonna take time.

You can't just drop a couple of extension cords in there and, and Bob's your uncle. You've gotta put in massive new resources, massive new cooling capabilities. And if Verizon can offer a stop gap for people that need to ramp up quickly, then that's valuable for them.

And you, you talked about the fact that 10 years ago they sold off all their data center assets. 10 years ago we thought the cloud was the way to go, right? Like we, everything was gonna be based in Corvallis and Reston and, and, uh, that was that, right?

We, we didn't need all of these other data center assets. Colos were, you know, on death's door. I can remember that cloud's gonna take my job.

And it wasn't until AI became a thing 18 months ago that people really started looking at the possibility of needing to host their own stuff again. And so I think Verizon is really kind of, they, they've picked that up and they were already headed in the edge computing direction to begin with. They were, I think it was more at the time, a, a solution in search of a real problem to utilize it.

I think AI is that problem, for lack of a better term. And, and the resources are available and ready to be consumed if you're intelligent about how you consume them. And you're not gonna be dropping these heavy LLMs on things.

You're probably, what you're gonna be doing is offering this like a cluster to tenants who wanna be able to distribute this workload, maybe run it overnight and get the results back the next day, as opposed to, I have to have this in the next hour. Right on. Yeah.

I think, uh, that is something that is going to happen. I I think the operators are going to already have the plans in place, but make these steps to figure out how they can quite simply take advantage of it. And they, they need to, because as we know, uh, the 5G space hasn't exactly been a, a huge revenue generator for the operators.

And now with ai, it cannot only fire up, you know, 5G and, you know, the next iterations of 5G as well as six G in terms of revenue, diversification and potential, but also any connectivity, fiber and so forth. All of these assets can take advantage of this AI potentiality. And I think this is something that is going to create new competitive dynamics.

And it's, it's just an interesting time to be watching all this. And I think, uh, this is also bringing to mind, as impressive as AI has been in some areas in terms of, you know, the large language model training and the capabilities that, uh, these GPU clusters could enable. And at the edge we can use CPU clusters.

In fact, NVIDIA offers CPU clusters. People sometimes might not know that. But I think it's interesting that when it comes to the edge, this is where we can see, again, a small language model approach and these other rightsized approaches that the operators can sell, uh, to, uh, the customers out there.

But what, what I was getting to in terms of, you know, what is limiting somewhat AI's capabilities is the fact that when you're looking at AI workload performance, it is slowed or delayed by 30% by the network. I think we've seen, uh, data points to that effect and similar ones. And so while we might have some very impressive heavy duty performance going on inside the data center, it's also the network within the data center as well as across the YAN and other parts that will quite simply are gonna be important to, you know, take full advantage of these AI capabilities.

And so this is where I think AI or gene ai, uh, specifically can make a difference. And that is in the area of advancing network automation. I know we just recently had a, an exclusive on tech field day with Nokia that focus on this very vital issue.

And I think we're seeing solutions and advances that can make a big difference here. So in essence, when we are looking at ai, it's about semantics and data, and which together can bring about new ways to express disruptive insights, sharp actions and so forth. And so what I think is going to happen is that semantics is about picking up everything that can be said amongst stakeholders.

That is certainly amongst, uh, network planners, network engineers, as well as the other teams that are going to be important in terms of making the network itself better perform, uh, capable, uh, that is, um, uh, all the more automated. And there have, there are existing barriers out there that are significant. There's been some incremental progress, but I think this is where we can see this, uh, making a catalyzing difference that is using gen AI natural language prompts to, you know, have a tent based networking become basically, uh, almost a norm, if you will.

And so we're seeing a lot of research that is being conducted around the idea of semantic spaces. And what this is linking to is that, uh, it's enabling, uh, what can be quite simply described in terms of network automation, the autonomous network. And as we've seen with the t uh, informs vision of the autonomous network, it's really the path to 0, 8, 0 touch, zero trouble, et cetera.

That is, that is, you know, a network that performs just as simply and as easily as, say, a local ATM machine except rip large. And we know that it's not gonna happen overnight, but I think this is an interesting approach, an interesting solution that can make a difference. That is, you know, through network automation, all the more impressive SLAs can be rolled out and fulfilled because of these AI workloads and all these other workloads that are also, uh, you know, expanding dramatically.

And so what we're seeing is, uh, vendors like Nokia coming out, um, with how they can make this happen. And already we see existing solutions such as the network, um, services platform, NSP bin Nokia offering these eight automation capabilities already being used by over a thousand network operators out there. So this is something that's in place and I believe can make a difference in terms of enabling that end-to-end IP network that allows the, um, automation to be extended throughout the network, across the customer's premises over aggregation and ex hole domains, as well as gateways of various kinds.

So this is something that's important. It has to be organization wide, and it has to be something that is enduring, that is something that is going to, uh, enable just that the, uh, uh, entire automation of the network, not just portions of it or at the individual level. Now, the good news is, is that the addition of gen AI capabilities allow the Nokia network services platform to correlate these data streams that I touched on in new ways, because it can provide reasoning that helps the platform understand what it sees, discover new relations across the networks, and identify root causes of event getting closer and closer to that idea of an attempt driven network.

Ultimately, I believe we're getting closer to using gen AI enabled platforms such as Nokia, NSP, that can act as an assistant to the customer's, network designers, planners, engineers, operations staff, and so forth. And quite simply use the platform language model towards a specific service provider organization, network language model that is exclusive to them and enable these, you know, possibilities. And so, Tom, with that, uh, from your view, I, I know that these barriers are real, but do you think this type of capability, this type of approach, could actually move the needle more in terms of making network automation happen?

Yes, and part of the reason why we run into the network automation problems that we have, it comes courtesy of, of our friend Mike Boong at Nokia, who was talking about this during that exclusive event where he said, you know, a lot of people have these concepts of the way that things just should be done, right? And so they go out and they, they try to automate, you know, uh, data entry tasks, you know, the, the easy things. I, I don't want to have to check this thing to program of VLAN or whatever.

And they eventually hit a barrier because once they've solved all the easy problems, the only thing left are hard problems. And you, you kind of have to pick the hard problem that you want to tackle. And eventually people just kind of give up, think about the way that some systems are basically forced to automation.

Now, an API is nice, but what if you don't have one? Well, you have to, the system has to log in, dump the config, analyze the config, figure out what needs to change. It has to go back in, it has to rewrite the config.

It's basically doing human things super fast without human interaction. And you're probably sitting there thinking to yourself, well, that's dumb. Well, it is dumb.

And the reason why is because the, the first step that we took for that is an an API an application programming interface. You know, it would be like if I was gonna paste something into a Word document, like I have to pull up the thing over here and I have to copy the text out and I have to paste it back over here and I have to change formatting, you don't have to do that, right? Like, you can just hit share and insert it into the document.

And that's an API, right? Is we're sending information to another program through a call and it shows up over there. That's how humans interact with it.

What Nokia is proposing is they want to extend that, they wanna create more things that allow the opera, the automation operations to run even faster. So that as we enable these things, as we create these new languages to do that or create new metadata, that allows us to understand it, that the automations don't need human interaction. It goes back to that whole thing of, you know, think about using, uh, a Rube Goldberg machine to automate a door lock, right?

You know, like I have the bar that opens it up and closes it up and things like that. Well, the next phase is to have a door lock that doesn't have a, a handle on it, because if it's all automated, why do I put a handle there other than for emergencies? And so we're, we're gonna slowly start removing these capabilities that we don't need anymore, like screen scraping, and we're gonna get to a point where the systems can run on their own and it works better that way.

Like the idea of having an operating system that I don't have console access to scares me. But there are a lot of people that don't know there's a terminal window in Mac OS or Windows or anywhere else because they're just used to the operating system working the way that it operates. Or, you know, moving into cloud networking.

Why do I ever need a console into this switch? I can do everything through the web. I can click here and program this stuff and take care of that.

And once I know how to do that, then I can have a system replicate that for me. I can feed it scripts, I can give it intent. And that's the whole purpose behind intent-based networking is not to remember, oh, this is a Juniper switch.

I need to use geno syntax. It's to say, I would like to configure this application to use these ports and these networks and do this thing, and then the system figures out how to do that for you. And that's true automation in my mind.

It's not the making my typing day a little bit easier type of thing. It's me making a decision based on business process that executes technical capabilities on the back end. That's how cloud functions, right?

Like, nobody goes in and you're like, okay, what VLAN am I supposed to use? No. They say, I need to deploy this application to support this many users in this location, and then the rest of it gets taken care of in the background.

And they don't know that. And I think that that's one of the reasons why using Gen AI to do these things is so important now because for, for better or for worse, AI is this magical idea that, that things can just happen and I don't need to deal with it. And you just, you know, you've seen the, the asking Chad GPT to give you summaries of things and stuff like that.

It just works well, if we can program the gen AI things to make other stuff just work, then we get over those humps that we get into and these automation journeys where people get stuck on problems that shouldn't be as hard to solve as they are, but that it, it ends up derailing the whole project. So I applaud Nokia for doing this. I just, I hope that there's more traction for this beyond just their operating platforms and their operating systems.

I hope that they can bring this to market enough that other people start to embrace it and that it really becomes a game changer. Yeah, no, I think those are very important points. I think one thing that's a difference maker here, it's using an existing platform applying gen AI capabilities, and it's already distributed a box a thousand plus operators.

So that I think, can in itself be a difference maker. But I think another key takeaway that was, uh, emphasized, or it was talked about and, uh, a good deal at 5G Americas was the fact that because AI is such a strategic business priority for almost all organizations, that there's a broadening in the decision making. And that is, hey, we see, you know, CXOs and other key decision makers saying, well, gen AI is making a difference in things like, you know, chat GPT props that could, you know, help summarize a say a webcast script and, and other, you know, capabilities as helping with customer service, uh, lowering coding requirements, uh, improving field tech performance.

Then, you know, why not other important areas such as network automation. And I think with Agen AI tying together all of these domain, all the, you know, uh, knowledge and data from these different domains into a natural language interface that allows the decision maker to take advantage of the capability can be just that massive breakthrough where, you know, even massive networks can be a great deal more automated and more intuitive and easier to support, uh, build and, and so forth. So now I think that is actually, you know, good news.

And, uh, with that, speaking of good news, don't forget, we have our tech field day coming up November 6th and seventh. And again, we'll be diving into how AI can even be all the better applied toward improving networking. Uh, Tom, now that, you know, we're getting, uh, closer to Halloween and all that good stuff, any thoughts to add on, you know, should AI be something that we should fear or something that we should embrace?

Uh, I, I don't fear it because I have the non conducting blade right here under my desk, so that if I have to cut it off, then I'm, I'm totally fine. I think that it just like any other tool, right? A a weed burner is a valuable tool for people who need to work outside.

It's a terrible tool for people who wanna use it as a flame thrower. Um, you've gotta figure out how to apply the AI properly in order to get the outcome that you want. And of course, the most important thing is do not ever let it become self-aware.

And, and if it ever does, then, uh, you know, uh, the Crystal Palace in, uh, in the mountains is gonna place divinity to hide out. And, uh, Ron, I'll save you a seat. Uh, shout out to, uh, the Terminator movie series.

Fair enough. Well, with that, uh, thank you everyone again for joining us again. Be sure to bookmark, uh, the 5G Factor.

It's on the Futureum Group, uh, website, as well as, uh, the Tech Field Day, uh, site. It's all interlink to the Futureum Group. And with that, thank you again, Tom and everybody have a great AI enabled 5G Day.