Thank you so much for joining us. As you guys know, Yigal is the Chief Technology Officer at AT&T. A combination of scheduling conflicts and war and airplanes and government funding have left us doing this virtually. We really appreciate AT&T making the effort to be part of our conference again this year. Thank you, Yigal, for doing this.
Thank you for having me. Good morning, everyone.
Yigal, I'm here joined on stage with my BCG counterpart, Simonas. I'll do maybe the first part, and Sim will do the second part. I guess, you know, Yigal, I know you're at an off-site right now, but, you know, one of the things that John Stankey, you know, really made a priority at AT&T was kind of slimming down and simplifying the AT&T business. I think that we heard that some of your competitors earlier today wanna follow that playbook. I guess from a technology standpoint, you know, what are the new priorities at AT&T? Is it wireless coverage? Is it wireless capacity? Is it fiber network availability? Is it fixed wireless access?
Is it, you know, slimming down the business even further through efficiency initiatives? You know, what are the kind of things that a guy in your seat is working on right now to fulfill the AT&T vision?
Yeah. Thank you, Dave. Maybe before I start answering that first question, I need to call everyone's attention to our Safe Harbor statement. I think you're seeing this on the screen. It says that some of the things that I'm gonna share today are forward-looking and as such, they're subject to risk and uncertainty, and the results may differ materially. I also want to remind everyone that we are in a quiet period of Auction 113, and as a result of this, unfortunately, I won't be able to discuss that. Now that we got that off the table, we can-
Well-
I can try to address your question.
That was a rookie mistake on my part, Yigal, and so I feel bad about me.
No, it's okay.
I'm glad we got that out of the way. Thank you.
Exactly. That's the important thing. Well, I think, Dave, obviously, I'm sure you're anticipating the answer that says it's all important. All of the characteristic of the network that you described are important. All need to be addressed, and obviously, we are addressing all of them. The way we think about this is like everything else in AT&T, is through the lens of our customers or customer-first perspective. If you ask customers, I'm sure they're gonna tell you that the first thing or the most important for them is to actually get on the Internet. So they're able to make a phone call or send a text message or do an email or whatever they need to do over the Internet. I think they also care about consistency. Is their experience over the network consistent?
Things like dropped calls. AT&T has the lowest dropped call rate in the U.S. That kind of tells you about how we think about reliability. We are all privileged here to serve first responders. The first responders of America. We're building and operating the FirstNet network. That tells you that the bar is even higher when it comes to reliability and how we think about this. Now, at the same time, you mentioned capacity. Capacity is really important with the growing demand for connectivity and AI and video all traversing the wireless network. As we're going through our wireless modernization, think about the fact that we are putting to work all of our mid-band. We are simplifying the mid-band spectrum. We are simplifying the configuration in all of our cell sites.
We are opening up the architecture so that can help us to drive efficiency up and cost down. So this is how we think about all of those characteristics and how we're serving customers. I think that it's also an interesting optimization problem. As such, AI is a great capability to help us to think through this in terms of how do you bring in different attributes as customer experience, devices, reliability, coverage, capacity, other use cases, and how that helps us to guide our investment and how we're thinking about proportionally investing in the various capabilities that you mentioned.
Thank you for that, Yigal. Those are the kind of the what's, but I'm interested in a little bit on the how's. You know, I think that one of the bigger events that's occurred with AT&T, and there's been a couple, one big one was the acquisition of the fiber asset from Lumen. Another was the acquisition of Spectrum from EchoStar. I think that there's a lot of question about how these pieces fit together. The fiber, the Fixed Wireless Access, and the wireless network, and which one takes priority. I'm assuming your answer is gonna be everything takes priority. But how, what is the real answer and how do you come to that answer in terms of what you decide to spend your time and energy on?
Yeah. Great question, Dave. Maybe zooming out a little bit to where you started. What we're actually doing is building two scaled network platform, right? We have a modern, open, nationwide, 5G wireless network, and we're also building a metropolitan multi-gig fiber network that is already covering 36 million living units, which no one else has that size of a fiber network and no one is building at the pace that we're building our fiber network. That leads to the true benefits that are two in my mind. One is owner economics. Our ability to build and operate both a wireless and a fiber network gives us a lot of flexibility in owner economics. The other thing is actually supporting our overall strategy, which is all around convergence.
The way we're thinking about this is that we have a converged network strategy. One of the things we like to say about our network is that we're actually building a fiber network that has different access technology hanging off it. Think about the ability for us to optimize our fiber routes and building in a way that supports three or more distinct use cases, consumer broadband, wireless backhauling, enterprise services. The magic doesn't stop there. Historically, all of those services, the way they would converge in our core network was deep in our core backbone. They reroute or traversed our technical spaces or central offices through different architecture and different infrastructure. We are changing this through our network modernization. We are pushing out all of that convergence to the edge of our network, to each one of our central offices.
Now you can imagine how each one of our services is actually running on the same converged and intelligent architecture and infrastructure. Now let's talk about fixed wireless because you brought this up. Clearly, when you have fiber assets in 32 markets, that will always be our lead offer. This is what we wanna take to our customer because I don't think anyone's going to argue with the fact that fiber is the best broadband technology. It's symmetrical, it's reliable, it's doable. However, there are different scenarios in which either we don't have fiber assets, or we know that we're gonna come into a new neighborhood or a new area, but that's going to take us a year or two to build our fiber then, or we need a catch product to get people off our corporate infrastructure, small-medium businesses.
There's an interesting segment of customers that would prefer to go with other type of services. For all of that, we have our fixed wireless service, which is a great product addressing all of the use cases that I just mentioned. You mentioned the fact that we acquired more spectrum. In a matter of weeks, we were able to push out through a short-term spectrum management lease all of the additional 3.45 throughout our network through 23,000 cell site. That added capacity, that opened up a lot of new geographies and created a more addressable market for fixed wireless capabilities.
There's been a conversation here a little bit today, some mixed feelings about the enterprise market opportunity. You know, as the largest enterprise provider in the United States today, this AI evolution from a B2B standpoint, you know, I'm an old guy. I watched the internet lead to tears with millions of bankruptcies. I saw the cloud evolve, and no one made money except Google and Amazon and Microsoft and Oracle. Now we're talking about this new AI internet that's supposed to be coming into being, and there's been some players who've been noisy about it. I would say that AT&T and others have been more quiet about it. Is it a real thing?
Is it a construction project that's a one-off, or is it, something that's real?
First of all, I think we're living in the most remarkable period, technology period that I can remember, and it's really exciting to be part of it. To your question, I will probably take this further out a little bit than just the enterprise space. I will tell you that I don't think we're sitting here today and seeing a completely different about how the network is operating or what kind of AI workloads we're seeing traversing. We're absolutely seeing signals. What I mean by this is, all of us are using AI capabilities on our smartphones. We may start wearing all kinds of AI devices. The characteristics of those applications and devices is a little bit different. They are heavier on what we call the upstream side of the network.
We have a couple of examples of autonomous cars or robotaxis that driving over our network. For sure, in the training phase, we're seeing the ratio of upstream be above 50%, which is something that we never saw before. We are absolutely seeing growing demand for high speed, low latency, services, 400G with the enterprise space. Absolutely seeing this happening. If you think about agents, I'm sure everyone heard about agentic AI, and with the proliferation of agents, how this is gonna start behaving over the network. They don't work 8:00 AM to 5: PM, they work 24/7. We don't know at what time of the day or the night they will operate. The traffic might be east-west, which is different than how traffic is behaving today.
I think there's all kind of exciting capabilities and demand that is starting to show up. Think about physical AI, autonomous cars, drones, humanoid robots, all kind of other capabilities. They're all gonna drive for different capabilities and characteristics from the network. What I am very pleased with is the fact that I believe we have all of the assets that allows us to be AI-ready or build an AI-ready network. We have the right holding of low band, for sure with the aid of the 600 MHz, which is very useful for upstream services over the wireless network. If you move to an open architecture, our ability to adapt in real time to different usage characteristic or patterns in real time is something that we're gonna get better over time.
Our symmetric broadband with our fiber services to people's home or offices is gonna be extremely important when you think about all of those upstream usages. Our ability and we're continuing investment in building high speeds and low latency capabilities in each one of our metros, bringing this up to 400 G and above this over time, is going to be critical for many of those enterprises to connect with hyperscalers. We already have fiber connection to 600 data centers in the U.S. That gives our customers the ability to see and run their AI workloads or capabilities in different locations. They just need us to be there with providing them high-performing networking, whether it's high speed, high capacity, and low latency. All in all, I think we're in a very exciting time.
Maybe another example to see how convergence is coming to life. I talked about those physical AI devices or our capabilities. Many of them may demand what we call deterministic network. We are uniquely positioned to provide a deterministic experience all the way from the device over the wireless network, through the transport, and follow the context for AI models no matter where they are. As I said, I think we have all of the assets in place, and we are uniquely positioned to really be ready for this, AI-ready, era.
Building on that, you've mentioned a lot about your capabilities facing outward. Internally, how have you used AI in the network operation space, and what kind of advantages are you seeing versus the complexity that has been present in the past?
First of all, I think we're all privileged to be in a company that actually established a Chief Data Office more than a decade ago. We went through big data to machine learning and AI, and obviously, when the new capabilities of generative AI showed up three years ago, the team was able to be one of the early movers to help us to adopt those technologies at scale across the companies. We all have access to capabilities, whether it's in software development, whether it's in day-to-day work. We all have access to the latest models and the latest capabilities. Across the company, no matter if it's in customer service or fraud detection or in finance or in other areas, capabilities are being built to the benefit of operating our company.
On the network side, we've been building machine learning and AI capabilities for years in the network space. No doubt that two things help us to accelerate this. One, obviously the advancement in AI capabilities. The other part is our move into an open architecture. It allows us to have more structured data model that's flowing up into models, and it helps us to take action, what we call in our terminology, closed loop, in an easier way, follow intent, and being able to build modules, software modules and AI module that helps us to operate our network, whether it's energy savings, which again, we've been doing for years, like what we call cell site sleep, where you take down carriers when you don't need them in order to save.
The ability to use newer model help us to increase this by 20% or 30% the efficiency of those models. Self-optimized networks. We build what I would consider probably the most advanced 3D ray-tracing model, which allows us to really have a model of all of the propagation of all of our cell site. Think about the disruption in our network, whether it's a fiber cut or a weather event. The ability of the network in very close to real time to readapt the cells and the ability of our network to readjust itself so there's no customer impact has been increased phenomenally through, again, the advancement of our models.
We are using this across the board, mainly in the RAN space because clearly this is the area that we are spending the majority of our capital, so we've been engaging on those models and capabilities for years. Again, we were able to take advantage through very unique skills that we have internally and the newer model to actually improve significantly existing models and create new ones to the benefit of our network.
You mentioned that you're building an AI-ready network. As we look forward into the future, there's conversations around 6G enabling AI moving to the edge. How is that, if at all, changing your thinking for the longer term?
Well, that was the conversation of the day at Mobile World Congress. I don't know how many of you are attending there. Here's how I think about this because I think it's an important question. I wanna unpack a couple of those conversations. First of all, about 6G. I think the cycle of Gs over the years was instrumental to the success of the wireless industry and the ability to connect 7 million or 8 billion people around the world to the Internet. However, we just described the fact that the advancement we're seeing in AI.
In a world in which every month or two a new language model or foundational model is getting dropped at us, I don't think we as an industry can talk about what we think our industry is gonna enable in 2030. That doesn't work with each other. If you think about the state of the architecture of wireless today, it's already based on software, it's based on openness, AI, cloud native infrastructure. None of those technology domains live in five- or 10-year cycles. I think the first thing we need to do as an industry is completely decouple the Gs or the cycle of the Gs and our ability to innovate. We need to move to a continuous innovation and continuous progress in being able to consume new capabilities and innovation as it becomes available.
That's the first thing that we have to do as an industry, and that's part of our Open RAN strategy. That is exactly what we're doing. To the second part of your question about how far is the edge, which I think that is the question, because there's no doubt that AI is gonna be embedded into wireless network and, we're gonna call it AI native and combining the the physical space with the intelligence of the network. This is all true. The question is, one, the moment we have a software layer, the moment we move away from purpose-built baseband or the infrastructure we have at the bottom of the tower to a compute-based platform and software, we disaggregate it both.
Over time, we have full flexibility to decide what is the right compute that we wanna use at that cell site to serve our customers, to support our spectrum or whatever we wanna do in terms of innovation. The second part of this is how far is the edge and how do you wanna serve some of the use cases with compute and intelligence? My view on this is probably a little bit different than what you may have heard, otherwise. I think in the U.S., where I think the number is just this year, $600 billion or $650 billion of investment just in data centers, just in 2026. I think the proliferation of compute and high-performing compute across the nation in all metros is just happening with a software layer on top of this, with the tools that developers need.
I am not sure that there's much value in extending that compute all the way to the far edge just to save another millisecond of latency or 2 ms of latency. I think the U.S. perspective on this is a little bit different. We wanna take advantage of our nationwide modern wireless network, our deep fiber build and being able to create that deterministic experience between whatever use cases comes and help them to intelligently connect to the right model that they use, the context or the infrastructure that they need, because that's gonna be heavily distributed across the U.S. I think the U.S. is a little bit different from other countries and other perspective just because of how deep our edge compute is being built.
Shifting gears for the last topic. We've talked about the terrestrial networks, thinking about the stars and the sky. What is the role of satellite, and how do you view Starlink as a potential competitor for your terrestrial network and the D2C capabilities, and how do you compete with those as well?
I think, first of all, this D2C as a technology evolution in our industry is one of the most important innovation that we're seeing. I still remember the day that the folks from AST SpaceMobile walked into our office seven, eight years ago with that concept that we're gonna have an antenna in the sky, and it's gonna work on our regular phones. Yes, some of us rolls our eyes and say, "What? Really?" Here we are. It's actually working. It is from our perspective. It is absolutely a complementary service and capabilities to the terrestrial network. I think it has a place. We have the largest wireless network in the U.S., but still we cannot cover every square inch of America.
Therefore, our ability to provide our customers with service no matter where they are, whether they are hiking in Grand Canyon or doing something else off the grid, is a great opportunity for us in terms of our strategy around seamless connectivity and enable connectivity to our customers no matter where they are. Is Starlink a competitive to terrestrial network? First of all, we are partnering with AST SpaceMobile, and we think the technology that they're gonna bring to life and the way they're building their antenna in the skies is very unique and we have a lot of trust in their architecture. I don't see any time soon in terms of the beam creation or indoor experience, I don't see how D2C from the sky becomes a true competitive offering. It is absolutely.
I don't think I'm telling you something you haven't heard. This is absolutely something that we in the industry are looking as a complementary capabilities that is needed, and it's a great innovation and that's gonna provide a lot of value and benefits to our customers.
Well, thank you, Yigal. I think those are the topics we wanted to discuss with you today. Really appreciate your time, and it's great speaking with you again.
Thank you so much for having me.