YOU ARE AT:CarriersAT&T at MWC: Open RAN, AI and more

AT&T at MWC: Open RAN, AI and more

AT&T network CTO Yigal Elbaz on AI, Open RAN and how MWC trends from past years are playing out

RCR Wireless News caught up with Yigal Elbaz, AT&T’s SVP of technology and network services and network CTO for AT&T Services, at Mobile World Congress in Barcelona, to get the latest on what’s happening with the carrier’s network.

We asked about AT&T’s progress on its Open RAN transition, how the carrier is utilizing artificial intelligence and building a network digital twin, and Elbaz’s impressions of this year’s MWC Barcelona. Here’s what the conversation covered.

On Open RAN: “Everything is going according to plan,” Elbaz said. Per previously public timelines, AT&T expects to have “largely completed the modernization” of its 5G network by 2027, and to have 70% of its 5G network traffic traversing open hardware by late 2026.

Elbaz outlined the four parts of AT&T’s Open RAN transition strategy. First, the “build and swap” work happening on the network side, where Nokia radios are being swapped out and at the same time, additional midband spectrum is being deployed. Elbaz said that this involves “thousands of sites, [but] minimal impact to customers” and that most of the new hardware being deployed is “open capable.”

“We are deploying our midband spectrum while we’re doing this, cleaning up the configuration of the sites and building 3D imagery of our cell sites—creating a digital twin that is then fed into our AI-based network planning platform,” he added.

The second piece of the transition is cloud RAN. AT&T has had commercial traffic flowing over cloud RAN for more than a year, using Ericsson’s solutions, and made its first V-RAN call in early 2024. It considers the move to cloud RAN a foundation of its Open RAN transition, and Elbaz said the C-RAN deployments are continuing.

The third piece is third-party radios, including radios from Fujitsu and Mavenir as well as Ericsson. The fourth aspect is the management system. “Typically in our industry, every vendor comes with its own management system. We are changing that. We’re using one management system,” he said. “All of the vendors going to integrate the same one.” That use of one management system also presents another way to improve efficiency and introduce more AI and machine-learning.

On artificial intelligence: Elbaz emphasized that AI is not new to AT&T, and that the company has been “leveraging AI for years in various parts of the network,” including by building a number of AI tools that it has built itself.

One of the ways AT&T uses AI is in network planning, he explained. “If you think about the complexity of how we are building our wireless network and fiber planning and the amount of attributes that come into this, AI is an excellent tool to do that.” Launching of new sites heavily relies on automation and AI, he added, and in addition, AT&T has “built our own AI-based energy savings [tools] that I would argue can compete with anything out there,” which he says has enabled 30% energy savings.

The move toward open networks ties in to AT&T’s use of AI. “In order to really take advantage of genAI, you really want to have data flowing through your system,” Elbaz said, adding that as AT&T moves toward open networks, it has more data flowing in ways that can have be leveraged by AI—including through the service management and orchestration (SMO) and through rApps—at increasingly granular time intervals.

“The SMO … and the usage of rApps is an excellent example of how we can now introduce more capabilities of AI and even things that we build ourselves today,” he said, adding: “There’s a lot of advantages that come with openness, with data exposure to the system that we can collect more easily and then we can apply different models.” AT&T is also working on building its own network foundation model, he said.

On network automation: Network automation, and the goal of autonomous networks, was one of the major themes at this year’s MWC Barcelona. Elbaz said that he thinks of network autonomy using TM Forum’s network autonomy framework, which is what AT&T judges its progress against.

“We are trying to benchmark ourself constantly against that model to see where we stand,” he said—but, he went on, it gets complicated and depends on what part of the network you are considering. Legacy infrastructure represents a barrier to automation, but on the other hand, work that AT&T has done in the past several years has enabled higher levels of autonomy in some parts of the network, he said. Others will take more time. And AI is integral to automation as well.

“We are taking advantage of all of the AI knowledge that we have,” Elbaz said. “And don’t forget that today, telco is probably one of the leading enterprises when it comes to use of AI. And … we’ve been one of the first movers. We took advantage of our relationship with Microsoft and Azure and we are very quickly created our own closed garden in Azure for our data.

“Our CDO team is constantly introducing new models into the backend. … Behind the scenes, there’s a lot of intelligence [and] different models, all based on APIs that makes it easy,” he said. “Through those APIs, we can build our own capabilities that we can apply for wireless planning, for wireless intelligence and different parts of the network. … The company’s doing extremely well in that regard.”

On building a network digital twin: AT&T already has a lot of information about its networks, Elbaz said, but its aggressive work on modernization and open networks is allowing the company to do site clean-up and “significantly simplifying” sites. In many cases, he said, AT&T is flying drones around sites to do 3D imagery that is then being fed into digital models.

“You create the imagery, you create the data that comes from those systems and that starts to help you to build a digital twin that is fed into our network planning platform that we’ve been using for years,” Elbaz said. That in-house built platform, then, is used by AT&T to help decide where to add capacity, build a new site, which spectrum to deploy at a specific location, and so on.

On MWC Barcelona 2025: Two years ago, he recalled, MWC Barcelona was all about APIs; last year, the emphasis was artificial intelligence.

In regards to the use of APIs, Elbaz referenced the news from the show that the three national U.S. carriers are all going to be support Aduna, an API joint venture launched in September by Ericsson, Google Cloud and a number of global operators, as well as the Open Gateway project as signs of progress on APIs in the past 24 months since they made a splash at MWC.

“I think this industry is innovating in a place that we haven’t seen before,” he said. “I do think that our ability to monetize our network in more ways than our standard of telecom model is an exciting opportunity for all of us. And there’s power in that unity, because no one wants to build an app or write an app for one telco or one region. Everyone wants scale. … Now we need to work the demand.” But being able to aggregate APIs and provide scale across multiple carriers, regions and devices means that the API ecosystem is maturing, which will mean it becomes more attractive to developers.

When it comes to AI at MWC, he reflected, “I think there’s a lot of AI still, but all of the messages are tempered. I think it’s a lot of business as usual, [with] some level of maturity, which I think is actually good. We don’t just use buzz to get excited. We all know that there’s focus happening. … I think what you’re seeing is that the industry is saying, okay, there’s an uber-cycle here across all industries, which is called genAI. And everyone understands that that will require a huge amount of data movements, and that data centers are getting more distributed. That is a classic area for us network people to be part of.

“I think a lot of the companies are asking us, okay, what is our role and how do we participate in this value creation? … How do we really monetize? How do we come together and then how do we really participate in this whole AI economy? I think it’s fascinating,” he said. “And I think what you’re seeing is different answers from different operators and that can go all [ways]. So it’s interesting to watch.”

ABOUT AUTHOR

Kelly Hill
Kelly Hill
Kelly reports on network test and measurement, as well as the use of big data and analytics. She first covered the wireless industry for RCR Wireless News in 2005, focusing on carriers and mobile virtual network operators, then took a few years’ hiatus and returned to RCR Wireless News to write about heterogeneous networks and network infrastructure. Kelly is an Ohio native with a masters degree in journalism from the University of California, Berkeley, where she focused on science writing and multimedia. She has written for the San Francisco Chronicle, The Oregonian and The Canton Repository. Follow her on Twitter: @khillrcr