Count AT&T CTO Igal Elbaz among the growing number of telecom executives expecting AI to completely revamp network development and control, with the added kick that this new autonomous overlord may also kick the telecom industry’s “G” fetish to the curb.

Elbaz explained to an audience during the recent NSR/BCG Global Connectivity Leaders Conference that AI was having an impact across all aspects of telecom networks, from the core all the way out to the end user device, though the timing of that impact is still being determined.

“I will tell you that I don't think we're sitting here today and seeing a complete difference about how the network is operating or what kind of AI workloads we're seeing traversing … but we’re absolutely seeing signals,” Elbaz said, noting those “signals” are coming from “AI capabilities on our smartphones.”

Those signals could change as “we may start wearing all kinds of AI devices,” Elbaz continued, adding that those devices will be “more heavy on what we call the upstream side of the network.”

“We have a couple of examples of autonomous cars or robot taxis driving over our network,” Elbaz noted. “For sure in the training phase, we're seeing the ratio of upstream be above 50%, which is something that we saw before.”

Elbaz’s comments echoed that of his CTO counterpart Yago Tenorio at rival Verizon, who previously told SDxCentral that AI systems that are being embedded in everyday devices will change consumption expectations from “a transactional communication that you have today with AI: so you ask a question, you get an answer; to more of a very intelligent assistant that is aware of everything around you without you needing to fill in the context every time you want an action.”

“That is what requires a strong uplink, because either, if it's like video, audio, or sensing, it means that there is information flowing to a processing point in a continuous way,” Tenorio added. “So that's the problem that the industry needs to solve.”

Elbaz said that AT&T has started to tackle this AI-driven challenge by using its open network architecture in support of AI-driven solutions.

“[The open network architecture] 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, following intent and being able to build modules, software modules, and AI modules that helps us to operate our network,” Elbaz said.

The result is a more efficient network deployment and operating model.

“We're using this across the board, mainly in the RAN (radio access network) space, because clearly this is the area that we were spending a majority of our capital, so engaging on those models and capabilities for years, and again, we were able to take advantage through a very unique skills that we have internally in the newer model to actually improve significantly existing models and create new ones to the benefit of all them,” Elbaz added.

AT&T’s IoT efforts is one of those areas garnering some of that benefit. Shawn Hakl, head of AT&T’s business products team, recently explained to SDxCentral how that long-simmering business segment was starting to ride an AI-powered service wave.

“As you think about AI and agentic AI, you can simplify that flow,” Hakl said of the carrier’s growing IoT business. “If you make the communications element and the security elements agent consumable, coupled with the use of agentic AI tools for workflow, you can capture more knowledge, you can scale it, and you can execute it much more automated than you could before.”

Get off the ‘G’ cycle

That combination of AI and software is also putting the squeeze on the wireless telecommunication industry’s network upgrade cycle.

This generational upgrade cycle has been curated along a decade-length timeline by near-mythical “G” connotations. This started with analog “1G” technology back in the 1980’s; the initial move to digital with “2G” in the 1990s; the beginning of wireless data with “3G” in the 2000s; broadband-like data and an IP-based architecture for “4G” in the 2010s; and the current “5G” world of spectrum aggregation-based higher performing networks.

Elbaz thinks that generational naming scheme could be coming to an end due to the near-constant upgrade cycle enabled by AI and software.

“I don't think we as an industry can talk about what we think our industry is going to enable in 2030. That doesn't work with AI,” Elbaz said. “And 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 that the first thing we need to do as an industry is completely decouple the ‘Gs’ or the cycle of the ‘Gs’ from 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.”

This path is starting to gain steam as technology vendors continue to push this more rapid approach to network upgrades.

Nvidia, as an example, recently released a telecom-focused report that revealed almost 80% of telecom operators expect to see AI-native networks get the jump on 6G deployment, while around two-thirds said AI is driving autonomous network build out.

Where is the connected edge?

Elbaz also downplayed the need to rush compute resources to far edge locations, an architectural construct that is tied toward supporting latency-sensitive applications and is part of the surge in data center investments.

“I think the proliferation of compute and high-performing compute across the nation, in all metros, is just happening, with the software layer on top of this, with the tools that developers need,” Elbaz said of this surge. “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 two milliseconds of latency. I think the U.S. perspective on this is a little bit different.”

Elbaz said that from AT&T’s perspective, “we want to 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 going to be heavily distributed across the U.S. I think the U.S. is a little bit different from other countries and our perspective just because of how deep our compute is being built.”