- Operators are increasingly chasing a goal of becoming AI companies
- But telcos have historically struggled to evolve outside their core competencies of voice and data
- There is one key area where they could play but they're stuck waiting for the right use cases to emerge
Pivot. Fail. Pivot. Fail. Could the turn toward artificial intelligence (AI) finally be the pivot that works out for operators?
Faced with lagging revenue, operators in recent years have been scrambling to find a clique that will make them cool again. First, there was 5G, which, with tools like network slicing, was supposed to make telcos more than just a dumb pipe. Then, with the rise of the cloud, operators tried to contort themselves into so-called “techcos,” a feat which proved easier said than done. And now, they’re chasing the latest fad – AI.
As we recently noted, telecom giants including Deutsche Telekom, SoftBank, Fastweb, Indosat, Singtel and Swisscom, are helping to build or investing in AI and AI factories. Telenor, for instance, has created an entirely new side business – Skygard – to function as an AI startup focused on innovating around its existing network capabilities.
And then there’s SK Telekom, which last year developed a proprietary telco large language model and in 2025 is aiming to transform itself into a self-described “AI company.”
But as operators buy a new hardware wardrobe to dress themselves up as AI cool kids, one key question looms: can they pull it off?
Well, it seems even the operators chasing the dream have their doubts. In a New Year’s address to employees, SK Telekom CEO Ryu Young-sang stated “The year 2025 is expected to be one of unprecedented crisis” with increased competition around AI.
Opinions elsewhere seem split. IDC Research VP John Byrne told Fierce that “Telcos can potentially make it work but only if they can 1) correctly identify the true opportunities where they have a “right to win”, i.e., a legitimate claim to be able to provide unique value, and then 2) build a GTM strategy and set of tactical plans that enables them to seize those opportunities.”
The catch, of course, is that both of those things have historically been hard for telcos, he said, primarily because of the complexity associated with moving beyond traditional voice and data services. That’s “part of the reason why potential revenue generators such as IoT and multi-access edge computing (MEC) have not moved the needle very much,” Byrne said. AI poses a similar challenge, he explained.
FiberLight CEO Bill Major, in contrast, said telcos are better off sticking to their core competencies to facilitate the AI boom with their networking expertise. “I believe that the concept of a telco becoming an AI company is premature and highly unlikely,” he told Fierce. “Traditional telcos have continued to chase that next thing, and when you try to be everything to everyone, you fall short more often than you succeed.”
While AI workloads are highly centralized today for model training, analysts have previously predicted processing to become more distributed in 2-3 years with the rise of inferencing workloads. Not to mention hyperscalers are scrambling to find enough space and power to expand their compute facilities. And as Fierce has previously noted telcos are perfectly poised to pounce on this opportunity.
Why? Well, as Major, Byrne and Recon Analytics founder Roger Entner, all pointed out, operators are sitting on a number of prime locations for edge data centers courtesy of their copper retirement efforts. These are perfectly placed to deliver AI inferencing closer to end users.
Byrne added “mobile operators also have thousands of towers and other cell sites that are increasingly capable of processing AI workloads and work really well for “AI-on-RAN” types of services that require radio and accompanying AI inferencing.” For example, he said services in the latter category might include connected cameras or vehicles utilizing AI-powered computer vision.
But again, the hang up is realizing the potential that’s out there. As Entner noted, promised revenue from mobile edge computing has failed to materialize. Case in point: Verizon, which has been chasing the MEC dragon since at least 2020. The operator originally said it expected meaningful revenue to materialize from 5G MEC in 2022 but later hedged to 2023 before admitting that slower-than-expected uptake would hamper its stated goal of hitting $2 billion in MEC and enterprise revenue by 2025.
“I think AI is big without a doubt,” Entner concluded. “But the use case for edge compute [for AI or anything else] has not really emerged yet.”