New AI services point to telcos moving back to the edge

Verizon, Telefónica and Rakuten this week demonstrated some of the different approaches telcos are taking to capture new revenue growth from enterprise AI usage.

In the US, Verizon is betting on adapting its existing investment in network services to the needs of enterprise AI traffic with AI Connect, a combination of products and services aimed at hyperscalers, cloud providers and global enterprises.

“We already have a funnel of well over a billion dollars and growing, simply by leveraging our existing infrastructure,” said Scott Lawrence, SVP and Chief Product Officer, Verizon, in a call with analysts.

Operators which have built up fiber and 5G networks as well as investing in data centers and compute power see an opportunity to package them up and benefit from the current AI boom. That includes earlier investments in MEC infrastructure. Meta and Google are among the list of hyperscaler, GPU-as-a-service and data center companies buying dark fiber or lit wave services from Verizon in order to fulfil current and predicted demand for AI products and services.

Enterprises are also “looking for ways to architect their network … [to] control where the workloads flow, and [to] … securely interconnect across a multi-cloud or distributed compute,” according to Lawrence.

In the next five to seven years Verizon, which was an early investor in multi-access edge computing infrastructure, says it expects AI workloads to help generate “explosive” growth in demand for networks, including at the edge.

“As AI shifts from training to deep inferencing deployment, the need for distributed computing will become increasingly more important for real-time decisions and predictions. We have thousands of distributed telco facilities that already have power, space and cooling available,” said Lawrence.

So far, growth in edge services across the industry has been lackluster. However, new research from IDC suggests GenAI may create a problem to which edge services will be the solution.

“Faced with the challenges of scaling GenAl inferencing, 40% of ClOs will rely on edge services from cloud providers to satisfy performance and data compliance requirements,” says the research company.

In addition, Verizon has stated that 40% of data centers are expected to face operational constraints by 2027, which in turn could shift workloads to the edge.

Yet it is still relatively early days for enterprise GenAI adoption, and it is not always simple for enterprise customers to manage.

An IDC cross-sector survey of 1,600 CIOs found that 30% are currently using GenAI extensively in their IT infrastructure. However, a total of 17% of those surveyed have experienced "significant disruption in [their] organization's competitive position or business operating model in 2024" as a result of using GenAI.

Given the difficulties that enterprises can face, some telcos are stepping in to offer GenAI services that go beyond connectivity.

Telefónica Tech, for example, has launched a GenAI platform that enables B2B customers to create customizable virtual assistants that can solve complex queries, automate repetitive tasks and optimize internal processes. The system sets out to be simple to configure while providing access to the main large language models (LLMs).

And in Japan, Rakuten Mobile has launched Rakuten AI for Business, which it describes as a GenAI tool optimized for Japanese culture, regulations and business customs. In addition to answering questions it helps with document creation, translation, brainstorming, analysis and research.

But when it comes to providing GenAI services, TM Forum Chief Analyst, Mark Newman, cautions that telcos will be more credible if they have gained in-house experience of the tools they are providing enterprises.

“Throughout the years, telcos have tried to launch services based on technology that they haven't yet mastered themselves. They are more likely to make a success of GenAI service platforms if they have already deployed it at scale with their own teams."