LONDON – European telecom leaders gave differing views on the AI market opportunity as it braces for further impact from hyperscaler AI investment.

Speaking at this week's FutureNet World event, BT Business CTO Colin Bannon noted “extraordinary innovation and volume growth” being driven by physical network architecture and traffic flows between data centers, while admitting AI is not something network operators had centrally planned for.

“AI hasn't necessarily been launched under the network. It's escaped,” Bannon said. “We talk to the people who build the networks for the hyperscalers, and they talk to their AI people every day, and they don't know they've got more money than God, right? And if they don't know, we don't know.”

With hyperscalers able to spend freely on AI infrastructure, operators can only sit back and watch as the cloud giants dictate traffic patterns, capacity needs, and ecosystem choices, according to Bannon. The CTO drew comparisons with radio wave charts and the like from the Search for Extraterrestrial Intelligence (SETI) Institute, as telecom tracks a new and “different kind of fingerprint on the network.”

“There are challenges there, whether it be ... Meta smart glasses, where you start to look at the asymmetric traffic flow turning the network from blow to suck, to where you're sort of beaming up your contextual environment," Bannon said. "From there, to swarms of AI agents in the network talking to each other with a different pattern, where you're actually realizing you're starting to get some traffic flow. That is, the wide area network, or the local area network, where they charted a large language model across multiple laptops … and you start to see elephant flows in your lab, which is an effective way to fill up the caches of your networking requirement.”

The CTO hopes for the efficiency of AI to blow away the cobwebs from legacy telco processes, and for AI itself to become “more powerful around unstructured data, rather than just structured data.”

Bannon also saw an opportunity for telecom around sovereign AI, arguing there is no “better partner to consider,” for as telcos shut down legacy networks, they potentially free up power capacity for any government or national body wanting local, sovereign AI compute.

The CityFibre view

CityFibre CTIO and former Juniper Networks chief networking strategist Neil McRae brought a more here-and-now approach to the topic, highlighting that the British alternative network (altnet) provider recently launched an 8.5 Gb/s symmetrical wholesale broadband service in response to increasing network demands.

“People asked us upon launch why we were launching such a fast speed. No one's asking the hyperscalers why they're building so many data centers with so much AI, right?” McRae said. “The belief is that there's a huge opportunity there, and our network is very much ready for that in terms of low latency, in terms of symmetry of traffic moves from download only to being a much more balanced network."

McRae believes that with all the hyperscaler hype of AI, everyone is taking basic connectivity for granted.

“We're amazing in scalability. We're amazing at redundancy. We beat the hyperscalers in redundancy - fact. Those are things that are important to customers,” McRae said.

The CTIO’s views echo a larger sentiment in the connectivity space. In a conversation prior to FutureNet World, Lee Myall, CEO of U.K. B2B network provider Neos Networks, told this title bluntly that “AI would be dead in the water without us.”

“Hyperscalers and neocloud providers are placing huge bets right now on the build-out of data centers,” Myall said. “Given their stringent deadlines, they need confidence that the supporting connectivity is available when and where they need it, and at a cost structure that works for the individual project.

“At the same time, consumption profiles are changing," Myall added. "Day-one capacity requirements are higher than they’ve ever been, with enterprise customers wanting live visibility of utilization, early warning before they hit thresholds, and the ability to scale bandwidth without going through a ticketing process every time.

"Meeting those demands requires telecom providers to think differently about how they package and deliver connectivity," Myall continued. "Scale, of course, remains important, but it’s also about making what we have quicker and easier for customers to consume on an ongoing basis. We need to shift away from static quote and order processes and more toward connectivity that is consumed like software.”

Telecom: Plumbers or partners in AI race?

Myall’s view is shared by Idir Fodil, SVP for international networks strategy and EVP deputy at Orange. FutureNet saw the exec espouse flexible, software‑like connectivity to meet customer demands for adaptive bandwidth when training AI models.

“For example, 100-gig to one-terabyte, but for a small amount of time, and this is where we are seeing this value, as we adapted our network to be programmable,” Fodil said.

Fodil noted Orange sees value operating on the edge where AI inference occurs, applying content delivery network (CDN) principles to AI workloads by distributing AI inference workloads, and possibly model replicas, across many edge or regional locations.

But while Orange is confident it’s an AI partner of choice for enterprise, Kishan Chuckun, head of communications business for Europe, the Middle East, and Africa (EMEA) at IT firm Cyient, wasn’t so sure.

“The honest truth is I don't think large enterprises see telcos as the AI partner,” Chuckyn told the FutureNet audience. “There are discussions, but most of the discussions that large enterprises are doing in the AI journey are with the hyperscalers, with the SaaS (software-as-a-service) applications. They’re with the AI native startups that are coming up.”

On a more optimistic note, Chuckun saw telcos playing a more integral role in “the agentic world,” becoming part of a trustworthy application process as opposed to “just plumbing.” This echoed Fodil's view that trust is in the DNA of all operators.

“We provide trusted AI, meaning that we provide with the right infrastructure, with the best quality, best security, allowing our customers to be sure about their data,” Fodil explained.

“If you look at regulated enterprises, telcos have the footprint and they have the trust of these enterprises, which I don't think many AI companies have," Chuckun added. "So that's something that enterprises are recognizing more and more, which is a huge potential now for telcos to leverage."