- US carriers are gearing up for the AI tsunami. European carriers? Not so much.
- Second-generation AI networks will carry three to 6 times more traffic than today’s edge infrastructure.
- Telecom has been seeking an exciting revenue stream. Wish granted!
Over the next few years, something genuinely extraordinary is going to happen at the edge of telecom networks. For service providers and carriers, this could be either the best thing that has happened to them this century — or a complete disaster. It all depends on what they do today.
Agentic AI is the catalyst poised to change everything—not merely by generating more data, though that is part of the challenge—but by shifting the way traffic moves around the network.
Let’s break it down:
First-generation AI networks are all about the core – the massive data centers and powerful AI factories sitting in the heart of the cloud, handling LLM processing. End users at the network edge, typically consumers, use their AI chatbots to send their queries to the cloud, and the AI does its inferencing thing and replies with the answer.
In data terms, the queries are small, but the replies are large; the traffic flow is asynchronous and point-to-point, from user to LLM and back again (like a hobbit, on a quest).
So far, so good. This is a classic telecom traffic flow, requiring the same network design that carriers already have in place to handle consumer video streaming services.
Second-generation AI networks, based on agentic technology, take this familiar and relatively predictable model and blow it to smithereens. In this new world, AI agents operate autonomously, taking on complex tasks without requiring human oversight. (What sort of tasks? All of them. From robotics to autonomous vehicles to medical diagnostics, the list is endless).
Après AI, le deluge
From a network architecture perspective, the key here is not what these agents do, but where they are located: out at the edge of the network, where the users, enterprises and industries they serve live.
But also, and this is important, AI agents are a chatty bunch, constantly sharing information peer-to-peer with their community of virtual co-workers out at the edge. And that changes the traffic flow from point-to-point (up and down, to the data center and back) to multipoint.
Picture, if you will, a logistics business run by hundreds of AI-powered autonomous robots and drones. They coordinate nonstop—moving goods, restocking, routing shipments—in a chaotic storm of traffic, ping-ponging between countless end nodes. It’s a relentless, bewildering, overwhelming flow of data and actions, all driven by agentic AI working at scale.
Now extrapolate this to every industry, everywhere, and throw in consumer applications, and you start to get a sense of the challenge facing carriers.
Beyond the chaotic nature of this traffic, volume will also be a major test. Just how much more edge traffic are we talking about before the end of the decade? In discussions over the last two months with senior executives from a dozen vendors and carriers, estimates have ranged from a best-case scenario of three times the current load to six times, and even "the sky’s the limit." But the truth is, no one really knows yet. For now, let’s just say “a lot.”
Let the good times roll out
The AI edge surge is fantastic news for the telecom industry.
Vendors like Ciena, Cisco, Ericsson, Huawei, Nokia, and Samsung can expect spending levels not seen since the fiber optic boom of the late 1990s. Hoorah!
Carriers should also be excited, yet not all of them are, with their reaction splitting roughly along geographic lines.
North American carriers that I have recently interviewed — including NTT Data, AT&T, Lumen and Verizon — identify the impending AI data tsunami as a revenue growth opportunity and are all planning significant infrastructure investments to meet the new demand. They’re going all in on edge.
Conversely, major European operators, such as Deutsche Telekom, Orange and Vodafone, are still dithering over their fresh croissants and teeny-tiny cups of espresso - hesitant to invest heavily in upgrades following the disappointing returns on their initial 5G deployments.
There’s still a lingering belief inside carriers in Europe and parts of Latin America that, with 5G, they were sold an expensive network they didn’t really need. The problem is that unless they start planning and deploying edge networks for agentic AI traffic now, they’ll face the opposite issue: a burning need, with no network to handle it.
Time to move
This is not the time for carriers to vacillate. Over the past decade, they have been grappling with an existential crisis. As profitability wavered, they found themselves torn between envy of hyperscalers that had profited immensely from networks financed by carriers and the pursuit of a range of new, often poorly thought-out, get-rich-quick strategies that fell outside their area of expertise — including building AI factories and developing OTT-style consumer applications.
In contrast, improving edge networks to support AI is undeniably a core—and edge—competency for operators, something they excel at, and hyperscalers struggle with (see search engine giant Google’s woeful Google Fiber effort).
The edge AI build-out is, of course, non-trivial. AI requires telco-level reliability, increased bandwidth and low latency, and also presents new challenges (super-agents and LLM routing, to name two). However, in about three years’ time, carriers that take on this challenge will re-ignite a business case that has languished in this century: revenue that scales in proportion to capacity — and that is surely worth going the last mile for.
Steve Saunders is a British-born communications analyst, investor and digital media entrepreneur with a career spanning decades.
Op-eds from industry experts, analysts or our editorial staff are opinion pieces that do not represent the opinions of Fierce Network.