Better telco economics today…

… but tomorrow still looks like dicey.

Telcos are having a better time of it, apparently. Orange’s latest quarterly scores look decent: 3.6% higher revenue; 4.6% higher profit (EBITDAaL); 2.5% lower cap-ex; 8.8% better cash flow (across three quarters). It has raised guidance on EBITDAaL growth from 4% to above 4.5%. As with AT&T last week and Verizon yesterday, the uptick is down to some commercial growth, typically in bundled services, and real financial discipline, including simpler operations, improved productivity, targeted savings, some operational AI upside, and, here and there, by taking the axe to staff headcounts. They are benefitting, then, from smarter sales, sharper focus, plus some historical belt tightening – rather than new demand or services. Orange credits operational efficiency initiatives for helping drive its EBITDAaL increase. Still, it looks like a win, and a quarterly snapshot that is broadly reflective of the tier-one telco market. All are growing earnings faster than their top lines, raising guidance or otherwise showing some momentum.

So telcos are taking better care of their existing businesses, it seems. And then there is the future stuff, on top. Orange said nothing in its press note or earnings call about AI infrastructure, say – and nothing very much about ‘sovereignty’, except for allusions in the repeat of the group PR messaging about its mooted joint bid for European MSS spectrum with Deutsche Telekom, Telefonica, and Vodafone. Nor did it rise to the Starlink bait – in the way both AT&T and Verizon did during their Q3 Q&A sessions with city analysts. Which reflects their different geographic remits, perhaps – mad America versus angsty everywhere-else – and how the AI game is playing out in terms of infrastructure build-out. Picking up on the coverage yesterday about “no-shortcuts” for SpaceX, chief Dan Schulman also spoke at some length during Verizon’s results call about its $4 billion in signed AI infrastructure contracts, including a new $2bn-plus deal with AWS – to go with its big Google Cloud contract, announced last quarter.

He also referenced an “active pipeline” of “additional” hyperscalers, neoclouds, and (big) enterprises “in negotiations [for] either dark or lit fiber”, plus “momentum with numerous partners to use our network hubs for AI inference edge computing”. The latter is “already turning into signed deals”, he said. But Schulman expects the financial impact from these activities to begin in 2027, with only modest success-based cap-ex anticipated next year – and lots of it tied to lifetime terms, rather than ARR accounting (“hundreds of millions of dollars”, just totting-up its first tranche). Which, very clearly, makes AI infra a prospective growth engine rather than a ready one, driving very much in the way of Q426/FY27 improvements. Equally, he couched his caution with enough city dope to reassure the markets that AI will pay out for telcos, calling it a “real business” and a “tremendous opportunity” with “strong” demand and “attractive margins” within a “rapidly growing TAM”.

“We’re just scratching the surface,” he said. Separately, T-Mobile, ahead of its scores this week, is talking about the same – albeit with more of an architectural view, versus a commercial one. John Saw, chief technology officer, has a worthwhile blog about how the network might evolve from connecting physical AI to helping supply the intelligence behind it. It is evaluating whether parts of US humanoid-maker Figure AI’s Helix S1 and S2 models can run on edge compute, rather than rely just on robot-edge processing. It is working with Serve Robotics on an edge-based AI supervision layer to help manage delivery-robot fleets, and with the city of San Jose on connected cameras that use edge AI to build a broader picture of traffic conditions. This is the future stuff, of the type Verizon says it has signed a handful of customers for on long-term deals.

Meanwhile, while we’re talking-CTOs, Verizon’s own Yago Tenorio made a related point at 6G Forum earlier this month. Physical AI, he argued, will change the traffic mix: AI agents, wearables, cameras and robots could push networks away from the traditional 8:2 or 9:1 downlink/uplink ratios towards “something closer to 1:1”. There is “very real uplink demand”, he said, which telcos won’t satisfy “just with incremental improvements” from top-down 6G drops. Which is the other side of the same story: telcos also taking care of their future telco business, even while the AI infrastructure rave-up spills over. If networks are to support machines that constantly see, sense, and respond to the physical world, they need not only distributed compute, but the capacity to carry data back into the network. Tenorio thinks this will require architectural change, not just another generational tech upgrade.

Telcos have their work cut out – and, actually, despite all the talk and strategy, the use cases (to justify the business cases) still aren’t very well defined. Is ISAC really the answer? Are ‘wearables’ the thing? The tech is advancing, the deals are landing, and the architectural requirements are becoming clearer. Will all of this add up to material new-growth for telcos? We might have some better telco economics today, but we still have uncertain telco economics tomorrow. Note, a longer version of this article is available here.