Q2 2026 feels like the breakthrough quarter for US telcos, maybe. We need to take another look at their European counterparts, and consider also different contexts. But having spent a bit of time with AT&T, T-Mobile, and now Verizon, there seems to be a pattern. The old telco story is improving: their investments are paying off, their subscriber additions look good, their churn is down, their margins are healthier. They are all talking about sharper pricing discipline, as well – versus expensive promotions for speculative acquisition.

And tellingly, they are not talking so much about AI just as a tool to automate their networks or improve their services. They are increasingly, and more confidently, positioning themselves as infrastructure providers for the AI economy. Some of this is prep-work, of course; all the mobile access build-out in metro regions, including purchase of new spectrum, is about putting AI to work at the edge, and mostly on the move – some time in 2027/28, and probably later. But those of them with proper fiber footprints – AT&T and Verizon, notably – are seeing quicker gains.

AT&T was interesting last week, talking about how to orchestrate 43 billion daily inference tokens between telco-geared language models in its IT ecosystem. It sounded like novel thinking, and right-headed. But the discussion that attended Verizon’s Q2 scores, actually from Friday (July 24), talked about the real fiber plumbing for AI – in ways the likes of Lumen, Zayo, and others have harped-on about for several quarters, at least. So it was good for Verizon, and probably good for all diversified mobile carriers, to hear about its $1 billion DCI deal with Google.

More than that, and maybe just to convince investors that AI will drive a new growth cycle, Verizon chief Dan Schulman went so far as to say that such deals will stack up in 2027, and create a new multi-billion-dollar revenue stream over the next five to ten years. Margins from its AI Connect portfolio – not just from mega-sized DCI projects, but also metro fiber, metro/edge data centers, and sundry edge services (including slicing and private 5G); all bundled and orchestrated as a programmable enterprise layer – will be “equal to or greater” than existing connectivity lines.

What is the quote from Schulman? “This is a very different revenue growth profile than Verizon has had in a very long time.” So a good second quarter, then, which might just herald a good couple of years, at least, for telcos – maybe not enough to post the kind of double-digital growth that the likes of AMD et al, higher up the AI stack, have been recording quarter after quarter, but enough to look like proper propulsion. Because AI needs networks, of some sort, just as it needs chips and models – and increasingly so as AI workloads leave the data centers. We will see.