Happy Friday. Here’s a good overview-interview with AT&T boss John Stankey about the US carrier’s five-year reinvention plan – which is interesting partly because of what Stankey isn’t saying. Remember your roots, he seems to say; even if you change your outlook. Which runs a little contrary, in tone at least, to all the ‘supercycle’ suckery of Verizon, Lumen, Nokia, Ciena and others. AT&T will take DCI business where it makes sense, but its work, rather prosaic, is to build better networks to put AI to work for consumers and enterprises. Simple; also quite powerful.

Stankey sounds grounded in conversation with Goldman Sachs this week – pulled apart and written about in the RCR piece. He is pretty explicit that he doesn’t want DCI to become “the mainstay” of the business – even if AT&T is back in the interconnect game, taking hyperscaler opportunities it finds them; but only as part of its essential customer proposition. Don’t knock it, he seems to say; telco networks have been built over decades, and cannot just be replicated – in software in the cloud, in hardware in low-Earth orbit – overnight. Whatever Musk says.

Satellite comms are useful, of course, but very niche – serving the two percent, out in the sticks. AT&T will address the D2D gap in cooperation with its old rivals, T-Mobile and Verizon, he says. And meanwhile, more or less, the discipline is just to keep going: to build lots of fiber, retire lots of copper, grab spectrum and upgrade and densify wireless, and then make those networks work harder, and as one, for enterprises and consumers. Which is the AI grid story, in other words – metro access for digital AI, plus whatever physical AI (vehicles, robots, drones) actually arrives.