Here’s a snapshot of the current health of the telecoms sector, ailing for years: network operator AT&T has just posted quarterly revenues of $33.5bn, up 3.6%, and annual revenues of $125.6bn, up 2.7%; network vendor Nokia has just reported quarterly sales of €6.13bn and €19.2bn for the same periods, both higher by about 3%. We might pick anyone; every listed firm is delivering seasonal reviews. But these are stalwarts of the telecoms game, and their latest scores, taken together, show how the mercury is rising, or not. And the thermometer says: old, tired, not brilliantly well, but also looking a little brighter at last. There’s growth, at least.  

And both their stories are framed by some kind of reinvention around AI, of course. AT&T has booked new spectrum from EchoStar for 5G/6G, but it has at least as much going on with fiber – sucking up Lumen’s consumer assets, and driving managed backbone services for enterprises. Most of Nokia’s growth is with the same: long-haul fiber companies, including the likes of AT&T and Lumen, serving AI workloads. The Finnish firm is in thrall, strategically and rhetorically, to this so-called hyper-scale AI ‘supercycle’ – which, by definition, must come full circle, or full spiral, post infrastructure build-out, and disappear back closer to zero again.

By contrast, the backbone infrastructure industry is booming – growing double-digits, at least. The dark fiber market is projected to grow at a compound annual rate of about 13% through 2031; the broader fiber optic cable market, lit and unlit, will expand by 14%; the data center networking market, for switching and routing gear, will jump 18%. And then, at the end of the line, you have the data center market itself, where annual cap-ex is expected to grow at a 21% per annum. And all of it is strung along by money-bags hyperscalers, who spent 180% more in the third quarter of 2025 than they did in the same period in 2024. The top four will spend $100 billion each in 2026, Bill Barney told PTC in Hawaii earlier this month.

Point is: all boats will rise, and yet telecoms is still anchored to the past. What to diagnose? What to prescribe? Mobile telecoms is mostly a last-mile story, and the old infrastructure touch-point for end-users. This AI revolution is young, as much about backbone infrastructure for now as it is about pattern-matching pyrotechnics. But user touchpoints have moved to the cloud – where the hyperscalers reside – with their frontier models, connected on telecoms networks. Maybe the new era of AI inference will bring some of that dynamism (and growth) to the edge – along optical backbones, via metro interconnects, into last-mile systems.