4
AT&T is stripping out its legacy copper network and making fiber and wireless the core of its business – with a converged service, enterprise push, and adjacent satellite integration all on the cards.
In sum – what to know:
Foundation – Fiber and wireless are the foundation, with AT&T targeting more than 60m fiber locations by 2030 while retiring copper and eliminating around $6bn in legacy costs.
Convergence – AT&T chief Stankey says the next phase, after bundling, is to manage all network technologies as a seamless connectivity proposition for consumers and enterprises.
DCI and D2D – AT&T is returning to data-center interconnect, but Stankey says it will supports its network and not be its mainstay – and insists D2D is a niche, albeit valuable, concern.
AT&T wants to emerge from this decade as a “metropolitan fiber provider with a kick-ass nationwide wireless network” – and little else, it seems. That was the message from John Stankey, president and chief executive at the firm, during a sit-down interview with Goldman Sachs Group this week (September 9). Well, almost; it is the carrier profile AT&T wants so far as its traditional customers go. Alongside, it is also developing a line in data center interconnect (DCI) to connect big AI engines in the cloud, and into metro centers at the edge. But Stankey said AT&T does not want DCI to be the “mainstay”; its modus operandi is to get traffic to customers, with interconnect fiber for support.
AT&T is spending heavily on fiber and wireless infrastructure, illustrated by its recent deals for Lumen’s consumer fiber assets and EchoStar’s wireless spectrum, while also dismantling its legacy copper network, accumulated over many decades. Unglamorous, and not so well covered, but the project will strip $6 billion of legacy costs, it reckons – as initially tabled at its ‘analyst day’ in early 2025, and as realised in the closure of central offices, leases, “chillers”. “[We will] leave this decade… [as] a metropolitan fiber provider with a kick-ass national wireless network, and that’s all [we’re] going to support. All the other stuff we’ve been hauling around for decades [won’t be] there anymore.”
Stankey went on: “The agility that comes into the business, the clarity of purpose, the ability to use that unified infrastructure… to bring products out that support those two fabrics – dense, rich fiber, with a fantastic wireless network on top of it: it is going to be a great company as a result of that.” Its installed copper base makes it the “fifth largest copper mine in the US right now,” he suggested. “Seriously. That copper needs to come out and it needs to be monetized – to pay for the work to shut down all that infrastructure, rip out the mainframes, and do all the things we need to do to leave this decade as a fantastic metropolitan fiber company with a kick-ass wireless network.”
The message was repeated again. “That’s a different AT&T…. And it’s a different growth profile and a different agility. And that’s why it’s exciting.” But, per the wider telco discussion, AT&T’s transformation is not just about replacing copper with fiber, but about building a common hardware foundation and software fabric for consumers and enterprises, and AI players as well. Its priority is convergence, he said – not just as bundled service offers, already paying out, but as multi-access comms as well. The latter point addresses the Musk-sized angst in the market about satellite comms, of course. But all the bundling, on the back of its twin terrestrial plays, should be noted, said Stankey.
Fiber locations
AT&T has gone from a low-20 million range for fiber locations in 2021, to 30 million today, including about four million additional locations via the Lumen deal; it has a target of 40 million by the end of 2026, and 60 million by the end of the decade, incorporating organic builds, acquired locations, and its open-access Gigapower venture with investment firm BlackRock. “For five years, we’ve been pretty deliberately investing, top of the industry… If we put the right assets in play, we can grow at a better rate, with better profitability… The goal is to [be] a scale provider in fiber, and we lead in deployment of fiber infrastructure over time, [and to] strengthen our already-strong wireless business.”
Last quarter, its ‘advanced connectivity’ revenues – a mixture of fiber, fixed wireless, and mobile – grew around five percent in the second quarter, to June 30, with EBITDA up by eight percent. “Growth is coming,” said Stankey. “The products, together, are a superior offering; we’re able to take share in places where we weren’t before, plus take natural growth in the market. This is largely on the back of putting two products together… [It means] higher household revenues, better margin performance as a consolidated account, lower churn, [stable] acquisition costs… That formula is good, and it’s got legs because we’re not done building footprint until the latter part of this decade.”
At that point, the physical build-out is “wrapped-up”, effectively. “We’ll be able to ride that chain all the way,” he said, suggesting the bundling phase on top represents the “early innings” in the convergence game. The whole sport gets more interesting in the middle and later sessions, he said, when convergence moves beyond bundles, discounts, and retention into the management of different access technologies. “We’ll see satellite services bundled in at some point in the very near future,” he said. Which is what every telco is currently working on, of course – from rivals Verizon and Lumen, to suppliers Ericsson and Ciena, to partners like Equinix and Megaport. But it ain’t so easy.
”It’s not the easiest task to manage the layers of a network between satellite, wireless, fixed broadband, Wi-Fi. To do that right, you start to do things to say, ‘look, I can give you a very seamless experience from a service perspective as well as an easy experience from an account relationship billing perspective’. When you start to do those things and you’re giving somebody always-on connectivity, always working, very consistent – which moves [the proposition] beyond just a discount… [to] a product that works better.” Stankey put forward the firm’s OneConnect product as evidence of early progress – specifically to turn home broadband and mobile into a single billable product.
Newer ‘innings’
In these later “innings”, AT&T (et al) will move “well beyond just an affordable price of putting the products and services together,” he said. We’ll address Stankey’s comments, well understood, about the D2D momentarily; but the same convergence philosophy extends to its enterprise play – per its quarterly scores for ‘advanced connectivity’ revenue. There was an admission, as well, that AT&T has let it slide. “Our effectiveness… lagged [versus] what we’ve done on the consumer side. But many of the plays were the same… [and] we’re eking out low single-digit growth… [and] doing a better job of getting fiber to the right places… [and] extending our distribution channels,”
He added: “Candidly, we lived… very much in the Fortune 300. To be more relevant in the mid-market and low end of the market, we had to do a little bit of development of capabilities. We’re now starting to show some [results]. We’re still early; we have more work to do, but that’s what gives me confidence we can grow business [sales].” Stankey said AT&T is now prepping a “second act”, part of its late-phase convergence innings, where the product itself gets more sophisticated. “That’s next – [about] how an [enterprise] customer can manage quality of service [and security] across layers of the network, whether [the session is] initiated on a wireless device or on a fixed connection.”
He added: “Those things are about ready to come out.” Again, same for Verizon and Lumen, and everyone – and those two firms, at least, might be further down the line. The pressure is on; enterprises want AI services, and don’t want to wrangle with telcos about how to get them – safely, and on time. “[Some] businesses are having to think differently about how they gain access into the cloud to get to tools, augmentation of bandwidth to do those things, and finding use cases that they need support with.” Meantime, there’s lucrative DCI work to connect data centers, as well. Verizon has just flagged DCI deals with Google Cloud and AWS (Corning for AWS, nominally), of course.
“Data-center to data-center and data-center to metro-interconnect is an area we haven’t played in very actively for the last decade or so,” said Stankey. “We’re back in that game in spots where it makes sense for us to do that. [But] I don’t want it to become the mainstay of our business.” The last reference might be a subtle broadside for Lumen, perhaps. He went on: “Our business should be built on the value of getting traffic to end-user customers… But as this infrastructure gets built out, there’s no reason we shouldn’t participate… where we have routes we’re happy to build for our own traffic… [and] at a level [where we can] drop traffic into any cloud provider’s infrastructure.”
Primary business
AT&T wants enough interconnect capacity to support its primary business, to connect its customers to their workloads – in other words. Which means making sure fixed and mobile traffic reaches the right cloud infrastructure with the right latency and reliability – easier on “symmetrical” fiber systems than on asymmetrical mobile ones, where downlink and uplink channels are differently specified, and broad spectrum holdings dictate variable performance. “In the wireless business, it’s going to require a little more tuning. Upstream is going to be much more important… Low-band spectrum [with an] engineered uplink… is going to be really critical.”
In July, AT&T completed a $23 billion deal for 30 MHz of 3.45 GHz mid-band spectrum, valuable for additional 5G capacity, and 20 MHz of 600 MHz low-band spectrum, extending coverage and strengthening uplink capability. AT&T started to deploy the 3.45 GHz spectrum under a short-term spectrum-manager lease before the transaction formally closed; by November 2025, it had nearly 23,000 cell sites, covering more than 5,300 cities in 48 states. It has selected Ericsson’s 600 MHz dual-band radios with 8RX to deploy the latter.
Stankey referenced use cases “like autonomy [and] robotics”, plus “video processing upstream”, going “into the cloud for analysis”. He said: “Those applications are being seen very actively right now in the market. Those are all good things for us.” On the flipside, and for the record here, AT&T’s own internal AI usage has focused on “meat and potato stuff” like AI-assisted “flows into call centers”, where “customer satisfaction is high, effectiveness is high”, plus for certain business support areas and “software development shops” (higher productivity, rather bottom line savings). AI is starting to be applied to churn forecasting and more dynamic / bespoke customer offers.
Stankey said: “When you do some of the meat and potatoes like managing customer service functions, handling customers better, being more efficient on engineering, you don’t need frontier models. You can do [it with] last year’s chips and some pretty good open source models. We’re getting good at apportioning workloads to say what requires the high-end investment in frontier models and important chips versus… more streamlined infrastructure. We feel good about the returns we’re getting as a result.” But all of this – the big infrastructure build-out, familiar convergence bundling, early internal AI workloads – lead to the same place: a converged network offer for a new AI economy.
Satellite furore
And it leads us back to D2D satellites, as well – about which there is such a noise in the market, presently. Stankey made the same point that the bosses of most legacy telcos have made: that satellite is a niche complement tech, and not an all-conquering replacement – as the Starlink messaging goes. Stankey responded: “There’s certainly roles for satellites beyond just mobile. But relative to [mobile, which is] the center of attention right now: 98 percent of [mobile] use cases are solved by infrastructure that’s out there and working today, much of which is built over the course of decades. We started putting robust infrastructure in high-rise buildings and stadiums 20 years ago.”
That has been a “consistent drumbeat”, he said, pounded-out by the traditional telecoms industry – to the point that “someone at a World Cup game [can post on] social media… flawlessly”. “That doesn’t just happen overnight,” said Stankey, and took aim at “the notion that really small cells, randomly deployed, are going to somehow fix that problem”. Good fiber is the key, and terrestrial mobile networks are closer to it than low-Earth orbit (LEO) networks. “Once you hit the fiber, nothing’s ever going to beat it. And that includes the short hop from a device to a cellular antenna – which is a lot shorter than a hop to something that’s 300 miles above us.”
He added: “The faster you can get that into a fiber, and the more actively you manage it across both fixed and wireless, the better the experience is going to be… There are use cases where satellites will be very well suited. But we address the bulk of them on [our] infrastructure.” Stankey referenced its planned three-way joint venture in the US with T-Mobile and Verizon to pool spectrum resources, develop technical specifications, collaborate with multiple D2D providers, and make satellite connectivity more seamless when terrestrial coverage disappears. “We’re all in the same position: we’re addressing 98% of the use cases. We want to make sure we can get the last two percent.”
A full transcript of Stankey’s interview with Goldman Sachs is available here.
