Is AT&T Asking You To Fund A Network For AI Traffic?

AT&T (T) is running its largest year ever for fiber expansion, a plan for 8 million new locations, of which over 4 million were acquired from Lumen, and asking shareholders to keep funding it. It expects $23 billion to $24 billion of capital investment in 2026, and the roughly $18 billion of buybacks and dividends is essentially all the free cash flow left. Management used to justify the build product by product; with the customers, it was already winning. The lead now is one blended subscriber number and a network positioned for AI traffic.

AT&T Once Sold The Build On What It Returned

Two years ago the earnings calls gave the receipts separately. The company delivered 403,000 postpaid phone net adds in a third quarter, with more than 200,000 AT&T Fiber net adds beside them.

Across those calls, management said the 5G and fiber investment was fueling its growth. A holder could read those two lines and tell which product was doing the work.

One Blended Subscriber Line Now Leads

The second quarter of 2026 call opened on a different figure. The company gained more than 1 million Advanced Connectivity subscribers, one line folding AT&T Fiber, fixed wireless, and postpaid phones together, with all three categories posting higher net additions year over year. Postpaid phone net adds are still disclosed, at 432,000, but they no longer carry the story.

Management describes agentic AI as reshaping network traffic and presents the same fiber and spectrum as the infrastructure those workloads will need. AT&T is also powering down its legacy copper network.

The blended lens does follow the business. The Advanced Connectivity segment carries over 90% of service revenue.

So Is The Fiber Expansion You Fund Still Earning?

On the company-wide numbers, yes. Adjusted EBITDA margin reached 39.1% in the second quarter of 2026, its highest since the company refocused on advanced connectivity at the start of the decade. Revenue over the trailing twelve months grew 2.6% against a three-year average of 1.6%. Management raised planned buybacks to roughly $10 billion in 2026, saying the stock does not reflect those fundamentals.

The market has not agreed yet. The stock is down about 7% over the past year while the S&P 500 gained close to 19%, though it remains up around 36% over two years.

Management still gives the per-product read, just not as the lead. Fiber ARPU fell 1.3% year over year in the second quarter of 2026 and was roughly flat excluding the lower-ARPU customers acquired from Lumen. Management expects more pressure there in the near term as converged discounts spread.

The pivot to the blended lead is made from strength, though leverage will test it. Net debt to adjusted EBITDA ended the second quarter of 2026 at 2.68 times and should reach around 3.2 times after the EchoStar spectrum purchase, which management expected to close by the end of July 2026. Management targets the two-and-a-half times range within about three years of that close. The line to watch when the third quarter of 2026 is reported is Fiber ARPU excluding the Lumen footprint. If the wider question is which heavy spenders turn investment into compounding returns, our screen of capital compounders ranks them on it.

So Do You Hold AT&T For What It Is Now?

The bet changed shape while you held it. AT&T is still a network business, and the numbers are better. You hold it for those numbers, not for a spending case that leans on what the network will be asked to carry and is harder to check between calls. If you would rather not judge one telecom yourself, the Trefis High Quality Portfolio selects on business quality instead of a story. That portfolio has a track record of outpacing the three major indices.