AT&T's Fiber Flywheel Meets Its Leverage Test
A strong converged-customer quarter collides with SpaceX anxiety, lower fiber ARPU, and a balance sheet still digesting the EchoStar spectrum deal.
T · Earnings Call · 2026-10-21
AT&T's third-quarter report is a paradox. The operating numbers were the best they have been in years: more than one million net adds across fiber, fixed wireless and postpaid phone; consolidated service revenue up 3.8%, the best in more than four years; adjusted EBITDA up 5.4%; a 38.7% adjusted EBITDA margin; and adjusted EPS up 20%. Yet the stock is down about 6% over the past 90 trading days and 18% from its September 15 peak. The tape is not arguing with the quarter. It is arguing with the second act.
The beat is real — and narrower than the headline
AT&T's Advanced Connectivity service revenue rose 6.1% and segment EBITDA 7.3%. Wireless service revenue grew 5%, helped by 4% postpaid phone ARPU growth, while postpaid phone net adds were 379,000 on stable 0.92% churn. Advanced home internet service revenue rose more than 23%, with 696,000 net adds — 416,000 fiber and 280,000 Internet Air. This is a converged-customer machine: 43% of advanced home internet customers also take AT&T wireless, and converged account growth is faster than in Q2. The financials confirm the base: total revenue of $31.6B in the latest filed quarter, up 2% year over year. The catch is that this growth is increasingly manufactured through price points, not price.The quiet demotion of fixed wireless
Perhaps the most revealing line on the call was John Stankey's answer on fixed wireless. After years of treating Internet Air as a strategic growth product, he described a future in which “there is going to probably be some kind of a peak and then a downward trajectory of fixed wireless” — John Stankey, Chairman and CEO · 2026-10-21 as better technologies settle into place. That is not a reversal — on the prior call he had already called FWA “a really important tool in the portfolio” — John Stankey, Chairman and CEO · 2026-07-22 but not the optimal long-term fixed technology — but it is a sharper framing. AT&T still sold 280,000 Internet Air net adds in Q3, yet the strategic emphasis has moved decisively to fiber and convergence. The keyword fixed wireless remains in the company's top themes, but its related phrase now points to a downward trajectory rather than an offensive weapon. Investors should read that as a signal: AT&T does not intend to buy broadband share with FWA if it can build or acquire fiber instead. Fiber, however, comes with its own arithmetic. Fiber ARPU fell 4% year over year, and management guided Q4 to a similar decline, citing introductory pricing, converged discounts, and lower-ARPU Lumen customers. Stankey framed this as deliberate: “what we want is a portfolio of customers that basically buy from us at all discrete price points” — John Stankey, Chairman and CEO · 2026-10-21. The company is optimizing total service revenue and EBITDA rather than ARPU. That has been the playbook for at least two quarters; on the April call, management said getting fiber penetration from 40% to 50% was worth pursuing “even if it means we take some ARPU to dilution” — John Stankey, Chairman and CEO · 2026-04-22. The difference now is that the dilution is visible in the reported numbers, while the offsetting lifetime-value benefit is still a promise.AI-grade connectivity is the new growth story
The freshest language on the call was not about consumer broadband at all. Stankey positioned AT&T as an on-ramp for AI workloads: “Over 90% of mapped U.S. data centers are within 10 mi of our fiber, and over 80% are within 1 mi” — John Stankey, Chairman and CEO · 2026-10-21. The company is working with AWS and Microsoft on high capacity fiber connections and talking about managing traffic across multiple cloud fabrics, with long haul and metro fiber as the bridge between AI workloads and end users. This is a plausible second growth engine: AT&T owns dense metro fiber, and AI inference needs low-latency connections to the edge. It also explains why management is comfortable spending through a leverage spike. But the evidence is still early. The call offered no AI revenue number, and the global tape has been de-rating the theme: the AI data center basket is down about 5.6% over the past 30 days in the Know Trend price history. AT&T is trying to attach itself to a story the market currently likes less, not more.The balance sheet is the bear case
The quarter's clean operating story runs into a messier capital-allocation story. The $23 billion EchoStar spectrum purchase pushed net debt to adjusted EBITDA to 3.13x from 2.68x, and management still targets 2.5x by 2029. The sale of AT&T Mexico — which produced a $2.6 billion non-cash impairment — and the fiber JV with Global Infrastructure Partners and CPP Investments are meant to help. So is EBITDA growth. But the timing matters. AT&T now expects capex to be about $1 billion lower per year, yet it kept free cash flow guidance unchanged. Pascal Desroches explained where the savings go:In plain English, the capex dividend is being recycled into device subsidies and interest, not dropping to the bottom line. The balance sheet remains the constraint: effective net cash of negative $116.1B at the latest filing, with interest coverage of 3.7x.We believe this positions us to reach 60 million+ locations with fiber by the end of the decade, with annual capital investment that is roughly $1 billion lower than we previously planned. ... we expect to reallocate this capital into the business primarily through working capital in support of customer acquisition investments, as well as higher interest payment.