SerpApi.Org
AT&T delivered its strongest internet growth quarter on record Wednesday, per AT&T's Q2 2026 earnings release, adding 646,000 broadband subscribers across fiber and fixed wireless in the second quarter of 2026 — a result that did more than beat Wall Street's forecast. It handed AT&T's management team a data-backed rebuttal to the claim that had weighed on the stock for most of the year: that satellite broadband, and Starlink specifically, is poised to erode the carrier's competitive position. CEO John Stankey was blunt on the earnings call Wednesday. Satellite, he argued, addresses roughly 2% of the traffic AT&T cannot reach terrestrially — a "corner case," not a replacement. What the record numbers show is that the far more consequential battle is being won on AT&T's own terrain: a fiber-and-wireless bundle that is creating customers who are harder to lose, more profitable to serve, and increasingly irrelevant to Starlink's orbit.
That bundling advantage — AT&T calls it convergence — is not incidental to the Q2 beat. It is the mechanism behind it.
AT&T's Best-Ever Internet Quarter, by the Numbers
Adjusted earnings per share came in at $0.65, beating the $0.59 analyst consensus by roughly 10%. GAAP EPS of $0.66 came in up from $0.62 a year earlier. Revenue rose 2.3% year over year to $31.6 billion, coming in slightly below the consensus estimate of approximately $31.75 billion — a narrow revenue miss that the EPS beat substantially overshadowed. Adjusted EBITDA grew 5.2%, and free cash flow improved to $4.7 billion from $4.4 billion in the year-earlier quarter.
On the subscriber side, postpaid phone net additions reached 432,000 — nearly 100,000 more than the roughly 338,500 analysts had forecast. Postpaid phone churn held at 0.86%. That churn figure matters disproportionately: in postpaid wireless, a 0.86% monthly churn rate translates to roughly 10% annual attrition, and the European fixed-mobile convergence playbook AT&T is following has demonstrated 30% to 50% reductions in that rate for carriers that successfully bundle broadband and wireless under one account.
The internet numbers were the headline. Fiber contributed 367,000 net additions; fixed wireless added another 279,000. Combined, the 646,000 figure broke AT&T's own record and marked what Stankey called the company's best-ever second quarter for fiber net additions. Advanced home internet service revenues grew more than 27% year over year, driven by fiber growth and the contribution of the Lumen acquisition that closed in the first quarter of 2026. AT&T now reaches more than 38.6 million total consumer and business locations with fiber and remains on track to exceed 40 million fiber locations by year-end 2026.
The Convergence Engine: Why Bundling Is AT&T's Technical Moat
The subscriber numbers matter. The convergence rate matters more.
By the end of Q2, 42.5% of AT&T's advanced home internet customers also held a postpaid wireless account with AT&T. Excluding subscribers in the recently acquired Lumen footprint — where AT&T is still working to convert new fiber customers — that rate rose to 45%. "When customers consolidate their internet access with us, we see lower churn, outstanding brand affinity, higher lifetime values," Stankey told analysts.
The underlying mechanism is not a simple loyalty discount. Fixed-mobile convergence — a strategy European carriers perfected over the past decade — works because it increases the switching cost for both services simultaneously. Canceling one account means disrupting both. Research from strategy consultancy Arthur D. Little documents that mature FMC operators in European markets achieve annual churn rate reductions of 30% to 50% compared with non-converged competitors. Translated to AT&T's scale — more than 100 million wireless connections — a meaningful churn reduction in a converged base is a durable revenue shield, not a promotional effect.
The technical architecture enabling this is AT&T's OneConnect platform, which binds fiber home internet and 5G wireless under a single account and billing relationship. On the fixed wireless side, AT&T's FWA product delivers broadband via the same 5G mid-band spectrum as its mobile network, using customer-premises equipment that requires no trenching, no cable installation, and no professional setup appointment. The tradeoff: fixed wireless capacity is spectrum-constrained and shared with mobile users, which is why AT&T deploys it as a complement to fiber coverage rather than a fiber substitute. In high-density areas, fiber's dedicated bandwidth and symmetric upload/download speeds — identical performance in both directions — make it structurally superior; in locations where fiber infrastructure does not yet reach, FWA extends coverage quickly and cheaply. AT&T's record quarter shows the strategy working in both channels at once.
AT&T closed its acquisition of Lumen Technologies' mass markets fiber business in the first quarter of 2026, ahead of schedule, adding approximately 1 million fiber customers and expanding AT&T's footprint into major new metropolitan areas across 11 states. The Lumen subscribers carry slightly lower average revenue per unit than AT&T's organic fiber base — fiber ARPU declined 1.3% year over year in Q2, primarily reflecting the full-quarter impact of those subscribers — but excluding that cohort, fiber ARPU was approximately flat.
What AI Agents Do to a Carrier's Infrastructure Argument
Stankey used Wednesday's call to advance an argument that goes beyond the current competitive cycle. Agentic AI, he told analysts, is changing network traffic "in volume, shape, symmetry, and criticality" — and that change is structurally beneficial to carriers with dense, low-latency terrestrial infrastructure.
The data behind that claim is specific. A May 2026 report from Cisco analyzed real-world traffic patterns and found that AI agents generate approximately 450% more total network traffic per task than a human performing the same work manually. The reason: an agent does not send a single request and wait for a response. It queries multiple systems, chains API calls, retrieves context from several sources, validates outputs, and iterates — each step generating a separate network transaction. Roughly 70% of that incremental traffic is AI inference, meaning machine-to-machine communication between the agent logic and the large language models powering it. The inference connection — what Cisco researchers called the "agent's spinal cord" — demands low latency and symmetric bandwidth to function reliably. Starlink's latency of 25 to 60 milliseconds is competitive with cable for most residential applications, but fiber's 5 to 20 millisecond latency and symmetric upload speed give it a structural edge for latency-sensitive AI workloads.
Cloudflare CEO Matthew Prince confirmed in June 2026 that automated traffic — driven primarily by AI agents — had surpassed human-generated traffic online for the first time in the internet's history, with bots accounting for 57.5% of HTML web traffic. Enterprise network traffic, already projected to grow 2.5 times over the next decade on a baseline basis, is now modeled to grow approximately 9 times once agentic AI adoption becomes widespread, according to the Cisco report. AT&T's case is that its fiber-forward investment thesis is not just about residential broadband; it is about owning the last mile of infrastructure when the traffic traversing that mile becomes dominated by machine intelligence running on symmetric, low-latency connections.
Shareholder Returns: Buyback Raised to $10 Billion
Against that backdrop, AT&T's board approved an acceleration of its 2026 share repurchase program to approximately $10 billion, up from the $8 billion target set earlier in the year. Stankey said the decision reflected a gap between the company's operating fundamentals and its stock valuation — AT&T stock had declined roughly 12% in 2026 heading into the Wednesday report, weighed by concerns over wireless competition and satellite broadband competition. The company expects to repurchase nearly $1 billion of stock in July alone.
Full-year guidance remained intact: AT&T still targets adjusted EPS of $2.25 to $2.35 and free cash flow of at least $18 billion. The company's commitment to return $45 billion or more to shareholders through dividends and buybacks by 2028 also stands. Combined planned share repurchases and expected dividend payments are expected to total approximately $18 billion in 2026 — essentially the company's full free cash flow outlook for the year.
On the balance sheet, CFO Pascal Desroches noted that net leverage ended Q2 at 2.68 times adjusted EBITDA. That figure is expected to rise temporarily to approximately 3.2 times following the pending acquisition of spectrum licenses from EchoStar — a deal AT&T expects to close by the end of July 2026. AT&T agreed to purchase approximately 50 megahertz of nationwide spectrum for approximately $23 billion in an all-cash transaction — roughly 30 MHz of 3.45 GHz mid-band and 20 MHz of 600 MHz low-band. The FCC approved the transaction in May 2026. Management expects leverage to return to its target range of approximately 2.5 times within three years of closing.
The mid-band spectrum at 3.45 GHz is particularly significant: it covers the sweet spot between the raw speed of millimeter-wave 5G and the building-penetration depth of low-band, making it the preferred band for urban 5G capacity expansion and for supporting AT&T's fixed wireless access network. Acquiring it in 400-plus markets nationwide addresses the primary spectrum constraint on AT&T's ability to scale FWA alongside its mobile business without the two products competing for the same limited bandwidth.
What Stankey Got Right — and What He Did Not Say — About Starlink
The most consequential moment of Wednesday's earnings call was not a financial number. It was Stankey's characterization of Starlink — and whether that characterization holds under scrutiny.
Stankey described satellite direct-to-device connectivity as a solution for the roughly 2% of traffic AT&T cannot reach with its terrestrial network, framing it as a "corner case" and reiterating that an MVNO deal with SpaceX's Starlink Mobile would not be in AT&T's strategic interest. He argued that Starlink "arrives very late" and cannot easily replicate the decades of fiber and wireless infrastructure AT&T has built.
The 2% framing has quantitative support. An independent capacity analysis commissioned by the Fiber Broadband Association and conducted by research firm Cartesian estimated that Starlink's current constellation can serve approximately 2 million U.S. housing units — roughly 1 unit per square mile. The average U.S. suburban and urban density is approximately 30 housing units per square mile, which means Starlink's current spectrum capacity is insufficient to serve dense residential markets at scale. The study projected Starlink's U.S. serviceability rising to approximately 5.5 million housing units by 2030 as SpaceX deploys larger V3 satellites via its Starship rocket.
But Stankey's framing carries a material omission. Wells Fargo analyst Steven Cahall initiated coverage of AT&T with an Underweight rating and an $18 price target on July 8, calling AT&T "least likely" among the three major carriers to strike a Starlink Mobile MVNO deal and "most at-risk to net add and postpaid phone account share loss." Cahall estimated a 20% probability of an AT&T-Starlink MVNO versus 30% for T-Mobile and 40% for Verizon. Critically, Cahall modeled that Starlink's capacity expansion from Starship V3 launches would begin pressuring fixed wireless access net additions industry-wide by 2028 — a time horizon AT&T's current convergence momentum does not render irrelevant.
J.P. Morgan analyst Sebastiano Petti offered a more measured read: Starlink represents "more of a lingering overhang on U.S. wireless than a near-term fundamental threat," with SpaceX's U.S. terrestrial buildout modeled to begin in 2028 and a direct-to-consumer service not expected until 2029. TD Cowen analyst Gregory Williams, meanwhile, suggested that AT&T — given its dominant fiber network position — could be the carrier most likely to negotiate a wholesale network leasing arrangement with SpaceX, should any of the major carriers choose that path. MoffettNathanson analyst Craig Moffett flagged a separate pathway: SpaceX acquiring additional radio spectrum in a government auction planned for 2027, then using that spectrum as negotiating leverage for a wholesale network deal with one of the wireless carriers.
The core tension is structural. Stankey is correct that Starlink cannot immediately challenge AT&T at scale in dense residential markets. The density constraint is real. But his MVNO argument — that not striking a deal with SpaceX protects AT&T's strategic position — assumes that a carrier which declines to partner with a fast-growing connectivity platform is better positioned than one that does. Wells Fargo's analysis suggests the opposite: without a Starlink deal, AT&T's valuation is $16 per share; with one, approximately $27 per share, per Cahall's model.
The Q2 results make Stankey's case more credible in the near term. They do not resolve it in the long term.
The Competitive Landscape: Convergence Against a Moving Target
AT&T's blowout broadband figures arrive as the three major U.S. carriers compete aggressively for a finite pool of broadband customers. T-Mobile has expanded its own fixed wireless access product and continues to gain postpaid phone share. Verizon expanded its fiber footprint to nearly 30 million homes after its Frontier Communications acquisition closed in January 2026.
A less-discussed threat is coming from cable companies. Spectrum Mobile and Xfinity Mobile — cable MVNOs that buy wireless capacity from Verizon at wholesale and resell it under cable brands — are projected to account for more than 50% of postpaid phone net additions industry-wide by 2027 to 2028, according to analysis from Recon Analytics. That dynamic, more than Starlink's current constellation capacity, may represent AT&T's most immediate competitive pressure in postpaid wireless.
AT&T accelerated the decommissioning of its copper network during the quarter, with the legacy segment posting a 26% year-over-year decline in service revenue and a roughly 46% decline in EBITDA. The transition is painful on reported financials but necessary: copper infrastructure maintenance is capital-intensive and generates diminishing revenue as customers migrate to fiber and advanced voice services.
The Q2 results extended AT&T's streak to five consecutive quarters of beating analyst EPS estimates, with Q1 2026 having shown 294,000 postpaid phone net additions and 584,000 total internet net subscribers, also clearing analyst estimates, with adjusted EPS of $0.57 against a $0.55 consensus. The back-to-back momentum, combined with management's raised buyback target and reiterated full-year guidance, gives AT&T's bull case fresh evidence as the company heads into the second half of the year.
AT&T shares were trading at approximately $22.92 in morning trading on Wednesday, July 22, 2026.
Frequently Asked Questions
Is Starlink actually a threat to AT&T's business right now?
In the near term, the threat is limited by physics. An independent capacity analysis by research firm Cartesian estimated that Starlink's current satellite constellation can serve roughly 2 million U.S. housing units — about 1 unit per square mile — against an average U.S. suburban density of approximately 30 housing units per square mile. At current scale, Starlink cannot serve dense residential markets where most AT&T fiber and wireless customers live. That capacity is expected to grow as SpaceX deploys larger V3 satellites via its Starship rocket, with the study projecting serviceability of approximately 5.5 million U.S. housing units by 2030. The medium-term risk is more credible: Wells Fargo modeled that Starlink's expanding capacity would begin pressuring fixed wireless access net additions industry-wide beginning in 2028 — and the carrier that declines to strike an MVNO deal with SpaceX faces a valuation gap of more than $10 per share compared with the scenario in which a deal is reached, according to Wells Fargo analyst Steven Cahall.
What is AT&T's convergence strategy and why does it matter to subscribers?
AT&T's convergence strategy — branded through its OneConnect platform — bundles fiber home internet and 5G wireless service under a single account, billing relationship, and promotional discount structure. By the end of Q2 2026, 42.5% of AT&T's fiber internet customers also held an AT&T postpaid wireless plan (45% excluding the recently acquired Lumen customers). The subscriber benefit is a lower combined monthly cost than purchasing the services separately. The competitive benefit for AT&T is meaningful: research from Arthur D. Little documents that carriers executing this strategy in mature European markets have reduced annual customer churn rates by 30% to 50%, because canceling one service means disrupting both. At AT&T's scale of more than 100 million wireless connections, even a modest churn reduction compounds into substantial revenue retention over time.
How does agentic AI affect AT&T's infrastructure investment case?
Agentic AI — autonomous software agents that execute multi-step tasks by chaining API calls, running inference queries, and managing machine-to-machine workflows — generates approximately 450% more network traffic per task than a human performing the same work manually, according to a May 2026 Cisco report analyzing real-world WAN traffic patterns. Roughly 70% of that traffic is AI inference: machine-to-machine communication between the agent and the large language model powering it. That traffic is persistent, bidirectional, and latency-sensitive — properties that favor fiber's symmetric bandwidth and 5 to 20 millisecond latency over satellite's shared capacity and 25 to 60 millisecond latency. Enterprise network traffic without agentic AI is projected to grow 2.5 times over the next decade; with widespread agentic AI adoption, that figure rises to approximately 9 times. AT&T's fiber-and-5G investment thesis holds that the carrier with the densest, lowest-latency last-mile network benefits disproportionately from this demand surge.
Should AT&T shareholders be more concerned about cable MVNOs or Starlink?
Both represent genuine competitive pressures, but over different timeframes. Cable MVNOs — Spectrum Mobile and Xfinity Mobile, which buy wireless capacity from Verizon and resell it under cable brands — are projected to capture more than 50% of postpaid phone net additions industry-wide by 2027 to 2028, according to Recon Analytics. That represents near-term share pressure AT&T and its peers are already experiencing. Starlink's competitive threat to AT&T's core fiber and wireless business is more distant, gated by capacity constraints that limit its near-term serviceability to roughly 2 million U.S. housing units. The risk becomes more concrete beginning around 2028, when SpaceX is expected to begin a U.S. terrestrial buildout and its expanded satellite constellation begins putting pressure on fixed wireless access growth industry-wide. Whether AT&T's convergence momentum is durable enough to withstand both pressure sources simultaneously is the central question for investors in the second half of 2026.
Related Articles
