AT&T's $250B Bet: Building the Highway Under America's AI Economy

AT&T just committed $250 billion over five years to become the infrastructure layer beneath AI workloads, while splitting its financial reporting to separate a fast-growing 5G/fiber engine from its declining copper legacy. With 10.4 million fiber subscribers, a new segment reporting structure, and the largest convergence network in the U.S., the 150-year-old company is executing the most aggressive network-infrastructure play in telecom.

T · Communication Services · March 28, 2026

S&P 500 Position

AT&T sits within the Communication Services sector, which comprises about 11% of the S&P 500. Within telecom specifically, AT&T ($204B market cap) is the largest U.S. integrated telecom operator, slightly ahead of Verizon (~$190B) and behind T-Mobile (~$280B) in market cap. T-Mobile leads in wireless subscriber momentum and has SpaceX satellite exclusivity; Verizon is in restructuring mode under new CEO Dan Schulman, cutting capex to $16-16.5B. AT&T is the only player aggressively building both fiber and wireless at scale simultaneously, giving it the most differentiated convergence positioning in the sector.

Index Weight: ~0.40% | Rank: Approximately #50-60 in the S&P 500 by market cap

Company Overview

AT&T is executing on a convergence thesis that no other U.S. carrier can replicate at the same scale. The company operates both the nation's largest fiber network — now covering 32 million locations with a target of 60+ million by 2030 — and a nationwide 5G wireless network covering over 322 million people. The strategic logic is simple: fiber backhauls wireless, wireless extends fiber's reach, and converged customers (42% of AT&T Fiber households now also subscribe to wireless) churn less and generate higher ARPU. This dual-network ownership gives AT&T what CTO Yigal Elbaz calls 'owner economics' — the ability to control costs and convergence points that competitors leasing infrastructure cannot match. The $250 billion five-year commitment announced in March 2026 formalizes the capex trajectory implied by two major 2025 deals: the $23 billion EchoStar spectrum acquisition (adding 3.45 GHz mid-band and 600 MHz low-band nationwide) and the $5.75 billion purchase of Lumen's mass-markets fiber business. Beginning with Q1 2026 results, AT&T will report under two new segments — Advanced Connectivity (5G + fiber, ~90% of revenue) and Legacy (copper voice/DSL) — giving investors a clean view of the growth engine versus managed decline. COO Jeff McElfresh has framed this as positioning AT&T to become 'the highway upon which commerce and AI workloads traverse.' That is the pitch: AT&T is not building AI models — it is building the physical network fabric that AI inference, autonomous systems, and edge compute require. The competitive positioning is favorable. T-Mobile leads in wireless subscriber momentum but lacks a meaningful fixed-line network. Verizon has fiber (Fios) but far less geographic reach and is cutting capex under new CEO Dan Schulman. AT&T is the only carrier building both at scale, and the regulatory environment — aided by the One Big Beautiful Bill Act's tax provisions — is accelerating the buildout.

Products & Revenue

AT&T's revenue engine runs on two pillars: wireless service revenue ($70.1B in FY2025, ~56% of total) and a fast-growing fiber broadband business embedded within Consumer Wireline. Mobility drives the majority of cash flow, with 120 million subscribers (91 million postpaid). Consumer Wireline fiber revenue hit $8.6B in 2025, growing 17% YoY — this is the highest-growth product in the portfolio and the centerpiece of the convergence strategy. Business Wireline continues to shrink as enterprise customers migrate off legacy TDM/VPN services, though advanced fiber and connectivity services within that unit grew 6.8%. The Latin America segment (Mexico wireless, ~24.7M subs) is small but growing. Equipment revenue (~$24.5B) represents device sales that flow through Mobility but are low-margin pass-through.

Mobility (Wireless Services + Equipment) (~75%): Nationwide 5G and LTE wireless service to 120M subscribers, plus handset/device sales. Service revenue grew 2.4% YoY in Q4 2025; equipment revenue surged 12.7% on higher device volumes.

Consumer Wireline (~11%): Fiber-to-the-home broadband (10.4M subscribers), AT&T Internet Air (fixed wireless access), and legacy DSL/voice. Fiber revenue grew 17% in FY2025. EBITDA margins expanded 320bps YoY to 38.4% in Q4.

Business Wireline (~11%): Enterprise ethernet-based fiber, IP voice, managed services, VPN, and legacy data services. Revenue declining 7-8% annually as legacy contracts roll off, partially offset by fiber/advanced connectivity growth of 6.8%.

Latin America (Mexico) (~3%): Wireless services in Mexico serving approximately 24.7 million subscribers. Growing subscriber base (+4.7% YoY) with favorable FX tailwinds in 2025.

Based on AT&T FY2025 10-K filing (period ending December 31, 2025). Revenue percentages are approximate based on reported segment data. Beginning Q1 2026, AT&T will report under new 'Advanced Connectivity' and 'Legacy' segments.

Leadership

John Stankey

CEO since 2020. Stankey has spent his entire 40-year career at AT&T, holding roles as CIO, CTO, Chief Strategy Officer, and CEO of WarnerMedia before becoming CEO in July 2020. He was elected Chairman in February 2025. He led the strategic pivot away from media (spinning off WarnerMedia, selling DirecTV) to refocus entirely on connectivity, and has overseen the $23B EchoStar spectrum deal and $5.75B Lumen fiber acquisition that define the current growth thesis.

Jeff McElfresh, Chief Operating Officer: 27-year AT&T veteran who previously ran AT&T Communications and the Technology & Operations group (90,000+ employees). He is the operational architect of the $250B investment plan and the convergence strategy, driving the split into Advanced Connectivity vs. Legacy reporting segments.

Yigal Elbaz, Network Chief Technology Officer: Leads network architecture and technology strategy across 5G, fiber, Open RAN, and satellite direct-to-cell. Currently chairs the ATIS Board of Directors. Driving the push to move service convergence from the core backbone to the network edge using cloud-native infrastructure.

Pascal Desroches, Senior EVP & Chief Financial Officer: Oversees the financial architecture behind AT&T's capital return program ($45B+ to shareholders through 2028) and the new segment reporting framework designed to separate growth metrics from legacy decline.

Shawn Hakl, SVP, AT&T Business Products: Leading AT&T's enterprise 'Connected AI' and edge compute strategy, including partnerships with Nvidia, Cisco, and AWS for industrial IoT and smart manufacturing use cases. Driving the agentic AI and IoT connectivity narrative.

Thaddeus Arroyo, Chief Strategy & Development Officer: Oversees corporate strategy, M&A, and business development. Played a key role in structuring the Lumen fiber acquisition and EchoStar spectrum deal that anchor the growth roadmap through 2030.

The AI Angle

AI runs on AT&T's pipes, not its servers

AT&T's AI strategy operates on two planes: using AI to run its own network more efficiently, and positioning its network as the essential transport layer for enterprise AI workloads. Internally, the company has deployed 'Ask AT&T,' a GenAI platform built initially on OpenAI's ChatGPT but interoperable with Meta's LLaMA and open-source Falcon models. Ask AT&T now has 100,000 internal users generating roughly 5 billion tokens per day, used for everything from coding assistance to contract analysis to deep research. The company built this in-house to keep sensitive data off third-party platforms — all LLMs are trained on AT&T's own internal data. Beyond Ask AT&T, the company uses NVIDIA CuOpt for field technician routing optimization, has built a full digital twin of its physical network (enabling GenAI to deploy workflows without building domain-specific telecom models), and runs AI-driven 'cell site sleep' algorithms that cut energy consumption 20-30% at individual sites. AT&T Active Armor, its consumer security product, uses ML to block or label over 2 billion robocalls per month. On the infrastructure side, AT&T is positioning itself as the connectivity backbone for enterprise AI at the edge. At MWC 2026 and NVIDIA GTC 2026, the company launched 'Connected AI,' a platform unifying 5G, IoT, and generative AI for smart manufacturing. It partnered with Cisco and NVIDIA on an edge AI solution combining AT&T's dedicated IoT core and RAN with Cisco's Mobility Services Platform and AI Grid. A preview of a last-mile enterprise AI connectivity offering with AWS is also in progress. SVP Shawn Hakl described the strategy as harmonizing AT&T's converged network with hyperscalers, neoclouds, and LLM makers so enterprise customers can route data from edge devices directly into their compute facilities of choice. AT&T is also a founding supporter of the GSMA's Open Telco AI initiative, contributing models trained on public telecom data to build purpose-built, open-source AI models for the industry — an acknowledgment that frontier LLMs are not built for telecom-specific network operations. CTO Elbaz noted that AI is already changing network traffic patterns, with AI wearables and autonomous vehicles driving upstream-heavy traffic ratios exceeding 50% during training phases, and agentic AI creating more continuous, unpredictable 'east-west' traffic flows. The risk is that AT&T remains a transport layer — essential but low-margin — while value accrues to hyperscalers and model makers sitting on top. The opportunity is that if AI inference workloads move to the edge as predicted (dwarfing centralized training loads), AT&T's 36-million-location fiber network and converged 5G architecture become the choke point that every enterprise AI deployment has to pass through. CTO Elbaz sees demand for 400G enterprise services already growing, and the $250B commitment is explicitly designed to meet this wave.

Financial Snapshot

Revenue (TTM): $125.6B — TTM (period ending Dec 31, 2025) | Net Income: $21.9B net income (includes $5.6B gain on DirecTV sale; adjusted net income closer to ~$15B)

Margins: Gross ~60%, operating 19.2%, net 17.4% (GAAP, elevated by DirecTV gain)

AT&T delivered its best year in a decade for broadband subscriber growth in 2025, with full-year adjusted EPS of $2.12 (up 9%) and free cash flow of $16.6B. The company returned over $12B to shareholders through dividends and buybacks — a 50%+ increase from 2024. Looking forward, management guides to $45B+ in shareholder returns through 2028, adjusted EPS of $2.25-$2.35 in 2026 with a double-digit 3-year CAGR, and capex of $23-24B annually. Cash taxes are expected at just $1-1.5B annually through 2028 due to the One Big Beautiful Bill Act, creating a significant FCF tailwind.

1-Year Performance

$29.10 as of March 28, 2026. The stock hit an all-time closing high of $29.10 on March 27, sitting near its 52-week high of $29.79.

AT&T shares have climbed roughly 27% from their 52-week low of $22.95 (hit January 27, 2026), driven by a strong Q4 2025 earnings beat ($0.52 adj. EPS vs. $0.47 consensus), the $250B investment announcement, and multiple analyst upgrades (Goldman Sachs raised target to $33, KeyBanc to $36). The stock has been a rare outperformer in 2026's risk-off environment, benefiting from its defensive dividend yield (~3.8%) and the perception that telecom infrastructure spending is a durable secular trend insulated from tariff and macro volatility.

Recent News

Fun Fact: AT&T's Bell Labs invented not just the transistor (1947) but also Unix (1969), the C programming language (1972), and information theory itself — Claude Shannon's 1948 paper was published as a Bell Labs employee. Today, AT&T's internal GenAI platform 'Ask AT&T' processes roughly 5 billion tokens per day across 100,000 employees — more daily token throughput than many standalone AI companies — all running on infrastructure kept entirely in-house to protect intellectual property, with interoperability across OpenAI, Meta LLaMA, and Falcon models so the company can hot-swap LLM backends per use case.