US Telecommunications: Broadband Competition and Household Access

E0. Market boundary: household connectivity, with company scope preserved

This audit studies the supply and purchase of internet connectivity in the United States, concentrating on fixed broadband and its interaction with mobile subscriptions. Fibre, cable and fixed wireless belong in the same household decision because each may provide a primary home connection. A mobile phone subscription may complement or sometimes replace that connection, but a mobile line is not a household. Business connections enter only where a provider’s published category includes them and the table says so.

The unit of competition is ultimately an eligible service address and a usable offer. National provider names do not establish that a household has several alternatives. Similarly, an additional subscription can represent switching, an additional residence, a business connection or a second service; it need not represent a newly connected household. The selected company panel therefore supports comparisons of reported commercial trajectories, not estimates of national penetration or market share.

Three boundaries govern the evidence. Company operational stocks and flows retain their original customer definitions. Financial analysis uses identified domestic service categories rather than adding group revenues with international or media businesses. Household and adult surveys supply participation context, with their populations and dates kept separate. The primary company records are AT&T’s schedule (S02), Charter’s results (S03) and Comcast’s domestic tables (S04).

E1. Market base: divergent trajectories and an acquisition break

At June 2026, the panel contains growing fibre and fixed wireless bases alongside declining cable broadband counts. The rows below are intentionally not summed: some include business connections, while others cover residential customers only.

Sources: AT&T S02, printed p. 8; Charter S03, Key Operating Results; Comcast S04, customer metrics. Stocks are end-quarter; net additions are quarter flows. The Charter subset must not be added to its total.

A longer consumer-only fibre series demonstrates why a single year-on-year growth rate is insufficient.

Source: AT&T S02, printed pp. 8 and 16, footnote 12.

Calculated from this series, 2025 consumer fibre growth was 11.5%, or 1.075 million net additions. During the first half of 2026, the stock rose by 1.738 million but reported net additions totalled 617,000. The 1.121 million residual is a stock-flow reconciliation item, not a precise acquired-customer estimate. AT&T separately disclosed a rounded 1.1 million Lumen fibre acquisition base adjustment. Treating the entire stock rise as customer acquisition through ordinary sales would materially overstate operating momentum.

The counterevidence is that consumer fibre net additions remained positive in every displayed quarter, including after the acquisition. The acquisition changes the scale of reported growth; it does not explain away the underlying positive flow. Conversely, this panel cannot establish how much fibre or fixed wireless growth came directly from cable customers. That requires switching or address-level evidence.

E2. Household costs: a headline price is a conditional subtotal

The macro environment gives mixed signals. August 2026 consumer prices were 3.4% above a year earlier, with electricity up 3.8%; July real disposable personal income increased 0.4% month-on-month. These observations describe household budget context after the company reporting quarter. They are not explanations of Q2 subscription changes. BLS S10, August CPI release; BEA S11, July income release.

The following calculations use public offers inspected on 18 September 2026, without supplying a test address. They illustrate billing structure across technologies; they do not rank equivalent services or claim nationwide availability.

Sources: AT&T S12, offer cards and footnotes; Spectrum S13, plan and WiFi terms; T-Mobile S14, Rely and connection-charge terms. Calculations are 12 times the stated monthly charge, plus an identified one-time charge. Unknown charges are unresolved, not zero. T-Mobile account-dependent provider fees and local taxes are excluded from these subtotals. Conditional reward cards are excluded. T-Mobile notes that AutoPay may not appear on the first bill, so its subtotals assume a full year of qualifying discounts rather than a guaranteed first invoice sequence.

The actionable distinction is between the cost to start service, the recurring first-year cost and the price after the promotion. A household can face different constraints at each stage. The paired T-Mobile rows also show why a discounted bundle cannot be compared with standalone internet without retaining the associated voice bill. Spectrum’s selected WiFi option adds $120 over a year; a household supplying an appropriate router would face a different basket. These are transparent scenarios, not observed average expenditure.

E3. Institutions: transparency helps only when the offer remains legible

The FCC’s December 2022 broadband-label rule supplies a useful institutional benchmark: its pricing provisions distinguish standalone price, introductory and subsequent rates, billing discounts, recurring charges and one-time fees. This audit uses that adopted framework to organise the cost comparison; it does not claim to reproduce every subsequent rule change or the present status of every programme named in the original rule. FCC/Federal Register S15, paragraphs 13–22.

The economic mechanism is reduced search cost. A household can compare offers more reliably when discounts and connection costs are separated from the service price. Disclosure alone, however, cannot create an eligible alternative at an address or make an unaffordable payment manageable. Public national advertising also does not exhaust eligible offers: T-Mobile describes a $15 monthly home-internet option for qualifying New York customers. It is a state-specific offer, not a national tariff or a claim that all applicants qualify. T-Mobile S17, New York eligibility FAQ.

Acquisition accounting creates a parallel transparency issue for industry analysis. A change in reported customer scale may reflect a transfer of ownership rather than new capacity or household adoption. Competition assessments should therefore date ownership changes, separate inherited bases from operating additions and avoid treating an announced transaction as completed. This audit makes no national concentration estimate.

E4. Earnings: mobile growth cushions broadband pressure unevenly

A direct revenue decomposition is more informative than comparing total corporate growth. The domestic residential categories below isolate the fixed–mobile relationship, while retaining each company’s reporting definitions.

Sources: Charter S03, residential revenue table; Comcast S04, Residential Connectivity & Platforms table.

The ratio of mobile service revenue gain to fixed broadband revenue decline is 90.2% for Charter and 33.9% for Comcast. This is an original arithmetic decomposition, not an estimate of mobile profitability or the causal effect of bundling. Charter’s two components decline by $19 million in combination. Comcast’s rounded components imply a $244 million decline, while its reported convergence total declines by $243 million; the $1 million discrepancy is retained as a source-rounding difference.

The strategic implication is conditional. Adding mobile revenue can protect a customer relationship, but a growing line count alone does not establish that a convergence strategy has preserved revenue, margin or cash generation. Retention discounts, wholesale network costs, device costs and customer-service expenses matter to that test; comparable incremental margins were not established here.

Investment mix also matters. Charter’s upgrade/rebuild capital expenditure rose from $457 million to $657 million while total capital expenditure was broadly flat at approximately $2.87 billion. The calculated category increase is 43.8%. A stable headline investment budget can therefore conceal substantial reprioritisation. Charter S03, Capital Expenditures.

E5. Participation: high connectivity averages conceal different usable access

Census QuickFacts reports that 91.0% of households had a broadband subscription and 95.5% had a computer in the 2020–2024 five-year ACS estimate. These are pooled period measures, not observations of September 2026. The broad broadband category must not be read as fibre or cable adoption, and device ownership does not measure whether a household has enough suitable devices for its activities. Census S07, Computer and Internet Use.

More recent primary survey evidence uses a different denominator. Pew’s 2025 NPORS surveyed 5,022 adults between 5 February and 18 June through web, mail and telephone, with population weighting. It reports 78% subscribing to home broadband. Smartphone dependence was 34% among adults in households earning below $30,000 and 4% among those earning at least $100,000: a calculated 30-percentage-point difference between published estimates. The audit does not test statistical significance or infer an income effect from that association. Pew S09, methodology and income table.

Historical Census evidence adds a dimension not captured by price alone: in 2021, broadband subscription was reported for 84% of households including a person with a disability, versus 92% of households without one. Those figures are context, not a current gap estimate. Census S08, printed p. 6.

The commercial and participation questions consequently differ. Winning an already connected household can raise a provider’s subscriptions without reducing exclusion. Conversely, effective entry-level service may improve participation without maximising speed-tier revenue. Suitable devices, usable interfaces and support deserve separate measurement. The available surveys identify where to investigate those constraints; they do not establish which intervention will work or how many subscriptions it will generate.

E6. Delivery: eligibility, activation and support shape the effective market

The customer journey links an advertised offer to an actual connection: address eligibility, plan selection, payment requirements, installation or gateway setup, in-home performance and continuing support. Failure at any stage can make a nominal alternative unusable. The cost basket shows payment and equipment conditions; it does not measure completion rates at those stages.

T-Mobile’s FAQ explicitly ties unlimited home-internet eligibility to the service address and network capacity, which can change as customers join. It also distinguishes provider fibre delivered through fibre partners from its fixed wireless service and describes a gateway-return requirement after cancellation. These are concrete examples of a supplier and service chain that extends beyond the retail brand. T-Mobile S17, eligibility, fibre and cancellation sections.

The operating question is how many eligible enquiries become working, retained connections at acceptable support cost. Address approval and the advertised price precede that outcome; gateway setup, in-home performance and help with faults determine whether the connection remains useful. Network passings alone cannot measure this conversion from infrastructure to sustained service.

E7. Competition: substitution is conditional and convergence works in both directions

The selected evidence supports a narrower and more useful conclusion than a national winner ranking. Fibre and fixed wireless are credible sources of additional home connections, while cable operators can expand in mobile. In Q2 2026 Charter added 406,000 mobile lines and Comcast 448,000 domestic wireless lines even as their broadband counts fell. Mobile lines and broadband customers cannot be netted against each other to produce a household-growth figure. Charter S03; Comcast S04.

T-Mobile’s latest inspected year-end disclosure provides historical scale: 8.5 million 5G broadband customers at December 2025, with 495,000 quarterly additions; it also separately identifies 141,000 acquired 5G broadband customers in Q3 2025. Its Q2 2026 headline addition of 277,000 refers to postpaid accounts, not home-broadband customers. T-Mobile S06, year-end results; T-Mobile S05, Q2 results.

Substitution should be tested on four conditions: eligibility at the same address, adequate performance for the household’s workload, a complete lifecycle bill and manageable switching costs. A service can pass one condition and fail another. This framework also makes the competition claim falsifiable: if households cannot order an alternative, or its performance and total cost do not meet their needs, national subscription growth does not establish effective local rivalry. No representative address panel was assembled here.

E8. Technology: advertised speed does not settle service equivalence

The price table deliberately retains different technologies and speed tiers. T-Mobile describes Rely’s typical download range as 170–354 Mbps and upload range as 12–55 Mbps, with performance depending on local conditions. Its terms also identify lower priority during congestion and further deprioritisation above 1.2 TB per month; that threshold is not presented as a hard data allowance. T-Mobile S14, Rely speed and network-management terms.

Those terms matter to substitution because a household’s constraint may be upload performance, simultaneous use or reliability rather than the maximum advertised download speed. The correct comparative question is whether the service meets a defined task at the times it is needed. A national average, a top-tier headline and a typical range answer different questions.

Cable also remains an evolving technology in this comparison. Charter states a 2027 target for completing its network evolution towards symmetrical and multigigabit services. That is a company target, not evidence that every current location already receives those capabilities. Charter S03, network evolution discussion. The observed spending shift in E4 makes the upgrade programme economically material, while completion, orderability and realised customer experience remain separate tests.

E9. Outlook: three observable paths through September 2028

The next 12–24 months should be assessed through repeated observations rather than a single market-growth forecast. The scenarios below are analytical possibilities, with no assigned probabilities.

A practical quarterly monitor would retain the same company category definitions, separate acquisitions and track the combined revenue decomposition. A complementary household monitor would record first-year and post-promotion bills at the same addresses, including payment and equipment conditions. Changes in survey participation should be evaluated only after confirming the new sample, question wording and reference period. Improving supplier results alone would not establish broader inclusion; improving adoption alone would not establish sustainable supplier economics.

E10. Critical success factors: what would validate or overturn the strategy

The audit identifies four testable priorities. Each connects at least two evidence modules, rather than restating a generic industry aspiration.

These indicators are a research and operating agenda, not claims that the relevant outcomes have already been measured. The first priority has a current financial baseline; the latter three require stronger address-level, service-quality and subgroup evidence before causal conclusions are justified.

E11. Assessment, method and limits

The United States evidence supports an industry assessment centred on the quality of competition and the economics of convergence. The selected providers exhibit different subscription trajectories, and the mobile revenue offset is measurable but incomplete. Household evidence shows why the resulting competitive activity cannot be treated automatically as an inclusion outcome. The analytical contribution is the connection between those three objects—provider growth, household price conditions and usable participation—while preserving their different units.

The audit uses purposive primary-source selection, not a statistically representative provider sample or a systematic literature review. Original publisher pages and tables were inspected through 18 September 2026. Company data extend through June 2026, except the expressly dated T-Mobile fixed wireless observation at December 2025. The AT&T consumer fibre sequence covers ten consecutive quarters. Census QuickFacts supplies a 2020–2024 period estimate; the 2021 Census subgroup report supplies historical context; Pew supplies separate 2025 adult-survey evidence. The inspected Census update of 6 August 2026 stated that the 2025 ACS one-year release date was still being determined. Census S16, August update.

Calculations are descriptive and reproducible: differences, ratios, stock-flow reconciliation and specified price subtotals. They are neither econometric estimates nor claims of causal switching. Published values retain source rounding. Financial categories are not harmonised beyond their stated scope, and non-GAAP measures should not be equated with accounting profit. Offer observations are public snapshots without address-level checkout, so they cannot establish a final bill, serviceability or a cheapest-provider ranking.

The principal unresolved evidence needs are a complete national supplier denominator, comparable local availability and performance, current household subgroup tables, representative installation, complaint and supplier-cost observations, and longitudinal switching and bill data. These limits preclude a national market-share table, a competition index, a national affordability rate and a causal estimate of fixed wireless displacement. The most valuable extension is an address-based panel linking orderability, actual costs, service experience and retention, followed by properly comparable household participation updates.

Funding and competing interests

Funding: No report-specific funding declaration is recorded.

Competing interests: No report-specific declaration is recorded.

Data availability and reproducibility

Download the observations (CSV), calculations (JSON) and source register (JSON). Source identifiers US-S01–US-S17 correspond to S01–S17 in the references. Observations retain periods, entities, units and locators; calculations separate inputs and outputs, preserving the acquisition and rounding differences. Original publishers retain responsibility for their reports and offers; public access does not imply an open redistribution licence. Use the correction form to report errors.