Starlink · SpaceX · Satellite Broadband · Direct-to-Cell
Starlink, Counted: The Filed Dominance, the Rivals, and the Mobile Question
SpaceX now files its numbers with the Securities and Exchange Commission: $11.387 billion of Starlink revenue in 2025, 13.6 million subscribers, two-thirds of everything active in orbit, and every class of customer in the file buying voluntarily. This report tests the claim that neither sovereign constellations nor mobile carriers can contest that position. The record carries it in broadband. In mobile it fails: the replacement network is an unclosed spectrum purchase, a 15,000-satellite license with zero satellites launched, and a bull forecast of 13 percent of the United States market by 2036.
The File in Brief
The question put to this desk was whether the satellite broadband business, and after it the mobile phone business, now belong to one company in a way no rival, commercial or sovereign, can contest, and whether that position rests on choice rather than compulsion. The record answers the two halves differently. On broadband, the position is filed, not asserted. SpaceX, public since June 12, 2026, reports audited Starlink revenue of $11.387 billion for 2025, 13.6 million consumer subscribers at September 30, 2026, and roughly two of every three active satellites in orbit. Its customers in every class, households, airlines, shipping lines, investors, and allied governments in wartime, buy voluntarily, at posted or negotiated prices, against alternatives or against nothing. No rival named in the record has a deployed consumer product, a funded route to SpaceX's launch economics, or a schedule that closes the gap this decade. On mobile, the record runs the other way. Starlink Mobile today is a coverage layer sold through the carriers it is said to replace; the replacement version requires a spectrum purchase that has not closed, a second-generation constellation that exists as a license and not as hardware, and phones that begin shipping in 2027. The strongest bullish forecast in the file, labeled a forecast, gives it 13 percent of the United States market by 2036. The thesis is carried on broadband and fails, on the current record, on mobile replacement. Both halves are documented below, with every forecast labeled and every number carrying its source.
The Dominance, Filed
One event reorganized the evidence in this file. SpaceX filed a Form S-1 on May 20, 2026 and listed on Nasdaq, ticker SPCX, on June 12, 2026. Since that date, the central figures of the Starlink business are reported figures, filed with the Securities and Exchange Commission, and the pre-filing estimates that once carried the argument can be checked against them. The filed series for consumer subscribers, defined in the S-1 as a unique service line on a starlink.com account and excluding managed enterprise and government customers on negotiated contracts, runs: 2.3 million at the end of 2023; 4.4 million at the end of 2024; 8.9 million at the end of 2025; 10.3 million at March 31, 2026; 12.0 million at June 30, 2026; and 13.6 million at September 30, 2026, with several hundred thousand enterprise and government customers additional. SpaceX's own broader count, which it calls active customers, runs slightly higher at each date and is not mixed with the filed series in this report. Growth at the margin is decelerating against expectations: third-quarter net additions of 1.6 million came in below the analyst consensus of roughly 2.29 million reported around the October 20, 2026 earnings release.
Revenue, as filed: Connectivity, the segment that is Starlink, took in $11.387 billion in 2025, 61 percent of SpaceX's consolidated $18.674 billion, with segment operating income of $4.423 billion. Consumer service supplied $7.208 billion of the segment; enterprise and government supplied $4.179 billion. The pre-filing estimates, Quilty Space at about $11.8 billion, Payload at $10.4 billion, Morningstar at $10.6 billion, Sacra at $11.4 billion, bracketed the filed figure. A growth rate of 83 percent for the segment circulates in secondary write-ups of the S-1; the segment tables support roughly 50 percent, and this report prints the lower figure. The service is sold, by the company's count, in 174 countries, territories, and other markets at September 30, 2026, of which its own availability file counts 133 of the 193 United Nations member states as launched. The distinction matters in section seven.
The most-quoted market figure in the file requires its caveat printed beside it. Ookla's 2025 Global Satellite Broadband Performance Report places Starlink at 97.1 percent of all global satellite Speedtest samples in the third quarter of 2025, with Viasat at 1.7 percent and HughesNet at 1.0 percent. Ookla states in the report that the findings provide relative guidance and scale, not exact customer figures. The number is a share of speed tests voluntarily run, not a share of subscribers, revenue, or traffic. Printed with its definition, it is consistent with everything else in this file. Printed as market share, it is not a measurement of the market.
Investors, the fourth class of customer, have opted in at scale and in the open. The valuation ladder, as reported: about $350 billion in a December 2024 tender, about $400 billion in July 2025, about $800 billion in an insider tender in December 2025, about $1.25 trillion for the combined entity after the all-stock acquisition of xAI in February 2026, and a price of $135 a share at the June 2026 initial public offering, a $1.77 trillion valuation on a $75 billion base deal, closing the first day at $160.95, up 19.2 percent, for a market value near $2.1 trillion. A greenshoe later took gross proceeds to about $86 billion. The public market now re-prices the position daily. Morningstar's discounted-cash-flow fair value for the company, about $780 billion, sits roughly 55 percent below the offering price; both figures are in the file, and the argument between them is now a market argument, conducted with money.
The Machine Underneath
The constellation, counted independently by astronomer Jonathan McDowell's tracking, the standard count in the field: at November 9, 2026, about 13,100 to 13,140 Starlink satellites launched all-time, 11,204 in orbit, about 11,190 working, and 9,704 in operational orbit, the difference being satellites raising, drifting, or lowering. About 1,900 have re-entered since 2019. Starlink is 65 to 66 percent of all active satellites in orbit. SpaceX's filed count at September 30, 2026 is over 11,100 broadband and mobile satellites. Direct-to-cell satellites are a subset of these counts, never additional: 674 launched, 637 in orbit, in late October 2026.
The production figures are filed or stated at their dates. Satellite production at Redmond, Washington ran at about 70 satellites a week on average from December 2025 through April 2026, as disclosed in the S-1. Terminal production at Bastrop, Texas was stated at 15,000 kits a day in March 2025. Launch cadence: 165 Falcon 9 orbital launches in 2025, a sixth consecutive annual record, 123 of them Starlink missions carrying more than 3,000 satellites, only 8 of the 165 on a new booster. Booster B1067 flew its 37th flight on August 25, 2026; the S-1 depreciates boosters over 25 flights. In the first nine months of 2026 the filings show 81 internal Starlink launches against 29 external Falcon launches. No price for a Starlink launch to itself is disclosed: the published customer price for a dedicated Falcon 9 is $74 million, the best on-record internal figure is Elon Musk's May 2020 statement of about $15 million marginal cost in the best case, and analyst ranges in 2026 run $15 million to $30 million per reused flight, labeled estimates. The structural fact does not need the precise figure: every rival in this file buys launch at customer prices, from someone, and Starlink does not.
The next capacity step is the file's clearest single point of dependence. The third-generation satellite, roughly 2,000 kilograms, designed for about 1 terabit per second of downlink capacity, about ten times the second generation, is too large for Falcon 9's fairing in useful numbers and depends on Starship. Starship has flown 14 times from April 2023 to September 28, 2026. Flight 14 was the first to reach orbit and the first revenue mission, deploying 26 operational third-generation satellites; the same flight had an engine shut down early on each stage, was shortened out of caution, and expended the ship in a Pacific splashdown. No Starship upper stage has been caught or reflown. Cumulative Starship spending exceeds $15 billion through 2025. Company guidance puts third-generation deployment in meaningful volume through the end of 2026 and 2027, with a few hundred satellites needed before customers feel the effect, possibly by mid-2027. The design target of 60 satellites per Starship launch is a target; 26 is the demonstrated number. Meanwhile the second-generation fleet, carried entirely by Falcon 9, is the whole of the operating business, and it is sufficient for the business as filed. The dependence is on the growth step, not on the present service.
The Rivals, One by One
Amazon Leo, renamed from Project Kuiper in November 2025, is the best-funded challenger in the file and the clearest schedule. It has launched 396 production satellites in 14 missions from April 2025 to July 2, 2026, 224 on eight Atlas V rockets, 72 on three Falcon 9s, 100 on three Ariane 6s, and none since July 2 through October. The service is not on general sale and has no published consumer price; an enterprise preview has run since November 2025, and terminals were still listed as awaiting FCC authorization on Amazon's site in late October 2026. Amazon's own January 2026 filing forecast about 700 satellites by July 2026; the actual figure was 396. The 2020 license required 1,616 satellites deployed by July 30, 2026. Amazon missed the deadline and asked for a 24-month extension. The FCC's order of June 5, 2026, DA-26-553, granted a limited conditional waiver instead: the license survives and the 2029 full-deployment deadline stands, but satellites launched after July 30, 2026 lose spectrum priority until the earlier of March 30, 2028 or 50 percent deployment. SpaceX opposed the waiver. Amazon has invested more than $10 billion and contracted 102 launches, the largest launch procurement in history, largely on rockets that had flown no Leo missions through October 2026; a New Glenn booster assigned to the first Leo mission exploded in a static fire on May 28, 2026. Amazon states initial service in first markets before the end of 2026; its mid-2026 commercial and Australian targets were already missed when it said so.
Eutelsat OneWeb is the only other operating global low-orbit network: about 654 satellites in orbit, about 648 active, live now, and sold only to enterprise, government, maritime, aviation, and backhaul customers through distribution partners. No consumer retail product exists. The current program is replenishment, 669 replacement satellites ordered from Airbus in tranches through September 2026, the first 32 delivered October 2, 2026, none launched by October 10. The group's financial state is printed in its own filings: revenue of 1,235.9 million euros for fiscal 2025-26, down 0.6 percent; a net loss of 457.3 million euros, the third consecutive annual deficit; low-orbit revenue of 297 million euros, about 24 percent of the group, against Starlink's filed $11.387 billion for the overlapping period. A 1.5 billion euro capital raise in late 2025 left the French state holding 29.65 percent, Bharti 17.88 percent, and the United Kingdom government 10.89 percent. Eutelsat dates OneWeb's continuity to the full commercial availability of the European sovereign system, IRIS², in mid-2032.
AST SpaceMobile is the direct-to-cell specialist. It has 12 working BlueBird satellites, five Block 1 from September 2024 and seven Block 2 from December 2025 to August 2026; one further Block 2 was stranded by a New Glenn upper-stage failure in April 2026 and de-orbited, a total loss part-covered by insurance. There is no continuous commercial service. The company deferred continuous service from late 2026 to early 2027 in July 2026 and now targets nationwide intermittent service in the United States and other markets by the end of 2026. Continuous coverage requires 45 to 60 operational satellites; the company targets about 45 in orbit by mid-2027. Its carrier position is contractual: a definitive AT&T agreement through 2030, a Verizon partnership with a $100 million commitment and a definitive agreement of October 2025, Vodafone through 2034, and a ten-year stc Group agreement with a $175 million prepayment. Cash at September 30, 2026 was $3.2 billion; third-quarter revenue was $42.3 million. In the same file, Starlink's filed 2025 revenue was $11.387 billion. AST also announced in November 2026 its own reusable rocket project, first flight targeted for late 2028, a target and not a date.
Two smaller files complete the commercial record. Lynk Global, the first direct-to-phone operator, has 12 satellites launched all-time with about six to eight working, providing intermittent store-and-forward text and emergency alerts through partners in island and rural markets; its public listing collapsed into litigation, and in August 2026 it merged with Omnispace and relaunched as Elveo Mobile, with SES as a strategic investor for an undisclosed amount. Its full constellation, cited variously at 640, 2,000, or 5,000 satellites, targets commercial service in 2027 and is not fully funded. Globalstar is the quiet incumbent: a second-generation constellation operating since the last decade, carrying Apple's Emergency SOS and Messages service since November 2022, with Apple taking 85 percent of network capacity under a wholesale agreement and committing $1.5 billion in November 2024. Its next-generation system of 48 plus 6 satellites is on order from MDA; the first eight replenishment satellites launched in August 2026; the FCC granted nationwide authority on October 8, 2026, with the 50 percent milestone not due until October 2032. In April 2026 Amazon agreed to acquire Globalstar for about $11.57 billion, expected to close in 2027 subject to antitrust and other clearances, and still pending in October 2026. If it closes, the second American satellite-to-device network becomes a subsidiary of the second broadband challenger.
Telesat Lightspeed, Canada's entry, has zero production satellites; two pathfinders are in assembly. Its financing closed in September 2024, its program expanded in August 2026 to 225 satellites at about 7 billion Canadian dollars, and its service date has moved from 2023 to late 2027 to the first quarter of 2028 to, in its October 30, 2026 update, mid-2028, with pathfinders no earlier than June 2027. All 225 satellites are claimed in orbit by the end of 2028 on 14 contracted Falcon 9 launches. The Canadian sovereign-adjacent system, like the American challenger, launches on SpaceX.
The Sovereign Projects
IRIS² is the European Union's flagship sovereign constellation and the test case for the sovereign half of the question. Deployed satellites: zero. Service: none. The concession was signed December 16, 2024 with the SpaceRISE consortium. An implementation agreement of August 7, 2026 expanded the system to 348 satellites, 330 in low orbit and 18 in medium orbit. The cost record: pitched at 6 billion euros with service in 2027 in 2022; 10.6 billion euros at the concession in December 2024; 15.6 to 15.7 billion euros as reported in August and September 2026, including a Le Monde investigation of September 11, 2026. The schedule record: first launches 2029; progressive initial services 2030; full constellation 2030 to 2032; commercial Ku-band service fully operational in mid-2032, per ESA and Eutelsat. The original 2027 service date has slipped three to five years on the record. IRIS² is a government secure-communications system, not a consumer broadband competitor; it enters service at the earliest six years after this report, into a market where the incumbent files $11.4 billion of annual revenue.
China SatNet's GuoWang is further along and further from service. The independent counts diverge and are printed as a range: 195 launched by September 14, 2026 on one tracker; about 222 in orbit after an October 9, 2026 batch on a second; 279 launched and 224 operational on McDowell's list at November 9, 2026. The plan, filed with the International Telecommunication Union in September 2020, is 12,992 satellites. No commercial broadband service exists and no Chinese satellite-internet license for the system was found as of September 2026; the operation is in testing. The reported deployment plan requires about 310 satellites by the end of 2026, about 1,300 by the end of 2029, and ITU milestones of about 1,299 by about September 2029 and about 6,496 by 2031-32, failing which the filing is reduced to the deployed count. The measured cadence from January through August 2026 was about 7.4 satellites a month. Trade analysis published in August 2026 concludes the 2026 and 2029 targets are out of reach without a sustained acceleration of about four times.
Shanghai Spacecom's Qianfan, or Spacesail, shows the same pattern in its own numbers. The May 2024 roadmap promised 648 satellites by the end of 2025 for regional service, 1,296 by the end of 2027 for global service, and 15,000 by the end of 2030. The actual count at the end of 2025 was 108. The restated 2026 roadmap, 324 by the end of 2026 and a first phase of 1,296 by 2027, is the 2024 plan shifted roughly a year and cut. McDowell counts 276 launched and 195 operational at November 9, 2026, with about 22 failures, including 17 early satellites stranded near 800 kilometers by thruster and gyroscope failures; launches paused for months and resumed in April 2026 after corrective work. No commercial service operates anywhere as of September and October 2026, and its first-quarter 2026 revenue was zero in its own capital-increase documents. Its international position is licensing-first: Brazil's Anatel authorized up to 324 satellites in February 2026 with commercial service targeted for the fourth quarter of 2026, and distribution frameworks are in place or in talks in Malaysia, Thailand, Kazakhstan, and about 30 countries. A Series B of about 7 billion yuan, about $1 billion, closed in August 2026 at a valuation of about 50 billion yuan, closed to foreign investors.
The Opt-In Ledger
The mechanism at the center of the commissioning question is that nobody in this ledger is conscripted. Each class of customer signs, and the file records the signatures. Consumers: 13.6 million of them at September 30, 2026 on the filed definition, each on a month-to-month retail subscription, supplying $7.208 billion of the filed 2025 segment revenue. There is no market in the file where Starlink is the compulsory provider; the subscription is bought against terrestrial alternatives where they exist and against nothing where they do not.
Aviation buys wholesale and gives the product away. United's September 2024 agreement covers more than 1,000 aircraft, the largest aviation connectivity agreement in the file; 635 or more aircraft were equipped by October 15, 2026, the service is free to MileagePlus members, and the full fleet is targeted for the end of 2027. Qatar Airways flew its first Starlink 777 in October 2024 and had 150 widebodies equipped by August 20, 2026, free gate-to-gate. Hawaiian was the first mover in February 2024 and had its entire Airbus fleet equipped by September 24, 2024; Alaska and Air France are on the same model, free behind a loyalty login. A dated compilation counts 21 airlines with at least one Starlink aircraft carrying passengers at October 26, 2026; it is a compilation, not a single primary source, and is filed as such. Maritime signed earlier and wider: Carnival Corporation completed rollout across 100 percent of its fleet, more than 90 ships, in May 2024; Royal Caribbean, Norwegian, MSC, Virgin Voyages, and Viking are also fleet-wide; Maersk equipped 330 or more owned containerships after a 30-ship pilot in October 2023; Hapag-Lloyd is fleet-wide; Mitsui O.S.K. Lines and Anglo-Eastern each run 200 or more vessels.
Government and military purchasing is the class where the mechanism is most often misdescribed, and the file is specific. Starshield, the government product line announced in December 2022 and running in part over the Starlink constellation, holds a first Space Force contract of September 1, 2023 with a $70 million ceiling for 54 mission partners, and a classified National Reconnaissance Office imaging contract of $1.8 billion signed in 2021 and reported by Reuters in March 2024 from five sources; the NRO has never named the vendor. SpaceX told investors on August 6, 2026 that it had won more than $6 billion in multi-year United States government Starshield contracts. Separately reported Space Force awards of May 2026, $2.29 billion and $4.16 billion, may overlap that figure and are not added to it in this report; the Wall Street Journal has reported more than $8 billion of Golden Dome awards to SpaceX in 2026. Circulating totals from procurement aggregators are not Department of Defense releases and are not used here.
Ukraine is the hardest case and the cleanest ledger. The first batch, March and April 2022, was 5,175 terminals under a USAID partnership: USAID purchased 1,508 and SpaceX donated 3,667, per the USAID Inspector General; USAID spent about $3 million on hardware and delivery. The scale grew to more than 20,000 terminals by August 2022, about 47,000 delivered by December 2023, and about 42,000 operating in March 2025, dated figures from sources that conflict across dates and are printed with their dates. SpaceX's own letter to the Pentagon of September 8, 2022 states that about 85 percent of the 20,000 terminals then in Ukraine had been paid for in whole or part by others, Poland, the United States, the United Kingdom, and non-governmental organizations, while SpaceX carried most of the service cost on the remainder. Poland is the largest contributor: about 9,000 terminals initially, 19,500 of the 47,000 delivered by December 2023, and funding for more than 20,000 terminals at about $50 million a year by 2024-25, per the Polish Digital Ministry. The Pentagon confirmed a contract with Starlink for Ukraine on June 1, 2023, price and scope withheld; Bloomberg reported 400 to 500 Pentagon-bought terminals and a December 2024 extension moving 2,500 terminals onto Starshield atop 500 already connected. In the hardest case in the file, the service was bought by allied governments through ordinary procurement after an initial donation phase. The donation phase did not last. The purchasing did.
The Closed Doors
The map has edges, and they are legal, not technical. Starlink's 174 markets are the markets that let it in. Russia has not certified the service and cannot legally be used; an April 29, 2026 government decree formally banned the import of foreign satellite terminals. China has granted no authorization; unlicensed foreign satellite communication is illegal, and maritime reporting in 2026 records fines on ships using Starlink in Chinese waters. Iran has not authorized it; possession is criminalized, smuggled terminals operate under jamming that cut throughput by up to about 80 percent in January 2026, and Starlink's own file lists Iran as coming soon, which in this context is a functional ban. North Korea has never permitted it. About 1.5 billion people live under states that bar the service outright. In those markets the sovereign systems face no competition, which is a different kind of having a chance: a captive one, held by law rather than won in a market.
Three open-market files show the friction in detail. India has licensed and not launched: a GMPCS license in June 2025, an IN-SPACe constellation authorization on July 8, 2025, distribution agreements with Jio and Airtel from March 2025, and 20 gateway sites built; as of October and November 2026, commercial spectrum assignment and Home Affairs security clearance were still pending, at the same stage as Jio Satellite and Eutelsat OneWeb. The binding constraint is India's process, not the applicant's readiness. South Africa has no license and no application: the Electronic Communications Act requires at least 30 percent ownership by historically disadvantaged groups for individual network licensees; the minister's December 2025 policy direction proposing an equity-equivalent alternative was answered by the regulator in May 2026 that a legislative amendment is required; a parliamentary answer of June 2026 states that ICASA has received no application from Starlink. SpaceX's offered substitute is an equity-equivalent package of about 500 million rand including free connectivity for more than 5,000 rural schools, offered June 14, 2025. Brazil shows a third instrument: in August 2024 a Supreme Court justice froze Starlink's Brazilian finances to secure fines owed by X, transferred about 11 million reais, about $1.9 million, from a Starlink account on September 13, 2024, and unfroze the accounts; Starlink complied with the X blocking order within a day after the regulator threatened its license. The service was never banned. A sovereign reached the company's money through its corporate sibling in a market where it operates legally, and the company complied.
Starlink Mobile, As It Stands
The mobile file must be read separately, because the numbers change character. The first-generation direct-to-cell constellation is complete: more than 650 satellites launched in 18 months per SpaceX's Progress Report 2025, 637 in orbit in late October 2026. The partner count in wide circulation is wrong in the direction of understatement: the Progress Report states 27 mobile network operator partners in 22 countries on six continents, with more than 400 million people in reach and more than 6 million monthly direct-to-cell customers. Twenty-two is the country count. Named partners include T-Mobile in the United States, Rogers in Canada, One NZ, Optus and Telstra in Australia, KDDI in Japan, Kyivstar in Ukraine, Salt in Switzerland, and Entel in Chile and Peru.
The flagship service, T-Mobile's T-Satellite, launched commercially on July 23, 2025 with text messaging, picture messaging, short audio clips, and satellite text-to-911 across more than 500,000 square miles of uncovered American territory, on any recent LTE phone. Curated application data followed on October 1, 2025: WhatsApp, Google Maps, AllTrails, AccuWeather, X, onX, and T-Life, with WhatsApp voice and video working inside the application. Native cellular voice is not live in late 2026; T-Mobile's own documentation lists voice calling as roadmap, and general browsing and streaming are not offered over handset satellite data. Pricing is inclusion on T-Mobile's top plans and $10 a month per line for everyone else, including AT&T and Verizon customers, by eSIM; text-to-911 is free to anyone on a compatible phone. The product, as it stands, is a coverage layer sold through a carrier, including to the customers of that carrier's rivals.
The replacement version rests on three dated objects. First, spectrum: the EchoStar purchase announced September 8, 2025, about $17 billion in cash and stock for 50 megahertz, amended in November 2025 to add up to 15 megahertz of unpaired AWS-3 for about $2.6 billion more, a package of about $19.6 billion and up to 65 megahertz. The FCC approved the assignment on May 12, 2026; the licenses moved into a trust on May 22, 2026; the transaction has not closed, final closing is targeted for November 30, 2027, and the FCC bars commercial use of the bands until closing. SpaceX had paid $856 million toward EchoStar debt interest by September 30, 2026. Second, a constellation: FCC Order DA 26-1078 of October 6, 2026 authorizes 15,000 next-generation satellites in nine shells at 326 to 335 kilometers, parts deferred, with milestones of 50 percent launched and operating by October 7, 2032 and 100 percent by October 7, 2035. It is a license ceiling with conditions, not hardware; none of the 15,000 has launched. Third, the regulator's wider frame: on October 29, 2026 the FCC voted 3 to 0 to propose expanding the Supplemental Coverage from Space framework and replacing its spectrum-lease requirement with a certification model, a proposal that opens a comment period and grants nothing, with final rules requiring a second vote. SpaceX's own-spectrum service, described by the company as 5G-quality, is a company target of early 2028, slipped from the end of 2027 on the August 4, 2026 earnings call. Compatible phones for the new frequencies begin shipping in about 2027 on Elon Musk's own September 2025 statement of a two-year chipset timeframe, not 2028-29 as widely repeated; the binding constraints are the spectrum closing and the unbuilt constellation.
The Hurdles, at Full Strength
The counter-case begins with the partner. T-Mobile's chief financial officer, Peter Osvaldik, speaking at investor conferences in August and September 2026, states that a satellite about 350 kilometers up arrives too weak to penetrate buildings, walls, or, in his phrase, even a Tesla windscreen reliably; that about 70 percent of mobile use is indoors or in modern vehicles; and that direct-to-device was 0.0002 percent of T-Mobile's network usage in its busiest summer months, a company figure reported by Light Reading, with some write-ups printing 0.002 percent and the company's range treated here as 0.0002 to 0.0003 percent. The attenuation figures cited from Ericsson Mobility Report data are 5 to 10 decibels for glass, 20 to 30 for concrete, 25 to 40 for low-emissivity glass, on top of about 51 decibels of additional path loss against a tower one kilometer away. A tower-industry-funded white paper of July 2026, its funding flagged here as the record requires, makes the same point from the handset side: a phone transmits under half a watt, omnidirectionally.
The independent measurement in the file agrees. García-Cabeza and colleagues at the Universidad Politécnica de Madrid and Weplan Analytics, working from crowdsourced United States measurements from October 2024 to July 2025 and published in IEEE Communications Magazine in June 2026, measured about 3 megabits per second per beam for data service outdoors, shared by everyone in a footprint tens of kilometers across, with median satellite signal 24 decibels below terrestrial on the same devices. Expansion to about 18 megabits per beam is conditional on regulatory and business decisions. The geometry is the argument: terrestrial networks reuse the same spectrum every few hundred meters and a satellite beam cannot. The spectrum arithmetic runs the same way. After EchoStar and the reported Grain 800 megahertz transaction, SpaceX holds about 76 to 80 megahertz of mobile-suitable spectrum, against about 400 megahertz for T-Mobile alone and about 1,000 megahertz across the three American carriers, figures stated by rival chief executives in October and November 2026 and corroborated in structure by the July 2026 industry analysis.
Two money figures in circulation fail checking, and the file prints the corrections. A widely repeated claim of $900 billion in additional financing for Starlink Mobile traces to a Pivotal Research note of October 12, 2026 in which the $900 billion covers SpaceX's broader expansion across satellite communications, launch services, and artificial intelligence infrastructure over a decade; the same note's mobile model is cumulative capital expenditure of $30.6 billion from 2027 to 2036, $22.5 billion of spectrum and $8.1 billion of satellites. A syndicated truncation rewrote the figure as mobile infrastructure cost, and the truncation, not the note, is what circulated. Rival chief executives' counter-estimates are likewise labeled interested: Verizon's Dan Schulman put a terrestrial-competitive build at $300 to $500 billion over seven to eight years on October 26, 2026; Deutsche Telekom's Tim Höttges said at least $200 billion on November 5, 2026. Musk replied on October 26, 2026 that it will take two or three years. None of the three numbers is a neutral costing, and this report uses none of them as one.
The Market-Size Fight
The two market sizings in the file do not measure the same thing, and the report prints them as what they are. SpaceX's $1.6 trillion connectivity total addressable market is in its IPO filings, broken down by Dell'Oro as about $870 billion of broadband and about $740 billion of mobile; it aggregates essentially the whole connectivity market outside China and Russia. Morningstar's Testing the Sky's Limits of June 8, 2026 sizes the realistic addressable market at about $129 billion as of 2025, about $84 billion of niche demand, rural broadband and remote business, and about $45 billion of carrier-partnership add-on, with no material share of core telecommunications, on the stated ground that satellite connectivity faces physical and technological constraints that are very difficult to overcome in a dense urban environment. Morningstar forecasts, and it is labeled a forecast, that Starlink revenue grows about 7.5 times by 2035, capturing about 45 percent of that ex-core market. One number sizes the ocean. The other sizes the part of the ocean a satellite can physically swim in.
The bullish mobile forecast, labeled, is Pivotal Research's of October 12, 2026, analyst Jeffrey Wlodarczak: Starlink Mobile reaches 160 million global subscribers by 2036, 50 million in the United States and 110 million international, equal to 13 percent of the United States mobile market and 1.5 percent international outside China; blended average revenue of about $22 a month; subscribers beginning in late 2027 and 2028; EBITDA positive in 2030; cumulative capital expenditure of $30.6 billion from 2027 to 2036; service built in part through Starlink terminals acting as small cells; initial target rural and suburban, with, in the note's terms, more limited opportunities in densely populated urban markets, and indoor calls initially reliant on Wi-Fi. Pivotal raised its SpaceX price target to $230, Buy, on the note. Even the bull case in the file is a 13 percent share, decade-long, Wi-Fi-assisted case. It is not a carrier replacement case, and its author does not present it as one.
The Verdict on the Record
The commissioning question asked whether sovereign constellations and mobile carriers have a chance against Starlink, and whether its position is one the world opts into. On broadband, the record carries the affirmative. The dominance is documented in audited filings rather than asserted in marketing: $11.387 billion of 2025 revenue, 13.6 million filed subscribers and rising by more than a million a quarter, two-thirds of everything active in orbit, a production line building 70 satellites a week and 15,000 terminals a day, and a launch operation that flew 165 times in 2025 while its nearest challenger bought tickets on it. The opting-in is documented class by class: households on month-to-month subscriptions, 21 airlines, whole cruise fleets, the world's largest container line, $75 to $86 billion of primary capital at a public offering, and allied governments paying through ordinary procurement for terminals in an active war, 85 percent of them paid by someone other than the operator. Against this, the commercial rivals have 396 satellites and no consumer price, 648 satellites and no consumer product, or 12 satellites and no continuous service; the sovereign projects have zero satellites and a 2032 service date in Europe, about 200 to 280 satellites and no license in China, and a Canadian system launching on the incumbent's rockets in 2028. No entry in the file closes the gap this decade.
On mobile replacement, the record does not carry it. What exists is a coverage layer of genuine value, text and curated data where no tower stands, sold through carriers at $10 a month. What is claimed beyond that depends on a spectrum transaction that closes, on the present schedule, on November 30, 2027; on a 15,000-satellite constellation authorized in October 2026 of which zero units have launched; and on physics that the operator's own launch partner states in public: a signal that does not reliably pass a windscreen, a beam shared across tens of kilometers at about 3 megabits per second measured, and 76 to 80 megahertz of spectrum against a thousand. The strongest honest version of the mobile thesis is Pivotal's: 13 percent of the American market by 2036, rural and suburban first, Wi-Fi carrying the indoor call. That is a business. It is not the end of the carriers, and the carriers' own conduct in the file, partnering, reselling, and selling the service to each other's customers, prices it as a business.
Two qualifications stand on the record beside the verdict. The closed markets are not a footnote: about 1.5 billion people live where the service is barred by law, and there the sovereign systems hold their markets the way monopolies held theirs before competition existed, by decree. And the broadband lead, while carried by the record, carries two documented exposures: growth decelerating at the margin against the expectations priced into a $1.77 trillion offering, and the next capacity step riding on a rocket that has reached orbit once, expended its ship doing it, and has never reflown a stage. The record as of November 12, 2026 supports a narrower and stronger sentence than the one proposed. In broadband, Starlink's dominance is structural, filed, and bought voluntarily by every class of payer in the file, and no rival in the record is positioned to displace it this decade. In mobile, it is a tenant and a wholesaler to the industry it is said to be ending, with a licensed path to something larger that opens, at the earliest, in 2028.
Sources and Method
This report was compiled from the primary record where the record exists in primary form: SpaceX's Form S-1 of May 20, 2026 and its 2026 quarterly filings; FCC Orders DA 26-1078 of October 6, 2026 and DA-26-553 of June 5, 2026 and the Commission's October 29, 2026 rulemaking; EchoStar's September 8, 2025 Form 8-K and its third-quarter 2026 10-Q; Eutelsat's fiscal 2025-26 results; the USAID Inspector General's Ukraine report; and Jonathan McDowell's independent satellite tracking. Company statements are attributed to the company. Analyst figures are attributed to the analyst and dated. Every forecast in the report is labeled a forecast at the point of use, with its maker named: Quilty Space, Payload, Morningstar, and Pivotal Research. Subscriber figures use the Securities and Exchange Commission definition from the S-1 throughout; SpaceX's broader active-customer count is identified where it appears and the two series are never mixed. Satellite counts print their definitions, in orbit, working, and in operational orbit being three different numbers. Where sources conflict, the conflict is printed rather than resolved by preference. Two claims in wide circulation were checked against their sources and corrected in the body: the partner count for direct-to-cell, which is 27 operators in 22 countries, and the $900 billion financing figure, which its own author's note assigns to SpaceX's whole expansion, with $30.6 billion modeled for mobile. The report is the desk's analysis of that record, stamped objective; the file the desk read is clipped below, each item stamped subjective and described by what it says.