> ## Content Index > Fetch the complete content index at: https://www.satelliteinsights.com/llms.txt > Use this file to discover other available public pages before exploring further. # Satellites Are the Cheap Part: Sizing SpaceX's Mobile Network - URL: https://www.satelliteinsights.com/spacexs-site-problem/ - Published: 2026-08-18T13:00:55.000Z - Updated: 2026-09-09T12:27:17.000Z - Description: SpaceX's mobile network is not limited by spectrum or satellites. It needs about 40,000 cell sites, and that buys a mid-market carrier. - Author: Glenn Canales - Tags: Starlink, Direct to Cell, #Deep Dive Contents [·Executive summary](#summary)[01What did SpaceX actually announce?](#s01)[02Why does frequency reuse decide this?](#s02)[03What are the limits of reuse from orbit?](#s03)[04Where does the 65 megahertz actually sit?](#s04)[05How large is the gap between demand and satellite supply?](#s05)[06Does the rooftop femtocell close the gap?](#s06)[07What would it take to carry a quarter of all US mobile traffic?](#s07)[08How many paying customers does 65 megahertz support?](#s08)[09What kind of carrier does this spectrum make?](#s09)[10Can a cable partnership solve the site problem?](#s10)[11What would have to be true for this to work?](#s11)[12What this means depending on where you sit](#s12)[13What to watch](#s13)[14What would change this view](#s14)[ Methodology and sources](#sback)[ Disclosure](#sback) Sources - Federal Communications Commission (FCC) filings, DA-25-197 and DA-26-36 - Cellular Telecommunications and Internet Association (CTIA) Annual Wireless Industry Survey 2025 - Starlink published V3 specifications - SpaceX, Mobile World Congress 2026 and the 4 August 2026 earnings call - Crown Castle and Zayo transaction disclosures, May 2026 - TMF Associates; Aetha Consulting Every figure pulled from a live search at time of writing. Assumptions are named as assumptions. Paid edition Research · Direct-to-Device # Satellites Are the Cheap Part Sizing SpaceX’s Mobile Network SpaceX says it will build a mobile network out of satellites and rooftop femtocells. Sized end to end, the network is limited by neither spectrum nor satellites. It needs 37,000 to 62,000 cell sites, and that buys a mid-market carrier rather than a fourth national network. Research report · **18 August 2026** · Glenn Canales, Principal · Satellite Insights LLC 447,605 US cell sites, year-end 2024 65 MHz SpaceX holdings post-EchoStar 25 MHz Of that, facing downlink 37–62k Cells for 20 million subscribers Executive summary **SpaceX's mobile network is constrained by cell sites, not by spectrum and not by satellites.** That conclusion comes from sizing the network end to end rather than arguing about megahertz, and it inverts the framing most coverage of the 4 August 2026 earnings call has taken. **1\. Capacity growth on Starlink is reuse growth, and nothing else.** The V3 satellite carries about ten times the capacity of V2 and has 10.7 times the beams. Per-beam throughput actually fell slightly over that generation, from roughly 500 to 488 megabits per second. **2\. Reuse from orbit is capped three ways.** By beam count rather than antenna size, by a fixed solar power budget divided across however many beams are formed, and by an aggregate emission limit that forces satellites to switch beams off as the constellation grows. SpaceX described the third of these to the Federal Communications Commission (FCC) in its own filings. **3\. The 2,048-beam figure is the broadband payload, not the mobile one.** The 65 megahertz is reused per cell on the ground and per beam from a mobile payload of roughly 768 beams at 2 gigahertz. The two are not interchangeable, and most coverage of this story treats them as one number. The ground layer delivers roughly two and a half times the capacity of the entire national direct-to-device (D2D) constellation, which is where the commercially relevant reuse happens. **4\. The constellation is the cheap part.** A 20 million subscriber network, roughly twice the current Xfinity Mobile base, needs about 290 satellites at an estimated 900 million dollars. **5\. The site layer is not.** That same business needs 37,000 to 62,000 cell sites at 1.9 to 3.1 billion dollars. Roughly 70 to 80 percent of the capital and effectively all of the execution risk sit in the terrestrial layer, and none of it is a space problem. **6\. Sixty-five megahertz is sufficient for that business.** Twenty million subscribers at 30 dollars a month is near 7.2 billion dollars a year of revenue against a build of 2.8 to 4.0 billion. Spectrum is not the binding constraint at any plausible commercial scale. **7\. Sizing the same network against national traffic share gives much larger numbers, and they are a ceiling rather than a target.** A quarter of all United States mobile traffic would need roughly 2,400 satellites and 308,000 to 513,000 cells, at 22.7 to 33.0 billion dollars. But a quarter of national traffic is about 165 million subscribers, close to 29 percent of every wireless connection in the country, and no operator is a plausible candidate at that scale. The 20 million subscriber business in findings 4 to 6 is 3 percent of national traffic. The two scenarios differ by a factor of 8, which is why traffic share is the wrong denominator for a business case and appears in this report only to establish the capacity envelope. **7\. What the thin holding cannot do is serve heavy users.** The median American consumes 14.5 gigabytes a month while the top decile exceeds 52 and accounts for close to 40 percent of all traffic. Twenty-five megahertz of downlink serves the first group well and the second not at all, which makes this a mid-market network by construction rather than by choice. **8\. The fastest route to the site layer is somebody else's poles.** Cable strand, or the independent small-cell platform spun out of Crown Castle in May 2026, supply pre-permitted mounting points with power and fiber already at them. Continues below ## Fourteen sections, five exhibits, the full sizing tables Satellites, gateways, cell sites, subscribers and capital, sized end to end. Plus the partnership analysis, what kind of carrier this spectrum actually makes, what it means depending on where you sit, and what would change this view. - 01What did SpaceX actually announce? - 02Why does frequency reuse decide this? - 03What are the limits of reuse from orbit? - 04Where does the 65 megahertz actually sit? - 05How large is the gap between demand and satellite supply? - 06Does the rooftop femtocell close the gap? - 07What would it take to carry a quarter of all US mobile traffic? - 08How many paying customers does 65 megahertz support? - 09What kind of carrier does this spectrum make? - 10Can a cable partnership solve the site problem? - 11What would have to be true for this to work? - 12What this means depending on where you sit - 13What to watch - 14What would change this view - Methodology and sources - Disclosure [Upgrade to read the full report](https://www.satelliteinsights.com/#/portal/signup) For paid members. Upgrade in a minute, and every Deep Dive and research report comes with it. Contents [·Executive summary](#summary)[01What did SpaceX actually announce?](#s01)[02Why does frequency reuse decide this?](#s02)[03What are the limits of reuse from orbit?](#s03)[04Where does the 65 megahertz actually sit?](#s04)[05How large is the gap between demand and satellite supply?](#s05)[06Does the rooftop femtocell close the gap?](#s06)[07What would it take to carry a quarter of all US mobile traffic?](#s07)[08How many paying customers does 65 megahertz support?](#s08)[09What kind of carrier does this spectrum make?](#s09)[10Can a cable partnership solve the site problem?](#s10)[11What would have to be true for this to work?](#s11)[12What this means depending on where you sit](#s12)[13What to watch](#s13)[14What would change this view](#s14)[ Methodology and sources](#sback)[ Disclosure](#sback) Sources - Federal Communications Commission (FCC) filings, DA-25-197 and DA-26-36 - Cellular Telecommunications and Internet Association (CTIA) Annual Wireless Industry Survey 2025 - Starlink published V3 specifications - SpaceX, Mobile World Congress 2026 and the 4 August 2026 earnings call - Crown Castle and Zayo transaction disclosures, May 2026 - TMF Associates; Aetha Consulting Every figure pulled from a live search at time of writing. Assumptions are named as assumptions. Full report **unlocked**. Fourteen sections and five exhibits. Section 01 ## What did SpaceX actually announce? **SpaceX said it will build a terrestrial mobile network, not simply a better satellite one.** On 4 August 2026, on its first earnings call since June's initial public offering, SpaceX told investors it intends to build a terrestrial mobile network and take customers from the three national carriers. President Gwynne Shotwell said the company will build out the terrestrial component of the spectrum it is acquiring from EchoStar, declined to give a capital expenditure figure, and suggested the enclosure that holds a Starlink broadband dish could instead house "these little femtocells" deployed as demand appears. Elon Musk described the same architecture from the other end: Starlink dishes on the roofs of houses and businesses, radiating in the mobile bands, with a clear path down to phones at street level. Coverage since has settled on two questions. How much spectrum does SpaceX have, and how many satellites can it launch? Both are answerable and both are secondary. The variable that decides whether a satellite network can substitute for a terrestrial one is spatial reuse, and it is the one number that does not improve with launch cadence. Section 02 ## Why does frequency reuse decide this? **Capacity is bandwidth multiplied by the number of times that bandwidth is independently reused, and reuse is where orbit loses to the ground by roughly three orders of magnitude.** Figure 1 · The same megahertz, reused once or many times over One beam shared by everyone beneath it SATELLITE One beam covers about 3,400 sq km TERRESTRIAL About 566 sectors fit that same ground Both panels are the same patch of ground. The grid is drawn at 55 cells for legibility against roughly 566 actual, so it understates the gap by about tenfold. Across the contiguous United States the counts are near 14,600 satellite beams against 1.34 million terrestrial sectors, a ratio close to 92 to 1\. Sites: CTIA Annual Survey 2025, year-end 2024\. Beam count derived from published per-satellite figures. Capacity in any radio network is the product of three terms: the bandwidth in use, the spectral efficiency of the air interface, and the number of times that same bandwidth is independently reused across the service area. Operators compete on the first two by a factor of two or three. They win or lose on the third by a factor of thousands. The United States had 447,605 cell sites in operation at year-end 2024, of which 166,264 were small cells, according to the most recent Cellular Telecommunications and Internet Association (CTIA) annual survey. Most macro sites are sectorized, which puts the number of independent reuses of a given block of spectrum on the order of a million. A low Earth orbit satellite beam covers thousands of square kilometers and reuses that block once. Beam width is set by orbital altitude and antenna aperture. Launching more spacecraft adds beams, which helps, but it does not shrink the footprint of any individual one. The clearest statement of this constraint came from SpaceX. At Mobile World Congress in March 2026, Starlink senior vice president Michael Nicolls positioned the service as complementary to terrestrial networks, telling the audience that satellite cannot provide the data density terrestrial networks have but can augment them where terrestrial cannot reach or needs additional capacity. Section 03 ## What are the limits of reuse from orbit? **Beam count sets the ceiling, not antenna size, and power and regulation cap it further.** SpaceX has now published enough hardware detail to close this argument arithmetically rather than argue it by analogy. The Version 3 broadband satellite supports 2,048 downlink beams and 2,048 uplink beams, against 192 downlink and 144 uplink beams on Version 2, and is designed for 1 terabit per second of downlink and 160 gigabits per second of uplink. Divide it out. Version 3 delivers about 488 megabits per second per downlink beam. Version 2, at the 96 gigabit figure SpaceX has used, delivered about 500\. Beam count rose 10.7 times. Capacity rose about 10 times. Per-beam throughput went slightly down. The entire generational leap is reuse. Not a better waveform, not higher order modulation, not more spectrum per beam. SpaceX multiplied the number of independently steerable spots and capacity followed one for one. On this architecture, capacity is beam count. The mobile payload is a separate and much smaller machine. Current direct-to-cell satellites reportedly carry three downlink antennas and one uplink antenna, each producing eight beams across two polarizations, for 48 downlink beams on a phased array of roughly 25 square meters. That figure comes from secondary technical coverage rather than a filing and should be treated as directional. For the next generation, Nicolls gave three multipliers: an array five times larger, sixteen times the beams, and four times the bandwidth per beam, using 20 by 20 megahertz channels with multiple-input multiple-output and the 5G New Radio non-terrestrial network standard. Two things fall out of those multipliers. First, sixteen times 48 is 768 downlink beams, spread across a field of view of roughly 2.6 million square kilometers, which is a beam about 66 kilometers across. A five-times-larger array is about 11 meters on a side, and at 2 gigahertz that could form a beam roughly 9 kilometers across. The antenna is not the limit. The beamformer is. SpaceX could make the spots five to seven times finer and does not. Second, sixteen times the beams multiplied by four times the bandwidth is 64 times the capacity. SpaceX claims 20 times the throughput per satellite. That shortfall, a factor of about 3.2, is the power budget showing up in the company's own arithmetic. A phased array splits one fixed power supply across however many beams it forms, so doubling the beams halves the power in each and costs spectral efficiency. This is the asymmetry that decides the argument. When a carrier splits a cell, it installs a new site with its own power feed, and reuse and power scale together. When a satellite splits a beam, it subdivides a power budget fixed by the solar array, and reuse and power scale against each other. Terrestrial densification is additive. Orbital densification is a zero-sum division of a fixed resource. There is a regulatory ceiling above the physical one. Section 25.202(k)(1) of the Federal Communications Commission (FCC) rules sets an aggregate out-of-band emission power flux density limit of -120 dBW/m2/MHz across all satellites in view, not per spacecraft. SpaceX won a waiver to -110.6 dBW/m2/MHz in March 2025, confined to the two 5 megahertz segments adjacent to its downlinks at 1985 to 1990 and 1995 to 2000 megahertz. Asked how it would comply without that waiver, SpaceX told the Commission that approaching roughly 1,500 satellites it would need compensatory measures including "reducing the number of active beams on a satellite," reducing beam power, or both, and estimated a resulting network throughput reduction of about 20 percent. Because the limit aggregates across the constellation, launching more direct-to-cell satellites does not deliver proportional capacity over a given patch of ground. Past a threshold it forces each satellite to turn beams off. Section 04 ## Where does the 65 megahertz actually sit? **The 65 megahertz is reused per cell on the ground and per beam from the mobile payload. It is not the band the 2,048-beam broadband figure describes.** One point of architecture is worth fixing before the numbers, because conflating it is the most common error in coverage of this story. The 2,048 beams and the 1 terabit per second are the **broadband payload**, operating in Ku, Ka, E, V and W band. The direct-to-phone service is a **different payload** at 2 gigahertz with roughly 768 downlink beams, and that is where the 65 megahertz lives. The two beam counts are not interchangeable, and applying the broadband figure to the mobile link overstates direct-to-phone capacity by roughly a factor of three. Whether the two payloads ride the same spacecraft bus is a separate question, and the answer is probably yes. Today's direct-to-cell satellites are already a variant of the V2 Mini platform rather than a distinct vehicle, and SpaceX states that technology developed for V3 will support its forthcoming mobile satellites. A V3-derived mobile variant is the likely path, and SpaceX has a separate pending application for a dedicated constellation of 15,000 satellites for direct-to-cell operation. None of that changes the arithmetic. What matters is that the mobile link has its own beam count, its own band and its own capacity, whatever bus carries it. Nor does any of this mean the broadband payload sits outside the mobile business. It is central to it. Every rooftop small cell fed by a Starlink dish is backhauled by the broadband payload, which is the architecture Section 06 examines. The claim here is narrow and specific: the 65 megahertz does not ride the broadband link, and the broadband beam count does not describe the mobile one. Figure 2 · Where the 65 megahertz actually sits Mobile payload2 GHz, 65 MHz per beam Broadband payloadKu, Ka, E band Rooftop dishbackhaul only Small cell65 MHz per cell Beam reuse is also weaker than cell reuse for three reasons already established above. Beams from a single aperture interfere with each other in a way ground cells do not, because distance, terrain, buildings and antenna downtilt isolate terrestrial sectors while a handset antenna cannot discriminate between overlapping beams from 530 kilometers. Beam reuse divides a fixed power budget, while every new ground cell brings its own power feed. And beam reuse is capped in aggregate by the out-of-band emission rule, so launching more spacecraft does not multiply usable beams over a given patch of ground indefinitely. Section 05 ## How large is the gap between demand and satellite supply? **Designed satellite capacity over the United States sits one to two orders of magnitude below measured national mobile demand.** Americans consumed 132 trillion megabytes of mobile data in 2024 across 579 million wireless connections, per the CTIA survey published in September 2025\. That averages to roughly 33.5 terabits per second sustained. Applying a peak-to-average ratio of 1.5 to 2.5 puts busy-hour demand between 50 and 85 terabits per second. That figure is measured traffic, and it has grown at about 32 percent annually. Figure 3 · Terabits per second, demand against satellite supply US mobile, busy hour estimate67.0 TbpsUS mobile, 2024 average33.5 Tbps15,000-satellite D2C fleet23.8 TbpsStarlink broadband, today7.1 Tbps1,200-satellite coverage shell1.9 Tbps Demand, measured Supply over the contiguous US, designed Demand derived from CTIA 2024 traffic. Satellite figures are designed specifications from SpaceX statements at Mobile World Congress 2026, apportioned by surface area share. On the supply side, Nicolls put the second-generation mobile spacecraft at more than 100 gigabits per second of downlink, with roughly 1,200 satellites required for contiguous global coverage. These are designed figures presented at a conference, not demonstrated throughput. The contiguous United States is about 1.6 percent of the Earth's surface, so an initial coverage shell puts the equivalent of about nineteen satellites' worth of capacity over the country at any moment. That is a capacity share rather than a headcount visible from any one location, and it is deliberately conservative, since satellites can steer beams toward land and away from ocean. From a single address the numbers are much smaller: roughly six satellites of that shell sit above a 25 degree elevation mask, about two above the 40 degree mask a handset link realistically needs, and only one of them is serving the phone at a time. Scale to 15,000 mobile satellites, the size of SpaceX's separate pending application for a dedicated direct-to-cell constellation, and the designed figure over the contiguous United States reaches roughly 24 terabits per second. Note that this is a different 15,000 from the Gen2 broadband cap the FCC authorized in January 2026, which is the figure used in Sections 06 and 07\. Every apportionment in this report uses that same pure area share, so the figures are comparable across sections and conservative throughout. Two caveats belong on that arithmetic. Apportioning by surface area understates the American position, because inclined orbits dwell longer at mid latitudes than a uniform split implies. Call that a factor of 1.3 to 1.5 in SpaceX's favor, which moves nothing about the order of magnitude. Second, the demonstrated side of the ledger is far behind the designed side: a crowdsourced measurement study of the current service estimates per-beam throughput at about 3.1 megabits per second. That is a single-sourced academic estimate and should be treated as directional. Satellite counts are the wrong unit for anyone trying to picture the service, though. What determines a subscriber's experience is the beam they happen to sit in. A second-generation mobile beam covers about 3,400 square kilometers, roughly 66 kilometers across. At the contiguous United States average population density of 42 people per square kilometer, that beam contains around 144,000 people. Over a suburb at 400 per square kilometer it contains 1.4 million. Over an urban core at 2,000, close to 7 million. All of them share one beam's designed 130 megabits per second. Take the average-density case and assume only one percent of those people are active at once. That is roughly 1,400 users on 130 megabits, or about 90 kilobits each. At three percent active, 30 kilobits. Those figures are illustrative rather than a service prediction, since real deployments concentrate beams where subscribers are, but they explain in one step why a terrestrial layer is not optional. No amount of spacecraft fixes a number that is set by how many people fit inside a 66 kilometer circle. Section 06 ## Does the rooftop femtocell close the gap? Figure 4 · Both paths draw on the same space segment Starlink satelliteone shared capacity pool Rooftop Starlink dishacts as the backhaul Rooftop small cell2 GHz, at street level Phonesame user The small cell changes which link carries the bits, not which resource pays for them. Architecture as described by SpaceX on the 4 August 2026 earnings call. Fiber-backhauled small cells escape this constraint; dish-backhauled ones do not. **Not against today's constellation. Against a mature V3 fleet, largely yes, and that is a correction to the version of this argument in circulation.** The terrestrial layer is what buys the reuse back, and on the face of it the rooftop plan is a reasonable way to get it. Buy back the reuse factor on the ground, where cells are small and numerous, and use the satellite only where nothing else reaches. The question is whether the backhaul the company described can carry it. If the small cell is fed by the Starlink dish it is bolted to, the traffic still transits the space segment. The architecture has moved the bottleneck from the direct-to-device link to the Ku-band and Ka-band feeder link and picked up spectral efficiency along the way. Whether that solves anything depends entirely on which constellation you are talking about. Against the current fleet it does not. Starlink's total constellation capacity is around 450 terabits per second globally across roughly 10,930 satellites in the active catalog as of 17 August 2026, of which about 10,700 are counted as working, and the share physically over the contiguous United States at any instant is roughly 7 terabits per second, already carrying a paying broadband base with congestion pricing in dense markets. Against a mature Version 3 fleet the picture changes materially, and this is where the popular version of the argument, including an earlier version of my own, gets it wrong. A full 15,000-satellite Version 3 deployment would put on the order of 240 terabits per second over the contiguous United States, about one and a half times the entire projected national busy-hour mobile demand for 2028\. Ku-band and Ka-band backhaul to a fixed directional antenna is a far easier link than 2 gigahertz to a handset, and Version 3 is where SpaceX has put its engineering. So the honest verdict splits. The reuse argument holds for the direct-to-phone link, and the published beam counts strengthen it. The claim that satellite backhaul can never feed a meaningful small-cell layer is a statement about today's constellation, not about the architecture. Section 07 ## What would it take to carry a quarter of all US mobile traffic? **Roughly 2,400 satellites and 308,000 to 513,000 cell sites. This section sizes a capacity ceiling, not a business plan: a quarter of national mobile traffic is about 165 million subscribers, close to 29 percent of every wireless connection in the country. The figures a commercial operator would actually build to are in Section 08 and are eight times smaller.** Both sets of numbers appear in this report and they answer different questions. The distinction is worth fixing before the tables: | | Subscribers | Share of national traffic | Cell sites | V3 satellites | | ------------------------ | ----------- | ------------------------- | ------------------ | ------------- | | **A commercial target** | 20 million | 3 percent | 37,000 to 62,000 | about 290 | | **The capacity ceiling** | 165 million | 25 percent | 308,000 to 513,000 | about 2,400 | Traffic share is the wrong denominator for a subscriber business. No operator is a plausible candidate for 29 percent of every connection in the country, and the largest carrier in the United States does not hold that share today. The envelope is worth establishing anyway, because it bounds the problem and shows where the capital sits. Read the rest of this section as the ceiling. Section 08 gives the floor a real business needs. Assume the competitive service is built exclusively on Version 3 satellites feeding rooftop small cells. Rolling 2024 traffic forward at decelerating growth of 30, 25, 20 and 18 percent gives roughly 304 trillion megabytes in 2028, about 77 terabits per second sustained and 154 at busy hour. That extrapolation is the author's, not a published forecast. The three segments size very differently. **Space segment.** At 1 terabit per second per satellite and roughly 1.6 percent of the fleet over the contiguous United States at any instant, which is the pure area share used throughout, the requirement is modest. Inclined orbits dwell longer at mid latitudes than a uniform split implies, so allowing 1.3 to 1.5 for that effect would cut these fleet figures by about a quarter. The conservative figure is used here. Manifest sensitivity matters, because 60 satellites per Starship flight is a designed figure. The demonstrated figure is 20, deployed on a suborbital trajectory on Flight 13 on 24 July 2026 and intentionally deorbited about twenty minutes later. No Version 3 satellite has yet reached an operational orbit. | Share of US mobile traffic | Tbps over the contiguous US | V3 fleet required | Flights at 60 each | Flights at 20 each | | -------------------------- | --------------------------- | ----------------- | ------------------ | ------------------ | | 5 percent | 8 | 486 | 8 | 24 | | 10 percent | 15 | 972 | 16 | 49 | | 25 percent | 39 | 2,431 | 41 | 122 | | 50 percent | 77 | 4,861 | 81 | 243 | | 100 percent | 154 | 9,722 | 162 | 486 | A quarter of all US mobile traffic needs about 2,400 satellites, well inside the 15,000 already authorized. On a five-year design life that fleet also carries a standing replacement burden of roughly 490 satellites a year, which is 24 flights annually at the demonstrated manifest just to hold the constellation flat. **Ground segment.** Version 3 carries 1.2 terabits per second of radio frequency backhaul across four quad-band antennas, roughly 300 gigabits per antenna link. Landing 39 terabits per second requires about 128 simultaneous links, or roughly 320 antennas allowing 2.5 times for handover and geographic diversity. SpaceX disclosed more than 100 United States gateway sites carrying over 1,500 antennas in July 2025\. Even full national replacement lands near 1,300 antenna equivalents. The binding item is fiber, at roughly 9 terabits per second into a single gateway site, which is a serious transport buy at twenty to forty locations but an ordinary procurement. **Terrestrial layer.** Assume 25 megahertz of downlink and a rooftop small cell achieving 5 bits per second per hertz in good conditions or 3 in mixed. | Share of US mobile traffic | Cells at 125 Mbps each | Cells at 75 Mbps each | Share of all Big Three sites | | -------------------------- | ---------------------- | --------------------- | ---------------------------- | | 5 percent | 61,600 | 102,700 | 14 to 23 percent | | 10 percent | 123,200 | 205,300 | 28 to 46 percent | | 25 percent | 308,000 | 513,300 | 69 to 115 percent | | 100 percent | 1,232,000 | 2,053,300 | 275 to 459 percent | The Big Three operate 447,605 sites, roughly 1.34 million sectors, on 279 to 376 megahertz. A 25 percent traffic share requires SpaceX to build between 69 and 115 percent of every cell site the three incumbents have deployed. Note that 166,264 of those 447,605 sites are already small cells, so the comparison is not small cells against macro towers; against the macro estimate alone the requirement runs 109 to 182 percent. The arithmetic driving that is the spectrum shape: 25 megahertz of downlink against 150 to 200 effective for an incumbent is roughly a sevenfold handicap, and the only currency that buys it back is cell count. **Capital.** SpaceX discloses neither Version 3 unit cost nor Starship marginal cost. Assuming 2 million dollars per satellite plus a 60 million dollar flight across 60 satellites gives about 3 million all-in, and taking 50,000 dollars per installed small cell as an industry-typical figure: | Share of US mobile traffic | Space segment | Terrestrial layer | Total build | | -------------------------- | ------------- | ----------------- | ----------------- | | 10 percent | $2.9B | $6.2B to $10.3B | $9.1B to $13.2B | | 25 percent | $7.3B | $15.4B to $25.7B | $22.7B to $33.0B | | 100 percent | $29.2B | $61.6B to $102.7B | $90.8B to $131.9B | The terrestrial range follows the cell range above: the lower figure assumes 5 bits per second per hertz, the upper assumes 3. Those unit costs are assumptions, not disclosed figures, and a factor of two error moves the space-segment column materially without changing the conclusion. Against combined industry capital expenditure that CTIA puts above 30 billion dollars annually, a 23 to 33 billion dollar build for a quarter of national traffic is nine to thirteen months of what the incumbents already spend. The distribution is the finding. Roughly 70 to 80 percent of the capital and effectively all of the execution risk sit in the terrestrial layer, and none of it is a space problem. The hard costs are site acquisition, power, landlord consent, permitting, zoning, and maintaining uptime across hundreds of thousands of locations SpaceX neither owns nor controls, since a customer who cancels their subscription removes a cell from the network. The host base compounds it. Roughly two million United States Starlink dishes could in principle support far more capacity than the requirement, but they sit where terrestrial alternatives are poorest, which is where demand density is lowest. Section 08 ## How many paying customers does 65 megahertz support? **Twenty million subscribers on 37,000 to 62,000 cells. Spectrum is not the binding constraint at any plausible commercial scale.** Share of national traffic is the wrong denominator for a subscriber business, and using it makes the build look harder than it is. A quarter of all United States mobile traffic is not a customer target: at average consumption it is roughly 165 million subscribers, close to 29 percent of every wireless connection in the country. Convert instead to subscribers. The average American smartphone used about 25 gigabytes a month at the end of 2025\. Extrapolating per-subscriber growth to 2028 gives roughly 45 gigabytes, which at a 7 percent busy-hour share is about 233 kilobits per second per subscriber during the busy hour. A 25 megahertz downlink cell delivering 5 bits per second per hertz carries 125 megabits per second, so it supports around 536 subscribers. At a more conservative 3 bits per second per hertz, about 321. | Subscriber base | Busy-hour Tbps | Cells at 125 Mbps | Cells at 75 Mbps | V3 satellites | | --------------- | -------------- | ----------------- | ---------------- | ------------- | | 5 million | 1.2 | 9,300 | 15,600 | 74 | | 10 million | 2.3 | 18,700 | 31,100 | 147 | | 20 million | 4.7 | 37,300 | 62,200 | 294 | | 40 million | 9.3 | 74,700 | 124,400 | 588 | | 60 million | 14.0 | 112,000 | 186,700 | 883 | Twenty million subscribers, roughly twice the current Xfinity Mobile base, requires 37,000 to 62,000 cells and about 290 satellites. At 30 dollars a month that is 7.2 billion dollars a year of revenue against a terrestrial build of 1.9 to 3.1 billion and a space segment near 900 million. Those unit costs carry the same assumptions flagged in the previous section, but the shape of the answer does not depend on them. So the spectrum is sufficient. Sixty-five megahertz does not cap the subscriber base at any plausible commercial scale, because capacity is spectrum multiplied by cells and cells scale with customers rather than with the national total. What the thin holding costs SpaceX is a multiplier rather than a ceiling: roughly seven times more cells than an operator holding 175 megahertz of effective downlink would need for the same customer count. Seven times more of 5,300 cells is 37,000 cells. Expensive, not impossible. For scale, the small cell platform spun out of Crown Castle in May 2026 holds more than 100,000 nodes. On this arithmetic that footprint alone would support somewhere between 32 and 54 million subscribers. The site layer for a 20 million subscriber business already exists, in third-party hands, today. Section 09 ## What kind of carrier does this spectrum make? **A mid-market carrier selling to the median user, not a premium national network.** Figure 5 · Spectrum depth, and which direction it faces T-Mobile376 MHzAT&T314 MHzVerizon279 MHzSpaceX65 MHz Total sub-6 GHz holdings SpaceX downlink SpaceX uplink and unpaired Carrier depth: TD Cowen post-transaction estimates as reported in trade press; other published methodologies place AT&T higher. SpaceX split assumes conventional pairing, which remains open in the transaction docket. SpaceX is acquiring up to 65 megahertz through the EchoStar transactions: 40 megahertz of AWS-4, 10 megahertz of Personal Communications Service (PCS) H-block, and between 5 and 15 megahertz of unpaired AWS-3 uplink depending on the license area. The 65 megahertz figure quoted throughout this report and everywhere else is therefore a ceiling rather than a uniform national holding, and in some markets the position is nearer 55\. The FCC granted the assignment applications on 12 May 2026, with the second step of the transfer expected around late 2027. Set against post-transaction estimates of national spectrum depth compiled by TD Cowen, T-Mobile at roughly 376 megahertz, AT&T at roughly 314 and Verizon at roughly 279, the holding is thin. The less obvious problem is its shape. As those bands are conventionally paired, only about 25 megahertz of the 65 faces downlink, with the balance on the uplink side. That is a sensible configuration for direct-to-device service, where the 200 milliwatt handset transmitter is the binding constraint and uplink bandwidth buys back link margin. It is a poor configuration for a terrestrial network, where downlink carries the overwhelming majority of traffic. The band direction is not settled: filings in the transaction docket reference reversing EchoStar's earlier election to designate 2000 to 2020 megahertz for downlink operations. Note also that the 20 by 20 megahertz channel plan Nicolls described for the next mobile generation matches AWS-4 rather than the 5 megahertz PCS G-block in use today. If that reading is right, the EchoStar purchase is as much about escaping the supplemental coverage power regime as it is about bandwidth. Then there are the handsets, where the picture has moved and most commentary has not kept up. AWS-4 and H-block are not supported in the installed base, a point made independently by TMF Associates and by Aetha Consulting, which put ecosystem support two to three years out before the replacement cycle begins. Musk has acknowledged a similar lag. That estimate is now better read as a statement about fleet penetration than about availability. Qualcomm announced the X105 modem at Mobile World Congress in March 2026, the first platform with native 5G New Radio non-terrestrial network support, and says it began sampling with commercial devices expected in the second half of 2026\. Trade reporting pairs it with the next flagship silicon generation, which puts it in handsets shipping late 2026 and into 2027, and Qualcomm has positioned the timing explicitly against SpaceX's 2027 service target. Chipset support is therefore arriving now rather than in three years. What still takes two to three years is the replacement cycle that turns chipset availability into a base large enough to sell against, and that is the constraint that matters commercially. The 15 megahertz of AWS-3 uplink is the interesting piece, because it pairs with AWS-4 downlink to form 3GPP Band 70, which does have some existing device support. Where the thin holding genuinely binds is not on subscriber count but on subscriber type, and that is the more consequential constraint. A 25 megahertz cell peaks near 125 megabits per second when empty, and under load with a few hundred attached devices individual throughput falls into the low single digits. A 2026 study of United States consumption puts the median user at 14.5 gigabytes a month while the top decile exceeds 52 and accounts for close to 40 percent of all traffic. SpaceX can serve the median user comfortably on this spectrum. It cannot serve that top decile competitively. That forces the business model rather than merely shading it. This is a value and mid-market proposition sold to light and moderate users at low average revenue per user, structurally closer to Xfinity Mobile or Mint than to a Verizon postpaid base. Any model assuming premium pricing on this spectrum position is assuming away the physics. The uplink shape points the same way: 40 megahertz of uplink against 25 of downlink is backwards for a terrestrial network and only makes sense if direct-to-device remains the primary use. Section 10 ## Can a cable partnership solve the site problem? **It solves site access, power and backhaul. It solves none of the spectrum, rural coverage or handset problems.** If the terrestrial site layer is the constraint, the fastest route through it is not to build it. It is to rent somebody else's. That is what makes the reported Charter conversation more significant than a wholesale negotiation, and it is worth being precise about what such a deal could be and what it could not. **What a partnership would actually provide.** A cable operator holds, already permitted and already powered, the exact asset the sizing says is missing. | Requirement from the sizing | What a cable operator already holds | | ------------------------------------------------------ | ----------------------------------------------------------------------------------------------- | | 37,000 to 62,000 cell sites for 20 million subscribers | Aerial strand passing 65 million homes at Comcast alone | | Site acquisition and landlord consent | Municipal franchise and pole attachment rights, decades old | | Power at every site | Cable plant power supplies distributed along the strand | | Backhaul independent of the satellite | Data Over Cable Service Interface Specification (DOCSIS) and fiber at every node | | Downlink depth beyond 25 megahertz | Citizens Broadband Radio Service licenses, roughly $1 billion bought across Comcast and Charter | | Retail, billing and device supply | More than 18 million mobile lines between the two | The strand is the decisive item. Comcast has deployed Samsung strand-mounted small cells specifically because they leverage existing DOCSIS infrastructure without building or acquiring cell sites, and there is precedent at scale in the roughly 19,000 strand mounts Sprint and Altice deployed on Long Island. Compare the two site models directly. A rooftop femtocell requires a willing homeowner, a subscription that keeps paying, a power arrangement, and a location determined by where people happen to want satellite broadband. A strand mount requires a work order. The spectrum fit is better than coincidence should allow. SpaceX's structural weakness is shape rather than depth, with roughly 25 megahertz facing downlink. Cable's Citizens Broadband Radio Service holdings at 3.5 gigahertz are pure capacity spectrum in exactly the dense areas where the reuse arithmetic bites hardest, and Comcast has said its licenses cover about 80 percent of its homes passed and roughly half the United States population. Comcast also sold its 600 megahertz low-band to T-Mobile, with final sale expected around 2028, which removes the coverage layer it once planned and leaves a gap that AWS-4 fits. **What a partnership would not provide.** Four limits deserve to be stated plainly, because the optimistic version of this story blurs them. | Problem | Solved by a cable partnership | | ------------------------------------------ | --------------------------------------------------------- | | Cell site acquisition at national scale | Yes, and this is the whole point | | Power and permitting per site | Yes, through existing franchise rights | | Backhaul that is not the satellite | Yes, but only inside cable footprint | | Downlink capacity depth | Partly, as a hosted tenant rather than a licensee | | Indoor coverage | Partly, through cable WiFi rather than the radio layer | | Rural and dead zone coverage | No, this is what SpaceX contributes, not what it receives | | Access to a national radio access network | No | | Handset band support for AWS-4 and H-block | No | The Verizon route is closed and Verizon has said so on the record. Addressing the Charter speculation on its second quarter 2026 earnings call, chief executive Dan Schulman said there is "no backdoor to our MVNOs in any structure," adding that Verizon does not intend to wholesale to a satellite operator and will approach direct-to-device through a joint venture with T-Mobile and AT&T. The residential agreement with Comcast and Charter is described as perpetual and irrevocable, with pricing amended as recently as January 2026, but perpetual for cable does not mean assignable to a third party. Citizens Broadband Radio Service licenses also stay where they are. Any arrangement is hosting and sharing rather than a transfer, and Priority Access Licenses carry assignment constraints. SpaceX would be a tenant on another company's spectrum inside another company's footprint, which is a materially weaker position than the 65 megahertz it is buying outright. Geography is a real mismatch rather than a detail. Comcast passes roughly 65 million of about 145 million United States homes, in non-contiguous urban and suburban clusters. Cable plant does not reach the rural areas where satellite coverage is differentiating. A cable partnership therefore supplies capacity where SpaceX needs capacity and contributes nothing where SpaceX has an advantage. That is a complement, not a combined national network. **Who is likely, and who is not.** Charter is the reported counterparty and the most plausible cable one: it has been the more aggressive of the two on Citizens Broadband Radio Service deployment, it has been developing a fiber-powered distributed antenna platform with Nokia to create dense supercells on its hybrid fiber coaxial plant, and the Cox combination adds footprint. Comcast is the better asset and the worse prospect. It offloads roughly 90 percent of Xfinity Mobile traffic onto more than 23 million of its own WiFi hotspots, crossed 10 million wireless lines in the second quarter of 2026, and improved its Verizon terms in January. A company that has already solved the problem with assets it owns has limited reason to take on a satellite partner, and considerable reason to be seen talking to one before its next wholesale renegotiation. The third candidate is the neutral host layer. On 1 May 2026, Crown Castle completed the sale of its small cell business to a fund managed by the private equity firm EQT, creating Arium Networks as a standalone company with more than 100,000 on-air or contracted small cells across 43 states, supported by a long-term fiber agreement with Zayo, whose network now spans about 224,000 route miles after absorbing Crown Castle's fiber business the same day. Those small cells already serve all three national carriers, which means the platform is spectrum-agnostic and carrier-neutral by construction. It is the only asset of the required type in the United States that is not owned by a company with a reason to say no. The three national carriers are not going to be the answer. They have declined wholesale, Verizon has ruled it out publicly, and they are building a shared direct-to-device vehicle of their own. An outright acquisition of a cable operator by SpaceX is also implausible on capital, regulatory and strategic grounds. **One channel SpaceX already has.** The EchoStar transaction was more than a spectrum sale. It carried a long-term commercial agreement under which Boost Mobile subscribers reach SpaceX's next-generation direct-to-device service through Boost's own 5G core, and EchoStar continues to offer mobile service through a wholesale arrangement with AT&T. That gives SpaceX a retail brand, a billing relationship and a subscriber base without building any of them, which is the one piece of the puzzle it is not missing. It does not solve site access, and Boost's base is a fraction of the 20 million subscriber target, but any analysis that treats SpaceX as starting from zero on distribution is wrong. **The shape a deal would take.** Not a mobile virtual network operator agreement, which is the frame most coverage has reached for and the one Verizon has already blocked. A neutral host arrangement: the site owner hangs SpaceX radios on its strand or poles, feeds them with existing fiber, and takes a per-site fee or an equity interest. SpaceX contributes the constellation, which the sizing says is oversized and comparatively cheap, and coverage outside the host's footprint. The host contributes the site layer, which the sizing says is roughly 70 to 80 percent of the capital and effectively all of the execution risk. No traffic touches a national carrier's radio access network, which is precisely why the structure survives Verizon's objection. Section 11 ## What would have to be true for this to work? **Site access at scale, on somebody else's poles, sold to a customer the incumbents price above.** The version of this business that closes is not the one in the headlines. It is a nationwide coverage and resilience layer, sold on the elimination of dead zones and on staying up when terrestrial networks fail, with terrestrial capacity added selectively where the unit economics are favorable and where backhaul already exists. Against roughly 600 billion dollars of annual national carrier revenue, taking a low single-digit share on that proposition is a substantial business. But the sizing points somewhere more specific than a coverage layer. Reuse physics says the satellites cannot substitute for a ground network, so a ground network has to exist. Sizing that network says the constellation is cheap and the site layer is everything. Converting it to customers says 65 megahertz supports a business of 20 to 40 million mid-market subscribers, needing something on the order of 37,000 to 125,000 cells. That is not a fourth national carrier and it is not a coverage novelty. It is a real company, built on somebody else's poles, selling to the median user rather than the heavy one. Whoever solves site access at that scale captures it, and that is a real estate and permitting problem rather than an aerospace one. The tell is where SpaceX is shopping. Bloomberg reported the company approached Charter Communications about routing a portion of its mobile traffic over Charter's terrestrial network, while all three national carriers have publicly declined a wholesale arrangement. If satellites and rooftop radios closed the gap on their own, there would be no reason to be negotiating for somebody else's ground network. Section 12 ## What this means depending on where you sit **The same finding points in different directions for different readers, and the divergence is the useful part.** **If you invest.** The asset that gets repriced by this analysis is not SpaceX. It is whoever owns pre-permitted, powered, fiber-fed vertical real estate in dense markets. A build of this size cannot be assembled through acquisition at any speed that matters, which makes site access the scarce input and its owners the beneficiaries whether or not SpaceX is the buyer. The cable operators, the tower companies and the independent small-cell platform spun out of Crown Castle all sit on that asset. On the incumbent side, the threat to carrier revenue is real but slower and smaller than the headlines imply: a mid-market network taking a low single-digit share is a margin story, not an existential one, and it arrives after handset support rather than before. The number to watch is not satellites launched. It is sites contracted. **If you buy connectivity for an enterprise.** Nothing here changes a procurement decision inside three years. Handset band support is two to three years out before the replacement cycle begins, and the terrestrial layer does not exist. What does change is the medium-term planning assumption. A credible fourth network, priced for the median user and weakest exactly where enterprise traffic concentrates, is a lever in a carrier negotiation before it is a supplier. Treat it as competitive pressure on incumbent pricing, not as a candidate on a shortlist, and revisit when a device maker announces the bands. **If you sell into the ground segment.** A terrestrial build in the tens of thousands of sites is a radio, backhaul and site-services procurement, not a space program. The gateway side is the nearer-term opportunity: landing tens of terabits per second means roughly nine terabits of fiber into a single gateway location, at twenty to forty locations, which is an ordinary transport buy at an unusual density. The small-cell radio order that follows is large enough to move a vendor's year, and it will go to whoever can supply a 2 gigahertz outdoor unit at strand-mount cost rather than macro-site cost. Both are commercial conversations available now, well before the network exists. **If you work in policy.** The interesting question is not whether SpaceX gets more spectrum. It is whether the site layer becomes a competitive bottleneck the way spectrum once was. If the binding constraint on a new entrant is access to poles, strand and rooftops rather than megahertz, then pole attachment terms, franchise conditions and neutral-host arrangements matter more to market structure than the next auction. The 30 November 2029 performance milestone is also worth reading closely: it is specified against outdoor users, which tells you what the regulator expects this network to be. **If you do business development.** The opening is not SpaceX. It is everyone who now has to answer a question they were not asked six months ago. Cable operators are being valued on an asset they built for video. Tower and neutral-host companies have a new class of tenant to price. Vendors need a position on whether a 2 gigahertz strand-mount product line is worth building. Those conversations are live now and they do not depend on whether SpaceX executes. Section 13 ## What to watch Whether SpaceX secures site access and fiber transport at scale, and in what legal form. Hosting or neutral host language rather than mobile virtual network operator language would signal a genuine infrastructure deal, because the two imply completely different economics. SpaceX radios appearing in Spectrum Access System registrations tied to cable or neutral host sites would be visible in public data before any announcement. If the conversation with Charter evaporates once cable signs improved wholesale terms, it was leverage rather than strategy. Whether the X105 modem actually ships in volume devices in the second half of 2026 as Qualcomm says, and whether any handset maker markets the bands rather than merely supporting them. Chipset availability is no longer the open question; installed-base penetration is. The payload count on the first orbital Starship deployment, which converts the 60-satellite manifest from designed to demonstrated or reveals the gap. Whether the company bids aggressively in the upper C-band auction the FCC intends to complete by July 2027, which is where real downlink depth would come from. The first independently measured throughput on second-generation mobile hardware, against the 150 megabit per second peak SpaceX announced in February 2026\. And the 30 November 2029 performance milestone, the first date on which a regulator decides whether this worked. Section 14 ## What would change this view Research that cannot be wrong is not research. Five things would move the conclusion materially, and three of them are observable within twelve months. **A Starship manifest well below 60.** The space-segment column assumes 60 satellites per flight, which is designed. Twenty is demonstrated, and suborbital. A sustained manifest near 20 triples flight counts and pushes the deployment timeline from roughly nineteen months to nearly five years at plausible cadence, which changes how much runway the incumbents have but not the destination. **Rooftop small cells failing to reach 5 bits per second per hertz.** The subscriber arithmetic assumes a spectral efficiency that residential rooftop installations at 2 gigahertz may not deliver in practice. At 3 bits per second per hertz the cell requirement rises by two thirds. At 2, this stops being a mid-market carrier and becomes a rural overlay. **The band pairing resolving toward downlink.** Filings in the transaction docket reference reversing EchoStar's earlier election on the 2000 to 2020 megahertz block. If more of the 65 megahertz ends up facing downlink than the 25 assumed here, the cell requirement falls proportionally and the heavy-user constraint softens. **Site access proving unavailable at any price.** The entire thesis rests on renting a site layer rather than building one. If cable, the neutral host platforms and the tower companies all decline, the build reverts to the rooftop plan and the cost and timeline estimates here are far too optimistic. **Handset penetration arriving faster than expected.** Chipset support is already closer than the two-to-three-year consensus implies, with the Qualcomm X105 expected in devices from the second half of 2026\. What remains is the replacement cycle. A carrier-driven acceleration, a device maker marketing the bands as a differentiator, or an aggressive handset subsidy would pull the whole timeline forward and with it the date at which the site build has to be finished. Back matter ## Methodology and sources Every figure was pulled from a live search at time of writing, and the full set was re-verified against live sources on 17 August 2026 before publication. Traffic, connection and infrastructure counts come from the CTIA annual wireless industry survey published September 2025, covering calendar 2024\. Satellite hardware specifications are SpaceX published figures and statements at Mobile World Congress 2026, and are designed specifications throughout, labeled as such. Constellation counts are as of 17 August 2026\. Spectrum depth estimates are TD Cowen figures as reported in trade press; methodologies vary and other published estimates for AT&T run higher. Regulatory items draw on FCC orders and public notices in the SpaceX supplemental coverage and EchoStar assignment proceedings. The sizing sections rest on the author's calculations. Capacity apportionment to the contiguous United States uses pure surface area share, 1.6 percent, consistently in every section. Inclined orbits dwell longer at mid latitudes, which would move the figures 1.3 to 1.5 times in SpaceX's favor; that adjustment is stated where relevant but is deliberately not applied, so every fleet figure here is the conservative one. Spectral efficiency assumptions for rooftop small cells, satellite and small-cell unit costs, the 7 percent busy-hour share, the per-subscriber consumption extrapolation and the 2028 demand extrapolation are stated assumptions rather than sourced figures, and are identified in the text. Claims resting on a single source are flagged where they appear. Primary sources: Qualcomm product announcements and trade coverage of the X105 modem; FCC International Communications Filing System, including DA-25-197, DA-26-36 and the direct-to-cell constellation application SAT-LOA-20250916-00282; Securities and Exchange Commission filings by EchoStar Corporation and Space Exploration Technologies Corp.; CTIA Annual Wireless Industry Survey 2025; published United States smartphone consumption studies from 2026; Starlink published Version 3 specifications; SpaceX statements reported by Fierce Network, SDxCentral, Light Reading and SpaceNews; Comcast, Charter, Crown Castle and Zayo corporate communications and earnings coverage; analysis from TMF Associates and Aetha Consulting. Back matter ## Disclosure Satellite Insights LLC has no commercial relationship with SpaceX, EchoStar, Comcast, Charter Communications, Cox, Crown Castle, Zayo, Arium Networks or any of the three national wireless carriers named in this piece. No position is held in any security discussed. This analysis rests entirely on publicly available information. ## This is what your membership pays for Independent analysis of the satellite communications sector from a forty-year enterprise ground-segment lens: low Earth orbit broadband, geostationary operator economics, ground segment, direct-to-device and sovereign connectivity. No hype. A point of view. If you know someone who would argue with these numbers, forward this to them. That is the most useful thing you can do with it. [Share the report link](https://www.satelliteinsights.com/spacexs-site-problem/) **Glenn Canales** is Principal of Satellite Insights LLC. Forty years in commercial and enterprise satellite communications, most recently as Senior Director of Enterprise Broadband Services at Viasat from 2016 to 2024, and previously at iDirect Technologies, PanAmSat and Intelsat, Sagenet, and as a communications supervisor in the United States Air Force. Satellite Insights LLC also advises operators, vendors and investors on enterprise ground-segment economics, multi-orbit service strategy and go-to-market questions, through retained and project engagements and through the expert networks. Reply to this post by email, or reach Glenn on LinkedIn. **Disclosure.** Satellite Insights LLC has no commercial relationship with SpaceX, EchoStar, Comcast, Charter Communications, Cox, Crown Castle, Zayo, Arium Networks or any of the three national wireless carriers named in this report. No position is held in any security discussed. Independent weekly intelligence on the satellite communications sector. No hype. A point of view.