Starlink Mobile Is Building the Cellular Network Backwards

SpaceX's latest spectrum deal has unnerved America's mobile operators. The threat is not that satellites will make cell towers obsolete, but that a company starting with the satellite layer might need a different mix of towers, spectrum and capital to compete.

On 9 October 2026, America's biggest mobile operators suffered a remarkable sell-off. T-Mobile fell by roughly 13%, AT&T by nearly 10%, and Verizon by almost 9%. Across the same trading session, shares in the companies that own mobile towers rose sharply. Crown Castle gained about 16%, American Tower more than 9%, and SBA Communications more than 7%. The contrasting reactions, documented in Reuters' account of the sell-off, were responses to the same development.

The previous day, SpaceX had announced an agreement to buy Grain Management's nationwide portfolio of 800 MHz mobile spectrum, a deal reportedly worth around $8 billion. Grain's announcement confirmed that the sale remained subject to approval by the Federal Communications Commission (FCC). SpaceX intends to use the frequencies as part of Starlink Mobile, a network combining satellite connectivity with terrestrial radio infrastructure.

The market appeared to be pricing in two possibilities at once: a new competitor for AT&T, Verizon and T-Mobile, and a new customer for the towers those competitors use. That apparent contradiction is a useful starting point. SpaceX is not trying to abolish the ground network; it is beginning with a large network in orbit and asking how much ground infrastructure must be added to turn satellite coverage into a competitive mobile service.

Satellite and terrestrial infrastructure may play different roles in rural, suburban and urban areas. Conceptual illustration by the author.

Coverage Is Not Capacity

There is an important difference between reaching a location and carrying the traffic generated there. A satellite hundreds of kilometres above Earth can communicate across a broad geographical footprint, making it particularly attractive for ships, remote settlements and roads where a conventional tower might serve very few paying customers. A mobile network in central Manhattan faces almost the opposite problem: enormous volumes of traffic concentrated within a few city blocks.

Terrestrial operators handle dense demand by dividing cities into many smaller cells, each using radio frequencies that can be reused elsewhere in the network. Satellites also use multiple beams and frequency reuse, but they cannot ordinarily match the density and short radio paths of thousands of nearby terrestrial cells. A large footprint that is excellent for reaching distant users becomes a constraint when too many of them want capacity at once.

The handset introduces a further complication. Ordinary phones have small antennas and limited transmission power. Sending a useful signal to a moving satellite is difficult enough outdoors; walls, roofs and other obstacles make reliable indoor service harder still. Starlink's existing satellite broadband coverage should therefore not be confused with universal, high-capacity direct-to-phone coverage. The two services use different equipment and place different demands on the network.

SpaceX's 8 October announcement describes the new purchase as up to 14 MHz of paired spectrum in the 800 MHz band. Low frequencies generally travel farther and penetrate buildings better than higher ones, although they offer less capacity for a given amount of spectrum. SpaceX proposes to use this band as a terrestrial coverage layer alongside its satellite-to-mobile constellation and other frequencies. It is an acknowledgement that reliable mobile service cannot be designed around satellite coverage alone.

Earlier spectrum agreements with EchoStar are another part of the strategy, but the assets should not all be treated as if they had already been transferred without conditions. EchoStar's June 2026 SEC filing describes licenses placed in a trust, with the subsequent acquisition by SpaceX then targeted for November 2027. The distinction matters because a portfolio of announced and contracted frequencies is not yet the same thing as an operational nationwide network.

Building the Mobile Network Backwards

Traditional mobile networks were built from the ground upwards. Operators acquired spectrum, installed radios on towers and rooftops, arranged backhaul, and expanded service outward from the areas with enough customers to support the investment. Coverage of rural territory was progressively added, sometimes for commercial reasons and sometimes because licenses or public policy required it. Incumbents already use different frequency bands, infrastructure sharing and satellite partnerships to manage those costs; their networks are hardly uniform carpets of towers.

SpaceX nevertheless starts from an unusual position. It already operates a large satellite constellation, manufactures its spacecraft, and controls the rockets used to launch and replenish them. Starlink supplies an internal customer for launch capacity, while repeated launches help improve the satellite business. That reinforcing relationship is explored in the Journal's review of Eric Berger's Reentry, which examines how reusable rockets became part of a larger industrial system.

With that satellite layer in place, SpaceX can approach terrestrial construction as an exercise in filling particular gaps. A remote area may need relatively little additional infrastructure for basic connectivity. A suburb may benefit from a modest number of terrestrial cells to improve indoor reception and absorb local traffic. A dense city will still require extensive infrastructure, because no amount of satellite coverage changes how many people are competing for capacity at street level.

This is the meaning of building a mobile network backwards. SpaceX begins with broad, though imperfect, coverage and attempts to add capacity and reliability where customers need them. Conventional operators began with terrestrial capacity in populated places and worked to extend coverage beyond them. Both approaches can lead to a hybrid satellite-terrestrial network; the question is whether their different starting points produce different costs when they arrive there.

Vertical integration gives SpaceX more freedom to coordinate satellite design, launch cadence, network operations and terrestrial deployment. It does not make those components free. Satellite capacity used by mobile phones cannot also be allocated without limit to broadband customers, and satellites must be replaced. Owning the rocket does not abolish the cost of manufacturing it or the opportunity cost of using a launch for one purpose rather than another. Any structural advantage must emerge from the economics of the complete network, not merely from the impressive list of assets SpaceX owns.

How Many Towers Can SpaceX Actually Avoid?

The tower companies' share-price gains are telling, but they should not be mistaken for evidence that SpaceX has already signed large numbers of tower leases. They express an expectation: if SpaceX wants to operate a competitive terrestrial network, existing sites may offer the fastest way to install its radios. In cities, towers, rooftops, indoor systems and small cells would all be candidates, depending on where coverage and traffic demand justify them.

The scale could be considerable. In an August 2026 assessment reported by Yahoo Finance, Bernstein analysts estimated that a standalone SpaceX mobile network could cost roughly $50 billion to $130 billion, including spectrum, across different build-out scenarios. The estimate predates the October Grain agreement and is not a SpaceX investment plan. It does, however, show how far the costs of a national network may extend beyond satellite launches and spectrum purchases.

The most revealing test may be outside the extremes of wilderness and Manhattan. In a remote area, the benefits of satellite coverage are comparatively obvious. In a crowded city, the necessity of ground infrastructure is equally obvious. Suburbs and smaller cities occupy the disputed middle: enough people to generate meaningful traffic, but not always enough to justify the same cell density as an urban centre. If SpaceX can provide acceptable indoor coverage and data capacity with fewer terrestrial sites there, the architectural difference begins to acquire economic meaning.

Even then, site counts alone would be misleading. A network with fewer towers but inadequate speeds, poor indoor reception or expensive satellite capacity is not more efficient in any commercially useful sense. The relevant comparison is the total cost of serving customers at a given level of reliability and performance. It must include spectrum, equipment, leases, backhaul, satellite replacement and the cost of carrying growing amounts of data.

There is a real possibility that SpaceX will discover the same constraints that shaped its competitors' networks. It could spend heavily on terrestrial infrastructure and still find itself facing the usual problems of expensive spectrum, limited urban sites and customers who expect their phones to work in lifts, shopping centres and railway stations. A hybrid architecture might be innovative without being cheaper. Conversely, savings in sparsely populated areas, combined with shared satellite capacity across broadband, business and mobile customers, could be valuable even if city networks end up looking conventional.

The Threat to Telecom Profits May Arrive Before the Subscribers

SpaceX does not need to replace Verizon or AT&T to hurt their economics. Mobile operators bear substantial fixed costs, and once capacity is installed the incremental cost of serving another subscriber is often modest until congestion demands another round of investment. An entrant that attracts a relatively small share of customers can still force an incumbent to offer lower prices, better bundles or more generous service terms across a much larger subscriber base.

Starlink Mobile's most plausible early selling point may be reach rather than peak speed. A rural household already paying for Starlink broadband could find a combined subscription attractive. A contractor moving between towns and remote sites might value fewer dead zones. Commercial fleets and people who travel frequently through poorly served territory could care more about continuity of service than about winning another urban speed test. These are possibilities, not yet proof of demand at prices that make a new network profitable.

Incumbents also have options. T-Mobile already offers T-Satellite with Starlink, which supports messaging and selected satellite-optimised applications on compatible phones where terrestrial coverage is unavailable. Its current limitations help illustrate the technological gap between emergency or occasional off-grid connectivity and a complete replacement for terrestrial mobile service. They also demonstrate that traditional operators can buy the satellite layer rather than build it themselves.

That makes the relationship between SpaceX and established carriers unusually ambiguous. SpaceX might operate its own consumer mobile business, sell satellite connectivity to other networks, share infrastructure or combine all three. A wholesale model could strengthen incumbents by improving their coverage. A retail model could undermine their pricing power. Neither requires SpaceX to replicate every part of a traditional operator, and a partnership may prove more profitable than a national build-out.

The market's reaction was therefore a repricing of future competitive possibilities, not a measurement of customers already lost. The tower companies might ultimately benefit from SpaceX's expansion, while mobile operators might find that incorporating satellite service is cheaper than competing against it. The outcome depends less on whether satellites work than on who owns the customer relationship and how the costs of the combined network are shared.

Global Satellites, National Rules

A satellite network can cross borders far more easily than a telecommunications license. The frequencies SpaceX has agreed to acquire in the United States do not give it permission to establish terrestrial mobile operations in France, India or Brazil. Spectrum allocations, network approvals and the treatment of direct-to-device services vary between countries. In many markets, partnerships with national operators may be the only commercially sensible route.

There is also a strategic dimension. Mobile and satellite systems are increasingly treated as infrastructure on which governments, emergency services and defence organisations may depend. A foreign-controlled service with powerful advantages in launch, satellites and connectivity can become both commercially attractive and politically sensitive. The concern resembles the risk examined in the Journal's essay on Europe's dependence on orbital infrastructure: access to an excellent service is not quite the same as retaining control over a capability when relations deteriorate.

Starlink Mobile may therefore develop into several different businesses under one technological roof: a direct mobile competitor in one country, a satellite wholesale partner in another, and a service constrained by national rules elsewhere. An integrated satellite system does not imply a single global commercial model.

What Would Prove the Strategy Works?

The test over the coming years will be less dramatic than the October share-price reaction. It will involve where SpaceX installs terrestrial equipment, how well ordinary phones work indoors, what happens when thousands of customers demand capacity in the same place, and how much the company has to spend to provide a credible alternative to established operators.

Three outcomes are possible. SpaceX might discover a meaningfully cheaper network architecture, using satellites to avoid substantial terrestrial investment without sacrificing service quality. It might become a conventional-looking mobile operator with unusual advantages in remote coverage and vertical integration. Or it might find that supplying a satellite layer to existing carriers offers better returns than competing with them for every customer.

The first outcome would be a genuine change in the economics of telecommunications. The second and third could still create substantial businesses and alter incumbent pricing, but they would not establish that building from space down is inherently superior. The difference will only become visible when network performance, infrastructure requirements and financial results can be compared rather than inferred from an announcement.

SpaceX has changed the starting point. Whether it has changed the cost of reaching the destination remains to be demonstrated.