SpaceX Joins AT&T in EchoStar’s Spectrum Sale

In May 2026, the FCC approved SpaceX’s purchase of roughly 65 MHz of mobile spectrum from EchoStar. AT&T won approval to buy 50 MHz from the same seller. Eight years earlier, SpaceX had sought permission to put broadband satellites overhead.

Mobile competition is reorganizing around control of usable spectrum, compatibility with ordinary phones and access to subscribers—not simply ownership of towers. SpaceX can buy a path toward mobile service without inheriting a carrier’s retail network; carriers can add satellite reach without owning a constellation. Neither arrangement gives one party the whole customer relationship by default.

Starlink’s first license answered a different question

In 2018, the FCC approved SpaceX’s plan for a 4,425-satellite Starlink broadband network and required it to launch and operate half the planned spacecraft by March 2024. The regulatory question was whether SpaceX could deploy a constellation to provide internet access. It was not whether Starlink could compete for the spectrum on which mobile carriers build handset service.

That design made orbital capacity the conspicuous asset. SpaceX had to put spacecraft in position, operate a network and reach users from above. Terrestrial carriers appeared, at most, as possible partners or rivals at the edge of that project. The phone in a customer’s pocket did not yet define the service.

SpaceX changed its position in the mobile stack when it agreed in 2025 to acquire EchoStar’s AWS-4 and H-block licenses for up to $8.5 billion in cash and up to $8.5 billion in stock. On May 12, 2026, the FCC approved EchoStar’s sale of approximately 65 MHz to SpaceX and 50 MHz to AT&T. SpaceX won approval for a 4,425-satellite broadband network in 2018 and for a mobile-spectrum purchase alongside an incumbent carrier in 2026.

SpaceX later agreed to acquire a low-band spectrum portfolio, describing a goal of making Starlink a major US mobile carrier. That is an ambition, not a description of service already delivered. But a satellite operator buying mobile licenses makes the old boundary—carriers buy spectrum, satellite companies supply coverage—less useful for describing the buyers.

The ordinary handset has become a gate to the network

A constellation can cover a place without serving the phone someone already owns. The handset has to support the relevant radio connection, and the operator needs spectrum rights under which that connection may run. Samsung’s 2023 announcement of a standardized 5G non-terrestrial-network modem addressed that device problem: it was intended to enable two-way communication between smartphones and satellites. Samsung’s announcement did not put the capability into every phone; speculation about its appearance in the Galaxy S23 did not materialize.

Vodafone supplied a more concrete demonstration in 2025 when it made a satellite video call on a standard smartphone through AST SpaceMobile satellites. A demonstration establishes that a particular phone and network can complete a call. It does not establish that the operator can offer the same performance, at commercial scale, wherever its satellites pass.

T-Mobile’s Starlink-powered T-Satellite subscription made another distinction visible. T-Mobile opened the service to US users who were not already its mobile customers, giving satellite coverage a retail route beyond one carrier’s subscriber base. At general availability, however, T-Satellite supported texts and location sharing over satellite, not the full set of services people expect from a terrestrial mobile network. A phone can be compatible with a useful satellite service without that service replacing its tower connection.

Device support therefore changes what satellite operators must negotiate. A network designed around a dedicated terminal asks customers to adopt another piece of equipment. A network seeking the ordinary handset must meet the phone where it is: in a standards ecosystem, on licensed frequencies and inside a service plan someone can buy.

Carriers still own the difficult part of the handoff

SpaceX and T-Mobile obtained FCC approval in 2024 to use Starlink for supplemental telecom coverage in remote areas. That arrangement attached orbital reach to an existing mobile operator rather than dispensing with one. T-Mobile supplied a route to users and a terrestrial network to serve them when a satellite link was unnecessary or unavailable.

Other carriers made different choices about the orbital supplier. Verizon partnered with AST SpaceMobile for broadband direct to phones, committing $65 million in prepayments and $35 million in debt. AT&T signed an AST SpaceMobile agreement running through 2030. Vodafone and AST planned a Europe-led satellite-to-smartphone constellation operated from Germany. Those commitments put carriers in the role of customer, distributor or operating partner, rather than treating every satellite signal as an attempt to bypass them.

The partnerships do not make the carrier’s position automatic. T-Mobile’s standalone satellite offer shows that a satellite-backed product can reach beyond the carrier’s existing subscribers. SpaceX President Gwynne Shotwell, meanwhile, said the company aims to add ground-based infrastructure to its satellite network for a “true mobile service.” SpaceX’s own stated plan places ground facilities inside the job it wants to perform. A satellite cannot make those facilities, a retail channel or an operating agreement disappear.

AST faces the complementary problem from the other side. Carrier agreements give it prospective customers, but AST must still deploy enough working spacecraft to serve them. Its constellation execution became a commercial constraint when it delayed satellite-to-phone service to 2027 after a Blue Origin launch mishap. The partner can sell the promise of coverage before the satellite operator can consistently supply it.

EchoStar’s sales exposed more than one way to acquire access

AT&T agreed to pay about $23 billion in cash for roughly 50 MHz of EchoStar licenses. SpaceX’s separate EchoStar agreement carried up to $8.5 billion in cash and up to $8.5 billion in stock for AWS-4 and H-block licenses. Those prices are not interchangeable per-megahertz comparisons: the transactions involve different license portfolios and deal terms. They do show an incumbent carrier and a satellite operator allocating substantial capital to the same class of strategic asset.

Other companies are paying to change what spectrum can be used for, rather than buying the same kind of mobile portfolio. Eutelsat and SES were expected to receive roughly $6.1 billion combined for helping clear 160 MHz of satellite spectrum for US wireless services. The operators’ part in that bargain is to vacate or reorganize use so someone else can put the frequencies to work. The payment is for making spectrum usable under a new arrangement, not for launching a better spacecraft. The mechanics of that spectrum reallocation matter because a frequency cannot serve two incompatible plans merely because both buyers can finance one.

Amazon chose an acquisition route, agreeing to pay $11.57 billion for satellite operator Globalstar, subject to replacement-satellite milestones. That transaction extends a platform company’s position in low-Earth-orbit connectivity; it does not, by itself, demonstrate an Amazon mobile service. Across the deals, companies are assembling different parts of a service—licenses, satellite capacity, device access and distribution—through purchases and contracts rather than through one established carrier template.

The regulator and the launch schedule can still stop the call

An FCC approval to transfer spectrum answers who may hold the licenses. It does not resolve every question about how a handset service will operate beside existing networks. AT&T and Verizon asked the FCC to deny a SpaceX waiver request, arguing that Starlink’s proposed service with T-Mobile could interfere with their networks. Their objection may protect incumbent interests, but interference is also a technical and regulatory constraint that a new service must answer. Permission to own frequencies is not blanket permission to use them in any configuration.

National access is separate again. SpaceX signed memoranda of understanding with Jio and Airtel in 2025, yet its India launch was still reported as stalled in October 2026. Agreements with local distributors did not settle the question of when Starlink could operate there. AST’s launch-related delay shows a different failure point: even where partners want service, the spacecraft have to reach orbit and work.

The tower still stands. For a satellite to serve the phone beside it, a regulator must permit the frequency, a modem must understand the signal, and someone must put the service on the customer’s bill.