Amazon has asked the Federal Communications Commission to authorize up to 5,105 satellites for a network that would talk directly to ordinary smartphones — voice, messaging, data and emergency services, no cell tower involved. Deployment would begin in 2028.
The satellite count is what travelled. It is the least informative number in the announcement.
The number that explains it
In April, Amazon agreed to acquire Globalstar in a cash-and-stock deal worth roughly $11 billion.
Globalstar operates a couple of dozen aging satellites. Nobody paid eleven billion dollars for those. What Amazon bought was mobile satellite service spectrum — L-band and S-band — licensed with global authorizations, plus the customer relationship built on top of it.
That is the whole story. Direct-to-device is not primarily a hard engineering problem anymore; several companies have demonstrated it. It is a spectrum problem. Talking to an unmodified phone from orbit means transmitting in bands phones already listen to, and those bands are spoken for everywhere on Earth. You either license them, partner with a carrier who holds them, or you do not have a business.
Monday's filing is what that purchase was for. Amazon is now describing the constellation the spectrum can carry.
How it differs from Starlink
This is the part worth understanding, because "Amazon takes on Starlink" flattens a real distinction.
SpaceX's US direct-to-cell service runs over T-Mobile's terrestrial mobile spectrum, under supplemental-coverage-from-space rules — a satellite acting as a cell tower on a carrier's own licensed airwaves. Expanding country by country means striking a deal with a carrier in each one.
Amazon's approach uses dedicated satellite spectrum it will own, authorized globally. Fewer negotiations, more control, and a service that is not contingent on a single carrier partnership per market.
Neither model is obviously superior. The carrier route reaches phones already in pockets today; the MSS route needs handsets with compatible chipsets, which is why Amazon's own materials emphasize "the growing number of smartphones from major manufacturers that have satellite-capable chipsets built in." But the two companies are making meaningfully different bets, not running the same play.
The Apple piece
Globalstar is the network behind Apple's Emergency SOS via satellite.
Under agreements announced with the April deal, Amazon Leo will power satellite features on future iPhone and Apple Watch models — Emergency SOS, Messages, Find My and Roadside Assistance — and Amazon will continue supporting Apple devices already running on Globalstar's existing network.
Apple was Globalstar's largest customer by a distance, having committed roughly $1.5 billion in 2024 across service prepayments and an equity stake in a Globalstar subsidiary. Acquiring the operator meant acquiring the arrangement. One of the more consequential outcomes here is that Apple's satellite features become, in practice, an Amazon service.
What Amazon says it is building
Five orbital shells, all in low Earth orbit:
| Coverage | Altitude | Inclination |
|---|---|---|
| Mid-latitude | 580 km | 55.7° |
| Mid-latitude | 560 km | 58.2° |
| Mid-latitude | 510 km | 56.3° |
| High-latitude | 570 km | 73° |
| Near-polar | 540 km | 84° |
The three mid-latitude shells concentrate capacity over populated regions; the steeper two extend service toward the poles.
Spectrum: L-band and S-band between satellites and phones. Ka-band and V-band for feeder links to Amazon's ground stations — effectively the backhaul.
Not bent pipe. Rather than simply relaying a signal to the ground, the satellites process it in orbit and carry traffic across the constellation over optical inter-satellite links. Amazon also cites digital beamforming and beam-hopping, dual-polarization reception, and adaptive modulation and coding. Translated: it is trying to squeeze more usable capacity out of a narrow, valuable slice of spectrum, and to avoid interfering with everyone else operating in it.
The context Amazon is quieter about
Leo is the network formerly called Project Kuiper. It currently has more than 390 satellites in orbit — enough, Amazon says, to begin rolling out fixed broadband across initial latitude bands later this year.
That is the honest scale check. The company is still standing up its first-generation broadband system, years behind its original schedule, and is now proposing a second constellation more than ten times the size of what it has managed to launch so far.
It also has partners lined up who are waiting: Vodafone, DirecTV, South Africa's Herotel, Australia's NBN.
What has actually happened
An application has been filed. The FCC has not granted anything, and applications of this size attract objections from every operator with a claim on adjacent spectrum.
Deployment starts in 2028 by Amazon's own timeline, and 5,105 satellites take years to launch. Anyone reading this as a service arriving soon is reading it wrong.
What is real today is narrower and more interesting: Amazon spent about $11 billion to own the spectrum that makes phone-to-satellite service possible without asking a carrier's permission, and it now has a public plan for what to do with it.
