Amazon Leo said in late July that a 5,105-sat direct-to-device system is too big to do entirely on its own, after the Globalstar agreement. The same month SpaceX, on its first earnings call, said it would build terrestrial small cells and compete with US mobile operators on EchoStar spectrum and a V2 mobile satellite. That is no longer launch plus a dish. That is a carrier.
Dish, D2C, compute file, small-cell carrier. Light Reading, AST 10 Aug call: Avellan partners with MNOs; SpaceX is becoming one.
X will keep posting launch videos. Post the meeting.
Who still has a meeting
- Order turbine
- Queue date
- Transformer
- First MW
The slip lives between queue date and first megawatt.
Disclosed / tracked · as of 22 Aug 2026
The carrier already has the thin product
Four products that used to be four companies
Starlink consumer broadband: about 10,000 satellites, about 10 million subscribers, the dish you already know. Starlink Direct to Cell: 650-plus dedicated birds, text and a short app list, T-Mobile as the US face, late-2027 language still on voice and full data. Starmind: an FCC number up to one million, a Vera Rubin NVL72 sentence on the August call, Starship as the assumed truck. Add EchoStar spectrum and terrestrial small cells and you have a fourth: a US network. No listed name is allowed to be all four. The closest attempts are Amazon (Leo plus AWS plus a Globalstar MSS bid plus a 2028 D2D file) and Rocket Lab (Electron plus a Neutron that has not flown plus an $8 billion Iridium deal that has not closed).
The competitive tell is who still has a room. AT&T and Verizon stay with AST because Avellan will not build a competing tower company. T-Mobile already sold the Starlink text feature and is in a future US MNO joint venture that AST says could add a third or fourth US home network. If the small-cell plan is real, that joint venture is also a defensive object: carriers pooling a satellite partner so they are not renting their hardest-to-serve square miles from a company that also wants their urban ones.
The fairing is the scarce input
SpaceX's own feed in mid-August was on the 100th Falcon of the year, plus Starship Flight 14 prep. Starcloud cannot book rideshares past 2028 and 2029. AST moved from New Glenn to three-at-a-time Falcon after a pad explosion. Amazon is on Vulcan, Ariane, a few Falcons, and a New Glenn book that has to become flights. Until Neutron flies, it is.
A company that sells the fairing, the dish, the text SIM, the future GPU hall, and the small cell needs counterparties who still have a product it does not want. Right now that is AST's unmodified-phone broadband arrays, Amazon's AWS-shaped enterprise terminal, Iridium's L-band voice that Rocket Lab just bid $8 billion for, and Overview's beam into a farm. Those are niches. They are also the only listed ways to express not-SpaceX without pretending a 391-sat constellation is a peer.
A dish is not a tower
There is no listed SpaceX share on this scan. It is a map of who still has a meeting after SpaceX decided to be a carrier and a computer. AST's 45-sat plane versus Starlink V2 mobile broadband in 2027. Amazon's consumer Leo in a named band this year. Neutron's first flight. The Iridium close in mid-2027. A Meta-metered Overview beam in 2028. If three of those five print on time, the industry is still a set of companies. If they slip while Falcon does 130 a year and Direct to Cell adds countries, the industry is a supplier list.
Cell tower on a green hill.
