10,800 Starlink satellites are in orbit. Amazon Leo has launched 314. That gap, roughly 10,400 satellites, is the short version of the LEO broadband race in 2026.
Amazon Leo holds the FCC license for the second-scale constellation, but had to ask for its 50% deployment deadline to be waived.
OneWeb is the quiet third player, 648 satellites, selling enterprise and government capacity rather than chasing consumers.
The constellation gap
In June 2026 the FCC waived the July 30 milestone that required Amazon Leo to have 1,618 of its 3,236 authorized satellites in orbit. As we wrote in July about AI-native constellation operations, satellite software autonomy has outpaced the physical deployment that feeds it.
Amazon had roughly 314 satellites in orbit when the order landed. It asked for a two-year extension. The FCC said no. Instead it granted a conditional waiver that removes the time limit on the 50% milestone but temporarily demotes the spectral priority of satellites launched after the deadline, until the company accelerates build-out. The full 3,236-satellite constellation still must deploy by 2029.
The arithmetic explains the ask. To reach 1,618 by mid-2026, Amazon would have needed to launch more than 1,300 satellites in under a year across a manifest built on Atlas V, Falcon 9, Ariane 6 and a delayed New Glenn. It did not. The production line works; the cadence does not yet bridge the gap.
| Starlink | Amazon Leo | OneWeb | |
|---|---|---|---|
| Operator | SpaceX | Amazon | Eutelsat Group |
| Satellites in orbit | ~10,800 | ~314 | 648 |
| Authorized | up to 42,000 | 3,236 (+4,500 Gen2) | 648 |
| Subscribers | 10M+ | 0 (pre-launch) | enterprise/wholesale |
| Built for | consumer scale | AWS-edge extension | B2B capacity |
The table masks a real structural difference. Starlink was designed first as a consumer ISP. Amazon Leo, as the KeepTrack deep-dive noted, is designed first as an extension of AWS, a satellite layer that connects AWS Ground Station, Direct Connect and regional centers so customers can compute where fiber does not reach.
Why the deadline matters for institutional capital
A constellation license is a promise to build, not a parking spot for spectrum. The FCC uses "use it or lose it" milestones precisely so nobody warehouses frequency. Amazon's waiver changed that bar. A satellite license that previously forced deployment now traded time for a spectral penalty.
For a principal evaluating a private space-asset bet, the read-through is direct. The waiver does not kill Leo; it reschedules it. What it changes is price. Satellites are worth less when a license is not a guaranteed timeline. Every month of delay pushes back the AWS-edge revenue case and delays the moment Leo becomes the actual second constellation rather than a second license.
The difference is between operating capacity and licensed capacity. Starlink turns 10 million customers into steady revenue. OneWeb sells lower-margin enterprise capacity. Leo still has neither. Yet this is already the third-largest commercial broadband network in orbit, a network with zero paying users.
Whether Leo lifts cadence toward the intermediate deployment thresholds the FCC set for this year.
Whether the spectral penalty pushes its launch schedule into earlier, smaller batches.
Whether subscriber pricing appears before the 2029 full-constellation deadline.
The gap between Starlink and Leo is not a technical failure; it is a negotiation for time. The 10,400-satellite gap is the entire trade. Those who licensed the band early now run scale; those who came late are negotiating for the right to do so.
