Kuiper

Amazon’s LEO constellation — announced as Project Kuiper and now marketed as Amazon Leo — is the only competitor to Starlink with a balance sheet deep enough to lose money for as long as it takes, and a distribution machine nobody else in the sector can match.

It is also years behind, facing a regulatory deadline it is widely expected to miss, and dependent on rockets that other people own. Both things are true at once, and that tension is the whole story.

The Deadline

Amazon’s FCC authorisation carries a milestone that shapes everything else about this programme: half the constellation — 1,618 satellites — must be in orbit by 30 July 2026, with the full 3,236 by July 2029.

Production launches did not begin until April 2025, when an Atlas V carried the first 27 operational satellites. To reach 1,618 from that standing start inside fifteen months would have required a launch cadence the available vehicles could not physically deliver. The gap is not marginal; it is an order of magnitude.

The realistic outcomes are an extension — the FCC has granted them before, and has an obvious interest in not killing Starlink’s only credible competitor — or a modification of the milestone. An outright licence revocation is close to unthinkable. But “Amazon will ask for more time” is an assumption, not a fact, and the regulatory response is the single most consequential open question about this constellation.

Launch: The Structural Weakness

Amazon has bought one of the largest commercial launch campaigns ever procured — roughly 80-plus launches across four providers:

Provider	Vehicle	Launches contracted
United Launch Alliance	Vulcan Centaur	38
United Launch Alliance	Atlas V	9
Arianespace	Ariane 6	18
Blue Origin	New Glenn	12 (plus options)
SpaceX	Falcon 9	3

Note the last row. Amazon was obliged to buy launches from its principal competitor — a fact that tells you more about the state of the launch market than any market-sizing report will.

The deeper problem is that three of those four vehicles were new and unproven when contracted. Vulcan, Ariane 6 and New Glenn all had to reach reliable operational cadence before they could fly Kuiper at rate, and none did so on the schedule Amazon’s milestone assumed. Amazon does not control its own launch, and that single fact explains most of the delay. SpaceX’s advantage is not that it builds better satellites; it is that it never has to wait in someone else’s queue.

The Product

• Standard terminal — roughly 11 inches square, targeting around 400 Mbps.
• Compact terminal — around 7 inches, targeting ~100 Mbps; positioned as the cheapest LEO terminal on the market.
• Enterprise terminal — targeting ~1 Gbps for business and government.

Amazon has stated it can produce the standard terminal for under $400 — a claim that matters enormously, because terminal cost, not satellite cost, is what has historically strangled consumer satellite broadband. Consumer hardware at scale is one of the few things Amazon demonstrably does better than SpaceX.

The Real Strategy: AWS

Reading Kuiper as a broadband ISP misses the point. Amazon’s structural advantage is that the constellation plugs directly into AWS.

For an enterprise customer — a mining operator, a shipping line, an energy major, a government — the pitch is not “internet in remote places”. It is data from a remote asset landing inside the cloud environment where it is already being processed, under one contract, one bill and one security model. That is a bundle SpaceX cannot construct, and the enterprise segment is where the margin lives.

Add Prime as a consumer distribution channel and Amazon’s logistics network for terminal fulfilment, and the competitive question becomes less “can Amazon catch up on satellites” and more “does being three years late matter in a market this far from saturated”.