Amazon Leo vs Starlink: Constellation Size, Speed, Price and Coverage Compared

The short answer

Starlink wins every comparison you can measure today. It has 11,102 satellites in orbit against Amazon Leo's roughly 391; it had 12.0 million subscribers in 167 countries, territories and other markets on June 30, 2026; SpaceX's Connectivity segment, primarily Starlink, had revenue of $11.4 billion in 2025 (SpaceX prospectus); and it launches on rockets its parent company owns. Amazon Leo, the constellation Amazon renamed from Project Kuiper in November 2025, has an FCC licence for 3,232 satellites, a factory, launch contracts with four providers, a stated investment of more than $10 billion, and no commercial service yet.

The interesting question is not who is ahead but whether the gap is structural or just early. Amazon has the balance sheet, the retail channel and AWS, and it has been willing to lose money for a decade to win a market before. SpaceX has a five-year head start, a launch cost floor nobody can match, and a network that is already the benchmark. This guide is organised around that tension.

Constellation size (verified counts)

Satellite counts are the most-quoted and most-misquoted number in this comparison. The figures below are from our curated constellation registry, verified against primary tracking sources with an explicit count date; the live catalogue totals are on How many satellites are in orbit?

The scale difference is not a matter of Amazon being behind schedule on a similar plan. Even fully deployed, Amazon Leo's 3,232 satellites would be less than a third of what Starlink flies today, and on January 9, 2026 the FCC raised SpaceX's second-generation authorisation to 15,000 satellites, on top of its first-generation licence. Amazon's bet is that a smaller constellation of more capable satellites in slightly higher orbits can deliver competitive service without matching SpaceX satellite-for-satellite.

Growth rate is the number to watch. SpaceX counted about 9,600 Starlink broadband and mobile satellites in orbit on March 31, 2026 and more than 10,200 in its August 2026 quarterly report; it launched 165 Falcon 9 rockets in 2025. It is also starting to launch its larger V3 satellites on Starship: the first 20 flew on a suborbital test on July 24, 2026 and reentered by design, and Flight 14, the first orbital Starship flight, deployed 26 on September 28, 2026, SpaceX reported. Amazon Leo went from a first pair of prototypes in late 2023 to production launches in 2025 and, by its own count, 396 satellites deployed in early July 2026: a constellation in build-out, and still a long way from 3,232.

Orbital shells and spectrum

Starlink flies in low shells: most of the first-generation constellation sat at about 550 km, and on January 1, 2026 SpaceX's Starlink engineering head said it would lower all of its satellites at about 550 km to 480 km over the course of 2026. Low altitude buys low latency and a self-cleaning orbit (a failed satellite decays within a few years without help) at the cost of needing more satellites to cover the same ground. SpaceX says the network can serve everywhere on Earth, including the poles, and that the constellation carried more than 23,000 inter-satellite lasers on March 31, 2026, which route traffic through orbit rather than relying only on ground stations.

Amazon Leo is licensed by the FCC for 3,232 satellites (the 2020 grant was for 3,236 in 98 orbital planes at 590, 610 and 630 km): a little higher than Starlink, which trades a few milliseconds of latency for a larger footprint per satellite and a slower natural decay. Amazon says its terminals use Amazon-designed chips, and its satellites are built at its facility in Kirkland, Washington.

Both networks use Ku- and Ka-band spectrum coordinated through the ITU and licensed nationally through the FCC, and both are parties to the increasingly crowded coordination between LEO constellations and geostationary operators. The FCC's 5-year post-mission deorbit rule applies to both; Starlink's lower shells make compliance easier by physics, while Amazon Leo's 590-630 km orbits require the onboard propulsion the satellites carry. Our debris and traffic-management guide covers the rules.

Speed, latency and coverage

Starlink: SpaceX prospectus (June 2026) and Form 10-Q for the quarter to June 30, 2026. Amazon Leo: Amazon's November 2025 terminal announcement, as reported by Via Satellite; Light Reading (July 2026).

The asymmetry in the table is the whole story. Starlink's numbers are medians SpaceX reports for a network with millions of users; Amazon's are design figures for terminals. A designed peak says nothing about what a customer in a busy cell will see once the network has users, which is the reality of shared satellite capacity that Amazon Leo will meet too.

Coverage is where the head start compounds. Starlink's laser links and dense shells give it working service over oceans, in polar regions and in countries with no local ground station. Amazon Leo will have to build all of that, though AWS ground stations and Amazon's global logistics give it a better starting position than any previous challenger. Our Starlink vs OneWeb comparison covers the enterprise-only alternative.

Price

Starlink sets prices by plan and by country and changes them often, so check starlink.com for your address rather than trusting a number on any third-party page, this one included. The one figure SpaceX reports to investors is the average: Starlink subscriber ARPU (average revenue per subscriber) was $66 a month in the second quarter of 2026, down from $85 a year earlier.

Amazon Leo has published no service plans or prices. Amazon's history (retail, AWS, Alexa) is of absorbing early losses to buy share, and it has two distribution levers no other constellation operator has: Amazon.com and its physical retail for terminals, and the Prime relationship. Whether it uses either for Leo is not yet announced.

The hidden variable is terminal cost. In his 2022 letter to shareholders, Andy Jassy said Amazon expected to produce its standard residential terminal for less than $400 each, using Amazon-designed baseband chips. A low monthly price only works if the hardware subsidy behind it is survivable.

Launch supply: vertical integration vs a multi-provider buy

This is the deepest structural difference. SpaceX launches Starlink on its own rockets. SpaceX launched 165 Falcon 9 rockets in 2025, 157 of them on flight-proven boosters, and does not charge its own constellation a list price: it capitalises Starlink launch costs rather than booking them as revenue. Starship, which SpaceX says will carry its larger V3 satellites, released 20 of them on a suborbital test flight on July 24, 2026 (they reentered by design); its first orbital flight, Flight 14 on September 28, 2026, deployed 26 into the constellation, SpaceX reported, so SpaceX has a second, larger vehicle on the way, though still in flight testing. The rocket company and the constellation company are the same company, and each subsidises the other.

Amazon buys launches. In April 2022 it booked up to 83 launches (38 on ULA's Vulcan Centaur, 18 on Arianespace's Ariane 6 and 12 on Blue Origin's New Glenn, with options for 15 more), on top of earlier Atlas V purchases, and added SpaceX's Falcon 9 after shareholders sued over the original supplier choice. Diversification protects against any one rocket being grounded; it also means Amazon's deployment pace is hostage to other people's factories. Records from our rocket pages:

The 2026 reality of that manifest is uncomfortable. Atlas V flew its last Amazon Leo mission on July 2, 2026; its remaining rockets are contracted to Boeing for Starliner. Vulcan has four flights and had its national-security launches paused after a solid-rocket-booster anomaly on February 12, 2026, and had not flown again as of October 9, 2026. New Glenn's third flight in April 2026 left its payload in too low an orbit, a vehicle exploded on the pad in May, and Blue Origin hopes to return it to flight by the end of 2026. That leaves Ariane 6 and, ironically, Falcon 9 carrying much of the near-term load. Amazon's multi-provider strategy was designed for exactly this kind of resilience, but it was not designed for three of the four providers to have problems at once.

Business models and government work

Starlink is a consumer business with enterprise, mobility and direct-to-phone lines on top, and a government business, Starshield, alongside. SpaceX's Connectivity segment, primarily Starlink, had revenue of $11,387 million and operating income of $4,423 million in 2025, its prospectus says, about 61% of SpaceX's $18.7 billion total, and it had 12.0 million subscribers on June 30, 2026. Its use by Ukrainian forces since 2022 made it the benchmark for resilient military communications, and SpaceX describes Starshield as a secure network designed for U.S. Government customers and national-security applications.

Amazon Leo is positioned less as satellite internet than as “AWS connectivity everywhere.” The pitch to enterprises is satellite data landing directly in AWS regions; the pitch to carriers and airlines is connectivity: Amazon Leo has contracts with JetBlue and Delta Air Lines and a deal with AT&T to connect business customers (Light Reading, July 2026). AWS's existing government cloud business is the natural entry to government work, and that door is opening: on August 13, 2026 the Space Force awarded $60 million across five companies, Amazon Leo among them, to prototype connections to the SpaceX-built data network backbone. Note the shape of that award: the Pentagon is deliberately cultivating a multi-vendor ecosystem on top of SpaceX's infrastructure, which is the opening Amazon needs.

Sovereignty is the other lever. Governments in Europe, India and elsewhere are reluctant to depend on a single U.S. company for critical communications; a second American constellation with different ownership does not solve that, but many buyers will still prefer a two-vendor strategy, and Amazon is the obvious second vendor for anyone who does not want OneWeb's enterprise-only model.

The honest state of Amazon Leo service

As of October 11, 2026, Amazon Leo is not yet commercially available. Production launches began in 2025, the in-orbit fleet is small relative to the licence, service plans and prices are unannounced, and Amazon said in July 2026 that it expects to launch service later in 2026 while serving beta customers. The FCC's condition was that half the constellation (1,616 satellites) be in orbit by July 30, 2026. Amazon asked in January 2026 for a 24-month extension or a waiver, saying it expected about 700 by that date; on June 5, 2026 the FCC granted a limited waiver, with conditions, and kept the July 30, 2029 deadline for all 3,232.

What has happened is real, though. The Kirkland factory is producing satellites; Amazon counted 396 deployed in early July 2026; it has shown terminals designed for up to 1 Gbps; enterprise beta customers are connected; and the Space Force is paying Amazon Leo to prototype connections to military networks. That is a constellation in build-out, not a paper project. It is also, five years after Starlink's public beta, still at the stage Starlink was at in 2020.

If you need satellite broadband today, the comparison is not close: Starlink is the answer almost everywhere it is licensed. If you are an enterprise or carrier planning a 2027-2028 procurement, Amazon Leo is a credible second bidder worth keeping in the process, if only for the leverage it gives you on the first one.

What to watch

• Amazon Leo's launch cadence given the state of its providers: track Vulcan, New Glenn and Ariane 6.
• A published Amazon Leo price and service launch. Amazon has said only that service starts later in 2026.
• Starship's Starlink V3 cadence. SpaceX says each V3 satellite is designed to offer one Tbps of downlink capacity; routine Starship flights would widen the gap faster than any Amazon launch schedule can close it. Starship tracker.
• Direct-to-cell. Starlink Mobile works with T-Mobile and other carriers in about 30 countries, SpaceX says; Amazon agreed in April 2026 to buy Globalstar for about $11.6 billion, a deal expected to close in 2027, to move into direct-to-device service. The first network that reliably connects unmodified phones at scale wins a market neither serves today.
• SpaceX's quarterly reports as a public company (SPCX): subscriber counts and ARPU every quarter. Space stocks.

Frequently asked