Amazon Leo is the satellite broadband network Amazon spent six years calling Project Kuiper. By late October 2026 it had nearly 400 satellites in orbit across 14 missions, more than 1,000 satellites built, and a promise on Amazon's Q3 2026 earnings call that the network remains on track to begin initial service this year.

It's also behind its own schedule, short of its FCC deployment milestone, and waiting on two rockets that weren't flying for most of 2026. So what is Amazon actually building, which numbers are hard facts versus Amazon targets, and how far is Leo from being a real Starlink competitor?

Key Takeaways

  • Amazon Leo is the renamed Project Kuiper, a low Earth orbit broadband constellation with an initial plan of more than 3,000 satellites at 590 to 630 km. The rebrand happened in November 2025.
  • About 396 production satellites have launched on Atlas V, Falcon 9 and Ariane 64, per Light Reading's October 2026 count. That's roughly 12% of the 3,232 satellites the FCC license requires.
  • The FCC gave Amazon a limited waiver on its July 2026 halfway milestone but kept the July 30, 2029 deadline for all 3,232 satellites, with a spectrum-priority penalty for late satellites.
  • Three terminals cover consumers to enterprises: Leo Nano at up to 100 Mbps, Leo Pro at up to 400 Mbps and Leo Ultra at up to 1 Gbps down, with direct links into AWS.
  • Inside Deep Tech's take: Leo's limit isn't satellite quality or factory output. It's launch cadence. Vulcan and New Glenn carry most of the backlog, and neither flew a Leo mission through October 2026.

What Amazon Leo is

Amazon Leo is a constellation of broadband satellites in low Earth orbit, plus the ground gateways, customer antennas and network software that connect them to the internet. Amazon renamed Project Kuiper to Amazon Leo on November 13, 2025. The program sits inside Amazon's Devices and Services organization, not Blue Origin, and is headquartered in Redmond, Washington.

The pitch is the same one Starlink made: fast, low-latency internet anywhere a terminal can see the sky. Low orbit matters because a satellite at roughly 600 km is about 60 times closer to the ground than a geostationary one, which cuts round-trip delay enough for video calls, gaming and cloud apps. Amazon's twist is the enterprise and AWS angle, plus the ability to bundle service with the rest of the Amazon ecosystem.

🛰️
Inside Deep Tech's take: Leo is best understood as an AWS edge network that happens to live in orbit. The consumer dish gets the headlines, but the Ultra terminal's private links into AWS are the product Starlink can't easily copy.

Amazon Leo at a glance

ItemAmazon Leo (Gen1)Status, Oct 2026
Former nameProject KuiperRenamed Nov 13, 2025
Licensed satellites3,232 in three shellsAbout 396 launched
Altitude590 to 630 kmDeployed at 465 km, raised to 630 km
FCC final milestone3,232 by Jul 30, 2029Unchanged by the 2026 waiver
Launch contracts100+ missions, 5 rocketsAtlas V, Falcon 9, Ariane 64 flown
Satellites builtUp to 5 per day capacity1,000+ built
ServiceConsumer, enterprise, governmentEnterprise preview since Nov 2025

Source: Amazon Leo FAQ; Amazon mission updates; FCC order DA-26-553; Light Reading production report.

How the constellation works

The first-generation system was originally authorized by the FCC in July 2020 at 3,236 satellites, later modified to 3,232 satellites. Satellites launch into a lower parking orbit and then use onboard electric propulsion to climb. Amazon's mission log says each batch is deployed at about 465 km and then raised to 630 km, the top of its operating range.

Each satellite relays traffic between customer terminals and ground gateways that connect to fiber. The network design lets Amazon hand enterprise traffic straight into AWS rather than onto the nearest public internet exchange, which is the basis of the Direct to AWS service described below.

Gen2, Polar and the direct-to-device layer

Leo is already licensed beyond Gen1. The same FCC order notes authorizations for a Gen2 system of 3,212 satellites and a Polar system of 1,292 satellites, each with a 50% milestone due February 10, 2032 and full deployment by February 10, 2035.

Separately, Amazon has filed for a direct-to-device network of up to 5,105 satellites, with deployment starting in 2028. Those satellites would use L-band and S-band spectrum to reach ordinary phones, Ka-band and V-band for backhaul, plus onboard processing and laser links. The pending Globalstar acquisition, covered below, is how Amazon plans to get that spectrum.

Amazon Leo terminals: Nano, Pro and Ultra

Amazon has designed three customer antennas, all flat electronically steered arrays, plus a separate aviation antenna. Amazon says the Ultra design eliminates moving parts and that its antennas run on a custom chip designed by Amazon Leo.

TerminalSizePeak downloadTarget customer
Leo Nano7 by 7 inchesUp to 100 MbpsPortable, consumer, IoT
Leo Pro11 by 11 inchesUp to 400 MbpsHomes, small business
Leo UltraEnterprise panelUp to 1 Gbps down, 400 Mbps upEnterprise, government
Aviation antennaAircraft fuselageUp to 1 Gbps down, 400 Mbps upAirlines

Source: Amazon Leo FAQ for Nano, Pro and Ultra; Amazon aviation antenna announcement for the aircraft terminal.

Leo Ultra is the interesting one. Amazon describes it as a full-duplex phased array built on Amazon custom silicon that can send and receive at the same time. Enterprise customers can connect it to AWS through Direct to AWS, which plugs into Transit Gateway and Direct Connect, or through Private Network Interconnect into their own networks. Preview customers test the network on production hardware and software.

☁️
Why it matters: a Leo Ultra site can reach an AWS VPC without traffic ever touching the public internet. For oil fields, ships, mines and military sites already running on AWS, that's a cleaner architecture than a consumer dish plus a VPN.

Launches: where the satellites actually came from

Amazon booked what it called the largest commercial procurement of launch vehicles in history, and it now says the manifest covers more than 100 missions across five rockets: Atlas V, Ariane 6, Falcon 9, Vulcan and New Glenn. The first production launch, KA-01 on an Atlas V, flew on April 28, 2025.

RocketSatellites per flightLeo flights flown (Oct 2026)Contracted / status
ULA Atlas V27 to 298 production flightsAll Atlas V missions flown
SpaceX Falcon 9243Completed
Arianespace Ariane 6432 to 36324 launches, LE-04 NET Nov 23
ULA Vulcan40038 launches, LV-01 Nov 13
Blue Origin New Glenn48024 launches, LN-01 undated

Source: Amazon mission updates; Amazon on doubling its launch rate for per-rocket capacity; Amazon Arianespace expansion; ULA Vulcan Leo 1 page; Aviation Week on New Glenn.

The math behind the roughly 396 production satellites is simple. Atlas V contributed about 224 across eight flights, Falcon 9 added 72 across three, and Ariane 64 delivered 100 satellites in three missions. Ariane 64's capacity also grew: new P160C boosters added more than 2 tonnes of lift and room for four more satellites per flight.

Vulcan and New Glenn are the backlog

The two rockets that matter most for the 2029 deadline haven't flown Leo yet. ULA's first Leo mission, LV-01, is a Vulcan VC6L with six solid boosters carrying 40 satellites, the first of 38 contracted Vulcan launches. Amazon lists it for November 13, 2026, pending range approval, and says the vehicle flies with an improved solid rocket booster nozzle design.

New Glenn is further back. Its first Leo mission, LN-01, was set to be the first of 24 New Glenn launches for Amazon, carrying 48 satellites. The vehicle being prepared for that mission was lost in an explosion at Launch Complex 36, and SpaceNews reports Blue Origin now expects its rebuilt pad to be ready by the first half of December for a return to flight this year. LN-01 has no confirmed date.

Amazon had planned 20-plus missions for year two of deployment. Amazon's mission log shows seven Leo launches in 2026, and none since LA-08 on July 2. The year will land far below plan even if Vulcan and Ariane both fly in November.

🚀
Inside Deep Tech's take: every Leo forecast is really a Vulcan and New Glenn forecast. Those two rockets account for 62 of the contracted missions and the biggest payloads per flight. Until both fly on a steady monthly rhythm, treat Amazon's deployment targets as aspirational.

For how the other side of the launch market is evolving, read Inside Deep Tech's SpaceX Starship V3 full guide. Starship is the vehicle SpaceX needs to deploy its own next generation of Starlink satellites.

SpaceX Starship V3: A Full Guide
Raptor 3, Flights 12 to 14, Starlink V3, cadence, licenses and the honest limits of SpaceX's next launch system.

Manufacturing and ground infrastructure

Satellite output is no longer the bottleneck. Amazon's Kirkland, Washington factory can build up to five satellites per day, and Light Reading reports the 172,000-square-foot facility had built more than 1,000 satellites by October 2026. In its FCC waiver request, Amazon cited a production capacity of about 30 satellites per week.

Amazon is also paying for launch-site capacity. It has committed more than $200 million in ULA facility upgrades, runs a 100,000-square-foot satellite processing facility in Florida and is investing at Vandenberg. Its $140 million site at Kennedy Space Center can support up to three simultaneous launch campaigns.

The ground segment is less visible but just as important. Gateways connect satellites to fiber, and Amazon's AT&T deal includes using Leo to extend AT&T's network, with AWS also involved in modernizing AT&T's infrastructure, according to the February 2026 joint announcement.

The FCC deadline and the waiver

Amazon's license required half of the Gen1 system, 1,616 satellites, by July 30, 2026, and all 3,232 by July 30, 2029. In a January 30, 2026 filing, Amazon told the FCC it had 180 satellites in orbit and expected about 700 by the July 2026 milestone. That wasn't close.

On June 5, 2026 the FCC's Space Bureau granted a limited waiver in order DA-26-553. Amazon didn't lose its license, and the bureau declined to move the final 2029 milestone. Instead, satellites launched after the missed milestone lose their spectrum priority against other systems from the same processing rounds for 20 months, until March 30, 2028, or until Amazon reaches 50% deployment. The order shortens that window to 15 months if Amazon certifies it has built the satellites and secured the launches needed for 50%. SpaceX opposed the waiver.

The gap is large. With about 396 production satellites up, Amazon still needs roughly 2,836 more to reach 3,232 by July 30, 2029. That's about 86 satellites a month for the next 33 months, or roughly two full Vulcan or New Glenn launches every month on top of Ariane 64 flights.

⚠️
Watch this: if Amazon falls well short in 2029, the FCC can reduce the authorized constellation to the number of satellites actually in orbit. The waiver bought time on the midpoint, not on the finish line.

Customers and partners

Amazon began an enterprise preview in November 2025, with a wider rollout planned for 2026. Amazon says it has signed agreements with customers and partners including JetBlue, Vanu, Hunt Energy Network, Connected Farms and Crane Worldwide Logistics. Larger distribution deals so far:

Notice the dates. Most consumer-facing deals start in 2027 or 2028. NBN Co's page still says Amazon planned Australian service from the middle of 2026, a date that has already slipped. Light Reading also reports Amazon executives have hinted at pairing Leo with Prime, though no Prime bundle has been announced.

The Globalstar deal and phone service

Amazon signed a definitive agreement to acquire Globalstar, the satellite operator that powers Apple's emergency SOS and messaging features. Globalstar shareholders can take $90.00 in cash or 0.3210 Amazon shares per share, with the cash portion capped at 40% of the total. Holders of about 58% of the voting power have approved it, and closing is expected in 2027, subject to regulatory approvals and Globalstar's HIBLEO-4 milestones.

The deal also includes an agreement with Apple covering satellite features on iPhone and Apple Watch. Combined with the planned 5,105-satellite direct-to-device system, it gives Amazon a path into the same phone-to-satellite market that SpaceX is chasing with Direct to Cell. For the broader trend, see Inside Deep Tech's look at satellite-to-mobile connectivity among the 7 space tech trends of 2026.

Starlink is the benchmark, and the gap is wide. SpaceX ended 2025 with more than 9 million Starlink customers after adding over 4.6 million in the year, and Fierce Network reports it reached 13.6 million subscribers in 174 markets by Q3 2026.

MetricAmazon LeoStarlink
Satellites in orbitAbout 392 (396 launched)More than 10,000
CustomersEnterprise preview only13.6 million (Q3 2026)
MarketsNone commercially live174
Launch providerFour outside providers plus Blue OriginSpaceX's own Falcon 9
Top terminal speedUp to 1 Gbps (Ultra)Varies by plan
Cloud integrationDirect to AWSPartner peering
Phone servicePlanned 5,105-satellite system from 2028Direct to Cell in service

Source: KeepTrack Kuiper count and Light Reading for Leo in orbit; GeekWire for Starlink's fleet size; Fierce Network for Starlink subscribers and markets; Amazon Leo FAQ and Amazon direct-to-device explainer.

Starlink's structural advantage is vertical integration. SpaceX flew more than 120 Falcon 9 Starlink missions in 2025 on its own reusable rocket. Amazon has to buy every launch from someone else, including its founder's separate rocket company. Leo's advantages are Amazon's balance sheet, AWS distribution and enterprise sales relationships, and a later design that could use newer silicon from day one.

For investors weighing the two, Inside Deep Tech's guide to space tech investment trends covers why connectivity infrastructure now attracts more capital than moonshots.

The money: what Leo costs Amazon

Amazon doesn't break out Leo spending as a segment, but its Q3 2026 call gave one clear signal. CFO Brian Olsavsky said Amazon will begin capitalizing certain Leo costs, including production and launch, in Q1 2027, later than he had predicted on earlier calls. Until then, the majority of Leo costs stay in operating expenses.

That means Q4 2026 results will still carry most of Leo's cash burn as expense, even as Amazon guided Q4 operating income to between $28.5 billion and $33.5 billion. The switch to capitalization in 2027 will make Leo look cheaper on the income statement without changing its cash cost.

Honest limits of Amazon Leo

The real limits as of November 2026:

  • Launch cadence. Amazon planned 20-plus missions in 2026 and flew seven, the last on July 2. Vulcan and New Glenn, which hold most of the backlog, hadn't flown a Leo mission.
  • Schedule slips. Leo missed its mid-2026 service target, per Light Reading, and NBN Co's mid-2026 Australian launch date has passed.
  • Regulatory clock. The FCC kept the July 30, 2029 full-deployment milestone, and late satellites lose spectrum priority until 2028.
  • Density. About 400 satellites can't deliver continuous global coverage. Initial service will be regional and capacity-limited until the constellation grows.
  • No public pricing. Amazon hasn't published consumer prices or data caps, so price competition with Starlink is still unproven.
  • Phone service is years away. The direct-to-device system starts deploying in 2028, and the Globalstar deal won't close until 2027.
📡
Inside Deep Tech's take: Leo isn't failing, it's late. The satellites work, the factory works and the enterprise product is differentiated. The 2027 test is whether Vulcan and New Glenn can deliver two heavy launches a month. If they can, Leo becomes a real second network. If not, it stays a niche enterprise service.

If orbital compute interests you as much as orbital connectivity, Inside Deep Tech's analysis of data centers in space covers where the two ideas overlap.

What to watch next

  • Vulcan LV-01 on November 13, 2026: the first 40-satellite Leo flight and the first test of ULA's fixed booster nozzle on a Leo mission.
  • Ariane 64 LE-04 no earlier than November 23, which would bring the Arianespace total to 136 satellites.
  • New Glenn return to flight after the LC-36 rebuild, which Blue Origin targets before the end of 2026.
  • Initial service launch, which Amazon says will begin before the end of 2026. Watch which countries and customer types go first.
  • Q4 2026 earnings in early 2027 for satellite counts, first revenue hints and the capitalization switch.

FAQ

Is Amazon Leo the same as Project Kuiper?

Yes. Amazon renamed Project Kuiper to Amazon Leo on November 13, 2025. The satellites, FCC licenses and launch contracts are the same program.

When will Amazon Leo service start?

Amazon said on its Q3 2026 earnings call that Leo remains on track to begin initial service this year. An enterprise preview has run since November 2025, and many airline and telecom deals begin in 2027 or 2028.

How many Amazon Leo satellites are in orbit?

About 392. Amazon had launched 396 production satellites by October 2026 and deorbited four, per Light Reading, and KeepTrack shows a similar count. Amazon describes it as nearly 400 satellites across 14 missions.

How fast is Amazon Leo?

Amazon rates Leo Nano at up to 100 Mbps, Leo Pro at up to 400 Mbps and Leo Ultra at up to 1 Gbps down and 400 Mbps up. Real-world speeds will depend on constellation density and network load.

Is Amazon Leo part of Blue Origin?

No. Leo is an Amazon program in its Devices and Services organization. Blue Origin is a separate company that sells Amazon launches on New Glenn.

How does Amazon Leo compare with Starlink?

Starlink is far ahead, with 13.6 million subscribers in 174 markets as of Q3 2026. Leo has a few hundred satellites and no broad commercial service yet, but offers direct AWS integration and three terminal sizes.

The bottom line

Amazon Leo has the satellites, the factory, the terminals and a growing list of airline, telecom and enterprise customers. What it doesn't have yet is launch cadence. With roughly 12% of its licensed Gen1 satellites in orbit and a firm 2029 FCC deadline, the next 12 months of Vulcan, Ariane 64 and New Glenn flights will decide whether Leo becomes a genuine Starlink rival or a well-funded enterprise niche.

For the launch system on the other side of that race, read Inside Deep Tech's SpaceX Starship V3 full guide.