The Space Broadband Margin War
Why Your Satellite Broadband Thesis Depends on Which Business You're Actually Buying
What This Means:
The space broadband market is not one race — it is three distinct businesses competing with incompatible cost structures, customer bases, and path-to-profitability timelines. Starlink’s residential average revenue per user (ARPU) of $90–120/month is already funding its next phase of enterprise expansion, while Amazon Leo (formerly Project Kuiper) has yet to prove it can convert $25 billion in infrastructure spend into durable margin. Meanwhile, direct-to-device (D2D) challengers led by AST SpaceMobile are betting on wholesale telco partnerships rather than consumer subscriptions — a model that changes who captures the economics. Investors holding positions in any of the three need to understand which business they actually own.
Scope note: This analysis focuses on the three highest-capitalization consumer and enterprise LEO broadband plays. OneWeb/Eutelsat and Telesat Lightspeed are excluded from this comparative framework; both are navigating financial restructuring or early deployment phases that introduce spectrum and market-share variables outside the scope of this analysis.
Three Models, One Market
There is a version of the space broadband story that gets told at every industry conference: satellite internet is finally ready, latency has fallen, coverage maps have expanded, and the demand is bottomless. That version is mostly true. What it omits is that three fundamentally different business models are now competing for the same subscriber dollar and they will not all get there on the same schedule, at the same margin, or through the same channel.
Starlink, operated by SpaceX, surpassed 10 million subscribers globally by February 2026, after crossing the 9 million threshold in December 2025, according to Broadband Breakfast and Quilty Space’s March 2026 Financial Overview adding approximately 1 million subscribers in just 53 days. Residential ARPU runs $90–120/month in the United States and $30–80/month in most international markets. Analyst estimates of Starlink’s 2025 revenue diverge materially: Quilty Space’s March 2026 forecast puts the figure at $10–12 billion across consumer and enterprise segments, while other industry trackers place the estimate considerably higher, one analysis citing figures above $18 billion, reflecting genuine methodological disagreement on how enterprise, mobility, and government contracts are accounted for. Enterprise and mobility contracts in aviation, maritime, government, and defense drive ARPU well above the residential average, with some segments reaching $1,500–1,700 annually per account. None of these figures come from SpaceX’s own public disclosures, the company does not report Starlink financials as a separate line item, which itself represents a persistent measurement gap for any investor trying to size the business. Quilty Space’s March 2026 forecast projects year-end 2026 subscriber count of 16.8 million and $11.3 billion in consumer revenue alone, with enterprise scaling from $584 million in 2024 to $1.68 billion by 2026.
Amazon rebranded Project Kuiper to Amazon Leo in November 2025 and initiated enterprise beta on April 8, 2026, with Verizon, AT&T, Vodafone, JetBlue, and NASA among the initial partners. The service offers three terminal tiers: standard enterprise terminals deliver up to 400 Mbps, while the premium Aviation Antenna reaches up to 1 Gbps download / 400 Mbps upload, with commercial availability targeted for mid-2026 — a timeline confirmed by CEO Andy Jassy in his April 2026 shareholder letter. A material regulatory constraint shapes that timeline: Amazon is required by the FCC to deploy at least half of its 3,236-satellite constellation — approximately 1,618 satellites — by July 2026. With approximately 1,500 satellites in orbit at beta launch and Amazon planning 20 additional launches in 2026, the company faces a narrow but specific execution window to meet compliance. FCC enforcement posture on constellation milestones has historically allowed phased remediation rather than immediate service suspension, but a compliance shortfall would introduce regulatory uncertainty into the mid-2026 commercial launch narrative and warrant close monitoring by investors. Truist Securities, in a June 2025 pre-beta analysis, projected that assuming a similar adoption pace to Starlink, Amazon Leo could deliver annual revenue of approximately $6 billion by 2030 — with internal Amazon projections cited in some analyses targeting a $20 billion revenue stream. That Truist estimate predates the enterprise beta launch and the April 2026 rebrand; it should be treated as a pre-commercial baseline, with post-beta analyst revisions likely to follow the first disclosed revenue recognition quarter. The gap between the $6 billion and $20 billion figures is not a typo — it reflects genuine analyst uncertainty about whether Amazon can monetize broadband at the premium margins Starlink has established, or whether it will run the service as a below-cost acquisition engine tied to Amazon Web Services (AWS) and Prime ecosystem retention.
Why Unit Economics Are the Scorecard That Matters
The framing of “subscribers gained” obscures the question that should matter more: at what cost was each subscriber acquired, and what does that subscriber contribute to margin over a multi-year period? Starlink’s advantage is not just first-mover scale — it is that SpaceX built and continues to launch its own rockets, manufactures its own satellites, and does not pay third-party launch costs at commercial rates. That vertical integration creates a structural cost floor that Amazon Leo, no matter how large its constellation, cannot fully replicate.
Amazon has contracted launch capacity from United Launch Alliance (ULA), Arianespace, and Blue Origin. As of late April 2026, ULA had completed six Atlas V missions for Amazon Leo, with three Atlas V missions remaining under contract before ULA transitions Amazon to its next-generation Vulcan Centaur rocket — a vehicle currently grounded following a February 2026 anomaly during a Space Force mission, with return-to-flight timing pending certification. Amazon’s remaining launch manifest also includes Ariane 6 and Blue Origin New Glenn, the latter of which experienced an orbital deployment anomaly in April 2026 that placed an AST SpaceMobile satellite in an incorrect orbit. These are not catastrophic program risks, but they illustrate a structural reality: Amazon Leo’s per-satellite deployment economics carry third-party margin and vehicle-risk baked in at the launch layer — a cost structure that Starlink does not face and one that will widen if Starship operationalizes at the cost curve SpaceX has projected.
That cost structure gap has real implications for subscriber pricing. Amazon Leo’s terminal tiers are competitive with Starlink on specs, but if Amazon prices aggressively to capture residential market share, it is burning through the $25+ billion capital expenditure in its constellation build without a clear path to the ARPU that would justify the investment on standalone economics. The Truist analyst frame — $6 billion in revenue by 2030 on pre-beta assumptions — implies Amazon Leo generates its primary value as a platform-enhancing asset that drives AWS contract retention and Prime ecosystem engagement, not purely as a broadband business. Goldman Sachs, in an April 2026 note, described Amazon Leo as aligned with Amazon’s broader AWS strategy, arguing the service “de-risks spectrum rights, supercharges plans, and improves returns” across Amazon’s cloud and connectivity ecosystem. For investors evaluating Amazon’s stock, that distinction matters: they are not buying a satellite internet company, they are buying an argument that broadband access makes Amazon’s other businesses stickier.
The next sections: Starlink's vertical integration cost floor, Amazon Leo's FCC compliance window and launch manifest risk, AST SpaceMobile's carrier contract threshold math, and the five specific 2026 data releases that will confirm or break each investment thesis. Paid subscribers get the full analysis, the Decision Questions framework, and the Related Decisions action checklist.
