On May 26, American Airlines announced it would equip more than 500 of its aircraft with Starlink Wi-Fi. The stock jumped 6%. The market’s reaction was instant, and telling.
SpaceX has now signed three of the four largest US airlines: United, Southwest, and American, covering over 2,300 commercial aircraft. Only Delta remains. An entire industry is migrating to one provider, and that provider happens to own the rockets that put the satellites up.
This is what a moat looks like in real time.
But to understand why Starlink won, and why conventional analysis keeps underestimating what it’s worth, you need to start with the graveyard.
The Graveyard of Good Ideas
The idea of connecting the world via satellite was not Elon Musk’s. In 1990, Craig McCaw and Bill Gates founded Teledesic with a plan to launch 840 satellites into low-Earth orbit and deliver global broadband. The investors were serious. The vision was correct. The company went bankrupt in 2002.
Teledesic wasn’t alone. Iridium went bankrupt in 1999 with $1.5 billion in debt. Globalstar filed in 2002. OneWeb, backed by SoftBank and Airbus, filed for bankruptcy in 2020 and had to be rescued by the British government.
The pattern is consistent: visionary concept, credible backers, correct market thesis, bankruptcy. All major attempts to build LEO satellite businesses failed at first. Except one.
What did SpaceX do differently?
The Vertical Integration Answer
Every company before Starlink had to buy launches. Teledesic was dependent on Boeing, both an investor and prime contractor, with all the misaligned incentives that implies. OneWeb paid Arianespace and later Roscosmos until Russia’s invasion of Ukraine stranded 36 of its satellites and forced a scramble for alternatives.
SpaceX builds the rockets. SpaceX manufactures the satellites. SpaceX makes the user terminals. The vertical integration that made Starlink possible is the same vertical integration that makes it nearly impossible to dislodge.
The cost differential is structural, not cyclical. A Falcon 9 launch costs SpaceX roughly $15–20 million at full capacity. Starlink launches 60 satellites per mission, around $250–350k per satellite in launch costs. A competitor buying launches commercially pays $50–60 million for the same rocket. The moat is baked into the cost of goods.
Starlink launched its first satellites in 2019. Beta services reached consumers by late 2020, ahead of most expectations. The $10 billion capital programme that followed has produced 9,600 satellites across 164 countries and 10.3 million subscribers.
At 61% of SpaceX’s total revenue, Starlink isn’t a feature. It’s the engine.
Why TAM Models Always Get It Wrong
When Starlink launched, conventional TAM analysis said: rural broadband. Underserved households without fibre. The addressable market was estimated at tens of millions of subscribers. A niche business.
That analysis correctly identified the initial market. It failed entirely to anticipate what came next.
Big ideas create markets that couldn’t be anticipated at launch. The original iPhone TAM was “smartphones.” The actual TAM turned out to be the entire global economy, restructured around a device nobody had imagined. Starlink’s TAM is working the same way.
Aviation: The Market Nobody Modelled
In-flight connectivity was, until recently, a pain point masquerading as a product. Gogo’s air-to-ground system was famously slow. Viasat’s satellite offering was better but plagued by congestion and latency. The market was worth a few billion dollars globally and growing slowly.
Starlink’s Aero Terminal delivers up to 1 gigabit per second per antenna. Gate to gate. The American Airlines deal covers its full Airbus narrowbody fleet. United signed in 2024. Southwest followed. Three of the four largest US carriers, over 2,300 aircraft. Only Delta remains.
American’s stock jumped 6% on the announcement. The market wasn’t pricing in better Wi-Fi. It was pricing in the strategic value of locking passengers into a seamlessly connected loyalty ecosystem, gate to gate. The in-flight connectivity market, perhaps $5 billion annually in 2015, is now part of Starlink’s TAM. Nobody modelled it.
Defence: Picking Sides in the Next Cold War
Nothing demonstrated Starlink’s strategic value more clearly than Ukraine. When Russian forces invaded in February 2022, Starlink terminals arrived within days. Ukrainian military commanders credited Starlink with maintaining battlefield communications when terrestrial infrastructure was destroyed. It became NATO’s communications backbone in the theatre.
Starlink has made an explicit geopolitical choice: no service to Iran, Russia, or China. This isn’t just ethics. It’s a strategic alignment with the Western alliance, and the defence contracts that follow are significant. Military service fees have surged. The Pentagon is a customer. NATO allies are customers. The defence TAM alone runs into the tens of billions annually.
The next cold war has as much to do with the economics of space as with terrestrial military capability. Starlink has chosen its side. The contracts are following.
The Copper Wire Moment
In emerging markets, mobile phones didn’t complement copper wire telephone networks. They made them redundant. Billions of people skipped fixed-line telephony entirely and went straight to mobile. The incumbent telecoms companies found their competitive advantage evaporating in a decade.
Starlink is the same moment for broadband. Where fibre is uneconomic and cable infrastructure absent, Starlink doesn’t compete with terrestrial broadband. It makes the question irrelevant. The incumbents facing this don’t yet know it — which is exactly what their counterparts in fixed-line telecoms said in 1995.
The Flywheel Nobody Can Replicate
Starlink generates $7.2 billion in adjusted EBITDA annually. That cash funds Starship development. Starship reduces the cost per satellite launched. Lower costs increase Starlink’s margins. Higher margins fund more Starship development. The flywheel compounds.
Amazon’s Project Kuiper is the only credible competitor. It has Blue Origin for launches and $10 billion committed. But Blue Origin’s New Glenn is years behind Starship in cadence and reusability. Kuiper buys launches at commercial rates while Starlink launches at marginal cost. The unit economics don’t close at the same subscriber price point.
OneWeb, now owned by Eutelsat, targets enterprise and government markets with a partial constellation and no launch infrastructure of its own. It’s a real business. It isn’t a flywheel.
The moat Teledesic couldn’t build in 1990, Starlink has built by owning the entire stack. It’s the same lesson as Issue 02: vertical integration isn’t just a business model. It’s a compounding competitive advantage that gets harder to close every quarter.
What This Means for the Space Economy
Issues 01 and 02 established two things: the $100/kg threshold is the unlock for the orbital economy, and SpaceX’s vertical integration means one company is positioned to profit from almost every layer of it.
Issue 03 adds the proof. Starlink is the proof that this model works. $11.4 billion in revenue, $7.2 billion EBITDA, 86% year-on-year growth, three of the four largest US airlines signed, the Pentagon as a customer, and a flywheel that compounds every time a Starship launches.
The markets that couldn’t be anticipated in 2015 are now Starlink’s fastest-growing segments. The markets that haven’t yet been imagined will follow the same pattern. TAM models will keep getting it wrong, because they always do with genuinely transformative infrastructure.
Creative destruction isn’t a theory. It’s a bill that comes due. Teledesic, Iridium, Globalstar, OneWeb all tried and failed because the conditions weren’t right. Starlink succeeded because SpaceX created the conditions itself.
That’s exactly what The View from Space tracks.