Amazon is buying rockets to buy back time

Amazon already had a big contract with Arianespace, the European launcher. On September 9th it made it bigger, raising its order for powerful Ariane 64 rockets from 18 to 24 flights, the latest in a series of purchases signed at France's International Space Summit. The knee-jerk reading is that Amazon's satellite-internet arm, known as Amazon Leo, needs more seats to space. The more telling one is that the company is spending heavily to compress the years its upstart constellation has lost to a rival whose head start it cannot buy back at any price.

The clue is in the timing. Arianespace has flown only three missions for Amazon Leo this year, depositing around 100 satellites in orbit. That is a trickle next to what the constellation is supposed to become — more than 3,200 spacecraft, whose name, dropped and replaced by "Leo" in a 2025 rebrand, still haunts its licence. The Federal Communications Commission granted the licence in 2020 on condition that Amazon put half the constellation, roughly 1,600 satellites, in orbit by July 30th 2026. Two months after that deadline, Amazon has perhaps 400 satellites across some 14 missions — around a quarter of the way there, and below even the 700 it promised the regulator in April it could deliver.

Amazon blew the deadline because it spent five years standing up a business. Its first production launch came only in April 2025, half a decade after authorisation; prototypes that flew in 2023 were deorbited and did not count. The FCC did not punish the miss. To the contrary, it granted a conditional waiver, refusing Amazon's request for a two-year extension but keeping the project alive.

The conditions are where the real economics live. For every satellite launched after the deadline, Amazon Leo loses its privileged claim on the radio spectrum — the scarce, contested resource that lets satellites talk to the ground. That priority is restored only if Amazon orbs half its constellation, or proves by October 2027 that it has built all the spacecraft and signed the launch contracts (even if they fly later), or waits it out until March 2028. In other words, the regulator has told the company it may be late, but it may not dawdle.

Why the leniency? Because the FCC has an interest of its own. Wallowing in a market dominated by one operator — SpaceX's Starlink, which has roughly 11,000 satellites in orbit, added about 1,300 during 2026 alone, and counts more than 10 million subscribers — is not a state a competition-minded regulator wants to entrench. A credible second entrant is worth a bent deadline. The waiver is, in effect, industrial policy dressed as accommodation: rules bent so that a significant challenger survives. It is a fair bet that a regulator would have been less forgiving to a would-be operator it did not believe could start.

The purchases are the arithmetic of that pressure. Amazon Leo spreads its business across five rocket families in a procurement worth more than $10bn, an arrangement meant to avoid any single point of failure. The trouble is that its two most eagerly awaited vehicles are grounded. The 38 Vulcan Centaur flights and 12 New Glenn flights it booked in 2022 were supposed to do the heavy lifting; both are out of service. That leaves Arianespace, Atlas V and Falcon 9 as the working rides, which is why the European order keeps growing. Simultaneously, Amazon paid $11.57bn in April for Globalstar, an older satellite operator, chiefly for its radio spectrum — insurance for direct-to-device services that would let phones reach satellites without special hardware, the same pitch Apple has made with Globalstar's blessing.

None of this asks much of Amazon the company. It booked roughly $201bn of revenue in a single recent quarter, and per-share earnings of $5.75 came in more than three times the analyst consensus; Leo's $10bn launch budget and $11.6bn acquisition are rounding errors against a business of that scale. For the stock, Amazon Leo is a large option on the fastest-growing corner of telecommunications, not a promise to earnings. What the launch orders and the waiver make clear is how the option is being exercised: capital standing in for decades of engineering time, and a regulator standing in for a level playing field.

The risk is not that Amazon is wrong that satellite broadband is a good business. It is that the good business is being built out by a competitor willing and able to spend even more, faster, and that the advantages being purchased — rockets, spectrum, regulatory patience — are the ones that erode. Globalstar's spectrum will age; launch contracts buy capacity, not subscribers. The six extra Arianespace flights are a sensible way to spend money, and they confirm a strategy that treats money as a substitute for time. Money is a good substitute. It is not a perfect one, and the gap to Starlink is now measured in thousands of satellites, not contracts.