If you want broadband internet beamed down from low-Earth orbit today, you have essentially one place to get it: Elon Musk. SpaceX’s Starlink has dominated the market for half a decade, and the only other provider, OneWeb, offers limited services. This has been a fine arrangement if you are Elon Musk, and a slightly awkward one if you are, say, an airline or a shipping company or a government that would prefer not to route all of its connectivity through one man’s mood board.
Soon there will be a second option, and it is Amazon, a company with some experience being the second option and then, eventually, not the second option. As Ars Technica’s Eric Berger reports, Amazon recently manufactured its 1,000th Leo broadband satellite at its factory in Kirkland, Washington, and is weeks away from offering commercial service for the first time. The debut will also mark Amazon’s first launch on the Vulcan rocket — and a second Vulcan, standing right next to the first, is already being prepped to carry more Amazon satellites this year.
Running the effort is Rajeev Badyal, who has the single most interesting résumé in the industry: he led Starlink in its early years, and Musk fired him eight years ago for moving too cautiously. There is a version of this story where that firing was obviously correct — SpaceX shipped fast and won — and a version where the guy Musk fired for being too slow now runs the second-largest satellite manufacturing operation on Earth, behind only SpaceX. Both versions are true. That is competition.
The interesting question, I think, is how you mass-produce a satellite, because for most of space history the answer was: you don’t. Satellites were bespoke objects, hand-built and tested under NASA standards written decades ago, procedures that take days and weeks per unit. That works if you are building one satellite. Amazon aims to build thousands of satellites, so it can’t work for Amazon.
Badyal’s account of solving this is worth quoting, because it describes the whole game: “It took us six months to make our first 27 satellites. Now we can easily make 27 a week.” Paul Palcisco, Amazon’s director of production, told Ars the company took an empty 170,000-square-foot building in December 2022 and turned it into a working satellite factory in 15 months — building the factory at the same time as the satellites it would produce, which is not the usual order of operations. “As soon as we opened the doors in that factory, we needed to change it,” he said.
A satellite now moves through the line in several weeks, and design changes can be incorporated in one to two. The production schedule is paced backward from the launch calendar, with a buffer of finished satellites in front of it, so that — in Palcisco’s words — there is never a rocket waiting on satellites. Think of it as the opposite of most manufacturing, where the factory waits for demand. Here the satellites wait for the rockets, which tells you something about which half of the space business is scarce.
On testing, Amazon did the thing every scaled manufacturer eventually does: it borrowed from consumer electronics. Palcisco said the company blended the old rigorous space standards with high-volume test procedures, getting tests that used to take days or weeks down to hours while maintaining or even increasing the strength of the screening. Vertically, Amazon keeps its highest-value intellectual property in-house — the propulsion system is designed and built entirely internally — and buys the commodity components from suppliers with decades of experience. This is roughly the Apple playbook, except the phone has a rocket engine.
The other thing striking about Badyal’s comments is the sheer shape of the bet. Amazon set the constellation’s architecture — capacity, capability, everything — within the first three months, then spent years and enormous capital building hardware and a factory before getting any real data back from space. “The biggest concern you always have is, hey, is this architecture going to work?” Badyal said. You commit to the design in 2018 and find out whether you were right approximately never-until-launch. Badyal described the required mindset as faith in the team, which is a polite way of saying you wire the money first and ask questions later.
And there is real money waiting on the other end. Satellite broadband has proven useful for everything from e-commerce to warfighting, and Amazon’s Chris Weber, vice president of business, has been signing contracts. Delta Air Lines agreed to use Amazon instead of Starlink on its planes — a decision that has reportedly upset Musk, to put it mildly, which is a phrase doing a lot of work in that sentence.
The backdrop to all of this is that Jeff Bezos owns a rocket company, Blue Origin, whose New Glenn rocket is supposed to help launch these satellites, and yet Amazon is debuting on a Vulcan, built by the Boeing-Lockheed joint venture. Even in a story about one company challenging Musk’s dominance, the actual plumbing of the industry remains gloriously entangled. “If this was easy, everybody else would be doing it,” as Badyal put it. Everybody else is not doing it. For about five more weeks, exactly one person is.