The emergency message an iPhone can send from a trailhead with no bars travels through a satellite owned by Globalstar. This week the company that operates that satellite moved a step closer to belonging to Amazon.
The Federal Communications Commission on Oct. 28 approved Amazon’s Amazon Leo unit to acquire the mobile satellite operator Globalstar for $11.5 billion. The agency concluded that the two companies’ assets complement each other and that the combination would strengthen Leo’s ability to enter the market for direct-to-device satellite service, the corner of the industry where a handset talks to a spacecraft with no dish and no terminal in between.
Globalstar runs the emergency satellite feature Apple offers on the iPhone and Apple Watch. Apple held a 20% stake in Globalstar before the transaction. Once the deal closes, Leo will supply the satellite service behind supported models, which puts Amazon inside a handset function that Apple sells as a safety feature rather than a convenience.
The numbers explain why Amazon bought the company instead of renting capacity from it. Leo has about 400 satellites in orbit and plans to deploy more than 3,000 in low earth orbit. SpaceX’s Starlink operates roughly 11,000. A constellation of 400 cannot cover the planet continuously, and continuous coverage is the whole requirement: the satellite has to be overhead at the instant the subscriber needs it, in the place where no tower reaches.
What Amazon is buying, then, is not a fleet. It is a set of licenses, ground infrastructure and handset integrations that took years to negotiate, attached to a customer that ships the feature to a very large installed base. That is quicker to acquire than to build, and it is the kind of asset that does not come up for sale often.
The regulatory logic was straightforward enough that the commission did not split the difference. A mobile satellite operator with a thin fleet and a hyperscaler with capital and launch capacity are complements in the literal sense: one has rights to spectrum and the operating experience that goes with it, the other has the money to fill an orbital shell and the cloud business to sell the resulting data path into.
The consumer-facing question is what happens to the Apple arrangement. Globalstar’s service is sold as a subscription bundled with hardware, and the terms behind it were negotiated with a supplier in which Apple held equity. Amazon will inherit the contract, not the relationship. Whether the feature stays a premium tier, expands into messaging or voice, or becomes an argument for buying an Amazon device instead is the part of this deal that will be decided in Cupertino as much as in Seattle.
Inside Amazon, the satellite purchase is arriving in the same week as a much smaller piece of news with a similar shape. The company has sharply cut back its internal artificial-intelligence lab, placing most of its Nova models into what amounts to maintenance mode, according to people familiar with the decision, with staff and computing capacity redirected to a single frontier-model effort led by Pieter Abbeel. A week earlier the company closed an 80-person artificial general intelligence research office in San Francisco.
Read together, the two moves describe a company narrowing its ambitions to the places where it can own the physical layer. Amazon does not need to win a model benchmark to sell satellite connectivity or cloud capacity. It does need spectrum, launch slots, ground stations and a service that works when a customer is in trouble, and those are bought with capital rather than with research taste.
The contrast with the rest of the industry is instructive. Companies that publish frontier models measure progress in evaluation scores and release cadence. Companies that own infrastructure measure it in megawatts, orbital planes and regulatory approvals, all of which move on a timescale of quarters and years. Amazon, which has spent a decade building the second business while talking mostly about the first, appears to have decided which one it is.
For Globalstar the approval ends a period of uncertainty that began when the transaction was announced. A mobile satellite operator of its size earns a modest return on capacity; the contract to carry Apple’s emergency service is what gave it strategic value well above its revenue. Being the supplier to a safety feature is a stable business. Being a unit of the company that owns the cloud behind the feature is a different one.
For the direct-to-device market the approval sets a reference price. Eleven and a half billion dollars for a small constellation, a set of licenses and one handset integration establishes what strategic satellite access costs when a hyperscaler decides it cannot wait. Rivals weighing their own constellations now have a number to argue against.
What remains is execution, and it is not trivial. Integrating a satellite operator into a cloud company means merging two engineering cultures that share almost no vocabulary, and doing it while the constellation is being expanded on a launch schedule that Amazon does not fully control. The licenses transfer, the satellites keep flying, and the phones keep asking for a connection.
The first test will be quiet: a subscriber in a place with no signal sends a message, and it arrives. Amazon will have paid $11.5 billion for the right to be the company standing behind that message. Everything else about the deal is a forecast.
