Amazon Paid $11.6 Billion to Prove a Point: Broadband Satellite and Direct-to-Cell Are Not the Same Network

In April 2026, Amazon agreed to buy Globalstar for about $11.6 billion. Ninety dollars a share, cash and stock, closing expected in 2027.

Stop and think about that number for a moment. Amazon is already building Amazon Leo, formerly Project Kuiper, a low-Earth-orbit constellation authorized for more than 3,200 broadband satellites, with thousands more filed for. It has a satellite factory in Kirkland, a launch manifest across five rocket families, and more than ten billion dollars already committed. So why would a company with that much orbital broadband capacity under construction turn around and pay eleven billion for Globalstar, an operator flying just 24 aging satellites?

Because broadband satellite and direct-to-cell are two different networks. Different spectrum, different antennas, different satellites, different physics, different regulators. Amazon’s own broadband fleet cannot do direct-to-cell, and the fastest way to fix that was to buy a company that already holds the right spectrum. If you understand why that acquisition made sense, you understand the whole distinction that most of the market still gets wrong.

What a broadband satellite actually does

A broadband satellite talks to a terminal, not to a phone. When you order Starlink or sign up for Amazon Leo, a dish arrives. Amazon calls its three terminals the Leo Nano, Pro, and Ultra. That dish is a phased-array antenna with real aperture, real gain, and mains power behind it. It points at the constellation, tracks satellites as they cross the sky, and holds a high-capacity link.

The frequencies tell the same story. Broadband satellites operate up in Ku-, Ka-, and V-band, tens of gigahertz, where there is enough bandwidth to deliver hundreds of megabits per second. Those frequencies only work because the terminal has a large, well-designed antenna to close the link. Take the dish away and the link collapses. That is the entire point of a broadband architecture: put the complexity and the power in a fixed terminal on the ground so the satellite can serve homes, offices, ships, and aircraft with serious throughput. Amazon Leo sits at roughly 590 kilometers and is being sold first to enterprises, then to residential broadband users. It is an internet-to-the-building product.

What direct-to-cell actually does

Direct-to-cell, D2C, or direct-to-device, D2D, throws that architecture out. There is no dish. The satellite has to close a link to an ordinary smartphone in someone’s pocket, a device with a fingernail-sized antenna and around a fifth of a watt of transmit power. The satellite becomes, in effect, a cell tower in orbit. Your unmodified phone sees it as another cell in the network and camps on it the way it would camp on a rooftop tower.

That is a brutally hard link to close, and it forces three consequences. First, you need low frequencies that a phone’s antenna and radio can actually use, in the mobile and mobile-satellite bands, not Ka-band. Second, the satellite carries a very large antenna to make up for everything the phone lacks. Third, first-generation service is thin by design: text and basic data before voice, because the link budget barely allows more.

This is not theory. Starlink’s direct-to-cell layer is roughly 650 satellites, and by coverage area it is already the largest 4G network on the planet, delivered through T-Mobile’s T-Satellite service as a low-cost add-on. It reached around ten million users inside its first stretch of commercial operation. It works on about sixty phone models with the right LTE release, with no app and no dish, purely because those satellites were built as orbital base stations and use mobile spectrum. Different mission, different machine.

The four things that separate them

Strip away the branding and the two systems diverge on four hard engineering axes.

• Spectrum. Broadband lives in Ku/Ka/V-band; D2C lives in low mobile and mobile-satellite service bands so that unmodified phones can transmit and receive.
• Ground segment. Broadband depends on a powered phased-array terminal; D2C has to work with a handset that has almost no antenna and almost no power.
• Link budget and waveform. Broadband optimizes for throughput to a high-gain dish; D2C optimizes for reaching a weak device at all, which caps early service at messaging and modest data.
• Regulation. Broadband uses fixed-satellite-service and gateway spectrum; D2C needs mobile-satellite-service rights or shared access to a mobile operator’s own bands, which is a completely separate regulatory and commercial fight.

A satellite designed for one of these jobs is the wrong tool for the other. You do not upgrade a broadband bird into a direct-to-cell bird with a software patch.

Why the Globalstar deal is the proof

Now the acquisition makes sense. Amazon Leo, as built, is a broadband constellation. To enter direct-to-device, Amazon needed something its own fleet does not have: globally harmonized mobile-satellite spectrum that phones can use without begging every mobile operator in every country for access. Globalstar holds exactly that, L- and S-band including Band n53, already licensed across borders, and it already powers Apple’s satellite features. Buying Globalstar let Amazon skip years of spectrum coordination and regulatory delay in one move, and it came bundled with an Apple agreement for iPhone and Apple Watch satellite services.

Then look at what Amazon filed next. In July 2026 it applied to the FCC for a separate constellation of more than 5,100 satellites purpose-built for direct-to-device, targeting deployment around 2028. Not a tweak to the broadband fleet. A second, distinct network, on distinct spectrum, for a distinct service.

That is the argument settled by a balance sheet. A trillion-dollar company decided that the cleanest path from broadband satellite to direct-to-cell ran through an eleven-billion-dollar acquisition and a whole new constellation. If the two were the same network, none of that would have been necessary.

Why this matters if you work in mobile

For operators, regulators, and network engineers, the broadband-versus-D2C line is not a trivia question. It decides which spectrum you defend at WRC, which partnerships actually extend your coverage, how you model interference into your own bands, and what you can honestly promise a subscriber standing in a dead zone. Confusing the two leads to bad spectrum strategy and worse roadmap decisions. The companies spending billions have already picked a side of the line. The professionals working around them need to see it just as clearly.