Amazon, which is encroaching on specialized conferences, has clashed with Elon Musk's Starlink.

Writing by: Little Gold Tooth

Amazon has struck again.

This time, it acquired the satellite communications company Globalstar in a deal worth at the hundred-billion-dollar scale, continuing the latter’s collaboration with Apple, and directly moving into the communication capability of “phone direct-to-satellite.”

This makes it hard not to think back to the competition that took place years ago between Jeff Bezos and Elon Musk over space—though we thought the story was basically over, it turns out it is nowhere near finished.

Amazon didn’t go head-on to compete on the number of satellites; instead, it chose to start from D2D (direct-to-device), a more fundamental capability, trying to open a gap from the communication approach itself.

If we put this acquisition in a broader context, it also fits Amazon’s usual path: it doesn’t rush to confront head-on, but rather enters from the periphery and gradually encircles. The recent series of moves in AI also repeats the same logic.

Amazon’s acquisition of Globalstar

Let’s first look at this acquisition.

Amazon’s offer could be described as full of sincerity—the acquisition price is $11.6 billion. This figure is 40 times Globalstar’s revenue last year. Globalstar shareholders can choose either to receive $90 in cash per share, or to receive 0.321 shares of Amazon stock. The cash payment cap is 40% of Globalstar shares.

Amazon’s move to acquire Globalstar may look, on the surface, like it’s buying a satellite company. But if you only understand it as “buying dozens of satellites,” you would be underestimating the true value of this deal.

The most valuable part of Globalstar is the spectrum resource in its hands—hard to replicate—and the direct-to-device capabilities already built around that spectrum.

Globalstar doesn’t have that many satellites. It’s on the scale of dozens: continuously updating and replacing them. But it has a spectrum band known as S-band/L-band, which is especially critical for “phone direct-to-satellite.” It is scarce and has high regulatory thresholds.

Spectrum is allocated globally on a licensing basis; it isn’t something you can just buy because you have money. According to U.S. federal filings, Globalstar is the only licensed MSS (Mobile Satellite Service) spectrum holder for the 2483.5–2495 MHz range; earlier FCC filings also described it as the operator of a single MSS system running on this band.

Globalstar’s satellite network mainly serves stable, low-data-volume communications that can directly connect mobile devices—Direct-to-Device (D2D), or direct-to-device.

This technology no longer requires devices to connect to terrestrial cellular base stations, making it crucial for supporting emergency services and providing connectivity in areas where cellular network coverage is insufficient.

Globalstar’s star customer is Apple.

For example, the satellite-based safety features such as “Emergency SOS” and “Find My” on current Apple iPhones and Apple Watch are backed by Globalstar’s D2D support.

Apple isn’t only a customer of Globalstar.

In 2024, Apple invested about $1.5 billion in Globalstar to support the expansion of its iPhone communications services. That deal also gave Apple about 20% of Globalstar’s shares.

At the end of last year, Globalstar also stated that, with Apple’s support, the new network it is developing will expand its satellite fleet from roughly 24 satellites currently to 54 satellites, including a small number of backup satellites.

While Amazon is acquiring Globalstar, its cooperation with Apple remains.

In other words, through this deal, Amazon not only gains valuable spectrum assets and opens up D2D phone direct-connection scenarios, but also starts out with a major customer like Apple.

Is the space race between Bezos and Musk over?

Amazon’s goal is very clear: to strengthen its satellite business, which is still in its early stage, and catch up with Elon Musk’s Starlink. This move comes at a time when SpaceX is pushing forward with its IPO plans.

Starlink has more than 10,000 satellites and currently serves over 9 million users worldwide. SpaceX has been deploying Starlink satellites at a very fast pace, launching dozens at a time, building the world’s largest satellite constellation. Starlink accounts for about 50% to 80% of SpaceX’s revenue.

Amazon is far behind—if we’re talking about satellites, it currently has only a bit more than 200.

Amazon was already pushing forward with its own low Earth orbit satellite project. The original name people were more familiar with was Project Kuiper, and now it is talking more explicitly as Amazon Leo. The license it received from the U.S. Federal Communications Commission (FCC) is to deploy 3,236 low Earth orbit satellites. Under the original regulatory milestones, Amazon would need to deploy half—1,618 satellites—by July 2026, and complete the rest by July 2029.

But the real-world progress isn’t that fast. In a recent Reuters report, it said Amazon has currently launched only about 243 satellites, which is far from the halfway deployment target by mid-2026; a report from The Verge a few days ago used nearly the same figure, writing that it is 241 satellites. These two numbers differ slightly, but they point to the same fact: it is very far from the FCC’s originally planned mid-term threshold.

So Amazon has already gone to the FCC to apply for an extension. Public materials show that it wants to push the original mid-term deployment deadline of July 30, 2026 back to July 30, 2028.

In other words, Amazon has essentially acknowledged that with its current pace, it can’t catch up to the original plan.

In terms of commercialization, Amazon’s new timeline is: start commercial services for Leo in mid-2026. Andy Jassy has said this publicly recently, and some media outlets also mentioned that Amazon has already conducted enterprise previews and signed up a batch of customers or partners in advance, including aviation, telecom operators, and government-related clients.

What mainly drags things behind is still launch capacity and the supply-chain timeline. Amazon’s problem isn’t just making satellites; it also includes tight rocket launch resources, manufacturing delays, and the fact that it doesn’t have a mature high-frequency in-house launch system like SpaceX. Even Blue Origin can’t meet launch demands, so to make up for the schedule it has to rely on third-party rockets—up to using SpaceX’s Falcon 9.

Although Amazon founder Jeff Bezos has already retired from Amazon for some time, his “space rivalry” with Musk has never ended. From the disputes between Blue Origin and SpaceX years ago to today’s low Earth orbit satellite standoff between Amazon and SpaceX, the competition is set to continue.

Amazon’s choice to enter from D2D is also rather subtle. After all, Musk’s Starlink mainly runs high-speed broadband services. Although it has already worked with telecom operators such as T-Mobile to develop D2D services, it has not “taken over everything.” Amazon is already behind Starlink in the satellite broadband area. Acquiring Globalstar will help it catch up in D2D spectrum planning and achieve a leap in D2D deployment.

Amazon’s ambitions

If the acquisition of Globalstar is more of a communications deal, then if you zoom out further, after 2026 Amazon is also taking frequent action in another highly watched battlefield—the AI space.

Interestingly, just as Amazon’s low Earth orbit satellite business seeks to break through through D2D, Amazon’s AI endeavor has also carved out its own distinctive approach.

A few days ago, an internal memo from OpenAI was leaked. One very eye-catching detail is that Amazon is mentioned repeatedly.

In the memo, Denise Dresser, OpenAI’s newly appointed Chief Revenue Officer, emphasized that while the partnership with Microsoft naturally laid the foundation for OpenAI, she also directly pointed out that such long-term binding limits OpenAI’s ability to serve enterprise customers “where they are,” and for many companies, that “where they are” is Amazon Bedrock.

She also mentioned that since the parties announced their collaboration at the end of February, the related demand has been “stunning.” In other words, in OpenAI’s own wording, Amazon has become an increasingly important enterprise distribution channel—far more than just a new “sugar daddy” type of investor.

Less than two months ago, Amazon and OpenAI had just announced a multi-year strategic partnership: Amazon would invest $50 billion in OpenAI; AWS would be the exclusive third-party cloud distribution provider for OpenAI’s enterprise platform Frontier; some of OpenAI’s capabilities would be offered to enterprise customers through Amazon Bedrock. The two sides will also jointly develop a “stateful runtime environment,” so that during model operation it continues to retain context, task state, and historical information—rather than starting from scratch every time a call is made.

And Amazon is not placing all its bets on just one line.

It invested in Anthropic earlier and more deeply.

In 2023, Amazon first announced an investment of up to $4 billion in Anthropic. By November 2024, the collaboration expanded further: Anthropic’s official statement confirmed that Amazon’s total investment would increase to $8 billion, while AWS would become its primary cloud and training partner, and Amazon would continue to hold the status of a minority shareholder. That means that with Anthropic, Amazon is betting on equity, cloud, and a self-developed chip ecosystem. With OpenAI, it’s betting on larger-scale capital, enterprise distribution, and platform lock-in. Although these two lines seem different on the surface, the underlying logic is consistent: make AWS into an infrastructure platform that top model companies can’t do without.

If this memo accidentally reminded the outside world of anything, it might be that Amazon’s role in the AI race is no longer that “old giant with a low profile” that many people once imagined.

Over the past period of time, the outside world has more easily focused attention on model companies like OpenAI, Anthropic, Google, and Meta, and Amazon has often seemed like it is standing in the back row.

But when you look back now, it has in fact quietly secured several very critical positions: on one hand, it locks in the enterprise customer entry through AWS; on the other, it carries model capabilities upward through Bedrock; it extends downward to the chip and computing-power layer through Trainium; and at the same time, it deeply binds itself through capital with several of the most important foundational model companies.

It may not always be the company that generates the loudest buzz, nor always the one at the center of public attention.

While the outside world is still watching whose model is stronger and whose products are hotter, Amazon has already been quietly stitching satellites, cloud, chips, enterprise distribution, and model collaborations into a bigger map.

When this map truly takes shape, people may realize that the rivalry between Bezos and Musk may never have been only about space.