The Phone Without a Tower: Satellites vs. the Carriers

Sponsored: links marked “Trusted partner” in this article go to a partner’s offer, and we are paid when readers respond to it.

A phone has always been only as good as the nearest tower. For forty years that single fact decided who sold you service, what you paid, and which companies got rich when the handset changed. This month it stopped being a settled fact.

On October 6 the Federal Communications Commission adopted an order allowing SpaceX to build a constellation of up to 15,000 satellites designed to talk directly to ordinary phones, flying at a record-low 326 to 335 kilometers across nine orbital shells and, crucially, without the spectrum lease from a ground-based carrier that the rules used to require.¹ Two days later SpaceX agreed to buy a nationwide block of 800 megahertz spectrum, the low-band frequencies that reach indoors, from Grain Management.² On the next trading day Verizon closed down 10.14%, and AT&T and T-Mobile were down 11.28% and 12.81% in afternoon trading.³

Apple’s answer to the question of what a phone will be arrived on October 23: the iPhone Duo, its first foldable, starting at $1,999.⁴ One company spent October changing the shape of the phone. The other spent it changing what the phone connects to. This report is about the second bet, what it can and cannot do today, and where the money has gone every time the phone business changed its mind about what a phone is.

How the Tower Became the Gatekeeper, and Who Got Paid Last Time

Start with the last platform shift, because it is the template. In 2007, the year the first iPhone went on sale, the world bought about 1.15 billion mobile phones and Nokia sold 435 million of them, a 37.8% share that reached 40% in the fourth quarter.⁵ Nokia was the obvious name. It agreed to sell its handset business to Microsoft in September 2013, and the sale closed in April 2014 for about $7.2 billion.⁶

The money went somewhere else. Qualcomm, which made the modem chips inside most of the new smartphones, reported revenue of $8.87 billion and net income of $3.30 billion for its fiscal 2007.⁷ By fiscal 2014 revenue was $26.49 billion and net income was $7.97 billion.⁸ The company whose name was on the phone shrank, and the company whose chip was inside it tripled.

The company whose name was on the phone shrank, and the company whose chip was inside it tripled.

That is the pattern worth carrying through this report. A new kind of phone creates a new bill of materials, and the suppliers of the parts the old phone did not need, the radios, the antennas, the power systems, the modems, grow with every unit sold regardless of whose logo is on the box. The brand fight gets the headlines. The supply chain gets the compounding.

Satellites entered that supply chain quietly. Apple put emergency satellite texting into the iPhone in 2022 through Globalstar, and in November 2024 it committed about $1.5 billion to the company, including $1.1 billion in cash and a 20% equity stake, to pay for new satellites and ground stations. Globalstar’s shares rose 31.4% that day.⁹ T-Mobile turned Starlink’s first direct-to-cell satellites into a consumer product called T-Satellite, which went commercial in July 2025 at $10 a month, open to customers of rival carriers as well.¹⁰ Both were add-ons to a phone that still lived on towers. The October decisions are different in kind.

Six Weeks That Redrew the Map

The FCC order is the foundation. It conditionally grants the application SpaceX filed in September 2025 for 15,000 next-generation direct-to-cell satellites, approves the EchoStar frequency bands SpaceX is acquiring for cellular use inside and outside the United States, lets the satellites keep using T-Mobile’s PCS G-block for now, and defers a few requests. It builds on the commercial direct-to-cell license the FCC first issued in November 2024.¹ The order sets deadlines: half the constellation operating by October 7, 2032, and the full system by October 7, 2035.¹¹

The spectrum behind it was assembled over the previous year. In September 2025 SpaceX agreed to pay about $17 billion for EchoStar’s AWS-4 and H-block licenses, up to $8.5 billion in cash and $8.5 billion in SpaceX stock, and to fund about $2 billion of interest on EchoStar debt through November 2027. SpaceX President Gwynne Shotwell said the deal “will advance our mission to end mobile dead zones around the world.”¹² EchoStar sold its other airwaves to AT&T, about 30 megahertz of 3.45 gigahertz and 20 megahertz of 600 megahertz across more than 400 markets, for about $23 billion in cash, in a deal announced August 26, 2025.¹³ A company that never built a competitive network of its own sold its licenses for roughly $40 billion in two transactions. That alone tells you what the scarce input is.

Then came October 8. SpaceX agreed to buy up to 14 megahertz of paired 800 megahertz spectrum nationwide from Grain Management, subject to FCC approval. SpaceX said the purchase “addresses one of the key remaining technical gaps” for Starlink Mobile to become a major U.S. carrier, and described the low band as a coverage layer that can reach indoors.² SpaceX’s stated plan is a hybrid network, satellites plus terrestrial infrastructure, with the 800 megahertz band handling indoor, rural and dead-zone coverage and mid-band spectrum near 2 gigahertz carrying capacity. In the company’s words, “Starlink Mobile can now become the first network operator to deploy both satellite and terrestrial spectrum.”¹⁴

✓ Trusted partner A former Wall Street tech trader who says he saw the smartphone boom coming early has now named the one stock he believes is best placed if a satellite-connected AI phone arrives →

The market’s verdict came on October 9. Verizon closed down 10.14% at $41.65, with about 71 cents of the drop explained by its ex-dividend date, and AT&T and T-Mobile were down 11.28% and 12.81% in mid-afternoon trading.³

Behind the deals is a company that now reports like a utility. SpaceX’s third-quarter results on October 20 showed revenue of $14.6 billion, up 161% from a year earlier, and net income of $4.4 billion, its first profitable quarter as a public company. Starlink ended the quarter with 13.6 million consumer subscribers, up from 12.0 million in June, paying an average of $69 a month, and connectivity revenue was $5.2 billion. SpaceX counted 11,100 satellites in orbit as of September 30, a figure that includes satellites slated to de-orbit, with coverage in 174 countries, and flew 35 launches in the quarter, including a Starship flight that deployed 26 of the larger V3 Starlink satellites.¹⁵ Starlink had 10 million subscribers at the end of the first quarter and passed 12 million in the second, a quarter in which it also raised prices and ended promotions.¹⁶ This is the balance sheet that will pay for 15,000 satellites.

What Direct-to-Cell Can and Cannot Do Today

Separate the network that exists from the one just approved. Today’s service is version one: about 650 direct-to-cell satellites, available in the United States through T-Mobile, with data speeds estimated at about 4 megabits per second.¹ T-Satellite costs $10 a month as an add-on, is included on T-Mobile’s top plan, and can be bought by customers of other carriers with an unlocked phone. It handles text messages, location sharing and a short list of apps such as WhatsApp and Google Maps. Native voice calls and general data are not supported yet.¹⁷

The carriers’ case rests on that gap. On September 9, T-Mobile’s outgoing chief financial officer, Peter Osvaldik, told Citi’s 2026 Global TMT Conference that Starlink is “not a competitive threat from a wireless perspective at all,” adding: “The physics just don’t allow direct-to-cell to be a substitute product.” T-Mobile said satellite accounts for 0.0002% of its network usage.¹⁸ Verizon’s chief executive, Dan Schulman, said the same day: “I do not see it as a viable competitor in the medium and even longer term.” At Goldman Sachs’ Communacopia conference that week he said satellite “just can’t compete without a terrestrial network in urban and suburban,” and that Verizon sees “absolutely no advantage in extending an MVNO to any LEO provider.”³ He also said: “We’ve seen zero discernible impact on our broadband growth from any LEO satellite provider, including Starlink.”¹⁸

They are right about the physics of version one. A satellite hundreds of kilometers up, sharing a sliver of borrowed spectrum, cannot match a tower a mile away. That is exactly why the October deals matter: each one is aimed at a specific item on the carriers’ list. The new satellites will fly lower, at 326 to 335 kilometers. The EchoStar bands give SpaceX exclusive spectrum instead of a slice of T-Mobile’s. SpaceX’s stated target for the new system is 5G-class service of 150 megabits per second per user, using the satellites together with small ground femtocells, a figure that comes from the company itself and that T-Mobile’s finance chief has dismissed.¹ And the 800 megahertz purchase answers the indoor problem directly, with a terrestrial layer, which is the one thing Schulman said satellite could not do without.²

Timing is the honest caveat. Testing on the EchoStar bands cannot start until the second step of that deal closes, scheduled for November 30, 2027, which is also when SpaceX projects its first version-two satellites will launch.¹ The Grain purchase needs FCC approval.² Nothing resembling a satellite phone plan that replaces a carrier contract is available to buy today, and the FCC’s own milestones run to 2035.¹¹

The Incumbent’s Bet: A Folding Screen

Apple’s October bet was a different reading of the same question. The iPhone Duo starts at $1,999 for the 256-gigabyte model, against $1,199 for the iPhone 18 Pro. Pre-orders opened October 16 and delivery estimates slipped into early 2027 almost at once, with new orders showing January 29, 2027.¹⁹ It went on sale October 23 with limited store inventory. Analyst Dan Ives expects about 10 million units over the next few quarters and up to 25 million in 2027, with demand “running well ahead of supply out of the gates.”⁴

Reviewers were cooler. The most common complaint was weight: 254 grams, five more than the iPhone 18 Pro Max. The Duo swaps Face ID for Touch ID, has no telephoto camera, and The Verge found it runs hot, calling it “one of Apple’s messiest first-generation products in a long time.” The 7.6-inch inner display with its matte coating drew praise; the 5.4-inch cover screen was called cramped. Engadget measured 28 hours and 48 minutes of local video playback.²⁰

None of that is a knock on Apple’s engineering, and the demand is real. The point is what the Duo is not. It is a new shape for a device that still depends on the same towers, the same carrier contracts and the same spectrum map as the phone it replaces. It changes the hardware. It does not change the network. The two Octobers, Apple’s and SpaceX’s, are bets on different questions, and only one of them is a bet on what a phone will be connected to in 2035.

Does a SpaceX Phone Exist? What Has Actually Been Said

No. There is no SpaceX phone for sale, and the company has not announced one. What exists is a trail of statements worth reading in order, because the order matters.

In October 2024, at a town hall in Pennsylvania, Musk said: “Man, I sure hope we don’t have to make a phone. That’s a lot of work. The idea of making a phone makes me want to die. But if we have to make a phone, we will. But we will aspire not to make a phone.”²¹^, 22 On January 30, 2026, replying on X to a post about Meta and SpaceX building phones, he wrote: “It would be a very different device than current phones.”²²

On July 1, 2026, the Wall Street Journal reported that SpaceX had shown investors an early prototype of a handheld AI device ahead of its June listing: slimmer than an iPhone, running a proprietary operating system built around xAI technology, with Qualcomm Snapdragon silicon inside, and with SpaceX telling investors the project was early and might never ship. Musk posted on X the same day: “Utterly false.”²³

✓ Trusted partner See the research behind this story →

Those are the facts: a conditional in 2024, a description in January of what such a device would be if it existed, and a flat denial in July. What changed in October is the context around them. After the spectrum deal, Deepwater’s Gene Munster wrote on X on October 9: “Elon says he doesn’t want to make a phone. I still think $SPCX eventually will.”²⁴ That is an opinion, not a product announcement.

SpaceX is also not alone in the race to connect ordinary phones from orbit. AST SpaceMobile has its BlueBird 12 and 13 satellites in orbit, had 48 satellites in production as of August 26, and could have another 15 satellites in orbit by early 2027 by observers’ estimates; a beta service is possible by spring 2027 on the same estimates, and its preferred rocket, Blue Origin’s New Glenn, exploded at Cape Canaveral on May 28, 2026.²⁵ Apple’s own route runs through Globalstar, which keeps about 85% of its network capacity for Apple and has ordered as many as 26 new satellites.⁹ The direction is not in dispute. The question is who builds the parts.

Where the Money Goes in a Platform Shift

Follow the revenue. Starlink’s 13.6 million subscribers at $69 a month are the base; the 15,000-satellite order is the capacity to sell those subscribers, and tens of millions of phone users beyond them, a second service.¹⁵ A TD Cowen survey found 34% of respondents would consider switching to a mobile service from SpaceX.¹¹ Whether or not that number holds, it is the kind of number that makes a carrier’s stock fall a tenth in a day.

The satellite operator captures the subscription. But the physical build runs through several categories of supplier, and history says that is where the compounding happens. Consider what a satellite-connected phone needs that a tower-connected one does not.

First, radio-frequency front ends and antennas tuned for new bands. Each slice of spectrum SpaceX has bought or been assigned, the AWS-4 and H-block frequencies, the 800 megahertz low band, the PCS G-block, has to be supported inside the handset by filters, amplifiers and antenna designs that were not in last year’s phone.¹^, 2 Second, modems. Every phone that talks to a satellite needs baseband silicon that understands non-terrestrial links; the device described in the Journal’s report ran on Snapdragon silicon, a reminder that even a company as vertically integrated as SpaceX buys its chips.²³ Third, the satellite side: solar arrays, phased-array panels, propulsion and radiation-hardened electronics for thousands of spacecraft, built on a schedule the FCC now enforces.¹¹ Fourth, launch and ground: the rockets that lift them, and the gateways and femtocells that tie orbit to the terrestrial network.¹ Fifth, spectrum itself: EchoStar turned licenses it never fully used into about $40 billion of proceeds from two buyers.¹²^, 13

Notice what is missing from that list: the brand. In the last shift, the brand with the biggest share in 2007 was sold to Microsoft in a deal that closed seven years later for about $7.2 billion, while its chip supplier tripled its revenue.⁵^, 6^, 7^, 8 A supplier with a modest market value and a part that a new category of device cannot be built without holds the asymmetric position in a platform shift, and it is usually priced as a component business right up until the moment it is priced as a platform business.

The Risks, What to Watch, and the Supplier’s Decade

The risks are real and specific. The physics argument is not wrong today; version one is a texting service at about 4 megabits.¹ The FCC grant is partial and conditional, with some spectrum requests deferred, and the exclusive bands cannot even be tested until the EchoStar deal’s second step closes on November 30, 2027.¹ The Grain purchase needs FCC approval.² The incumbents have cash, towers and the wholesale lever: Verizon has already said it sees no advantage in extending an MVNO to any satellite provider.³ Rival constellations face launch risk, as AST’s dependence on a rocket that exploded in May shows.²⁵ And the phone at the center of the story does not exist, by the account of the one man who would know.²³ A supplier priced for a device that never ships is just an expensive component maker.

What to watch, in order. The carriers’ next quarterly reports, for any change in the satellite usage figures and the MVNO stance they gave in September.¹⁸ Apple’s next results, for the first hard read on Duo supply, which preorder wait times cannot give.¹⁹ The FCC’s decision on the Grain 800 megahertz transfer.² Starship launches carrying V3 satellites, the first of which deployed 26 in the third quarter, because 15,000 satellites is a launch-cadence problem.¹⁵ AST SpaceMobile’s next BlueBird launches and the spring 2027 beta it is aiming for.²⁵ November 30, 2027, when the EchoStar step two is due to close and testing on the exclusive bands can begin.¹ And the FCC’s own clock: half the constellation by October 7, 2032, all of it by October 7, 2035.¹¹ Any handset announcement from SpaceX, a partner or a rival would pull the timeline forward. None has been made.

Step back from the personalities and the week of October 6 looks like a regulatory and financial fact pattern, not a rumor. A company with 13.6 million paying subscribers, a profitable quarter and 11,100 satellites in orbit was cleared to build a network that connects to ordinary phones, bought the spectrum to reach indoors, and watched the three incumbents lose a tenth of their value in an afternoon.¹⁵^, 3

The phone that would make full use of that network is not here. The carriers say the physics will keep it that way. They may be right for years. But the last time the phone changed, the fortunes were not made by buying the famous name after it was famous. They were made by working out, early, which parts the new device could not be built without, and owning the companies that made them.

That is the work a report like this can only begin. The categories are clear: radio front ends and antennas, modems, satellite components, launch and ground equipment, spectrum. Picking the single company inside those categories that a satellite-connected AI phone would depend on most is a different job, and at least one analyst believes he has done it.