SpaceX Picks Qualcomm To Put Starlink On Your Street Corner

On November 4, 2026, Qualcomm sent out a release from San Diego that takes Starlink somewhere it has never been: your street corner. SpaceX selected Qualcomm Dragonwing platforms for Starlink Mobile, and the job is a nationwide 5G network in the United States, built on the ground.

Sawyer Merritt flagged it on X. His summary of the announcement: under a multi-year agreement, SpaceX plans to deploy tens of thousands of cellular sites across the US powered by Qualcomm's Dragonwing infrastructure technology, with deployment beginning in 2027. Starlink posted alongside Qualcomm that Starlink Mobile is building the first fully integrated US nationwide satellite and terrestrial network that will keep people connected on their phones no matter where they are.

For years most people kept Starlink in a small box. A dish on a ranch. A dish on a camper van. If you lived in a suburb, SpaceX never came up when the phone bill arrived. My read is that box is gone. SpaceX already has over 11,000 satellites in orbit, and now it is ordering radios for poles and rooftops.

The bill you already pay

By the third quarter 2026 figures, the average Verizon postpaid account costs $168 a month. T-Mobile's averages $155.73. Over a year that comes to between about $1,870 and $2,016, and your call still drops on the highway and in the basement.

One caution. Those are account averages. A family of four on one plan counts as one account, so a single line usually costs less. The machine behind those bills is huge. Verizon has 94.5 million postpaid phones across 34.47 million accounts and $34.3 billion in total quarterly revenue. AT&T has 75.28 million postpaid phones bringing in $17.76 billion of quarterly wireless service revenue. T-Mobile has about 35 million postpaid accounts and $19.4 billion in quarterly service revenue.

The number I care about most is churn, the share of customers who leave each month. AT&T sits at 0.92%. Verizon at 0.94%. T-Mobile at 1.09%. About one in a hundred. People complain about their carrier and stay. That inertia is the hardest thing SpaceX has to beat.

On SpaceX's August 4, 2026 earnings webcast from Bastrop, Texas, President Gwynne Shotwell put the big three at roughly $600 billion a year and said SpaceX expects to win over quite a few of their customers with better service. Treat that as her framing. Verizon's and AT&T's quarterly totals annualize to about $266 billion combined, so check her number before you repeat it.

What does SpaceX charge today? Starlink consumer lines hit 13.6 million by September 30, 2026, at an average of $69 a month. That figure is for home internet. At the All-In Summit in September 2025, Elon Musk pitched one account covering the dish on your roof and the phone in your pocket, the way you have an AT&T or Verizon account now. SpaceX has published no Starlink Mobile price and has no signup page. Any savings on your bill stays hypothetical until a price exists.

Where coverage runs out

On August 14, 2026, the FCC released its broadband report, covering 343.68 million people with June 2025 maps. Outdoors, basic 4G LTE reaches 99.7% of people, which still leaves 1.15 million without it. Raise the bar to 5G at 35 megabits per second and outdoor coverage falls to 95.2%. That leaves out 16.5 million people.

Now get in a car. Per that report, fast in-vehicle 5G covers 85.5% of people nationwide and 52.7% in rural America. Run 85.5% against 343.68 million and nearly 50 million Americans lose fast 5G when they pull onto the road.

Satellites close the outdoor gap quickly. Picture a direct-to-cell satellite as a stadium floodlight about 330 km, roughly 205 miles, overhead. It sees every highway. Nobody outside stays dark.

The wall a satellite can't cross

Walk into a concrete apartment building and shut the door. A floodlight does not shine through a roof and drywall. After hundreds of kilometers, the signal is too weak to reach your couch. Strand Consult, citing Barclays, says 80% or more of mobile traffic happens indoors.

Then add crowds. Picture one block in Manhattan or downtown Austin with thousands of people streaming video. A satellite beam covers a huge footprint, and everyone inside it shares the same bandwidth. Ten thousand city phones fill it fast.

Verizon CEO Dan Schulman made this exact point on Verizon's October 26, 2026 call. He said satellites just can't compete with a terrestrial network in suburban and urban areas, where he put about 98% of Verizon's revenue, and called a ground network somewhere between 100 and 1,000 times more efficient. On satellite physics alone, I think Schulman is right.

So SpaceX is building what he says it needs. In Qualcomm's release, Starlink senior vice president Michael Nicolls named the problem: "We shouldn't live in a time where phone calls are still dropping whether you're at home or on the go."

Roof and pillars

Here is how I picture the plan. The satellite fleet is the roof. It covers the continent and erases dead zones. The tens of thousands of Qualcomm-powered ground sites are the pillars. They sit on street corners, push signal through your window and carry the heavy city load. Satellites alone cannot handle a dense block. Ground sites alone cannot cheaply reach every empty highway. SpaceX wants both inside one company.

Sixty-four beams on one panel

Qualcomm's release named the building blocks. At the top of the tower, the radio unit runs on Qualcomm's QRU100 chip, which powers 64T64R massive MIMO. T means transmit, R means receive: 64 transmit paths and 64 receive paths on one panel.

An old antenna shouts in every direction like a megaphone. A 64T64R panel works more like 64 aimed spotlights and 64 directional microphones on one corner. Stream a video three blocks away and those paths line up their waves into a tight beam aimed at your window, while steering dozens of other beams to other phones at once. Qualcomm says the design improves coverage, capacity and spectral efficiency. Each site reaches farther, handles more callers and squeezes more data from every slice of spectrum.

Follow the cable down the tower to the QDU100 chip running Layer 1 software. Layer 1 is the physical layer, the step that turns raw radio waves into digital bits. Qualcomm supplies that bottom layer, and PCMag's Rob Pegoraro noted it supports SpaceX's own Layer 2 and Layer 3 code on top. My read: Qualcomm sells the silicon and the lowest software layer, and SpaceX writes the brains above it. That matches how SpaceX builds everything else.

Qualcomm also says both chips maximize performance per watt. ## Already running abroad

This gear has shipped before. Qualcomm unveiled Dragonwing on February 25, 2025. Viettel in Vietnam deployed the QRU100 radios. NTT Docomo in Japan used the Dragonwing X100. Rakuten Mobile launched QRU100 in February 2026. In San Diego, Qualcomm tested a panel on a 10-story roof and showed narrow, high-capacity beams.

At a November 4 briefing covered by PCMag, Qualcomm's Durga Malladi said, "This is the first time we have a US-based deployment that's coming up." Reuters quoted Qualcomm calling the SpaceX contract its biggest 5G deal yet. Malladi on size: "We're not doing small sales over here." On site count: "If you look at how many sites the existing carriers have, it's in the same ballpark."

A clean sheet of paper

Legacy carriers keep decades of older gear alive on their towers. SpaceX has no legacy ground gear at all. Malladi said that gives you the ability to have a significantly more modern network: "It's like a clean sheet of paper."

Qualcomm pitches this as a road to the next generation, too. In the release, Malladi said the companies are advancing 5G today with an opportunity to lay the foundation for AI-native 6G, and Qualcomm CEO Cristiano Amon posted on X that morning about a path to AI-native 6G. ## Count the sites

Here is where I slow down. Per the Wireless Infrastructure Association, America had 254,850 macro cell sites at year-end 2025, plus nearly 200,000 outdoor small cells. On October 13, 2026, Deutsche Telekom CEO Timotheus Höttges said, "We have 85,000 mobile antennas in America," and said Elon Musk has to build those first.

So when Malladi says "same ballpark," compare it to one carrier, not the whole country. Tens of thousands of sites against T-Mobile's 85,000 makes a real city network. It will not match every tower in America, and I don't think it has to, because the satellites cover the empty miles.

Where do the sites go? In October, Bloomberg reported SpaceX was in talks with tower landlords like Crown Castle for at least 80,000 cell sites. As of November 4, nothing is signed. Those leases are the next thing I'm watching.

The cost fight

Estimates for a national ground network run from about $80 billion to $500 billion. TD Cowen says over $80 billion. Joe Madden at Mobile Experts and Sam McHugh at BNP Paribas each say at least $100 billion. Schulman said on October 26 it will take a minimum of 4 to 6 years and gave a range of $300 billion to half a trillion over seven or eight years. The same day, Musk replied to a clip of Schulman on X: "Will take two or three years."

The incumbents have more arguments. Verizon spokesman Rich Young told Reuters on October 8 that tons of spectrum doesn't matter without a network: "It's just empty airwaves." T-Mobile CTO John Saw wrote on LinkedIn on October 9 that even after the Grain deal, Starlink Mobile would have roughly 80 megahertz, some of it uplink only, against T-Mobile's nearly 400.

The usual shortcut is closed. New carriers normally rent capacity from the big three while they build. Schulman said he sees "absolutely no benefit whatsoever to extending an MVNO deal to any satellite provider." On October 28, T-Mobile CEO Srini Gopalan said, "I see no room for an MVNO with SpaceX." So SpaceX builds or waits. That is why the Qualcomm order exists.

Elon time versus the calendar

I have tracked Musk's timelines for years. His clock is always aggressive, and big hardware takes longer than the first post. SpaceX's own executives sound more careful. On October 20, Shotwell said the priority is flying the mobile V2 constellation through 2027, with service on SpaceX's own spectrum aimed at early 2028. CFO Bret Johnsen targeted the first half of 2028 for 5G-quality satellite service.

The calendar: Qualcomm base stations start rolling out in 2027. The Grain 800 megahertz deal awaits FCC approval, targeted for the first half of 2027. The EchoStar spectrum's second closing is targeted around November 30, 2027. Neither big spectrum deal has closed. Commercial service on SpaceX's own airwaves points to 2028.

Meanwhile, Shotwell said Starlink Mobile passed 22 million monthly users through partner carriers at quarter end, up 37% in one quarter.

What changes for you

Before November 4, the skeptics' best line was Young's: spectrum without a network is empty airwaves. SpaceX answered with a purchase order for 64T64R radios on tens of thousands of US sites, starting from a blank slate in 2027.

For a typical household, I expect nothing to change on the bill in 2027. The first real test is a price. When SpaceX posts one, put it next to the $155.73 or $168 account averages and ask whether one account for dish and phone beats what you pay now.

Longer term, I think cars, robots and AI agents will need an unbroken link between ground and orbit just as much as phones do. Legacy carriers are bolting satellite partnerships onto networks built for yesterday's phones. SpaceX is building the roof and the pillars in one company. Three years or six, my read is that the big three just got a fourth builder with real hardware on order.