Now, the Financial Times reports that SpaceX Chief Operating Officer Gwynne Shotwell told institutional investors during the IPO roadshow that the company is considering selling mobile contracts directly to consumers and may eventually build its own terrestrial U.S. mobile network.
According to a Financial Times report, SpaceX has told investors it’s considering launching a retail Starlink mobile service in the US. Instead of simply partnering with wireless carriers, the company could begin selling mobile plans directly to consumers, putting it in direct competition with Verizon, AT&T, and T-Mobile.
News reports this morning say Spacex may buy T-mobil for access to their network.
This may well be s negotiating strategy. For access. Building their own network would be costly and might take years. Shared service w an existing network would be much easier. But why would existing networks want to admit a new competitor?
All of the majors have sub-leased capacity and there are a host of low-cost competitors: Mint, US Mobile, and others. But doesnt he have enough to just buy one of those MVNOs or something?
To be a cell carrier he would have to have some sort of arrangement, it certainly couldn’t be done just with the satellites. The cell companies have to erect multiple-multiple towers because of call volume, and there just isn’t enough bandwidth for satellites flying overhead to handle it all with the one, two, or three that might be in position at any given time.
Broadcasting from a satellite to a million satellite dishes? Easy. Broadcasting from a million satellite dishes to a satellite? Hard. Probably impossible.
Valuable service for certain situations: oil platforms, military units in combat, remote logging and mining camps. Crowded urban settings? Maybe not.
They usually add additional antennas to existing towers when possible. The bandwidth problem is addressed by reducing the range of the antennas and then adding additional towers between existing ones.
Yes, Starlink satellites are close enough to Earth, orbiting in Low Earth Orbit (LEO) at around 340 to 350 miles (550 km) up. This close proximity allows them to act as “flying cell towers,” delivering low-latency connections that easily support real-time voice calls, text messaging, and internet browsing directly on standard smartphones. [1, 2, 3, 4, 5]
Here is how the direct-to-cell service works and what it delivers:
How It Functions: Starlink’s direct-to-cell (DTC) satellites are equipped with large, advanced phased-array antennas that can form narrow beams to close the link with the weak transmission power of an ordinary, unmodified cell phone on the ground.
Supported Devices: You do not need a special satellite phone or extra hardware. The service operates on 4G LTE/5G bands, integrating automatically with your existing mobile carrier (such as T-Mobile, which offers this service) when you lose traditional terrestrial cell coverage.
Latency & Speed: Because they are in LEO rather than traditional high-altitude geostationary orbits (which are over 22,000 miles up), round-trip data delays (latency) are very low. Users typically experience latencies in the 25 to 50 millisecond range, making real-time communication viable and fast.
Current Capabilities: The technology is primarily designed to provide supplemental coverage from space, eliminating “dead zones” in rural areas or places where terrestrial towers have been destroyed. It currently provides basic global text messaging, SOS emergency features, and voice calls, with continuous upgrades rolling out for expanded data speeds. [1, 8, 9, 12, 13]
Could you tell me what mobile carrier you currently use, or what part of the country you are looking to travel to or get coverage in? This will help determine if the Starlink direct-to-cell service is already available to you.
You on’t know yet, but AI is only as good as the dataset it has been fed. There’s a lot of “behind the scenes” stuff here, but I’ll skim the highlights for you.
A cell tower, except for those upgraded in the last couple years, can handle maybe 150 simultaneous voice calls. (This is improving, sometimes dramatically, as they change to VoIP and/or upgrade trunk lines back to the CO. But most aren’t like that yet.)
The reason you are asked by the airline (and mandated by the FCC) not to use cell service when aloft is that you can “grab” as many as 1,000 individual circuits on individual towers depending on whether you’re over a dense city and your altitude and so on.
Following from #2, cell phones only reach 2-4 miles in cities (buildings, electric infrastructure), but in the country up to 25. So by having 5,000 towers in greater Chicago it can handle the load - both ways. But if you try to do that with a couple of satellites overhead, it’s too much traffic all at once. (5000>4, if that helps.) A Starlink can handle a couple hundred simultaneous calls (a bit more more with some bandwidth tricks), so trying to run “solely” on satellites is a non-starter, at least where people live in density. (Aka: customers).
Which is why I say its a terrific system for people in rural, even ex-burban settings, (military, oil rigs, and similar) and you can have some traffic from urban areas, but you’re better handling as much as you can through a traditional ground-based cellco because you simply can’t have that many customers with density , at least until they can convert everything to digital (which takes less bandwidth/radio frequency) and put more satellites overhead (and there is a limit to that by orbital separation needs.)
I am not saying this can never be done, or can’t be done; of course it can, but it is, with current technology, self limiting. Better to route as much traffic to ground as can be handled with ground/fiber/coax, and keep the other resource for where you need it. Which is why partnering/buying an existing cellular infrastructure would be logical.
PS: I consulted with Motorola and United to try to get “live cell service” and “live TV” aboard aircraft around 1996-97. Didn’t happen then for a variety of reasons, most notably cost and their unwillingness to “just do it.” A “test” program would never reach critical mass, and we weren’t willing to fund a test we knew would fail. Eventually, long after I was gone, somebody else with somebody else did make it happen, or at least parts of it, and good for them. We knew it was a good idea, the technology just wasn’t quite there for it. (Or the market. It was 1996, after all.)
Don’t listen so much to AI summaries. Sometimes they’re right, often they’re just remixed garbage. This one, for instance, has been trained on what I will call “Musk Predictions”. What else there, after all?
If you really want an AI job well done you reach for Claude.
So Claude
I’ll check current Starlink Direct-to-Cell specifics before answering, since this area has moved fast.Short version: your core thesis is right and still endorsed by SpaceX itself. Several of the specific numbers and the “circuit” framing are dated. Here’s the breakdown.
What still holds
The central claim — satellite direct-to-cell is capacity-limited, best as a gap-filler for rural/maritime/remote, and complementary to (not a replacement for) dense terrestrial networks — is exactly the position SpaceX takes publicly. At MWC 2026 their Starlink SVP said the system cannot provide the data density that terrestrial networks can, and positioned satellite as complementary, useful for mountaintops, oceans, and regions terrestrial can’t reach. So point 3 and point 4 are essentially correct and aging well.
Your prediction in point 4 (“partnering/buying existing cellular infrastructure would be logical”) is now actively playing out. SpaceX bought 65 MHz of nationwide mid-band spectrum from EchoStar (FCC-approved May 2026), and on June 26, 2026 Gwynne Shotwell told IPO investors SpaceX plans to challenge AT&T, Verizon, and T-Mobile directly and may build its own terrestrial cellular network. The thing they’re still missing is towers — which is your whole point.
What’s outdated or wrong
The “circuit” / 150-simultaneous-calls framing is dated. LTE/5G isn’t circuit-switched the way AMPS/early digital was. VoLTE rides on packet data; capacity per sector is governed by spectrum, MIMO, and sectorization, not a fixed circuit count. The “grab 1,000 circuits across towers from altitude” model is legacy thinking. The real reason for the airborne-phone rule is a fast phone at altitude registers with too many towers at once and creates handoff/signaling churn — but framing it as seized circuits is anachronistic.
“A couple hundred calls per satellite” is the most outdated number. Gen1 (the ~650 deployed sats) is text-focused, but the architecture moved. Gen2/V2 satellites support thousands of spatial beams and roughly 20x the throughput of a first-gen satellite, and SpaceX claims the second-gen satellite has nearly 100x the data density and ~16x the beams, with per-satellite throughput exceeding 100 Gb/s down / 50 Gb/s up. V3 on Starship is projected at another big step. So the per-satellite ceiling is rising fast.
“Convert everything to digital (less bandwidth)” is anachronistic. It’s already all digital. The actual constraint isn’t analog→digital conversion — it’s spectrum × beams × satellite count, bounded by orbital slots. You named the orbital-separation limit correctly; that’s the durable one.
The cleaner way to state your density argument
The reason satellites lose on density isn’t a fixed small call count — it’s beam footprint. A satellite beam covers tens of km on the ground, so all that capacity is shared across a huge area and population. Terrestrial wins because you pack cells tightly and reuse frequency over and over. That geometric fact is permanent regardless of how many beams or satellites SpaceX adds, which is exactly why even SpaceX frames it as fill-in, not replacement. That’s the strongest version of your argument, and it survives all the spec upgrades.
So: thesis sound, conclusion vindicated by SpaceX’s own moves, but drop the circuit/analog framing and the specific per-satellite call numbers — those are the stale parts.
As I tried to explain, Starlink is great where there are no customers (rural, RV parks, military, etc) and maybe not so great where lots of people live:
And the space-based network is already starting to experience some major strains — as [some experts have long predicted](https://futurism.com/starlink-internet-connection-problem). It's a lesson some Starlink customers are finding out the hard way. As *TechDirt*'s Karl Bode highlighted, SpaceX is now charging some users so-called "demand surcharges" of up to $1,500, simply because their address is within a high demand area.
"Starlink is too congested to handle meaningful load at scale so they're quietly hitting people with $750-$1500 'demand surcharges,'" Bode wrote. Even worse, those who are trying to contest these charges are facing a "black hole," since SpaceX doesn't "invest in customer service."
Starlink has a place. It’s not gonna take over the cell phone business without, you know, buying a cell phone company. This doesn’t auger particularly well for using it to communicate with hundreds of autonomous robotaxis zipping around the urban core. But these kinds of pesky facts never seem to get much traction with Musk enthusiasts.
I admire and detest the founder, and my husband and I have the internet capabilities we need to work intensively and connectedly in the Mexican countryside because of Starlink.