The question arrived on X as a link to a spectrum filing. Elon Musk answered it in one line on Monday: building SpaceX’s own mobile network across the United States, he wrote, “will take two to three years.”
He offered nothing further. Not the number of cell sites, not the cities that would come first, not whether a customer would buy service from SpaceX directly or from the carrier that currently resells its satellite coverage. For a company that publishes launch schedules years ahead, the reply was notable for what it left out.
The pieces are mostly assembled. In September 2025, SpaceX agreed to buy EchoStar’s AWS-4 and H-block spectrum licenses for about $17 billion in cash and stock, a transaction that also obliged the buyer to cover $2 billion of interest for the seller. Those licenses carried buildout conditions: cover a set share of the population by a set date, or return the airwaves to the government.
That requirement is what turns a two-year promise into a longer project. Land mobile spectrum is licensed on deadlines, and the Federal Communications Commission has shown little patience with holders that miss them. EchoStar itself spent years asking for extensions before it sold.
Direct-to-cell service, the part SpaceX already operates, is the cheap half of the business. The company began working with T-Mobile in 2022, moved from emergency texts to a commercial offering in July 2025, and treats every launch as another tower in orbit. A satellite pass costs a fraction of a ground site, and the marginal customer already owns the handset.
A terrestrial carrier is a different trade. It needs towers on leased land, fiber backhaul to each of them, radios tuned to its own frequencies, roaming agreements wherever coverage stops, certification from handset makers, and a network operations center that answers when a caller cannot reach emergency services.
Each of those is a contract, and contracts take longer than rockets. Site acquisition alone typically runs eighteen months in dense markets, according to people who work in tower development, because a single objection from a municipal board can restart the process.
The competitive field is no longer empty. AST SpaceMobile has spent six years building satellites that talk directly to unmodified phones, backed by AT&T and Verizon, and has launched its first commercial craft. Its approach mirrors SpaceX’s: put the tower in orbit and let the carriers pay for the capacity. Neither company has yet shown it can carry voice traffic at the density of a city.
Analysts who follow the sector treat three years as the optimistic case. Dish Network took roughly a decade to assemble a national network from scratch and still fell short of the coverage commitments attached to its licenses, a history regulators cite in every proceeding about new entrants. Building a fourth national carrier from a standing start in two years would be faster than any comparable effort in the industry’s recent memory.
The capital is the other question. Starlink, the broadband unit, is the group’s cash engine, and Musk has said for years that it would eventually be listed without naming a date. A ground network would consume cash in the years before it produces any, at the same time that the launch business is absorbing spending on Starship.
T-Mobile has reason to watch closely. The partnership gave the carrier a satellite service to sell in rural counties where it owns no spectrum of its own, and gave SpaceX a retail channel it could not build quickly. If SpaceX becomes a carrier, the two companies move from suppliers to competitors, and the wholesale terms between them become the first thing to renegotiate.
Analysts said the more likely near-term shape is a hybrid: SpaceX sells capacity wholesale to carriers and keeps direct billing for a narrow set of customers, such as maritime and enterprise accounts, where no retail brand is needed. That model requires less capital and no stores.
Regulators will have a say in the pace regardless. Any terrestrial build needs approval to use the purchased spectrum for mobile service, and the FCC’s review includes interference studies with adjacent licensees, a process measured in months rather than weeks.
The wholesale route also fits the spectrum. The H-block licenses SpaceX bought sit next to frequencies that T-Mobile and Dish already use, and a network built on adjacent airwaves can share radios and sites with an incumbent. Buying access to someone else’s towers is cheaper than building a portfolio of them, and it compresses the schedule that Musk just described.
For customers, the practical question is simpler than the corporate one. Satellite service today works best outdoors and in places where nothing else does. A ground network would work everywhere, including indoors, which is where most calls are made and where the satellite product cannot yet compete.
Musk has a record of setting deadlines that slip and then arriving. He said in 2016 that SpaceX would send a craft to Mars by 2018. The company has not done so, and it now dominates the launch market anyway. The two-to-three-year figure should be read in that light: a direction of travel, dated by a man who dates everything.
