Up to 7 MHz Twice: What SpaceX's Reported $8 Billion Actually Buys
Headlines say 14 MHz and walls. The FCC paperwork and building-loss physics show two strips of up to 7 MHz each, a band not allocated for supplemental coverage from space, and a price near 2.2 times what T-Mobile's books implied.
Two strips of radio, each up to seven megahertz wide, sit 45 MHz apart: 817 to 824 MHz and 862 to 869 MHz. One carries a phone's signal up to the network, the other carries it back down. The FCC's own yardstick for a "broadband" network on that pair is a single LTE carrier 5 MHz wide in each direction. SpaceX agreed on Oct. 8 to buy both strips from Grain Management, reportedly for about $8 billion in cash, and Grain had owned them for 58 days.
The prize is low-band spectrum, which travels far from a tower, sold under a terrestrial license rather than a satellite one. It does not push a signal through a roof much better than the 2 GHz band SpaceX already has, and the FCC's own order says the band is not allocated for satellite supplemental coverage. This is a tower-and-rulebook bet wearing a satellite headline.
What 14 megahertz paired actually means
SpaceX describes "up to 14 megahertz of paired spectrum in the 800 MHz band." The arithmetic is simple: up to 7 MHz up plus up to 7 MHz down. In the FCC's July order approving the T-Mobile and Grain swap, the agency described 4.85 to 14 MHz of 800 MHz spectrum across 3,224 counties and 730 market areas, covering approximately 100% of the U.S. population. Fourteen is the ceiling.
The licenses have an odd history. They came to T-Mobile through Sprint, and the Justice Department required T-Mobile to divest them. DISH was offered them and did not buy; its option expired on April 1, 2024. On May 30, 2025, T-Mobile agreed to sell them to Grain affiliates for $2.9 billion in cash plus Grain's 600 MHz licenses, according to T-Mobile's quarterly filing. That closed on Aug. 11; Grain's announcement of the SpaceX sale, 58 days later, has no price. The $8 billion comes from the Wall Street Journal, citing people familiar with the matter, via Reuters; neither company has confirmed it.
SpaceX's own post on X:
Post by @SpaceX
The post says "low-band" and "major mobile carrier." The price, the ground plan and the timetable are not in it. Converge Digest noted the announcement "did not specify a terrestrial rollout schedule, infrastructure partners, service pricing or a commercial launch date." The FCC's approval earlier this week of the 15,000-satellite Gen2 constellation, built for 2 GHz, is a separate matter for the company that listed this year.
Markets did not wait for details. In after-hours trading Thursday, T-Mobile, Verizon and AT&T fell around 6%. By 11:02 a.m. EDT Friday, per the quote table published with Reuters' market report on The Globe and Mail (data as of 11:02 a.m. EDT), T-Mobile was down 10.42% at $153.46, AT&T down 7.2% and Verizon down 6.68%, while SpaceX was up 1.77%. The earlier text of the Reuters report tallied 9.1%, 7.4% and 5.9%. Deutsche Telekom, which holds 54% of T-Mobile US, fell 7.1% in Frankfurt. The tell is the towers: Crown Castle was up 11.73%, SBA 5.86% and American Tower 7.06%. Investors are reading a deal that needs steel in the ground. These numbers move.
Low-band buys range from a tower, not a way through a roof
Start with what low frequencies really do. At the same distance, a signal at 865 MHz loses 7.26 dB less in free space than one at 1995 MHz. Translated: a tower reaches about 2.3 times as far, or covers roughly 5.3 times the land, in open air only; real terrain shrinks that. Bernstein, as summarized by BigGo, estimated Grain's spectrum "could cut the number of needed sites by about 30%." It is the logic of why 5GHz Wi-Fi gives out sooner than 2.4GHz, at larger scale.
Walls are a different story, and the one SpaceX leans on. Its statement says the new band "will provide a coverage layer that ensures Starlink Mobile's signal penetrates through obstacles, such as walls." Low-band does penetrate better in principle; the question is how much. The ITU's standard building-entry-loss model gives a median loss for a traditional building of 14.2 dB at 865 MHz and 15.0 dB at 2 GHz, a gap of 0.7 dB. For a modern thermally efficient building, with metallised glass or foil-backed panels, it flips: 31.2 dB at 865 MHz against 30.5 dB at 2 GHz. ITU measurements show loss rising about 1 to 3 dB per GHz in most buildings, so roughly 1 to 3 dB across this gap. The wall costs 14 dB or more. The frequency edge is a rounding error beside it.
Then there is geometry. A satellite is overhead, not at the horizon, and the same model adds 0.212 dB for every degree of elevation at the building face. That is 9.5 dB at 45 degrees and 19.1 dB straight up, in either band.
| Path and building (editor's computation, ITU-R P.2109-2 median) | 865 MHz | 2 GHz |
|---|---|---|
| Traditional, horizontal | 14.2 dB | 15.0 dB |
| Traditional, 45 degrees | 22.2 dB | 23.5 dB |
| Traditional, 90 degrees | 31.5 dB | 32.9 dB |
| Thermally efficient, horizontal | 31.2 dB | 30.5 dB |
The ITU calls the model "not site-specific," and the 10th-to-90th percentile spread for a traditional building at 865 MHz runs from 5.7 to 25.0 dB, so any one house can differ a lot. And there is an idealized point from textbook physics: with a fixed-size satellite antenna and a fixed-gain phone antenna, received power does not depend on frequency at all. The range advantage of low frequency is a tower benefit. That is derivation, not measurement. SpaceX's case has force, though: the band suits towers, and Apple lists it for the iPhone 16.
The second limit is capacity. A full 7 MHz lane holds one 5 MHz LTE carrier each way, with 2 MHz to spare. A preprint measuring Starlink's direct-to-cell beta, which uses a 2x5 MHz slice of PCS spectrum, estimated downlink capacity of 3.1 Mbps per beam, an upper bound, at 0.61 bps/Hz. Scaled to a 7 MHz lane, that is about 4.3 Mbps per beam: arithmetic, not a measurement. The FCC's own 2036 satellite benchmark of 600 kbps/MHz/beam works out to 4.2 Mbps over 7 MHz. Split among 50 simultaneous users, that is about 85 kbps each.
Analyst Vish Nandlall, in a social media post quoted by RCR Wireless, called the market reaction "a crazy market reaction to a relatively small amount of radio spectrum," and put the point plainly: "the spectrum provides coverage rather than much capacity." Tim Farrar of TMF Associates told CNBC, as quoted by TechRadar, that "to get reliable building penetration in urban areas, SpaceX would have to deploy towers on the ground." JPMorgan said SpaceX would likely need large cell towers unless it secures a mobile virtual network operator deal. SpaceX's own plan, per Reuters, includes "ground-based radios and antennas on towers and rooftops."
The rulebook says this band is not allocated for supplemental coverage from space
The FCC's order, at paragraph 50, says: "The 800 MHz Enhanced Specialized Mobile Radio Service (ESMR) band, for which Grain would be acquiring licenses from T-Mobile, is not allocated for SCS." SCS is supplemental coverage from space, the FCC's framework for satellites to fill gaps on carriers' spectrum. Its 800 MHz allocation covers 824 to 849 and 869 to 894 MHz, just above the Grain block.
Grain asked for waivers. The FCC deferred them "until Grain identifies a D2D partner and files the relevant applications," and said the public may comment on any waiver requests when that happens. The satellite path is conditional and untested. To use the satellite benchmarks instead, Grain had to finish a competitive partner-selection process by Nov. 5 and file complete satellite applications by Dec. 4. How those dates apply now that SpaceX is the buyer, coverage has not answered. Public-safety and business radio channels at 851 to 862 MHz sit directly beside the block. A Light Reading column in April argued the FCC has never set a power limit for satellite use of this band "because no one has asked until now"; that is opinion.
The terrestrial rules are firmer. Grain "or any future assignee" must offer reliable service to at least 33% of the population or 25% of the geography of each license by July 1, 2029, and 66% or 50% by July 1, 2034. Miss the final deadline and the license terminates automatically, with no way back.
No closing date has been announced, and no filed application has been reported in the sources reviewed. The FCC's goal is to act within 180 days of the notice accepting an application, with a clock that can pause. Recent history ran over: 224 days from the first notice for SpaceX and EchoStar (168 from the amended notice), and 407 days for the T-Mobile and Grain swap. FCC Chairman Brendan Carr, responding to the deal, said: "We are going to see more investment, more opportunities for a range of providers, and America leading the world in next-gen technologies — to the benefit of consumers all across the country."
The price pays for permissions
The usual yardstick is dollars per MHz-pop: per megahertz, per person covered. Using a U.S. population of 341.8 million:
| Deal | Spectrum | Price | Per MHz-pop |
|---|---|---|---|
| SpaceX from Grain (reported) | 14 MHz, low-band | about $8 billion | $1.67 |
| Grain from T-Mobile (book basis) | 14 MHz, low-band | about $3.6 billion | $0.75 |
| SpaceX from EchoStar | 65 MHz, 2 GHz and AWS | $19.6 billion (contract value) | $0.88 |
| AT&T from EchoStar | about 50 MHz, mid and low | about $23 billion | $1.35 |
The reported price is 2.2 times what T-Mobile's books imply for the same licenses two months earlier, and about 1.9 times the contract value of the EchoStar deal, up to $11.1 billion of which was paid in SpaceX stock valued at $212 a share. It is also about 1.24 times AT&T's blended price. If average depth is 10 MHz rather than the maximum 14, the price works out to $2.34, so the ratios to the EchoStar and AT&T deals are floors, while the ratio to T-Mobile's books is unchanged. The gap of about $4.41 billion is not Grain's profit, since Grain also gave up its 600 MHz licenses.
What changed in between is the rulebook. Grain asked the FCC for a satellite path in March 2026, and the FCC granted a conditional one on July 1. T-Mobile's filing says it "may receive a share of certain future proceeds" from Grain's monetization of the licenses, and nothing public says whether this sale triggers it or how much it could be. The evidence points one way: the price pays for permissions and a possible satellite path, not for megahertz alone.
Nandlall's arithmetic, from the same post, shows what that demands. At 20 million subscribers, recovering $50 billion over 10 years requires $21 per subscriber per month, before operating costs, financing or any return. The flip side worries carriers: for every 100 million subscribers, a $1 cut in the monthly bill removes $1.2 billion in annual revenue.
If you pay AT&T, Verizon or T-Mobile, nothing changes yet
Your bill and coverage are the same today, and there is no price, launch date or ground schedule. SpaceX targets 2027 for next-generation satellites, per Teslarati. T-Mobile is both SpaceX's direct-to-cell partner and the licenses' seller. Gwynne Shotwell, SpaceX's president and COO, said on the August earnings call: "I anticipate us to be able to acquire quite a few of their customers because I think our service will be better."
One check you can do yourself: whether your phone lists LTE Band 26 or 5G n26. Band 5 alone does not cover 817 to 824 MHz. SpaceX says "most existing mobile devices already support this underused band," which is plausible for recent flagships and untested for the rest.
Verizon's spokesman Rich Young told Reuters: "This company can have tons of spectrum available, but if they do not have the network to use it, it doesn't matter. It's just empty airwaves." The FCC applies the same test to everyone. On the terrestrial route, whoever holds these licenses must cover a third of the population, or a quarter of the land, in every license area by July 1, 2029, license by license, or the final deadline moves up a year.
