Where the rules get written

The FCC has already permitted a satellite operator to build a terrestrial network

On 12 May 2026 the FCC’s Wireless Telecommunications Bureau and Space Bureau approved both EchoStar transactions. The order moves about 115 MHz of mid-band spectrum to SpaceX and AT&T, and most of its length is conditions. One sentence in it settles the architecture question the market has been arguing about (Exhibit 1).

The Commission described the two deals as “collectively surpassing $40 billion in deal value” (news release; order in GN Docket No. 25-302, DA 26-471). The parties’ own filings say $19.617 billion for SpaceX and $20.25 billion for AT&T, $39.87 billion between them. The two figures reconcile once the $2.4 billion escrow is added.

The consequential part of the decision is one sentence under the heading Tech-Neutral Innovation. It allows SpaceX to use its new spectrum “flexibly for terrestrial, space-based, and hybrid network architectures”. The permission carries conditions. The Commission attaches what it calls “first-of-its-kind performance obligations”, requiring “meaningful, reliable connectivity to the public — whether D2D, terrestrial, or both — within years”. Neither the permission nor the obligation prefers one architecture over the other.

That symmetry matters. Whoever holds this spectrum decides how to use it, and the test it must pass says nothing about how the signal reaches the handset. In our view the regulatory question is closed: SpaceX may build a terrestrial mobile network on its own spectrum if it chooses. What remains is whether it can afford to and whether the physics allows it.

The FCC has answered whether a satellite operator may build a terrestrial mobile network on its own spectrum, and the answer is yes. The questions left are financial and physical, and no regulator will settle them.

The rest of the order is conditions on other parties. EchoStar must fund the $2.4 billion escrow, AT&T’s 600 MHz build-out is brought forward, and Boost Mobile survives through a hybrid MVNO arrangement with AT&T. The performance terms deserve a closer look than the release gives them, because the Commission has since used them twice.

The order does not ask SpaceX for a terrestrial build-out. The rules would have required 75% of the population of each licence area to be covered for the AWS-3 and H block licences and 70% for AWS-4; the order waives that in favour of a test a satellite can pass. From 30 November 2027, in each licence area and each block, SpaceX has two years to show a downlink signal-to-noise ratio of 5 dB and service available 70% of the time over 90% of the area, with uplink throughput for the worst-served users of at least 100 kbps per MHz; four years to lift availability to 80% and throughput to 200; and nine years to reach 10 dB, 90% and 300, with spectral-efficiency floors of 300 and then 600 kbps per MHz per beam on the downlink. Missing the first test brings the final one forward by a year, and missing the second by two years more; a licence that fails the final test terminates automatically. The order also notes that the 3GPP standards do not yet support satellite operation in the 1915 to 1920 MHz block, and tells SpaceX to ask for a waiver near the milestone if it needs one.

The 800 MHz order applied the same test to a second band

Seven weeks later the Wireless Telecommunications Bureau used the same template on a band SpaceX did not yet own. On 1 July 2026 it approved T-Mobile’s sale of its 800 MHz licences, the 817 to 824 and 862 to 869 MHz pair, to affiliates of Grain Management, and Grain’s sale of 600 MHz licences back to T-Mobile (DA 26-653, WT Docket 25-178). This is the band the Sprint merger judgment had ordered T-Mobile to sell, which DISH declined to buy and which drew no qualifying bid at auction in 2024; part 4 has that history and the prices. Grain had come to the Commission in 2025 describing utilities and regional operators as its customers, and in March 2026 it changed its plan to satellite direct-to-device. The Bureau allowed the change on conditions. Grain had to appoint a neutral adviser and open a competitive process for a satellite partner by 7 August 2026, finish it and report the outcome by 5 November, and have complete satellite applications filed by 4 December. If it did, the satellite metrics of the EchoStar order apply to the band on the same three dates: 30 November 2029, 2031 and 2036. If it did not, a terrestrial build-out applies instead: 33% of the population or 25% of the geography of each licence by 1 July 2029 and 66% or 50% by 1 July 2034, with automatic termination behind it. Because the 800 MHz band is not allocated for supplemental coverage from space, the Bureau also deferred every waiver the band needs for satellite use until the partner was named (Exhibit 2).

The partner was named on 8 October 2026, when Grain announced a definitive agreement to sell the whole portfolio to SpaceX, subject to the Commission’s consent, and said in its release that the agreement aligns with the July order’s process (Grain release). The bids and the other participants are not public; Bloomberg had reported AST SpaceMobile’s interest in August, and that report is unverified. On the record, the regulator wrote a satellite performance test for one operator’s spectrum in May, wrote the same test into a second band in July and required that band’s holder to find a satellite operator to meet it, and the operator found was the one already holding the first band. Two days before the Grain announcement the Space Bureau and the Wireless Bureau had authorised that operator’s next constellation: 15,000 direct-to-device satellites in nine shells between 326 and 335 km, half of them due in orbit by 7 October 2032 and all by 7 October 2035, with the 2000 to 2020 MHz band turned to downlink use by waiver so that both 2 GHz blocks point the same way, the 2020 to 2025 MHz uplink request deferred, and the out-of-band power question of 2024 settled (DA 26-1078; part 3). The 800 MHz band is not among that constellation’s frequencies. Serving it from orbit will take an application the Grain order expects by 4 December.

The regulator now writes a satellite performance test rather than a build-out, and it has written the same one twice. Whoever holds the 800 MHz band and the EchoStar bands answers to one calendar, and its first date is 30 November 2029.

Congress had settled a year earlier where the next tranche of spectrum comes from.

Congress has restored auction authority to 2034 and ordered an 800 MHz pipeline

Congress restored the FCC’s auction authority in the reconciliation act signed on 4 July 2025, and this time it runs to 30 September 2034. The statute also attaches a pipeline of at least 800 MHz, including at least 100 MHz of upper C-band at 3.98 to 4.2 GHz that must be auctioned by July 2027. The FCC’s own auction release credits the statute for both the authority and the pipeline. The enacted text, section 40002, and its six deadlines are set out clause by clause in part 2 of the 6G series.

The first auction under the restored authority followed within a year, and it was the first of any kind in four years. Auction 113 opened on 2 June 2026 with AWS-3 spectrum at 1695 to 1710, 1755 to 1780 and 2155 to 2180 MHz. The 200 licences on offer were the ones left over after the 2014 defaults, and the proceeds go to the “rip and replace” reimbursement programme (FCC, 2 June 2026). The next sale is Auction 115, which opens on 27 April 2027 with 160 MHz of upper C-band (Exhibit 3).

Auction 113 cleared at $2.658 per MHz-POP

The opening release framed Auction 113 as housekeeping: leftover licences, with proceeds to a reimbursement fund. The results read differently. The 200 licences sold for $3,572,889,200 in gross winning bids across 1,344,395,555 MHz-POPs, which works out at $2.658 per MHz-POP (our calculation). Verizon, bidding as Cellco Partnership, paid $3,162,445,000 for 82 licences, 88.5% of the total (Exhibit 4).

Two things stand out. The first is the concentration: no other bidder came near Verizon’s $3.16 billion, and T-Mobile, which took more licences than Verizon at 102, paid $277.8 million for them. The second is that SpaceX bid at all. Until June 2026 the company appeared in the American terrestrial record only as a tenant. Its two licences cost $8,490,200, or $0.367 per MHz-POP, the lowest price paid in the auction (Exhibit 5).

One caveat governs every use of that clearing price. Auction 113 sold a leftover inventory concentrated in dense markets, so its $2.658 per MHz-POP cannot usefully be set beside a nationwide price such as the C-band’s $0.852 or the $0.887 SpaceX agreed with EchoStar. A fair comparison would need a population-weighted match market by market, which this part does not attempt. What the figure does show is that in 2026 a national carrier paid about $3 per MHz-POP for mid-band capacity in the markets it wanted, and that price belongs beside the C-band figure whenever the latter is quoted as the benchmark.

The midterm election cannot reach the permissions already granted

The mechanics of the Commission are widely misdescribed. The Congressional Research Service’s standing account of the agency (CRS R45699) sets them out, and they are simpler than the commentary suggests. There are five Commissioners on staggered five-year terms, and no more than three may come from one party. The President designates the Chair from among the sitting Commissioners on the first day of an administration, without a separate Senate vote. A Commissioner whose term has expired may stay on as a holdover, and three Commissioners make a quorum. None of those rules turns on an election.

The membership does change, though not on election night. A nomination to one of the Commission’s vacant seats was announced in August 2026 and had not been confirmed when Axios reported it on 10 August (Axios, 10 Aug 2026). An August report does not give the seat count in September.

Election day is 3 November 2026. All 435 House seats are contested, as are 35 in the Senate (33 regular Class 2 seats and two specials), by the count of the published election trackers. None of that changes who sits on the FCC, because the terms and the appointment mechanism run on their own statutory calendar. The result can change five other things.

• Committee chairs. Who chairs Senate Commerce and House Energy & Commerce.
• Confirmations. How willing the Senate is to confirm nominees to the two vacant seats.
• Appropriations. The agency’s budget.
• Legislation. Whether any spectrum legislation gets beyond the 2025 act.
• Oversight. How hard Congress looks at NTIA and BEAD.

Those levers act on the pace of policy; the permissions already granted are out of their reach. We know of no scheduled proceeding that would reopen the core direct-to-device permissions before 2028. The SCS framework in GN Docket 23-65 is adopted, the EchoStar transfers are approved and the auction pipeline is written into statute. The FCC’s composition turns over with the presidency, and the next presidential election is in November 2028, the period in which SpaceX’s performance obligations and AST’s leases also come due. The dates coincide; this part does not forecast from them.

No scheduled proceeding reopens the core direct-to-device permissions before 2028. The election can change committee chairs, confirmations and the pace of legislation. It cannot touch the waivers, the SCS framework or the statutory pipeline.

A proposed unlicensed regime would need no spectrum position at all

Separately, and with far less attention, the Commission circulated a Notice of Proposed Rulemaking for its open meeting of 6 August 2026. ET Docket No. 26-169, Unleashing Unlicensed Spectrum for Direct-to-Device, is described in the fact sheet the Commission published on 16 July 2026. The proposal would let a device transmit to a satellite from three unlicensed bands at ordinary Part 15 power of up to 36 dBm EIRP, more than 225 MHz in all (Exhibit 6). The notice also asks whether power flux-density, out-of-band and aggregate limits should apply when several satellite operators share a band.

If the proposal becomes a rule, a class of direct-to-device service arrives that needs neither an auction nor a lease. Nothing in the proposal suggests it could carry voice. It may carry the sensor and asset traffic on which several business cases in this market depend, which is why a proceeding about unlicensed bands belongs in a strategic model that otherwise tracks only licensed ones.

The same notice records why the Commission is interested. By its own count, more than $40 billion has moved through the US space economy for direct-to-device in eighteen months. It cites SpaceX’s $19.6 billion EchoStar purchase, AST’s $550 million L-band agreement, and Amazon’s acquisition of Globalstar and Rocket Lab’s of Iridium as “collectively totaling nearly $20 billion”. Those are the transactions the earlier parts cover from the commercial side. The notice remains a circulated draft and its technical limits are proposals; it is a signal of direction and nothing firmer.

The six factors in a PESTLE frame

Exhibit 7 puts the findings of the series into a PESTLE frame, one factor per row, with the document, docket or dataset each rests on.

Terms used

• MHz-POP. One megahertz of spectrum covering one person. Dividing a price by the MHz-POPs it buys puts auctions and private sales of very different sizes on one scale.
• SCS, supplemental coverage from space. The FCC’s framework, adopted in GN Docket 23-65, under which a satellite operator may serve ordinary phones on spectrum that a terrestrial carrier is licensed to use.
• PFD, power flux-density. The power a satellite’s signal delivers per square metre at the ground. The FCC limits it to protect terrestrial users of the same and adjacent bands.
• NPRM. A Notice of Proposed Rulemaking: the FCC’s draft rules and questions, published for comment before any rule is adopted.
• Part 15. The FCC’s rules for unlicensed devices, which set power limits such as the 36 dBm EIRP proposed in ET Docket 26-169.
• SINR. Signal to interference plus noise ratio, in decibels: how far a wanted signal stands above everything else at the receiver. The FCC’s satellite tests ask for 5 dB on the downlink to begin with and 10 dB by the final test.
• The 800 MHz band. Here, the 817 to 824 and 862 to 869 MHz pair that T-Mobile sold to Grain and Grain has agreed to sell to SpaceX. The 800 MHz of bandwidth the 2025 statute orders auctioned is a quantity rather than a band.