Sovereignty

Scale vs. Sovereignty - Deep Dive Weekly

A newly public SpaceX filed for a 100,000-satellite Gen3 constellation. Japan, Indonesia, Nepal and the UAE answered with sovereign connectivity.

Glenn Canales

Glenn Canales

— 17 min read
Share
Scale vs. Sovereignty - Deep Dive Weekly
SpaceX Gen3: 100,000-Satellite FCC Filing — Deep Dive | Satellite Insights Weekly

Satellite Insights Weekly

A Newly Public SpaceX Files to Put 100,000 Satellites Overhead

The most significant event of the week is a filing, not a launch: SpaceX — public on the Nasdaq since June 12 — has asked the FCC for up to 100,000 “Gen3” satellites and a slice of greenfield W/D-band spectrum, reframing Starlink as AI-era infrastructure. Japan, Indonesia, Nepal and the UAE spent the same week buying the opposite thesis: sovereign control as the product.

Glenn Canales — Principal, Satellite Insights LLC · 40+ years in satellite communications; former Sr. Director, Enterprise Broadband Services, Viasat (2016–2024); iDirect, PanAmSat/Intelsat; USAF Communications Supervisor (TS/SCI). Disclosure: The author is a former Viasat employee. Viasat is referenced in this edition only as a likely objector in an FCC proceeding; nothing here reflects Viasat's positions.
100,000
Gen3 satellites SpaceX asked the FCC to authorize
proposed ceiling · designed
~$920M
Japan J-LEO maximum subsidy (USD equiv.)
≈ ¥148B · program cap
~€567M
MDA Space's cash offer for ~70% of CLS
≈ C$920M · stake value
4
Sovereign-connectivity deals landed this week
Japan · Indonesia · Nepal · UAE

Most significant event of the week

SpaceX's July 6 FCC filing for a 100,000-satellite Gen3 constellation is the first major strategic signal from a public SpaceX — a spectrum land-grab and an AI-infrastructure repositioning rolled into one authorization request, and one wholly contingent on Starship reaching high-cadence flight. Everything else this week was the rest of the industry deciding whether to compete on scale or on sovereignty.

Card 01 · Lead

SpaceX files for a 100,000-satellite “Gen3” VLEO constellation — its first big move as a public company

What did SpaceX actually ask the FCC for, and why does the timing matter?

The extractable answer: On July 6, 2026, SpaceX filed FCC application SAT-LOA-20260630-00264 seeking authority to deploy up to 100,000 third-generation (“Gen3”) satellites in very-low Earth orbit, plus access to greenfield W- and D-band spectrum (92–275 GHz). The filing frames the system as infrastructure for “human connectivity and AI-fueled progress” and never once uses the word “Starlink.” It is roughly a 10× increase over the current in-orbit fleet.

The numbers (all designed / proposed, not demonstrated)

  • DesignedUp to 100,000 satellites across two VLEO shell bands — 323–327.5 km and 473–477.5 km, inclinations 26°–96.9°. Jonathan McDowell's database parses this as ~20 shells of ~5,000 satellites each.
  • Designed~1 Tbps downlink per satellite (≈10× Gen2) and 160–200 Gbps uplink (≈22× Gen2). This is per-satellite aggregate design capacity, not per-user throughput.
  • Designed~2,000 kg dry mass per satellite — launchable only on Starship; a Falcon 9 could loft roughly two. (Some outlets use ~800 kg V2 Mini as the comparison baseline — minor discrepancy flagged.)
  • Single sourceA “~4 Tbps combined RF + laser backhaul” per-satellite figure appears in only one outlet (BigGo Finance) and could not be corroborated. Treat as unverified pending the primary filing text.

Designed vs. demonstrated — the gap that frames the story

The 1 Tbps headline is an authorization-ceiling design figure. Demonstrated Starlink performance, per Ookla data cited by Via Satellite (July 9, 2026), is a median download at or above 100 Mbps in every U.S. state except Alaska, against advertised “up to 400+ Mbps,” with median U.S. upload still hovering near the FCC's ~20 Mbps broadband floor. The distance between that reality and the Gen3 pitch is the analytical center of this card.

The public-company backdrop: SpaceX has been listed on the Nasdaq (SPCX) since its June 12, 2026 debut — the largest IPO on record. That makes the Gen3 filing the first major strategic signal from a public SpaceX, and it reads like one: the AI framing that ran through the pre-listing period (the AI1 orbital-data-center reveal on June 8; the separate ~1M-satellite “Starmind” filing in January) now plays to a capital-markets audience. The 100,000-satellite application itself landed weeks after the debut — on a day SPCX traded down ~7%, below its IPO-day close.
Three entities, three different things — keep them separate: (1) SpaceX the parent is public (SPCX). (2) Starlink did not get its own IPO — it went public bundled inside SpaceX as the “Connectivity” segment; there is no separate Starlink ticker, S-1 or scheduled spin-off. (3) The private tender/secondary valuations from late 2025 and early 2026 (~$800B in December; secondary-market marks around $1T in April) were price-setting events, not public offerings. Conflating any of these is the most common error in current coverage.
Why it matters

This is three moves at once: a spectrum land-grab claiming greenfield W/D-band before Amazon Leo and China can; an AI-infrastructure repositioning now playing out in full public-market view; and a competitive pre-emption against every rival constellation. But an FCC application is not an approval — Gen2 took years and remains contested — and the whole edifice rests on Starship. The next real catalyst is August 6, 2026: SpaceX's first quarterly earnings as a public company and its first insider-lockup expiration land the same day. Expect formal objections from Amazon, Viasat, EchoStar, NASA and astronomy groups during the comment window.

↓  PUBLIC PREVIEW ENDS HERE  —  in Ghost, place the /paywall split on the line below this point. Everything above (headline, standfirst, “most significant event,” and the lead card's extractable answer) stays public and indexable.
The Sovereign-Connectivity Answer

Card 02

Japan moves to award its ~$920M “J-LEO” sovereign D2D program

Why is a G7 government spending nearly a billion dollars to avoid depending on Starlink?

Japan's Ministry of Internal Affairs and Communications (MIC) is selecting a winner for J-LEO, a domestically owned and operated direct-to-device network, with a public reveal expected in the July–August window.

  • JPY / USDPer Japan's CIAJ implementation manual (Advanced Television, July 6, 2026): a maximum subsidy of ¥148 billion (~$920M USD), capped at 50% of eligible pre-tax costs; with required private matching the overall eligible budget could approach ~¥296 billion (~$1.8–2B USD).
  • MIC service bar: smooth video calls on regular smartphones 16+ hours a day, nationwide rollout by March 2029, and all network and data control kept fully domestic.
  • DesignedFront-runner is Rakuten Mobile + AST SpaceMobile, helped by a June 24 subcommittee approval of 700 MHz for satellite D2D — a “platinum band” only Rakuten holds. But AST had ~10 Block 2 satellites in orbit in July 2026 and needs 45–60 for continuous Japan coverage: early service would be intermittent into 2027.
Why it matters

The competing camp is KDDI–SpaceX, whose “au Starlink Direct” passed 4 million users in its first year. J-LEO is the clearest signal yet that, for a major economy, data sovereignty is overriding “good-enough” foreign service — the exact counter-thesis to Card 01.

Card 03

Space42 and Skylo run standards-based D2D over a GEO bird, Thuraya-4

Can geostationary satellites — not just LEO — do direct-to-device?

On July 9, 2026, Space42 (ADX: SPACE42) and Skylo announced successful testing of two-way SMS and SOS on standard Android devices over the GEO Thuraya-4 satellite, using Skylo's 3GPP-compliant NTN platform — no modified SIM, no operator core changes. Commercial rollout is targeted for end-2026, pending regulatory approval.

  • Demonstrated vs. roadmapSMS and SOS were demonstrated. Voice, location sharing, IoT data and automotive connectivity remain designed/roadmap.
  • Evolving claimNote the sequencing: a May 11, 2026 announcement had claimed a bi-directional voice call over Thuraya-4; the July test specifically validated two-way SMS/SOS on standard Android. Track this as an incremental, shifting claim set.
Why it matters

It reinforces that GEO operators can play in standards-based D2D, and it advances the UAE's sovereign NTN ambitions — Space42 was formed from the Bayanat–Yahsat merger.

Incumbents Execute: Software-Defined GEO & Owned Ground

Card 04

Es'hailSat orders a fully software-defined GEO satellite from Thales Alenia Space

If GEO is “dying,” why is Qatar still ordering a new one?

On June 30, 2026, Thales Alenia Space announced a contract with Qatar's Es'hailSat for Eshail-3 / Türksat-Biruni, a fully software-defined GEO satellite on the Space INSPIRE platform, shared with Türksat and covering Europe, Africa, Central Asia and the Middle East.

  • Undisclosed valueContract value not disclosed. For scale: Thales Alenia Space (Thales 67% / Leonardo 33%) posted €2.36B consolidated revenue in 2025.
  • The satellite can be reprogrammed in orbit and host government payloads — the reconfigurability that lets a GEO bird chase demand instead of being locked to a fixed plan.
Why it matters

The GEO order book is being sustained by software-defined, sovereignty-framed satellites, not traditional fixed-payload birds. This is the strategic bet incumbents are making against LEO — flexibility and national hosting over raw capacity.

Card 05 · Ground-segment franchise

Neo Space Group and Hughes deploy JUPITER 3 gateways across three countries

Who actually owns the ground network — and who just supplies it?

On July 6, 2026, Saudi PIF-owned Neo Space Group (NSG) and Hughes (an EchoStar company) deployed three JUPITER 3 ground gateways — in Jeddah, Sintra (Portugal) and the Canary Islands — to support NSG's connectivity services.

  • DemonstratedNSG reported more than 100 Mbps in initial field testing with peak speeds up to ~200 Mbps — a demonstrated floor and a demonstrated peak, not a design claim.
The franchise distinction, drawn cleanly: NSG owns the gateway infrastructure — sovereign infrastructure ownership aligned with Vision 2030. Hughes is the technology / system vendor supplying the JUPITER 3 ground system. Owner and vendor are not the same role, and conflating them is exactly the imprecision this publication exists to correct. See the franchise cluster: The Great Ground-Segment Carve-Out and The Independent & Regional Ground-Segment Layer: A 2026 Map.
Vendor-risk flagWatch the vendor's balance sheet. This deployment was announced July 6, 2026 — the same day Hughes's creditors engaged restructuring counsel ahead of a $1.5B debt maturity due August 1, and the same day its CEO resigned (Charlie Ergen assuming the role). No Hughes projects or commitments have been reported canceled, and Hughes was excluded from the concurrent Dish DBS Chapter 11 filing — but the maturity is unresolved and Chapter 11 for Hughes itself remains on the table. This is where the ownership-vs-vendor split earns its keep: NSG owns the gateways outright and is insulated from Hughes's balance sheet; Hughes's role as technology vendor is where the exposure sits — a live diligence question for anyone depending on JUPITER 3 ground systems.
Why it matters

Another sovereign wealth fund is buying and owning ground infrastructure outright rather than renting capacity — extending the ownership-vs-vendor split that defines the ground-segment map. (Aviation/IFC is the underlying use case here, so it's context, not the lead subject.)

Challengers Scale, Downstream Consolidates

Card 06

China's Qianfan passes 238 satellites on a disciplined launch cadence

How cheaply is China actually building its megaconstellation?

China launched two Qianfan (Spacesail) batches on July 4–5 — 18 satellites via Long March 6A and a 20-satellite group via Long March 8A, “the largest group to date” — bringing the total to 238. Guowang continues in parallel.

  • CNY / USDPer China-in-Space, each Qianfan satellite weighs 300 kg (flat-pack design, krypton Hall-effect thruster) and costs under 10 million yuan (~$1.47M USD, as of July 4) — a citable benchmark on Chinese mass-manufacturing economics.
  • EstimatesBut “cheap” is relative — and Qianfan isn't. SpaceX publishes no official unit costs, but on analyst estimates Qianfan's ~$1.47M / 300 kg runs higher per satellite than a Starlink V2 Mini (730 kg, ~$400K–$800K est.) and roughly level with the far larger projected V3 (~1,500 kg, ~$1.2M est.) — and ~4–6× more expensive per kilogram. The V2 Mini estimate itself spans a 2× range across sources, so treat the comparison as directional.
  • Targets: 324 satellites by end-2026; Phase 1 is 1,296; full constellation 15,000 by 2030. Qianfan is eyeing international markets (Brazil, Malaysia, Kazakhstan, Türkiye) around Q4 2026.
Why it matters

China's dual megaconstellations keep deploying on schedule despite earlier 2025 thruster/gyro issues (resolved April 2026). But the contrarian read is in the unit economics: Qianfan is not a per-unit cost leader. On the available estimates it costs more per satellite — and several times more per kilogram — than Starlink. China's “compete on scale” thesis is real, but it rests on state backing, launch cadence and strategic will, not a manufacturing-cost advantage.

Card 07

MDA Space moves downstream, buying ~70% of France's CLS

What does a satellite manufacturer want with a data-analytics firm?

On July 8, 2026, MDA Space (TSX/NYSE: MDA) entered a firm, irrevocable offer to acquire ~70% of France's CLS (Collecte Localisation Satellites) for ~€567M (C$920M) cash; French space agency CNES retains ~30%.

  • EUR / CAD / USDStake value ~€567M ≈ C$920M; MDA upsized a bought-deal equity offering to ~US$819M to fund it. CLS expects €286M (C$465M) 2026 revenue, ~22% average annual growth since 2023; it reported ~€203M in 2025.
  • Scope noteCLS includes a satellite-IoT business (keeping it in scope here); its EO-analytics core is adjacent context. The deal is expected to roughly double MDA's recurring revenue.
This fits the consolidation thread tracked in The Next Domino: Satcom's Most Likely M&A Moves — but note the direction of travel: hardware makers buying recurring, service-oriented revenue, not operators buying operators.
Why it matters

It signals a downstream/analytics land-grab — manufacturers moving from project work toward recurring geointelligence and IoT revenue, tying hardware constellations to AI-driven analytics distribution.

Card 08

The sovereign-connectivity cluster: Indonesia and Nepal sign in

Where is the “neutral alternative” pitch actually landing?

  • Univity–Telkomsat (Indonesia, July 6): France's Univity signed an MoU with Telkomsat to explore VLEO architecture and 5G D2D with nationally hosted ground infrastructure. Telkomsat has now taken three D2D partners in ~18 months (Starlink resale, Equatys/Space42, now Univity). DesignedUnivity plans two demonstrator satellites in 2027, operational deployment from 2028.
  • Rivada–Mercantile (Nepal, July 9): Rivada Space Networks signed Mercantile Space for its planned 600-satellite gateway-less “Outernet.” Company claimRivada says it has now “lined up over US$20 billion of business globally” — that's claimed contracted pipeline, not revenue, and Rivada has not launched operational satellites.
Why it matters

Southeast and South Asia are becoming the proving grounds where “neutral / sovereign” positioning is the explicit differentiator against American (Starlink) and Chinese (Spacesail) incumbents. Neutrality is being sold as the product — the through-line connecting Cards 02, 03 and 08.

Methodology & caveats

  • Live-figure discipline: every market cap, valuation, share price, deal value and satellite count was pulled from live July 2026 sources at compile time — not memory or prior sessions.
  • Starlink in-orbit count varies by source/date — ~10,700 (Reuters/CNBC) to 10,722 working / 10,738 tracked (McDowell, July 2) to ~10,777 (HighSpeedInternet, July 10). Flagged as a launched-vs-working-vs-operational definitional gap; the ~10× Gen3 comparison uses this range, not a single number.
  • Market-size spread (figures vary by source): satellite D2D estimates diverge widely — ~$0.57B (2025)→$2.64B (2030) per MarketsandMarkets; ~$5.03B (2026)→$13.80B (2031) per Mordor; ~$4.15B (2026) per Coherent; Omdia forecasts ~$12B smartphone-D2D revenue by 2030. These reflect different scope and methodology, so the range — not any single figure — is the honest read.
  • Single-source items flagged in-text: Gen3 “~4 Tbps RF+laser” (BigGo only); Rivada's “>$20B lined-up business” (company claim, pipeline ≠ revenue).
  • Designed vs. demonstrated enforced on all performance/capacity claims. Currency labeled throughout (USD/EUR/CAD/JPY/CNY); conversions approximate.
  • Scope: military/defense procurement, aviation and maritime treated as context only, never lead subjects. Previously heavily-covered threads (SES–Intelsat, Amazon–Globalstar, SpaceX–EchoStar, Rocket Lab–Iridium, the FCC Space Modernization Order, upper C-band auction, SpaceConnect) deliberately excluded except where a net-new development warranted it.
  • Corporate-status note: SpaceX has been public since June 12, 2026 (Nasdaq: SPCX). This edition treats the Gen3 filing as a post-IPO move and keeps SpaceX (public) distinct from Starlink (no separate listing) and from prior private secondaries.
  • Time-sensitive — Hughes (Card 05): the $1.5B Hughes debt maturity is due August 1, 2026; the CEO resignation and restructuring-counsel engagement are dated July 6. This situation is pre-deadline and fast-moving — re-verify Hughes's status at publish time, as it may change materially before the Tuesday date.
Satellite Insights Weekly · Deep Dive · July 14, 2026
© 2026 Satellite Insights LLC. Compiled from live sources; figures verified at time of publication.

Read more