SpaceX’s next Starlink system is beginning to look less like a better broadband constellation and more like a layer of orbital computing infrastructure.
Elon Musk says each Gen3 satellite will target roughly 10 terabits of connectivity in both directions, draw 250 kilowatts of power and carry a SpaceX-designed Nvidia Vera Rubin NVL72 computer.
Those are targets, not demonstrated on-orbit results. Even so, they put a startling scale on the hardware behind SpaceX’s application for as many as 100,000 Gen3 satellites.
The timing is important. The FCC accepted that application for filing on September 18, moving it into formal review.
Musk supplied the new per-satellite details two days later.
The Federal Communications Commission lists SpaceX’s Gen3 proposal in public notice SAT-02038. The filing asks for authority to operate up to 100,000 non-geostationary satellites across two very-low-Earth-orbit altitude bands.
The lower group would fly between 323 and 327.5 kilometers. The upper group would sit between 473 and 477.5 kilometers, with orbital inclinations ranging from 26 degrees to 96.9 degrees.
Accepted for filing does not mean approved. It means the FCC found the application complete enough to enter the formal review process, where petitions, oppositions and technical questions can still shape or stop the proposal.
The notice also preserves the Commission’s right to return an application after deeper examination. SpaceX has cleared the administrative starting line, not the finish line, and the company cannot treat the requested 100,000-spacecraft system as authorized.
Musk put the new hardware targets plainly in his same-day post:
Connectivity per sat will be more like 10Tb in both directions and there is a path to 100+Tb.
Each sat is 250kW and will have a SpaceX-designed Nvidia Vera Rubin NVL72 computer.
— Elon Musk (@elonmusk) September 20, 2026
The combination is more revealing than any one number. Ten-terabit connectivity describes the communications ambition, 250 kW describes the power class, and the NVL72 reference points to a substantial onboard AI-compute role rather than a satellite acting only as a relay.
The FCC Gen3 application record identifies the proposal as file SAT-LOA-20260630-00264, call sign S00887. SpaceX filed it in July and describes the system as infrastructure for consumers, enterprises, governments and billions of AI-powered devices.
The application seeks extremely low latency and multi-gigabit symmetrical service. It also reaches into frequencies as high as 275 GHz, including W- and D-band spectrum that could support enormous capacity but is technically demanding to use through the atmosphere.
Flying low helps latency and reduces the distance those high-frequency links must travel. It also means atmospheric drag can bring failed spacecraft down faster, though managing a fleet on this scale would create an orbital-traffic challenge unlike anything operating today.
The 100,000 figure is a systemwide ceiling, not 100,000 satellites in every orbital shell. SpaceX is asking for flexibility to distribute the fleet across the proposed altitudes and inclinations while keeping the total under that cap.
Tesla North connected Musk’s statement with SpaceX’s current launch plan. Starship Flight 14 is expected to carry 26 production Starlink V3 satellites, beginning the move from development articles toward hardware intended to add paying network capacity.
SpaceX previously described each V3 satellite as adding about 1 Tbps of downlink capacity. Musk’s new 10-terabit statement appears to describe a broader Gen3 connectivity target, and his reference to a path beyond 100 terabits pushes even further out.
The distinction matters. None of those larger figures has been demonstrated in orbit, and the 100,000-satellite filing remains under review.
Starship must also reach a launch cadence capable of deploying spacecraft far larger and more power-hungry than today’s V2 Mini satellites.
That leaves SpaceX with three enormous jobs at once: win regulatory approval, prove the Gen3 hardware in orbit and make Starship routine enough to build the constellation at useful speed.
If the targets hold, Starlink Gen3 will not be merely a larger version of the current network. It will be a communications and computing platform in orbit, built on a scale that only Starship is supposed to make practical.
Join the conversation!
Please share your thoughts about this article below. We value your opinions, and would love to see you add to the discussion!