NextFin News - SpaceX wants to expand Starlink to more than 100,000 Version 3 and later satellites, tying that buildout to a larger claim: artificial intelligence will require far more orbital bandwidth and, eventually, computing capacity in space. The proposal, laid out in Elon Musk’s recent remarks, lands just after SpaceX’s initial public offering raised about $75 billion at a valuation of roughly $1.75 trillion.
This is not about selling more satellite internet subscriptions. It is about trying to move Starlink from a connectivity product into infrastructure for AI, with satellites that could carry data-center-like workloads in orbit. On the surface this looks like a bigger broadband network; the real issue is whether SpaceX can turn orbital capacity into a business with pricing power across communications, defense and AI compute.
The technical case rests on Starlink V3. Gulf News said the satellites are expected to be about the size of a Boeing 737 and deliver roughly 1,024 Gbps of downlink capacity per satellite, more than 10 times the throughput of V2 models. SpaceX is also pitching the network as a way to extend direct-to-cell service into remote areas where fiber and towers are expensive, slow to build or commercially uneconomic.
If those performance targets hold, the change is structural. A constellation of that scale would operate less like a retail telecom service and more like a distributed utility for moving data, supporting mobile connectivity and potentially hosting computing tasks. The beneficiaries are clear: enterprise customers, defense users and AI operators that need low-latency links across geographies. The pressure falls on terrestrial providers in hard-to-serve regions and on any incumbent that assumes AI demand will stay concentrated in land-based server farms.
But the math does not add up yet on the hardest part of the story: orbital AI economics. SpaceX has filed plans for as many as 1 million satellites that could function as AI data centers, each potentially delivering 120-150 kW of power using technology derived from Starlink V3. That is not a routine network upgrade. The real trade-off is between theoretical global coverage and the cost, power management, launch cadence, maintenance burden and regulatory friction required to keep such a system operating at scale. Whether this works depends on whether SpaceX can verify not just throughput, but that customers will pay enough for space-based compute to justify manufacturing tens of thousands of satellites and replacing them over time.
The immediate constraint is orbital congestion. Concerns over space junk, collision risk and interference with astronomy are not side issues; they go to the operating limits of the model. Every additional launch, maneuver and deorbit cycle increases complexity in what would already be one of the most crowded low-Earth-orbit systems ever assembled. FCC approval and international coordination will decide how much of this vision can be deployed, and regulatory timelines do not compress simply because a company has fresh capital.
The $75 billion IPO matters because this is a capital-intensity story before it becomes a margin story. The proceeds give SpaceX room to scale satellite production, Starship launches and any future space-based AI infrastructure, but a $1.75 trillion valuation means investors are already paying for more than rockets and broadband. They are paying for a multi-line infrastructure company whose next test is proving that higher throughput can become durable revenue growth rather than a larger capital bill.
The risk nobody is talking about enough is execution sequencing. SpaceX has to mass-produce V3 hardware, launch at pace, manage congestion, win regulatory clearance and then show that demand for orbital bandwidth and compute is real, not merely adjacent to the AI boom. More than 100,000 satellites in the sky is a concrete industrial target; a profitable orbital AI economy still needs to be proved.
Explore more exclusive insights at nextfin.ai.
