SpaceX Wants 100,000 Satellites. This Time, It's Not Really About Your Wi-Fi.
The Setup: A Constellation Ten Times the Size of Today's
Starlink's current constellation, roughly 10,000 satellites strong, already blankets the globe with broadband. That would normally be the ceiling of an ambitious plan. Instead, SpaceX has filed an application with the FCC for a third-generation system, called Gen 3, designed to eventually span up to 100,000 satellites — an order-of-magnitude leap that dwarfs the existing network entirely.
The Breakthrough: Built for People, But Also for Machines
According to the FCC filing, Gen 3 satellites would sit in two tightly stacked shells in very-low Earth orbit, at roughly 325 and 475 kilometers up, and would use frequencies stretching from the traditional Ku and Ka bands all the way into the largely untapped W and D bands between 92 and 275 GHz. The company describes the goal as ultra-low-latency, multi-gigabit broadband delivered to orders of magnitude more users worldwide.
But the filing's most telling line isn't about consumers at all. SpaceX describes Gen 3 as intended to serve not just households, businesses and governments, but billions of AI-powered devices — going so far as to say the system will function as the communications backbone of the AI age. Each Gen 3 satellite reportedly weighs around 2,000 kilograms and can only be launched aboard the still-in-development Starship rocket, promising roughly ten times the downlink capacity and 22 times the uplink capacity of the current fleet.
Why It's Bigger Than It Looks
Most satellite internet coverage focuses on a simple pitch: connectivity for the billions of people who don't have reliable broadband. Gen 3's filing reframes the story. If SpaceX genuinely expects billions of AI-connected devices to need constant, low-latency bandwidth, then satellite internet stops being a backup option for rural households and starts becoming infrastructure for a machine-to-machine economy that doesn't care where the nearest fiber line is.
That's also a financial story. Analysts have offered sharply different views of what this could mean for SpaceX's newly public satellite arm, with some projecting AI-related revenue could climb from roughly $3.2 billion to $322 billion by 2030 — a hundredfold jump that only makes sense if satellites become genuine AI infrastructure.
The Part Nobody Talks About: Regulatory Approval Isn't a Formality
A filing is not a launch schedule. Getting FCC approval for 100,000 satellites invites exactly the kind of scrutiny that has slowed past mega-constellation proposals: orbital debris risk, interference with other satellite operators, and spectrum-sharing fights with rivals who don't want to cede the W and D bands without a fight. None of that is settled by a filing alone.
Conclusion: The Network Is Racing Ahead of the Rocket
Gen 3 depends entirely on Starship actually flying reliably at scale, something that has not yet been proven on the timeline SpaceX has repeatedly promised. Still, the filing signals where the company thinks demand is heading — not just more people online, but more machines needing to talk to each other from anywhere on Earth, all the time.
References:
1. https://www.pcmag.com/news/starlink-gen-3-to-span-100k-satellites-for-gigabit-broadband-but-also-ai
2. https://aphnetworks.com/news/31887-starlink-gen-3-span-100k-satellites-gigabit-broadband-also-ai
3. https://finance.biggo.com/news/83a23cc2-b4ab-4fe3-82a9-455493765c32