SpaceX just asked the FCC for permission to launch a million satellites. Yes, you read that right. A million.
To put that number in perspective, humanity has launched about 15,000 satellites total since Sputnik in 1957. SpaceX wants to multiply that—by itself—in the coming years. The goal? Build a space-based network powerful enough to run the world’s AI systems.
It sounds absolutely insane. But then again, so did reusable rockets ten years ago.
What Exactly Are They Proposing?
Starlink already has around 6,000 satellites beaming internet to people in remote areas. That’s impressive, but it’s basically a warmup compared to what this new filing describes.
The idea is to create a massive mesh network in low Earth orbit—essentially a computing backbone floating above the planet. Right now, AI companies like OpenAI and Google rely on huge data centers packed with servers. These facilities eat electricity, generate enormous heat, and tend to cluster in places with cheap power and cold weather.
Space solves some of those headaches. Solar panels work great up there. The vacuum handles cooling naturally. And a satellite network can serve users anywhere on Earth without the latency problems that plague people far from data centers.
Is This Even Possible?
Technically? Maybe. SpaceX has proven it can manufacture and launch satellites at a pace nobody else matches. Their costs keep dropping. And they’ve got the rockets.
Practically? There are serious questions.
Astronomers are already frustrated with Starlink messing up their observations. A million satellites would make that problem exponentially worse. We’re talking about fundamentally changing what the night sky looks like.
Then there’s the debris issue. Every satellite is a potential collision. Managing traffic for a million objects—even with AI coordination—pushes the limits of what’s been done before. One bad chain reaction, and you could make low Earth orbit unusable for decades.
The Money Angle
Here’s why SpaceX is doing this: AI infrastructure is becoming the most valuable real estate in tech.
Microsoft, Amazon, and Google spend billions every year building and running data centers. The AI computing market is projected to hit hundreds of billions annually within the decade. If SpaceX can capture even a fraction of that spending, we’re talking about revenue streams that dwarf Starlink’s current $6 billion.
And unlike selling internet to consumers, enterprise AI clients have deep pockets and long-term contracts. That’s the kind of business that makes investors very happy.
The Geopolitical Reality
There’s another layer here that doesn’t show up in the FCC filing. Control over AI infrastructure means leverage.
If SpaceX builds the backbone that powers global AI, whoever controls those satellites has significant influence over who gets to use that capability—and who doesn’t. Countries without their own space-based alternatives might find themselves dependent on a network controlled by a single American company.
China’s clearly thinking about this. Their satellite ambitions have accelerated dramatically. The space-based AI infrastructure race might already be underway, even if most people haven’t noticed yet.
What Happens Now
The FCC has to review the proposal. They’ve generally supported satellite expansion, but nothing at this scale has ever crossed their desk. Environmental reviews, international spectrum coordination, safety assessments—it’s going to take a while.
Maybe it gets approved. Maybe it gets scaled back. Maybe the whole thing turns out to be too ambitious even for SpaceX.
But the fact that they’re even asking tells you something about where this industry is headed. The ground isn’t big enough anymore. The next frontier for tech infrastructure might genuinely be space.
Source: Federal Communications Commission
