Space FCC Brawl: Musk vs. Legacy Titans

Photo: kovop / Shutterstock

Elon Musk has asked regulators for permission to put one million AI data center satellites in orbit—but the gap between the paperwork and actually doing it is wide enough to fly a Starship through.

Story Snapshot

SpaceX filed a Federal Communications Commission application to deploy up to one million orbital data center satellites.
The plan leans on Starship launches, solar power, and radiative cooling to run artificial intelligence in space.
Engineers warn the heat, radiation, and cost problems are far beyond anything yet proven.
The million-satellite number is a legal ceiling, not a concrete build schedule.

What SpaceX Has Actually Put On Paper

SpaceX did not just tweet a dream; it filed a formal application with the Federal Communications Commission asking to operate up to one million satellites as “orbital data centers.” The filing describes satellites in low Earth orbit between 500 and 2,000 kilometers, using narrow orbital shells about 50 kilometers thick. Each satellite would be solar powered and tied into the existing Starlink laser-link network, turning the whole swarm into a giant computing cloud above the atmosphere.

The company says these orbital data centers would dodge land, power, and cooling limits that now choke artificial intelligence growth on the ground. SpaceX claims near-continuous sunlight in those orbits means the satellites can run on solar power more than 99 percent of the time, with batteries as backup. Vacuum in space would act as the ultimate heat sink, with radiative cooling panels dumping waste heat straight into the cold black sky. On paper, it sounds clean, cheap, and limitless.

The Million-Satellite Number And The Starship Bet

The one-million figure has shocked people, but it is more legal maximum than firm promise. The application sets a ceiling so SpaceX does not have to go back and ask for more capacity later if the idea works. There is no timeline in the filing for actually reaching one million satellites. SpaceX expects to start with pilot tests on upgraded Starlink Version 3 hardware and then ramp up only if demand and Starship economics line up.

Everything hinges on Starship, the giant reusable rocket SpaceX is still proving out. The filing and outside analyses assume Starship can loft huge masses of hardware at very low cost, enabling thousands of launches over time. American conservative common sense says this is where the risk really sits: if Starship reaches airline-like reuse, the numbers begin to pencil out; if it stalls, the dream turns into a very expensive science project. Even friendly analysts admit that launch costs today are far above the “hundreds of dollars per kilogram” level most experts see as needed for orbital data centers to compete with land-based ones.

Physics Does Not Care About Hype Or Stock Price

The biggest pushback does not come from politicians or short sellers; it comes from engineers with calculators. Running powerful artificial intelligence chips creates extreme heat. One detailed analysis points out that SpaceX’s proposed AI-1 satellite design would need to throw off more than twice the heat of the International Space Station using about a quarter of the radiator area, meaning roughly eight times more heat rejection per square meter than any spacecraft has ever achieved.

Radiation is the second brick wall. The most advanced commercial AI chips, like those from Nvidia, are built for data centers, not space. In low Earth orbit, high-energy particles can flip bits in memory every single 90-minute orbit, corrupting computations unless protected. Musk has promoted a future radiation-hardened “Tera F” chip for space use, but there is no public proof it exists in working form today. Until either new hardware appears or massive error-correction software proves itself at scale, putting today’s AI brains into orbit looks more like a demo than a dependable utility.

Money, Green Narratives, And Conservative Skepticism

On the money side, serious people are using serious numbers. One equity research firm has warned that the full one-million-satellite vision could reach into the trillions of dollars over time, far above any “cheapest place to run AI” claim Musk has floated. Gartner analysts have gone even further in tone, branding the idea “peak insanity” given today’s launch prices and the need to design, build, and replace hardware on five-year cycles.

SpaceX’s own stock offering paperwork backs up that caution. In its S-1 filing for the initial public offering, the company admits that orbital AI compute relies on “unproven technologies” and “may not achieve commercial viability.” That sort of language is not woke hand-wringing; it is a legal shield written by lawyers who know the Securities and Exchange Commission will pounce if investors are misled. For a reader with a conservative bent, this is the key tell: when the marketing sizzle and the sworn risk disclosures disagree, trust the sworn part.

Space Crowding, Governance, And What Comes Next

There is also the simple question of how much hardware the sky can swallow. Today, there are about fourteen thousand active satellites. SpaceX alone wants permission to add up to one million more, and other firms have begun to file for large constellations of their own. Policy outlets like Brookings warn that flooding orbit with data centers risks more collisions, more debris, and more fights over who owns the sunlight and radio spectrum those machines use.

Regulators have not slammed the door. The Federal Communications Commission accepted the SpaceX filing and moved it into public comment, with outside observers suggesting the agency may use the process both to gather feedback and to build a record strong enough to say no if the plan cannot be squared with safety. The core facts are clear: Elon Musk can file for one million satellites. Whether he can launch, cool, and pay for them fast enough to make space the cheapest home for artificial intelligence is a different question—and physics, not politics, will have the final vote.