Fifty-one thousand, six hundred satellites. That is the number Blue Origin put in its Federal Communications Commission filing on Thursday for Project Sunrise, a proposed megaconstellation of orbital data centers designed to run AI workloads in space.

The company’s New Glenn rocket has completed two test flights.

The Pitch

Blue Origin argues that solar-powered satellites eliminate the land costs, grid constraints, and water consumption that plague terrestrial data centers. The filing describes a constellation spread across sun-synchronous orbits between 500 and 1,800 kilometers, with 300 to 1,000 satellites per orbital plane connected by optical inter-satellite links. Traffic would route to Earth through TeraWave, Blue Origin’s planned broadband network — with the first 5,000-plus satellites slated for deployment by the end of 2027.

The company even asked the FCC for a waiver from milestone rules that would require half the constellation deployed within six years of approval. In other words: we need more time than the rules allow for a technology that, according to Gartner analysts, doesn’t exist yet and would likely prove unreliable for practical data center operations.

The Race Nobody Asked For

Blue Origin is not the first to file for orbital compute. SpaceX proposed its own space data center constellation earlier this year — which Amazon promptly petitioned the FCC to block. Now Bezos wants in on the same idea his company tried to kill when a rival proposed it.

Meanwhile, astronomers are growing louder. A study published in Nature found that planned megaconstellations could spoil more than 96 percent of images from space telescopes, according to NASA-affiliated researchers. With roughly 14,000 active satellites currently in orbit and over a million more proposed, the night sky is becoming prime real estate.

The technology for orbital data centers does not yet exist. But the FCC filings, apparently, write themselves.

Sources