SpaceX’s Ambitious Plan for Orbiting AI Data Centers Faces Scrutiny
Elon Musk’s SpaceX is pushing the boundaries of data center technology with plans to launch orbiting facilities powered by solar energy and dedicated to artificial intelligence processing. These satellites, envisioned to be longer than the International Space Station (ISS), are sparking both excitement and concern among space experts, and astronomers.
Scale and Capabilities
The proposed satellites are designed with exceptionally large solar arrays to capture abundant solar energy, powering high-density AI processing. While the initial rendering didn’t provide exact dimensions, each satellite significantly exceeds the length of the Starship V3 rocket (124.4 meters or 408 feet) and dwarfs the ISS, which spans 109 meters (357 feet). Musk clarified that the presented image represents a “rough approximation” of a “mini” version, suggesting future iterations could be considerably larger.
Current designs aim for 100 kilowatts of AI computing power, with future versions targeting a “megawatt-range” capacity. To put this in perspective, 1 megawatt can power between 200 and 700 homes, depending on energy consumption. SpaceX’s ultimate goal is to launch up to 1 million satellites, potentially offering thousands of gigawatts of AI compute, mirroring the scale of terrestrial AI data centers like those planned by xAI, which aim for 1,000 megawatts (1 gigawatt) of capacity.
Technical Challenges and Concerns
Despite Musk’s confidence that the project requires “no new physics,” the plan is not without its challenges. Hugh Lewis, a space debris expert and professor of astronautics at the University of Birmingham, highlights the need for constant maneuvering to avoid collisions with space junk and other satellites. He estimates approximately 40,000 maneuvers per day are currently needed across all constellations to avoid objects 10 cm and larger, potentially reaching 100,000 maneuvers daily with a significantly expanded fleet. He warns that even with these efforts, collisions among active satellites are likely due to residual collision probabilities.
Astronomers and environmentalists are also raising concerns about light pollution and potential interference with radio astronomy observations. Samantha Lawler, a professor of astronomy at the University of Regina, noted that the scale of the project was underestimated. The International Astronomical Union fears the satellites will radiate heat that interferes with observations.
Another concern revolves around the potential release of ozone-depleting chemicals during the atmospheric re-entry of retired satellites. SpaceX has responded to the Federal Communications Commission (FCC) stating it will begin with a small constellation to study potential atmospheric impacts before scaling up, and aims to make the satellites too faint to be visible to the naked eye or telescopes.
Chip Manufacturing and the Terafab
A significant hurdle for SpaceX is the large-scale manufacturing of the AI chips required for these satellites. Musk announced plans to build a new factory, dubbed the “Terafab,” in Austin, Texas, to produce cutting-edge processors, including GPUs and memory chips, for both SpaceX and Tesla. While currently reliant on suppliers like Samsung and Micron, Musk stated, “We either build the Terafab, or we don’t have the chips.” Building chip factories in the US typically takes three to five years and requires tens of billions of dollars in investment, and the industry is currently facing a chip shortage.
Competition and Future Outlook
SpaceX is not alone in pursuing orbiting data centers. Blue Origin and the startup Starcloud are also developing their own constellations involving tens of thousands of satellites. Musk believes the cost of deploying AI into space will soon be lower than terrestrial AI deployment, potentially within two to three years.