The vision of orbiting data centers has attracted major investments from tech giants, with Nvidia CEO Jensen Huang declaring at the March GTC conference that "space computing, the final frontier, has arrived." SpaceX has acquired xAI and is planning a constellation of space-based data centers, while Google announced Project Suncatcher with Planet, aiming to launch two satellites equipped with TPU AI chips by early 2027. Startup Starcloud has filed an FCC proposal for an 88,000-satellite constellation dedicated to orbital computing.
Proponents point to significant advantages: abundant solar energy, free cooling thanks to the cold of space, and immunity from Earth-based disruptions like earthquakes, floods, or protests. These fleets, each housing racks of AI-grade GPUs, would connect via free-space optical links and communicate with Earth through microwave links.
However, the technical obstacles are substantial. The article notes that placing sophisticated computing hardware in orbit introduces challenges around heat dissipation, radiation hardening, and maintaining high-bandwidth connections to ground stations. Launch costs, while declining, remain significant for constellations numbering in the thousands of satellites.
Regulatory approval also presents a major bottleneck. The FCC must weigh orbital debris concerns, spectrum allocation, and international coordination before approving massive constellations like Starcloud's proposed 88,000-satellite network. Existing satellite internet constellations have already faced criticism for cluttering low Earth orbit.
Critics argue that terrestrial data centers, powered by renewable energy and benefiting from rapid advances in cooling efficiency, may prove more practical for most AI workloads. The economics of transmitting data to and from orbit add latency that could negate the benefits for many applications.